Dic-Of-Social-Research.Pdf

Jupp-Prelims.qxd 3/8/2006 10:37 AM Page i

The Sage Dictionary of Social Research Methods Jupp-Prelims.qxd 3/8/2006 10:37 AM Page ii Jupp-Prelims.qxd 3/8/2006 10:37 AM Page iii

The Sage Dictionary of Social Research Methods

Compiled and edited by Victor Jupp

SAGE Publications London ●● Thousand Oaks New Delhi Jupp-Prelims.qxd 3/8/2006 10:37 AM Page iv

First published 2006

Apart from any fair dealing for the purposes of research or private study, or criticism or review, as permitted under the Copyright, Designs and Patents Act, 1988, this publication may be reproduced, stored or transmitted in any form, or by any means, only with the prior permission in writing of the publishers, or in the case of reprographic reproduction, in accordance with the terms of licences issued by the Copyright Licensing Agency. Inquiries concerning reproduction outside those terms should be sent to the publishers.

The format of this dictionary was originated by Eugene McLaughlin and John Muncie and was first used in the SAGE Dictionary of Criminology (Sage, May 2001).

SAGE Publications Ltd 1 Oliver’s Yard 55 City Road London EC1Y 1SP

SAGE Publications Inc. 2455 Teller Road Thousand Oaks, California 91320

SAGE Publications India Pvt Ltd B-42, Panchsheel Enclave Post Box 4109 New Delhi 110 017

British Library Cataloguing in Publication data

A catalogue record for this book is available from the British Library

ISBN13 978 0 7619 6297 7 ISBN13 978 0 7619 6298 4 (pbk) ISBN10 0 7619 6297 2 ISBN10 0 7619 6298 0 (pbk)

Library of Congress Control Number 2005923706

Typeset by C&M Digitals (P) Ltd., Chennai, India Printed on paper from sustainable resources Printed in Great Britain by Athenaeum Press, Gateshead Jupp-Prelims.qxd 3/8/2006 10:37 AM Page v

Contents

List of Contributors vii

Editor’s Introduction xi

The Sage Dictionary of Social Research Methods 1-326

Subject Index 327

Author Index 332 Jupp-Prelims.qxd 3/8/2006 10:37 AM Page vi Jupp-Prelims.qxd 3/8/2006 10:37 AM Page vii

List of Contributors

Editor

Victor Jupp, Northumbria University, UK

Project Administrator

Susan Doberman, Northumbria University, UK

Administrative Support

Angela Brady, graduate, Northumbria University, UK David Doberman, graduate, Warwick University, UK Rachael Beth Moss, graduate, Northumbria University, UK

International Advisory Board

Martin Bulmer, University of Surrey, UK Amanda Coffey, Cardiff University, UK Norman Denzin, University of Illinois, USA Nigel Gilbert, University of Surrey, UK Martyn Hammersley, The Open University, UK Yvonna Lincoln, Texas A&M University, USA Jonathan Potter, Loughborough University, UK Catherine Kohler Riessman, Boston University, USA Clive Seale, Brunel University, UK David de Vaus, La Trobe University, Australia

Contributors

Lee Barron, Northumbria University, UK Russell W. Belk, Northwestern University, USA Jupp-Prelims.qxd 3/8/2006 10:37 AM Page viii

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

Roy Boyne, University of Durham, UK Angela Brady, Northumbria University, UK David Brockington, University of Plymouth, UK Martin Bulmer, University of Surrey, UK John Bynner, University of London, UK David Byrne, University of Durham, UK Paul Bywaters, Coventry University, UK Ellis Cashmore, Staffordshire University, UK Amanda Coffey, Cardiff University, UK Anthony Columbo, Coventry University, UK Louise Corti, University of Essex, UK Iain Crow, University of Sheffield, UK Julia Davidson, University of Westminster, UK Pamela Davies, Northumbria University, UK Martyn Denscombe, De Montfort University, UK Derek Edwards, Loughborough University, UK Nigel G. Fielding, University of Surrey, UK Uwe Flick, University of Applied Sciences, Berlin, Germany Jeremy J. Foster, Manchester Metropolitan University, UK Sally French, The Open University, UK Philip Gardner, Cambridge University, UK Jeanette Garwood, Leeds Metropolitan University, UK Maureen Gillman, Northumbria University, UK Luca Greco, University of Paris III (Sorbonne Nouvelle), France Martyn Hammersley, The Open University, UK Jamie Harding, Northumbria University, UK Rom Harré, Georgetown University, USA Alexa Hepburn, Loughborough University, UK Claire Hewson, Bolton Institute, UK Dick Hobbs, University of Durham, UK Mark Israel, Flinders University, Australia David Jary, University of Birmingham, UK Victor Jupp, Northumbria University, UK Vince Keddie, Department for Education and Skills, UK Aidan Kelly, University of East London, UK Robert V. Kozinets, Northwestern University, USA Richard Lampard, University of Warwick, UK Gayle Letherby, Plymouth University, UK Curtis Le Baron, Brigham Young University, Utah, USA Ana Lopes, University of East London, UK Eugene McLaughlin, City University, UK Craig McLean, University of Newcastle upon Tyne, UK Bernd Marcus, Chemintz University, Germany Steve Miles, Liverpool University, UK David L. Morgan, Portland State University, USA Rachael Beth Moss, University of Liverpool, UK

viii Jupp-Prelims.qxd 3/8/2006 10:37 AM Page ix

LIST OF CONTRIBUTORS

George Moyser, University of Vermont, USA John Muncie, The Open University, UK John Newton, Northumbria University, UK Paul Oliver, University of Huddersfield, UK Geoff Payne, formerly University of Plymouth, UK Sarah Pink, Loughborough University, UK Helen Poole, Coventry University, UK Jonathan Potter, Loughborough University, UK Paul E. Pye, University of Teesside, UK Deborah Reed-Danahay, University of Texas at Arlington, USA Karl-Heinz Renner, University of Bamburg, Germany Catherine Kohler Riessman, Boston University, USA Lyn Richards, Director, Research Services, QSR, Australia Margaret Rowe, Northumbria University, UK Andrew Rutherford, University of Keele, UK Roger Sapsford, University of Teesside, UK Mark N.K. Saunders, Oxford Brookes University, UK Thomas A. Schwandt, University of Illinois, USA John Scott, University of Essex, UK Christina Silver, University of Surrey, UK Teresa Smallbone, Oxford Brookes University, UK Mark J. Smith, The Open University, UK Paul E. Spector, University of South Florida, USA Robert A. Stallings, University of Southern California, USA Thomas Staufenbiel, University of Osnabrueck, Germany Graham Steventon, Coventry University, UK Reinhard Suck, University of Osnabrueck, Germany Maggie Sumner, University of Westminster, UK John Swain, Northumbria University, UK Nick Tilley, Nottingham Trent University, UK Adelinde Uhrmacher, University of Rostock, Germany Paul Valentine, University of East London, UK David de Vaus, La Trobe University, Australia Margaret Wetherell, The Open University, UK Chris Wharton, Northumbria University, UK Malcolm Williams, University of Plymouth, UK

ix Jupp-Prelims.qxd 3/8/2006 10:37 AM Page x Jupp-Prelims.qxd 3/8/2006 10:37 AM Page xi

Editor’s Introduction

Organizing themes

Although presented in alphabetical order, the concepts covered in this Dictionary were selected on the basis of several key themes which are embraced within the term ‘social research’. These are:

(1) Philosophy of science, for example issues of ontology (what is the essential nature of reality?) and epistemology (whether or how we can gain knowledge of that reality). (2) Research paradigms, for example positivism (which in general terms is taken to include the scientific study of some objective social reality) and constructionism (which is concerned with the study of ways in which the social world is constructed through social interactions). (3) Research designs, for example the experiment (the attribution of outcomes to the controlled administration of a ‘treatment’ to one group and not another) and social survey (the systematic collection of data from or about units of analysis, usually individuals, often using sampling techniques). (4) Specific aspects of data collection, for example participant observation (participating in a group in a covert manner in order to study that group) and specific aspects of data analysis, for example multivariate analysis (a set of statistical techniques to examine the relationships between several variables). (5) Issues to be addressed when carrying out research, for example ethics (what standards should be adopted, say in relation to obtaining informed consent from subjects?) and politics and research (the extent to which research is contributing to the oppression of certain groups in society). (6) The role of research in terms of function, for example policy-related research (research to evaluate the impact of social policies) and in terms of context, for example marketing research (the systematic collection of data about consumers of products and services in order to make informed decisions).

Structure of the contributions

The term ‘Dictionary’ is used to be consistent with the Sage Publications series of Dictionaries but, as with others such as The Sage Dictionary of Criminology it is more encyclopaedic in nature. Each of the contributions is structured Jupp-Prelims.qxd 3/8/2006 10:37 AM Page xii

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

according to a standardized format. First, there is a very brief definition of the concept. Second, this is followed by a longer elucidation of distinctive features, which could include historical background, disciplinary background (for example, sociology, psychology, economics), key writers, applications (where appropriate) as well as main features. Authors were encouraged to think in terms of writing a critical and reflective essay. Therefore, for each concept, there is an evaluation in which authors raise some of the key issues and problems relating to the concept under consideration. The issues and problems which are raised are those chosen by each author rather than as a result of prescriptions laid down by the editor. It is the sections on distinctive features and on evaluation which give the publication its encyclopaedic character. For each entry, cross- references are made to associated concepts within the Dictionary. Some of these are associated by ‘similarity’ and ‘mutuality’ and others because they represent ‘challenges’ and ‘rivalry’ to the concept under consideration. The cross-references facilitate a mapping of concepts in terms of similarities and differences as described below. Finally, a brief list of key readings is provided.

How to use the Dictionary

The text can be used as a conventional dictionary or encyclopaedia to clarify the meaning of a term. However, more usefully it can be used in almost textbook fashion as a means of learning about the field of social research, and in the construction of an essay or dissertation, by making use of the cross-referencing provided by the associated concepts. The latter provide a mechanism for mapping connections between concepts in terms of similarities and differences. Associated concepts relating to any given definition have been chosen to direct the reader not solely to other concepts that share common features or underlying themes and principles but also to concepts that differ – often sharply – in terms of such features, themes and principles. The features of two of the definitions in this Dictionary can be adapted to assist in this endeavour. First, network analysis is a technique that examines the relationships between units of analysis. It was in part based on sociometry, a method founded upon asking children about their friendships. Network analysis is now more sophisticated and permits the examination of the strengths of relationships and the degree of density and interconnectedness of networks. By following cross-references it is possible to construct and examine networks of concepts: which concepts relate to one another, how they relate in terms of closeness, strength of relationship, similarities and differences. The second key concept is called the constant comparative method, which is a form of analysis in qualitative research and includes the process of minimal and maximal comparison of units of analysis in order to further understanding. Minimal comparison involves examination of cases which are as similar as possible and maximal comparison involves examination of cases which are as different as possible. This idea can be adapted to further the understanding of the territory of social research by listing the ways in which certain concepts in the network are similar and how other concepts differ. In this way the breadth and depth of social research can be uncovered.

xii A-Jupp-3373.qxd 3/8/2006 10:41 AM Page 1

setting or social group, and the relative standing ACCESS of different social actors within the field. There are also occasions when access will need Definition to be negotiated with a variety of different The process of gaining and maintaining entry social actors within the setting. This can mean to a setting or social group, or of establishing paying particular attention to the role(s) that working relations with individuals, in order the researcher adopts during social research, that social research can be undertaken. and the ways in which these roles are managed throughout the research encounter. This role management can include adopting certain Distinctive Features styles of dress or personae in order to fit into Access is part of the initial phase of social the setting. research, and is usually negotiated at the start Complex and protracted access negotia- of the research project. Access can also be tions are most often associated with qualita- conceptualized as a continuous process (Lee, tive or ethnographic work (de Laine, 2000). 1993). Relationships may need to be renegoti- However, there may be similar processes ated throughout the course of research. It is involved in gaining access to secondary data, sometimes unclear exactly what access is documentary sources, or large research sam- required once research is under way. ples. Establishing trust and gaining a working There are a variety of practical and theo- familiarity of the research field are essential retical matters associated with ensuring components of undertaking social research of research access. There are certainly practical all kinds. issues involved in gaining entry to settings or data, and to establishing rapport with Evaluation research participants. These might include formal approaches or applications (by letter Access is not a single event to be undertaken or direct contact), arranging initial meetings only at the beginning of the research process. and providing descriptions of the research However, advice tends to focus on the initial for potential settings and participants. It is negotiation of access, rather than on the main- usually appropriate to offer reassurances of tenance and renegotiation of access over time. confidentiality and trust as part of initial Access is more than seeking formal permission access negotiations. or gaining informed consent. It is also part of Identifying and establishing rapport with a more general process of active engagement key gatekeepers or (informal) sponsors within with settings and social actors, and of recog- a setting can be important to gaining and nizing the need to work at ethical research maintaining access (Hammersley and relationships (Denscombe, 2002). Atkinson, 1995). This can entail developing a How much information to disclose during theoretical or analytical appreciation of the access negotiations is a matter for careful A-Jupp-3373.qxd 3/8/2006 10:41 AM Page 2

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

consideration. It may not be possible to through change interventions involving a inform potential research participants about process of collaboration between researchers the full extent of the research, particularly and participants. The process is seen to be both if an exploratory study is envisaged, with educational and empowering. Action research new research questions emerging over time. should not be confused with evaluation Equally, full disclosure may not be desirable, research which attempts to measure the perhaps because this may jeopardize access, impact of interventions without the active col- or potentially change the behaviour of infor- laboration of participants. mants. Sometimes it is easier to give a fuller account of the aims of the research once Distinctive Features initial access has been secured and a level of rapport and trust established. However, In a systematic review based on UK health- where possible, those concerned should be care settings, Waterman et al. (2001) identi- told about any implications of the research fied two distinguishing features of action from the outset. Deception should certainly research: first, the cyclic process, and second, be avoided. the research partnership. The action research cycle begins with the analysis of a social situation or the identification of a problem. This Amanda Coffey is typically followed by the formulation of some kind of intervention (for example, Associated Concepts: confidentiality, covert nurses in a clinic may change the way they research, dangerous research, ethics, gate- carry out some aspect of professional prac- keeper, impression management, informed tice), which is then evaluated. The planning, consent, organizational research, participant action, reflection and evaluation may lead to observation, reflexivity, research bargain, unob- new rounds of intervention and evaluation. trusive measures, validity of measurement The research element often focuses on the process of change and the achievement of Key Readings planned objectives. The second distinguishing feature of action de Laine, M. (2000) Fieldwork, Participation research is a partnership or collaboration and Practice: Ethics and Dilemmas in between the researcher(s) and the researched. Qualitative Research. Thousand Oaks, CA: In traditional research there is a clear separa- Sage. tion between the researcher(s) and the Denscombe, M. (2002) Ground Rules for researched, which is seen as essential to pre- Good Research. Milton Keynes: Open serving objectivity. In action research, how- University Press. ever, there is a deliberate attempt to involve Hammersley, M. and Atkinson, P. (1995) participants as a way of promoting change and Ethnography: Principles in Practice. London: Routledge. as a device to reduce the social distance Lee, R. M. (1993) Doing Research on Sensitive between researchers and subjects. This Topics. London: Sage. involvement is felt to be both educative and empowering. As Greenwood and Levin (1998: 6) say: ‘Action research aims to increase the ability of the involved community or organization members to control their own destinies ACTION RESEARCH more effectively and to keep improving their capacity to do so.’ Definition It is clear that underpinning these distinguishing features are certain value com- Action research is a type of applied social mitments. The first of these is what could research that aims to improve social situations be called a democratizing motive, which

2 A-Jupp-3373.qxd 3/8/2006 10:41 AM Page 3

ANALYSIS OF VARIANCE (ANOVA)

reverses the conventional relationship between Evaluation should also take into account how researchers and researched, which some action research aims to change the way people action researchers see as elitist and exploita- do things. Its outcomes are not necessarily tive. The second commitment is to partici- ‘findings’ in the conventional sense of theoret- pation. Drawing on a long heritage of ical progress, but in terms of new practices, pragmatist scholarship Peter Reason (1994) changed behaviour patterns or improvements has argued that traditional social science in organizational processes. alienates subjects from their own under- Given the rapidly changing nature of orga- standings of the world. Action research nizational settings and the continuing pressure attempts to transcend this alienation through on organizational members to improve their the active involvement of people in trans- performance it is likely that action research has forming organizations or social groups. The a considerable contribution to make in the third commitment is to forms of inquiry that management of change. As Waterman et al. rely on subjective understanding. Action (2001: 57) suggest, action research has the researchers believe that it is only through potential to go ‘beyond an analysis of the status action and reflection that participants can quo to directly consider questions of “what understand their social situation and through might be” and “what can be”’. formulation of interventions arrive at new understandings. John Newton Evaluation Associated Concepts: applied research, If judged by the standards of conventional emancipatory research, evaluation research, academic research, action research might messy research, participatory action appear to be unscientific. The close and research, policy-related research, practitioner collaborative relationship between researchers research and researched, for example, could be seen as a source of bias because the researcher is Key Readings no longer independent of the research setting. The flexible design features of action Greenwood, D. J. and Levin, M. (1998) research projects might also be an anathema Introduction to Action Research: Social to the mainstream social researcher. In con- Research for Social Change. London: Sage. Reason, P. (ed.) (1994) Participation in Human trast to the clear specification of research Inquiry. London: Sage. questions or hypotheses to be found in Waterman, H., Tillen, D., Dickson, R. and conventional empirical studies, action de Koning, K. (2001) ‘Action research: a research is characterized by a fluid and ongo- systematic review and guidance for assessing process of formulation, implementation, ment’, Health Technology Assessment, 5 (23). adaptation and evaluation in which the identification of stages or project milestones is often difficult. Research design in action research is evolutionary rather than specified beforehand in a research protocol. ANALYSIS OF VARIANCE (ANOVA) Since there is such a marked difference between conventional and action research Definition it would probably make more sense to judge the latter on its own terms. Such a judgement A set of procedures that estimate and attribute would need to consider the action research variance in a data set to different sources process, with its emphasis on participation and determine the probability, under the to bring about change and the role of reflec- null hypothesis, of obtaining the differences tion and self-evaluation in that process. between the variance estimates by chance.

3 A-Jupp-3373.qxd 3/8/2006 10:41 AM Page 4

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

Distinctive Features matrix algebra terms, a commonality is evident. Indeed, the same matrix algebra equa- In ANOVA, variance attributable to differ- tion is able to summarize all three of these ences between groups of scores is compared analyses. As regression, ANOVA and ANCOVA with an average variance attributable to can be described in an identical manner, differences between the scores within each clearly they follow a common pattern. This group. Between-group variance is defined common pattern is the GLM (general linear with respect to differences between the modelling) conception. It is said that regres- group means, and within-group variance is sion, ANOVA and ANCOVA are particular defined with respect to differences between instances of the GLM or that the GLM the individual scores and their group mean. If subsumes regression, ANOVA and ANCOVA. the nature of the groups influences the scores Unfortunately, the ability of the same matrix more than chance fluctuation, then the algebra equation to describe regression, between-group variance estimate will exceed ANOVA and ANCOVA has resulted in the inac- the within-group variance estimate. If this curate identification of the matrix algebra difference between the variance estimates is equation as the GLM. However, just as a sufficiently large, then the null hypothesis particular language provides a means of that all group means are equal is rejected. expressing an idea, so matrix algebra provides The t-test compares the means of two exper- only one notation for expressing the GLM. imental conditions. However, when there are more than two groups or conditions, more than one t-test is needed to compare all of the condi- Andrew Rutherford tions. Unfortunately, the likelihood of obtaining a significant result by chance increases with the Associated Concepts: experiment, general number of statistical tests carried out (that is, linear modelling, inferential statistics, multi- hypotheses tested). A Type 1 error is committed variate analysis when the null hypothesis is rejected erro- neously. Therefore, the Type 1 error rate Key Readings increases with the number of statistical tests applied to any data set. ANOVA was developed Keppel, G., Saufley, W. H. and Tokunaga, H. to assist in the analysis of data obtained from (1992) Introduction to Design and Analysis: agricultural experiments with any number of A Student’s Handbook, 2nd edn. New experimental conditions without any increase York: Freeman. in Type 1 error. ANOVA procedures appropriate Kirk, R. E. (1995) Experimental Design: Procedures for the Behavioral Sciences for an extensive variety of experimental designs . Pacific Grove, CA: Brookes/Cole. are now available (e.g. Kirk, 1995). Rutherford, A. (2001) Introducing ANOVA and Nevertheless, despite ANOVA being devel- ANCOVA: A GLM Approach. London: Sage. oped for use in experimental research, it may be applied to any data organized by categories. However, as with any statistical procedure, interpretation of the ANOVA results will depend upon the data collection ANALYTIC INDUCTION methodology and the conformity of the data to the statistical assumptions underlying the Definition analysis (e.g. Rutherford, 2001). A research strategy of data collection and Evaluation analysis which explicitly takes the deviant case as a starting point for testing models or theo- When regression, ANOVA and ANCOVA ries developed in research. It can be character- (analysis of covariance) are expressed in ized as a method of systematic interpretation

4 A-Jupp-3373.qxd 3/8/2006 10:41 AM Page 5

ANALYTIC INDUCTION

of events, which includes the process of the development of grounded theory. Three generating hypotheses as well as testing types of results are obtained with analytic them. Its decisive instrument is to analyse the induction: forms of activities (how something exception or the case that is deviant to the is done normally), accounts of self-awareness hypothesis. and explanations, and motivational and other reasons for specific behaviours are all Distinctive Features analysed and presented. Analytic induction does not start from This procedure, introduced by Znaniecki in conventional definitions or models of what is 1934, of looking for and analysing deviant studied. As it implicitly assumes that hypothe- cases is applied after a preliminary theory ses, theories and models are not immediately (hypothesis, pattern or model) has been devel- perfect, it is a strategy to refine interpretative oped. Analytic induction, above all, is oriented conclusions from data. By searching for nega- to examining theories and knowledge by tive cases and by testing models and theories analysing or integrating negative cases. against them, it is a strategy to define their The procedure includes the following limits and to make explicit under what time, steps: (1) a rough definition of the phenome- local and social conditions they are not valid. non to be explained is developed; (2) a hypo- Therefore, analytic induction is a way to gen- thetical explanation of the phenomenon is eralize and to delimit qualitative, case-based formulated; (3) a case is studied in the light findings. of this hypothesis to find out whether the Analytic induction has been criticized as hypothesis corresponds to the facts in this it does not – as originally intended by case; (4) if the hypothesis is not correct, Znaniecki (1934) – provide a means for estab- either the hypothesis is reformulated or the lishing causal laws and universals. There are phenomenon to be explained is redefined in a question marks against the generalization way that excludes this case. Thereafter, the of case studies and external validity in gen- researcher actively searches for negative eral. Nevertheless, analytic induction has cases to discredit the hypothesis, model or its own importance as a procedure for assess- typology. ing and developing analyses by the use of Practical certainty can be obtained after a negative cases. small number of cases has been studied, but the discovery of each individual negative case by the researcher or another researcher Uwe Flick refutes the explanation and calls for its refor- mulation. Each negative case calls for the Associated Concepts: case study method, redefinition of concepts and/or the reformu- constant comparative method, deviant case lation of hypotheses. Further cases are stud- analysis, grounded theory, hypothesis, induc- ied, the phenomenon is redefined and the tion, validity hypotheses are reformulated until a universal relation is established. Key Readings Evaluation Cressey, D. R. (1950) ‘Criminal violations of financial trust’, American Sociological Analytic induction is not based on enumera- Review, 15: 733–48. tive argumentation. As it focuses on the Flick, U. (2002) An Introduction to Qualitative single deviant case to test a more or less gen- Research, 2nd edn. London: Sage. eralized model, it is a genuinely qualitative Lincoln, Y. S. and Guba, E. G. (1985) way of assessing the stability and limitations Naturalistic Inquiry. London: Sage. of research findings. With its emphasis on Znaniecki, F. (1934) The Method of Sociology. testing theories it goes one step further than New York: Farrar & Rinehart.

5 A-Jupp-3373.qxd 3/8/2006 10:41 AM Page 6

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

and social practices that draw directly or ANTI-RACISM RESEARCH indirectly on the conviction that there are racial groups which have distinct physical or Definition cultural characteristics which are usually but Research that focuses on studying the belief not only defined in negative terms. Social (and social practices) that there are racial scientists distinguished racial prejudice, groups which have distinct physical or which is hostile and negative attitudes cultural characteristics usually, but not only, towards minority races and ethnic groups, defined in negative terms. racial stereotyping and racial discrimination, defined as unjustified negative or harmful Distinctive Features action towards minority races and ethnic groups. The conclusions were that racial prej- The desire to categorize people into racial udice and racial stereotyping premised upon types based on physical appearance (or skin notions of white superiority were widespread colour) has a long tradition in Western society and pervasive in the United States and the (see Banton, 1987). Darwin’s evolutionary UK. There was also evidence of racial dis- theory of 1859 instituted the idea that there crimination across a variety of social and was a racial hierarchy based on different economic settings, and of racial violence, the species or races with natural, discernible most extreme form of discrimination. biological characteristics. Darwin also talked Researchers realized that one problem with of the ‘civilized races of man’ exterminating survey and questionnaire data was that it only and replacing the ‘savage races’ and was at provided information on the racism that peo- one with Galton in demanding control of how ple were willing to admit to. In addition, as a different races breed (eugenics). For many result of social changes, it became apparent decades, in a variety of national settings, that fewer people were willing to express positivist ‘race scientists’ researched for the overtly racist attitudes to researchers. This genetic base of race. These investigations posed a key question: were white people were originally encouraged by European becoming less prejudiced or were an increas- colonization and the development of the ing number of people much more knowing European and American slave trade and and therefore guarded about revealing their reached their high point in Nazi Germany, true sentiments on racial matters to with extensive government resources being researchers? Researchers concluded that allocated to verify scientifically the superiority subtle more flexible forms of racism had of the Ayran race. However, social scientists replaced the old-fashioned racism expressed increasingly argued that we needed to put by previous generations of respondents quotation marks around the word ‘race’ to (McConahay et al., 1981). This is likely to be indicate that we are dealing with a social the case because in several countries the construction of individuals and groups, rather direct expression of racist views has been than an established unproblematic scientific outlawed. fact (Wetherell, 1996). This rejection of the One of the key proponents of this argu- scientific validity of the concept of race ment is US social psychologist David Sears opened up the possibility of more sophisti- (see Sears, 1988; Sears et al., 1999). His work cated research designs examining how and on the new racism describes a more elusive, why ‘race’ matters. abstract symbolic language of race that avoids A substantial programme of research has blatantly negative racist statements in favour been carried out based on standardized social of political code words and symbols. This new surveys and tests with the acknowledgement racism being expressed is partly based on a in the United States and the UK in the 1970s view of racial discrimination as being out- that white racism was a serious social prob- dated and puts the onus of achievement and lem. ‘Racism’ can be said to refer to beliefs equality on blacks and other minority groups.

6 A-Jupp-3373.qxd 3/8/2006 10:41 AM Page 7

ANTI-RACISM RESEARCH

The new racism asserts that it is black people’s A separate sociological research programme own deficiencies that are the cause of their has focused on analysing official data to identify problems, not the history of slavery segrega- racial and ethnic inequalities, most noticeably tion, discrimination, prejudice and racism in the labour market, government agencies, that is assumed to have come to an end. The housing, education, the media and criminal new racism is thought to be most obvious in justice. This has concentrated minds on struc- white people’s views on affirmative action tural and systemic rather than individual and and various social problems. Conversations group dimensions of racism. In the UK context on these topics are often framed by an unspo- this debate received renewed focus as a result ken subtext of racist attitudes and negative of the conclusion of the judicial inquiry into the associations. Researchers have devised sophis- murder of black teenager Stephen Lawrence ticated modern racial prejudice scales to that the murder investigation was hindered examine various covert forms of contempo- by ‘institutionalized racism’ within the rary racism (see Sears et al., 1999; see also Metropolitan Police (see Macpherson, 1999). Bobo, 1999; Dovidio and Gaertner, 1998). Originally coined in 1967 by the US Black Wetherell and Potter (1992) have argued Panther Stokely Carmichael and Charles V. that the modern racism approach in social Hamilton, the inquiry defined institutional psychology is limited by its dependence on racism as: ‘the collective failure of an organiza- the prejudicial model of racism. They argue tion to provide an appropriate and professional that racism should not be narrowly equated service to people because of their colour, cul- with a particular psychological complex of ture, or ethnic origin. It can be seen or detected feelings, thoughts and motives. Their inter- in processes, attitudes and behaviour which views focused on how the taken-for-granted amount to discrimination through unwitting discourse of white New Zealanders rational- prejudice, ignorance, thoughtlessness and racist izes and justifies Maori disadvantage, how stereotyping which disadvantage minority inequality is normalized and rendered ethnic people’ (Macpherson, 1999: 28). This unproblematic and conflicts subdued. finding has provided a challenge to social scien- Sociologists, suspicious of the psychologizing tists to develop the methodological tools to of racism, have undertaken ethnographic research the British government’s programme case studies of community attitudes towards of action to identify and eradicate ‘institutional race or have accidentally uncovered everyday racism’. Given the nature of the Macpherson talk about race to be a central issue in the inquiry, it is not surprising that a considerable course of research. Some of the most inter- focus of research is on the rules, procedures esting if controversial sociological studies and guidelines that produce discriminatory out- have involved research on white racists comes in the criminal justice system. groups (see Blee, 2002). A substantial body of sociological research findings also exists on Evaluation the media and racism. Research on race and the news media, for understandable reasons, Contemporary racism takes a bewildering tends to concentrate on studying how visible variety of forms and is spawning a new ethnic minorities are negatively portrayed generation of research questions that require and crudely stereotyped by white-controlled a rethinking of conventional methodologies. media organizations. A particular focus is This is also generating critical writings on the analysing the discourses used by sections of ethical dilemmas associated with the doing of the news media to racialize public debate anti-racism research (Twine and Warren, about law and order and other social prob- 2000). Researchers need, for example, to lems (Law, 2002). The possibilities for work through how their racial and ethnic researching media and racism have increased backgrounds influence the analytical lens dramatically with the development of the through which they view their research Internet. subjects. There are also ‘who’s side are we

7 A-Jupp-3373.qxd 3/8/2006 10:41 AM Page 8

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

on?’ questions and the temptation to expose Sears, D., Sidanius, J. and Bobo, L. (1999) the views of racist respondents to advance a Racialised Politics: The Debate about particular political agenda. Finally, there is Racism in America. Chicago: University of the controversial question of whose racism is Chicago Press. researched? To date the primary focus has Twine, F. W. and Warren, J. W. (eds) (2000) been on the racism of lower socio-economic Racing Research and Researching Race: white groups. Researching institutional Methodological Dilemmas in Critical Race racism will require, amongst other things, Studies. New York: New York University accessing powerful government agencies and Press. Wetherell, M. (1996) ‘Group conflict and the multinational corporations. In addition, there social psychology of racism’, in M. is also a lack of research on what we might Wetherell (ed.), Identities, Groups and describe as the rationales and dynamics of Social Issues. London: Sage. pp. 175–234. non-white racism. Wetherell, M. and Potter, J. (1992) Mapping the Language of Racism: Discourse and the Eugene McLaughlin Legitimation of Exploitation. London: Sage.

Associated Concepts: critical research, discourse analysis, discursive psychology, emancipatory research, Internet research, media APPLIED RESEARCH analysis, official statistics, politics and social research Definition

Key Readings Research that focuses on the use of knowledge rather than the pursuit of knowledge for Banton, M. (1987) Racial Theories. its own sake. A motivation behind applied Cambridge: Cambridge University Press. research is to engage with people, organiza- Blee, K. M. (2002) Inside Organised Racism: tions or interests beyond the academic disci- Women in the Hate Movement. Berkeley, pline and for knowledge to be useful outside CA: University of California Press. the context in which it was generated. Bobo, L. (1999) ‘Race, interests and beliefs about affirmative action’, American Behavioural Scientist, 41 (7): 985–1003. Distinctive Features Dovidio, J. F. and Gaertner, S. L. (1998) ‘On This engagement with the ‘outside world’ (for the nature of contemporary prejudice: the example, government departments, commer- causes, consequences and challenges of cial organizations, pressure groups) gives aversive racism’, in J. L. Eberhardt and applied research some distinctive characteris- S. T. Fiske (eds), Confronting Racism: The tics. Bickman and Rog (1998), for example, Problem and the Response. London: Sage. argue that basic and applied research differ in Law, I. (2002) Race in the News. London: purposes, context and methods. Although the Palgrave. Macpherson, Sir William (1999) The Stephen differences are presented in dichotomous Lawrence Inquiry. London: TSO. terms, the authors suggest that in reality they McConahay, J. B., Hardee, B. B. and Batts, V. should be seen as continua. The differences (1981) ‘Has racism declined in America? of purpose can be described in terms of the It depends upon who is asking and what goals of knowledge production. For the basic is asked’, Journal of Conflict Resolution, 25: researcher the production of knowledge is an 563–78. end in itself whereas for the applied Sears, D. (1988) ‘Symbolic racism’, in P. Katz researcher knowledge is used to further other and D. Taylor (eds), Eliminating Racism. ends or goals. For example, a basic researcher New York: Plenum Press. might be interested in understanding how

8 A-Jupp-3373.qxd 3/8/2006 10:41 AM Page 9

APPLIED RESEARCH

customers make purchasing decisions with other economic phenomena) which frequently no other concern than the process of decision inform government decision makers. Outside making itself. An applied researcher might be government, university-based economists, have interested in the same kind of behaviour but a long-standing reputation for their forecasts on be primarily concerned with how the find- economic growth (Begg and Henry, 1998). ings of the research could be used, say, to Psychology, too, has always recognized the increase the sale of certain commodities. practical application of its knowledge base. In This points to an important difference in World War II psychological assessment tech- the context of applied and basic research. niques were used to recruit pilots, and after Applied research is often initiated by some- the war returning members of the armed one other than the researcher (it could be a forces brought home emotional problems government department, a pressure group or which were managed by a growing number a commercial organization). Basic research, of clinical psychologists. Today, the discipline in contrast, is more often than not conducted of psychology has become institutionalized as by the person(s) who formulated the topic or a professional activity with a variety of research question in the first place. Where applied specialisms, such as clinical, educa- research is ‘other initiated’, the researcher tional, forensic and industrial. Each specialism may have less control over various aspects of has its own conferences and peer-reviewed the research process (for example, the design journals under the umbrella of the relevant of the study, its scope and timeframe and, national association for example, the perhaps, whether the results will be made American Psychological Association and the publicly available). Where research is ‘self- British Psychological Society. initiated’ there is greater opportunity to ‘set Sociology, in contrast, has been slower to one’s own agenda’ and be less constrained develop as a form of practice, though this is about how the research is managed and perhaps more the case in the UK than the conducted. Even so, the standard conditions United States. Even in the US, however, the attaching to externally funded projects mean President of the American Sociological that basic researchers are also subject to a Association opened his 1980 presidential number of ‘other-initiated’ requirements (for address with the remark that ‘the stance of example, compliance with ethical guidelines). our profession toward applied work … has There are no specific research methods asso- been one of considerable ambivalence’ (Rossi, ciated with either basic or applied research, but 1980: 890). Various explanations have been Bickman and Rog (1998) suggest that applied put forward for this stance. The ‘opt out’ view researchers are more likely to pay greater atten- argues that sociologists have always seen their tion to issues of external validity. This does not discipline primarily as a form of scholarship imply a lack of attention to internal validity on in which the subject is seen as an accumula- the part of applied researchers; it is simply a tion of literature to be learned, debated and matter of usefulness. Applied researchers will developed. As such, sociologists have not been want to show that their results can be used to concerned to promote or advertise their work address a problem or issue in the ‘real world’. to others. The result, Martin Albrow has suggested, is that sociology has become some- Evaluation thing of a ‘subterranean mystery’ (Albrow, 2000); a craft rather than a profession. The Different social science disciplines have ‘sell out’ view argues that sociology’s identity engaged in applied research with different is borne out of radicalism. The ‘debunking levels of enthusiasm. Some, like economics, motif’ inherent in some views of the sociolog- have established a close relationship between ical enterprise has meant that sociologists the development of theory, techniques of mea- have been reluctant to become embroiled surement and statistical analysis in order to in applied research because this would entail explain the workings of the economy (and an endorsement of existing power structures

9 A-Jupp-3373.qxd 3/8/2006 10:41 AM Page 10

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

and would turn sociologists into state-funded listings of individuals, groups, institutions or ‘apparatchiks’. whatever. Sometimes a further sub-sample is These ‘anti-applied’ views are not taken in which case the phrase multi-stage completely dominant, however. A volume of sampling is used. essays published by the British Sociological Association called for an ‘active sociology’ Distinctive Features (Payne and Cross, 1991). Its editors argued that no academic discipline can afford to With area sampling, data are collected from become too introspective or passive towards or about all individuals, households or other the world outside academia. What is units within the selected geographical areas. required, they say, is ‘an applied sociology … Area sampling is sometimes referred to as [which] … will be an active participative soci- block sampling, especially in the United ology which engages with society’ (1991: 2). States. It is also a form of cluster sampling.

Evaluation John Newton Area sampling is used a great deal in coun- Associated Concepts: action research, criti- tries where there are no adequate population cal research, econometrics, evaluation lists, which are replaced instead by maps. research, emancipatory research, policy- However, the boundaries on such maps must related research, politics and social research, be clearly defined and recognizable, both at practitioner research the stage of sampling and at the stage of data collection. This form of sampling is not Key Readings appropriate where the aim of the research is to follow a cohort of individuals over time Albrow, M. (2000) Review of Steele, S. F., because of the population changes within the Scarisbrick-Hauser, A. M. and Hauser, areas that have been selected in the sample. W. J., ‘Solution centred sociology: addressing problems through applied sociology’, Sociology, 34 (3): 596–7. Victor Jupp Begg, I. and Henry, S. G. B. (1998) Applied Economics and Public Policy. Cambridge: Associated Concepts: cluster sampling, Cambridge University Press. cohort study, cross-sectional survey, longitu- Bickman, L. and Rog, D. J. (1998) Handbook dinal study, sampling, social survey of Applied Social Research Methods. London: Sage. Payne, G. and Cross, M. (eds) (1991) Sociology Key Reading in Action: Applications and Opportunities for Moser, C. A. and Kalton, G. (1971) Survey the 1990s. London: Macmillan. Methods in Social Investigation. London: Rossi, P. H. (1980) ‘The challenge and oppor- Heinemann. tunities of applied social research’, American Sociological Review, 45 (Dec.): 889–904.

ATLAS.ti

AREA SAMPLING Definition

Definition One of a number of Computer Assisted Qualitative Data Analysis Software (CAQDAS) A form of sampling in which the clusters that programs designed to facilitate the manage- are selected are drawn from maps rather than ment and analysis of qualitative data. It was

10 A-Jupp-3373.qxd 3/8/2006 10:41 AM Page 11

ATLAS.TI

originally developed as an exercise to support Searches can be filtered in a variety of ways grounded theorizing. and the Supercode tool allows search expressions to be easily saved and re-run. Some of Distinctive Features the more sophisticated search operators are based on the nature of links between codes. Like all CAQDAS programs, Atlas.ti is a tool A number of tools facilitate the organiza- for facilitating analysis rather than a method tion and management of team projects. These in itself and therefore can feasibly be used include a tool to merge projects, a facility to support a number of methodological or allowing shared data to be accessed by theoretical approaches. It supports the man- several projects (‘Hermeneutic Units’), sup- agement and analysis of textual, audio and port for East Asian and right-to-left languages, visual data and enables the creation of easy ways to back up and move projects and (semantic) networks to facilitate theory build- XML project export. ing processes. Atlas.ti also enables the integration of As well as the code and retrieve, organiza- quantitative and qualitative aspects of tional (for example, demographic variables) projects offering a flexible word count facil- and search (for example, Boolean, Proximity ity, the results of which can easily be etc.) tools available in most CAQDAS programs, exported to statistical or spreadsheet pack- Atlas.ti provides a number of additional tools ages. It is also possible to import (and export) which increase flexibility and facilitate analytic demographic information. development. The code margin view is fully interactive, Evaluation enabling codes and memos to be easily and quickly accessed, edited, merged, replaced The evaluation of any CAQDAS package must and linked. Not only can codes be grouped take into consideration a number of factors – into (for example, theoretical) families, but such as methodology, theoretical approach, the new Super Families tool enables addi- type of data, size of data set, project aims and tional hierarchical collections to be created requirements. Therefore certain packages and a particular code can belong to any may be particularly useful for certain types of number of families. The Autocoding tools are studies and researchers are advised to inves- also particularly flexible. tigate the various options before choosing. In addition to being able to directly code Atlas.ti is clearly amongst the more sophis- multi-media data, objects (for example, Excel, ticated options available, offering a variety of PowerPoint) can be embedded into rich text powerful and flexible means by which to format (rtf.) files. These objects can then be explore, work with and interrogate qualita- edited in-context as the functionality of the tive data. As such, it enables analysis to go corresponding application becomes available beyond code and retrieve processes. within Atlas.ti. Perhaps its most frequently cited advan- The Networking tool is extremely versatile, tage is that, in comparison to other CAQDAS allowing quotations (segments of text), codes, software, Atlas.ti keeps the researcher very documents and memos to be linked to each ‘close’ to the data. For instance, when navi- other in a variety of ways. It is also gating around coded data, if required, quota- possible to create hyperlinks between quotations will be illustrated in their source tions to, for example, track a story or sequence context – allowing easy in-context re-coding. of events within or between data files. As a Atlas.ti is therefore often seen by researchers presentation tool, the Networking facility is as particularly well suited for ‘grounded’ flexible, allowing, for example, quotations to approaches to qualitative analysis. be displayed as illustrations of research find- Atlas.ti is a multi-faceted software and the ings or theoretical processes to be visualized. user may employ only a small proportion of The Query tool allows both simple and the tools available for any given project. Its sophisticated interrogation of the data set. size and flexibility are both its advantage and

11 A-Jupp-3373.qxd 3/8/2006 10:41 AM Page 12

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

its disadvantage. It can be used easily at a occurs in longitudinal research when the basic level but possibly needs a more experi- subjects studied drop out of the research for enced and confident researcher to make inno- a variety of reasons, which can include: vative use of specific tools for investigating unwillingness of subjects to continue to par- and representing complex relationships. ticipate in research, difficulties in tracing original respondents for follow-up (for example, because of change of address) and non- Christina Silver availability for other reasons (for example, death, serious illness). A survey of major lon- Associated Concepts: CAQDAS, coding, The gitudinal studies in the United States found Ethnograph, grounded theory, NUD*IST, that the average attrition rate was 17 per cent qualitative research, QSR NVivo (Capaldi and Patterson, 1987, cited in Sapsford and Jupp, 1996). Therefore, quite a lot of Key Readings cases may be lost. A second sense of attrition relates to the Atlas.ti website http://www.atlasti.de/ loss of data from secondary sources in the CAQDAS Networking Project website process of collection. A commonly used http://caqdas. soc.surrey.ac.uk/ Coffey, A., Holbrook, B. and Atkinson P. example is the case of crime statistics, which (1996) ‘Qualitative data analysis: tech- are the end product of processes of decisions nologies and representations’, Sociological by victims to report crimes to the police, Research Online, 1 (1). police decisions about recording and deci- Fielding, N. and Lee, R. (eds) (1991, 2nd edn sions by various agencies in the criminal 1993) Using Computers in Qualitative justice process which may or may not result Research. London: Sage. in prosecution, conviction and the imposition Kelle, U. (ed.) (1995) Computer-Aided of a particular sentence. Using the British Qualitative Data Analysis: Theory Methods Crime Survey (BCS) (a random household and Practice. London: Sage. survey of criminal victimization) as a base Lewins, A. (2001) ‘CAQDAS: Computer line, it is estimated that only 45 per cent of Assisted Qualitative Data Analysis’, in the crimes recorded by BCS are reported to N. Gilbert, Researching Social Life, 2nd the police, only 24 per cent are recorded as edn. London, Sage. pp. 302–23. crimes, 2 per cent result in a conviction and Weitzman, E. and Miles, M. (1995) A Software 0.3 per cent result in a prison sentence Source Book: Computer Programs for (Home Office, 1999). Thus there is a consid- Qualitative Data Analysis. Thousand Oaks, erable loss of data between the base source CA: Sage. and subsequent measures.

Evaluation The problem of attrition is not merely the ATTRITION problem of a reduction in sample size but, more importantly, it raises the possibility of Definition bias. Those who ‘drop out’ may have particu- The ‘wearing away’ or progressive loss of lar characteristics relevant to the research data in research. aims (for examples, see Sapsford and Jupp, 1996: 9). The same point applies to the crime Distinctive Features statistics example: certain crimes may be less likely to be reported, recorded or result in Attrition occurs when cases are lost from a a conviction. For example, Kelly (2000), sample over time or over a series of sequen- reviewing studies of attrition in rape cases, tial processes. One form of sample attrition found that fewer than 1 per cent of these cases

12 A-Jupp-3373.qxd 3/8/2006 10:41 AM Page 13

AUDITING

result in conviction. Researchers thus need to Also, a form of evaluation called programme be aware of the problem of attrition and the and performance auditing is routinely threats to representativeness that can result. performed at state and national levels. As defined by the Comptroller General of the United States (US GAO, 1994), a performance Maggie Sumner audit is ‘an objective and systematic examination of evidence … of the performance of a gov- Associated Concepts: bias, cohort study, ernment organization, program, activity, or longitudinal study, official statistics function in order to provide information to improve public accountability and facilitate Key Readings decision-making’. A programme audit is a sub- category of performance auditing in which a Capaldi, D. and Patterson, G. R. (1987) ‘An key objective is to determine whether pro- approach to the problem of recruitment and gramme results or benefits established by the retention rates for longitudinal research’, Behavioural Assessment, 9: 169–77. legislature or other authorizing bodies are being Home Office (1999) Digest 4: Information on achieved. Evaluators of social programmes and the Criminal Justice System in England and performance auditors share a professional Wales. London: Home Office. interest in establishing their independence and Kelly, L. (2000) ‘A war of attrition: recent warranting the credibility of their judgements. research on rape’, Trouble and Strife, 40 Second, an auditing procedure has been (http://www.cwasu.org/warattrition1.htm). suggested as a means to verify the depend- Nathan, G. (1999) ‘A review of sample attri- ability and confirmability of claims made in a tion and representativeness in three longi- qualitative study (Lincoln and Guba, 1985; tudinal surveys’, Government Statistical Schwandt and Halpern, 1988). A researcher is Service Methodology, Series No. 13. advised to maintain an audit trail of evidence London: Government Statistical Service. documenting the data, processes and product Sapsford, R. and Jupp, V. (1996) Data (claims) of the inquiry. A third-party inquirer Collection and Analysis. London: Sage. then examines that the audit trail can attest to the appropriateness, integrity and depend- ability of the inquiry process and also the extent to which claims made are reasonably AUDITING grounded in the data. Third, auditing has entered the scene of social science theory. At issue is the prolifer- Definition ation of audit practices in all spheres of A procedure whereby an independent third human activity – management, education, party systematically examines the evidence social services, healthcare and so forth – of adherence of some practice to a set of influenced largely by the ideology of New norms or standards for that practice and Public Management (NPM). NPM empha- issues a professional opinion. sizes a programmatic restructuring of organizational life and a rationality based on Distinctive Features performance standards, accountability and monitoring. By being submitted to formal The concept and practice of auditing is mani- audit procedures the work of organizations is fest in several ways in the social sciences. First, held to be more transparent and accountable. in social programme evaluation, the general idea of auditing has informed the process of Evaluation meta-evaluation – a third-party evaluator examines the quality of a completed evaluation Whether auditing and evaluation are (or against some set of standards for evaluation. ought to be) comfortable bedfellows can be

13 A-Jupp-3373.qxd 3/8/2006 10:41 AM Page 14

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

debated. Some observers have argued that measurement that strongly shapes conceptions evaluation and performance auditing differ in of knowledge, politics and ethics. For example, the ways each conceives of and accomplishes some scholars argue that auditing (and associ- the aim of assessing value. Some of the dif- ated practices such as total quality manage- ferences noted between the two practices ment, performance indicators, league tables, include the following. First, auditors address results-oriented management, monitoring sys- normative questions (questions of what is, in tems) is not simply a set of techniques but a light of what should be) while evaluators are system of values and goals that becomes more concerned with descriptive and impact inscribed in social practices thereby influenc- questions. Second, auditors work more inde- ing the self-understanding of a practice and its pendently of the auditee than evaluators do role in society. Thus, to be audited, an organi- with their clients. Third, auditors are more zation (or practice like teaching or providing exclusively focused on management objec- mental health care) must transform itself tives, performance and controls than evalua- into an auditable commodity – auditing thus tors. Fourth, auditors work with techniques reshapes in its own image those organizations for generating evidence and analysing data and practices which are monitored for perfor- that make it possible to provide quick feed- mance (Power, 1997). Others argue that audit back to auditees, while evaluations often culture or society promotes the normative ideal (though not always) have a longer time frame. that monitoring systems and accountability Fifth, although both auditors and evaluators ought to replace the complex social-political base their judgements in evidence, not on processes entailed in the design and delivery of impressions, and both rely on an extensive kit social and educational services and the of tools and techniques for generating evi- inevitably messy give-and-take of human inter- dence, they often make use of those tools in actions. Still others contend that the growing different ways. Finally, auditors operate influence of an audit culture contributes to under statutory authority while evaluators the disappearance of the idea of publicness as work as fee-for-service consultants or as traditional public service norms of citizenship, university-based researchers. Other observers representation equality, accountability, impar- have argued that the practices of auditing and tiality, openness, responsiveness and justice are evaluation, although they often exist inde- being marginalized or replaced by business pendently of one another, are being blended norms like competitiveness, efficiency, produc- together as a resource pool for decision makers tivity, profitability and consumer satisfaction. and managers in both public and private organizations. In this circumstance, an amal- gamated picture is emerging of professional Thomas A. Schwandt objectives (for example, placing high value on independence; strict attention to documenta- Associated Concepts: applied research, tion of evidence), purpose (for example, com- cost–benefit analysis, critical research, evalu- bining normative, descriptive and impact ation research, meta-analysis, performance questions) and methodologies (for example, indicator, politics and social research, process making use of a wide range of techniques). mapping, secondary analysis, social indicators A variety of criticisms based in empirical and conceptual investigations are directed at the audit society, audit culture, or the culture of accountability as the latest manifestation of the Key Readings infiltration of technological, means–end and Lincoln, Y. S. and Guba, E. G. (1985) instrumental rationality into the forms of every- Naturalistic Inquiry. Beverly Hills, CA: Sage. day life. Auditing is viewed as an example of a Power, M. (1997) The Audit Society: Rituals of key characteristic of modernity – that is, the Verification. Oxford: Oxford University drive for efficiency, perfection, completion and Press.

14 A-Jupp-3373.qxd 3/8/2006 10:41 AM Page 15

AUTOETHNOGRAPHY

Schwandt, T. A. and Halpern, E. S. (1988) their own subjectivity and life experiences Linking Auditing and Metaevaluation. (usually within the context of fieldwork). For Newbury Park, CA: Sage. literary critic Mary Louise Pratt (1992), US General Accounting Office (1994) autoethnography is a mode of self-and group Government Auditing Standards. representation on the part of colonial or post- Washington, DC: US General Accounting colonial subjects that is informed by repre- Office. sentations of them by others who are more Wisler, C. (ed.) (1996) ‘Evaluation and audit- dominant. In literary theory, autoethnogra- ing: prospects for convergence’,New phy is frequently viewed as a form of Directions for Evaluation, No. 71. San counter-narrative. Francisco, CA: Jossey–Bass. Memoirs, life histories and other forms of self-representation are analysed as autoethno- graphies when they deal with topics such as transnationalism, biculturalism or other forms AUTOETHNOGRAPHY of border crossing, and local cultural practices. Ethnographers make use of such texts to Definition explore these issues in the context of life experiences and cultural constructions of these. A form of self-narrative that places the self Ethnographers have also adopted the term within a social context. It includes methods autoethnography to label forms of self- of research and writing that combine autobio- reflexivity. For example, Arthur Bochner and graphy and ethnography. The term has a dual Carolyn Ellis (2002) are sociologists who advo- sense and can refer either to the ethnographic cate an ‘emotional sociology’ that incorporates study of one’s own group(s) or to autobio- personal narrative as a method to avoid the graphical reflections that include ethno- objectification of more scientific methods of graphic observations and analysis. research by erasing boundaries between the self of the researcher and that of the Distinctive Features researched. Anthropologists such as Reed- Danahay (1997) use the concept of autoethno- In autoethnography there is both self- graphy to analyse the uses of self-writing reference and reference to culture. It is a among anthropologists and among ‘natives’. method that combines features of life history and ethnography. The term autoethnography Evaluation has been used both by qualitative researchers in the social sciences, who emphasize the Autoethnography, as a method of research and connections between ethnography and writing, requires reflective and critical autobiography, and by literary critics who are approaches to understandings of social and mainly concerned with the voices of ethnic cultural life and the relationship between the autobiographers. Autoethnography can be self and the social. Autoethnographic texts are associated with forms of the following: first, most compelling when they synthesize objec- native anthropology or self-ethnography – in tive (outsider) and subjective (insider) points of which those who previously were the objects view, rather than privileging the latter. of anthropological inquiry come to undertake Because there is such a wide range of work ethnographic research themselves on their currently labelled ‘autoethnography’, ranging own ethnic or cultural group; second, ethnic from creative nonfiction to ethnic autobiogra- autobiography – in which autobiographers phy to the testimonials of postcolonial sub- emphasize their ethnic identity and ethnic jects, the task of evaluation is challenging and origins in their life narrative; and third, auto- depends in large part on the intent of the biographical ethnography – a reflexive writer/researcher. Questions are frequently approach in which ethnographers analyse raised about authenticity and voice in such

15 A-Jupp-3373.qxd 3/8/2006 10:41 AM Page 16

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

texts, regarding the authority of the speaker. context, this form of research and writing is Personal accounts of fieldwork are, for example, critically evaluated on the basis of how well often used to evoke the ‘ethnographic author- it synthesizes the subjective experience of ity’ (Clifford, 1983) of the ethnographer. participant(s) in social and cultural life and Autoethnography calls attention to issues of the structural conditions in which their lives power, and can be most effective when it fore- take place. grounds the complex and nuanced relationships between researcher and researched, dominant and subordinate, in the context of Deborah Reed-Danahay individual experience and socio-cultural structures of beliefs and control. Associated Concepts: cultural research, As a method, autoethnography is frequently ethnography, intellectual craftsmanship, life contrasted to ‘objective’ scientific or positivist history interviewing, oral history, reflexivity methods that call for the researcher to set him- or herself apart from the object(s) of research. In ethnography, this would require a stance of Key Readings distance from the self, from those who are Bochner, A. P. and Ellis, C. (eds) (2002) being studied and from the social contexts of Ethnographically Speaking: Autoethnography, the research situation. Those who advocate an Literature, and Aesthetics. Walnut Creek, autoethnographic approach argue that reflex- CA: Alta Mira Press. ivity about oneself and about the research sit- Clifford, James (1983) ‘On ethnographic uation, that is, being aware of one’s position in authority’, Representations, 2: 118–46. the context of research rather than denying it, Pratt, M. L. (1992) Imperial Eyes: Travel is vital to a full understanding and is not com- Writing and Transculturation. London: pletely at odds with forms of ‘truth’ or valid- Routledge. ity. Although there are variations in the degree Reed-Danahay, D. (ed.) (1997) Auto/ to which autoethnographers emphasize their Ethnography: Rewriting the Self and the own experience or that of the ethnographic Social. Oxford: Berg.

16 B-Jupp-3373.qxd 3/8/2006 10:41 AM Page 17

uncontested central component of social BIAS science methodology. In recent years, however, especially under the influence of construction- Definition ism and postmodernism, there has been grow- Generally regarded as a negative feature of ing debate, especially among qualitative research, as something that can and should researchers, about the meaning and usefulness be avoided; occasionally the term is used in a of these terms (see, for example, Lather, 1986; neutral or even a positive sense, referring Kvale, 1989; Harding, 1992; Altheide and simply to the fact that the researcher has Johnson, 1994). In part, this reflects the fact adopted a particular angle of vision. that they had previously often been interpreted in ways that depended on a form of positivism that is now largely discredited. The latter pre- Distinctive Features sented research, when properly executed, as Even in its negative sense, there are broader producing conclusions whose validity follows and narrower interpretations of the term. automatically from the ‘givenness’ of the data Sometimes it refers to any systematic deviation on which they are based. from the truth, or to some deformation of On this view, the course that inquiry should research practice that produces such deviation. take is clearly defined and, as a result, devia- Thus, quantitative researchers refer to ‘mea- tion from it – whether caused by prior com- surement bias’ and to ‘sampling bias’, by which mitments or by some other source of error – is they mean systematic failure in measurement also straightforwardly identifiable. What is or sampling procedures that produces erro- required to avoid bias is for researchers to be neous results. The contrast here is with random objective; in other words, they must pursue (or haphazard) error. However, another influen- research in the way that ‘anyone’ would pursue tial usage of the term ‘bias’ refers to a particu- it who was committed to discovering the truth, lar source of systematic error: a tendency on whatever their personal characteristics or the part of researchers to collect data, and/or to social position, appealing only to data that are interpret and present these data in such a way observable by ‘anyone’. as to favour false results that are in line with The influence of positivism meant that a their presuppositions and/or their political and clear distinction was not always drawn practical commitments. This may consist of a between, on the one hand, a researcher having positive tendency towards a particular, but potentially biasing commitments, for example false, conclusion; or it may mean the exclusion particular political views, and, on the other, from consideration of some set of possible con- these commitments impacting negatively on clusions that happens to include the truth. the research process. In other words, researchers were (and sometimes still are) Evaluation described as biased simply because they have commitments pertaining to the topic on ‘Bias’ is part of a set of terms – ‘validity’ and which research is being carried out. This fol- ‘objectivity’ are others – that were once an lows from the false assumption that, in order B-Jupp-3373.qxd 3/8/2006 10:41 AM Page 18

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

to be objective, a researcher must strip away were relying had an unacceptable chance of all his or her assumptions until the bedrock of being wrong and might therefore lead them empirical givens is reached, and then build up astray. In short, they did not take the proper true knowledge from that foundation solely methodological precautions to avoid error, for by logical means. Modern philosophy of sci- example by assessing the relative validity of ence has rendered this view indefensible. alternative interpretations. It is worth pointing out, however, that, In conclusion, then, while the abandon- in effect, some of the critics of the concepts ment of positivism requires us to recognize of truth, objectivity and bias have taken over that research will inevitably be affected by this idea. They rely on it in their denial that the personal and social characteristics of the research can ever be unbiased (see, for exam- researcher, and that this can be of positive ple, Gitlin et al., 1989). If we adopt a more real- value as well as a source of systematic error, istic conception of what objectivity and bias it does not require us to give up the guiding involve, this sceptical conclusion can be principle of objectivity: in other words, the avoided. Nevertheless, it must be recognized commitment to avoiding bias. that, once we abandon positivism, error becomes much more difficult to identify. Given that there is no bedrock of absolute givens, and Martyn Hammersley no method that guarantees valid findings, what constitutes systematic deviation from the ratio- Associated Concepts: autoethnography, nal pursuit of inquiry (that is, bias) is not constructionism, error, measurement, posi- always clear. In the course of inquiry about any tivism, postmodernism, qualitative research, topic, we necessarily take other matters for quantitative research, validity granted; and in the absence of a foundation of absolute givens these can only be matters Key Readings about which we believe our knowledge to be sound but less than absolutely certain. If we Altheide, D. L. and Johnson, J. M. (1994) did not make such assumptions, we would ‘Criteria for assessing interpretive validity have no ground at all on which to stand, and in qualitative research’, in N. K. Denzin we would indeed lapse into a thoroughgoing and Y. S. Lincoln (eds), Handbook of scepticism. Judgements have to be made, then, Qualitative Research. Thousand Oaks, CA: about the validity of presuppositions; but in the Sage. pp. 485–99. absence of any prospect of absolute proof. Gitlin, A. D., Siegel, M. and Boru, K. (1989) Where, previously, procedural error was ‘The politics of method: from Leftist thought to be a matter of logic, it now becomes ethnography to educative research’, Inter- deviance from communal judgements about national Journal of Qualitative Studies in Education, 2 (3): 237–53. what is and is not reasonable behaviour in pur- Hammersley, M. and Gomm, R. (2000) ‘Bias suit of knowledge in the relevant context, with in social research’, in M. Hammersley these judgements being open to dispute and to (ed.), Taking Sides in Social Research. subsequent revision. London: Routledge. ch. 6. So, while all research is not necessarily Harding, S. (1992) ‘After the neutrality ideal: biased, there is always the potential for bias. science, politics and “strong objectivity”’, Furthermore, some of this will be non-culpable, Social Research, 59 (3): 568–87. in the sense that the researcher could not Kvale, S. (ed.) (1989) Validity Issues in have known that what was being relied on Qualitative Research. Stockholm: Students was erroneous or dysfunctional. At the same Literature. time, some systematic error will be culpable, Lather, P. (1986) ‘Issues of validity in openly in that researchers were in a position to rec- ideological research’, Interchange, 17 (4): ognize that an assumption on which they 63–84.

18 C-Jupp-3373.qxd 3/8/2006 5:40 PM Page 19

Other features include: support for dynamic CAQDAS data (that is, documents can be edited from within the project software, whilst still main- Definition taining the integrity of coding and other point- Software specifically designed for qualitative ers); categorizing records or parts of records by research, supporting detailed exploration, demographic and other attributes, to inform coding and analysis of textual and other subsequent analyses; editing memos, notes and non-numerical data. The acronym is for reports, which may be available for coding and ‘Computer Assisted Qualitative Data searching; filtering and grouping documents or Analysis Software’. Such software is more categories; text searching (sometimes with wild commonly referred to simply as ‘QDA soft- cards, sometimes storing finds as coded data); ware’ or ‘qualitative software’. graphical modelling tools for displaying and exploring theories and relationships (semantic networks or free-form drawings), whose items Distinctive Features are live to the data. QDA programs vary widely in functions. A checklist of the primary tools currently Evaluation available in one or more packages is as fol- QDA software appeared in the 1980s, long lows. First, importing text files, sometimes after programs for statistical analysis, and writing them in the program, sometimes also took longer to become normalized. Qualita- accessing associated pictorial, tape or tabular tive records are complex and understood in data. Usually these files are in plain text, context, by methods emphasizing interpreta- though recent programs allow formatting and tive processes, rather than mechanical ones. rich text appearance. Second, coding passages Early programs based on the manual methods of text or non-textual material, as user- of coding and retrieval were often rejected as created categories, and managing the cate- adding technological challenges to an already gories (adding, deleting, merging their coding, unsatisfactory method of exploring data. ordering them in index systems), sometimes Since then, the range and user-friendliness of storing weightings for codes. Third, accessing tools has increased, as have the number of data coded at a topic, reporting and some- programs and rate of revision. Use of QDA times re-contextualizing coded segments and software is expected or required across many in some programs also coding this text on to disciplines, and software clearly supports further categories. Fourth, automating coding qualitative methods of access and exploration and generation of codes, usually by text of data not possible by manual means. search. Some programs automate coding by document outline and some support command scripts for automating other processes. Lyn Richards C-Jupp-3373.qxd 3/8/2006 5:40 PM Page 20

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

Associated Concepts: Atlas.ti, coding, The rather than the use of random assignment. Ethnograph, NUD*IST, QSR NVivo, qualita- A case study can involve a single case tive research, SPSS (for example, a community study or a ‘socio- biography’ of a member of a deviant sub- Key Readings culture) or a number (possibly quite large) of cases (for example, in the analysis of the Any print descriptions of programs are nec- conflict behaviour of different types of work essarily out of date; use websites for cur- groups). rent information. A central website is The main thrust of a case study can be provided by the UK-based program carry- descriptive, exploratory or explanatory (Yin, ing the CAQDAS acronym (http://caq 1984). Exploratory case studies may provide das.soc.surrey.ac.uk/). initial analysis of a phenomenon that will Alexa, M. and Zuell, C. (1999) A Review of Software for Text Analysis. Spezial Band 5, then be systematically explored in other stud- Mannheim: ZUMA. ies, possibly by the use of another approach, Bazeley, P. (2002) ‘Computerized data analy- such as a sample survey. Or they may follow sis for mixed methods research’, in on from survey work to provide a more A. Tashakkori and C. Teddlie (eds), detailed account of particular findings. A Handbook of Mixed Methods for the Social descriptive study will attempt to provide a and Behavioural Sciences. Thousand Oaks, full portrayal of the case or cases being stud- CA: Sage. ied. An explanatory case study will attempt to Richards, L. and Richards, T. (1994) ‘Using provide an account of what caused a particu- computers in qualitative analysis’, in lar phenomenon observed in the study. Norman K. Denzin and Yvonna S. Lincoln Case studies have a key place in policy (eds), Handbook of Qualitative Research. research. They are frequently undertaken to Thousand Oaks, CA: Sage. provide examples of good practice in the deliv- Weitzman, E. and Miles, M. B. (1995) ery of a specific policy or programme, or they Computer Programs for Qualitative Data may be undertaken as part of an evaluation Analysis. Thousand Oaks, CA: Sage. project, providing examples of the impact of a policy. In these, and other instances, a key factor affecting the success of the study will be the criteria for selection of the cases to be stud- CASE STUDY METHOD ied. For the results to be persuasive they will normally need to be based on cases that provide a report of the operation of the policy in Definition a range of settings. There may, however, be An approach that uses in-depth investigation occasion to test the policy in a ‘critical-case’ of one or more examples of a current social setting, which presents the most difficult phenomenon, utilizing a variety of sources of circumstances for it to succeed. data. A ‘case’ can be an individual person, an event, or a social activity, group, organization Evaluation or institution. The main criticism of the case study method Distinctive Features is that in most circumstances the individual cases are not sufficiently representative to A major feature of the case study, accord- permit generalization to other situations. ing to Hakim (2000), is its flexibility. It can Efforts to overcome this perceived weakness range from a simple narrative description to a include increasing the number of cases so as very rigorous study achieving experimental to improve their representativeness, and pro- isolation by the selection of cases on the basis vide for comparative analysis within the case of the presence or absence of key factors study (Bryman, 1988). But, as Yin (1984)

20 C-Jupp-3373.qxd 3/8/2006 5:40 PM Page 21

CASIC

argues, this issue affects other methods as well. How is it possible to generalize from CASIC an individual experiment? In both instances the generalization involves the statement of a Definition theoretical proposition, which in turn will be Acronym for Computer Assisted Survey tested through the use of further case studies Information Collection. The term encompasses and other methods. This, for example, is often the technological developments which swept the use of case studies in the study of organi- through data collection in professional social zations, where organizational theory has been survey research organizations during the last developed on the basis of one or a small num- 15 years of the twentieth century, particularly ber of cases. The depth and rigour of the in North America and Western Europe. analysis will be the crucial issue here. Where negative or critical cases are used to test and Distinctive Features develop theories, the term analytic induction is sometimes used. In professional survey research organizations, As was noted above, case studies are fre- surveys are now normally conducted by inter- quently exploratory in nature and are linked viewers using portable laptop computers, into with other methods. In these instances the which the survey questionnaire has been pro- case study may be viewed less as a vehicle for grammed using a package such as BLAISE, generalization than as a form of pilot study. now the standard for public sector surveys. Also, in policy research it may be not neces- Such programs take the interviewer through sary to have numerous cases in order to iden- the survey questions from screen to screen, tify the negative impact of a phenomenon, according to the rules for the routing of ques- policy or programme. The main factor here tions programmed into the questionnaire. The will be whether the criteria for the selection interviewer sees one question at a time, then is of the case or cases will provide a robust test. taken to the next question to be asked of the respondent, determined by the program. Very Note complex questions can be asked, with specific thresholds (for example, on income, savings, The above material does not represent the state benefits and financial circumstances gen- views of the Department of Education and erally). Unlike the traditional social survey, Skills, UK. however, there is no paper copy of the questionnaire for the interviewer; the whole Vince Keddie process is electronic. At the end of the process, data is downloaded to the survey organization Associated Concepts: analytic induction, for editing and analysis, usually remotely from community study method, evaluation the interviewer’s home. Such editing and research, exploratory research, good practice analysis is typically carried out by software studies, policy-related research, validity of packages such as SPSS. generalization The first type of CASIC instrument was CATI (Computer Assisted Telephone Interviewing), carried out from central loca- Key Readings tions with interviewers sitting at IT consoles, Bryman, A. (1988) Quantity and Quality in scrolling though a CASIC questionnaire as they Social Research. London: Unwin Hyman. asked the questions. The interviewer did not Hakim, C. (2000) Research Design. London: leave the office. With the availability of rela- Routledge. tively lightweight portable computers, the Yin, Robert K. (1984) Case Study Research: development of CAPI (Computer Assisted Design and Methods. Beverly Hills, CA: Personal Interviewing) became possible, which Sage. is the most common mode of delivering CASIC

21 C-Jupp-3373.qxd 3/8/2006 5:40 PM Page 22

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

questionnaires in large survey organizations For example, a pictorial model is a way of today. There are other forms, such as CASI portraying possible relationships between (Computer Assisted Self Interviewing), where several variables and in this sense should be (perhaps for one section of a survey) the treated as an integrated set of hypotheses portable is turned around and respondents fill which are worthy of further investigation. in the section themselves, according to instruc- Lines are drawn between variables to indi- tions which appear on the screen in front of cate potential relationships and when arrow- them. This is particularly suitable for ensuring heads are added they indicate causal a better response on private or confidential direction. The production of such pictorial matters. models, no matter now crude, can form part of exploratory research and they subse- Evaluation quently act as guides to further inquiry. Their shape and content can be based on theoriz- CASIC has facilitated a greater degree of tech- ing, previous research findings or, sometimes, nical complexity in the survey data collection hunches. process. Survey researchers are trained to write Models can also take the form of a statisti- questionnaires in programs such as BLAISE, cal equation. An example is the multi- which means that all interviewers administer ple regression equation whereby values of a them in a standardized fashion, thereby reduc- dependent variable (say, income) are pre- ing error emanating from the interview dicted from the values of several independent process. variables (say, terminal age of education, CASIC programmes are also being devel- number of qualifications, father’s social oped for Web surveys, although this mode of class). A residual term is usually included in research faces many more challenges in rela- the equation to represent the degree of variation to sampling. tion in the dependent variable which cannot be explained by the combination of indepen- Martin Bulmer dent variables. The multiple regression model is founded on the assumptions that all variables are linear and that the effects of all the Associated Concepts: coding, interview, independent variables add up. This is known questionnaire, social survey, SPSS as addivity. A third form of model is a graphic repre- Key Reading sentation of relationships between variables. Couper, M. (1988) Computer Assisted Survey Lines between the variables indicate relation- Information Collection. New York: Wiley. ships, arrowheads indicate causal direction, but further, scores are added to the lines to provide numeric representations of the strength of relationship between paths of vari- CAUSAL MODEL ables when all other variables are held constant. An example is a path diagram, based on path analysis in which path coefficients pro- Definition vide the numeric value of strength of rela- A representation of a set of relationships in tionship. Although they are related, path which the causal connections between sev- analysis has one big advantage over ordinary eral variables are examined simultaneously. multiple regression analysis, namely that it is possible to examine direct and indirect Distinctive Features pathways to a dependent variable and therefore calculate the direct and indirect effects Causal models can take several forms and which independent variables have on the may be used at different stages in research. dependent variable.

22 C-Jupp-3373.qxd 3/8/2006 5:40 PM Page 23

CAUSALITY

Evaluation experience it may be concluded (almost certainly falsely) that doing A will always Unlike exploratory pictorial models, those produce B or (more circumspectly) that it based on multiple regression and on path often produces B. Even the character of this analysis are concerned with testing a theoreti- simple form of causation has been the subject cal model against the data. They seek to assess of much, largely unresolved, philosophical the degree of ‘fit’ between model and data. dispute. However, the complexities are multi- Their value over individual hypotheses is that, plied when we extend the meaning of ‘causal- out of recognition of the complexity of the ity’ to refer to situations involving feedback social world, they can look at the effects of sev- processes, and/or where we are concerned eral variables simultaneously. However, such with how factors cause outcomes that them- models cannot by themselves demonstrate selves involve human action. causality. The notion of causality is very much an in-built assumption and inferences about causal direction come not from statistical analy- Distinctive Features sis but from the researcher him- or herself. Researchers have varied in their attitudes A much more fundamental critique is towards the concept of causality. In recent that social science research should not be times, many of those engaged in quantitative founded on the principle of causality. This research have aimed at producing causal critique comes from those who work within a knowledge, while recognizing that establish- different paradigm, especially one that seeks ing a correlation between two or more vari- explanation-by-understanding. ables can never demonstrate with absolute certainty that they are causally related. The Victor Jupp ways in which quantitative researchers have sought to deal with this problem have Associated Concepts: causality, exploratory included path analysis, where correlations research, hunting (snooping and fishing), are interpreted in terms of a structural model hypothesis, path analysis that is designed to identify possible causal relationships, for example on the basis of temporal ordering and sometimes also a Key Readings theoretical rationale. Along with this, various Blalock, H. (1961) Causal Inferences in Non- precautions are taken to avoid designating as experimental Research. Chapel Hill, NC: causal what are in fact spurious relationships University of Carolina Press. generated by associations between some Sapsford, R. and Jupp, V. (1996) Data extraneous factor and both the independent Collection and Analysis. London: Sage. and dependent variables. Qualitative researchers have varied in their attitudes towards causation. Some have argued that in observing patterns of social CAUSALITY interaction causal relationships can be directly perceived and documented. This Definition might seem to involve a form of naive realism, of the same kind that confuses correlation A complicated concept that seems to have with causation. As Hume argued: while we its source in a relatively simple experience may see events follow one another, we do not common even for very small children: that see that they must do so. However, a more actions of particular kinds bring about partic- sophisticated version of this argument would ular sorts of result. Note that what is involved rely on the idea that, even if there is a sense here is not just a singular event, but a recur- in which we cannot eyeball the causal link rent experience. And on the basis of that between actions and events, the ‘inference’ to

23 C-Jupp-3373.qxd 3/8/2006 5:40 PM Page 24

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

such links will in many cases be beyond than establishing relationships that occur reasonable doubt – in the sense that it would whenever given conditions are met. Some of be questionable only if we were prepared to those with a more qualitative orientation have doubt much of what we routinely take to be rejected this approach, insisting that universal- knowledge about the world. istic laws can be identified, notably through analytic induction or via some form of com- Evaluation parative analysis (see, for example, Ragin, 1987). Here the aim, often, is to identify the Other qualitative researchers have denied that necessary and sufficient conditions for some causality operates in the social realm, arguing type of event happening. that the task of research is, instead, to under- It has also occasionally been argued that the stand the meanings which are constructed by sequence and pacing of the impact of causal participants in it. And this links to the argu- factors needs to be taken into account in any ments of some philosophers that human analysis: that the same set of factors operating actions must be understood in terms of in a different order, or at a different pace, will reasons rather than causes (see, for example, produce different outcomes (Becker, 1998). Peters, 1958). In part, this is a reaction against Such thinking is close to that of many histori- what is seen as the determinism that is ans (and, interestingly, also to that of epidemi- implicit in the notion of causality: the idea ologists), who distinguish between different that the actions of human beings are simply a ways that causal factors can operate: for exam- product of the various causal factors operating ple, as an underlying or predisposing factor, or on them. Such determinism is opposed on two as a trigger stimulating a particular sort of out- grounds. First, it is argued that it is at odds come. Here, the focus is usually more idio- with our experience of frequently choosing to graphic than nomothetic; in other words, the act in one way or another. Secondly, deter- aim is to explain why some particular event minism is opposed on ethical grounds. Some happened when it did, rather than to develop see it as encouraging the treatment of people or test a theory about what causes what across as objects, as manipulable – a view that is many situations. Even so, any identification of often believed to serve the interests of the a causal relationship carries implications about powerful. Others argue that it makes people what would and would not happen in other no longer responsible for their actions, and situations; this so-called ‘counterfactuality’ is thereby legitimates unethical behaviour. It essential to the distinction between correlation should be said, though, that most qualitative or sequence, on the one hand, and causation, researchers find it difficult to avoid the use of on the other. words and phrases that seem to carry at least The concept of causality is a difficult one, some sort of causal import: for instance, the then, but it is one on which most social verbs ‘influence’, ‘shape’ and ‘structure’. research cannot avoid relying. A further complication is that the reasons people give for their behaviour often involve Martyn Hammersley reference to putative causal relationships. Many social scientists have insisted on the multi-causal character of human social life: Associated Concepts: analytic induction, that what happens is always a product of many causal model, correlation, hypothesis, idio- factors. Some quantitative researchers have graphic, nomothetic, path analysis, positivism sought to deal with this by measuring the relative contributions of different factors, rec- Key Readings onciling themselves to the discovery of causal laws that are probabilistic in character; in other Abbott, A. (1998) ‘The causal devolution’, words, laws that indicate the likelihood that Sociological Methods and Research, 27 (2): one type of event will cause another, rather 148–81.

24 C-Jupp-3373.qxd 3/8/2006 5:40 PM Page 25

CENSUS

Becker, H. S. (1998) Tricks of the Trade. Governments, particularly those of Europe Chicago: University of Chicago Press. and North America, have collected informa- Hage, J. and Meeker, B. F. (1988) Social tion about their population through censuses Causality. London: Unwin Hyman. for several centuries. A census was first con- MacIver, R. M. (1942) Social Causation. ducted in Sweden in 1749, in the United New York: Harper and Row. States of America in 1790 and in the United McKim, V. R. and Turner, S. P. (eds) (1997) Kingdom in 1801. The key purpose of the Causality in Crisis: Statistical Methods and Census in the United States is to divide the the Search for Causal Knowledge in the Social 435 seats in the House of Representatives Sciences. Notre Dame, IN: University of fairly between the fifty states, based on the Notre Dame Press. size of their population. However, there are Peters, R. S. (1958) The Concept of also other purposes: for example, to allocate Motivation. London: Routledge & Kegan Paul. funds to geographical areas and to identify Ragin, C. C. (1987) The Comparative Method. groups in need of services. In the UK, the Berkeley, CA: University of California link to political representation is not so Press. direct, although the boundaries of parliamen- Taylor, R. (1967) ‘Causation’, in P. Edwards tary constituencies are periodically reviewed (ed.), The Encyclopaedia of Philosophy. in the light of changing population figures. New York: Macmillan. p. 127. Census data are used primarily for social purposes, particularly to determine the level of funding allocated by central government to local authorities. The UK Census collects data by sending CENSUS one form to every household. In 2001, there were a number of questions for one person to Definition answer on behalf of the household and 35 to A census can have two meanings. One is an be completed by each individual. The 2000 attempt to collect data from every member of Census in the USA used two forms: a short the population being studied rather than one consisting of seven questions was sent to choosing a sample. The other is a specific most households, while a longer one was form of social survey organized by govern- used in approximately one in six cases. ments with the aim of collecting information from every household in the country. Govern- Evaluation ment censuses are organized at regular intervals – most commonly every ten years – A census with a 100 per cent response rate with the information collected affecting polit- has an advantage over a sample in that there ical structures and social policies. are no concerns as to whether the people who take part are representative of the popu- Distinctive Features lation. However, if not everybody responds to the census, there remains a difficulty in seek- Taking the first meaning of a census, some ing to establish whether the non-respondents populations can be easily identified, such as would have given different information, or every delegate to the United Nations or every expressed different opinions, to the people athlete who competed at the most recent who took part. Olympic Games. However, censuses can also Government censuses internationally tend be taken of less obvious groups. Research to have response rates of over 95 per cent. that collected data from every person enter- However, the 2000 Census in the USA ing a specified supermarket between 2 pm achieved a response rate of only 67 per cent, and 4 pm on a specified day would be a form although this was an improvement on the of census. 65 per cent achieved in the 1990 Census. This

25 C-Jupp-3373.qxd 3/8/2006 5:40 PM Page 26

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

was the first time that the response rate had with their increasingly mobile populations. risen from one census to the next, possibly as However, these difficulties are relatively a result of a programme of advertising, pro- small when compared to the situation in motions and special events emphasizing the other countries, where a national census importance of taking part. The Census would be an impossible undertaking. A Bureau undertook a coverage measurement condition of conducting a comprehensive survey and, on the basis of this, decided not population survey is that at least a large to adjust the census figures. majority of people can be found at any one The UK census has traditionally achieved time, with resources available to pay enu- a higher response rate – over 97 per cent merators to collect the data and research staff of the population is believed to have been to analyse it. These conditions are clearly not covered by the 1991 Census. However, non- met in countries where there is severe response was not evenly spread between dif- poverty, war or internal conflict. ferent geographical areas and different groups: in some cities, over 20 per cent of all young males were believed to have been Jamie Harding missed. A Census Validation Survey did not establish the extent and distribution of the Associated Concepts: bias, data archives, under-enumeration. empiricism, error, official statistics, quantita- Similar problems emerged with the 2001 tive research, secondary analysis, social survey UK Census, but there were better techniques in place to try to overcome them. This time Key Readings approximately one million people were Periera, R. (2002) ‘The Census Coverage thought to have been missed by the official Survey – the key elements of a one count. From previous census and other number census’, Population Trends, 108: publicly available data it again became clear 16–31 at http://www.statistics.gov.uk/ that young men were particularly likely to articles/population_trends/censuscoverage_ have been missed by the count, but so were pt108.pdf other groups, such as older women. US Census Bureau (2002) Census 2000 Attempts were made to correct the 2001 Basics. US Department of Commerce at: data through the Census Coverage Survey. A http://www.census.gov/mso/www/c2000b sample of approximately 300,000 households asics/00Basics.pdf was taken from areas that were stratified by an de Vaus, D. (2002) Surveys in Social Research, index estimating the number of ‘hard to count’ 5th edn. London: Routledge. groups. Information was collected by interview, which offered more opportunity to establish whether people had been accidentally excluded than had been the case with the CHI-SQUARE TEST original self-completion form. Interviewers were instructed to make as many calls as was necessary to contact a household member. The Definition differences between the data revealed by the A statistical test of significance which is used to Census Coverage Survey and that obtained in compare observed frequencies with expected corresponding areas by the Census was used frequencies. to ‘impute’ the number and type of households missing from the Census nationally. Distinctive Features The ‘imputing’ of individuals to add to the Census data has caused some unease and is The chi-square χ2 test is used for analysing symptomatic of the increasing difficulties in data where one has counted the frequency conducting such surveys in Western countries, (number of cases or respondents) in different

26 C-Jupp-3373.qxd 3/8/2006 5:40 PM Page 27

CHI-SQUARE TEST

categories. For example, one might have categorical variable and has measured the recorded the number of men and the number frequency of some event at each level of that of women who have died within a certain variable. For example, suppose one has period from lung cancer. There are two vari- recorded the number of road accidents on ables, ‘sex’ and ‘cause of death’. Both are each day of the week over a six-month category variables, as people are divided into period. ‘Day of the week’ is a categorical vari- mutually exclusive groups or categories. In able with seven levels (one for each day). One this example, each variable has two levels; might ask whether accidents are related to sex has the levels of ‘male’ and ‘female’ and days. If there were no relationship between cause of death has the levels ‘from lung accident frequency and day, the total number cancer’ and ‘not from lung cancer’. (There of accidents would be divided equally between can be more than two levels of a variable: if the days. The one-sample chi-square is used one had ‘death from lung cancer’, ‘death to test whether the observed frequencies from other cancers’, ‘death from any other differ from what would be expected if the cause’ there would be three levels of the variables were unrelated. cause of death variable.) For the chi-square test to be valid, the When the observed frequencies have been entries in the cells have to be independent obtained, it is possible to construct a cross- which means no respondent or case can tabulation or contingency table like this: occur in more than one cell of the table. So N, the total number of cases in the table, must Cause of death equal the number of participants. (If one has repeated measures, the McNemar test may be From lung Not from lung appropriate.) Sex cancer cancer It is commonly stated that for the test to be Male valid the expected frequencies in the cells of the table must be 5 or more in at least 20 per Female cent of the cells. (But some authorities say this is not essential.) To meet this require- The entries in each cell of the table would be ment, one may be able to merge rows or the number of cases, so the number of males columns or have to collect more data. who died from lung cancer would be entered With a 2 × 2 contingency table, Yates’ in the top left cell. correction for continuity can be applied in The researcher is usually concerned with calculating chi-square but many authors do whether there is a relationship between the not recommend it. two variables, for example, whether there is a relationship between sex and likelihood of dying from lung cancer. The two-sample chi- Evaluation square test is used to answer such a question. The chi-square test is a valuable tool for The test involves calculating the frequen- analysing frequency data, but is restricted to cies, known as the expected frequencies, studying no more than two variables at a time. which would be found if there were no rela- When there are three or more categorical vari- tionship between the variables and comparing ables, it is necessary to use more complex them with the observed frequencies. If the procedures such as log-linear analysis. difference between the expected and the observed frequencies is sufficiently large, it can be concluded that there is a relationship Jeremy J. Foster between the variables. It is important not to confuse the two- Associated Concepts: coding, correlation, sample chi-square with the one-sample descriptive statistics, log-linear analysis, vari- chi-square, which is used if one has a single able analysis

27 C-Jupp-3373.qxd 3/8/2006 5:40 PM Page 28

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

Key Readings Ward’s algorithm that is recommended for interval-scaled variables, start with the Howell, D. C. (2002) Statistical Methods for finest possible classification in which each Psychology, 5th edn. Pacific Grove, CA: single object forms one cluster. They then Duxbury. proceed to fuse objects into fewer and Tabachnick, B. G. and Fidell, L. S. (1996) Using Multivariate Statistics. New York: fewer clusters until all objects are grouped HarperCollins. into one large cluster. Divisive-hierarchical algorithms start the other way round, that is, they begin with all objects in one cluster. Non-hierarchical procedures start with a predetermined number of clusters, CLUSTER ANALYSIS for example one based on hypotheses or some pre-specified classification. Definition Probabilistic techniques like latent class analysis do not assign objects determinis- A group of statistical algorithms used to classify tically into clusters but yield different objects on the basis of their similarity with assignment probabilities for each object respect to a set of attributes. Objects can be per- within each cluster or latent class. sons, groups or any other entities of interest for Probabilistic methods differ from hierar- the social sciences. Attributes, for example chical and non-hierarchical algorithms in traits of persons or leisure preferences of socio- their underlying statistical assumptions cultural groups, are measured so that every (see Rost and Langeheine, 1997). object can be described by a profile that repre- (4) Several suggestions exist to determine sents the respective values of an object the number of clusters when using hier- on these attributes. Cluster analyses generate archical algorithms that do not end with a numerical classification or taxonomy that the ‘correct’ numerical classification. A comprises groups or ‘clusters’ of objects in such common rule when using Ward’s algo- a way that profile differences between objects rithm is to consider the increase in error within a cluster are minimized and profile dif- sum of squares after each fusion step. ferences between clusters are maximized. That number of clusters, say for example four, can be chosen after which a Distinctive Features sharp increase in error sum of squares results in the next step, that is when the Cluster analytic techniques are used in many four are fused into three clusters. different sciences whenever objects are to be However, determination of a cluster classified. Carrying out a cluster analysis solution based on this criterion is usu- involves the following main steps: ally not that unequivocal, so that addi- (1) Object attributes, for example personality tional criteria should be used (see traits of individuals, that seem to be rel- Everitt et al., 2001). evant for a given theory must be (5) Once a certain cluster solution is selected, selected. it should not be accepted at face value but (2) To calculate (dis)similarity between the evaluated against different criteria, for objects on these traits an adequate example concerning its replicability across measure must be chosen, depending on different algorithms and samples. the attributes’ measurement level and theoretical considerations. Evaluation (3) The calculated (dis)similarities between the objects serve as a starting point for run- A single best strategy for carrying out a cluster ning a cluster algorithm. Agglomerative- analysis does not exist, since there are no def- hierarchical cluster methods, for example inite rules for selecting the ‘right’ attributes,

28 C-Jupp-3373.qxd 3/8/2006 5:40 PM Page 29

CLUSTER SAMPLING

similarity measures, cluster algorithms and then areas within regions randomly selected, for determining the correct number of clus- then several levels of smaller areas, then finally ters. However, certain strategies have proven households within the smallest area. For exam- to be more valuable under certain conditions. ple, a study may define a number of towns typ- Facing this relative indeterminacy, it is essen- ical of the different types of towns found in a tial to replicate and validate a cluster solution nation-state and then conduct a random sample using different criteria and methods. In this from the registered residents in each town. In context, the interpretability and usefulness of some survey research designs all members in a a cluster solution should also be considered. cluster chosen at the final stage will be studied, for example, all children in a class chosen for inclusion in the study. Karl-Heinz Renner Evaluation Associated Concepts: discriminant function analysis, factor analysis, multivariate analy- Random sampling is an ideal standard that it sis, quantitative research, validity is often not possible to attain because of practical and ethical issues. Cluster sampling may Key Readings be justified on pragmatic (including cost) grounds, especially where the consequences Bailey, K. D. (1994) Typologies and Taxonomies: of erroneous inferences are considered to be An Introduction to Classification Techniques. slight. They may also be justified theoreti- London: Sage. cally where relationships between those in Everitt, B. S., Landau, S. and Leese, M. selected clusters are characteristics of the (2001) Cluster Analysis. London: Arnold. cluster that may influence the individual Rost, J. and Langeheine, R. (eds) (1997) subject (as in multi-level or structural effects). Applications of Latent Trait and Latent Class The major problem of cluster sampling is Models in the Social Sciences. Münster: that the sample does not fully represent the Waxmann. population from which it is drawn. It is generally the case that individuals within the clusters are more like each other than those outside their cluster. Thus, cluster sampling CLUSTER SAMPLING reduces and thus does not fully represent the variety that would be found in a true random Definition sample. In multi-stage sampling the larger the number of higher level units sampled the A method of survey sampling which selects more representative the final sample will be clusters such as groups defined by area of res- of the population. idence, organizational membership or other group-defining characteristics. Aidan Kelly

Distinctive Features Associated Concepts: area sampling, proba- Cluster sampling is often used where a bility (random) sampling, sampling, social complete list of subjects is impossible or survey, stratified sampling impractical to construct. Cluster sampling is a two- (or more) stage process whereby clusters Key Readings of individual units are first defined and selected and then samples of individual units are taken Barnett, V. (1991) Sample Survey: Principles and from each of the defined clusters. Large-scale Methods. London: Edward Arnold. household surveys may involve multi-stage Kish, L. (1965) Survey Sampling. New York: cluster sampling where regions are determined, Wiley.

29 C-Jupp-3373.qxd 3/8/2006 5:40 PM Page 30

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

Evaluation CODING Coding is an issue for other types of research Definition than social survey research. For example, in all the forms of qualitative research (such as The process by which observations recorded observation, analysis of documents and texts, in the course of social research – typically in ethnographic interviews, collection of trace a social survey questionnaire – are trans- or unobtrusive measures) there is a need to formed from raw data into categories and categorize or segment the data. Typically this classifications, which then become the is done after data collection and in the pursuit subject of quantitative data analysis. Coding of theory formation rather than theory testing. involves the act of measurement, for in clas- There is increasing interest in software sifying answers to a question, one is trying to packages for the analysis of qualitative data, measure the underlying social variable which which involves the use of coding. In all forms the survey question intends to tap. of research the risk is that the researcher imposes inappropriate and invalid categories Distinctive Features and codes on the data. In professional social survey research many survey questions are asked in fixed-choice Martin Bulmer format, the answers to which are pre-coded into categories. The respondent’s answer is Associated Concepts: CAQDAS, mea- assigned by the interviewer to one of these surement, operationalization, qualitative data categories when their response is given. archives, qualitative research, quantitative Other questions, often of a more complex research kind, may be coded after the interview by office coders, working from a coding frame, Key Readings which sets out the rules according to which answers are to be classified. This involves Bateson, N. (1984) Data Construction in knowledge of the range of responses possible, Social Surveys. London: Unwin Hyman. and for more complex issues, like the mea- Rose, D. (2003) A Researcher’s Guide to the surement of social class, reliance upon other National Statistics Socio-economic Classifica- classifications as the building blocks for the tion. London: Sage. major coding decisions. Thus in the case of social class, occupation is commonly used in the UK as the basis for assigning a person to a particular social class, and this COHORT STUDY relies upon the official Classification of Occupations to determine the occupation, industry and employment status of the Definition person concerned. Prospective cohort studies re-investigate Coding is a neglected topic in social groups of people who share some social research, and deserves more attention. characteristic. Bateson argues that it should be approached in terms of data construction, and questions asked Distinctive Features about how the researcher is constructing the phenomenon being studied. Questions of the Cohort study takes its name from two intel- validity of measurement are also critical in lectual traditions. Cohorts were Roman scrutinizing the results of coding. The wider army units of about 600 men. The medical topic of measurement error is closely linked to profession adopted this classical reference the efficacy of coding. to describe studies of groups of patients

30 C-Jupp-3373.qxd 3/8/2006 5:40 PM Page 31

COHORT STUDY

undergoing particular treatments, and it is shorter than the life-long perspective of many here that the term is most frequently encoun- social science projects, so there is less chance tered today. In both cases, the common of people dropping out. It is in patients’ self- theme is how many survivors there were. interest to continue treatment, because of the Although there are more complicated implied promise of improved health. In social cohort designs, the best cohort study design science research, the benefits are less obvi- in health research started with two matched ous. In health research, not only do hospitals groups of people. After initial assessment, the provide regular supplies of patients as poten- experimental group received a measurable tial recruits, but back-up facilities for keeping treatment: their medical outcomes were track of samples are already in place in the followed through repeated tests, preferably form of hospital administrative personnel and until patients were healed. The other, ‘con- medical records, whereas social science trol’, group did not receive the treatment: any projects are usually set up from scratch with emerging differences could plausibly be more limited funds. Those running health attributed to the effects of the treatment. studies – doctors – have high social status, Cohort studies thus combine the core charac- whereas the patient role has low status: teristics of an experiment (although not all patients are more likely to comply with a intervening effects can be controlled by medical study than are the general public matching the two groups) with those of the dealing with mere academic social scientists. longitudinal study, in which claims of causal- This may explain why most of the larger and ity may be justified by showing that one thing better-known social science cohort studies, occurs before another (the treatment causes like the National Child Development Study recovery because it happens first). (NCDS) in theUK, had their origins in public Whereas medical researchers mainly use health. the term cohort study, social scientists see Variations on the core design include retro- cohort studies as a special case of longitu- spective (or historical) designs, where records dinal studies, trend studies or time series are used to construct a sample whose mem- often using these labels interchangeably. The bers share prior common characteristics. For methods are actually similar. Medical usage example, the Australian Gulf War Veterans usually involves stronger notions of causality, Health Study (almost uniquely successful in whereas social science usage is more recruiting every veteran and all but two of the exploratory, historical and involves more matched control group) was established a variables in its explanatory model. Cohort decade after the conflict. Concurrent or and longitudinal studies usually re-study the prospective studies use larger samples with no same people: trend studies and time series control group, sometimes adding new mem- collect quantitative, often socio-economic, bers as required: the US Multicenter AIDS data from a series of different samples over Cohort Study started with 4954 men in 1985, time. The terminology used owes more to adding another 668 between 1987 and 1991. conventions of academic disciplines than essential differences. Evaluation The greater popularity of cohort studies among doctors is partly due to the simplicity The key attraction of cohort studies is the of the core design. It is easily explained, and time component, enabling interactions of vari- used with simple statistical techniques. It also ables to be better understood than in single suffers less from one of the key problems of point studies like the one-off (cross-sectional) sociological and psychological cohort studies, survey (Davies, 1987). While a time series maintaining continued participation of the study based on successive independent sample. samples is cheaper and easier to organize (for In medical research this is less problem- example, the British General Election Studies, atic. The length of cohort studies is often the Social Attitudes Surveys, the Labour Force

31 C-Jupp-3373.qxd 3/8/2006 5:40 PM Page 32

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

Survey), variation in findings may in part be Marmot, M. and Wilkinson, R. (1999) Social attributable to sampling. Time series studies Determinants of Health. Oxford: Oxford are, however, much better than comparing University Press. respondents of different ages within a single Payne, G. and Roberts, J. (2002) ‘Opening sample: this can produce long-term seriously and closing the gates: recent develop- misleading results, such as the false picture of ments in British male social mobility’, social mobility drawn by the Nuffield College Sociological Research Online, 6 (4). study (Payne and Roberts, 2002). Ruspini, E. (2002) An Introduction to Apart from high costs, the disadvantage of Longitudinal Research. London: Routledge. the prospective cohort study is waiting for data. Studies of how childhood experience impacts on adult disease risk or social outcome involve waiting literally a lifetime. Meanwhile, COMMUNITY STUDY METHOD researchers change their priorities and improved research techniques (or medical Definition treatments) become available. After 50 years in which it had to abandon its Northern Ireland An approach that uses a range of research element, and having been run by three strategies and methods to study communities entirely different organizations using six in a holistic manner, usually with the close different data collection agencies, the NCDS involvement – and sometimes participation – (Bynner and Parsons, 1997) retains only two- of researchers in those communities. thirds of its original sample (not surprisingly, the frequent absence of systematic associa- Distinctive Features tions being reported for the NCDS data is beginning to raise doubts about its quality). The community study method has its roots in The attrition rate, the loss of participants social anthropology and was subsequently at during the study, is a major defect. The the heart of the work of the Chicago School Aberdeen Birth Cohort Study of people born of Sociology in the 1920s and 1930s. The in 1921 had contact with only six in ten of its Chicagoans studied a whole range of social sample 70 years later. If those who drop out issues, especially those which were crime- have different social characteristics from based, but in the context of the community. those who remain, any findings will not be Their research used statistics to map social applicable to the general population. conditions in Chicago but also complemented this with a range of qualitative methods – such as participant observation, life history Geoff Payne interviews and documentary analysis – to capture the subjective experiences of life in Associated Concepts: attrition, causality, the Chicagoan communities. The method was cross-sectional survey, experiment, longitudi- also central to the study of politics and power nal study, one-shot design, panel study, in middle America in the 1950s, for example prospective study, retrospective study in the work of Lynds and Lloyd Warner, and also to examine the effects of urban decline on working class family life in the East End of Key Readings London. In the twenty-first century there Bynner, J. and Parsons, S. (1997) It Doesn’t Get has been a revitalization of interest in the com- Any Better. London: Basic Skills Agency. munity via the communitarianism movement Davies, R. (1987) ‘The limitations of cross and also participatory action research (PAR). sectional analysis’, in R. Crouchley (ed.), A key feature of the community study Longitudinal Data Analysis. Aldershot: method is a holistic approach, that is, there is Avebury. pp. 1–15. an assumption that the groups or communities

32 C-Jupp-3373.qxd 3/8/2006 5:40 PM Page 33

COMPARATIVE METHOD

are more than or different from the sum of their parts. Therefore it is necessary to study COMPARATIVE METHOD the community in totality. A wide range of methods is used, mainly but not exclusively Definition qualitative in nature. Methods include partic- The selection and analysis of cases that are ipant observation, interviews, life histories, similar in known ways and differ in other documents. More recently the use of videos, ways, with a view to formulating or testing small area statistics and modelling techniques hypotheses. have been used. Distinctive Features Evaluation The comparative method is reflected in different A number of issues surround the community research styles. For example, the use of quasi- study method. Some of these are practical experimental designs to evaluate policy initia- issues of research design and data collection, tives involves the comparison of the two groups for example, how to gain access, how to avoid or areas, one of which receives the policy initia- marginality and how to study communities in tive whilst the other does not. The two groups a holistic manner when they are divided and are compared before and after the introduction when there are conflicts between sections of of the initiative with regard to the feature the community – acceptance by one section at which the policy is aimed. This two-way can involve rejection by another. comparison (of two groups on a before–after There are also contentions and disputes basis) facilitates some evaluation of the effec- about the concept of ‘community’ in itself. tiveness of the policy initiative. The comparative For some, it is much more sensible and realis- method in social surveys is illustrated by the tic to think about social networks rather than way in which categories of people in a sample community. The notion of social networks is (for example, men and women: old and young) viewed as being much more valid in a frag- are compared in terms of their having or not mented and postmodern social world. having an attribute (for example, a scale which measures ‘fear of crime’). Qualitative, ethno- Victor Jupp graphic work emphasizes the development of generalizations and of grounded theory by the Associated Concepts: access, covert systematic comparison of cases in terms of their research, cultural research, ethnography, similarities and their differences. The cases fieldwork, network analysis, methodological which are compared include interactions, social pluralism, participatory action research meanings, contexts, social actions and cultural groups. Because this is a continuous, rather than a once-and-for-all activity, it is sometimes Key Readings referred to as the constant comparative method. Other methods to use comparison include con- Bell, C. and Newby, H. (1971) Community tent analysis (the comparison of documents), Studies. London: Allen & Unwin. historical analysis (comparison of time periods) Bulmer, M. (1984) The Chicago School of Sociology. Chicago: University of Chicago and analysis of official statistics (comparison of Press. areas, groups or time period in terms of social Crow, G. and Allan, G. (1994) Community Life. indicators). Comparison permeates different Hemel Hempstead: Harvester Wheatsheaf. stages of research. For example, in research Wellman, B. and Berkowitz, S. D. (eds) (1988) design it is illustrated in the ways in which sam- Social Structures: A Network Approach. ples are selected for study so as to compare Cambridge: Cambridge University Press. sub-sets on some variables while holding others Whyte, W. F. (1988) Learning from the Field. constant. At the analysis stage different areas of Beverly Hills, CA: Sage. social groups can be compared on a variable at

33 C-Jupp-3373.qxd 3/8/2006 5:40 PM Page 34

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

the same point in time (cross-sectional analysis) several observable variables into an overall or the same areas of groups can be compared on score. Composite scores can be computed a variable at different points in time (longitudi- either by using simple sums or averages or by nal analysis). Sometimes comparison takes place performing transformations on the original after publication of results, for example, in variables. examining the conclusions of projects from different parts of the world to consider how Distinctive Features different societies deal with domestic disputes. Single indicators are usually insufficient to Evaluation provide researchers with satisfactorily reliable and valid measures of complex theoretical Comparison is an essential part of research constructs. It is therefore often necessary to methodology. It involves deliberately seeking or combine multiple indicators of the same con- anticipating comparisons between sets of obser- struct into a composite. Examples of such vations. Without comparison with a base-line or composites include scores on psychological against a control group it is not possible to reach tests made up of multiple items, measures of plausible and credible conclusions. However, socio-economic status composed of various this requires the clear establishment of a base- observable indicators, or the consumer price line and the ability to control for factors that the index based on a weighted combination of researcher does not wish to vary. prices of various goods – to name only a few. If the single indicators are scored on the Victor Jupp same scale and can be assumed to be of roughly equal relevance as, for example, in Associated Concepts: constant comparative most psychological tests, it is often possible to method, content analysis, cross-sectional simply compute the sum or arithmetic mean survey, secondary analysis, longitudinal of the raw scores as the composite score. study, quasi-experiment If scales are identical but the significance across indicators is not (for example, combin- Key Readings ing counts of more or less severe crimes into an index of criminal conduct), a weighted Bryman, A. (2004) Social Research Methods, combination may be more appropriate. 2nd edn. Oxford: Oxford University Press. Furthermore, if the individual items are Hopkin, J. (2002) ‘Comparative methods’, in measured on different scales (for example, D. Marsh and D. Stoker (eds), Theory and protocols of behavioural observations and Method in Political Science, 2nd edn. teacher ratings of aggressiveness combined London: Palgrave. pp. 249–67. into an index of aggressive behaviour), it is Kennett, P. (1973/2001) Comparative Social necessary to transform the single item scores Policy: Theory and Research. Milton Keynes: into comparable measurement units before a Open University Press. composite score can be computed. OEyen, E. (ed.) (1990) Comparative Metho- The next steps in the development of com- dology: Theory and Practice in International posite measures are to assess how the single Social Research. London: Sage. indicators are related to each other and how the composite relates to outside variables. The use of multiple indicators is a necessary pre- COMPOSITE MEASUREMENT condition for performing the former task but not a sufficient one. Structural analyses Definition provide the researcher with the information to decide which of the single variables are Measurement of an unobservable variable or adequate and which should rather be dropped construct by means of aggregating scores on from the composite. Further evidence on the

34 C-Jupp-3373.qxd 3/8/2006 5:40 PM Page 35

CONFIDENTIALITY

construct and criterion-related validity of inferences based on the measure may then be CONFIDENTIALITY obtained from external validation studies. Definition Evaluation The principle that information about partici- Composite measures usually yield more pants in research is private and should only desirable psychometric properties than single be revealed with their consent. indicators. However, in deciding which indicators should be retained for the final Distinctive Features composite, the test constructor may be faced When people allow researchers to investigate with a dilemma: highly intercorrelated indi- them, they often negotiate terms for their cators lead to homogeneous measures with a agreement. Participants in research may, for relatively unequivocal meaning. On the other example, consent on the basis that the infor- hand, higher correlations with outside vari- mation obtained about them will be used ables can often be obtained by combining only by the researchers and only in particular only weakly correlated variables (cf. Ghiselli ways. The information is private and is et al., 1981). Thus, there is some trade-off voluntarily offered to the researcher in confi- between construct validity and the goal of dence in exchange for possibly not very much predicting external criteria. It is possible to direct benefit. While social science research distinguish between observed indicators as participants might be hurt by insensitive data effects of constructs (for example, behaviour collection, often a more significant danger is as a manifestation of a personality trait) and posed by what happens to data after it has indicators as causes of the construct (for been collected. example, education, income etc. as determi- In general, three different arguments are nants of socio-economic status), and it is also used to justify maintaining confidentiality. possible to demonstrate that high intercorre- Consequentialist arguments examine the lations between indicators are desirable only results of an ethical practice and may in the former case. To summarize, every deci- consider what would happen if the practice sion in the construction of a composite did not exist. So, interviewees might be reluc- measure should be guided by careful theoret- tant to reveal secrets to social scientists if ical considerations but also be informed by they thought that the information might be additional empirical analyses. freely disseminated to third parties. Researchers who break confidences might Bernd Marcus not only make it more difficult for themselves to continue researching but, by damaging the Associated Concepts: factor analysis, indica- possibility that potential participants will tor, measurement, multiple indicators, perfor- trust researchers, might also disrupt the work mance indicator, quantitative research, of other social scientists. reliability, scaling, validity The second justification for confidentiality is rights-based. Beauchamp and Childress (2001) argued that our right to privacy rests Key Readings on the principle of respect for autonomy. Ghiselli, E. E., Campbell, J. P. and Zedeck, While some matters cannot or should not be S. (1981) Measurement Theory for the concealed, people should have the right, as Behavioral Sciences. San Francisco, CA: far as is possible, to decide what will happen Freeman. to them. In research, they should be able to Nunnally, J. C. and Bernstein, I. H. (1994) maintain secrets, deciding who knows what Psychometric Theory, 3rd edn. New York: about them. Finally, fidelity-based arguments McGraw–Hill. rest on the view that researchers owe loyalty

35 C-Jupp-3373.qxd 3/8/2006 5:40 PM Page 36

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

to the bonds and promises associated with stripping data of individual identifiers may research. By offering a promise of secrecy, be compromised by improved capacities to researchers offer to give allegiance and agree manipulate multiple, linked data sets. While a at minimum to keep silent or possibly even to survey might not include individual names or do more to guard a confidence. other unique identifiers, it may include suffi- In some research projects, negotiations cient identifying attributes to allow a person’s around confidentiality may be fairly straight- identity and/or various sensitive attributes to forward. Some researchers operate in relatively be inferred. There are two major statistical predictable contexts where standardized assur- ways to disguise or conceal the identities of ances may be included in a covering letter with individuals whose attributes are reported in a questionnaire. However, other work takes data sets. The first involves altering the data place in informal and unpredictable environ- and the second requires restricting access to ments, where agreements need to be negoti- the data (Brady et al., 2001). Data alteration ated with individuals and groups and may allow data to be disseminated more renegotiated during the course of lengthy field- broadly, but may affect the confidence that work. A further complication may arise if the people can place on particular aspects of participant has commercial interests to protect the data. Conversely, restricting access may and the resources and expertise to ensure that create inconveniences and limit the pool of these protections are stipulated in any agree- researchers that can use the data, but gener- ment. However, obligations of confidentiality ally permits access to greater data detail. cannot be considered absolute and in some situations – such as when researchers uncover Evaluation gross injustice – researchers might contem- plate disclosing information received under an Researchers should evaluate their position in implied or explicit assurance of confidentiality. relation to the law. Some researchers in the In some cases, researchers may be forced United States, Canada and Australia have by government officials or courts to disclose received statutory protection for specific data, breaching assurances of confidentiality. kinds of data. Even without statutory protec- There are two kinds of measures that can be tion, researchers have refused to reveal infor- taken to preserve confidentiality. The first is mation to courts (McLaughlin, 1999). The methodological, the second legal. Researchers reasons for their decisions and the point at have acted to protect the confidentiality of which they decided they could no longer research participants and their activities cooperate with the legal system varied either by not recording names and other data considerably. Picou undertook a longitudinal at all, or by removing names and identifying study of the social impact of the 1989 Exxon details of sources from confidential data at Valdez oil tanker disaster. Exxon subpoenaed the earliest possible stage. These precautions his files but Picou (1996) persuaded the court offer the advantage of helping guard data to restrict access to data used for published against theft or improper disclosure by other papers to an expert sociologist retained by members of a research team. Some researchers Exxon who was only to use the data for have reported sending files out of the jurisdic- statistical analysis. In Canada, a Master’s tion, and avoiding using the mail or telephone student investigating the death of AIDS system so that data could not be intercepted. patients was subpoenaed by the Vancouver In qualitative research, identifiers such as Coroner to appear at an inquest. Ogden won names, geographical clues and vernacular his case on the basis that the information terms can also be removed in the writing up had been obtained in confidence, confiden- stage. However, it can be difficult to hide the tiality was essential to the research relation- identity of some people from themselves, ship, that the research was socially valuable their peers, investigative journalists or offi- and that the harm of breaching confidential- cials. In quantitative research, practices of ity outweighed the benefit to be gained by

36 C-Jupp-3373.qxd 3/8/2006 5:40 PM Page 37

CONSTANT COMPARATIVE METHOD

disclosure (Palys and Lowman, 2001). Of theories from collecting and analysing course, researchers may be vulnerable if they empirical data. It stresses that comparing all refuse to disclose information where ordered the data throughout the analytic process is by a court. In 1972, a Harvard political scien- the most elucidating way to knowledge. tist, Samuel Popkin, spent eight days in jail while an American sociology graduate Distinctive Features student, Rik Scarce, spent 159 days in jail in 1993. Glaser (1969) suggests the ‘constant comparative method’ as a procedure for interpreting empirical material. It basically consists of four Mark Israel stages: ‘(1) comparing incidents applicable to each category, (2) integrating categories and Associated Concepts: access, covert their properties, (3) delimiting the theory, and research, data archives, ethics, exploitative (4) writing the theory’ (Glaser, 1969: 220). It is research, gatekeeper, informed consent, poli- similar to the procedures suggested in later tics and social research, research bargain versions of grounded theory by Strauss and Corbin (1990), but puts more emphasis than Key Readings these on interpretation and comparison than on (open, axial and selective) coding. For Beauchamp, T. L. and Childress, J. F. (2001) Principles of Biomedical Ethics, 5th edn. Glaser, the systematic circularity of this New York: Oxford University Press. process is an essential feature. ‘Although this Brady, H. E., Grand, S. A., Powell, M. A. method is a continuous growth process – each and Schink, W. (2001) ‘Access and confi- stage after a time transforms itself into the dentiality issues with administrative next – previous stages remain in operation data’, in M. Ver Ploeg, R. A. Moffitt and throughout the analysis and provide continu- C. F. Citro (eds), Studies of Welfare ous development to the following stage until Populations: Data Collection and Research the analysis is terminated’ (Glaser, 1969: 220). Issues. Washington, DC: National This procedure becomes a method of constant Academy of Sciences. pp. 220–74. comparison when interpreters take care that McLaughlin, R. H. (1999) ‘From the field to they compare codes over and over again with the courthouse: should social science codes and classifications that have already research be privileged?’, Law and Social been made. Material that has already been Inquiry, 24 (4): 927–65. coded is not ‘done with’ after its classification Palys, T. and Lowman, J. (2001) ‘Social but is continually integrated into the further research with eyes wide shut: the limited process of comparison. confidentiality dilemma’, Canadian The constant comparative method has been Journal of Criminology, 43 (2): 255–67. a starting point for more recent developments. Picou, J. S. (1996) ‘Compelled disclosure of Several authors suggest the minimal and max- scholarly research: some comments on high imal contrasting of individual cases for a com- stakes litigation’, Law and Contemporary parative interpretation of interviews or Problems, 59 (3): 149–58. biographies (see Flick, 2002 for an overview). The cases are structured according to recon- structive criteria in order to develop a typology from such interviews. Biographies with maxi- CONSTANT COMPARATIVE METHOD mal similarities are classified as groups, which are labelled as empirical types in the further Definition proceeding. For each type, specific everyday situations are distilled from the material and Part of the methodology developed by Glaser analysed across the individual cases. and Strauss (1967) for building grounded Developing ideal types and comparing them

37 C-Jupp-3373.qxd 3/8/2006 5:40 PM Page 38

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

with real types (of existing cases) is another G. J. McCall and J. L. Simmons (eds), elaboration of the constant comparative Issues in Participant Observation. Reading: method. This strategy includes the following Addison–Wesley. pp. 216–28. steps: after reconstructing and contrasting the Glaser, B. G. and Strauss, A. L. (1967) The cases with one another, types are con- Discovery of Grounded Theory: Strategies structed; then, ‘pure’ cases are tracked down. for Qualitative Research. New York: Compared to these ideal types of processes, Aldine. the understanding of the individual case can Strauss, A. L. and Corbin, J. (1990) Basics of be made more systematic: after constructing Qualitative Research. London: Sage. further types, this process comes to an end by structure-understanding that is the understanding of relationships pointing beyond the individual case. The main instruments CONSTRUCTIONISM are the minimal comparison of cases, which are as similar as possible, and the maximal Definition comparison of cases, which are as different as possible. They are compared for differences A perspective that considers facts, descriptions and correspondences. The comparisons and other features of ‘objective reality’ to be become more and more concrete with respect inescapably contingent and rhetorical. This is to the range of issues included in the empiri- a more recent formulation of constructionism cal material. The endpoints of this range (without the ‘social’) which follows the tradi- receive special attention in the maximal com- tional view of social constructionism as a parison, whereas its centre is focused in the perspective wherein people are seen as pro- minimal comparison. duced (constructed) through social interaction rather than through genetic programming and Evaluation biological maturation. The constant comparative method, contrasting cases and developing a typology from con- Distinctive Features trasting cases represent three ways to general- There are a number of different traditions of ization in qualitative research. Generalization constructionist thought. In feminist theory here is the gradual transfer of findings from there are arguments for the constructed rather case studies and their context to more general than essential nature of gender categories. In and abstract relations, for example a typology. educational research constructionism describes In the process of building grounded theories, a theory that emphasizes the activity of the these methods are as important as, for example, learner in constructing knowledge. In sociology theoretical sampling. Berger and Luckmann’s hugely influential book from 1966, The Social Construction of Uwe Flick Society, emphasized the way people experience reality through different constructions. Associated Concepts: case study method, Gergen (1999) distinguishes between some comparative method, grounded theory, theo- of the different constructionisms that have retical sampling developed in psychology: Constructivism has been used to describe the work of thinkers such as Jean Piaget and Key Readings George Kelly. There is a strong emphasis on Flick, U. (2002) An Introduction to Qualita- the constructions as something mental (like tive Research, 2nd edn. London: Sage. images or pictures) produced through inter- Glaser, B. G. (1969) ‘The constant compara- action with objects or some features of the tive method of qualitative analysis’, in external world.

38 C-Jupp-3373.qxd 3/8/2006 5:40 PM Page 39

CONSTRUCTIONISM

Social constructivism has been used for constructing the world rather than reflecting work where the constructions are something directs the researcher towards studying how mental, as with constructivism, but they are descriptions are made to seem objective and generated as much through social relation- independent of their producer, and how they ships and conversation as through interaction are used in particular actions. Potter reviews with objects. Lev Vygotsky and Jerome a range of procedures through which descrip- Bruner are key exponents of this perspective. tions can be made objective, including: Another variant of social constructivism (1) Category entitlements. Descriptions are comes in the form of social representations constructed as coming from a category theory, developed by Serge Moscovici. Its that is credible or knowledgeable in a emphasis is on mental representations and way that is relevant to the claim. how these are used to construct reality. (2) Stake inoculation. Descriptions are Social constructionism is distinct again for constructed as coming from someone giving a central role to discourse in the con- whose stake in that talk is counter to struction of self and the world. Gergen (1994) what you would expect when making usefully highlights some basic assumptions the claim. ‘I used to be sceptical, but for a social constructionist science. These experience of [the death penalty/workers’ include the following. First, the terms we use cooperatives/alien abduction/etc.] leads to describe ourselves and our worlds are not me to think …’ dictated by their objects. Second, the terms (3) Corroboration and consensus. Descrip- through which we understand the world and tions are constructed as corroborated by ourselves are social artefacts, produced over an independent witness (preferably from historic time through exchanges between an appropriate category) and/or something people within cultures. Third, any account that everyone agrees on. ‘Two or three of the world or self is sustained not by its other people started to look scared …’ objective validity but by social processes. (4) Active voicing. Descriptions are con- Finally, language derives its significance in structed using quotations and reports of human affairs from the way it functions thoughts to present the views and impres- within relationships. sions of others as corroborating, or to A major impetus behind modern construc- show the vivid or unexpected nature of tionism comes from sociology of scientific what is described. ‘Karen turned to me knowledge (SSK) where a series of highly and said “What the hell’s that?”’ detailed studies of various scientific and engineering controversies highlighted the The general point is that descriptions of contingency of scientific facts and technolog- things or people are constructed (how are ical artefacts. In contrast to standard views accounts built?) and constructive (how do of science, SSK researchers suggested that the accounts build the world?). Potter has devel- products of scientists are fabricated through oped this line of argument as a major feature social interaction between specific individu- of discursive psychology that treats all descrip- als in accordance with ad hoc criteria in idio- tions as constructions oriented to action. syncratic circumstances in which scientists Edwards, Ashmore and Potter (1995) have work in an opportunistic manner. generated considerable debate in defending a Potter (1996) has attempted to systematize radical version of constructionism. the field of constructionist thinking in discourse analysis, sociology of scientific knowl- Evaluation edge and post-structuralism. He focuses on constructionism as a practical position that Critics of constructionism tend to be highlights particular research topics. Rather concerned that it denies the ‘real’ – whether than searching for a fundamental philoso- this is ‘inner’ reality of the unconscious, of phical justification, the idea of discourse phenomenological experience, of the essence

39 C-Jupp-3373.qxd 3/8/2006 5:40 PM Page 40

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

of gender or of cognitive processes; or ‘outer’ Distinctive Features reality – of social structures, institutions and Content analysis was used as a way of study- inequalities. The former critiques tend to ing the contents of newspapers and it was come from psychoanalysis, psychologists developed most systematically during the taking a traditional cognitivist approach and 1940s in a series of studies on political critical psychologists embracing critical real- propaganda in the mass media. Under the ism. The latter critiques tend to come from var- direction of Harold Lasswell, this research ious forms of structural sociology, including aimed to uncover the intentions and interests Marxism. behind particular media products and to The value of constructionism lies in its show how these intentions and interests recognition of the way in which objects enter shaped the presentation of information and into social life through descriptions and in ideas in the media. They aimed to show, for focusing research on topics that previously example, how newspaper and radio presenta- had seemed natural, objective or universal. tion of information about wartime events could produce a distinctive message, or set Alexa Hepburn of messages, that would have the effect of producing particular attitudes and prefer- Associated Concepts: deconstruction, ences among the members of the audience. discourse analysis, discursive psychology, Content analysis formed a part, therefore, of phenomenology, realism, relativism a particular model of mass media production and effects in which it was combined with Key Readings the survey analysis of readers and audiences. Over the following decades, content analysis Edwards, D., Ashmore, M. and Potter, J. was systematized as a distinctive formal (1995) ‘Death and furniture: the rhetoric, method, separate from these particular politics and theology of bottom line argu- assumptions, that could be used as a step in ments against relativism’, History of the the analysis of any kind of documentary Human Sciences, 8: 25–49. source. Gergen, K. J. (1994) Realities and Relationships: In content analysis, the contents of a Soundings in Social Construction. Cambridge, document are analysed by the frequency MA: Harvard University Press. Gergen, K. J. (1999) An Invitation to Social with which particular categories of meaning Construction. London: Sage. are used. The aim is to identify clear and Potter, J. (1996) Representing Reality: Discourse, coherent categories that highlight salient Rhetoric and Social Construction. London: aspects of the message conveyed and to use Sage. objective and reliable methods of calculating their relative significance in the overall message. The categories used will vary from study to study, according to the nature of the CONTENT ANALYSIS material and the theoretical presuppositions of the researcher. Categories in content analy- Definition sis will, therefore, include such diverse things as positive, negative and neutral expressions A method of analysing the contents of docu- concerning political parties, women, medical ments that uses quantitative measures of the treatment, asylum seekers, God and street frequency of appearance of particular ele- crime. Very often, the aim is to devise cate- ments in the text. The number of times that a gories that grasp favourable or unfavourable particular item is used, and the number of attitudes or representations of these. For each contexts in which it appears, are used as mea- category, the particular words, phrases or sures of the significance of particular ideas or images that exemplify it must then be speci- meanings in the document. fied. The researcher takes the frequency with

40 C-Jupp-3373.qxd 3/8/2006 5:40 PM Page 41

CONTENT ANALYSIS

which the words, phrases and images appear complex ways – they may read them more or as indicators of the salience of the category of less superficially, they see them in the light of meaning. what they have been told by other people, and Holsti (1969) has said that such sets of they very often hear only what they want to categories must meet the formal criteria of hear. For this reason, the actually received mes- comprehensiveness, exhaustiveness, mutual sages among readers and audiences may be exclusiveness and independence. The compila- very different from the message disclosed by tion of a set of categories depends upon the a piece of content analysis. Content analysis range of materials from which they are may disclose the range of possible received selected. Content analysis must aim to cover messages, rather than specifying any one as as wide a range of relevant sources as possible. the message. A set of categories must be exhaustive in that Nor can content analysis give a completely they allow every relevant item to be classified valid picture of the intended message. into one of the categories; and the categories Producers of documents have their own agen- must be mutually exclusive in that it is not das and concerns, shaping their intentions and possible to classify any item into more than interests, and the original inspiration behind one category. These criteria reduce the ambi- content analysis was to disclose the effects of guity of classification. The criterion of inde- producer intentions. The process of produc- pendence means that the classification of any tion of a document, however, is a social one item must not affect the classification of process in which many different agents are any others. Each classification decision is involved – each with their own intentions – taken on its own terms. and in which the process itself has a logic and Once the initial categories have been chosen momentum that ensures that the final, inter- and the principles of classification have been nal meaning of the document cannot simply decided, it is possible to apply them to a par- be taken as an expression of the intentions of ticular discourse or set of documents and to a particular group of producers. generate a quantitative picture of the range of Content analysis, therefore, is a useful and meanings conveyed in the document. important tool of documentary analysis, providing objective and rigorous methods for Evaluation investigating social meanings. Its results, however, correspond neither to the intended Although content analysis is built around the nor to the received messages of the docu- construction and use of objective and reliable ment, and it must be used with caution. procedures, the validity of its conclusions are no more unambiguous than those of more qualitative approaches. The meaning of a John Scott particular word, phrase, or image depends upon its context. A simple word, such as Associated Concepts: coding, diary, docu- ‘freedom’, can mean any of a large number of ment analysis, media analysis, quantitative different things depending on the context and research, textual analysis the time at which it appears. Any judgement of context relies on a body of background knowledge and assumptions. Key Readings Content analysis discloses, at best, the Berelson, B. (1952/1971) Content Analysis on ‘internal’ meaning of a document: it discloses Communication Research. New York: the meaning that the text would convey to a Hafner. reader who employed reading techniques sim- Holsti, O. R. (1969) Content Analysis for the ilar to those used by the researcher. For most Social Sciences and Humanities. Reading: readers and audiences this is not the case. MA: Addison–Wesley. Audience research has shown that people see, Weber, R. P. (1990) Basic Content Analysis, hear and read documents in a variety of 2nd edn. Beverly Hills, CA: Sage.

41 C-Jupp-3373.qxd 3/8/2006 5:40 PM Page 42

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

terms of questions such as ‘What’s going on?’ CONVERSATION ANALYSIS and ‘Why that now?’. Then, after locating a structurally recurring phenomenon in the Definition conversation, the analyst will collect the The detailed study of the methods used by same phenomenon in the whole corpus and participants to achieve social practices such possibly also in other contexts. as greetings, giving directions and telling Harvey Sacks (who died prematurely in stories in a variety of contexts (formal, infor- 1975) is considered the founding father of mal). The term ‘conversation’ does not refer this field. In his lectures at Irvine and UCLA so much to a given type of speech (a genre) as (1964–1972) and his collaborative research to the idea of ‘talk-in-interaction’ as a social with Goffman, then Garfinkel, Schlegoff and activity. Jefferson, he developed a radically empirical approach to the study of social behaviour in context, which had relevance to the whole Distinctive Features field of sociology. As a field of research, conversation analysis In the field of conversation analysis, emerged in the 1960s and 1970s in the United conversation is a way of approaching social States and evolved from the work of a group order. Conversation is far from the apparently of sociologists drawing inspiration from unorganized and random object it might seem ethnomethodologist Harold Garfinkel. at first. It is a locus for social interaction Conversation analysts use audio and/or where speaker-change recurs (or occurs), par- video recordings of carefully transcribed talk- ties talk at a time; transitions with no gap and in-interaction. Recording and transcribing data no overlap are common. This orderliness is drawn from ‘natural’ situations (that is, with- exhibited by speakers according to what has out intervention by the researcher) is a been described by Sacks, Schlegoff and methodological advantage for social studies. It Jefferson (1974) as turn-taking organization. guarantees a more reliable corpus than note Turn is the fundamental unit of description in taking, enables the analyst to study the same conversation analysis. It can be defined as the fragment over and over and makes it easier to length of time a speaker holds the floor. share interpretations with other researchers. Length and turn constructional units are There are no a priori categories or hypotheses constantly negotiated by speakers as they involved in the process of analysing the interact. A turn can be anything from any conversational data since they are themselves audible sound, to a single word, a clause, a ‘research-generating’. Contrary to other social sentence, a narrative. Another feature of this sciences with an exogenous approach of type of organization is projectability. Through context, conversation analysis uses an endoge- the turn-taking system, hearers can detect first nous approach. In this respect, it is not possible completion of the current speaker’s because an interaction takes place in a hospi- turn. This point is called transition relevant tal or a tribunal that it should be interpreted as place, that is the moment at which change of necessarily ‘formal’. Rather, it is the partici- speaker may take place. Turn-allocational pants’ behaviours (and not the decision of the techniques centre around three options: analysts) within this specific interaction that (1) a current speaker may select a next will set off particular elements of the context. speaker (for example, asking a question); This only becomes relevant for its interpreta- (2) if (1) is not used, then parties may self- tion if some of its features have a bearing on select in starting to talk; the structure of the interaction (size and order- (3) if (2) is not used, then the current ing of turns, type of sequence …). speaker may continue to talk. Analytical inquiry of corpora tries to account for what has occurred in a conversation by One fundamental kind of sequential orga- reconstructing the members’ perspectives in nization is constituted by adjacency pairs

42 C-Jupp-3373.qxd 3/8/2006 5:40 PM Page 43

CORRELATION

(question/answer, greeting/greeting, offer/ analysis (cognition in context, communication acceptance …) with the following features disorders, work organization; grammar in (Schegloff and Sacks, 1973): interaction). (a) the sequence is composed of two adjacency turns issued by two different Luca Greco speakers; (b) given the first, the second is expectable Associated Concepts: constructionism, (conditional relevance rule); empiricism, ethnomethodology, grounded (c) when the second part is missing (for theory, hermeneutics, interpretive reper- example, the answer in the ‘question/ toires, narrative analysis, phenomenology, answer’ pair), its absence is pointed out videography by one of the speakers; (d) this mechanism provides a frame for Key Readings interpretation: by producing a second pair, speakers display their understanding Atkinson, J. M. and Heritage, J. (eds) (1984) of what the first pair is actually doing. Structures of Social Action. Cambridge: Cambridge University Press. In adjacency pairs, the second pair often Sacks, H. ([1964–72] 1992) Lectures on presents two alternative courses of action (for Conversation (2 vols) (ed. by Gail example, invitation/acceptation or refusal). Jefferson, with an introduction of Preference is the term that researchers have Emmanuel A. Schegloff). Oxford: Basil chosen to indicate the series of possibilities Blackwell. where the unmarked alternative, usually Sacks, H., Schegloff, E. A. and Jefferson, G. short and simple (for example, ‘yes’), is (1974) ‘A simplest systematics for the referred to as ‘preferred’; whereas the ‘dis- organisation of turn-taking for conversa- preferred’ alternative can be marked by tion’, Language, 50: 696–735. pauses, hesitations, excuses, mitigations and Schegloff, E. A. and Sacks, H. (1973) ‘Opening justifications. up closings’, Semiotica, 8: 289–327. Preference is also at work in repairs Schegloff, E. A., Jefferson, G. and Sacks, H. performed by speakers when they detect a (1977) ‘The preference for self-correction problem concerning speaking, hearing or in the organisation of repair in conversa- understanding during a conversation. tion’, Language, 53: 361–82. Schlegoff, Jefferson and Sacks (1977) have noticed that speakers (in certain situations) tend to prefer ‘self-repair’ versus ‘other-repair’. So that when a speaker notices a trouble CORRELATION source in the other speaker’s turn, he may have recourse (rather than directly using repair) to ‘repair initiations’ such as ‘What?’, Definition ‘You mean X?’, ‘Do what?’, generating a self- Correlation refers to the linear relationship repair in the next turn of the addressee. between two variables. The correlation coefficient is a measure of the association between Evaluation two numerical variables, usually denoted as x and y. It is a symmetrical relationship: if x is Conversation analysis is currently applied in correlated with y, y is correlated with x. a range of disciplines sharing an interest in social interaction (anthropology, linguistics, Distinctive Features psychology); with interesting issues concerning methodology (use of video technology, The value of the correlation coefficient ethnographic fieldwork, quantification) and lies between +1 and −1. A positive coefficient

43 C-Jupp-3373.qxd 3/8/2006 5:40 PM Page 44

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

indicates that a high value of x tends to be by x, but this should not be interpreted as associated with a high value of y and a nega- meaning that 41 per cent of y is caused by x. tive coefficient indicates that as the value of x increases the value of y is likely to decrease. Evaluation A coefficient of +1 is a perfect positive correlation between x and y, while a coefficient Correlation does not indicate causality. This of −1 is a perfect negative correlation (some- means that x may cause y, or y may cause x, times referred to as an inverse correlation). or both may be caused by another, interven- A coefficient of 0 means that there is no rela- ing variable (z), or by a set of other variables tionship between the two variables. (z1, z2, ... zk). The influence of an interven- The Pearson product moment correlation ing variable can be controlled for by comput- coefficient (r) is calculated for data that are at ing a partial correlation coefficient. the interval level of measurement at least. In addition to calculating the correlation The correlation coefficient is calculated by the coefficient, it is also advisable to view a scatter formula diagram to gain a better appreciation of any relationship, or its absence. Sometimes other information about the units can explain a cor- (x − x¯)(y − y¯) relation or the lack of it (For example, a trend r = (x − x¯)2 (y − y¯)2 with time, or correlations between subgroups rather than individual units.) There may be a linear relationship between variables, where change in one variable takes The size of the resulting coefficient gives place at the same rate as change in another some indication of the strength of the associ- variable. Alternatively, there may be a curvi- ation. Its statistical significance can be read linear relationship, in which the rates of from a book of statistical tables or, if the num- change between variables varies. For exam- ber of pairs is greater than 100, by using the ple, people are more likely to need medical standard error of r. The degrees of freedom attention when they are very young and very equal the number of pairs minus two. Where old than at other times in their lives. If one the data are at the ordinal level of measure- were to plot a scatter diagram of medical con- ment, Spearman’s rank order correlation tact with age this would show a relationship (rho) can be used. that is higher at the two extremes than in the Where a data set containing a number of middle. Such a relationship would yield a variables is involved, it is possible to compute very low or negligible correlation coefficient, a correlation matrix showing the correlation but would be more apparent from a scatter coefficients of every suitable variable with plot. Truncation of one variable (that is, only every other suitable variable. When examining measuring a small part of the possible range) the significance levels of a large matrix, how- will reduce the correlation. Lack of correla- ever, the researcher needs to remember that tion may be misleading for this reason. some coefficients are likely to be statistically Small samples cannot detect low correla- significant by chance. tions. Large samples may enable low correla- If the correlation coefficient is squared it tions to be detected. A low correlation only gives the proportion of the variance of one indicates a slight relationship, but if it is variable that is ‘explained’ by the other vari- significant it shows that there is an association. able. Since a proportion multiplied by 100 If a coefficient is not statistically significant gives a percentage, this enables one to know it may be due to chance, and should not be what percentage of y is explained by x. Thus, interpreted. if r = 0.64, then r2 = 0.41, and we can say that Care should be taken not to assume that a 41 per cent of the variation in y is explained correlation between units such as geographical

44 C-Jupp-3373.qxd 3/8/2006 5:40 PM Page 45

COST–BENEFIT ANALYSIS

locations means that a similar correlation forensic evidence in police investigations have applies to the individuals within those areas. often balanced the costs of achieving a foren- This is the ecological fallacy. sic identification with the likelihood of a ‘positive’ outcome, such as an arrest. In the UK the importance of cost–benefit Iain Crow analysis has increased significantly in recent years as a result of three major factors: first, Associated Concepts: causality, descriptive central government control of certain public statistics, ecological fallacy, hypothesis, infer- agencies has increased; secondly and related ential statistics to this, there has been an impetus to allocate scarce resources effectively; and thirdly, Key Readings there has been a move towards evidence- based approaches to social interventions Fielding, J. and Gilbert, N. (2000) ‘Interval allowing for such analysis to be made. data: correlation and regression’, in Agencies within the UK criminal justice Understanding Social Statistics. London: Sage. ch. 8. system have been particularly concerned Norusis, M. J. (1992) ‘Measuring linear asso- with such analysis because it has been ciation’, in SPSS for Windows, Base System increasingly subject to central government User’s Guide, Release 5.0. Chicago: SPSS control and research. Most importantly Inc. ch. 17. perhaps, increases in crime rates have Pallant, J. (2001) ‘Correlation’, in SPSS resulted in increasing costs throughout the Survival Manual. Milton Keynes: Open system. Costs of criminal justice inputs such University Press. ch. 11. as police service strength, prison numbers and the use of community penalties, must ultimately be balanced against the effects on crime rates. In this way, agencies and organizations are able to make informed decisions COST BENEFIT ANALYSIS – about how best to employ scarce resources. Definition Evaluation Also known as cost-effectiveness analysis, cost–benefit analysis involves a comparison Typically, cost–benefit analysis allows for of the cost of inputs of an intervention or comparison between the inputs and out- initiative, with the outputs and outcomes of puts/outcomes of an initiative but without that intervention. It is used in all areas of taking into account any of the processes that strategic decision making. The focus here is occur in between, or external factors that may on cost–benefit analysis by public bodies impact on a research project. Furthermore, it rather than commercial enterprises. may fail to acknowledge any negative impact or unmeasured positive impact that an initia- Features tive may have. For example, cost–benefit analysis of expensive high-visibility policing Cost–benefit analysis is typically employed in operations may reveal a temporary and spatial applied research, especially evaluations, in benefit to crime rates (the police deter offend- order to inform an organization’s decisions on ers in that area whilst they are present), but resource allocation between varying options. may ignore the fact that offenders are Accordingly, cost–benefit analysis often forms displaced, that police are removed from other the basis of best-practice or best-value guides (potentially beneficial) duties, or positive for central government agencies such as the impacts on the public’s confidence in the police. For example, evaluations of the use of police and reductions in the fear of crime.

45 C-Jupp-3373.qxd 3/8/2006 5:40 PM Page 46

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

It is therefore implicit that cost–benefit variables. Note that the presence of analysis in all areas of research is used only covariance, in the absence of solid theory as part of a wider evaluation of the impacts of and/or research design, does not imply the a project. Nevertheless, in an age of account- presence of a causal relationship. ability, it is important to be able to demonstrate that public funds are being deployed Distinctive Features wisely and effectively. Three basic types of covariance exist. Positive covariance exists where as one variable Helen Poole increases in value, the second variable increases in value as well. Negative covariance Associated Concepts: applied research, is where two variables tend to decrease in auditing, evaluation research, good practice value together. Null covariance is the state of studies, organizational research, performance observed independence between two vari- indicator, process mapping ables. Note that the observation of independence in a sample might not reflect the reality in the population. Furthermore, if the under- Key Readings lying relationship between two variables is not linear, measures assuming linearity might esti- Dhiri, S. and Brand, S. (1999) Analysis of Cost and Benefits: Guidance for Evaluators. mate no relationship. (Crime Reduction Programme Guidance There are two general tests of covariance. Note 1.) London: Home Office. One establishes the existence of a relation- Laycock, G. (2002) ‘Methodological issues in ship observed in a sample of a larger popula- working with policy advisers and practition, while the other estimates the strength of tioners’, in N. Tilley (ed.), Analysis for this association. Assuming the presence of a Crime Prevention. Devon: Willan. random probability sample, it makes sense to discuss the first before moving on to the second. Basic analysis of covariance simply establishes the existence of a relationship, which is necessary before considering the COVARIANCE strength of an observed relationship. It is common to employ a chi-square test (χ2) Definition when the level of measurement of the variables under study is either nominal or ordi- A condition where two measures vary nal. For interval, ratio and other continuous together, such as the level of educational measures, the F test or t-test are used to test attainment and income. In general language, for significance. this might also be referred to as association or Once the existence of a relationship has correlation. In the case where the analysis is been established (or in cases where one has conducted on a random sample, covariation access to the universe under study), one can be tested statistically to determine the moves on to estimate the strength of the probability that an observed relationship is relationship. Again, the particular measure not real but the product of random chance employed is a function of the level of analysis in the sample. Additionally, an assortment of of the underlying variables. Yule’s Q is a measures of association exist that estimate basic test for nominal data; under certain cir- the degree of covariance, the selection of cumstances it is advisable to employ Lamda. which depends upon certain parameters of With ordinal data, Gamma and Tau are com- the measures under study. The most common monly employed to estimate the strength of parameter that affects the measure of covari- association, while Pearson’s r is common to ance is the level of measurement of the two estimate relationships at the interval level.

46 C-Jupp-3373.qxd 3/8/2006 5:40 PM Page 47

COVERING-LAW MODEL OF EXPLANATION

Evaluation this model, the defining feature of a scientific explanation rests on the operation of general One limitation of most statistical estimates of scientific laws and particular ‘initial condi- covariance is that they are restricted to the tion statements’ (together known as the particular properties of the two variables explanans) which logically entail the pheno- under analysis. Comparisons across different menon thus explained (the explanandum). paired sets of variables are difficult, as most estimations of covariance are dependent Distinctive Features upon the scale of measurement for the con- stituent variables. It is possible to sidestep The basic D–N model is as follows: this problem in multivariate regression models by comparing the beta coefficient for (1) law(s); several independent variables, which is stan- (2) statement(s) of particular initial condi- dardized by measuring the change in the tions, which point to the applicability of standard deviation of the dependent variable the chosen law(s) to the case in hand; for every one standard deviation in the inde- (3) the phenomenon explained (and pre- pendent variables. dicted), that is as the outcome, as a deduction from, the above. The laws involved are couched as universal David Brockington conditional statements: ‘For any a, if a, then b’. For example, ‘Water heated at the pressure at Associated Concepts: causality, correlation, sea level boils at 100 oC’. descriptive statistics, inferential statistics, Whilst the D-N model thus identifies sci- measurement, multivariate analysis, proba- ence as involving deterministic laws, ‘probability (random) sampling, variable analysis bilistic laws’ (inductive–statistical or I–S model – asserting ‘if a, then a certain proba- Key Readings bility of b’) can also be accommodated as can Schroeder, L. D., Sjoquist, D. L. and ‘functional explanations’ (involving self- Stephan, P. E. (1986) Understanding regulating systems) in some forms. Regression Analysis: An Introductory Guide. A crucial distinction made on the basis of Sage Series in Quantitative Applications the model is between scientific and non- in the Social Sciences. London: Sage. scientific prediction: rather than being Wildt, A. R. and Ahtola, O. T. (1978) Analysis unconditional, scientific prediction is condi- of Covariance. Sage Series in Quantitative tional on the occurrence of the relevant initial Applications in the Social Sciences. conditions. The model is proposed as provid- London: Sage. ing a unified account of the role of explanation, prediction and test in science, based on a suggested ‘logical symmetry’ of explanation and prediction. COVERING-LAW MODEL OF Evaluation EXPLANATION Criticism of the covering-law model has Definition focused particularly on its claims to provide a universal, defining model of scientific expla- (Also known as ‘deductive–nomological nation. There is now general acceptance that explanation’ or D–N), a model of scientific subsumption under empirical ‘covering laws’ explanation particularly associated with the is a sufficient or a necessary condition for philosophers Carl Hempel and (in a more satisfactory scientific explanation. Is, using particular form) Karl Popper. According to Boyle’s Law, a change in the pressure of a gas

47 C-Jupp-3373.qxd 3/8/2006 5:40 PM Page 48

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

explained by a change in volume? Explanation for Max Weber, a half-way house between usually depends on an associated theoretical (and blending of) explanation by covering embeddedness for laws and often causal laws and more particular explanations. assumptions beyond mere ‘constant conjunc- For all the controversies, explanation and tion’. Also, successful claims to explanation prediction in terms of covering laws (whether can be formulated without these always deterministic or statistical) remains a strong being dependent on empirically observable feature of the practical world of science. In regularities or straightforwardly the basis of social science, for example, even in the con- prediction. One suggestion made (for exam- spicuous absence of a body of agreed univer- ple, in ‘scientific realism’ as outlined by sal laws, hypothesis testing (including the Bhaskar, 1978) is that scientific explanation testing of ‘null hypotheses’ and ‘causal mod- instead involves the identification of underly- elling’) can be seen as an explicit or implicit ing causal ‘powers’ or ‘mechanisms’ that use of the covering-law model. may not be associated with straightforward empirical covering laws or allow prediction. For example, while Darwinian theory pro- David Jary vides a highly important explanatory account of biological evolution this is couched in Associated Concepts: causal model, causal- terms of general mechanisms (for example, ity, falsification, hypothesis, hypothetico- ‘natural selection’), which do not predict deductive model, ideal type, positivism, particular variations or new species. realism In the social sciences in particular (not least in historiography, in relation to which Hempel especially argued for his model) it is Key Readings apparent that significant kinds of explanation Bhaskar, R. (1978) A Realist Theory of Science, exist which have cogency and validity but are 2nd edn. Hemel Hempstead: Harvester. dependent neither on covering law nor on Hempel, C. (1965) Aspects of Scientific general mechanisms. These include ‘inten- Explanation. New York: Free Press. tional explanations’ (a person’s reasons for Popper, K. (1934, English translation, 1959) undertaking a specifiable course of action) The Logic of Scientific Discovery. London: and ‘particular historical explanations’ (for Hutchinson. example, ‘the Danes plundered England Wright, G. von (1971) Explanation and before they settled’) and also, at least in some Understanding. London: Routledge & Kegan versions (for example, where conscious Paul. design is involved) ‘functional explanations’. Thus, while the covering-law model may be seen as adequately portraying one form of scientific explanation there is no widespread acceptance that it provides an adequate over- COVERT RESEARCH all model of physical science, or still less, of the range of explanatory modes (especially Definition modes of ‘meaningful understanding and interpretation’) employed in the social sci- Research that is undertaken without the con- ences and humanities (see Wright, 1971). sent or knowledge of respondents. This type However, ‘ideal types’ (for example, ‘rational of social research is most strongly associated action’, ‘perfect competition’), as heuristic with participant observational work where a aids to explanation or as parts of formal the- researcher joins a group or organization ory, allow exploration of the factors involved assuming a covert role in order to observe in empirical ‘departures’ from these ‘ideal- first hand the functioning and daily life of the ized’ forms and may be seen as providing, as group.

48 C-Jupp-3373.qxd 3/8/2006 5:40 PM Page 49

COVERT RESEARCH

Distinctive Features difficulty is one of accurately interpreting events and successfully translating these Covert research has a strong tradition in soci- events into academic writing, what Pearson ological and criminological work. Early (1993) has referred to as ‘the requirement to ethnographers such as Whyte (1943) wished carry the narrative back home’ (p. xvii). The to explore the social world using naturalistic problem is that the ‘reality’ may be lost in the techniques in the same way that anthropolo- researcher’s interpretation. gists have sought to understand different Covert research is ethically problematic as cultures through first-hand experience. it transgresses ethical codes of conduct such as Covert observation has been used with those produced by the American Sociological difficult-to-access or deviant groups. Gaining Association, British Sociological Association access can be difficult. This is often achieved and the Social Research Association. One of the through a key informant who is a part of the key ethical principles in research is that of group or organization and may know the real informed consent, the understanding that purpose of the researcher’s presence, and such those participating in research should be a person can act as a gatekeeper allowing fully informed about its nature and purpose access to participants and events. Adopting a before they agree to participate. Indeed some believable and acceptable role with the group researchers gain written consent. Covert can also be problematic, and sometimes researchers cannot gain informed consent from researchers choose to covertly observe an participants as the work is based upon deceit organization that they are already a part of. A and dependent for its success upon participants good example of this is Holdaway’s (1983) being unaware of the researcher’s real role. account of police work conducted covertly Those conducting this type of research would while he worked as a police officer. Managing argue that this is justified given what the the role and developing a successful exiting research adds to our understanding of the strategy are also important issues, given that complexities of the social world. the researcher’s presence is built around deceit and close relationships may be established during the fieldwork with participants. Julia Davidson There is also a danger that researchers become too immersed in the group they are studying Associated Concepts: access, dangerous and lose objectivity. Further, there are practi- fieldwork, ethnography, fieldwork, gate- cal difficulties associated with data recording keeper, informed consent, naturalistic data, and researchers must recall events and infor- organizational research, participant observa- mal conversations at a later time or carry a tion, validity of measurement concealed tape or video recorder.

Evaluation Key Readings Bryman, A. (2001) Social Research Methods. Covert observational research can provide Oxford: Oxford University Press. a rich account of experience and context. Hammersley, M. and Atkinson, P. (1995) However, relatively little covert research is Ethnography: Principles in Practice. undertaken by social scientists as it is time- London: Routledge. consuming. Given the high cost of fieldwork Hobbs, D. and May, T. (eds) (1993) Inter- and doubts about the validity (which are preting the Field: Accounts of Ethnography. often misplaced) of such research, key fund- Oxford: Clarendon Press. ing bodies are reluctant to support this type Holdaway, S. (1983) Inside the British Police. of work. Oxford: Blackwell. Ethnographers’ claims to capture the fab- Norris, C. (1993) ‘Some ethical considera- ric and reality of life through covert observations on field work with the police’, in tional methods have been criticized. The D. Hobbs and T. May (eds), Interpreting

49 C-Jupp-3373.qxd 3/8/2006 5:40 PM Page 50

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

the Field: Accounts of Ethnography. Oxford: Resistance: The Culture and History of a South Clarendon Press. African People; Michael Taussig’s Shaminism, Pearson, G. (1993) ‘Talking a good fight: Colonialism and the Wild Man; Michelle Fine authenticity and distance in the ethnogra- and Lois Weiss’s The Unknown City; Kathleen pher’s craft’, in D. Hobbs and T. May Stewart’s A Space on the Side of the Road. (eds), Interpreting the Field: Accounts of While difficult to characterize in terms of a Ethnography. Oxford: Clarendon Press. single set of features, critical ethnographies in Whyte, W. (1943) Street Corner Society. the main are marked by several shared dispo- Chicago: University of Chicago Press. sitions: a disavowal of the model of ethnographer as detached, neutral participant observer; a focus on specific practices and institutions more so than holistic portraits of CRITICAL ETHNOGRAPHY an entire culture; an emancipatory versus a solely descriptive intent; and a self-referential form of reflexivity that aims to criticize Definition the ethnographer’s own production of an Studies of social practices and cultural insti- account. This last feature may perhaps be the tutions which specifically aim to be critical of most telling for what constitutes a critical the taken-for-granted social, economic, cul- ethnographic investigation. Critical social tural and political assumptions and concepts science, in general, aims to integrate theory (e.g., family, work, self, agency, power, con- and practice in such a way that individuals flict, race, class, gender) of Western, liberal, and groups become aware of the contradic- middle-class, industrialist, capitalist societies. tions and distortions in their belief systems and social practices and are motivated to Distinctive Features change those beliefs and practices. A critical theoretical approach to social investigation Critical ethnographies are focused, theorized links hermeneutic (interpretive) and explana- studies that aim to change awareness of spe- tory interests to normative concerns. cific social institutions or practices, or of life However, this approach is never innocent; it itself. They may engage in ideology critique is never merely a theory ‘about’ the world. or demystification, showing, for example, Because critical social scientists assume that interests hidden behind, or vested in, cultural their very ways of theorizing the world con- meanings and practices or revealing forms of stitute the ways we access the world (theories domination or power. They may employ the provide the categories through which we technique of juxtaposing Western practice think about and experience the world), they against the ‘other’ practice in order to probe interrogate (and frequently disrupt and and make problematic understandings decentre) their way of theorizing by means of assumed in each. Theoretical sources for cri- reflexive critique. tique are multiple and include ideas drawn from cultural Marxism, critical theory of the Evaluation Frankfurt School, praxis theory, cultural studies, feminist studies, racialized and queer Critical ethnographers argue that conditions epistemologies, and the broad postmodern of social, economic and political inequality, critique of advanced capitalist societies. Some false ideology and distorted self-understanding examples of the rich variety of studies are real characteristics of society. They claim informed by these different and sometimes they can document and re-present these overlapping theoretical frameworks include morally unacceptable conditions. Yet, they Paul Willis’s Learning to Labour; Peter also acknowledge that all social inquirers par- McLaren’s Schooling as a Ritual Performance; ticipate in regimes of power/knowledge – they Jean Comaroff’s Body of Power, Spirit of are part of formal modern institutions that

50 C-Jupp-3373.qxd 3/8/2006 5:40 PM Page 51

CRITICAL RESEARCH

exercise power through the production and whether quantitative or qualitative – that management of knowledge about society. make claims of scientific objectivity. Rather, Thus, critical ethnography, particularly in its critical social research attempts to reveal the encounter with postmodern insights regard- socio-historical specificity of knowledge and ing deconstructing the social construction and to shed light on how particular knowledges production of knowledge, is a rich site for reproduce structural relations of inequality examining the meaning, limits and authority and oppression. of social theorizing. Distinctive Features Thomas A. Schwandt Critical social research, in the sense of offering critiques of social order, has a long history Associated Concepts: critical research, encompassing the likes of Aristotle, Socrates, ethnography, emancipatory research, Hobbes and Marx. It is currently evident in hermeneutics, postmodernism, reflexivity, the work of a wide variety of social critics, value-free research including Marxists, feminists and post- structuralists and is often ‘hailed’ through Key Readings attaching the adjective of ‘critical’ to any number of existing disciplines or methodolo- Carspecken, P. F. (2002) ‘The hidden history gies, as in ‘critical sociology’, ‘critical dis- of praxis theory within critical ethno- course analysis’, ‘critical anthropology’, graphy and the criticalism/postmod- ‘critical psychiatry’, ‘critical criminology’, ernism problematic’, in Y. Zou and E. T. ‘critical ethnography’ and so on. Yet, unlike Trueba (eds), Ethnography and Schools. other methodological traditions, the defining Lanham, MD: Rowman & Littlefield. characteristics of a critical approach have Kincheloe, J. and McLaren, P. L. (2000) rarely been spelt out. Its key features remain ‘Rethinking critical theory and qualitative elusive, though they are generally understood research’, in N. K. Denzin and Y. S. Lincoln (eds), Handbook of Qualitative Research 2nd to involve a critique of conventional research edn. Thousand Oaks, CA: Sage. agendas and methodologies by pointing out Levinson, B., Foley, D. and Holland, D. (eds) their (frequently hidden) ideological and (1996) The Cultural Production of the oppressive nature. Its intent is to expose Educated Person: Critical Ethnographies of enduring structures of power and domina- Schooling and Local Practice. Albany, NY: tion, to deconstruct the discourses and narra- State University of New York Press. tives that support them and to work as Marcus, G. E. (1998) Ethnography Through advocates for social justice. Thick and Thin. Princeton, NJ: Princeton The term ‘critical theory’ is widely associ- University Press. ated with the Frankfurt Institute for Social Marcus, G. E. and Fischer, M. M. J. (1986) Research founded in 1922. In particular, Anthropology as Cultural Critique. Chicago: members of the ‘Frankfurt School’ – such as University of Chicago Press. Adorno, Marcuse and Horkheimer – provided a damning critique of scientism and rationalism by revealing the significance of values and beliefs as key constituents of culture. CRITICAL RESEARCH Technical and one-dimensional forms of reason had subverted and eclipsed critical Definition reasoning about moral and political values. So, they argued, it was precisely the ethos of value A generic term usually applied to any neutrality, so revered by academics in 1930s research that challenges those conventional Germany, that had created the moral vacuum knowledge bases and methodologies – to be filled by fascism. The solution lay in

51 C-Jupp-3373.qxd 3/8/2006 5:40 PM Page 52

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

reconnecting reason with questions of ethics, disciplines within which, or against which, nature and morality. Value neutrality was a it works. As a result, it has been condemned dangerous illusion, to be avoided at all costs. as an ‘empty rhetorical shell’ (Hammersley, This emergent critical paradigm was fur- 1995). It does, however, continually remind ther developed by the fracturing of main- us that social research should not, indeed stream paradigms in the social sciences cannot, stand on the sidelines of society. associated with the civil rights, student Research is a value-laden activity, from the protest, anti-war, counter-culture and feminist choice of research subject and the questions movements of the 1960s. Empirical science, to investigate through to the interpretations structural functionalism and notions of a ‘com- and publication of results. Critical research mon value system’ came to be increasingly cri- is not afraid to reveal its theoretical, political tiqued as the tools of a ‘power elite’. Notions of and ideological underpinnings. It actively scientific neutrality, especially when practised engages with and challenges dominant through American corporate and military poli- assumptions and ‘taken for granted’ ways of cies, were condemned as little more than a knowing. Its ‘critical’ credentials are espe- convenient cover for systematic forms of con- cially clear when it is employed to reveal the trol at home and subversion abroad. Orthodox hidden agendas, partialities and limitations social scientists were singled out as mere of ‘official’ research and when it is part of an ‘technicians of power’ (Gouldner, 1971). The ongoing process of, and advocacy for, radical time had come, as Howard Becker put it, to social change. As a result it is always likely to ask ‘Whose side are we on?’. suffer from mainstream political arguments At this time a related research style evolved that attempt to discredit it as subjective, over- which was designed to sustain social criticism politicized and thus not ‘real research’. It is and to facilitate radical social change. Lehmann no surprise that it is consistently denied any and Young (1974), for example, talk explicitly significant research funding. of a ‘conflict methodology’ in which the research agenda is shifted from the powerless John Muncie (the usual subjects of conventional research) to exposing the exploitations and scandals of the Associated Concepts: critical ethnography, powerful. Research as exposé and as advocacy deconstruction, emancipatory research, femi- for the powerless becomes indivisible. Critical nist research, standpoint research, politics social research would not, however, claim any and social research, value-free research one direct political affiliation. Rather, it is committed to delving beyond surface appearances in order to reveal oppressive and discrimina- Key Readings tory social processes and structures and thereby suggests ways in which the oppressive Gouldner, A. (1971) The Coming Crisis of can be challenged through praxis (Harvey, Western Sociology. London: Heinemann. 1990). In this senses it is less a specific method Hammersley, M. (1995) The Politics of Social and more a theoretical, empirical, reflexive and Research. London: Sage. Harvey, L. (1990) Critical Social Research. engaged process informed by various ‘stand- London: Sage. point’ commitments to social justice and Hudson, B. (2000) ‘Critical reflection as human rights (Hudson, 2000). research methodology’, in V. Jupp, P. Davies and P. Francis (eds), Doing Evaluation Criminological Research. London: Sage. pp. 175–92. Critical social research does not occupy a sin- Lehmann, T. and Young, T. R. (1974) ‘From gle well-defined space. It is theoretically and conflict theory to conflict methodology: methodologically diverse. Its meaning is an emerging paradigm for sociology’, probably only captured through the particular Sociological Inquiry, 44: 15–28.

52 C-Jupp-3373.qxd 3/8/2006 5:40 PM Page 53

CROSS-SECTIONAL SURVEY

Security registers, to a ‘multi-stage, dispro- CROSS-SECTIONAL SURVEY portionately, stratified clustered’ sample in which sub-samples of the main sample may Definition be boosted for certain purposes and the data Any collection of data from a sample of indi- collected in from whole groups of individuals viduals (or groups) at a particular point in time in a limited number of locations rather than as a basis for inferring the characteristics of across the country as a whole. Another the population from which the sample comes. common type of design dispenses with the A cross-sectional survey of a population can be probability sampling principle and simply one-off or repeated at regular intervals thereby constructs the sample from individuals providing a means of ‘monitoring’ changes in selected purposefully to meet certain quotas the population in response to societal and considered important to the subject matter of policy change. A population census is in effect the survey, such as gender, social class, edu- a ‘bench mark’ social survey in which all pop- cational level, type of housing and so on. ulation members form the sample. Because such samples are selected on non- random principles, it is not strictly possible to estimate the standard error of the sample for Distinctive Features them and consequently the generalizability of Cross-sectional sample surveys are described the survey results is limited to this extent. as ‘descriptive’ or ‘analytic’ depending on whether the prime purpose is to provide Evaluation descriptive estimates of the population’s parameters or to test hypotheses about the relation- Social surveys are powerful tools in social ships between the variables encompassed by it. science that offer a quick and efficient means Analytic surveys build into the design a wide of collecting information about a range of vari- range of variables to which statistical models ables in a sub-sample of the population. can be applied. Such approaches require fairly However, to do this effectively is expensive, strong assumptions about temporal sequenc- because of the time and cost involved in drawing, which the data are often unable to support. ing up sampling frames, selecting a sample There is always the problem of ‘endogeneity’, and then carrying out the fieldwork. Much whereby it is uncertain when investigating the survey work, however, is not conducted to relationship between two variables, such as these demanding standards and is typically educational attainment and educational aspira- done on a one-off basis by students for disser- tions, as to which one is a function of (or caus- tation purposes or for organizations who want ing) the other. However, hypothesized causal quick snap-shots of particular groups of relationships that are not borne out by survey people, such as employees or local residents data provide a basis for falsification of hypothe- and so on. In these instances, the generaliza- ses. For example, if we hypothesize that girls tions drawn from the survey data are often dif- are discriminated against compared with boys ficult to sustain. Nevertheless, the results do in relation to certain occupational opportuni- provide basic descriptions of the people ties, and the survey shows no relationship involved in the survey, which gives some indi- between gender and occupational opportunity, cation of the prevalence of behaviours and atti- then the hypothesis falls. tudes in the particular group. Effective surveys of a population need to be based on a sample that is drawn on strict John Bynner probability lines, that is to say, each sample member has a known probability of being Associated Concepts: census, cluster sam- selected. Designs can range from simple pling, cohort study, falsification, hypothesis, random samples of a population identified longitudinal study, panel study, probability from a sampling frame, for example, Social (random) sampling, quota sampling

53 C-Jupp-3373.qxd 3/8/2006 5:40 PM Page 54

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

Key Readings culture. Cultural research is disparate in that its subject matters are extremely diverse, rang- Hoinville, G. and Jowell, R. (1978) Survey ing from the historical relevance of the works Research Practice. London: Heinemann. of Beethoven to the everyday consumption Kish, L. (1995) Survey Sampling. New York: of soap operas to the impact of museums on Wiley. Moser, C. A. and Kalton, G. (1971) Survey social exclusion. Cultural research can span Methods in Social Investigations. London: from the almost entirely abstract discussions Heinemann. as to the impact of globalization to highly Oppenheim, A. N. (1992) Questionnaire Design, focused policy-oriented analyses of art gallery Interviewing and Attitude Measurement. provision in a local community. New York: Basic Books. Cultural researchers deploy many of the de Vaus, D. (2002) Surveys in Social Research, skills associated with social research more 5th edn. London: Routledge. generally, from ethnographic methods to large-scale quantitative research. But they may equally use the skills associated with, say, the art historian or the expert in heritage CULTURAL RESEARCH or museum studies. At the policy end of the spectrum cultural research has become Definition increasingly interested in the social agenda and in particular questions of social inclusion A cross-disciplinary endeavour that applies a and urban regeneration. Such approaches range of methods concerned with researching have tended to favour a more quantitative the ways in which social groups express their approach. This may involve large-scale ques- place in the world and hence the ‘meaning’ of tionnaires on cultural attendance and atti- culture in all its diverse forms. tudes or the collection of secondary data illustrating the impact of cultural investment Distinctive Features on employment. At the other end of the spectrum many cultural researchers are con- The impact and meaning of culture has long cerned with how texts produce meaning, been an issue that has fascinated, and to a through, for example, advertising, and have degree, mystified, researchers from a wide therefore deployed a semiotic analysis. range of disciplines including Sociology, Cultural research should not be assumed Cultural Studies, Politics, Economics, Human to be a by-product or off-shoot of Cultural Geography and the Humanities. Cultural Studies. Indeed, many cultural researchers research is concerned with the properties asso- would disassociate themselves from what ciated with a concept, namely culture, that is they wrongly or rightly perceive to be constantly being revised by courtesy of the the lack of a method or methodology in fact it is itself the product of on-going social Cultural Studies (Gray, 2003). In this sense, change. But at the heart of what cultural cultural research has no discernible discipli- research seeks to find out is the way in which nary identity. It does not subscribe to a spe- societies or groups see the world and their part cific methodological dogma or disciplinary in that world and, perhaps most pertinently, agenda, but is rather a collection of diverse how such groups express their place in that approaches to a range of broadly related world. In this sense, culture is the ‘production phenomena. and circulation of meaning’ and cultural research is the study of how that meaning in Evaluation all its diverse forms is constructed. A key issue lying at the heart of the pro- The distinctiveness of cultural research as an duction of meaning is undoubtedly the rela- approach in its own right continues to be tionship between high culture and popular undermined by its apparently disparate

54 C-Jupp-3373.qxd 3/8/2006 5:40 PM Page 55

CULTURAL RESEARCH

nature. Despite the richness of a multi- because, by its very nature, culture is a disciplinary perspective, cultural researchers multi-faceted phenomenon that demands a remain in a constant battle to establish the more triangulated approach. worthiness of an area of research that does not In many respects cultural research finds necessarily fit very neatly into traditional aca- itself in a difficult position – unaccepted by demic disciplinary boundaries. This lack of a the academic mainstream both for its sup- disciplinary identity exacerbates a situation posed deficiencies in methodological rigour in which cultural research is not apparently and for an alleged lack of a genuine academic collected in a coherent or systematic fashion. tradition. Nonetheless, cultural research can Meanwhile, public policy makes particular be said to be making genuine in-roads on demands of its research to the extent that there both counts. As long as cultural researchers is a danger that the outcome of such research can convince policy makers and funders alike will be trimmed to suit those demands. There of the validity of their work, as well as the is a particular concern to establish the eco- benefit of a more sophisticated and long- nomic value of the cultural or arts sector, and sighted approach, then there is no reason to construct dispassionate economic impact why such progress cannot continue apace. studies, but as authors such as Johnson and Thomas (2001: 215) have suggested, ‘there is also a need to pursue an altogether wider Steven Miles research agenda on the economic effects of the arts. This agenda would focus more on the Associated Concepts: media analysis, measurement and valuation of the impact of methodological pluralism, policy-related the activities of the arts sector on the enjoy- research, qualitative research, quantitative ment, appreciation and human capital of research, visual methods participants.’ At a policy level cultural researchers often find themselves in the position of having to Key Readings ‘prove the case for culture’ through quantita- Alasuutari, P. (1995) Researching Culture: tive data that can make an immediate policy Qualitative Method and Cultural Studies. impact. Cultural researchers are therefore in London: Sage. danger of finding that their own creativity is Gray, A. (2003) Research Practice for Cultural increasingly hamstrung by a policy-oriented Studies. London: Sage. agenda. This tendency to research the wrong Johnson, P. and Thomas, B. (2001) ‘Assessing things for the wrong reasons may mean cul- the economic impact of the Arts’, in tural research will continue in its struggle to S. Selwood (ed.), The UK Cultural Sector: establish a coherent identity. An over-reliance Profile and Policy Issues. London: Policy on quantitative data is a particular concern Studies Institute. pp. 202–16.

55 D-Jupp-3373.qxd 3/8/2006 5:40 PM Page 56

out fieldwork on illegal activity can result DANGEROUS FIELDWORK in the researcher being suspected of being an informant or agent of regulatory or law Definition enforcement organizations. Indeed, few field- Forms of data collection in which there is a workers working with criminals have not at risk to the physical or mental well-being of some time been accused of ‘working with the the researcher. other side’, and this clearly places the fieldworker at risk from coercion or retribution. Distinctive Features Researching law enforcement itself can also be dangerous, as the researcher may be Fieldwork can be dangerous in a number of exploring activity that is corrupt, illegal or ways, and danger may be an accidental or morally ambiguous. The danger here is simi- deliberate aspect of life in the field. When lar to that posed by other deviant groups who researchers enter the field they enter places, need to reduce the possibility of negative spaces and cultures that are not their own, consequences, either deliberate or otherwise and this makes the fieldworker vulnerable emerging from the data. In transgressive due to unfamiliar hazards that can emerge fields the researcher needs to build up empa- naturally. The field itself can throw up dan- thy and trust with individuals in the field. gerous physical obstacles and hazards, partic- Whether based upon overt or covert tactics, ularly if the setting is especially removed acceptable language, appearance and a range from the kind of comforts and bourgeois sen- of indigenously based cultural competencies sibilities that most academics take for need to be acquired in order to reduce danger granted. For similar reasons the fieldworker to manageable levels. runs the risk of behaving inappropriately, In addition, the object of certain types of or of using language that is foreign and gen- fieldwork is more dangerous than others. For erally offensive to indigenous sensibilities. instance, researching essentially dangerous This latter category can render the field- jobs, pastimes, sports or hobbies places the worker vulnerable to anger or attempted fieldworker at a similar level of risk to that retribution by populations who are subjected posed to indigenous participants. However, if to the fieldworker’s gaze, or those on the the fieldworker is new to the endeavour, or is periphery of that gaze. incompetent or less than expert in an activity Researching transgressive fields can also that may be either central or peripheral to the be dangerous, for as part of immersion in the investigation, danger is increased. In such set- field the fieldworker may become implicated tings danger is an integral part of the research in illegal behaviour, or may even be required and the fieldworker, who may or may not pos- to participate in crime, and so becoming vul- sess the necessary competencies, will in effect nerable to police action. Similarly, carrying be researching danger. D-Jupp-3373.qxd 3/8/2006 5:40 PM Page 57

DATA

However, the cultural and physical Hobbs, D. (2001) ‘Ethnography and the contexts within which certain fields reside study of deviance’, in P. Atkinson, can also be dangerous, even if the object of A. Coffey, S. Delamont, J. Lofland and the research is not in itself hazardous. For L. Lofland (eds), Handbook of Ethnography. instance, carrying out a study of an urban London: Sage. pp. 204–19. school, nursery or social work department Lee-Treweek, G. and Linkogle, S. (eds) may produce few, if any, obviously dangerous (2000) Research Methods, Danger and occurrences that are integral to the targeted Ethics. London, Routledge. setting. However, the fieldworker, in order to contextualize these activities, will face the ethnographic realities of living in the area, travelling to work, shopping and drinking in DATA local pubs, and generally experiencing day-today pressures. In this respect even the most Definition competent fieldworker is likely to face those hazards and dangers that are everyday reali- Data are observations about the social world. ties for inner city residents. Data, the plural of datum, can be quantitative or qualitative in nature. The initial view of Evaluation the concept of data tends to be numerical, but qualitative data, for example descriptions of While the dangers of fieldwork should never interactions, are also common in social sci- be ignored, it is important to remember that ence. Marsh (1988: 42) points out that ‘data … the fieldworker has volunteered for the task, is [sic] produced, not given’; that is, and will eventually leave the field, unlike researchers choose what to call data, it is not the subjects of the research who experience just ‘there’ to be ‘found’. the daily rigours of the field as a permanent lived reality. Consequently, while a consider- Distinctive Features ation of personal safety in the field should always be given a high priority, the reflexive When data take the form of variables, they researcher will acknowledge both the tempo- can be discrete or continuous. Discrete data rary nature of being in harm’s way, and describe single properties, they are not con- attempt via the experience of fieldwork to tinuous in nature; legal offences (larceny, extract some notion of life in its permanent assault, homicide), types of household shadow. (nuclear family, extended family, friends or individual) and environmental hazard types (acid rain, oil spills, agricultural chemicals) Dick Hobbs are all examples of discrete data. A datum from one of these discrete groups will have Associated Concepts: access, confidential- one of these properties. Continuous data are ity, covert research, fieldwork, exploitative composed of properties that can be measured, research, impression management, partici- such as infant body weight, time to complete pant observation, research bargain anagrams, or distance covered by racing snails during ten minutes. At least theoretically, these continuous data can be measured to the Key Readings level of minute units, for example milligrams Coffey, A. (1999) The Ethnographic Self. for infants, milliseconds for anagram solving, London: Sage. and millimetres for racing snails. Ferrell, J. and Haamm, M. (1998) In addition to ‘variable’, a further impor- Ethnography at the Edge. Boston, MA: tant concept is that of attribute. Attributes Northeastern University Press. are ‘word categories’, and they generate

57 D-Jupp-3373.qxd 3/8/2006 5:40 PM Page 58

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

non-numerical data. The category of legal of data collected will be dictated by the kind offence can have the labels such as those given of analysis the researcher is able to undertake, above. They cannot have numbers meaning- whether with the help of others, or alone fully given to them, as they are not measures (Robson, 2002). One of the most unfortunate with numerical values. However, once a num- outcomes for inexperienced researchers is the ber has been assigned to a category it can be collection of data that cannot be analysed. very tempting to subject the data set to inferential statistical analysis. This temptation should be resisted, since the results of such Jeanette Garwood analysis are generally meaningless. For example, on a Likert scale question on fear of crime, Associated Concepts: chi-square test, data analysed by gender of respondent, women archives, diary, indicator, Internet research, could produce a mean value of 3.48, and men measurement, qualitative research, quantita- a score of 3.12, and this difference could be tive research, trace measures, unobtrusive statistically significant, but meaningless. It measures, variable analysis would be possible to investigate the relationship of gender to fear of crime, however, by Key Readings using a chi-square analysis, where each cate- Marsh, C. (1988) Exploring Data: An gory forms a column, and male and female Introduction to Data Analysis for Social form the rows, with the number of partici- Scientists. Cambridge: Polity Press. pants responding within each category each Robson, C. (2002) Real World Research, 2nd contributing a datum. edn. Oxford: Blackwell.

Evaluation Data can also be qualitative in a far wider sense, for example interview transcriptions, DATA ARCHIVES letters to newspapers, minutes of public meetings, television advertisements, televi- Definition sion interviews, videos of children playing, or in school classes, street art, graffiti, CCTV Data archives are resource centres that footage, and Internet sites. acquire, store and disseminate digital data for Data can be raw, and unprocessed, such as secondary analysis for both research and an interview transcription or survey scripts, or teaching. Their prime function is to ensure they can be processed, such as summary data long-term preservation and future usability of from questionnaires or the average of reaction the data they hold. times for different experimental conditions in an experiment. Data can also be things that Distinctive Features others have not noticed as important, or pre- designed into a study, as in the case of trace Data archiving is a method of conserving and unobtrusive methods. For example, data expensive resources and ensuring that their can be the number of tram or bus tickets sold research potential is fully exploited. Unless on a particular route, before and after a crime preserved and documented for further reduction intervention, or the wear of floor research, data that have often been collected at coverings at an exhibition as an indicator of significant expense, with substantial expertise the relative popularity of exhibits. and involving respondents’ contributions may The general goal in collecting data is the later exist only in a small number of reports idea of being able to analyse them in some that analyse only a fraction of the research way after their collection. Therefore the kind potential of the data. Within a very short space

58 D-Jupp-3373.qxd 3/8/2006 5:40 PM Page 59

DATA ARCHIVES

of time the data files are likely to be lost or by funding the Qualitative Data Archive become obsolete as technology evolves. (Qualidata) at the University of Essex, UK, in The data archiving movement began in the 1994. In 1995 a history data service was estab- 1960s within a number of key social science lished, devoted to the archiving and dissemi- departments in the United States which stored nation of a broad range of historical data. original data of survey interviews. The move- More recently, in 2003, the UK Economic and ment spread across Europe and in 1967 a UK Social Data Service (ESDS) was established to data archive (UKDA) was established by the provide a long-term and integrated strategy for UK Social Science Research Council. In the preserving, processing and disseminating all late 1970s many national archives joined types of social science data with an enhanced wider professional organizations such as the focus on supporting and training users. International Association of Social Science Typically, social science data archives Information Service and Technology (IAS- acquire a significant range of data relating to SIST), established to promote networks of society, both historical and contemporary, data services for the social sciences and to from sources including surveys, censuses, foster cooperation on key archival strategies, registers and aggregate statistics. procedures and technologies. Data archives typically collect numeric The first data archives collected data of spe- data, which can then be analysed with the cific interest to quantitative researchers in the use of statistical software. Numeric data social sciences. In the 1960s these were largely result when textual information (such as opinion poll or election data, but as the trend answers to survey questions) has been coded for large-scale surveys grew, by the late 1970s or they may represent individual or aggre- the UKDA began to acquire major government gated quantities, for instance earned income. surveys and censuses. Because of their large The demand for access to digital texts, images sample sizes and the richness of the informa- and audiovisual material has meant that tion collected, these surveys represent major material, from in-depth interviews and field research resources. Examples of major British notes to recorded interviews, as well as the government series include the General open-ended survey questions, is now avail- Household Survey and the Labour Force able for computer analysis. Survey. In the United States, key government Finally, data sets from across the world survey series include the Survey of Income are available through national data archives’ and Program Participation, the Current reciprocal arrangements with other national Population Survey and the National Health data archives. Interview Survey. A key concern for a data archive is to By the 1990s the UKDA collection had ensure that the materials they acquire are suit- grown to thousands of data sets spanning a able for informed use and meet demand. All wide range of data sources. Well-known UK- materials deposited are selected, evaluated based academic large-scale survey series and must meet certain criteria, such as include the British Social Attitudes Survey being documented to a minimum standard. and the British Household Panel Survey. Acquisitions policies should be flexible and In the United States, the General Social responsive to changes in both data and infor- Survey, the Panel Study of Income Dynamics mation needs of the research communities and and the National Longitudinal Survey of 1972 in the rapid changing climate of technology. are examples of long-running survey series. It is the responsibility of data archives to Major cross-national series include the World keep up with technological advances, by Values and European Values Surveys and the monitoring hardware and software develop- International Social Survey Program Series. ments and migrating their collection accord- In the 1990s the research community ingly. When technology changes, the data in recognized the needs of qualitative researchers their holdings are technically transformed to

59 D-Jupp-3373.qxd 3/8/2006 5:40 PM Page 60

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

remain readable in the new environment. and sign an agreement to the effect that they Computer programs are maintained to allow will not attempt to identify individuals when data to be easily transformed from an in- carrying out analyses. house standard to the various formats required by users. Data processing activities include first Louise Corti checking and validating, for example examining numeric data values and checking that Associated Concepts: census, confidential- the risk of identifying individuals is minimal. ity, official statistics, qualitative data archives, Second, metadata (data about data) are pro- qualitative research, secondary analysis, SPSS duced with the aim of producing high-quality finding aids and providing good user docu- Key Readings mentation. Metadata cover information ESDS website (December 2003) (Online): describing the study and the data, and for www.esds.ac.uk/ numeric data, files creating variable and IASSIST website (December 2003) (Online): value labels. A systematic catalogue record is http://datalib.library.ualberta.ca/iassist/ created for all studies, detailing an overview ICPSR (December 2003) Guide to Social of the study, the size and content of the data Science Data Preparation and Archiving, set, its availability and terms and conditions ICPSR website (Online): www.icpsr.umich. of access. User guides contain further infor- edu/access/dpm.html mation on how the data were collected, the Ryssevik, J. and Musgrave, S. (2001) ‘The original questionnaires and how to use the social science dream machine: resource data. discovery, analysis, and delivery on the Web’, Social Science Computing Review, Evaluation 19 (2): 163–74. Data supplied by data archives can be used for reporting of statistics, such as basic frequency counts, and for more in-depth sec- DECISION-MAKING RESEARCH ondary analysis. Secondary analysis strengthens scientific inquiry, avoids duplication of data collection and opens up methods of data Definition collection and measurement. Re-using A process whereby the context, background archived data enables new users to: ask new and dynamics of political decisions are questions of old data; undertake comparative analysed. Practically it seeks to understand research, replication or restudy; inform how and why a particular legislative proposal research design and promote methodological or government policy has come to fruition. advancement. Finally, data can provide significant resources for training in research and Distinctive Features substantive learning. Users typically request data in a particular Key characteristics of this type of research format, such as a statistical or word process- include its varied nature in addition to the ing package. These days, data can be accessed various possible strategies that are afforded via instant Web download facilities or can be to the researcher. One could, for example, dispatched on portable media such as CD- decide to conduct elite interviews, analyse ROM. The twenty-first century has seen a official documents and public records, move towards sophisticated online analysis or study the political biographies of the rele- tools, where users can search and analyse vant decision makers (although the latter data via a Web browser, such as Nesstar. would involve waiting until the diaries of the Users are typically required to be registered decision maker are published, which may be

60 D-Jupp-3373.qxd 3/8/2006 5:40 PM Page 61

DECISION-MAKING RESEARCH

several decades after the event, or which may and the institutionalist writings of Paul indeed never happen). Pierson (historical institutionalism). These A clear example of this type of approach scholars argue that legislative decisions are can be seen in decision making in the strongly influenced by the main EU institu- European Union. If one were to research the tions, which include the Commission, the EU’s Common Fisheries Policy (CFP), it would Council of Ministers (and the Council), immediately become apparent that the decision- the Parliament and the European Court of making process is wide and diverse. Policy is Justice. As soon as these institutions were not solely initiated by the institutions of the created, they began to assume powers that EU. Rather, decision making takes place the individual member states had not envis- within and between a complex web of inter- aged. Consequently, they have gradually connected institutions and stakeholders at the assumed greater competence in legislative EU (supranational), national and sub-national proposals and policy initiation. In terms of levels. In order to overcome this complexity, fisheries policy, which is now under the researchers are often required to make certain exclusive competence of the EU, the decision- methodological choices, for instance selecting making powers of member states have now one level of analysis in preference to another. been largely emasculated, resulting in an Such choices have caused controversies inability on the part of individual states to set within the academic community. For inter- the policy agenda. governmental (sometimes termed state- Although not yet fully developed, one centric) scholars such as Andrew Moravcsik, could make the case that the decision-making the primary focus for decision making must process is highly influenced by sub-state be the nation-state. In many respects this type groupings in the form of pressure groups, or of research strongly mirrors classical realism, even by transnational alliances. The former although there are several marked differences may include European regions; the latter may between the two. The main thrust of the inter- include pressure groups. Because of the governmental argument, however, is that nature of the EU, many regional representa- political decisions taken to integrate within tives or pressure groups can simply bypass a the EU are solely the preserve of the member member state and lobby the Commission states. At times of history-making decisions, directly. The clearest example of this type of such as the signing of the Maastricht Treaty, research has been forwarded by Gary Marks, the principal decisions are taken by the mem- Liesbet Hooghe and Kermit Blank with their ber states. Accordingly, controversies sur- writings on multi-level governance. Decision- rounding the CFP could ultimately be making is but one of a number of processes resolved by individual EU members. Should it in Europeanization, which also includes be to the taste of a particular member state policy initiation, implementation and adjudi- government to withdraw from the policy, then cation. The decision-making process is said to it would be free to do so. When undertaking take place in a highly fragmented environ- decision-making research, therefore, the ment in which individual states, EU institu- researcher should primarily focus on the tions, pressure groups and sub-national levels bargains reached between individual states, of government all play an important role. particularly in the Council of Ministers. Decisions on fisheries policy could therefore Many scholars, however, refute the notion be described as occurring in an environment that individual countries drive the decision- that includes all four of the major institutions, making process. A contra stance to the inter- in which all 25 member states are repre- governmental view of the decision-making sented, as well as including the lobbying of process is that the institutions of the EU drive individual regions and various pressure the policy agenda. This line of argument is groups. As this arena is so eclectic, it tends to most strongly enunciated in the supranational be less elegant than intergovernmental or writings of Ernst Haas (neo-functionalism) state-centric research approaches.

61 D-Jupp-3373.qxd 3/8/2006 5:40 PM Page 62

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

Methodological choices need to be made state-centric v. multi-level governance’, by the researcher depending on the decision- Journal of Common Market Studies, 34 (3): making perspective that the researcher deems 341–78. to be most convincing. An intergovernmental Moravcsik, A. (1999) The Choice for Europe: researcher may interview elites in a particular Social Purpose and State Power from Messina national government; a supranational scholar to Maastricht. London: UCL Press. by contrast may conduct elite interviews in Pierson, P. (1996) ‘The path to European the Commission or Parliament. Furthermore, integration: a historical institutionalist a multi-level governance scholar may decide analysis’, Comparative Political Studies, 29 (2): 123–63. to conduct interviews in a combination of the two in addition to speaking to pressure groups and representatives of sub-national governments. At this moment, however, there is no clear consensus on how one can best study DECONSTRUCTION the decision-making process in both the EU and international context. Definition Evaluation A style of thought developed by the French philosopher Jacques Derrida. On the one Clearly all of the particular research strate- hand, Derrida argues against the philosophico- gies outlined above have merit. A problem metaphysical tradition for its presupposition of that they all face to a greater or lesser degree, pure transcendence. On the other hand, he however, is that of access, which is a problem also argues against the non-philosophical encountered by most scholars of contempo- (empiricist) tradition of the human and social rary life. It could be the case that a contem- sciences for failing to acknowledge the neces- porary political scientist or sociologist will sity of a momentary transcendence. Decon- simply be unable to fully understand the real struction, then, is the inscription of the motives underlying decisions that are made ‘beyond’ of the transcendental into the empir- by politicians. A historian, relieved of the ical, and is thus the inscription of absence into baggage of everyday controversies when the presence by arguing that subjects and objects relevant embargo on official documents has are a play of traces (present-absence). elapsed, may be better able to comprehend why particular decisions were made at a given point. Distinctive Features Deconstruction provides a complex response to Craig McLean some of the major twentieth-Century strands in philosophy and social theory, notably phe- Associated Concepts: access, document nomenology, psychoanalysis and structuralism. analysis, elite interviewing, historical analy- Philosophers and social theorists have tradi- sis, mixed-methods research, policy-related tionally assumed an epistemology that involves research, process mapping the presence of a transcendent individual subjectivity and a universal objectivity. Hence language becomes the medium or tool that Key Readings represents and defines both the inner and Haas, E. B. (1968) The Uniting of Europe: outer worlds. On this traditional view, lan- Political, Social and Economic Forces. guage is secondary to the things it defines, like Stanford, CA: Stanford University Press. ‘thoughts’ and ‘objects’. Marks, G., Hooghe, L. and Blank, K. (1996) By contrast, deconstruction ‘is’ the disrup- ‘European integration from the 1980s: tion of all metaphysical-ontological questions

62 D-Jupp-3373.qxd 3/8/2006 5:40 PM Page 63

DECONSTRUCTION

beginning ‘What is … ?’ The paradox of this by being the person who does the uttering. In opening definition is that, by deconstruction’s written utterances however, the source is standards, it would have to disrupt itself. But confirmed via the person’s signature. Thus, the event of this disruption would be that this the signature signifies a person’s presence paradox is the definition of definition itself. (and therefore the presence of their inten- On the one hand, it is because of deconstruc- tions) in their absence. tion that we can define; on the other hand, it Derrida argues, however, that the repetition is because of deconstruction that any ‘defini- (iterability) of the signature implies that it can tion’ is thus a violent reduction (simplification both be imitated and imitates itself. The signa- of complexity) of the very term it attempts to ture is neither entirely inside nor outside of define. Thus, we cannot begin with a ‘defini- the text that it is supposed to be authenticat- tion’ of deconstruction – because we have ing, but rather it is on the edge of the text. This already begun; we are already enmeshed in suggests that the ‘origin’ of the text (and there- complexity and difference. fore the author’s intentions) is never fully pre- Derrida’s work seems to condemn this sent within it. This non-presence, this absence move of the definition as the familiar ‘meta- of the origin and its intentions, is part of the physical’ move towards an unattainable structure of language. origin. He suggests that any starting point Derrida also suggests that both Austin and will be complex or multiple – undecidable, an Searle erase things from their analyses ‘originary trace’. The attempt to assign some of context – the ‘parasitic’, ‘fictional’, ‘non- simple origin, e.g. God, humanness, to this serious’, ‘abnormal’, ‘absence’ (mainly of complexity is itself a secondary move, an intentions) – the ‘infelicities’ of speech acts. invention that occurs in response to that com- The reason they do this is because they are plexity. For example, for Derrida the ‘person’ seeking ‘ideal’ contexts, but Derrida argues is not an origin, rather it arises out of the very that this ‘deprives such theory of precisely attempt to assign a simple origin to some pri- what it claims for itself: seriousness, scien- mordial complexity. tificity, truth, philosophical value etc.’ (Derrida, When Derrida refers to language he does 1988: 72). If context were to be determinable – so with respect to its ‘play’ – the ‘problem of as Austin and Searle want it to be – then it language’ (Derrida, 1974: 6) arises in not could only be fully known to itself. Only when seeing it as a game whose play of differences it is realized that the context is undecidable and involves a necessary and originary absence. can never be fully known, as Derrida suggests, But presence and absence of what? Derrida’s does it become possible to at least begin to work shows that the history of metaphysics understand how to understand it. has been a history of the privilege of presence For researchers wanting to adopt a decon- over absence (the presence of subjectivity structive stance the task would be to apply and objectivity – and all other privileged Derrida’s rather esoteric thought to practical terms in a binary opposition) – what Derrida issues of analysis. The issue of analysing is calls ‘logocentrism’. Derrida’s main task has inseparable from what it is that we think we are always been the deconstruction of the logo- analysing – a participant’s thoughts or inten- centrism intrinsic in metaphysics. tions, a social context, or a piece of text? To give a brief example, Derrida’s (1988) Deconstruction would always take the latter: analysis of Austin and Searle develops three we are first and foremost enmeshed in textual interconnected arguments related to the ‘sig- traces. A necessary and originary absence (of nature’, the ‘event’ and the ‘context’. Austin subjectivity and objectivity etc.) is the conse- used the notion of the utterance-origin to sug- quence of this beginning. gest that we can refer to the person who does So what would a deconstructive analysis the actual uttering, in both verbal and written look like? Deconstruction, among other things, utterances. For verbal utterances it is simply is the demonstration of certain paradoxes,

63 D-Jupp-3373.qxd 3/8/2006 5:40 PM Page 64

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

contradictions, or ‘aporias’ of language. For moral distinctions such as that between right Derrida, these contradictions are not simply and wrong, good and evil, will be blurred. a contingent aspect of language, but are neces- Jurgen Habermas raises a typical example sary for language to function at all. An exam- of this type of criticism. He claims that decon- ple here would be to take the binary struction ‘levels’ the crucial distinction opposition of necessary/contingent itself: if the between philosophy and literature. Derrida’s thought of contingency is central to the decon- (1998) response to this type of criticism would struction of metaphysical eternal and infinite be that he sees decisions and responsibilities necessity, then we can say that the thought as things that arise out of plurality and com- of contingency is a ‘necessary contingency’. plexity. Only in the context of this undecide- That is, if that which is necessary is contin- ability can any decision be taken. gency, then the necessary is never simply Similarly, previous attempts at incorporat- necessary; and likewise, because if it is contin- ing deconstruction into critical approaches in gency that is necessary (as it must be), then psychology typically bypass the more radically the contingency is never simply contingent. anti-foundationalist features of a Derridean Deconstruction is not the deconstruction of deconstruction, and the sense in which it pro- metaphysics as if metaphysics existed prior to vides a resource for showing us that identities its deconstruction; rather, metaphysics can and realities can be constituted in order to be only exist through this deconstruction. recognizable as not constituted. So we can surmise that to deconstruct is to inhabit with no fixed assumptions about the unchanging presence or absence of subjectiv- Alexa Hepburn ity and objectivity. So we analyse texts rather than the presumed subjectivity or objectivity Associated Concepts: constructionism, dis- of that which lies behind the text. This means course analysis, epistemology that, like Derrida, we need to work from inside the text – where we are now, these texts Key Readings in front of us, and we need to be sensitive to the production of features of context, culture Bennington, G. (2000) Interrupting Derrida. and history that re-create meaningful utter- London: Routledge. ances. But we cannot assume in advance that Derrida, J. (1974) Of Grammatology (trans. we know what these things ‘are’ – to do this is G. C. Spivak) Baltimore, MD: The Johns to completely overlook them. In addition, if Hopkins University Press. Derrida, J. (1978) Writing and Difference we want a textual analysis that escapes the clo- (trans. A. Bass). London: Routledge. sures of metaphysical language we need to Derrida, J. (1982) Margins of Philosophy think of language in terms of its function rather (trans. A. Bass). London: Prentice-Hall. than its mere presence. And we need a non- Derrida, J. (1988) Monolingualism of the Other, cognitive approach to semiotics and analysis – or the Prosthesis of Origin (trans. P. Mensah). grammatology replaces semiology. Stanford, CA: Stanford University Press. Hepburn, A. (1999) ‘Derrida and psychol- Evaluation ogy: deconstruction and its abuses in critical and discursive psychologies’, Theory Deconstruction is neither theory nor method and Psychology, 9 (5): 641–67. in the traditional sense of these words; rather it is the process of calling into question the stability of binary oppositions such as theory/ method, invoking the value of plurality or multiplicity. However, this often leads to the DEDUCTION criticism that we are left with no grounds upon which to make a decision – crucial See Hypothetico-Deductive Model

64 D-Jupp-3373.qxd 3/8/2006 5:40 PM Page 65

DEMOGRAPHY

equivalent to a ‘death’, for example, the DEMOGRAPHY dissolution of first marriages. Demographic methods for examining such Definition processes are referred to as survival analysis The study of population structure, processes techniques. One such technique is the propor- and change, with analyses of fertility, nup- tional hazards model. Cox’s semi-parametric tiality, mortality and migration featuring version of this technique builds upon the life prominently. The tools used in demographic table, concentrating on variations between analyses include simple but effective mathe- population sub-groups in the risk of some matical techniques, some specific to demog- event, as opposed to variation over time in the raphy, alongside more sophisticated, generic underlying hazard rate (risk) for the whole forms of statistical analysis (Rowland, 2003; population (Trussell et al., 1992). Hinde, 1998). In addition to more sophisticated event his- Demographic research uses data from tory analysis techniques (for example, for the censuses, surveys, vital registration systems analysis of repeated events) and other recently and population registers. While demographic developed statistical methods (for example, analyses of official statistics are often macro- multi-level models), demographers frequently level, micro-level demographic analyses are utilize standard statistical techniques (regres- equally significant, with retrospective sion, logistic regression, etc.), as well as rela- surveys and longitudinal studies being partic- tively simple measures (for example, sex ularly valuable sources of micro-level life ratios, parity progression ratios). They also history data. create measures and techniques for specific demographic purposes, for example the historical back-projection technique used by Wrigley Distinctive Features and Schofield (1989) and, more generally, tech- When examining phenomena such as fertility niques used for forward population projec- and mortality, rates are an important tool. tions and forecasts. Age is central to demographic research, hence the use of age-specific rates is often Evaluation appropriate, with measures such as the total period fertility rate and life expectancy being Good demographic research often involves summaries of such rates across a range of the application of sophisticated mathemat- ages. In demography, the term standardiza- ical or statistical techniques to high quality, tion usually refers to the process of adjusting appropriately structured survey data. data to take account of age structure, to facil- However, demographic processes have eco- itate meaningful comparisons between differ- nomic, social, cultural and psychological ent populations. dimensions, thus theoretical accounts of such The life table is a key demographic tool, processes may draw upon various disciplines. consisting of a set of values (rates, numbers of Additionally, many contemporary demogra- cases, etc.) for a succession of age categories. phers recognize the value of qualitative data The life table for a particular period (for (for example, in relation to processes involv- example, a year) contains the age-specific ing decision making). mortality rates for that period and further In countries such as the UK and the United columns of figures corresponding to a hypo- States, the availability of microdata from thetical cohort of people experiencing those recent Censuses has opened up new research rates (for example, the proportion of the horizons. However, the coverage and quality cohort surviving to each age, the life expec- of data available for population research tancies of cohort members who reach that are uneven; population data for many less- age, etc.). More generally, a life table can be developed countries remain limited, notwith- applied to any process that ends in something standing international endeavours like the

65 D-Jupp-3373.qxd 3/8/2006 5:40 PM Page 66

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

Demographic and Health Surveys, and historical disposition and level of intelligence. Key demographic research often relies on the variables with which to describe groups or availability of parish registers and tax-related societies include social indicators about edu- returns. cation, social class, ethnicity, gender and demographic structure. Sometimes descriptive data are derived Richard Lampard from primary sources, for example social surveys conducted by a researcher for his or Associated Concepts: causal model, census, her own purposes. Alternatively, data can be cohort study, forecasting, longitudinal study, derived from secondary sources and applied multivariate analysis, official statistics, prospec- and adjusted to the researcher’s aims. These tive study, retrospective study can include official statistics collected by government departments as an aid to decision Key Readings making. Data archives also provide important sources of secondary data. Hinde, A. (1998) Demographic Methods. In the main, analysis using descriptive London: Arnold. statistics is one-variable or univariate as Rowland, D. T. (2003) Demographic Methods and Concepts. Oxford: Oxford University opposed to multivariate. Multivariate analy- Press. sis looks at the relationship between several Trussell, J., Hankinson, R. and Tilton, J. variables simultaneously, often with a view to (eds) (1992) Demographic Applications of constructing or testing causal models. By con- Event History Analysis. Oxford: Clarendon trast, univariate analysis seeks to describe Press. individuals or social groups using one variable Wrigley, E. A. and Schofield, R. S. (1989) The at a time. This can be done by a frequency dis- Population History of England, 1541–1871: tribution which counts the number of people A Reconstruction. Cambridge: Cambridge in social class categories or the number who University Press. answer survey questions in a particular way. The frequency in each category can be converted to a percentage which provides a standardized basis for comparison between DESCRIPTIVE STATISTICS categories and also between individuals or social collectives. Frequency distributions Definition can be portrayed visually as bar charts or pie charts. Basic statistical methods, and measures Other descriptive statistics include mea- derived from them, which summarize large sures of central tendency or average (means, sets of data so that descriptive statements can medians and modes) and measures of disper- be made about individuals, social groups and sion (standard deviation and variance). The societies. Sometimes descriptive statistics are latter indicate how closely scores are bunched contrasted with inferential statistics. The around an average and, therefore, tell us about latter are estimates of features of a population the overall shape of a frequency distribution (say indicators of multiple deprivation) for any given variable. derived from a sample of that population. Evaluation Distinctive Features The use of descriptive statistics alone puts lim- Descriptive statistics are used to describe itations on any analysis. Such statistics facili- individuals, social groups and societies on tate the description of an individual, group or key variables. Variables relating to individu- society on a range of variables. They also allow als can include type of personality, attitudinal ‘mapping’, for example in examining the social

66 D-Jupp-3373.qxd 3/8/2006 5:40 PM Page 67

DEVIANT CASE ANALYSIS

and geographical distribution of crime rates in whatever rule or pattern of conversational a city. However, descriptive statistics do not by interaction is currently being proposed. themselves provide an analysis in terms of relationships between two or more variables, Distinctive Features as in a causal model or any form of multivariate analysis. In this sense they cannot cap- Conversation analysis (CA) involves listening ture the complexity of the social world. Never- to recordings of talk-in-interaction, along- theless, the examination of descriptive side transcriptions, looking for robust pat- statistics is a vital preliminary stage in such terns of social interaction. In particular, CA advanced analyses in terms, say, of deciding looks for rules of conversational sequencing, which variables to include in a causal model such as in ‘adjacency pairs’, invitation– or assessing whether certain statistical charac- refusal sequences, story-prefacing and so on. teristics can be assumed (for example, whether Deviant case analysis is a procedure for values are normally distributed). checking the validity and generality of what- There is a further critique which can be ever phenomena emerge from that inductive levelled at the use of descriptive statistics, one method. It bears comparison to hypothesis which applies to quantification and statistical testing and falsification, as general method- analyses in general. This relates to the funda- ological principles (cf. Mitchell, 1983). mental epistemological question of whether However, it has special features that make it it is possible to measure the dimensions of an essential ingredient in CA. individuals’ psyches or features of the social In CA, rules of conversational sequencing world. are not statistically observed regularities, nor are they causal connections between one event (or turn-type) and another. Rather, they Victor Jupp are normative patterns, in the sense that they are demonstrably used, and oriented to, by Associated Concepts: causal model, data, conversational participants. This means that epistemology, hunting (snooping and fishing), exceptions to a conversational rule or pattern inferential statistics, measurement, multivari- are not simple phenomena to spot, such as by ate analysis, qualitative research, quantitative using a content coding scheme of some kind. research, variable analysis For example, it is a normative matter that greetings will be reciprocated, questions Key Readings answered, invitations responded to, and so on. However, when such expectable second Black, T. R. (2001) Understanding Social turns are absent, we are not forced to alter the Science Research, 2nd edn. London: Sage. analysis. Rather, norms are typically observed Blaikie, N. (2003) Analysing Quantitative in the breach (Garfinkel, 1967), in that absences Data. London: Sage. or other non-normative responses are heard Diamond, I. and Jeffries, J. (2000) Beginning and responded to as such. Statistics: An Introduction for Social What makes a case ‘deviant’ is not merely Scientists. London: Sage. that it contains an oriented-to breach of some conversational norm. Rather, it is a case that seems to disconfirm the developing analysis, which will already recognize the normative DEVIANT CASE ANALYSIS status of the rule or pattern being identified. Yet the importance of deviant cases, beyond Definition their serious status as potential disconfirma- tions, is that they are often examples that do The examination, in conversation analysis turn out to demonstrate participants’ norma- (CA), of examples that do not seem to fit tive orientations.

67 D-Jupp-3373.qxd 3/8/2006 5:40 PM Page 68

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

Stephen Clayman and Douglas Maynard deviant case analysis are matters of research (1995) define three uses of deviant cases policy and practice. As such, deviant case in CA: analysis provides an essential corrective to the tendency to look for confirming exam- (1) deviant cases may turn out, upon analy- ples. Its special relevance in CA lies in the sis, to instantiate the same normative nature of its phenomena as normatively orientations as are proposed for the oriented-to, rather than frequent or causal, standard cases; patterns of social interaction. (2) a new analysis may be proposed that now takes full account of both the (apparently) deviant cases, and also the main set; Derek Edwards (3) a deviant case may turn out to exhibit a different kind of order in its own right, Associated Concepts: analytic induction, such that it properly belongs in a differ- causality, constructionism, conversation ent collection. analysis, discursive psychology, epistemology, The classic example of the second type, hypothesis, induction, interpretive reper- where a revised analysis is proposed, is an toires, transcription early CA study by Emanuel Schegloff (1968), on the ways in which telephone conversa- Key Readings tions routinely begin. Using a corpus of 500 Clayman, S. E. and Maynard, D. (1995) examples, Schegloff had proposed a ‘distribu- ‘Ethnomethodology and conversation tion rule for first utterances’, to the effect analysis’, in P. ten Have and G. Psathas that the person answering the call is, (eds), Situated Order: Studies in the Social expectably, first to speak. One case was Organization of Talk and Embodied Activities. found to deviate from the rule. This led to Washington, DC: University Press of a questioning and radical revision of what America. pp. 1–30. the rule proposed. Rather than proposing Garfinkel, H. (1967) Studies in Ethno- a rule in which call answerers make first methodology. Englewood Cliffs, NJ: Prentice turns, which might be considered first parts Hall. of a greeting exchange (for example, ‘hello’, Mitchell, J. C. (1983) ‘Case and situational answered by ‘hello’), a revised rule could be analysis’, Sociological Review, 50 (3): 273–88. proposed which accounted for all 500 cases. Psathas, G. (1995) Conversation Analysis: The This was the summons–answer sequence, in Study of Talk-in-Interaction. London and which the first utterance of ‘hello’ (or what- Beverly Hills, CA: Sage. ever) was now understood as the second part Schegloff, E. A. (1968) ‘Sequencing in conver- of a different kind of pair, the answer to the sational openings’, American Anthropologist, summons, which was the phone ringing. 70: 1075–95.

Evaluation

Deviant case analysis bears comparison to DIARY hypothesis testing, but the basic procedure remains inductive, what conversation ana- Definition lysts call ‘unmotivated looking’ (Psathas, 1995). There need be no hypothesis prior to, A document, generally written for personal nor as a criterion for, data collection. Nor do use rather than for publication, that records we have to endorse induction, or empiricism, events and ideas related to the particular as general philosophical theories of knowl- experiences of the author. Researchers may edge and perception. Rather, induction and sometimes commission the writing of diaries

68 D-Jupp-3373.qxd 3/8/2006 5:40 PM Page 69

DIARY

related to specific events and ideas as a way Some sociologists have solicited the of acquiring relevant data. production of diaries geared to their research interests. Respondents are asked to keep a Distinctive Features diary for a specified period and are given a greater or lesser amount of guidance on what The keeping of diaries is closely related to the events and ideas are to be included in the growth of literacy and to a cultural orientation diary. Such diaries differ from the more spon- towards self-inspection and reflexivity. In taneously produced diaries, but they are Britain it was not until after the Reformation more likely to contain relevant and usable that diary keeping became at all common. data. Researchers into time budgeting have Among the wealthier and more literate classes, often used this method. diaries (as against account books and other forms of memoranda) began to be produced Evaluation in large numbers from the second half of the seventeenth century. The best known – A diary cannot be used as a reliable source in largely because they were eventually social research unless the conditions of its published – are those of John Evelyn and production are considered. Even when they Samuel Pepys. Both wrote mainly about public have not been written for publication, diaries affairs, though Pepys does describe aspects of may be written with other people in mind: the his private life, and they seem to have been author invariably considers the possible read- cast as memoirs, political commentaries and ers and shapes his or her material accordingly. autobiographies. Because of their greater liter- Indeed, some diaries are produced with publi- acy and their greater concern for religious self- cation (and posterity) explicitly in mind. This is reflection, clergymen were prolific diarists, a particular issue where diaries have been and many clerical diaries have also survived solicited or commissioned by researchers. through publication. Prominent examples are Many diaries do not survive the deaths of those of Josselin, Woodforde, Kilvert, Fox and their authors or are not likely to be made avail- Wesley. The everyday diaries of ‘ordinary’ able to researchers. This problem has been people – often little more than appointments given a new dimension by the growth of elec- books – are common, but rarely survive in a tronic diaries as these are rarely archived and form suitable for research use. Many contem- so no running record is retained. It is porary diaries are electronic. They are kept on extremely difficult to get an adequate sample computer and may be stored in specialist proof naturally occurring diaries covering issues grams that combine diaries with address of interest to a social researcher. books, contact lists and e-mail facilities. For large numbers of people, the everyday paper diary is being replaced by a handheld elec- John Scott tronic device. Diaries are often assumed to be written Associated Concepts: content analysis, dis- spontaneously and contemporaneously with course analysis, document analysis, historical the events that they describe. Typically, the analysis, qualitative data archives, secondary diarist is assumed to write his or her diary at analysis, trace measures, unobtrusive measures the end of each day, so giving an insight into the immediate subjectivity of his or her experience. However, a diary account – like any Key Readings account – is the outcome of a continuous, Allport, G. W. (1942) The Use of Personal Docu- ongoing reflection. It is written at various ments in Psychological Science. New York: times, is edited and rewritten periodically, Social Science Research Council. and is likely to be compiled in the light of Fothergill, R. A. (1974) Private Chronicles. some larger goal or intention. Oxford: Oxford University Press.

69 D-Jupp-3373.qxd 3/8/2006 5:40 PM Page 70

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

Matthews, W. (1950/1967) British Diaries. (3) Participatory research is research with, Gloucester, MA: P. Smith. rather than on, people. Ponsonby, A. (1923) English Diaries. London: (4) The research process is itself seen as a Methuen. process of change, most often characterized as a process of empowerment. (5) The research can also be a process of change through influencing, for instance, professional policy and practices, by DISABILITY RESEARCH reflecting the opinions of service users. Definition Emancipatory research, in the area of disability at least, has its roots in the growth of The examination of structural, environmental the Disabled People’s Movement and the and attitudinal barriers to full participation development of a social model of disability. and citizenship experienced by people with The emancipatory paradigm takes the adop- impairments, and the resistance by disabled tion of a social model of disability as the basis people and their supporters to institutional for research production. It can be argued that discrimination. emancipatory research, unlike participatory research, is not a research methodology Distinctive Features as such, but rather part of the struggle of disabled people to control the decision-making Disability has generally been defined in an processes that shape their lives and to individualistic, medicalized way as an internal achieve full citizenship. Changing the social condition of the individual, and most research relations and conditions of research can be on disability reflects this orientation. Many seen as part of the wider struggle to remove disabled people, on the other hand, view dis- all forms of oppression and discrimination in ability in terms of social, physical and attitudi- the pursuit of an inclusive society. nal barriers which could be removed if only Emancipatory research goes further than the political will to do so were present. participatory research by aiming to change Disability research has developed in two the social relations of research production, related directions: participatory and emancipa- with disabled people taking complete control tory approaches. Participatory approaches have of the research process. Although certain developed from general qualitative method- features of participatory and emancipatory ology, often by non-disabled researchers who research may overlap, one common confusion wish to break down the traditional hierarchical is the equating of emancipatory research with researcher–researched relationship. This can the qualitative paradigm. There is no reason be seen, for instance, in the referral to people inherent within the nature of emancipatory who are the subjects of the research as ‘co- research why it should adopt a qualitative researchers’ or ‘research participants’. General methodology, as long as the research agenda is characteristics of participatory methodology generated by disabled people themselves. For include the following: instance, emancipatory research into the (1) The research aims to reflect, explore and housing stock and, in particular, accessibility disseminate the views, feelings and of housing for disabled people is likely to take experiences of research participants the form of a quantitative survey to produce from their viewpoint. statistics to influence housing policies. (2) Participation, however, goes further to Participatory and emancipatory are, there- include the involvement of research fore, two distinct, though by no means participants in the design, conduct incompatible, research paradigms. There are and evaluation of the actual research shared beliefs within the two paradigms, but process. the differences also need to be recognized.

70 D-Jupp-3373.qxd 3/8/2006 5:40 PM Page 71

DISASTER RESEARCH

In general, disability research continues Key Readings to be developed both in terms of the metho- Barnes, C. (2003) ‘What a difference a dology adopted, for example narrative decade makes: reflections on doing approaches, and in terms of the arenas of “emancipatory” disability research’, social life examined, for example childhood Disability and Society, 18 (1): 3–18. and sexuality. Barnes, C. and Mercer, G. (eds) (1997) Doing Disability Research. Leeds: The Disability Evaluation Press. Moore, M., Beazley, S. and Maelzer, J. (1998) Key questions need to be addressed in an eval- Researching Disability Issues. Milton uation of disability research. First, does the Keynes: Open University Press. research promote disabled people’s Oliver, M. (1992) ‘Changing the social rela- control over the decision-making processes tions of research production’, Disability, that shape their lives? The question can be Handicap and Society, 7 (2): 101–15. directed at the decision-making processes within the research, and the empowerment of the participants through their involvement in research. Consideration of the effectiveness of the research in relation to this question hinges DISASTER RESEARCH crucially on the constraints over control. Ultimately, the question can only be answered Definition by disabled people themselves. A second question is: does the research address the con- The study of the behaviour of individuals and cerns of disabled people themselves? This the actions of groups and large-scale organi- question can be addressed in relation to the zations as they attempt to deal with extreme, disabled research participants, but also to the life-threatening events. broader population of disabled people. In terms of emancipatory intentions, a key ques- Distinctive Features tion is: does the research support disabled people in their struggle against oppression The wide range of events studied, the diver- and the removal of barriers to equal opportu- sity of disciplines involved and the variety of nities and a full participatory democracy for different research methods employed are fea- all? This, of course, is the most difficult ques- tures of disaster research. Sudden-onset tion to answer. Whether research can deliver events causing widespread physical destruc- anything in terms of contributing to the trans- tion and social disruption, such as tornadoes, formative process required by the emancipa- hurricanes and floods, have been most fre- tory paradigm remains an open question quently studied. So-called man-made or tech- (Barnes, 2003). Questions of the connection nological disasters, such as nuclear reactor between research and social change underpin accidents, massive power outages and airline fundamental controversies in current debates disasters, have received increasing attention. in disability research. More recently chronic events involving mass deprivation, such as drought and mass star- vation as well as conflict situations such as John Swain and Sally French civil wars, ethnic cleansing and acts of ter- rorism, have become subjects for some Associated Concepts: anti-racism research, researchers. critical research, emancipatory research, ethno- Disaster research is a multi-disciplinary graphy, participatory action research, politics endeavour carried out independently by and social research, practitioner research, researchers trained in different disciplines. qualitative research, value-free research Disciplines represented include sociology,

71 D-Jupp-3373.qxd 3/8/2006 5:40 PM Page 72

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

geography, political science, anthropology, have become increasingly more common. psychology and social psychology, public Studies of the popular culture of disasters, a administration and management science. As topic of intermittent interest throughout the one might expect, both the topics studied and history of the field, have become more fre- the methods employed to study them are quent and have dealt with subjects such as diverse. The field is most closely identified disaster humour, images of disasters in films, with the case study method, focusing on indi- and disaster memorials. vidual and organizational actions just before, during and immediately after the onset of Evaluation disasters. Qualitative or field methods such as depth interviewing and, where opportuni- Several unresolved issues currently confront ties present themselves, non-participant researchers in this field, some long-standing observation, are commonly used in these but others of more recent origin. Most funda- studies. Survey research methods are used mental is the issue of the appropriate subject to study pre-disaster behaviour, such as matter and boundaries of the field itself. The household and organizational preparedness seemingly esoteric question ‘What is a disas- and responses to warnings and predictions, ter?’ is actually about what types of events as well as post-disaster adjustments and should be studied and for what purpose. There recovery. Content analysis has been is increasing enthusiasm for viewing disasters employed in studying organizational records not as ‘acts of God’ or ‘acts of nature’ but and news media coverage of disasters. More as centrally and primarily involving human recently it has been used in studies of the agency, that is, as the consequences of prior popular culture of disasters. Secondary data human decisions and actions; for including as on a variety of social indicators have been disasters harmful chronic conditions and used in statistical analyses of the long-term undesirable long-term trends; and for seeing consequences of disasters. Historiographic governments as not only part of the solution to methods have been employed to study a few problems of disasters but also as one of the early disasters using both archival and pri- causes of disasters. vately held documents. Disaster research Related to this are issues of the appro- using surveys with panel and longitudinal priate theoretical framework for disaster designs as well as controlled experiments research and, although still mostly latent and under laboratory conditions are rare, but unrecognized, of the proper balance between there are notable examples of both in the lit- researchers’ roles as detached, unbiased erature. In general, methods used in disaster observers and as experts whose special insights research reflect those extant in the disciplines create a moral obligation to actively partici- in which researchers have been trained. pate in the amelioration of the unwanted Although studies of how individuals and consequences of disasters. Most disaster organizations react in the short term to the research over the years has been atheoretical. actual onset of disasters remain common, Although some researchers recently have increasing attention in recent years has been attempted to develop the concept of sustain- devoted to studies of longer-term adjustments ability, borrowed from environmental studies, such as mitigation, the physical and social into an overarching theoretical framework, changes made during periods of normalcy one that might also facilitate more cross- with the intent of reducing losses in future disciplinary research, most disaster research disasters, particularly economic costs. Newer remains loosely grounded in the theoretical research foci include examination of vulnera- traditions of the disciplines noted above. bility and the vulnerable in disasters, especially These recent developments within the along lines of class, ethnicity and gender. Studies field have been overrun by current events. of the role of women in disasters in particular The terrorist attacks in the United States in

72 D-Jupp-3373.qxd 3/8/2006 5:40 PM Page 73

DISASTROUS RESEARCH

2001 have created new theoretical, method- Key Readings ological, and practical challenges. It is no Drabek, T. E. (1986) Human System longer possible to exclude acts of violence in Responses to Disaster: An Inventory of their various forms from the definition of dis- Sociological Findings. New York: Springer- aster. Disaster research is increasingly sub- Verlag. ject to centralized screening processes aimed Quarantelli, E. L. (eds) (1998) What Is a at reducing duplication of both qualitative Disaster? Perspectives on the Question. field studies and survey research instru- London and New York: Routledge. ments. These challenges also include the Perry, R. W. and Quarantelli, E. L. (eds) increasing scrutiny of all research by institu- (2004) What Is a Disaster? New Perspectives tional review boards concerned with the use on Old Questions. Philadelphia, PA: Xlibris. of human subjects. Stallings, R. A. (ed.) (2002) Methods of There remains a disparity between the Disaster Research. Philadelphia, PA: world’s disasters and the world of disaster Xlibris. research. While a majority of disasters, Tierney, K. J., Lindell, M. K. and Perry, R. W. including the severest of them, occur in (2001) Facing the Unexpected: Disaster developing countries, the vast bulk of disaster Preparedness and Response in the United studies have been conducted in a relatively States. Washington, DC: Joseph Henry small number of industrialized nations. This Press. disparity primarily reflects the current con- Wisner, B., Blaikie, P., Cannon, T. and Davis, centration of disaster researchers in a handful I. (2003) At Risk: Natural Hazards, People’s of developed countries. Vulnerability and Disasters, 2nd edn. Ethical questions such as ‘Disaster research London: Routledge. for whom?’ and, ‘Disaster research to what end?’ are raised by some of these recent developments but remain little discussed. Traditionally, disaster research either explicitly DISASTROUS RESEARCH or implicitly has been a vehicle for improving the quality of emergency services provided Definition by governments and nonprofit organizations. This ‘top-down’ philosophy is reflected in Research that is either not accomplished or is the proliferation of university-based degree low in validity. programmes in emergency management. Increasing attention to categories of people Distinctive Features who are especially vulnerable in disasters has been accompanied by the rise of a competing There may be several reasons for the non- philosophy. A ‘grass roots’ strategy is implied, completion of a research project. For example, one that would empower those especially vul- access to the research field or to the research nerable in disasters – women, the poor, indige- subjects may be denied by gatekeepers. Also, nous populations – enabling them to improve the nature of the research topic may be such their overall situation. that it becomes too dangerous to continue with fieldwork. Even where a project is completed its findings and conclusions may not enter the Robert A. Stallings public domain because of promises made to the research subjects via a research bargain, or Associated Concepts: case study method, because sponsors of research refuse to sanc- content analyses, ethics, fieldwork, GIS, lon- tion publication as a result of not being happy gitudinal study, politics and social research, with conclusions or because they see them as sensitive research, social indicators a threat to their interests.

73 D-Jupp-3373.qxd 3/8/2006 5:40 PM Page 74

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

Research can be low in validity because of Key Reading issues of measurement, explanation or gener- Jupp,V. R. (1996) Methods of Criminological alization. Measurement validity refers to the Research. London: Routledge. ch. 5. extent to which a researcher measures what he or she intends to. Where a researcher uses a faulty measure or one that fails to capture the full extent of the concept under consideration the findings will be low in validity. Even DISCOURSE ANALYSIS if the measures can be trusted it is possible that the researcher may fail to provide an Definition adequate explanation by not including all relevant variables or by falsely treating corre- Detailed exploration of political, personal, lations between variables as sufficient evi- media or academic ‘talk’ and ‘writing’ about dence of a causal relationship. This is the a subject, designed to reveal how knowledges issue of validity of explanation, or internal are organized, carried and reproduced in par- validity. Even where reasonably valid conclu- ticular ways and through particular institu- sions have been reached they will be specific tional practices. to the subjects and the contexts under consideration. However, researchers typically Distinctive Features aim to generalize and make wider claims. Discourse analysis is a generic term covering Where such wider claims are not justified the a heterogeneous number of theoretical research will be low in validity of generaliza- approaches and analytical constructs. It tion, or external validity. This may be derives, in the main, from linguistics, semi- because the research subjects are not typical otics, social psychology, cultural studies and or representative of the wider population post-structural social theory. It is primarily a about which researchers wish conclusions to qualitative method of ‘reading’ texts, conver- be made. sations and documents which explores the connections between language, communica- Evaluation tion, knowledge, power and social practices. The above issues underline the importance In short, it focuses upon the meaning and of decision making about research in order structure (whether overt or hidden) of acts of to anticipate, as far as possible, the potential communication in context. problems of carrying out and publishing Numerous types of discourse-based research and the various threats to the valid- research can be identified. For example, a ity of conclusions. No research is perfect in branch of linguistics might be concerned execution and validity is always relative. with providing systematic accounts of con- Decision making about research is about versational exchanges in particular settings, trade-offs between the strengths and weak- whilst some areas of psychology might nesses of different research strategies and explore the effect of discourse structure designs in order to maximize the likelihood of on recall and memory. However, the broad accomplishing research and maximizing the appeal of discourse analysis to social validity of its findings. researchers lies in its ability to reveal how institutions and individual subjects are formed, produced, given meaning, con- Victor Jupp structed and represented through particular configurations of knowledge. Potter and Associated Concepts: access, dangerous Wetherell (1994) identify three key concerns fieldwork, gatekeeper, politics and social which characterize the research practice of research, validity discourse analysis:

74 D-Jupp-3373.qxd 3/8/2006 5:40 PM Page 75

DISCOURSE ANALYSIS

(1) locating talk and texts as social practices; their explicit and implicit discourses can be (2) identifying processes of action, construc- achieved in a number of ways. For example, tion and variability; in deconstructing policy documents the (3) recognizing the rhetorical or argumenta- researcher might ask: tive organization of talk and texts. • What are the conditions out of which this Much of this derives from the work of text emerged? What are the social, cul- Foucault (1972) and his ‘archaeology’ of the tural and political conditions which made knowledges, ‘interpretive repertoires’ and this text possible? practices of medicine. He used the concept of • What traces of other texts (intertextuality) discourse to counter positivist and scientific are evident in the text? claims to the ‘truth’ and to identify the mech- • How consistent, contradictory or coherent anisms whereby some versions of ‘truth’ is the text? How are contradictions come to be accepted and internalized, whilst managed? other readings are marginalized, discredited • How are people, objects and thought or discarded. Power is realized not through categorized? Who and what are included/ overt coercion but by a steady acceptance by excluded? all that a coherent text underlies all apparent • Who and what are viewed as normal, contradictions and paradoxes. Discourses are natural and common sense? ‘practices which form the objects of which • Are there any gaps, silences or ‘absent they speak’ (Foucault, 1972: 49). Through presences’? interacting with text/talk we interpret experi- • What is presented as legitimate/illegitimate? ence through structures already available to • Who are assumed to be the primary readers us and in doing so lend those structures a of the text? What assumptions are being solidity and normality which is difficult to made about the audience? move outside of (Mills, 1997). In different • What are the likely social effects of the ways all of us are regularly addressed by dis- text? courses that position and place us. They • What alternative readings might be made remind us of who we are and what might be by different social groups? expected of us in different social situations. These questions force the researcher to look Neither do discourses operate in isolation. beyond the immediate message of a text to At one and the same time we may be posi- reveal how it produces and disseminates par- tioned as a customer, a citizen, a consumer, a ticular ways of knowing. Importantly this does member of the public, a taxpayer, a worker, a not simply end at a level of identifying multi- public service user and so on. Neither are our ple interpretations but in gaining insights into subject positions fixed and stable. In turn this how some discourses come to be taken as raises the questions of contradiction and con- more legitimate than others. In other words, testation. Whilst Foucault’s analysis tends discourses not only have tangible material to disempower the subject as a passive recip- effects, but can be contested and resisted by ient of discourse, some Marxist and feminist counter discourses and social practices. scholars have identified the negotiated, contingent and dynamic nature of discursive Evaluation activity. Discourses can also be refused, contested and critiqued (Mills, 1997). It is easier to trace the theoretical underpin- The revealing of how text and talk produce nings of discourse analysis than to identify identifiable subject positions is one of the and describe the formal processes of actually basic skills of the discourse analyst. Data carrying out such research. This is partly due may be collected from any number of sources to the often intuitive and reflexive nature from legal statutes and media reports to of the skills involved. How then do we know diaries and personal testimonies. Identifying that one ‘reading’ is to be preferred over

75 D-Jupp-3373.qxd 3/8/2006 5:40 PM Page 76

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

another? Discourse analysts would respond by arguing that no social research can ever be DISCRIMINANT FUNCTION fixed and objective. A single ‘true’ reading of ANALYSIS social realities, by whatever method, is an impossibility. Discourse analysis is then just Definition as valid as any other. Moreover, despite such concerns over the relativity of this knowledge, A form of multivariate analysis which uses a the process of producing a reading allows us number of independent variables, measured to discuss, compare and challenge multiple at interval level, to predict the category or readings and the processes through which group to which a person might belong. they were arrived at. The methodology of discourse analysis then emphasizes the impor- Distinctive Features tance of being reflexive and open to competing knowledge claims. The avoidance of formulaic The discriminant function is a multiple analyses and commitment to be open to the regression equation in which category or new insights that a particular text/talk might group membership is treated as the depen- provide remain vital. The importance of this dent variable. It is also a categorical variable method will remain in its linking of method- whereas the independent or predictor vari- ological challenges about how we acquire ables are measured on an interval scale. For knowledge to epistemological challenges example, if the research aim is to predict about the nature of knowledge itself. It will voting preference in terms of either continue to play an important role in encour- Republican or Democrat the dependent variable will be treated as having only two values aging critical readings not only of conven- = = tional research methods but also of (say, Republican 1, Democrat 0). A range established modes of understanding. of independent variables will then be introduced to discriminate between those voting Republican and those voting Democrat. The John Muncie analysis can also indicate the relative importance of different independent variables Associated Concepts: constructionism, in predicting category or group membership. conversation analysis, critical research, Discriminant function analysis (DFA) can be deconstruction, discursive psychology, episte- used where the dependent variable involves mology, interpretive repertoires, postmod- more than two categories. ernism, textual analysis Some researchers use DFA as an alternative to multiple regression to predict the values of a discriminating variable, especially Key Readings where there is interest in the explanatory power of a categorical variable such as Foucault, M. (1972) The Archaeology of gender. For example, assume that the aim is Knowledge. London: Tavistock. to predict respondents’ incomes from social Mills, S. (1997) Discourse. London: class, occupational status, level of educational Routledge. Phillips, N. and Hardy, C. (2002) Discourse qualifications and gender. The analysis can Analysis. London: Sage. proceed by using the discriminant function to Potter, J. and Wetherell, M. (1994) predict gender from social class, occupational ‘Analysing discourse’, in A. Bryman and status and level of educational qualifications. R. Burgess (eds), Analysing Qualitative The percentage explained by the combination Data. London: Routledge. of these variables will be given by R-squared. Wetherell, M., Taylor, S. and Yates, S. (eds) If income is subsequently introduced into the (2001) Discourse Theory and Practice. equation and there is a significant increase in London: Sage. R-squared this would indicate that income is

76 D-Jupp-3373.qxd 3/8/2006 5:40 PM Page 77

DISCURSIVE PSYCHOLOGY

an important variable and that men and psychology that uses cognitive theories and women differ in terms of income when the experimental methods. Most specifically, dis- other variables have been allowed for. cursive psychology is a research perspective that considers psychology as an object in and Evaluation for interaction. More generally, it is a distinct form of discourse analysis in its own right. The value of discriminant function analysis is that it can incorporate categorical variables as Distinctive Features dependent variables in multivariate analysis. The limitations relate to statistical assump- Discursive psychology (DP) focuses on how tions which must be met: for example, the psychology is used by people in everyday and discriminating variables must be on an inter- institutional settings as a part of descriptions, val scale and also it is assumed that groups constructions and orientations. It asks what being studied are drawn from populations psychological categories and constructions within which the discriminating variables are are used for, and studies the practices in normally distributed. which they are used. Rather than trying Logistic regression analysis is sometimes to explain actions as a consequence of inner used as an alternative to DFA because it entities and processes (cognitivism) it focuses requires fewer assumptions. It also predicts on their role in performing actions through dichotomous or categorical variables, but this talk and texts. While theory and method in is usually in the form of whether something cognitive psychology presume an out-there will occur, for example whether individuals reality that provides input to cognitive opera- with particular characteristics are likely to tions, discursive psychology focuses on the graduate or offend. way both ‘reality’ and ‘mind’ are constructed by people conceptually, in language, in the course of their execution of various practical Victor Jupp tasks (Edwards, 1997; Potter, 1996). Take the example of scripts. In cognitive Associated Concepts: causal model, general psychology scripts are mentally encoded tem- linear model, hypothesis, hunting (snooping plates that guide action in situationally appro- and fishing), log-linear analysis, prediction, priate ways. In DP the topic becomes SPSS how script formulations are used by people themselves in situated discourse to perform Key Readings particular actions (for example, to present a speaker as acting in a normative manner or, Byrne, D. (2002) Interpreting Quantitative conversely, as driven by specific psychologi- Data. London: Sage. cal dispositions). The orderliness (or not) of Sapsford, R. and Jupp, V. (1996) Data Collection and Analyses. London: Sage. what someone has done can be constructed in the interaction for the purposes of the interaction. Rather than scripts existing as cognitively represented entities driving interaction, then, in discursive terms they are DISCURSIVE PSYCHOLOGY inseparable from action, bound up with the practical and moral world of accountability Definition (Edwards, 1997). Consider the following example from a rela- A strand of discourse analysis developed out tionship counselling session, where ‘Connie’ is of the work of Billig (1996), Potter and characterizing the solidity of her marriage (in Wetherell (1987) and Edwards and Potter contrast to the disruption caused by her part- (1992) as an alternative to the type of ner’s recent affair):

77 D-Jupp-3373.qxd 3/8/2006 5:40 PM Page 78

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

1 Counsellor: Whe:n: (.) before you moved over terms (3a) to construct a version of the here how was marriage that presents it as unexceptional 2 the marriage. (3b). This description is delivered in response 3 (0.4) to a question in a counselling session (1a) and ↑ ↓ − 4 Connie: O h. (0.2) I (.) to me: all alo:ng, works to counter her husband’s account of (.)right their marriage having long-standing problems 5 up to now, (0.2) my marriage was rock solid. (1b). Overall, it helps construct her partner as 6 (0.8) Rock solid.= We had argu- to blame for the marriage problems and as ments like the one who may need to change in response 7everybody else had arguments, to counselling (2). (0.4) buthh (0.2) 8 to me there was no major problems. Evaluation (Edwards, 1997: 154) DP is a relatively new social science perspec- Note the way having arguments is here script tive. Its success has come from the provision formulated as characteristic of all marriages of a rigorous way of working with records of (and thereby discountable as an explanation interaction (rather than the output measures of the specific problems that led this couple to and variable relations that are the staple of counselling). more traditional social and cognitive psychol- The theoretical and analytic development ogy). It has been particularly successful in of DP enables it to do justice to the status explicating interaction in complex multi-party of practices as practices. That is, it focuses settings, whether everyday or institutional. on the way discourse is oriented to action, There is some controversy over whether DP situated and co-constructed in stretches of will complement or supplant more traditional interaction, and given sense through the cat- forms of psychology. There is live debate over egories and formulations of participants. its relation to, for example, perspectives such DP’s analytic approach is guided by three as psychoanalysis (Billig, 1999). Currently most broad themes: clarity results from pressing a thoroughgoing DP perspective and seeing where more tradi- (1) Discourse is situated. Discourse is (a) occa- tional psychological notions are required. Its sioned in the conversation analytic sense; success in providing rich analyses of classically that is, talk and texts are reflexively related individual, private notions such as emotion to their position in interaction (Hutchby suggests that it can be developed further. and Wooffitt, 1998) and (b) rhetorically ori- Questions arise as to the relation of DP to ented, often designed to counter actual or conversation analysis. Do the two perspectives potential alternative versions (Billig, 1991). complement one another, or is there important (2) Discourse is action-oriented. DP focuses conflict between their conceptualization of the on how discourse is involved in actions nature of cognition? There is no clear resolu- and practices as people do their jobs, live tion to this question at present (see Te Molder their relationships, relate as profession- and Potter, 2004). als and clients and so on. A recent theme in DP has been the develop- (3) Discourse is constructed. DP asks two ment of a discursive psychology of institutions. kinds of constructionist question: (a) how This is focused on the role of psychological is discourse itself constructed out of categories and orientations to do specific insti- metaphors, rhetorical devices, categories, tutional business. For example, how do a caller interpretative repertoires and so on? and and a child protection officer collaboratively (b) how is discourse used to construct ver- use psychological constructions in performing sions of the world in the service of partic- important tasks at the start of calls to a child ular activities? protection helpline (Potter and Hepburn, 2003)? Take the script example above. Connie’s description of her marriage uses particular Jonathan Potter

78 D-Jupp-3373.qxd 3/8/2006 5:40 PM Page 79

DOCUMENT ANALYSIS

Associated Concepts: constructionism, social functions, interpretations, effects and conversation analysis, discourse analysis, uses that may be associated with them. interpretive repertoires, naturalistic data, reflexivity Distinctive Features Key Readings A wide range of documents is available to the Billig, M. (1991) Ideologies and Beliefs. social researcher. Personal items such as London: Sage. diaries, letters, aide-mémoires, shopping lists Billig, M. (1996) Arguing and Thinking: A and photographs produced as part of everyday Rhetorical Approach to Social Psychology, 2nd existence may form a private documentary edn. Cambridge: Cambridge University record, evidence of the way lives are lived and Press. how the social world is engaged with by indi- Billig, M. (1999) Freudian Repression: viduals and social groups at different times Conversation Creating the Unconscious. and in different places. Life is usually Cambridge: Cambridge University Press. recorded by a birth certificate, accompanied Edwards, D. (1997) Discourse and Cognition. by a passport or identity card and may be con- London and Beverly Hills, CA: Sage. cluded by a death certificate. Although involv- Edwards, D. and Potter, J. (1992) Discursive ing intensely personal meanings, these Psychology. London: Sage. documents are produced by or on behalf of Hutchby, I. and Wooffitt, R. (1998) the state and like so much documentary mate- Conversation Analysis: Principles, Practices rial the information carried there is part of and Applications. Cambridge: Polity Press. public documentation. Potter, J. (1996) Representing Reality: Dis- course, Rhetoric and Social Construction. Places of work, commercial and public London: Sage. organizations and educational institutions Potter, J. and Hepburn, A. (2003) ‘I’m a bit are amongst the many economic, social and concerned – early actions and psychologi- cultural organizations that produce docu- cal constructions in a child protection ments that are of interest to research. These helpline’, Research on Language and Social may range, for example, from contracts of Interaction, 36: 197–240. employment to till receipts and records of Potter, J. and Wetherell, M. (1987) Discourse educational achievement and offer informa- and Social Psychology: Beyond Attitudes and tion about people’s lives, work and leisure Behaviour. London: Sage. and the social and economic relations they Te Molder, H. and Potter, J. (eds) (2004) Talk enter into. and Cognition: Discourse, Mind and Social The continuing development of technol- Interaction. Cambridge: Cambridge ogy and the recent expansion of telecommu- University Press. nications and the mass media have added to the amount and number of forms that documents take, from the hand production of written and visual documentation to the DOCUMENT ANALYSIS mechanical production of printed material and reproduction of mass media documents. Definition Although document analysis is usually of the written word other forms of communication The detailed examination of documents pro- must be accounted for (Prior, 2003: 5). duced across a wide range of social practices, Radio, cinema and television generate a pro- taking a variety of forms from the written word lific twenty-first century display of visual, to the visual image. The significance of the textual, oral and sonic document forms. As documents may be located in the historical with all documents, these have a variety of circumstances of production, in their circula- social functions, including information, leisure tion and reception of the item and also the and social control functions. A television

79 D-Jupp-3373.qxd 3/8/2006 5:40 PM Page 80

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

advertisement, for example, is a ubiquitous A content analysis of documents will tend example of a mass media document, loaded towards a systematic and enumerative with commercial and social information, approach in order to quantify the frequency styles, ideas, attitudes, values, persuasions of elements within documents (such as words and ideologies and circulated widely, or images) or the quantity of similar types of through multiple broadcasts and receptions. document. Typically, this kind of analysis is Internet websites represent commercial, gov- concerned with the manifest content of a ernmental, educational and other organiza- document and is usually associated with the tional interests and exist alongside personal positivist tradition of social research. Textual websites and individually managed ‘blogs’, analysis on the other hand is usually thought the frequently updated web journals, that of as being part of the qualitative and inter- make up the pages of virtual documentation. pretivist tradition. Here emphasis is less on So there is a wide range of documents out the amount and frequency of occurrences there waiting to be analysed, and the social and more on interpreting the meaning the researcher will pose questions about docu- document might have. Semiology, the study ment availability and reasons for retention of signs, identifies words and especially but will also recognize that not all social images as signs that offer complex meanings facts have been documented and that not all or significations beyond the surface of the documents are available to research. text. Linguistic analysis explores the use and meaning of words and phrases. The analysis Evaluation of a television advertisement as a social document might require both enumerative A series of essential criteria is applied to the and textual analysis to give an indication of analysis of documents. Questions about the its frequency and potency. document’s authenticity, credibility, repre- Further areas of inquiry investigate docu- sentativeness and meaning are the usual ment interpretation, recognizing that differ- starting points (Scott, 1990). Establishing the ent people will interpret or ‘decode’ authentic nature of a document, that it documents in various ways which may be dif- was produced by the author or authorizing ferent to the producer’s or ‘encoder’s’ inten- body ascribed to it, leads into wider issues of tions. Documents may have an effect, change the nature and circumstances of the docu- or reinforce a belief, attitude or form of ment’s production. An exploration of the behaviour, but in certain circumstances the encoding process, the procedure of selection recipient of the document message may and putting into place the words, images or choose to use the message for other purposes. other elements that make up the content of For example, the health warning appearing the document is important not least in a con- on a packet of cigarettes may provide a sideration of any meaning the document wealth of documentary information and may come to have (Hall, 2001). The credibil- make an unambiguous statement about the ity of the document as evidence hinges on nature of the product, but this message might the truth and accuracy of its reference and be interpreted by the user and acted upon how widely it represents the phenomena the in a variety of ways. This is often character- researcher is investigating. For example, ized as a ‘field of action’, an area of activity questions of authenticity regarding an centred on the document, where the docu- inscription on a public building or monu- ment message as intended by the producer ment commemorating a historical event may may become diffuse amongst a social group be few but questions of why, how and or may be interpreted and reinterpreted whose interest the document serves may be through discussion and use (Prior, 2003: 2). many. The cutting, pasting and reassembling of

80 D-Jupp-3373.qxd 3/8/2006 5:40 PM Page 81

DOCUMENT ANALYSIS

electronic documents through computer Key Readings technology is a good example of the latter. Hall, S. (2001) ‘Encoding/decoding’, in M. G. Durham and D. M. Kellner (eds), Media Chris Wharton and Cultural Studies Keyworks. Oxford: Blackwell. Associated Concepts: content analysis, Prior, L. (2003) Using Documents in Social discourse analysis, diary, idiographic, netno- Research. London: Sage. graphy, qualitative data archives, textual Scott, J. (1990) A Matter of Record. Cambridge: analysis, unobtrusive measures Polity Press.

81 E-Jupp-3373.qxd 3/8/2006 5:42 PM Page 82

Evaluation ECOLOGICAL FALLACY The dangers of drawing conclusions about Definition individuals based on ecological correlations were highlighted by a number of studies dur- The failure to distinguish between a correla- ing the 1930s in the United States. These tion where the unit of analysis is indivisible have been considered in an article in the (for example, an individual person) and one American Sociological Review which demon- where the statistical object is a group or area. strated the mathematical relationship between ecological and individual correla- Distinctive Features tions (Robinson, 1950). Robinson explains that while individual correlation and ecologi- An ecological correlation exists where the cal correlation both depend on the within- researcher is interested in the prevalence of areas individual correlations, they do so in a particular phenomenon in a group or area. different ways: Ecological correlations are often used because data about individuals are not avail- The individual correlation depends upon the able, and mounting a study to collect such internal frequencies of the within-areas individ- data would be costly. However, the researcher ual correlations, while the ecological correlation should not then make inferences about the depends on the marginal frequencies of the behaviour of individuals based on the eco- within-areas individual correlations. … The logical data. For example, there may be a within-areas marginal frequencies which deter- positive correlation between levels of unem- mine the percentages from which the ecological ployment in police force areas and the crime correlation is computed do not fix the internal frequencies which determine the individual cor- rate in those areas. This is not, however, the relation. Thus, there need be no correspondence same as comparing unemployment and between the individual correlation and the eco- offending amongst a number of individuals. logical correlation. (Robinson, 1950: 354) To confuse the two types of correlation is an ecological fallacy. (For further comment on The size of ecological correlations is affected this example see Farrington et al., 1986). by the number of sub-areas. For example, the While a relationship may be found between size of an ecological correlation can depend the levels of unemployment and the levels of on whether the unit involved is a county, a crime in police force areas, it does not follow city, a district and so on. The size of the eco- that if individuals in those areas were sam- logical correlation between two variables for pled a correlation would necessarily be a county may be affected by the fact that the found between levels of unemployment and correlation for the county is based on corre- levels of offending for the individuals in the lations for local authorities within those samples. counties. E-Jupp-3373.qxd 3/8/2006 5:42 PM Page 83

ECONOMETRICS

The danger of confusing individual and Robinson, W. S. (1950) ‘Ecological correlation aggregate data is noted (again by Farrington, and the behaviour of individuals’, 1992) in discussing trends in juvenile delin- American Sociological Review, 15: 351–7. quency. It is pointed out that whereas a great Stack, S. and Gundlach, J. (1992) ‘The effect deal is known about correlates of juvenile of country music on suicide’, Social Forces, offending based on studies of individuals, it is 71 (1): 211–18. aggregate variables relating to neighbour- hoods, such as measures of social disorgani- zation, that are most useful in explaining variations in regional or national delinquency ECONOMETRICS rates. The two should not, however, be confused. Another example can be found in a Definition study in which two American researchers reported that the greater the airtime devoted The study of economic theory, and the appli- to country music, the greater the white sui- cation of economic principles using a mathe- cide rate (Stack and Gundlach, 1992). The matical or statistical approach. study was based on data for 49 metropolitan areas in the United States. Although further Distinctive Features analysis of the data showed that the association was independent of divorce rates, Traditional classic econometric analysis com- poverty and gun availability, because the prises a number of steps. data were based on metropolitan area rates, it does not necessarily follow that individual Step 1 Stating a hypothesis. The hypothesis people who listen to country music more can be based upon theory and/or empirical than others are more likely to commit suicide findings. It can be a result of intuition, ques- than those who listen to it less. Even were tions arising from prior reading or work-based such an individual level correlation to be findings. However, the aim of econometrics found, this does not mean that listening to is to address the hypothesis utilizing a quan- country music causes people to commit sui- titative analysis as opposed to a qualitative cide; consideration also needs to be given to approach. the possibility that the suicidally inclined find solace in listening to country music. Step 2 Stating any prior assumptions about the hypothesis in a mathematical form. Once we have stated the hypothesis we need to Iain Crow give it form. In econometrics this is based around the use of mathematical equations. Associated Concepts: causality, correlation, qualitative research, variable analysis Step 3 Specifying the nature of the econometric model. Specifying assumptions about our hypothesis, using equations, assumes that Key Readings we can exactly model the phenomena we are Farrington, D. P. (1992) ‘Trends in English studying. The complex interrelationships, juvenile delinquency and their explana- which exist in the most simplest of economic tion’, International Journal of Comparative phenomena, make this assumption highly and Applied Criminal Justice, 16 (2): 151–63. questionable. Therefore in order to accom- Farrington, D. P., Gallagher, B., Morley, L., modate this we derive an econometric model St Ledger, R. J. and West, D. J. (1986) which allows us to introduce these imper- ‘Unemployment, school leaving and fections into the study. This is done by adding crime’, British Journal of Criminology, an error term to the equations used in the 26 (4): 335–56. model.

83 E-Jupp-3373.qxd 3/8/2006 5:42 PM Page 84

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

Step 4 Data collection. Data collection in the evolution of complementary disciplines econometrics is vital for success. Success such as mathematics and statistics and the does not solely mean gaining results from our development of computer hardware and analysis which support the hypothesis. software. Success here should be read in broader terms, Whatever the type of approach chosen to that is, having confidence in the results, and perform an econometric analysis there are a in the validity and meaning of the conclu- number of issues that need to be considered. sions. Econometrics tends to be based upon First, there is the question of whether it is large amounts of data, although considerable possible to reduce economic phenomena to a breakthroughs in theory and principles have mathematical or statistical form which been found using small data sets. The general reflects reality. Second, there is the issue of rule is ‘bigger is better’. whether it is possible to adapt the mathematical form into an econometric model ready Step 5 Utilization of statistical/mathemati- for testing. Simply adding an error term to an cal procedures aimed at answering the equation may be viewed as a rather naive hypothesis. Once the data have been col- approach in attempting to construct an lected an analysis has to be undertaken. The econometric model. Third, there are general two most popular forms of analysis in research issues regarding the ability to econometrics are multiple regression and gather data, whether samples are representa- time series analysis. However, there are tive and whether results are generalizable. other techniques, and the researcher should Econometrics places a good deal of impor- strategically choose in advance which tech- tance on the ability of a derived model to niques are best suited to answer the hypoth- accurately forecast and predict what will esis, and the level of confidence required. In happen in the future given a certain set of the main, econometric models are multivari- circumstances. The validity of forecasts and ate models. predictions is very much dependent on whether these questions and issues can be Step 6 Utilization of findings. This is for the addressed successfully. purposes of forecasting, adding to knowledge or theory and formulating or evaluating policy. In this respect there can be differences Paul E. Pye between econometrics and other disciplines. Some disciplines are only concerned with Associated Concepts: causal model, fore- finding out whether there is a relationship casting, hypothesis, modelling, multivariate between the dependent variable and the analysis, prediction, quantitative research, independent variable(s), and the nature of regression analysis, SPSS, time series design that relationship. Econometrics takes this further by utilizing this information and predicting events, which may follow as a result Key Readings of it. Carter Hill, R., Griffiths, W.E. and George, G. (2001) Undergraduate Econometrics, 2nd Evaluation edn. New York: Wiley. Maddala, G. S. (2001) Introduction to This model of econometrics is known as the Econometrics, 3rd edn. New York: Wiley. classical model. Alternative models or Wojciech, W., Charemza, W. and Deadman, approaches do exist. Wojciech, Charemza and D. F. (1992) New Directions in Econometric Deadman (1992) highlight the shortcomings Practice: General to Specific Modelling, of the classical model and suggest a range of Cointegration and Vector Autoregression. alternative approaches which are in line with Aldershot: Edward Elgar.

84 E-Jupp-3373.qxd 3/8/2006 5:42 PM Page 85

ELITE INTERVIEWING

first-hand information about their personal ELITE INTERVIEWING backgrounds, outlooks and motivations. Secondly, interviews might also be under- Definition taken where the elites serve as experts. It is The use of interviews to study those at the often the case that such individuals have ‘top’ of any stratification system, be it in sport, unique experiences as ‘insiders’, enabling them academia, social status, religion, beauty or to comment upon events or evidence, provide whatever. In practice, however, elite research interpretations and suggest fruitful lines of focuses mainly on political and economic nota- further inquiry. Thirdly, elites may serve bles. The study of elites touches on some of as gatekeepers who, formally or informally, the major and perennial issues of social analy- control access to needed data sources. These sis. The views and activities of generals, busi- may be official records, internal reports, or nessmen, politicians and church leaders have even other individuals. Interviewing the per- been of concern to social thinkers since the son at the top may simply be aimed at gain- earliest days of Western thought. ing institutional rapport and ‘unlocking the door’. Elite interviewing requires careful prepa- Distinctive Features ration. First, the specific individuals must be The study of elites exhibits a considerable located and selected. This may be problem- diversity of research techniques, reflecting atic even in ‘open’ milieux. Furthermore, the many substantive purposes such study there is an array of varied selection proce- serves, as well as the wide range of evidence dures (for example, positional, reputational, elites typically generate. These include writ- decisional), each with its own well-known ten and printed materials, such as speeches, virtues and faults. Often, however, the elite diaries and autobiographies. There are also may be readily identifiable, and simply be behavioural data, such as voting records, sec- ‘important’ individuals worth studying. retarial diaries and judicial decisions. Then Gaining quality access to particular individ- there are the much-used personal and bio- uals once selected can also pose problems. In graphical statistics in standard publications. part it is, again, a matter of the relevant elite However, a further source is now used quite and legal milieu. It also depends on the image frequently – the personal interview. Of of the potential interviewer, the research pro- course, this is not always a viable option. ject or even of the host institution or social sci- Elites can pose considerable problems of ence in general. Often, one help can be some access for scholars even in relatively open link with the interviewer that the interviewee cultural milieux. values. Perceived ideological, social or even Where elite interviewing is possible, its sartorial or gender similarities may also have value will depend on the particular research the same effect. On the other hand, links and objectives. In some cases, interviewing would similarities can handicap the interviewer as a be laboriously inefficient, such as for ques- neutral ‘outsider’. tions of elite social composition, where pub- Access is also influenced by the research lished sources would normally be better. In agenda, some agendas being very sensitive in other cases, elite interviews may serve only in particular circumstances or to particular a secondary role, for example, to gain access to elites. This may lead to outright (and reveal- private documents. But certainly there are ing) refusals or to a reduced level of coop- times when interviewing can be appropriately erativeness that may not be immediately used as the principal means of inquiry. apparent. In general, it is a cost–benefit situ- There are three broad needs that face- ation where the interviewees especially to-face contact can help fulfil. Above all, weigh the pluses and the minuses of the through interviewing, elites may provide exchange. Such considerations bear on the

85 E-Jupp-3373.qxd 3/8/2006 5:42 PM Page 86

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

selection and preparation of the interviewer. find less structured formats are generally In some instances, professional interviewers preferable. Here, the researcher can define have been successfully employed. More typi- the agenda in broad terms, but the respon- cally, however, the researcher acts as inter- dents are allowed scope to respond to its con- viewer because of the substantive and technical stituent parts in their own particular terms. demands of the situation. Interviews also tend to be one-offs. The typi- This means that, even for the scholar– cal elite individual has a very demanding interviewer, there is a heavy premium upon schedule that rarely permits extended discus- personal prior preparation. He or she will be sion, although surprisingly many make them- assumed to be, within limits, an expert by the selves available. Only exceptionally, such as elite respondent. This entails familiarization for retired elites, is time not a pressing issue. with the values, terminology and significant The most important ingredient for the reference points likely to be used by respon- success of elite interviewing is trust and rap- dents. Without this capacity to demonstrate port. Usually this requires a stance of empa- familiarity, valuable data gathering opportu- thetic neutrality but, on occasion, a challenging nities may be lost and, not least, trust and style may be needed to get some respondents rapport may be compromised. In general, to open up. However, scholars (unlike media sound preparation will always be rewarded interviewers) have few sanctions at their and poor preparation revealed. disposal, and hence threatening stances can Elite interviews, as noted, potentially mean failure. In general, experience shows serve varied purposes. They may also take that elite interviews can be highly variable in varied forms. Some, for example, involve quality, ranging from the unproductive and lengthy in-depth encounters over a period of uninformative to the essential and highly several meetings while others are brief one- insightful. It is a matter of using interviewing offs. Equally, the format may differ according appropriately, preparing well and being to the degree of structure or standardiza- aware of the relevant problems. tion employed – the order in which topics are raised, the way specific questions are asked, and whether closed or open-ended responses George Moyser are allowed. Some are fully structured, while others have limited, or almost no, structure. Associated Concepts: access, decision-making research, elite selection, ethnographic interviewing, gatekeeper, interview, politics and Evaluation social research, research bargain The choice as to type of interview is a matter of weighing different considerations. Issues need to be raised, such as whether the end- Key Readings product of the study is to emphasize the qual- Dexter, L. A. (1970) Elite and Specialized itative and particular, or the quantitative and Interviewing. Evanston, IL: Northwestern general; technically, how much control the University Press. researcher is willing to cede to the intervie- Field, G. L. and Highley, J. (1980) Elitism. wee, and how much procrustean structure London: Routledge & Kegan Paul. the latter will tolerate; practically, whether Marcus, G. E. (ed.) (1983) Elites: Ethnographic the needed interviews will be very numer- Issues. Albuquerque, NM: University of ous. In general, considerations tend to mili- New Mexico Press. tate against the fully structured interview. Moyser, G. and Wagstaffe, M. (eds) (1987) It is precisely because elite individuals are Research Methods for Elite Studies. London: important repositories of expertise, with Allen & Unwin. unique experiences, and often-sophisticated Parry, G. (1969) Political Elites. London: intellectual capacities, that many scholars Allen & Unwin.

86 E-Jupp-3373.qxd 3/8/2006 5:42 PM Page 87

ELITE SELECTION

being a ‘power elite’ or a ‘ruling class’ for the ELITE SELECTION rest of society? To answer such questions, particular elite Definition individuals must first be identified and then A form of research which seeks to identify selected for further analysis. But as substan- and select elite individuals in social group- tive questions of influences and power are ings, communities, organizations and society involved, there has been considerable disas a whole in order to address a range of agreement about the appropriate selection questions, for example, about their social net- methods to be used. In the debate, three works, their influence and their power. generic methodological approaches have emerged, each reflecting different views about the nature of power and how it should Distinctive Features be located and observed. Theory and method, Social theorists have long recognized that in other words, are here closely interwoven. human society manifests a fundamental inequality in the way that it is organized. Selection by position A common view of Whether from differences of ability or from modern societies is that power is institution- such collective social imperatives as the need alized – it is a function of formal offices for authority, there is an observed tendency within certain social institutions. The power- for society to produce individuals who are espe- ful, then, are those individuals who occupy cially influential, dominant or noteworthy – key positions in government, party commit- ‘top people’ in their chosen sphere. Various tees, commercial and industrial corporations classical theorists, such as Pareto, Mosca and and public bureaucracies. On this basis, the Michels have argued that these individuals members of the elite are identified and constitute a single elite group with a certain selected by determining which formal offices collective self-consciousness, ideological are powerful. Such decisions can be difficult, coherence and commonality of action. As particularly at the margins and in relatively such, the elite would form a particularly ‘closed’ societies and sub-cultures (see dominant element in society. Much analysis, Moyser and Wagstaffe, 1987). But, in general, therefore, has been done to test these views information is often available that can make empirically. the process relatively straightforward. As a Today, theorists tend to recognize that result, positional analysis is commonly used there can be many elite groups, one for each as a selection method (see Mills, 1954, for sphere of social action, in education, sport, example). It is not, however, uncontroversial. the arts, even criminal activity. These have all Even in modern societies, power is not received some attention – social celebrities always solely a function of formal office, but and the glitterati as elites, for example. But, also of other less tangible factors as suggested following the classical tradition, most atten- by the phenomenon of the eminence grise, the tion is still reserved for those elites who have person of ‘real’ power behind the façade of the greatest potential influence over society the throne. As a result, individuals may be as a whole – top individuals in government, selected into the elite who should be the economy and military affairs (Putnam, excluded, and other individuals left out who 1976; Scott, 1990). How are they recruited, should be included. renewed and replaced? What are their views and social backgrounds? What influences do Selection by reputation To try to remedy they exert in shaping the goals of society, in such problems, selection by reputation has maintaining and integrating society? Above been employed, using a method originated by all, to what extent do these individuals form Hunter (1953). This approach treats power as a unified elite, tied together through social being potentially either formal or informal. For and ideological networks, and capable of Hunter, therefore, the elite is best uncovered

87 E-Jupp-3373.qxd 3/8/2006 5:42 PM Page 88

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

by the broader notion of reputation than by also can be used to gauge the reputation of institutional position. The method centrally the original positional elites or even, conceiv- involves a set of expert judges who select the ably, to de-select them. elite by their general reputation of power. There is no single method for elite selection The elites are then, in turn, asked to nomi- that is free of criticism. Each is associated with nate others for inclusion. One key potential certain substantive views of power, and each weakness of this method is the reliance it could arguably produce different results. In places on the views of the judges. This may this situation, it would perhaps seem most be ill-informed, confused or even deliberately appropriate to rely on a combination, or syn- biased about the structure of power. Some thetic approach. This might be the closest one researchers have indeed found that different can get to the idea of a neutral technique in panels of judges produce different lists of which the choice of a particular selection elite individuals. method keeps open what the findings about the elite in question will turn out to be. Selection by decision making A third selection method focuses on the analysis of George Moyser particular decisions, or events, on the view that the essence of power is the capacity to make the tough big decisions. As developed by Associated Concepts: access, critical research, Dahl (1961), its advantage is the recognition decision-making research, messy research, that an individual’s scope of power mixed-methods research, organizational may be circumcised or limited to a particular research, snowball sampling, triangulation arena. Hence, the possibility arises of many different elite groups rather than one that is Key Readings all-embracing. However, this method may Dahl, R. A. (1961) Who Governs? Yale: Yale tend systematically to underplay the latter pos- University Press. sibility, producing an overly fragmented view Mills, C. Wright (1954) The Power Elite. New of the elite by linking power too closely to York: Oxford University Press. involvement in particular disparate decisions. Moyser, G. and Wagstaffe, M. (1987) Research It also has to face the problem that one ‘face’ Methods For Elite Studies. London: Allen & of power is the capacity to prevent issues com- Unwin. ing up for decisions (as with judges), which Presthus, R. V. (1964) Men at the Top: A might produce a different list of elites. Study in Community Power. New York: Oxford University Press. Evaluation Putnam, R. D. (1976) The Comparative Study of Political Elites. Englewood Cliffs, NJ: In recognition that none of the above seems Prentice Hall. to be adequate in itself, researchers such as Scott, J. (1990) The Sociology of Elites. Presthus (1964), for example, have combined Cheltenham: Edward Elgar Publishing. selection methods employing two of them together. One approach is to start with positional analysis and augment it with reputational selection. This is known as the EMANCIPATORY RESEARCH ‘snowball’ technique. Individuals are initially selected by position. Then they become the Definition judges who nominate additional individuals by reputation. Those who receive multiple Research that seeks to empower the subjects nominations in turn offer names until the of social inquiry. It is now commonly recog- process is complete, or too few names are nized that power is a fundamental aspect added to justify continuation. The technique of all research relationships. Traditional

88 E-Jupp-3373.qxd 3/8/2006 5:42 PM Page 89

EMANCIPATORY RESEARCH

research processes have been criticized for their This leads some to argue that researchers objectification of respondents. One response should be aware of the power that they hold has been to argue for an emancipatory research and should make themselves vulnerable and try process: one which recognizes this power to equalize their relationships with respondents imbalance in research and aims to empower which would allow for empowerment through respondents through research. research (for example, Stanley and Wise, 1993). This may involve including respondents in the Distinctive Features construction of the research design, investment of the researchers’ self and life experience in Traditional research processes argued for the research encounters and involving respondents objective and value-free production of knowl- in the checking of data before the final presen- edge: a ‘scientific’ social science. From this tation of the research report. Although consid- perspective the research process is value-free, ered less often in methodological literature, coherent and orderly – in fact ‘hygenic’ issues of power and empowerment within (Stanley and Wise, 1993; Kelly et al., 1994). research teams are also important to consider if Amongst other criticisms of this approach is the aim is for an empowering research experi- the viewpoint that power is a fundamental ence for all (Kelly et al., 1994). aspect of all research relationships. It is useful here to draw on the work of Giddens Evaluation (1985), who argues that power is in one way or another an aspect of all relationships and Issues of power are complex within research. makes the distinction between two types of As suggested above, there is an assumption that resources involved in power – control of the researcher is always in control of the material resources (for example, money, time research situation and is the one who holds the research ‘tools’) and control of authoritative balance of power, but it is often more compli- resources (for example, as in holding the sta- cated than this in reality. It is important not to tus of researcher). Within research it is usu- over-passify research respondents, not least by ally the researchers who have the time, assuming that they are always vulnerable resources and skills to conduct methodologi- within research. Some respondents do not feel cal work, to make sense of experience and disempowered by either their life experience or locate individuals’ experiences in historical by the research relationship and it may be and social contexts. The researcher usually patronizing of the researcher to assume that has control over, for example, the construc- the respondent needs to be empowered by the tion of the questionnaire, the order in which process. Also, research relationships are fluid the questions are asked in a qualitative inter- and jointly constructed and at times during the view, the frequency and timing of visits to a research process it is the researcher who might research site, and thus holds the associated feel vulnerable and/or at a disadvantage. This status that this brings. Furthermore, it is the may be the case when researching individuals researcher who is more often than not who are older, more experienced, more knowl- responsible for the final analysis and presen- edgeable and/or when undertaking research tation of the data. Thus, researchers ‘take with people with sexist, racist, homophobic away the words’ of respondents and have the (and so on) views and attitudes (for example, power of editorship. So, it is important to Collins, 1998). In addition to the emotional acknowledge that researchers often have danger suggested here it is important to also the objective balance of power throughout the acknowledge that research can be physically research project and have control of both the dangerous for researchers. material and authoritative resources. With all Even when aiming for an emancipatory of this in mind, Stacey (1991: 144) argues that research process it is an illusion to think that, ‘elements of inequality, exploitation, and in anything short of a fully participatory even betrayal are endemic to [research]’. research project, respondents can have anything

89 E-Jupp-3373.qxd 3/8/2006 5:42 PM Page 90

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

approaching ‘equal’ knowledge (about what is Stacey, J. (1991) ‘Can there be a feminist going on) to the researcher. It may also be sim- ethnography?’, in S. B. Gluck and D. Patai plistic to assume that an approach which (eds), Women’s Words, Women’s Words, includes the respondents at all levels is ulti- Women’s Words: The Feminist Practice of mately empowering for respondents. ‘Participa- Oral History. New York: Routledge. tory’ research can entail disparate levels of Stanley, L. and Wise, S. (1993) Breaking Out input from research respondents at different Again: Feminist Ontology and Epistemology. stages of the research process and some may London: Routledge. not wish this type of involvement. Thus, there may be a tension between the desire to give respondents a voice and the making of knowledge, not least because individuals may not nec- EMPIRICISM essarily possess the knowledge (or have the desire) to explain everything about their lives. Definition Arguably, equalizing relationships within research could also be seen to be exploitative in Empiricism began as a philosophical doctrine that it could encourage isolated individuals to in the seventeenth century in the writings of come forward and reveal aspects of their expe- Locke, Berkeley and Hume. In the twentieth rience that they later regret. Furthermore, mak- century it had an enormous influence on sci- ing people feel powerful within the research entific method and was the basis of logical process does not necessarily change the emo- positivism (itself sometimes called logical tional and material circumstances of their lives. empiricism) (Gillies, 1993: 17–21).

Gayle Letherby Distinctive Features In empiricism knowledge is only validated Associated Concepts: anti-racism research, through sense experience, or in more recent critical research, dangerous fieldwork, dis- versions through the surrogates of scientific ability research, disaster research, exploita- instrumentation (which in the social sciences tive research, feminist research, participatory would include survey questionnaires and action research, politics and social research interview data). Its importance to scientific method in the natural and social sciences lies in the centrality of emphasis placed on Key Readings empirical hypothesis testing. Thus if we formulate a hypothesis such as ‘industrialization Collins, P. (1998) ‘Negotiated selves: reflec- leads to worker alienation’, this is only mean- tions on “unstructured” interviewing’, ingful if it can be verified empirically; Sociological Research Online, 3 (3) www. socresonline.org.uk/socresonline/3/3/4.html anything less is metaphysical speculation. Giddens, A. (1985) The Nation State and Moreover empiricists (unlike realists) eschew Violence. Cambridge: Polity Press. claims of causal necessity, because (after Hunter, F. (1953) Community Power Structure: Hume) it is maintained that although event A A Study of Decision Makers. Chapel Hill: may precede event B in time, we cannot be University of North Carolina. sure A brought about B. In social science this Kelly, L., Burton, S. and Regan, L. (1994) principle is exemplified by the social survey ‘Researching women’s lives or studying where the strength and direction of associa- women’s oppression? Reflections on what tion between variables is expressed, but no constitutes feminist research’, in M. necessary function claimed. Maynard and J. Purvis (eds), Researching In the nineteenth and early twentieth cen- Women’s Lives from a Feminist Perspective. turies empiricism was a radical and even lib- London: Taylor and Francis. erating movement for philosophy of science,

90 E-Jupp-3373.qxd 3/8/2006 5:42 PM Page 91

EMPIRICISM

providing a demarcation between meaningful Falsification requires a clear criterion of (that is, testable) scientific statements and what counts as a falsification of a hypothesis, philosophical speculation (Philips, 1987: 39). a condition usually only obtaining in chem- Testability allowed adjudication between sci- istry or physics and usually in the laboratory. entific theories and thus for progress to be The social sciences are mostly not based on demonstrated. The approach was enormously laboratory experiments, but rely on data gen- influential in mid-twentieth century social erated from experiments in open or semi-open science, particularly in the United States and systems, survey data or interpretive data. In especially in the survey and experimental the first two cases such data are probabilistic approaches pioneered by the ‘Columbia and thus are dependent on inductive reason- School’ (Oakley, 2000: 168). Large-scale longi- ing, that is, that the characteristics of a prop- tudinal experiments were set up to demon- erly constructed sample can be inferred to a strate whether or not social programmes had population. Nevertheless, technical difficulties worked and there were clear criteria about and the problems of operationalization aside, what counted as ‘working’. Yet ironically it might be claimed that a sample from a whilst empiricist methods were reaching defined population is immune from Popper’s their zenith in US social science, in philoso- critique, because the researcher does not phy of the natural sciences empiricism intend claiming that all xs are ys, merely that (specifically logical positivism) was undergo- a percentage of xs are ys. The degree of cor- ing a two-pronged attack from the philoso- roboration is a function of the strength of pher Karl Popper (Popper, 1959). association. The first part of Popper’s argument con- In the social survey the emphasis on prob- cerned the criterion of meaningfulness in ability and variable analysis has been a last- empiricism. Testability in empiricism requires ing legacy of empiricism and is retained in an observation language (to describe those other non-empiricist approaches such as things observed) quite separate from a theo- realism. However, Popper’s first objection retical language. The problem is that the has more telling implications. As Pawson descriptions themselves must be the result of points out (1989: 39–41), most variables in prior cognitive processes. Thus in science, social science are derived from unobserved according to Popper, there are no statements latent constructs (for example, class, ethnic- about the way the world is that are theoreti- ity etc). In other words they require prior cally neutral. An observation language there- theorizing, the data do not ‘speak for them- fore cannot be successfully differentiated from selves’ but are the result of a selection a theoretical one. process that will contain both theoretical and Popper’s second and more famous criti- empirical elements. Most social researchers cism was of the verification principle. For accept this position nowadays and this has empiricists a hypothesis was verified if there resulted in modified and more sophisticated was positive empirical evidence to confirm it. versions of empiricism. It is, for example, This raises the question of whether further accepted that there may be hidden variables positive evidence increases the level of cor- that are producing (causing) something to roboration. Popper maintained that it did not. happen, but as Hubert Blalock has argued No matter how many confirming instances (1961: 10), by definition these cannot be are observed there is no certainty of truth, observed or measured. The best that can be but just one disconfirming instance will done is to note covariation and that x and y falsify the hypotheses. On the basis of this vary together in a predictable way. According (quite correct) logical reasoning Popper built to this version of empiricism, embraced by an alternative falsificationist philosophy of most statisticians, the ability to successfully science, itself the subject of sharp criticism predict outcomes obviates the requirement later (see Lakatos and Musgrave, 1970 for a to demonstrate causality beyond statistical review of this debate). association.

91 E-Jupp-3373.qxd 3/8/2006 5:42 PM Page 92

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

Evaluation Pawson, R. (1989) A Measure for Measures: A Manifesto for Empirical Sociology. London: The influence of empiricism remains strong in Routledge. the social survey, though arguably this is more Philips, D. (1987) Philosophy, Science and to do with the emphasis on the centrality of Social Inquiry: Contemporary Methodo- measurement, associational techniques and logical Controversies in Social Science and statistical testing than adherence to a philo- Related Applied Fields of Research. Oxford: sophical doctrine. Empiricism, in the form of Pergamon. positivism, has long been the subject of criti- Popper, K. R. (1959) The Logic of Scientific cism in social science, usually by interpre- Discovery. London: Routledge. tivists who associate it with (what they see as) Williams, M. (2000) Science and Social Science: an inappropriate use of the methods of sci- An Introduction. London: Routledge. ence to study the social world (Williams, 2000: 87–104). Yet ironically something much closer to the spirit of nineteenth-century empiricism is found in just those interpretivist approaches that prioritize the experiences of EPISTEMOLOGY actors in the generation of theory (Bryman, 1988: 119). Grounded theory in particular pro- Definition ceeds from a close observation of the social A field of philosophy concerned with the world to systematically build theory, whilst possibility, nature, sources and limits of phenomenological approaches seek to bracket human knowledge. off the researcher’s prior constructs in favour of concepts generated by those studied. In each, something like an ‘observation lan- Distinctive Features guage’ is used to describe actions and events As distinct from ontology (the study of the prior to their being theorized. essential nature of reality), epistemology is concerned with whether or how we can have Malcolm Williams knowledge of reality: questions that have concerned philosophers since, at least, the Associated Concepts: falsification, grounded Ancient Greeks. Criteria for what counts as theory, methodology, ontology, phenomenol- knowledge (rather than mere belief) normally ogy, positivism, postmodernism, realism, rel- include reference to truth and to the justifi- ativism, subjectivity cation for it. A lucky guess that happens to be true might not count. Historically, a central epistemological Key Readings debate has been between empiricism and rationalism. Empiricism holds that knowledge Blalock, H. (1961) Causal Inference in derives from the external world, albeit medi- Nonexperimental Research. Chapel Hill, ated through sensory perception. Valid knowl- NC: University of North Carolina Press. Bryman, A. (1988) Quantity and Quality in edge is based a posteriori upon the ‘facts’ Social Research. London: Routledge. derived from systematic observations and Gillies, D. (1993) Philosophy of Science in the reflection upon them: inductive reasoning Twentieth Century. Oxford: Blackwell. from observational data allows the formula- Lakatos, I. and Musgrave, A. (eds) (1970) tion of general laws, which can be tested and Criticism and the Growth of Knowledge. verified through further observation. One crit- Cambridge: Cambridge University Press. icism is that this provides no guarantee of Oakley, A. (2000) Experiments in Knowing: truth, since it is always possible that future Gender and Method in the Social Sciences. observations will not provide verification of a Cambridge: Polity Press. theory. ‘Fallibilism’ or ‘falsificationalism’

92 E-Jupp-3373.qxd 3/8/2006 5:42 PM Page 93

EPISTEMOLOGY

(Popper, 1959) suggests that, whilst the truth Critical realism (‘transcendental realism’ is inherently uncertain, it is possible to know and sometimes just ‘realism’), associated par- and discard what is false. Thus, knowledge is ticularly with the work of Bhaskar (e.g. 1975), always provisional, consisting of ‘falsifiable derives from a critique of both empiricism and conjectures’: those conjectures that stand up idealism. Critical realism accepts epistemolog- to empirical testing can be regarded as nearer ical relativism (knowledge is always context- the truth than those that do not. For Popper, bound) but argues for recognition of a reality what distinguishes scientific knowledge is which exists independently of human activity that it specifies the empirical criteria upon and which should properly be the object of which it could be shown to be false (hence he inquiry. Such inquiry is not limited to what is attacked Marxism as unscientific). observable (as it is for empiricists), although Rationalism is the view that human ratio- this is important, but should be directed nality and cognitive processes make onto- towards an understanding of the deeper causal logical knowledge possible through logical structures which shape not only reality itself deduction from a priori principles. Knowledge but also the context and meaning of human is not limited to observable phenomena but perceptions. Critical realism thus re-asserts the encompasses a deeper reality which underpins importance of ontology and criteria of truth observable appearances. This is an idea that and suggests (in contrast to postmodernism) goes back to Plato. More recently, some forms that some forms of knowledge may have more of structuralism (for example, the work of validity than others. Weber’s concept of verste- Levi-Strauss) are predicated upon the view hen has been put forward as exemplifying this that there are deeper causal structures which type of approach. are not capable of direct empirical observation, but nevertheless underpin cognition as Evaluation well as language and human society. To the extent that rationalism locates The possibility of certain knowledge has knowledge in the rationality of the knowing increasingly been questioned both within phi- subject, rather than external phenomena, it losophy and by scholars from other disciplines, can be considered as a form of idealism. who draw attention to the social contexts of Some forms of idealism go further and main- knowledge production as well as the philosoph- tain that knowledge of reality is only avail- ical problems. A number of influential writers able through perceptual or theoretical grids, on the sociology of science in the 1960s and so, strictly speaking, there can be no evi- 1970s pointed out that the reality of science dence of any ontological reality outside of is much messier than is suggested by either those perceptions. The assumption of any empiricism or fallibilism. Claims about the intrinsic ability to grasp the true nature of unique rationality of science are seen to be reality thus can also be seen as questionable. reflections of conventional assumptions which ‘Reality’ cannot be known independently of are also (particularly in the work of Michel our perceptions and assumptions: it is a con- Foucault) seen as inextricably intertwined with struction of the knower. Constructivism the exercise of power, ‘disqualifying’ other underpins a range of perspectives in the knowledges. This latter point has also been social sciences, including symbolic interac- argued by feminists who point to the ways tionism, phenomenology, post-structuralism, in which women have been construed as irra- postmodernism and some forms of feminism. tional and to the marginalization of women as These are predicated on the idea that knowl- producers of knowledge and as objects of edge is contextual and that it varies over time inquiry. However, there is no single feminist and space and between social groups. All epistemology: rather feminists have adopted a knowledge is viewed as an outcome of the variety of epistemologies which Harding (1991) conditions of its construction: epistemology, characterizes as empiricism, standpointism in effect, gives way to hermeneutics. (which has much in common with critical

93 E-Jupp-3373.qxd 3/8/2006 5:42 PM Page 94

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

realism, see Cain, 1990) and postmodernism: result. In an experiment variation in a there is no one feminist epistemology. dependent variable which cannot be accounted Constructivist epistemologies, particularly for by the values of the independent variable(s) postmodernism, accept and even celebrate the is regarded as error. It can be caused by the existence of multiple ‘knowledges’ and meth- observation and the measuring equipment, by ods. However, a frequent criticism of these the experimental design, by the data processing approaches, from critical realists in particular, techniques, by variables not included in the concerns the danger of a slide into complete investigation, or any kind of random effects. relativism in which there are no epistemological criteria of truth or for judging between Distinctive Features different knowledges. Thus contemporary epistemological debates continue to be character- A major distinction is between systematic ized by contestation, many of the arguments and random error. A systematic error or bias reflecting much older arguments in philosophy. is a shift of the results common to all individual data points. For example when body weights are sampled from fully dressed Maggie Sumner children then the weight of the clothes is a systematic error. Associated Concepts: constructionism, In statistics, error denotes the difference empiricism, falsification, hermeneutics, between a true value of a parameter and hypothetico-deductive model, induction, an approximation to that value; mostly it is ontology, phenomenology, positivism, ratio- called sampling error. For example, the differ- nalism, realism, Verstehen. ence between an expectation (that is, a population mean) and a sample mean is a point in Key Readings question. For mathematical reasons the difference is Bhaskar, R. (1975) A Realist Theory of in most cases expressed as a quadratic devia- Science. Leeds: Leeds Books. tion. The best thing to happen is that the expec- Cain, M. (1990) ‘Realist philosophy and tation of this quadratic difference depends on standpoint epistemologies: feminist crimi- the sample size in such a way that the limit nology as a successor science’, in L. is zero for increasing sample size. If in the Gelsthorpe and A. Morris (eds), Feminist above example we are granted a true random Perspectives in Criminology. Milton Keynes: sample then Open University Press. pp. 124–40. Harding, S. (1991) Whose Science? Whose Knowledge? Milton Keynes: Open E(M−µ)2=σ2/N (1) University Press. Popper, K. (1959) The Logic of Scientific where M is the sample mean (regarded as a Discovery. London: Routledge. random variable because the sample is ran- Williams, M. and May, T. (1996) Introduction dom) of a sample of size N taken from the to the Philosophy of Social Research. random variable X whose expectation is and London: UCL Press. variance is σ2. (E is the expectation operator). For most other parameters of random variables similar laws obtain. Equation (1) contains two messages – one good, one bad. The ERROR good thing is that error can be kept as low as is necessary by increasing the sample size Definition (this is called consistency). The bad thing is that the increase of precision is rather slow. In general, error denotes the deviation of an Put another way, to halve the statistical error empirical observation from a true or optimal requires four times the sample size.

94 E-Jupp-3373.qxd 3/8/2006 5:42 PM Page 95

ESTIMATION

Moreover, one has to consider the possibility of bias. The sample may be biased in some ESTIMATION way; the sampled population may not be identical with the target population. The Definition expectation of the squared deviation of an Statistical estimation is the process of infer- θ estimator, say T, of some parameter can be ring the true value of a variable in the popu- shown to fulfil: lation from a sample drawn from that population. E(T−θ)2 = E(T−ET)2 + (ET−θ)2 (2)

The left-hand side of Equation 2 is called the Distinctive Features mean square error. The first term on the right of Many problems in social research require the Equation 2 is the variance of the estimator (that estimation of some unknown quantity. For is, Equation(1) in case of the sample mean), the example, we might wish to know the propor- second is the square of the bias. While the first tion of voters favouring candidate A or usually approaches zero with increasing N, the the effect of an anti-smoking commercial on second does not. The only way to keep bias the attitude of adolescents. We can gather the small lies with the choice of the estimator T, or relevant data (approval ratings) from a sam- with the experimental design, in particular with ple of subjects (voters). Based on the sample the chosen method of sampling. results, we can look for a ‘best guess’ for the true value in the population (of all voters). Evaluation Statistical estimation is concerned with this problem of inferring a population parameter If one proposes a theoretical model for a set from a sample statistic. of data then usually there will be some dis- More formally, we assume that a random crepancy between model and data. To mend sample of N observations X , X , …, X is this fact an ‘error term’ is introduced. This 1 2 N selected from the population. We denote the move, however, is meaningless unless this unknown population parameter as θ and the error term behaves appropriately. Conditions point estimator as T. The observations and T pertinent to this situation are, for example, as a function of the X are random variables. expectation zero, or variance independent of i The probability distribution of the estimator values of other variables involved in the is called the sampling distribution of T, its model (homoscedasticity), or normal distribu- standard deviation being the standard error tion. Above all, the error term should be suf- SE . We will first discuss some of the proper- ficiently small compared to the main terms T ties a good estimator should possess. (with high probability). Numerous statistical An estimator is said to be unbiased if the techniques such as the least squares method expected value of the estimator is equal to are based on error minimization. the parameter, E(T) = θ. It can be shown that =Σ the sample arithmetic mean M i(Xi)/N is an Reinhard Suck and Thomas Staufenbiel unbiased estimator of the population mean µ, E(M) =µ. On the other hand, the variance S2 Σ − 2 Associated Concepts: bias, estimation, = i(Xi M) /N of a random sample is not an inferential statistics unbiased estimator of the population variance σ2. The bias is E(S2) −σ2 =−σ2/N and can 2 =Σ − easily be removed by using s i(Xi Key Reading M)2/(N–1) instead of S2. Stuart, A., Ord, K. J. and Arnold, S. (1999) Efficiency denotes the accuracy of the esti- Kendall’s Advanced Theory of Statistics: mation, which corresponds to the variance of Classical Inference and the Linear Model. the sampling distribution of T. The smaller

New York: Oxford University Press. the standard error of the estimator SET, the

95 E-Jupp-3373.qxd 3/8/2006 5:42 PM Page 96

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

more efficient is T in the estimation of θ. Evaluation Commonly the efficiency of T is evaluated by Statistics derived from samples are used to comparing it with another unbiased estimator estimate population parameters. How good T’. The relative efficiency of T is then defined these estimates are depends on how the sam- as SE ’/SE . In a normally distributed popula- T T ple has been drawn, in particular the size of tion the sample mean and median Md are the sample and the degree to which it is both unbiased estimators of µ, but M is more representative of the population. Representa- efficient than Md. If the distributions are not tiveness can be improved by stratification of normal, this relation can be reversed. the sample. A third criterion for evaluating estimators is consistency. An estimator is consistent when the variance of the sampling distribution Thomas Staufenbiel and Reinhard Suck tends towards 0 with increasing sample size. The sample mean and many other statistics Associated Concepts: bias, error, inferential are consistent estimators. The probability that statistics, prediction, probability (random) sam- they come close to their population counter- pling, stratified sampling parts increases with larger sample sizes. The last property of an estimator to be con- Key Reading sidered here is sufficiency. T is said to be sufficient if it makes use of all the information Stuart, A., Ord, K. J. and Arnold, S. (1999) available in the sample. This is true of the sam- Kendall’s Advanced Theory of Statistics: ple mean or variance as in their computation Classical Inference and the Linear Model.

they take into account all Xi. The median, New York: Oxford University Press. mode or range on the other hand are statistics, which do not depend on all observations and are therefore not sufficient estimators. There is another approach called interval ETHICS estimation. It consists of determining an interval, which contains the parameter θ with a given probability. Its advantage compared to Definition point estimation is that the length of the inter- A field of moral philosophy dealing with the val quantifies the precision of the estimation. standards by which behaviour should be regu- For example, the 95 per cent confidence inter- lated. In terms of the ethical issues faced by val around the mean of a larger sample taken social researchers, most learned societies and from a normally distributed population is given relevant professional bodies publish codes of ± × by M 1 1.96 SEM. The probability that this ethics which provide rules, standards or guid- θ confidence interval contains is 0.95. ance on what is and is not acceptable practice. There are a number of general methods that produce estimates which possess many Distinctive Features of the above mentioned desirable properties. Two widely used methods of estimation are Research is generally justified in terms of the least squares and maximum likelihood estima- search for new or better knowledge, which tion. For example in multiple regression a is regarded as a social good or benefit. Ethical variable Y is regressed on a set of k predictors problems in research arise from the tensions = β + β Xi by means of the linear model Y 0 1 between this objective and the rights and inter- + + β + β X1 … k Xk E. The beta weights i are ests of individuals and groups which may be (ordinary) least squares estimates that mini- affected. Research is not carried out in a social mize E2. If the statistical assumptions are met, vacuum and there are a number of groups with these weights are unbiased, consistent, suffi- a direct interest in it. Most obviously, these cient and the most efficient estimators. groups include those who commission and

96 E-Jupp-3373.qxd 3/8/2006 5:42 PM Page 97

ETHICS

fund research, who have their own agendas in purposes or any other form of deception. determining what should be studied. They may A related principle is that of respect for the sometimes have a vested interest in research right to privacy: researchers should not that will produce the findings they want or is intrude on privacy to any greater degree than quick and cheap, rather than high quality, or is required by the research. There are partic- an interest in publicizing welcome results and ular issues here in relation to intrusive ques- playing down results that are less happy from tions and particularly questions on sensitive their point of view. A second group is the wider issues, such as sexual behaviour or illegal research community, which has an interest in activity. There are also legal aspects to this, in not jeopardizing future research by allowing terms of data protection legislation. A fourth research activity to get a bad name. A third key principle of ethical practice is confiden- group is those who may be affected directly or tiality, providing assurance that the informa- indirectly by policies formulated on the basis tion collected will be used only for research of research findings. Finally, but far from least, purposes. Generally, this goes together with there are those who participate directly in the assurances of anonymity: that is the individ- research as providers of the data which the ual will not be identifiable from the way in researcher seeks. Here the researcher must which the findings are presented. balance the need to obtain valid data against the rights of individuals and groups to privacy Evaluation and autonomy. Much of the debate on the ethics of social Discussions of the ethics of social research fre- research has been concerned with these last quently focus upon highly problematic pieces issues, the relationship between researcher of work, one of the most widely discussed of and researched. One basic principle is that which is Humphreys (1970) and, in particular, research should not bring harm to respon- his use of covert observation (see Bryman, dents, but longer-term consequences of 2001: 477 for a brief description). However, research participation are not always easily ethical dilemmas arise in all social research calculated. A second fundamental principle and are not confined to specific cases. of ethical social research is that of informed Because ethical dilemmas are endemic to consent, which respects the right of people to social research, the publication of guidelines know that they are being researched, the pur- has been an important activity for both learned poses of the research and what will be societies of academic researchers, for example, expected of them. People can then consent or the American Sociological Association, British not. There are issues here, of course, about Psychological Society, and professional asso- who can give informed consent: the most ciations, for example, in the UK, the Social obvious issue relates to children and the age Research Association and the Market Research at which they can be considered capable of Society; and these cover both qualitative giving valid consent (in practice, most codes research and the collection of statistical data. of practice regard parental consent as neces- All of the guidelines mentioned are available sary for people under 16). Related problems via the Internet, using search engines. can occur in a variety of situations in which However, these guidelines are just that. the validity of consent is questionable, for They are not mandatory. There is some dis- example where gatekeepers of research agreement among social researchers about the access have considerable power over poten- extent to which ethical principles are near tial participants so that people may be fearful absolute values that should never be infringed. of the consequences of non-participation (for It can be argued that it is sometimes impossi- example, in institutions such as prisons). The ble to do research without infringing ethical principle of informed consent would also sug- principles to some degree: for example, where gest that covert research is not acceptable, access to certain groups might be impossible nor is concealment of the researcher’s real without some degree of deception about the

97 E-Jupp-3373.qxd 3/8/2006 5:42 PM Page 98

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

researcher’s real purposes or where explaining method in itself and therefore can feasibly be those purposes might affect the validity of the used to support a number of methodological responses. It would follow that the researcher or theoretical approaches. However, it is only must make some personal moral judgements possible to work directly with textual data about the right balance between the need to when using The Ethnograph. know and the rights of others. The Ethnograph facilitates the qualitative data analysis process in a number of key Maggie Sumner ways, enabling the researcher to: create and manage data files; code data according to user-defined conceptual themes and group Associated Concepts: access, confidentiality, codes into families; organize data according gatekeepers, informed consent, validity to demographic information; note analytic thoughts as they occur; and having coded, Key Readings search the data set in a number of ways. In Barnes, J. A. (1979) Who Should Know What? addition, the code mapping function visually Harmondsworth: Penguin. represents how data have been coded using a Bryman, A. (2001) Social Research Methods. margin view visible whilst coding. Oxford: Oxford University Press. Textual data can be pasted direct from Bulmer, M. (ed.) (1982) Social Research a word processing application into The Ethics. London: Macmillan. Ethnograph where it will be automatically Bulmer, M. (2001) ‘The ethics of social formatted, or previously saved data files can research’, in N. Gilbert (ed.), Researching be imported. In addition, it is possible to tran- Social Life. London: Sage. pp. 45–57. scribe data directly into the editor window Homan, R. (1991) The Ethics of Social and the spell check function facilitates this Research. London: Pearson Education. process. Humphreys, J. (1970) The Tea Room Trade: As well as being able to apply factual vari- Impersonal Sex in Public Places. Chicago: ables to whole data files (for example, inter- Aldine. view transcripts), The Ethnograph includes Punch, M. (1994) ‘Politics and ethics in powerful organizational autocoding tools for qualitative research’, in N. K. Denzin and Y. S. Lincoln (eds), Handbook of Qualitative structured data. In focus group data, for Research. London: Sage. pp. 83–97. example, if formatted correctly, upon file import each speaker-section will be automatically coded. It is then easy to apply demographic variables to those coded segments. At any stage coded data can be retrieved THE ETHNOGRAPH (VERSION 5.0) and reviewed, allowing the researcher to focus on a particular code, or group of codes, Definition in isolation from the rest of the data set. One of a number of Computer Assisted Qual- The search tool interrogates the dataset itative Data Analysis Software (CAQDAS) pro- using standard Boolean operators to assist grams designed to facilitate the management investigating relationships between codes and analysis of qualitative data. Originally and/or factual variables. It provides easy developed in 1985 by the sociologist John ways to filter searches based on, for example, Seidel, it is now marketed through Qualis a particular speaker in focus group data, or a Research Associates. particular file, to which a particular or combination of variables applies. Distinctive Features Clearly labelled output following a search is usually given in the form of segments for Like all CAQDAS programs, The Ethnograph each data file and also displays other codes is a tool for facilitating analysis rather than a present in the retrieved data. The segments

98 E-Jupp-3373.qxd 3/8/2006 5:42 PM Page 99

ETHNOGRAPHIC INTERVIEWING

can then be printed and saved for importation Key Readings as data to be re-analysed. The CAQDAS Networking Project website The memo tool is versatile, whereby all http://caqdas.soc.surrey.ac.uk/ memos (whether attached to a line of text, a The Ethnograph website http://www.qualis data file, or the project) can be viewed research.com together or sorted and retrieved according to Fielding, N. and Lee, R. (eds) (1991, 2nd a number of attributes. edn. 1993) Using Computers in Qualitative Research. London: Sage. Evaluation Kelle, U. (ed.) (1995) Computer-Aided Qualitative Data Analysis: Theory Methods The evaluation of any CAQDAS package must and Practice. London: Sage. take into consideration a number of factors – Weitzman, E. and Miles, M. (1995) A including methodological and theoretical Software Source Book: Computer Programs approach, type of data, size of data set, pro- for Qualitative Data Analysis. Thousand ject aims and requirements and certain prac- Oaks, CA: Sage. ticalities. Therefore, certain packages may be particularly useful for certain types of studies and researchers are advised to investigate the various options before choosing. ETHNOGRAPHIC INTERVIEWING The Ethnograph is a package designed to handle small to medium-sized data sets in the application of a few basic tasks of qualitative Definition data analysis. Its attractiveness lies in its rela- A form of interviewing conducted in the con- tive simplicity compared to other more sophis- text of a relationship with interviewees with ticated CAQDAS software. It is fairly easy to whom the researcher has, through an ongo- learn and therefore may be particularly useful ing presence, established relations of rapport for teaching students. The software manual is and respect sufficient for a genuine ‘meeting very well written and provides methodological of minds’ and that enable a mutual explo- and step-by-step help. ration of the meanings the interviewee It lacks tools such as a text search (across the applies to their social world (Heyl, 2001). whole data set) function, and the broader range of sophisticated search facilities and visual Distinctive Features networking and modelling tools available in some CAQDAS packages. However, many soft- Ethnographic interviewing originated in cul- ware users claim that the basic coding, clearly tural anthropology, hence its emphasis on labelled, flexible retrieval and memoing tools duration and frequency of contact, on the are all they have time to use and that a lot of quality of the relationship with respondents, analytic work is anyway conducted outside the and on the meaning of actions and events to software and away from the computer. respondents. These emphases are consistent The Ethnograph may have some compati- with ethnography’s aim to ‘grasp the native’s bility problems with some operating systems point of view’ (Malinowski, 1922: 25). (W2000, WME, XP). Users should check com- Ethnographic interviews are normally con- patibility with their system before installing ducted in unstructured, in-depth format with the software. people from a particular culture, or who share particular experiences. As well as anthropology, the method is practised in fields includ- Christina Silver ing sociology, education, and psychology. What distinguishes it procedurally from other Associated Concepts: Atlas.ti, CAQDAS, cod- in-depth interviews is the centrality of rap- ing, NUD*IST, qualitative research, QSR NVivo port based on relatively long-term contact,

99 E-Jupp-3373.qxd 3/8/2006 5:42 PM Page 100

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

the investment of time in each round of can produce objective knowledge, can no interviewing and the kind of openness on the longer be unproblematically accepted. Indeed, researcher’s part that stimulates an even- the analytic tradition with which sociological handed relationship. Thus, a key feature is ethnography is most often associated – the idea that the researcher is there to learn symbolic interactionism – itself contained the from the respondent rather than impose an seeds of a more sophisticated perspective on external frame of reference, epitomized in ethnographic interviews in its vision of them Spradley’s (1979: 34) representation of the as sites of emergent interaction. Cultural researcher’s posture as being that ‘I want to anthropology, too, traced a path away from know what you know in the way that you positivist assumptions during the 1980s and know it ... Will you become my teacher and 1990s (Behar, 1996). Key texts on ethno- help me understand?’ This means that, more graphic interview methods that accommo- than other methods, interviewees are date the new perspectives include Kvale’s empowered to shape the interview’s content (1996) stage-based approach (with epistemo- and even the focus of the research, and that logical and ethical concerns arising at each the interviews can elicit strong emotions. stage), Holstein and Gubrium’s (1995) ‘active A supplementary feature is the frequent interview’ approach (where, for example, interest in documenting interviewees’ culture interviewers attune to signs of perspectival in their own language (Spradley, 1979: 24). shifts by interviewees) and Mishler’s (1986) Life history interviews can be seen as a form power-oriented distinction between three of ethnographic interview as they often share types of interviewer/interviewee relationship these emphases. (informants and reporters; research collabo- Spradley (1979) offers perhaps the most sys- rators, learners/actors and advocates). tematic treatment of the method. He devotes detailed attention to criteria for ‘locating’ Evaluation informants, and suggests what makes a ‘good’ informant (see also Johnson, 1990). He tracks Consistent with ethnography more generally, the interview’s stages step-by-step, from ethnographic interviewing raises many episte- explaining the project and interview to asking mological and ethical issues. Prominent are questions, turn-taking in talk, and various concerns about the relationship between devices that facilitate communication. He researcher and research ‘subjects’, and about discusses documenting the interviews (not the status of the accounts offered in interviews. simply a matter of audio recording), advises on These concerns are inter-related and, as well as question design, profiles appropriate types of arising from epistemological debates, they have (anthropological) analysis (domain, taxonomic, been animated by revelations about the early componential) and discusses writing up. While fieldwork that gave rise to some of the princi- attuned to concerns about the status of ethno- ples of ethnography and exemplars of its meth- graphic interview data that have latterly ods. Sexual relationships sometimes developed become prominent, some find Spradley’s between researchers and ‘the natives’, and approach rather rigid. there is also testimony that data from ethno- Ethnographic interviews are distinguished graphic interviews could sometimes be called from other types of interview analytically by to account because subjects guessed what their focus on cultural meanings. The early interpretation researchers favoured and ‘fed practice of ethnographic interviewing was them’ what they wanted to hear (Freeman, premised on the ‘naturalistic’ convention 1998). Such contingencies may not be common that, if rapport was achieved, full and frank but suggest the need to take into account the response would be forthcoming. There was motives both of researcher and subject when also an assumption that research subjects had interpreting the data from ethnographic inter- unique insight into their own culture. These views. As Heyl (2001: 370) observes, while premises, necessary to a belief that ethnography there are now many competing perspectives on

100 E-Jupp-3373.qxd 3/8/2006 5:42 PM Page 101

ETHNOGRAPHY

ethnographic interviewing (see, e.g., Mishler, 1986; Kvale, 1996; Reinharz, 1992), there is ETHNOGRAPHY agreement about its high level goals – to ‘listen well and respectfully’, to be aware of the Definition researcher’s role in ‘the co-construction of A research method located in the practice of meaning during the interview process’, to be both sociologists and anthropologists, and aware of how ongoing relationships and social which should be regarded as the product of context affect the interview process, and ‘to a cocktail of methodologies that share the recognize that dialogue is discovery and only assumption that personal engagement with partial knowledge will ever be attained’. A the subject is the key to understanding a par- reflexive posture is necessary when conducting ticular culture or social setting. Participant and analysing ethnographic interviews. That observation is the most common component can, indeed, be regarded as affirming the cen- of this cocktail, but interviews, conversa- trality the method has, since its origin, placed tional and discourse analysis, documentary on the relationship between interviewer and analysis, film and photography, life histories interviewee. all have their place in the ethnographer’s repertoire. Description resides at the core of Nigel G. Fielding ethnography, and however this description is constructed it is the intense meaning of social Associated Concepts: cultural research, life from the everyday perspective of group epistemology, ethics, ethnography, inter- members that is sought. subjective understanding, interview, life history interviewing, naturalistic data, reflex- Distinctive Features ivity, Verstehen Closely linked with, but actually the product of, various forms of fieldwork, ethnography Key Readings is a particularly valuable resource for researchers seeking to unpack cultures or Behar, R. (1996) The Vulnerable Observer. social settings that are hidden or difficult to Boston, MA: Beacon Press. locate. For instance, the researcher may seek Freeman, D. (1998) The Fateful Hoaxing of to understand the culture of places and Margaret Mead. Boulder, CO: Westview. spaces such as the factory, or the school, or Heyl, B. (2001) ‘Ethnographic interviewing’, alternatively the ethnographer may be inter- in P. Atkinson, A. Coffey, S. Delamont, ested in processes such as childhood, ageing, J. Lofland (eds), Handbook of Ethnography. sexuality or death. Whatever the focus of the London: Sage. pp. 368–83. ethnographer, the method is marked out by Holstein, J. and Gubrium, J. (1995) The Active the intensity of the relationship between the Interview. Thousand Oaks, CA: Sage. researcher and the field, and in particular, the Johnson, J. (1990) Selecting Ethnographic researcher and his/her informants. The latter Informants. London: Sage. is, of course, a natural bona fide resident of Kvale, S. (1996) InterViews. London: Sage. the social setting that is being studied, while Malinowski, B. (1922) Argonauts of the Western Pacific. London: Routledge. the researcher is to varying extents an out- Mishler, E. (1986) Research Interviewing: sider whose motivation for being there is Context and Narrative. Cambridge, MA: research-led. Harvard University Press. Ethnography is based on the assumption Reinharz, S. (1992) Feminist Methods in Social that every social group is distinctive in its Research. New York: Oxford University own right, and in order to explore this dis- Press. tinction researchers must engage with the Spradley, J. (l979) The Ethnographic group on its own ground. With participant Interview. New York: Holt Rinehart. observation at the core of ethnographic

101 E-Jupp-3373.qxd 3/8/2006 5:42 PM Page 102

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

research, this requires the ethnographer This role enabled access to the social setting to adopt a role within the setting being stud- while serving a useful purpose for those he ied that enables a smooth blending with the was attempting to observe. As Humphreys rules, codes and expectations of the locals, himself acknowledges, an overt strategy and that is pragmatic, serving the instrumen- would have resulted in failure. Many ethno- tal objectives of the study. Ethnographers graphers, however, retain both overt and have found roles for themselves in a multi- covert strategies, for social life generally tude of formal institutions such as schools, demands a degree of reticence regarding our the police, factories, churches, prisons and motives at the best of times, and researchers businesses. They have also negotiated roles are no exception. Therefore, in gaining access within a wide range of settings and cultures and maintaining a role in the field, will all or such as street gangs and poor neighbour- none of the group know the ethnographer’s hoods. All of these fieldwork settings require identity, or the reason for her or his presence? the construction of a practical and convincing Ethnographers typically collect data in the role or ‘front’ that enables data to be gleaned form of field notes and seek to induce analy- with a varying degree of personal impact on sis by way of appreciation. Appreciation is at the field setting. Strategically this involves the heart of any ethnographic analysis, which varying levels of involvement from complete utilizes categories made up of description, observer, where the researcher remains unob- interviews, documents or visual presentation trusive and observes the group in action, as appreciative cues and exemplars. From this through to the complete participant who is a ethnographic description, theory is derived full member of the social group, and is fully from applying various interpretive tech- involved in the group’s activities (see Gold, niques which are based upon presenting a 1958 for a full discussion of his four-part convincing account of the observer’s under- model of the ethnographer’s role). standing the particular social rules of the cul- When ethnographers totally adopt, in the ture (Glaser and Strauss, 1967). eyes of their critics, the culture of the researched, they are often accused of ‘going native’, or becoming one of the researched Evaluation group. The accusation here is that during Ethnography is an extremely broad church, fieldwork the ethnographer has abandoned and the range of methodological stances the role of researcher and forgotten the adopted by ethnographers extend across the- research task at hand, the disadvantage being oretical, political and technological divides that the academic product, the ethnography, (see Atkinson et al., 2001). requires a measure of distance through which Ethnography is an ideal method for deriv- theoretical work can emerge. ing meaning from social lives that otherwise The adopted role may be covert or overt, might go hidden or unnoticed. As a method it and both carry advantages and disadvantages. is intense and often makes considerable per- The adoption of a covert role often enables sonal demands upon the researcher. But as an access to fields that would otherwise be closed, academic product, the output of ethnography and can enable a relatively unimpeded, albeit can be a vivid document with human reso- intense and stressful immersion in a given nance impossible to recreate by the applica- culture. Inevitably the covert role has been tion of other methodologies. criticized on ethical grounds, and the names of certain writers will forever be associated with so called ‘unethical’ research. For Dick Hobbs instance, in his illuminating and controversial study of male sexuality, Humphreys adopted Associated Concepts: access, autoethno- the role of ‘Watch Queen’ in male public toi- graphy, critical ethnography, cultural research, lets that were being used for sexual activity. ethnographic interviewing, fieldwork,

102 E-Jupp-3373.qxd 3/8/2006 5:42 PM Page 103

ETHNOMETHODOLOGY

idiographic, microethnography, netnography, application. From this he drew the radical participant observation, qualitative research, conclusion that, as a result, they cannot pro- videography vide a scientific explanation for behaviour. Instead, values and norms must be seen as Key Readings constructs that are used by actors to account for what they and others do. Moreover, other Atkinson, P., Coffey, A., Delamont, S., explanatory resources used by conventional Lofland, J. and Lofland, L. (eds) (2001) forms of sociology, such as the notion of Handbook of Ethnography. London: Sage. interests, have much the same status. Glaser, B. and Strauss, A. (1967) The Contrary to the way in which ethnomethod- Discovery of Grounded Theory. Chicago: ology is sometimes understood, Garfinkel’s Aldine. response to this problem was to try to find a Gold, R. (1958) ‘Roles in sociological field observations’, Social Forces, 36: 217–23. more rigorous, scientific basis for the study of Humphreys, L. (1970) Tearoom Trade. human social life. He did this by redefining the Chicago: Aldine. focus of the sociological project, drawing on phenomenology. He argued that the orderliness of social life must be a product of collective human activity: of our ability to interpret situations and act on those interpretations. ETHNOMETHODOLOGY Moreover, we do this despite the fact that there are always many different ways of interpreting Definition any scene. We ‘read’ the behaviour of others for what it tells us about how they understand a An approach to studying the social world situation, what they are intending to do, and we developed by Harold Garfinkel in the early act on the basis of those ‘readings’. Furthermore, 1950s which focuses on how social order is in acting we indicate to others our understand- created, ongoingly, in and through the prac- ings, and they will in turn act on the basis of tices by which people make sense of what oth- their understanding of us; and so on. ers are doing, and display that understanding In this way, social situations are self- through their actions. In doing so it recom- organizing: their character is created continu- mends a re-specification of the focus of socio- ously in and through the actions that make logical inquiry, away from a concern with them up. But how is it possible for actors to explaining the causes, development and/or read one another’s behaviour in this way and effects of social processes or institutions. thereby to coordinate their actions? Garfinkel’s answer to this question is that, in making sense Distinctive Features of the situations they face, ordinary people engage in practical reasoning that is methodical Garfinkel started from a problem which he in character. If it were not, others would not be detected in efforts to create a theoretically able to understand it, and the social world informed scientific sociology in the United would not be intelligible. And he claims that it States in the late 1940s and early 1950s. The follows from this that what social scientists dominant theoretical framework at this time should be focusing on, if they are interested in was Parsons’s social theory. This focused on the problem of social order, are the methods ‘the problem of order’, and Parsons explained employed by actors in the practical reasoning the orderliness of social action as produced through which they continuously constitute, by the socialization of actors into the values and simultaneously display the orderliness of, and norms characteristic of their society. the social world. This is his re-specified focus However, Garfinkel raised questions about for sociological inquiry. this theory by noting that values and norms Ethnomethodology has inspired a range of do not include instructions for their own different kinds of research. The best known

103 E-Jupp-3373.qxd 3/8/2006 5:42 PM Page 104

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

is conversation analysis. This is concerned cost of re-specifying the focus of social with documenting how people who are inquiry in ethnomethodological terms is engaged in talk-in-action coordinate and con- considerable: it means that social scientists struct it in an orderly fashion. For example, cannot address many of the questions that there has been investigation of the ways in they have previously been interested in, many which certain types of utterance, such as of which relate to pressing social problems. questions, make relevant other types of utterance, such as answers; and how the production of, or failure to produce, the succeeding Martyn Hammersley action is monitored and responded to. There has also been a body of work, significantly Associated Concepts: causality, conversation different in character, and carried out by a analysis, covering-law model of explanation, later generation of Garfinkel’s students, that hermeneutics, methodology, phenomenology, is devoted to documenting work practices. positivism Here the emphasis is on the details of how particular kinds of work – notably but not Key Readings exclusively that of natural scientists and Heritage, J. (1984) Garfinkel and Ethnometh- mathematicians – gets done and how what is odology. Cambridge: Polity Press. produced is tied to the processes of its pro- Hutchby, I. and Wooffitt, R. (1998) duction (see Lynch, 1993). Conversation Analysis: Principles, Practices and Applications. Cambridge: Polity Press. Evaluation Lynch, M. (1993) Scientific Practice and Ordinary Action. Cambridge: Cambridge There has been a great deal of criticism of University Press. ethnomethodology, much of it misdirected Lynch, M. and Bogen, D. (1994) ‘Harvey because based on a false understanding of the Sacks’s primitive natural science’, Theory, nature and rationale of the enterprise. There Culture and Society, 11: 65–104. are, though, some problems which it faces. Indeed, these can be discerned in internal criticisms of each version of ethnomethodological analysis by those committed to oth- EVALUATION RESEARCH ers. For example, Lynch and Bogen (1994) criticize the work of Sacks, the founder of conversation analysis, for a ‘primitive empiri- Definition cism’ that does not take account of the con- The systematic identification and assessment stitutive character of social scientific inquiry of effects generated by treatments, pro- itself. And it is true that conversation analy- grammes, policies, practices and products. sis entails a rather restrictive conception of what can and cannot be studied scientifically. Distinctive Features At the same time, though, ethnomethodological studies of work do not display the impres- Formal evaluations are conducted for various sive cumulation of knowledge that is one of reasons: to inform policy and practice, to the most attractive features of conversation inform resource allocation, to hold policy analysis. Furthermore, abandonment of any makers accountable, to inform consumers and distinction between the orientation of science to inform decisions about programme and and that of other sorts of activity – which is policy continuation. characteristic, for example, of the work of Social policies and programmes have been a Lynch – leads to doubts about the nature of major focus of evaluation research. Social poli- the whole enterprise, given Garfinkel’s origi- cies and programmes comprise theories that nal project. Finally, it must be noted that the the introduction of some new measure or

104 E-Jupp-3373.qxd 3/8/2006 5:42 PM Page 105

EVALUATION RESEARCH

policy will bring about an alteration in a evaluation. ‘Realists’ emphasize the presence regularity that would be observed in the of various layers of social reality that need to absence of that policy or programme. In most be understood in programmes and also stress cases the policy or programme begins with generative accounts of causal forces, whose some pattern of events, behaviours or con- release through programmes needs to be ditions that are considered undesirable or grasped to make sense of outcome patterns problematic. The purpose of the policy or pro- produced. gramme is then to alter that pattern to reduce or remove the problem. A given pattern of Evaluation school failure, crime, cancer, road accidents, or unemployment might, for example, comprise There have been some sharp debates over the presenting problem. The policy or pro- evaluation methodology, in particular between gramme will then be concerned with reducing experimentalists and realists. ‘Experimenta- or removing some pattern of school failures, lists’ attempt to hold all other conditions bar crime, cancer, road accidents or unemploy- programme or policy presence constant and to ment. In a few cases programmes and policies measure the difference made by the pro- attempt to pre-empt problems that might other- gramme in relation to the problematic event, wise be expected, for example in relation to behaviour or condition being targeted. At best pollution, crime or traffic congestion. In such this is achieved by randomized controlled cases the aim of the policy or programme is to trials (RCTs). RCTs are deemed to achieve avert the emergence of a problematic pattern ‘internal validity’ – the conclusive association that would otherwise be expected. Evaluations of the treatment with the effect. They allocate of policies and programmes comprise tests of subjects randomly to experimental and control the policy and programme theories. Where conditions and compare the changes in each products are being evaluated rather than poli- before and after the introduction of the treat- cies or programmes, the theory being tested is ment to the experimental group. RCTs are con- that the product enables users better to achieve cerned to rule out any factor other than the their objectives. policy or programme for any change observed There are many methodological approaches in the experimental group. Replication is to evaluation, largely reflecting the method- undertaken in the interests of trying to achieve ological divisions in the social sciences ‘external validity’ – the generalization of find- more generally. Qualitative and quantitative ings to populations beyond those sampled for methods are used singly and in combination. experimental purposes. Meta-evaluations ‘Constructivists’ stress qualitative methods aggregate findings from series of experimental and construe programmes as constituted in studies to achieve large sample sizes that and through the negotiations of stakeholders, enable conclusions to be drawn with greater including programme architects, deliverers confidence. The Campbell Collaboration con- and participants. ‘Experimentalists’ stress ducts ‘systematic reviews’ of suites of rigorous quantitative methods and construe pro- evaluations, favouring those that are or that grammes as variables to be isolated with approximate RCTs in order to draw out effects to be observed. ‘Theory-driven’ evalu- dependable findings. ators use both quantitative and qualitative Realists stress the importance of understand- methods and fall into various camps, all advo- ing ‘mechanisms’ – the causal forces released or cating that programmes theories be identified inhibited by programmes; ‘contexts’ – the con- and tested. Evaluations using ‘theories of ditions necessary for mechanisms to be acti- change’ or ‘programme logic’ methods expli- vated; and ‘outcomes’ – the changes in cate the chains of events presumed in pro- regularity brought about by the activation of grammes and test their presence in practice. the mechanisms in the contextual conditions Other theory-driven evaluation theorists in which the policy or practice is introduced. stress the inherently normative character of Programmes and policies tend to trigger

105 E-Jupp-3373.qxd 3/8/2006 5:42 PM Page 106

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

multiple mechanisms and to operate in Key Readings multiple contexts, with varying and changing Campbell, D. and Russo, J. (1999) Social outcomes. The products of realist evaluations Experimentation. Thousand Oaks, CA: are tested suites of context–mechanism– Sage. outcome pattern configurations (CMOCs). As Chelimisky, E. and Shadish, W. (1997) in theory in the natural sciences, CMOC the- Evaluation for the 21st Century. Thousand ories are not proven. Efforts to test them Oaks, CA: Sage. however promise generalized middle range Mark, M., Henry, G. and Julnes, G. (2000) theories that become ever stronger as they Evaluation. San Francico, CA: Jossey– are further tested. The stronger theories have Bass. greater truth-content and hence are more Pawson, R. and Tilley, N. (1997) Realistic dependable as premises for action. The theo- Evaluation. London: Sage. ries embodied in policies and programmes include not only those of policy makers. Practitioners and participants will also hold programme theories that can usefully be EXPERIMENT articulated and tested. Social science may yield yet further conjectures about the ways in which programmes activate changes in Definition causal mechanisms. For realists, evaluation A research design used to draw causal infer- review involves canvassing any available ences regarding the impact of a treatment research that can be brought together to yield variable on an outcome variable. tests of realist programme theory. There is no a priori preference for experiments. Indeed, Distinctive Features realists are sceptical of experiments that disregard mechanisms and the contextual condi- An experiment normally involves the ran- tions for their activation. domized allocation of cases to experimental Whilst experimentalists take definitions of and control groups, exposing only the experi- programmes, projects, policies and interven- mental group to a treatment whilst control- tion measures at face value and attempt to ling the influence of extraneous factors. isolate their associated changes in the prob- Analysis seeks to determine whether the lem patterns at issue, realists are more con- experimental group changes significantly, cerned with the ways in which underlying over time, on the outcome variable in com- causal mechanisms come to be activated by parison with the control group that is not measures introduced, in ways often not exposed to the experimental treatment. The acknowledged or recognized by programme classic experimental design focuses on two designers and in contingent ways according variables: the treatment variable (also called to context. Experimentalists try to establish the intervention or the independent variable) what works and what does not work. Realists and the outcome variable (also called the try to establish what works for whom, in dependent variable). Its purpose is to evalu- what circumstances and how. ate whether the treatment variable has a causal impact on the outcome variable. This is achieved with five key design elements. Nick Tilley First, there is a time dimension in which cases are (normally) measured on the out- Associated Concepts: applied research, audit- come variable at least twice – once at the ing, experiment, field experiment, messy beginning and once at the end of the experi- research, meta-analysis, policy-related research, ment. Second, the design uses at least two process mapping, quasi-experiments groups of cases (the experimental and control

106 E-Jupp-3373.qxd 3/8/2006 5:42 PM Page 107

EXPERIMENT

groups). Third, cases are normally randomly participants to particular experimental allocated to these groups at the beginning of interventions. The investigator actively inter- the experiment to ensure that the two groups venes and exposes particular experimental are initially equivalent. The two groups are groups to specific interventions. An example then measured on the outcome variable. of a field experiment is the New Jersey Fourth, the treatment variable is selectively income maintenance experiment which pro- administered to the groups – the experimen- vided different levels of guaranteed income tal group is exposed to the treatment and the to different groups to explore the impact on control group is denied it. Fifth, apart from various behaviours (Orr, 1999). this differential exposure the experiences of The natural or quasi-experiment takes the two groups should be identical through- advantage of ‘naturally’ occurring events out the experiment. Finally, at some point rather than the active intervention by an inves- after the treatment(s) have been adminis- tigator. For example, an investigator, knowing tered, the two groups are re-measured on the that gambling is about to be legalized in a par- outcome variable. The amount of change ticular city could obtain measures of family between the beginning and end of the exper- well-being before and after the legislative iment is computed for both groups. If the change. A control group could be obtained by amount of change in the experimental group selecting a similar city where gambling is significantly different from that in the con- remained illegal and examining whether family trol group this difference is attributed to the well-being changed more in one city than the causal impact of the treatment or interven- other. The difficulty with this type of design is tion. This conclusion is drawn because the that it can be difficult to match the ‘experi- two groups should have been identical to mental’ and ‘control’ groups and to eliminate begin with and, with the exception of the the influence of other external changes. treatment, have been exposed to identical conditions throughout the experiment. As the Evaluation treatment is the only difference between the groups any differences between the groups at Generally speaking, the experimental design the end of the experiment are attributed to has been seen as providing a more power- the causal impact of the treatment. ful way than non-experimental designs to The settings in which experiments are con- uncover causal relationships. However, it ducted vary from laboratories to real-life encounters important shortcomings in social social settings. Laboratory settings have the research. advantage of being able to control many of One difficulty arises from practical and the extraneous factors and can help ensure ethical considerations. It is frequently neither that both the experimental and control ethical nor feasible to intervene in people’s groups have the same experiences throughout lives to see what happens. Experiments can the experiment. However, because of con- also face methodological shortcomings. While cerns about artificiality and the fact that experiments provide a powerful way of identi- many social experiments last for extended fying the causal impact of an intervention this periods, laboratory experiments are not comes at a cost. By removing the influence of widely used in social research. Real-life exper- all other factors through random allocation to iments involve either field experiments or experimental and control groups the experi- natural quasi-experiments. Field experiments ment has only a limited capacity to arrive at are conducted in real-life settings but retain explanations as to why one factor has the effect many of the essential features of the classic it does. Nor is it a good way of building a experimental design. In a field experiment the picture of how a set of factors produce a par- investigator creates experimental and control ticular outcome. Experiments typically focus groups through the randomized allocation of on the impact of just one or two factors.

107 E-Jupp-3373.qxd 3/8/2006 5:42 PM Page 108

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

A number of factors can threaten the Orr, Larry (1999) Social Experiments: Evaluating internal validity of experimental results because Public Programs with Experimental Methods. they can create changes in the outcome inde- Thousand Oaks, CA: Sage. pendently of the experimental intervention. de Vaus, D. (2001) Research Design in Social This makes it difficult to evaluate the extent to Research. London: Sage. which change is due to the intervention. Such factors include the influence of other events occurring during the experimental period (the history effect), change due to participants grow- EXPLOITATIVE RESEARCH ing older (maturation effect) and change due simply to being selected for and participating in an experiment (selection and testing effects). Definition Experiments can also be afflicted by measuring Social research often involves an intrusion instruments becoming less appropriate over the into people’s lives and ethical guidelines exist course of the experiment – especially in experi- (for example within universities or from pro- ments where maturational change is likely fessional organizations) to try to ensure that (instrument decay effect). Change over the individuals are not exploited in any way course of an experiment can also be attributed during the research process. However, histor- to the statistical phenomenon of regression to ically and to date there are examples of the mean and to experimental drop-out. Control research that exploits respondents physically, groups can minimize the impact of these prob- emotionally, sexually and so on. lems. Since both experimental and control groups should be affected equally by most of Distinctive Features these issues, it is possible to isolate the influence of the experimental intervention by focus- Defining the ethics of social research is coming, not on the absolute amount of change, but plicated; social research is a dynamic social on differences in the levels of change between and human process that often involves an the experimental and control groups. intrusion into people’s lives. Researchers While it can be difficult to implement the have a responsibility to their respondents and classic experimental design in social research need to ensure that the design, process and it is nevertheless a useful model to bear in delivery of the product of research is ethical mind when designing research. The logic of and not exploitative. Universities and profes- the design is useful and can inform the devel- sional organizations produce ethical guide- opment of less rigorous, quasi-experimental lines for researchers and increasingly designs. researchers have to present reports on their method and approach to internal and external ethics committees. The potentiality for harm David de Vaus caused by taking part in research may be more obviously evident in medical or psy- Associated Concepts: causality, ethics, eval- chological experiments, but exploitation of uation research, field experiment, hypothesis, respondents is possible in all social research. informed consent, messy research, quasi- For example, a researcher may ask respon- experiment, validity dents to divulge views that are considered by many to be deviant and/or may be asking respondents to revisit events and experiences Key Readings that make them distressed. It is possible to Campbell, Donald T. and Stanley, Julian C. argue that if respondents freely agree to be (1963) Experimental and Quasi-Experi- part of research then they have some respon- mental Designs for Research. Chicago: sibility for the relationship. However, Rand McNally. researchers who study people who are

108 E-Jupp-3373.qxd 3/8/2006 5:42 PM Page 109

EXPLOITATIVE RESEARCH

(arguably) particularly vulnerable (for example in physical, emotional and sexual danger children, intellectually disabled adults) and also. who undertake covert research – that takes Feminists and others have argued that in place without the knowledge or consent of order to be ethically sound and non-exploitative those being studied – need to think even research should be ‘for’ rather than ‘of’ the more carefully about the possible exploitative people being studied. Oakley (2000) gives an aspects of the research process. In addition, example of a research project that could be although a researcher – through their own described as ‘research of women’ rather reflexive practice and adherence to ethical than ‘research for women’. She cites a large guidelines – may collect the data in an ethi- research project on the social origins of cally sound, non-exploitative way the way, depression and notes that the study resulted the findings are used (which is not always in a convincing explanation of the relation- within the researcher’s control) may still neg- ship between women’s depression and their atively affect respondents. oppression. But there was no concern with Although not often considered within whether or how women defined themselves research reports and ethical guidelines, the as depressed, but only with how the state of research process can also be exploitative of women’s mental health could be exposed and members of the research team: for example, fit into a system of classification developed research staff are often employed on short- by a profession of ‘experts’ on mental health term, insecure contracts and there is much (psychiatrists). Also, the researchers did not anecdotal evidence of abuse of power with begin with a desire to study the situation of research teams. women or set out to give women a chance to understand their experience as determined by the social structure of the society in which Evaluation they lived. The primary aim of the data was In the second half of the twentieth century to study depression and women were theorists and researchers began to argue that selected as respondents because they are doing sociology (and we could add all social easier (and therefore cheaper) to interview, science) is inevitably a political endeavour. being more likely than men to be at home Shulamit Reinharz (1983: 95) compared tradi- and therefore available during the day. tional research to rape, where researchers The response of some researchers to this is ‘take, hit and run’ and ‘intrude into their sub- to argue that researchers should not aim to jects’ privacy, disrupt their perceptions, utilize represent the ‘other’ (that is, should not false pretences, manipulate the relationship, undertake research with people unlike them- and give little or nothing in return’. Histori- selves) in order to limit the possibilities of cally there are many examples of researchers exploitation. However, there are problems exploiting respondents physically, emotionally here, not least because academia is not repre- and sexually in order to get the ‘results’ that sentative of all groups (for example, in rela- they want. tion to gender, ethnicity, age, dis/ability), Yet, it is important to note that even within which could mean that the experience of research that aims to be non-exploitative the some groups remains unconsidered. researcher usually has control over the research process from design through to analysis and so holds the objective balance of Gayle Letherby power. Thus, as Judith Stacey (1991: 144) argues, ‘elements of inequality, exploitation, Associated Concepts: confidentiality, covert and even betrayal are endemic to [research]’. research, disability research, emancipatory Having said this, we need also to be aware research, ethics, feminist research, informed that researchers are not always ‘all powerful’ consent, politics and social research, sensitive and that research may place the researcher research

109 E-Jupp-3373.qxd 3/8/2006 5:42 PM Page 110

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

Key Readings throughout the research. The exploratory researcher does not approach their project Homan, R. (1991) The Ethics of Social according to any set formula. She/he will be Research. London: Longman. flexible and pragmatic yet will engage in a Kimmel, A. J. (1988) Ethics and Values in broad and thorough form of research. Those Applied Social Research. London: Sage. Oakley, A. (2000) Experiments in Knowing: engaged in exploratory research are con- Gender and Method in the Social Sciences. cerned with the development of theory Cambridge: Polity Press. from data in a process of continuous dis- Reinharz, S. (1983) ‘Experiential research: covery. Thus, exploratory research leads to a contribution to feminist theory’, in what Glaser and Strauss (1967) appeal for G. Bowles and R. D. Klein (eds), Theories in The Discovery of Grounded Theory, that is ‘a of Women’s Studies. London: Routledge. rhetoric of generation’ (1967: 18), where Stacey, J. (1991) ‘Can there be a feminist exploration is best interpreted as exploration ethnography?’, in S. B. Gluck and D. Patai for discovery. This specifically differs from (eds), Women’s Words, Women’s Words, exploration simply as investigation, which is Women’s Words: The Feminist Practice of rather too general a meaning for exploratory Oral History. New York: Routledge. research in social science, and from exploration as innovation, which is in contrast, rather too narrow and focused. Instead, exploration constitutes a distinct form of discovery. EXPLORATORY RESEARCH

Definition Evaluation Exploratory research is a methodological Exploratory research suffers from some fun- approach that is primarily concerned with damental misrepresentations and misunder- discovery and with generating or building standings. Exploratory research is often seen theory. In a pure sense, all research is rather simplistically as one initial stage in the exploratory. In the social sciences exploratory developmental process of conducting a sys- research is wedded to the notion of explo- tematic research inquiry. In this sense ration and the researcher as explorer. In this exploratory research has become synony- context exploration might be thought of as a mous with the notion of ‘feasibility study’ or perspective, ‘a state of mind, a special per- ‘pilot study’, both of which imply a prior or sonal orientation’ (Stebbins, 2001: 30) toward sequential stage in a research programme and approaching and carrying out social inquiry. thus a limited appreciation of exploration. It is also misleading to use exploration as a syn- Distinctive Features onym for qualitative research. Although it can involve qualitative inquiry, this reading In presenting social inquiry in the above way, narrows the meaning of exploratory research Stebbins draws on C. Wright Mills’s ‘On and undermines the notion of exploratory Intellectual Craftsmanship’ (Mills, 1959: research as being concerned with the devel- 195–226). This seminal description of the opment of theory from data. lifestyle of the social scientist as explorer is a Aside from misrepresenting exploratory rich and detailed account of how to do research through using inappropriate short- exploratory research. Stebbins also develops hand synonyms there are further issues that the notion of ‘concatenated exploration’ relate to the poor understanding of explo- (2001: 12) to emphasize how exploration ration. Exploratory research as exploration- refers to the overall approach to data collec- for-discovery is perhaps the most appropriate tion, not only at the beginning but also sense of exploration to convey here. Two other

110 E-Jupp-3373.qxd 3/8/2006 5:42 PM Page 111

EXTENDING SOCIAL RESEARCH

senses are referred to above – investigative and innovative. Neither of these forms of EXTENDING SOCIAL RESEARCH exploration properly represents exploratory research in the social sciences. However, fur- Definition ther contentious issues arise from the confu- An approach which involves the researcher sion that often emerges surrounding making findings available for implementation exploratory research as a distinct form of dis- of policy and practice and being involved in covery. The tendency in social sciences such implementation rather than solely con- toward verification of theory incumbent with cerned with the design and completion of deductive and confirmatory procedures is research. sometimes used as a platform to discredit generation of theory through exploratory research. Confirmatory research has dis- Distinctive Features tinctly different goals to scientific exploration in the social sciences. The goal of the former The current emphasis on the usefulness is to test hypotheses. The quality of the of social research is nothing new. Indeed, research design is key in turning the hypo- some argue that the current momentum and thetical into the conclusive so that generaliz- level of activity associated with the idea of ability follows. In crude terms confirmatory evidence-based policy (or evidence-based research is valid, reliable and representative, practice) is unprecedented, certainly in the it settles and confirms where exploratory UK. However, despite these and other moti- research allegedly remains inconclusive, vations for enhancing the impact of social sci- messy and vague. For those who remain to be ence research on the social world it examines, persuaded that such criticisms are unfair and models of the social research process still inappropriate, the associated concept of commonly end at the point when findings are ‘grounded theory’ is called upon to counter reported and this leads to others suggesting such criticisms. Credibility, plausibility and that social research has little impact on trustworthiness are hallmarks of Glaser and public policy. Strauss’s exposition of grounded theory, Extending Social Research (see Letherby which as noted above, is in itself an appeal and Bywaters, forthcoming 2006) questions for exploratory research. this end point and its associated apoliticism. From this perspective the research process is extended to encompass the tasks of making Pamela Davies findings available, and their application and implementation in policy and practice; these Associated Concepts: grounded theory, elements are seen as an integral part of the hunting (snooping and fishing), hypothetico- research process, as much part of the respon- deductive model, intellectual craftsmanship, sibility of the researcher as setting the aims methodology and deciding the methods, rather than detached activities which take place after the research has been completed. Key Readings Extending Social Research thus extends Glaser, B. G. and Strauss, A. L. (1967) The the boundaries of the traditional ‘design, Discovery of Grounded Theory: Strategies funding, access, fieldwork, analysis and out- for Qualitative Research. London: Sage. put’ model of research and involves not only Mills, C. Wright (1959) The Sociological a further stage – continued involvement and Imagination. Harmondsworth: Penguin. further developments – but also involves at Stebbins, R. A. (2001) Exploratory Research in every aspect of the process a reconsideration the Social Sciences. London: Sage. of what is meant by the research product(s).

111 E-Jupp-3373.qxd 3/8/2006 5:42 PM Page 112

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

Traditionally and to date, this is largely ‘From research to practice, programs and thought to be the research report and the pre- politics’ in which they argue that researchers sentations and publications that we write. can and should be influential in ‘determining Extending Social Research argues for the the impact of their study’. However, here too process to include active engagement with there is a clear distinction drawn between the the impact of research. This goes far beyond researchers’ role and that of ‘managers, reporting findings as clearly as possible or policy makers and practitioners’ (p. 177) and making recommendations for policy makers the focus is on the task of making recommen- or practitioners to consider. This way of dations, a language which itself reinforces the thinking about research has an influence sense of distance between researcher and throughout the process, from the way application. In their conclusion, Dartington research is set up, the process of engagement and Scott (p. 188) recognize that ‘research is with funders, collaborators and respondents rarely a linear process and it often takes the throughout the research programme and the researcher down unexpected pathways’. The way researchers engage with processes of impact of the research may start very early on, information giving, application and imple- and be an aspect of the research itself, even in mentation. Extending Social Research influ- studies that are not thought of as ‘political’. ences every stage, includes further work for Extending Social Research does not suggest and with research funders and respondents, that the researchers’ engagement with the and necessitates not only a revision of the tra- political aspects of research and the research ditional research model but a challenge to process is unique but that all researchers (not many accepted and expected academic work- just those engaged in explicitly identified ing practices. ‘action’/’political’ research) have a responsibility to engage with the political aspects of the Evaluation research process and the political potential of the work they do. Such is the current emphasis on evidence- However, Extending Social Research is not based policy and practice that it may be nec- always easy. For example, the closeness that essary to justify the charge that currently it necessitates with funders, collaborators and most accounts of the research process end respondents may prove difficult when with the production of a report. Most, if not researchers have difficult things to say and/or all, research funders require that bids include don’t deliver findings that funders expected more or less detailed accounts of how find- or hoped for. Furthermore, despite the con- ings are going to be made known. However, cern of academic research funders to pro- this expectation amongst funders is at mote applied and practice-based research, odds with standard accounts of the research other academic pressures, including the need process taught to successive generations to secure further research money and pub- of social science students. The final stage lishing research findings, means that time of producing ‘outputs’ is usually described spent on Extending Social Research can feel by teachers and authors as ‘writing up’ like time wasted. This means that, for some, ‘giving talks’ or ‘reporting and presenting the applied and practice-based research is some- findings’. what of a poor relation and a distinct type of Some texts and some models do clearly research apart from more theoretically identify the issue of the impact of social informed traditional research. research as an issue for the research process: that is, those that attempt to develop action and/or emancipatory research approaches. Gayle Letherby and Paul Bywaters And even those not adopting these models sometimes explicitly address the issue of Associated Concepts: action research, impact. For example, Darlington and Scott applied research, emancipatory research, explo- (2002: 177) include an Epilogue entitled itative research, policy-related research, politics

112 E-Jupp-3373.qxd 3/8/2006 5:42 PM Page 113

EXTENDING SOCIAL RESEARCH

and social research, research design, writing Hood, S., Mayall, B. and Oliver, S. (eds) research (1999) Critical Issues in Social Research. Milton Keynes: Open University Press. Letherby, G. (2003) Feminist Research in Key Readings Theory and Practice. Milton Keynes: Open University Press. Darlington, Y. and Scott, D. (2002) Letherby, G. and Bywaters, P. (2006) Qualitative Research in Practice: Stories Extending Social Research: Application, from the Field. Milton Keynes: Open Implementation and Publication. Milton University Press. Keynes: Open University Press.

113 F-Jupp-3373.qxd 3/8/2006 5:44 PM Page 114

factor, all the correlations between the factor FACTOR ANALYSIS and the variables should be high.) The outcome of factor analysis depends on Definition the variables that have been measured, the A set of procedures used to simplify complex respondents who yielded the data and the sets of quantitative data by analysing the method of factor analysis used. When plan- correlations between variables to reveal the ning to carry out factor analysis there need to small number of factors which can explain be at least 100 respondents and twice as the correlations. many respondents as there are variables. The respondents should be heterogeneous on the Distinctive Features variables being studied because if they are not the correlations between variables will be Factor analysis is used to reveal the mathe- low. The correlations should be linear, and matical constructs underlying the responses this can be seen if scatterplots are examined. to a set of variables such as tests or question- In exploratory factor analysis, one usually naire items. So it has clarified the nature of begins with a principal components analysis intelligence and personality, and can indicate, which yields a set of uncorrelated components. for example, whether a questionnaire that The aim of factor analysis is to explain the cor- claims to measure stress at work is measur- relation matrix with as few factors as possible. ing a single dimension. The number to extract can be determined from Exploratory factor analysis is employed to a scree plot or by examining the eigenvalues of identify the main constructs that will explain the principal components analysis. Factor the intercorrelation matrix, a table showing analysis is then run with the number of factors the responses to each variable correlated set to the value obtained from the principal with the responses to every other variable. components analysis. Usually one allows rota- Confirmatory factor analysis is where one tion of factors, using a method such as Varimax. tests whether hypothesized factor loadings fit There can be advantages in having oblique an observed intercorrelation matrix. or correlated factors, which are obtained if A component or a factor explains the vari- one uses a procedure such as Oblimin. In ance in the intercorrelation matrix, and the deciding whether to use orthogonal or amount of variance explained is the eigen- oblique factor rotation, one can apply ortho- value for the factor. An aim in factor analysis gonal rotation and see whether it provides a is to obtain simple structure. This means that simple structure solution. if there are two or more factors, each factor Communality is the proportion of the vari- should have high correlations with a few ance in each variable which the factors variables and the remaining correlations explain. A factor loading is the correlation of should be close to zero. (If there is only one a variable with a factor. A loading of 0.3 or F-Jupp-3373.qxd 3/8/2006 5:44 PM Page 115

FALSIFICATION

more is frequently taken as meaningful when sense, the empirical confirmation or falsification interpreting a factor. The naming of factors is of propositions is a routine aim in many areas a subjective process: one examines the vari- of social science. In a stricter sense (as asso- ables with high loadings on the factor and ciated especially with the philosopher Karl selects a name that summarizes the content Popper, 1934/1959), the falsification and the of these variables. ‘falsifiability’ of propositions is advanced as providing a superior criterion of science (‘fal- Evaluation sificationism’) than ‘verification’ and ‘verifiability’ (science as verified or potentially Factor analysis is of great value in elucidating verifiable knowledge – the ‘verification prin- sets of correlational data, but it is a group of ciple’) as in logical positivism. procedures so the researcher has to decide which ones to use and these decisions will Distinctive Features affect the outcome. It is a mistake to suppose that carrying out factor analysis is like arith- According to Popper, given the ever-present metic in that there is one ‘right’ answer, and possibility of new and potentially refuting all others are ‘wrong’. Factor naming is an evidence, an inductive universal generaliza- entirely subjective procedure. tion can never be finally verified, whereas a Modern computer packages allow one to single non-supporting (‘falsifying’) occur- perform in a few seconds a factor analysis that rence can refute a hypothesis. Thus, a single would have taken many hours of hand com- black swan refutes the general hypothesis putation. But this means that it may be used that ‘all swans are white’. On this view, sci- inappropriately. Researchers should refer to ence is best defined in terms of the ‘falsifia- one of the texts on multivariate analysis to bility’, rather than the ‘verifiability’, of its ensure the data are suitable for factor analysis theories and hypotheses. This also has the to be applied and for guidance on which of the merit of acknowledging the essential provi- many procedures available to use. sionality of scientific knowledge. For Popper, the ‘falsifiability’ of a disci- Jeremy J. Foster pline’s propositions provides the decisive criterion of demarcation between science and non-science. A virtue of the viewpoint is that in Associated Concepts: composite measure- recognizing the importance of hypotheses and ment, correlation, quantitative research, scal- theories in science, and of changes in scientific ing, SPSS knowledge, it also captures the ‘critical spirit’ of science. Hence, this position is sometimes Key Readings also referred to by Popper as critical rationalism. Kline, P. (1994) An Easy Guide to Factor A related feature of falsificationism or crit- Analysis. London: Routledge. ical rationalism, according to Popper, is its Tabachnick, B. G. and Fidell, L. S. (1996) encouragement of bold hypotheses since the Using Multivariate Statistics. New York: more unexpected the initial conjecture or pre- HarperCollins. diction, the more severe the tests, and the more powerful will be the explanatory contribution to science where these hypotheses withstand falsification (in Popper’s terms, FALSIFICATION receive ‘corroboration’).

Definition Evaluation The empirical refutation of a scientific Although it has attracted (and still finds) a hypothesis or proposed law. In a general good deal of support among social scientists

115 F-Jupp-3373.qxd 3/8/2006 5:44 PM Page 116

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

and also in social policy circles (and is are often independent of particular theories regarded as counting especially against ‘un- and ‘falsification’ often serves well enough falsifiable’ social theories such as Marxism as a pragmatic approach. However, this is a and Freudianism), critics of falsificationism long way from the claims associated with challenge its cogency on a number of counts: ‘falsificationism’. (1) that ‘the facts’ which are put forward as the basis of the ‘independent’ test of David Jary theories and hypotheses are themselves ‘theory-laden’ – experiments (including Associated Concepts: covering-law model of their instrumentation), for example, are explanation, hypothesis, hypothetico-deductive both constituted by and interpreted using model, positivism, realism theories; (2) in practice, in science, and contrary to Key Readings the position that can be termed naive falsificationism, it transpires that a single Blaug, M. (1980) The Methodology of refutation is rarely decisive since suc- Economics or How Economists Explain. cessful theories are not lightly discarded; Cambridge: Cambridge University Press. (3) the attempt (see Lakatos and Musgrave, Feyerabend, P. (1971) Against Method. 1970) to replace naive falsificationism London: New Left Books. with a sophisticated falsificationism, in Lakatos, I. and Musgrave, A. (1970) Criticism which an overall judgement is made and the Growth of Knowledge. Cambridge: between progressive and degenerating Cambridge University Press. scientific research programmes, fails to Popper, K. (1934, English trans., 1959) The overcome the problems of falsification- Logic of Scientific Discovery. London: Hutchinson. ism, for if no single observation is decisive and scientific research programmes are rejected and accepted holistically, falsification no longer provides a clear- cut rule of thumb in the day-to day pro- FEMINIST RESEARCH cedures of science, or any clear overall demarcation between science and non- Definition science. What makes research ‘feminist’ is a con- For many commentators (notably Feyerabend, tentious question. A view of feminist research 1971), the procedures suggested as mandatory as ‘on, by and for’ women has been seen as for science by falsificationists simply fail to fit problematic and debates have focused upon the past and present activities of science, and whether feminist research can be defined if used strictly would be likely to cripple it. through a distinctive choice of topic, methods, What Feyerabend most objected to was the methodology or epistemology. ‘over-rationalized’ and dogmatic character of Popper’s account of science and his claims for Distinctive Features a single ‘scientific method’. Where does this leave ‘falsification’? As The term ‘feminism’ encompasses a variety of Blaug (1980: 42) has remarked: ‘once granted strands of thought. However, feminists can be that completely certain knowledge is denied said to share recognition of the significance to us there is nothing uncomfortable about of gender in the organization of social life, of the profoundly theoretical nature of our way asymmetries of power in gender relations of looking at facts about the real world’. and a political commitment to revealing and In practice (and at odds with claims for negating women’s subordination. Since the ‘incommensurability’ and ‘relativism’) ‘facts’ 1970s, feminists have engaged in a critique of

116 F-Jupp-3373.qxd 3/8/2006 5:44 PM Page 117

FEMINIST RESEARCH

‘malestream’ assumptions in research across a and epistemological foundations of conventional range of social science (and other) disciplines. science. This approach corresponds with Historically, social (and also natural) science what Harding (1991) calls ‘feminist empiri- and research had been largely carried out by cism’, which she contrasts with ‘stand- men and focused upon men, representing the point feminism’ and ‘postmodern feminism’, resulting knowledge as universal and objec- both of which present greater challenges to tive. Thus conventional science had ignored conventional science. the issue of gender and had ignored or mar- For many feminists, the issue was not about ginalized women, (mis)representing women ‘adding women in’, but rather of developing a through stereotypical assumptions (for exam- more appropriate approach to researching for ple, casting women as ‘irrational’ or ‘emo- women. Research agendas needed to start tional’ in contrast to the rationality and from women’s experiences, asking questions objectivity claimed for scientific knowledge). relevant to women’s real lives, including those Beyond critique, there has been much less private and personal aspects marginalized in agreement between feminists about what the ‘malestream’ research (recognizing that the characteristics of feminist research might and personal is political). Hence, feminists stressed should be. the importance of recognizing subjectivity, not The idea current in the early 1980s that only that of the ‘researched’ (as recognized in feminist research should be on, by and for the phenomenological tradition) but also that women has come to be seen as problematic. of the researcher. The claims of science to Recognizing the significance of gender may ‘objectivity’ were questioned, together with make it important to study men as gendered the associated idea of the researcher as a beings who play an important role in detached observer of social life, as if the women’s lives. Nor is feminist research con- researcher’s own personal biography and fined to research by women, since not all involvement do not impinge upon the produc- women are feminists. The important charac- tion of knowledge. The conventional view of teristic of feminist research therefore seemed the detached, objective researcher was recog- to be that it is research for women (for exam- nized as leading to a view of research as a one- ple, Duelli Klein, 1983), although there are way process, in which those who are different interpretations of what this might researched are treated as objects: such research mean. It does mean that there is an explicit involves an exercise of (abusive) power which political dimension to feminist research, but was seen as ethically and politically inappro- there are debates about whether this implies priate to feminist research. Women should be distinctive approaches to method (and, treated as ‘knowers’ and as experts in their although qualitative methods, especially own lives. Knowledge could thus be viewed as depth interviewing, have often been pre- a co-production between researcher and ferred, in practice, feminist research has used researched, based on relationships of reciproc- a variety of methods, including quantitative), ity and shared experience. methodology and epistemology. Standpoint feminism takes the view that An initial response to the critique of mas- research is necessarily a political act: knowl- culinist science was an argument for the edge is necessarily produced in particular con- adoption of non-sexist research practice, for texts and from particular perspectives. Women example the avoidance of sexist language (for ‘know’ from their particular experiences and example, using the term ‘men’ to include their knowledge must be regarded as valid women) and avoiding generalizing to women knowledge. Some have gone further to argue from findings based on men. ‘Adding women that women’s knowledge is not merely differ- in’ would, on this view, lead to ‘better’ sci- ent from the knowledge produced by masculin- ence by producing a more complete knowl- ist science, but is actually ‘better’ knowledge: edge of reality. However, it is an approach women (like other oppressed groups) have that leaves unquestioned the methodological both to ‘know’ from the perspective of their

117 F-Jupp-3373.qxd 3/8/2006 5:44 PM Page 118

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

oppressors (who have power to define the rigour is abandoned. Rather, the aim is to world) and know from their own perspective produce good and useful knowledge (Cain, (knowledge which has hitherto been excluded 1986) on the basis of engagement with epis- or subordinated). Standpointism holds that temological and methodological issues, which feminist research carried out on this basis pro- are core to debates in social research more vides a new ‘successor’ science. generally. There are a number of problems with standpointism. One is that, if knowledge is necessarily produced from particular stand- Maggie Sumner points, then why should women be seen as having particular access to ‘truth’? Secondly, Associated Concepts: anti-racism research, there are problems about whether there is a critical research, emancipatory research, ‘women’s standpoint’: as black feminists in epistemology, intellectual craftsmanship, particular have pointed out, ‘women’ are not politics and social research, postmodernism, a unitary category and women’s experiences realism, relativism, standpoint research are fragmented by racial, generational and other differences such as sexual orientation. Key Readings Postmodern feminists argue that these criti- Cain, M. (1986) ‘Realism, feminism, method- cisms mean that, inevitably, notions of ology and law’, International Journal of absolute truth must be abandoned: rather Sociology of Law, 14: 255–67. there must be recognition of a multiplicity of Duelli Klein, R. (1983) ‘How to do what we relative ‘truths’ from the diverse experiences want to do: thoughts about feminist of women (and men). As with postmodernism methodology’, in G. Bowles and R. Duelli in general, this seems to slide into epistemo- Klein (eds), Theories of Women’s Studies. logical relativism. However, some writers (for London: Routledge. pp. 88–104. example, Cain, 1986; Stanley and Wise, Harding, S. (1991) Whose Science? Whose 1993), whilst accepting that knowledge is Knowledge? Milton Keynes: Open fractured amongst a multiplicity of stand- University Press. points, have argued a case for a (modified) Letherby, G. (2003) Feminist Research in ‘successor science’, which holds out the pos- Theory and Practice. Milton Keynes: Open sibility of ‘better’ knowledge (though not University Press. absolute truth) produced through personal Stanley, L. and Wise, S. (eds) (1993) Breaking and theoretical reflexivity about the assump- Out Again: Feminist Ontology and tions and frameworks utilized in research. As Epistemology. London: Routledge. Cain (1986) points out, this is a position close to epistemological realism.

FIELD EXPERIMENT Evaluation Criticisms of feminist research hinge around Definition its overtly political stance, the rejection of An experiment conducted outside the labora- objectivity and issues about the relationship tory, in a ‘natural’ setting. between method and experience. However, it could be replied that all research is produced Distinctive Features from particular standpoints and that feminists endeavour to make their standpoints explicit. Field experiments are contrasted with labora- Thus, an explicit political orientation and tory experiments. Laboratory experiments recognition of the problems of ‘objectivity’ are conducted in closed, controlled condi- does not mean that theoretical and intellectual tions. Field experiments are conducted in

118 F-Jupp-3373.qxd 3/8/2006 5:44 PM Page 119

FIELDWORK

open, natural settings. Field and laboratory respects. This is often unreasonable, and experiments are undertaken both in the phys- partially explains quite frequent variations in ical and social sciences. Laboratory experi- findings from study to study. ments attempt to create conditions in which Field experiments do not always attempt hypotheses about causal powers can be tested random assignment. In both the physical and in idealized conditions conducive for their social sciences they may also use observa- expression. Field experiments are sometimes tional and quasi-experimental methods. At conducted where laboratory experiments best, they aim to identify and explain varia- cannot be undertaken because of the nature tions according to theoretically specified of the phenomena being examined, for exam- sub-groups. ple in relation to many issues in climatology. Field experiments are also conducted where the potential application of laboratory-based Nick Tilley findings to natural conditions is important, for example in trials of medical treatments Associated Concepts: applied research, applied to real, human patients as against causality, evaluation research, experiment, non-human animals in laboratories. hypothesis, quasi-experiment, validity In the social sciences, the scope for laboratory experimentation is more limited than in Key Readings the natural sciences, with the exception of some areas of psychology. Hence field experi- Campbell, D. and Russo, J. (1999) Social ments are quite common. Field experiments Experimentation. Thousand Oaks, CA: are especially important in evaluation research, Sage. where the effects of new policies and practices Robson, C. (2001) Real World Research. are in question. Oxford: Blackwell.

Evaluation

Achieving the control applied in laboratories FIELDWORK in field settings is highly problematic. In an effort to isolate the effects of a given treat- Definition ment, random assignment to experimental and control conditions is often attempted. The research practice of engaging with the The only difference for the experimental worlds of others in order to study them at group comprises the treatment applied, and a close quarters. This is carried out in order comparison of changes in the experimental that the fieldworker can gain an understand- and control groups is made in the expectation ing of the everyday operations and mecha- that this will reveal the independent effects nisms of a particular way of life, and the of the treatment. Random assignment, how- meanings that members of that culture ever, is often not practicable, especially attribute to these everyday occurrences. where communities are the focus of intervention or where treatment groups cannot Distinctive Features be isolated from experimental ones. Random allocation promises a method of achieving The natural playing out of cultural life is the internal validity. It does not deal with many object of the fieldworker’s trade, and this can threats to external validity. Populations from only be achieved by making close observa- which random allocation is made to treat- tions from within an environment that is ment or control are always spatio-temporally natural to the observed rather than to the specific and generalizations to other popula- observer. To varying extents the fieldworker tions have to assume similarity in all relevant is a stranger in the field, working in an alien

119 F-Jupp-3373.qxd 3/8/2006 5:44 PM Page 120

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

environment, and the extent of the strangeness Chicago School as iconically urban, gritty and experienced by the fieldworker will vary soaked in the dangerous mystique of the city according to the specific culture under study, as a social laboratory, often ignores other and the background of the fieldworker. more mundane complementary methodolo- Strangeness and distance are distinct features gies and concerns of Chicago’s sociologists. of nineteenth-century social anthropology, Yet although the Chicago School did indeed where the racial and colonial context within utilize a wide range of methodologies, their which researchers practiced made any shar- fieldwork-based studies have an enduring ing of perspective with the studied popula- quality that has inspired generations of field- tion at best extremely unlikely. With workers (see Deegan, 2001). fieldworkers working at such a distance from their informants, native cultures were usually Evaluation regarded as exotic, essentially inferior and in decline. The collection of data often consti- The close involvement of fieldworkers in the tuted the amassing of cultural artefacts, with lives of their subjects poses a myriad of ethi- the fieldworker remaining remote from the cal problems and in many examples of field- realities of the everyday lives of the natives work it is the researcher’s proximity to (Wax, 1972). problematic, ambiguous or outright criminal As anthropologists developed a more self- behaviour that constitutes the core of the conscious approach during the 1930s and issue (Hobbs, 2001). Although the roles 1940s, the degree of participation in the lives adopted by fieldworkers can vary enor- and cultures of the studied population mously, common to all is an academically increased. This process was paralleled in contrived attempt to advantageously situate sociology via the influence of the Chicago the researcher in order that he or she might School, which combined the theoretical address a social setting and the culture of rigours of European theory and classical those who naturally reside there. anthropology with a concern for engaging The shapes, patterns and trajectories of with the social problems of a rapidly evolving action observed and experienced in the field urban setting. Chicago’s fieldworkers, if they are recorded as field notes, which are usually were not already part of the culture they taken as soon as possible after the observed were studying (for instance, Nils Anderson action (Emerson et al., 2001). Field notes are used his former-membership of hobohemia the immediate products of fieldwork, and to study transient workers), became members consist of observations and reflections that and associates of groups that were then sub- are representative of time spent in the field. jected to close scrutiny filtered through the Consequently, field notes are selective experience of the fieldworker The Chicago accounts which, when accumulated over a fieldwork tradition, typified by Robert Park’s period of time, will form the core of an acad- exhortation to his students to ‘go get the seat emically framed account of a culture. It is via of your pants dirty in real research’, has field notes that fieldworkers will seek to con- proved to be as enduring as it is has been struct ‘thick description’ (Geertz, 1973), evocative. With echoes of Henry Mayhew’s which will enable cultural interpretation to journalistic journeys into nineteenth-century take place, and the methodologies applied by London, and the campaigning writing of fieldworkers have had to evolve with the set- Jacob Riis and Lincoln Steffens, fieldwork- tings they study, and new fields are con- based studies of hobos, dancehalls, street stantly emerging (Hess, 2001). gangs and slums laid the basis for a rich Fieldwork constitutes deliberate attempts research tradition that was later joined by at understanding a culture, and the utility of studies of drug users, jazz musicians and a fieldwork has enabled a window onto dis- multitude of occupations, social groups and parate social worlds that would otherwise cultures. The aura that surrounds the have remained hidden or obscured. The

120 F-Jupp-3373.qxd 3/8/2006 5:44 PM Page 121

FOCUS GROUP

evolution of fieldwork from an archaeological by the researcher. A focus group typically enterprise linked closely to the rigours of consists of a tape-recorded discussion among colonial rule, into an interpretive device six to eight participants who are interviewed where the fieldworker is an essential compo- by a moderator. nent in the action has been a richly produc- tive process for the social sciences. Distinctive Features As a method for collecting qualitative data, Dick Hobbs focus groups emphasize learning about the thoughts and experiences of others. When the Associated Concepts: access, cultural participants in a group interview share an research, ethics, ethnography, naturalistic interest in the discussion topic, their inter- data, participant observation, qualitative action can provide information about how research, thick description they relate to the topic and to each other. The group dynamics in these participant-centred Key Readings conversations allow researchers to hear how people explore the discussion topic. Hence, Deegan, M. (2001) ‘The Chicago School of the most effective focus groups consist of par- Ethnography’, in P. Atkinson, A. Coffey, ticipants who are just as interested in the S. Delamont, J. Lofland and L. Lofland topic as the researchers are, which helps to (eds), Handbook of Ethnography. London: produce a free-flowing exchange. Sage. pp. 11–25. Emerson, R., Fretz, R. and Shaw, L. (2001) When the participants are mutually inter- ‘Participant observation and fieldnotes’, ested in the discussion, their conversation in P. Atkinson, A. Coffey, S. Delamont, often takes the form of sharing and compar- J. Lofland and L. Lofland (eds), Handbook ing thoughts about the topic. That is, they of Ethnography. London: Sage. pp. 352–68. share their experiences and thoughts, while Geertz, C. (1973) The Interpretation of also comparing their own contributions to Cultures: Selected Essays. New York: Basic what others have said. This process of shar- Books. ing and comparing is especially useful for Hess, D. (2001) ‘Ethnography and the devel- hearing and understanding a range of opment of science and technology studies’, responses on a research topic. The best focus in P. Atkinson, A. Coffey, S. Delamont, groups thus not only provide data on what the J. Lofland and L. Lofland (eds), Handbook participants think but also explicit insights of Ethnography. London: Sage. pp. 246–57. into why they think the way they do. Hobbs, D. (2001) ‘Ethnography and the study One particularly powerful strategy in of deviance’, in P. Atkinson, A. Coffey, focus group research is to bring together par- S. Delamont, J. Lofland and L. Lofland ticipants with a common background with (eds) (2001) Handbook of Ethnography. regard to the discussion topic. This homoge- London: Sage. pp. 204–19. neous group composition makes it easier for Wax, M. (1972) ‘Tenting with Malinowski’, the participants to engage in sharing and American Sociological Review, 37: 1–13. comparing, especially on topics that may involve what others consider to be deviant behaviour. For example, a group of women who are commercial sex workers can discuss FOCUS GROUP aspects of their lives that they would ordinar- ily keep secret from most other people. Definition Because homogeneous focus groups can discuss topics that might be considered taboo in A method for collecting qualitative data other contexts, they are a popular method for through a group interview on a topic chosen research in areas such as substance abuse,

121 F-Jupp-3373.qxd 3/8/2006 5:44 PM Page 122

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

family violence and sexual behaviour. It is with questions that generate lively exchanges important to note, however, that this strength can provide information about a range of depends on careful recruitment procedures to experiences and opinions that might be diffi- ensure a group composition where the partic- cult to observe outside such a discussion. ipants share a common set of experiences or Focus groups also have a set of correspond- beliefs with regard to the discussion topic. ing weaknesses, which lead to situations where Another advantage of creating groups other methods of collecting data would be where the participants share similar interests preferable. Individual rather than group inter- or experiences is an increased ability to carry views would be preferable when there is a need on their own conversation, with less active for greater depth and detail about personal guidance from the moderator. Such groups experiences or beliefs, because one-on-one con- are frequently termed ‘less structured’ since versations allow more time to generate richer the participants are free to pursue their narratives. Collecting qualitative data through own interests in their own ways. This style participant observation would be preferred over of focus group is especially useful for focus groups when there is a need to under- exploratory research where the goal is to stand behaviour in context. In addition, focus learn the participants’ perspectives. It is less groups and other types of interviews provide useful, however, when the research team has only verbal and self-report data, so they are no a strong set of predetermined objectives. substitute for observing how people actually In that case, a ‘more structured’ approach behave in realistic settings. Finally, focus groups would emphasize the role of the moderator as typically follow other qualitative methods in a discussion leader, who would use a fixed set relying on small, purposefully chosen samples of research questions to guide the group’s that generate theoretical insights, which makes conversation. survey research preferable for studies that Taken together, these two dimensions of require generalizability. ‘group composition’ and ‘interview structure’ In just two decades, focus groups have generate a variety of research design options moved from being almost unknown in the for focus groups. This flexibility makes it pos- social sciences to become a popular method sible to use focus groups for a wide range of for collecting qualitative data. This wide- purposes throughout the social sciences. spread use clearly demonstrates their value. At the same time, however, this relative new- Evaluation ness also means that focus groups have substantial, unexplored potential. This suggests a Compared to other qualitative methods, it is future for focus groups that builds on well- the interaction around a predetermined topic established procedures at the same time as it that makes focus groups unique. Although uncovers new uses for this method. individual qualitative interviews also concentrate on well-defined topics, they do not pro- David L. Morgan vide the group interaction that is the source of data in focus groups. In particular, the process of sharing and comparing in focus Associated Concepts: ethnographic inter- groups often leads the participants them- views, marketing research, purposive sam- selves to explore the topic in ways that the pling, qualitative research researcher did not anticipate. Compared to participant observation as a means of collect- Key Readings ing qualitative data, focus groups have the advantage of providing concentrated observa- Barbour, R. S. and Kitzinger, J. (1999) tions on the topics that are of most interest to Developing Focus Group Research. the researcher. In particular, a focus group Thousand Oaks, CA: Sage.

122 F-Jupp-3373.qxd 3/8/2006 5:44 PM Page 123

FORECASTING

Krueger, R. A. and Casey, M. A. (2000) Focus remains the same, and it is in essence based Groups: A Practical Guide for Applied upon the scientific school of inquiry. Research, 3rd edn. Thousand Oaks, CA: Sage. The forecasting process is one of: Morgan, D. L. (1997) Focus Groups as Qualitative Research, 2nd edn. Volume 16 in (1) defining the problem; the Sage Publications series on Qualitative (2) reviewing the information on the phe- Research Methods, Thousand Oaks, CA: nomena to be looked at; Sage. (3) formulating a hypothesis, and research Morgan, D. L. and Krueger, R. A. (1998) The strategy (quantitative, qualitative or Focus Group Kit (6 vols). Thousand Oaks, both) in order to address the hypothesis; CA: Sage. (4) gathering data; (5) analysing the results; (6) taking a decision as to whether to accept or reject the hypothesis; (7) forecasting into the future as a result of FORECASTING this decision.

Definition Evaluation The process through which a forecast is derived. A forecast is the estimate of a value Does history repeat itself? Does the past impact or condition in a future time period, for on the future? To a certain extent the answer to example, next year’s level of inflation, these questions is yes. For example, the level expected demand for a product or service in of sales a company makes in previous time 2010, whether a patient will develop chronic periods will have an effect on the demand for heart disease, students’ examination results its products. This may be due to spares and for the next semester. replacements to maintain the product, high levels of customer satisfaction with the prod- Distinctive Features uct, brand loyalty and other market features such as monopoly of supply. There are events, All forecasts, and thereby forecasting, work however, which may cause a ‘blip’ in history. on the basis that history will repeat itself, and These blips cannot be planned for, and do not that it will continue to do so. In addition to occur with any degree of regularity. For exam- this, forecasting is also involved with looking ple, the attack on the World Trade Centre had at the cause and effect relationship that may a devastating effect on the airline industry. Any exist between the independent variables and previous forecasts as to the number of aero- the dependent variable. Multiple regression, planes required to meet passenger demand had time series analysis econometrics and opera- to be seriously downgraded. Indeed a number tional research methods are examples of of airline companies went into liquidation due some of the statistical techniques available. to existing market conditions and the general Qualitative methods such as focus groups, public’s decision not to fly. and the Delhi experiment might also be used However, rail and coach business within to produce forecasts. the United States increased considerably as a Forecasting can range from the very simple direct result. Their forecasts of demand had to the very complex; it depends upon what to be revised significantly upwards. question you are attempting to answer. There A good deal of care and attention must be are many ways to derive a forecast, none is given when establishing a cause and effect superior to any another, and it is a matter of model. Social phenomena are particularly getting the right tool for the right job. complex and make forecasting all that more dif- However, the process involved in forecasting ficult. This problem is made more complex by

123 F-Jupp-3373.qxd 3/8/2006 5:44 PM Page 124

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

the fact that most of the statistical techniques Key Readings used assume linear relationships between the DeLurgio, S. A. (1998) Forecasting Principles variables. In reality this is not necessarily so. and Applications. New York: Irwin/ McGraw–Hill. Paul E. Pye Makridakis, S. (1998) Forecasting: Methods and Applications, 3rd edn. New York: Wiley. Associated Concepts: causal model, econo- Wright, G. and Goodwin, P. (1998) Forecasting with Judgment metrics, focus group, hypothesis, modelling, . New York: Wiley. Yaffee, R. A. (2000) Introduction to Time multiple regression, operational research Series Analysis and Forecasting. London: management, prediction, SPSS, time series Academic Press. design

124 G-Jupp-3373.qxd 3/8/2006 10:42 AM Page 125

Evaluation GARBOLOGY The use of such trace measures is valuable in Definition studying sensitive topics, such as excessive alcohol consumption, where reactivity is In the narrow sense, garbology is the study of likely to be high. It is also a quick and cheap the contents of garbage cans (or dustbins) in way of collecting data. However, there is the order to make some assertions about the problem of making valid inferences about lifestyle of those who deposit garbage of dif- forms of behaviour from an examination of ferent kinds. In a wider sense, it is the study waste or other traces. For example, vodka of all materials discarded by individuals or bottles found in a dustbin may have been groups, or within communities and societies deposited by alcoholics drinking in a nearby in order to uncover social and cultural forms. park and not by the resident to whom the dustbin belongs. Garbology is an intriguing Distinctive Features and innovative form of research but is unlikely to generate measures that are suffi- The study of waste products has roots in ciently strong by themselves. It is perhaps archaeology but more recently it has become best used in triangulation with other methods associated with those who are concerned with of research. ecological and ‘green’ issues. In the social sciences it has not developed a strong presence. However, where it has been employed it has Victor Jupp been linked to the arguments for unobtrusive, trace or non-reactive measures. The problem Associated Concepts: covert research, of reactivity, for example in asking individuals trace measures, triangulation, unobtrusive questions or observing them, is that those indi- measures, validity of measurement viduals may respond or act in particular ways as a result of being studied. They may manage the impressions they give or respond in what they believe to be socially desirable ways. So, Key Readings for example, in a questionnaire study of drink- Garwood, J., Rogerson, M. and Pease, K. ing habits it is likely that people will under- (2002) ‘Sneaky measurement of crime state the amount they consume, especially if and disorder’, in V. Jupp, P. Davies and they are heavy drinkers. A non-reactive way of P. Francis (eds), Doing Criminological collecting data could involve counting the Research. London: Sage. pp. 157–67. number of empty bottles in their garbage cans Webb, E. J., Campbell, D. T., Schwartz, R. or examining discarded supermarket till D. and Sechrest, L. (2000) Unobtrusive receipts to see how much alcohol they buy. Measures, rev. edn. London: Sage. G-Jupp-3373.qxd 3/8/2006 10:42 AM Page 126

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

Even if they are convinced of this, they may GATEKEEPER not be willing to engage in such additional and voluntary activities due to the time Definition and resources involved, the sensitivity of the The person who controls research access. For research topic or because of concerns about example, the top manager or senior executive confidentiality. Gatekeepers often have dif- in an organization, or the person within a ferent views to researchers regarding the group or community who makes the final deci- amount of time for which it is reasonable for sion as to whether to allow the researcher their colleagues to be involved in research, access to undertake the research. those issues that are sensitive, as well as issues of confidentiality. It is therefore impor- Distinctive Features tant that these are recognized and taken into consideration when negotiating access Gaining access to undertake social research is (Saunders et al., 2003). often problematic. Friends, contacts and col- In order to help secure a gatekeeper’s sup- leagues and others may be willing to vouch port, the researcher may offer something in for a researcher and the value of the research return for granting access. Within formal and act as research sponsors. However, organizations this is commonplace, often unless permission has been granted by a gate- taking the form of a summary report of the keeper from within the group, community or research findings or an opportunity to influ- organization in which it is planned to under- ence the research to ensure that it incor- take the research, it is unlikely that access porates issues that are pertinent to the will be allowed in practice. organization (Buchanan et al., 1988). For less Gatekeepers for research in formal organi- formal groups, such bargains are more varied zations such as firms, schools, political parties due to the diverse and often informal nature and the like are normally senior managers or of these research settings. It should be noted executives within that organization’s hierarchy. that some writers on research methods do not In addition, employee representatives such as recommend this approach to securing gate- trade union officials may also act as gatekeep- keeper support (Bryman, 2001). ers within firms. The seniority of such people means that they control both the researcher’s Mark N. K. Saunders physical access to the organization and influence the degree of support the researcher is given subsequently by others within that orga- Associated Concepts: access, confidential- nization (Saunders et al., 2003). ity, covert research, ethics, organizational It is more difficult to ascertain who the gate- research, research bargain keepers are for research in other settings, such as those involving communities, gangs and less Key Readings formal groups. Such people often take on the dual role of research sponsor and gatekeeper, Bryman, A. (2001) Social Research Methods. both vouching for the researcher and the value Oxford: Oxford University Press. of their research and controlling access to the Buchanan, D., Boddy, D. and McCalman, J. group (Bryman, 2001). A key characteristic of (1988) ‘Getting in, getting on, getting out, such gatekeepers is the high degree of respect and getting back’, in A. Bryman (ed.), they command within their group. Doing Research in Organisations. London: Routledge. pp. 53–67. Evaluation Saunders, M. N. K., Lewis, P. and Thornhill, A. (2003) Research Methods for Business Before allowing access, gatekeepers need to Students, 3rd edn. Harlow: Financial be convinced of the value of the research. Times/Prentice Hall.

126 G-Jupp-3373.qxd 3/8/2006 10:42 AM Page 127

GENERAL LINEAR MODELLING

power. However, there are ways around the GENERAL LINEAR MODELLING variable requirement. The GLM described by Equation (2) is linear with respect to both Definition parameters and variables. The implementation of a set of procedures =β +β ε that attempts to accommodate data on a ran- Yi 0 1Xi + i (2) dom dependent variable in terms of a linear combination of fixed and/or random variables However, the equation (the model) plus a random variable (the error). =β +β2 ε Yi 0 1 Xi + i (3) Distinctive Features is linear with respect to the variables, but β The general linear modelling (GLM) concep- not to the parameters, as 1 has been raised to tion is that data may be accommodated in the second power. Linearity with respect to terms of a model plus error, as illustrated the parameters also would be violated if any below, parameters were multiplied or divided by other parameters or appeared as exponents. data = model + error (1) In contrast, the equation

=β +β 2 ε The model Equation 1 is a representation of Yi 0 1X i + i (4) our understanding or hypotheses about the data. The error component is an explicit is linear with respect to the parameters, but not 2 recognition that there are other influences on with respect to the variables, as X i is Xi raised the data. These influences are presumed to to the second power. However, it is very simple = 2 be unique for each subject in each experi- to define Zi Xi and to substitute Zi in place 2 mental condition and include anything and of Xi. Therefore, models such as described everything not controlled in the experiment by Equation (4) continue to be termed linear, or study. The relative size of the model and whereas such as those described by Equation error components is used to judge how well (3) do not. In short, linearity is presumed to the model accommodates the data. apply only to the parameters. Models that are The term general in GLM simply refers not linear with respect to their parameters are to the ability to accommodate variables that described specifically as nonlinear. As a result, represent both quantitative distinctions that models can be assumed to be linear, unless represent continuous measures, as in regres- specified otherwise, and frequently the term sion analysis, and categorical distinctions linear is omitted. that represent experimental conditions, as in Traditionally, the label linear modelling ANOVA. This feature is emphasized in was applied exclusively to regression analy- ANCOVA, where variables representing both ses. However, as regression, ANOVA and quantitative and categorical distinctions are ANCOVA are particular instances of the employed in the same GLM. GLM, it should be no surprise that consider- GLMs can be described as being linear ation of the processes involved in applying with respect to both their parameters and these techniques reveals any differences to predictor variables. Linear in the parameters be more apparent than real. McCullagh and means no parameter is multiplied or divided Nelder (1989) distinguish four processes in by another, nor is any parameter above the linear modelling: (1) model selection, (2) para- first power. Linear in the predictor variables meter estimation, (3) model checking and also means no variable is multiplied or (4) the prediction of future values. While such divided by another, nor is any above the first a framework is useful heuristically, McCullagh

127 G-Jupp-3373.qxd 3/8/2006 10:42 AM Page 128

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

and Nelder acknowledge that in reality these stepping and ridge regression (Neter four linear modelling processes are not so et al., 1990) have been developed to reduce distinct and that the whole, or parts, of the the number of GLMs that need to be consid- sequence may be iterated before a model ered. After model selection, parameters must finally is selected and summarized. be estimated for each GLM and then model Usually, prediction is understood as the checking engaged. Again, due to the nature forecast of new, or independent, values with of non-experimental data, model checking respect to a new data sample using the GLM is likely to detect more problems, which will already selected. However, McCullagh and require remedial measures. Finally, the nature Nelder also include a more general account of the issues addressed by non-experimental of prediction, where the values fitted by the research make it much more likely that the GLM (graphically, the values intersected by GLMs selected will be used to forecast new the GLM line or hyperplane) are taken as values. instances of prediction and part of the GLM In contrast, a seemingly concise analysis summary. As these fitted values are often of experimental data occurs. However, con- called predicted values, the distinction sideration reveals identical GLM processes between the types of predicted value is not underlying a typical analysis of experimental always obvious, although there is greater data. For experimental data, the GLM standard error associated with the values selected is an expression of the experimental forecast on the basis of a new data sample design. Moreover, most experiments are (for example, Neter et al., 1990; Pedhazur, designed so that the independent variables 1997). translate into independent (that is, uncorre- With the linear modelling process of pre- lated) predictors, so avoiding multi-colinearity diction so defined, the four linear modelling problems. The model checking process con- processes become even more recursive. For tinues by assessing the predictive utility of example, when selecting a GLM, usually the the GLM components representing the exper- aim is to provide a best fit to the data with the imental effects. Each significance test of an least number of predictor variables (for exam- experimental effect requires an estimate ple, McCullagh and Nelder, 1989). However, of that experimental effect and an estimate of the model checking process that assesses best a pertinent error term. Therefore, the GLM fit employs estimates of parameters (and esti- process of parameter estimation is engaged mates of error), so the processes of parameter to determine experimental effects, and as estimation and prediction must be executed errors represent the mismatch between the within the process of model checking. predicted and the actual data values, the cal- The misconception that this description of culation of error terms also engages the linear general linear modelling refers only to regres- modelling process of prediction. Conse- sion analysis is fostered by the effort invested quently, all four GLM processes are involved in the model selection process with correla- in the typical analysis of experimental data. tional data obtained from non-experimental The impression of concise experimental studies. Usually in non-experimental studies, analyses is a consequence of the experimen- many variables are recorded and the aim is to tal design acting to simplify the process of identify the GLM which best predicts the GLM selection and only attending to certain dependent variable. In principle, the only model checks. sure way to select the best GLM is to examine every possible combination of predictors. As it takes relatively few potential predictors Andrew Rutherford to create an extremely large number of possible GLM selections, a number of predictor variable Associated Concepts: causal model, econo- selection procedures, such as all-possible- metrics, forecasting, multivariate analysis, regressions, forward stepping, backward path analysis, prediction, regression analysis

128 G-Jupp-3373.qxd 3/8/2006 10:42 AM Page 129

GIS (GEOGRAPHICAL INFORMATION SYSTEMS)

Key Readings mapping has excellent descriptive benefits through pattern generation, the power of GIS Cohen, J., Cohen, P., West, S. G. and Aiken, lies in two attributes; first, its ability to process L. S. (2003) Applied Multiple Regression/ and manipulate large volumes of data that Correlation Analysis for the Behavioral would be difficult or impossible using manual Sciences, 3rd edn. Mahwah, NJ: LEA. McCullagh, P. and Nelder, J. A. (1989) techniques (Heywood et al., 1998), and second, Generalised Linear Models, 2nd edn. its capability for spatial analysis involving com- London: Chapman & Hall. plex interrogation of relationships between spa- Neter, J., Wasserman, W. and Kutner, M. H. tial phenomena (Demers, 2000; Schuurman, (1990) Applied Linear Statistical Models: 2004). Indeed, as Schuurman (2004) points out, Regression, Analysis of Variance and GIS can generate information or knowledge by Experimental Designs, 3rd edn. Homewood, the visualization of both clear and precise and IL: Irwin. imprecise data thus allowing room for intuition Pedhazur, E. J. (1997) Multiple Regression in in analysis and thereby uniting the science and Behavioural Research, 3rd edn. Fort Worth, social process. TX: Harcourt Brace. Although GIS has its roots in geography, Rutherford, A. (2001) Introducing ANOVA and having developed from cartography, it is now ANCOVA: A GLM Approach. London: Sage. ubiquitous across a range of disciplines and per- vades many aspects of modern life from business to medicine, archaeology to policing. Bernhardsen (2002) notes that its major benefit GIS (GEOGRAPHICAL INFORMATION is its ability to link information from widely scattered sources, such as data obtained from SYSTEMS) maps, space satellite images, photographic and video information taken from low flying air- Definition craft, statistical data and even data electroni- cally transmitted via the Internet and other Generic computer-based facilities comprising electronic sources. Questions that may be posed hardware and software that process spatial of GIS include ‘the particulars of a given loca- data into digital information concerning, and tion, the distribution of selected phenomena, used to make decisions about, a particular the changes that have occurred since a previous geographical location. Processes include data analysis, the impact of a specific event, or the collection and inputting, the transformation relationships and systematic patterns of a of different types of spatial information, data region’ (Bernhardsen, 2002: 225). storage, organization and retrieval, data manipulation and analysis and data presentation in tabular, graphic or map form. Evaluation

Distinctive Features It is important to remember that GIS can only ever represent a simplified view of the piece of The underlying methodology of GIS comes world that is its subject matter and a particular from overlay mapping, which is the process of application will generally only reflect the spe- building data in a series of maps, each repre- cific purposes or research motives of its users. senting a specific type or theme of information, Definitions for GIS analysis of spatial phenom- which may be, for example, spatial, such as ena can, for example, vary according to the streets or trees, or statistical, such as census socio-political agenda of those commissioning data. The maps are then placed over one or carrying out the research (Schuurman, 2004). another so that geographical coordinates align Furthermore, as with any technology, what to present a composite picture of a particular comes out of the system is only as good as what geographical location and the interrelated is put in and therefore the quality and accuracy characteristics that pertain to it. However, while of data used is paramount. While the power of

129 G-Jupp-3373.qxd 3/8/2006 10:42 AM Page 130

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

GIS lies in its capability to represent spatial and Distinctive Features object relationships visually in ways that allow Studies of good practice are a form of policy patterns to be interpreted, the outcome is usu- research, which is characterized by the pursuit ally to establish cause and effect and this can be of knowledge in order to facilitate action particularly problematic where cause and effect (Hakim, 2000). Research to identify good prac- may be subject to the influence of phenomena tice can range from highly systematic reviews that have not featured in the data gathering of the effectiveness of particular procedures, process. for example in health services through well- It is also important to note that while GIS documented case study materials, and audit is valuable in its capacity to generate visual- and inspection reports, to ad hoc examples pro- izations of spatial relationships in a way that vided by the organizations themselves. The allows users to interpret patterns, this should identification of good practice may form part of not obviate the contribution of other method- a wider strategy of performance improvement ologies, such as local knowledge, in discerning such as benchmarking (for example the Best patterns when used in conjunction with visual Practice Benchmarking recommended to com- mapping (Schuurman, 2004). Depending on panies in England by the Department of Trade its application, GIS is best seen therefore as a and Industry). useful complementary tool alongside other A key feature of studies of good practice is forms of empirical research. a particular emphasis on the need for effective dissemination. A study among local pol- Graham Steventon icy makers in the United Kingdom (Wolman and Page, 2000) found that nearly all those Associated Concepts: causation, demogra- surveyed stressed the value and importance phy, disaster research, nomothetic of learning from the experience of others. But it was also acknowledged that generally there Key Readings was not a great deal done to ensure that such information was obtained and used. Bernhardsen, T. (2002) Geographic Information Even in those areas where there is a tradi- Systems: An Introduction, 3rd edn. New tion of rigorous scientific research, it is naïve York: Wiley. to assume that when research information is Demers, M. N. (2000) Fundamentals of made available it is somehow accessed by Geographic Information Systems, 2nd edn. practitioners, appraised and then applied in New York: Wiley. practice. There is a need to understand the Heywood, I., Cornelius, S. and Carver, S. barriers to the adoption of new practices and (1998) An Introduction to Geographical the necessity to develop strategies to bring Information Systems. Harlow: Pearson. about changes in behaviour. Schuurman, N. (2004) GIS A Short More and more, those responsible for dis- Introduction. Malden, MA: Blackwell. seminating good practice are devoting considerable resources and energy to developing more effective ways of ensuring that the findings of such work reach the target audience, GOOD PRACTICE STUDIES and are acted upon. The dissemination process is often IT-based. Definition Evaluation Studies of good (or ‘best’) practice aim to identify examples of successful activities and Because there is such a wide variability in the operations in organizations that can be adopted type of research that generates evidence of by others to improve their performance. good practice, it is not possible to take a

130 G-Jupp-3373.qxd 3/8/2006 10:42 AM Page 131

GROUNDED THEORY

single view on the value of such studies. Associated Concepts: applied research, audit- There has, in the past few years, been a sub- ing, case study method, cost–benefit analysis, stantial interest in evidence-based policy, that evaluation research, messy research, policy- is, drawing on the findings of good quality related research, process mapping research to inform the development of public policy (Davies et al., 2000). In the UK organi- Key Readings zations such as the National Health Service Centre for Reviews and Dissemination, and Davies, H. T., Nutley, S. M. and Smith, P. C. the Evidence for Policy and Practice (eds) (2000) What Works? Evidence-based Information and Co-ordinating Centre (EPPI Policy and Practice in Public Services. Centre), by carrying out systematic reviews Bristol: The Policy Press. of the quality of the research involved, pro- Hakim, C. (2000) Research Design. London: Routledge. vide reassurance that the practice that is Wolman, H. and Page, E. C. (2000) Learning being disseminated is well founded. from the Experience of Others: Policy Similarly, case studies of good practice are Transfer among Local Regeneration often produced as one of the outputs of well- Partnerships. York: Joseph Rowntree designed evaluation studies, which aim to Foundation. assess the extent to which a policy or programme achieves its objectives. Or, they will be documented in an inspection or audit process, whereby they are assessed in terms of agreed criteria that define the required GROUNDED THEORY attributes. In such circumstances, the case studies should provide reliable examples of Definition the desired practice. In other instances, where the method of Grounded theory is an approach to research study is not so rigorous, it will not be possible that was developed in response to concerns to be as sure that the recommended practice is over the predominance of quantitative meth- indeed all that it is claimed to be. It may be that ods in social sciences and the tendency for the activity is of value, or can provide pointers research to be undertaken to test existing to effective action, but the basis for the recom- grand theories. Glaser and Strauss (1967: mendation may not be very secure. p. vii) perceived that there was an ‘embar- This points us to a further cautionary note. rassing gap between theory and empirical Even if the recommended practice is well research’. They proposed instead an inductive founded it may not necessarily be applicable process in which theory is built and modified in all circumstances. Extraneous factors (for from the data collected. example, substantial differences in the range of problems clients present to a social ser- Distinctive Features vices agency) may mean that the organization adopting the recommended practice will not The interaction between data analysis, theory achieve the same results as that which pro- building and sampling is central to the devel- vides the exemplar. opment of grounded theory. As soon as sufficient data has been collected to begin Note analysis, it is used to generate categories. Glaser and Strauss discuss the constant com- The above material does not represent the parative method, whereby sections of the views of the Department of Education and data are continually compared with each Skills, UK. other to allow categories to emerge and for relationships between these categories to Vince Keddie become apparent.

131 G-Jupp-3373.qxd 3/8/2006 10:42 AM Page 132

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

Concrete categories are then modified into a danger that an approach that is described as more abstract concepts. Theory building grounded theory will simply become a form involves arranging these concepts into a logi- of data analysis, with no real theoretical con- cal scheme, which can be modified as further tent. It is important, therefore, that data are data are collected. The researcher is always used to guide theorizing but not to place ready to add new concepts as they emerge limits on it. from the data. Data continue to be collected Grounded theory is sometimes assumed in relation to a category until it is ‘saturated’, to mean that the researcher approaches that is, new data coming in are not contribut- the topic without any preconceived concep- ing anything new to the analysis. tual framework and that the framework The method by which cases are selected is formed entirely by the data. In reality, for analysis is referred to as theoretical sam- such an approach seems impossible – to pling. The aim is not to achieve a representa- choose to research a topic at all, a researcher tive sample: cases are chosen as it becomes must have some prior knowledge and/or clear that they can contribute to theory gen- preconceived ideas about it. A more realistic eration. Negative case analysis is an impor- approach is to acknowledge that a researcher’s tant element of the sampling strategy: cases perspective can shape their inquiry without that do not fit the emerging theory should be this perspective simply being applied to the examined in order that the theory can be data. modified appropriately. The process of searching for falsifying evidence, and modi- fying theory until no further disconfirming Jamie Harding evidence can be found, is referred to as analytic induction. Associated Concepts: analytic induction, The type of theory generated by this constant comparative method, induction, process tends to be middle range, that is, methodology, theoretical sampling somewhere between grand theory and a working hypothesis. Existing theory is not discounted when using a grounded theory Key Readings approach, but is considered after data analy- Glaser, B. G. and Strauss, A. L. (1967) The sis rather than before. When data analysis is Discovery of Grounded Theory: Strategies complete, the researcher examines a number for Qualitative Research. New York: Aldine of existing theories to establish which fits best de Gruyter. with the grounded theory that has been gen- Pidgeon, N. (1996) ‘Grounded theory: theo- erated. The existing literature on the subject – retical background’, in J. T. E. Richardson particularly unpublished documents – may (ed.), Handbook of Qualitative Research become part of the analysis, rather than pre- Methods. Leicester: British Psychological determining the nature of the research. Society. Pidgeon, N. and Henwood, K. (1996) Evaluation ‘Grounded theory: practical implications’, in J. T. E. Richardson (ed.), Handbook of As it is often difficult to theorize beyond the Qualitative Research Methods. Leicester: immediate phenomena being studied, there is British Psychological Society.

132 H-Jupp-3373.qxd 3/8/2006 10:42 AM Page 133

assumed to explain physical phenomena: by HERMENEUTICS describing their external characteristics and seeking to identify empirical regularities in Definition their behaviour. Rather, it was argued, what The study of how we understand the commu- people say and do can only be understood nications, actions and products of other ‘from the inside’ – against the background of human beings – especially those of past times how those people think and feel about the or other cultures. But it also implies a partic- world in which they live. Indeed, it was ular set of views about what such under- sometimes suggested that this provides a standing involves, one that stresses the role of deeper kind of knowledge than that available inner life experience, culture and/or imagina- to natural scientists. tion on the part of the interpreter – as against Dilthey is the most important nineteenth- attempts to found the historical and social century theorist of ‘understanding’ or Verstehen. sciences on a method that begins from, or tests He believed there was a need for a philo- hypotheses against, description of the external sophical psychology, or hermeneutics, which features and/or behaviour of human beings. would ground the methodology of the social and historical sciences; and he set out to pro- Distinctive Features vide it. His earliest formulations portrayed Verstehen as a psychological process whereby Originally this term referred to the interpre- historians project themselves imaginatively tation of ancient texts, especially of obscure into the lives of people in the past, ‘re-creating’ or ambiguous passages in these. Under the their experience, so as to capture the mean- influence of the Renaissance, and the revolu- ings embodied in texts and artefacts. His later tion in historiography which it stimulated, account portrayed it as cultural interpreta- there was recognition that in order to under- tion, rather than as a subjective process. It stand primary historical texts it is necessary involves locating what is to be understood in to take account of the distinctiveness of the terms of objective cultural meanings that cultures in which they were produced. This constitute the context in which it occurs. Part emphasis on cultural discontinuity, on the of Dilthey’s concern here was to recognize dangers of misinterpreting others by viewing historical diversity while yet resisting episte- them as if they were members of one’s own mological relativism. And his ideas have community, came to be central to much had considerable influence within the social thinking about the nature of historical and sciences, notably through Max Weber’s social inquiry in the nineteenth century, verstehende sociology. especially in Germany. It was argued that In the twentieth century there was an social phenomena could not be approached important shift in orientation by some of in the same way that natural scientists were those writing about hermeneutics, especially H-Jupp-3373.qxd 3/8/2006 10:42 AM Page 134

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

H.-G. Gadamer. He rejected the idea that the at best, Verstehen can only supply hypotheses past could ever be understood ‘as it was’, in for testing, since it does not involve any rig- the way that Dilthey and others assumed. He orous form of validation. Subsequently, some insisted that what is unknown can never be versions of structuralism and post-structuralism simply grasped or reproduced, since any have questioned the very possibility of under- interpretation necessarily draws on the par- standing. Indeed, for some there is an obliga- ticular socio-historical resources available to tion to challenge any claim that understanding the interpreter. There is no way in which we has been or could be achieved: to celebrate can escape reliance on these, because without or at least highlight the endless deferral of them there could be no understanding at all. meaning or the incommensurability of cul- Where Dilthey saw Verstehen as a method tural forms. that is essential to the scientific character of social research, Gadamer treated it instead as a fundamental aspect of human-being-in-the- Martyn Hammersley world. And, from this point of view, it does not differ from lay forms of understanding; Associated Concepts: causality, covering- nor does it have the procedural, fixed charac- law model of explanation, cultural research, ter of a ‘method’. epistemology, historical analysis, methodol- For advocates of this ‘philosophical herme- ogy, positivism, reflexivity, Verstehen neutics’, interpretation is always necessarily a matter of dialogue. What is involved is the Key Readings interpreter beginning from assumptions that are a product of his or her own socio-cultural Caputo, J. D. (1987) Radical Hermeneutics: location, while at the same time opening Repetition, Deconstruction and the Herme- these up to challenge in the course of inter- neutic Project. Bloomington, IN: Indiana pretation. Moreover, for Gadamer, the purpose University Press. is not simply to provide knowledge of other Dostal, R. J. (ed.) (2002) The Cambridge cultures but rather to advance our under- Companion to Gadamer. Cambridge: standing of issues that are of universal con- Cambridge University Press. cern, whether to do with ethics, politics, or Makkreel, R. A. (1975) Dilthey: Philosopher philosophy. of the Human Studies. Princeton, NJ: The hermeneutic tradition carries implica- Princeton University Press. tions both for what sorts of knowledge can be Palmer, R. E. (1969) Hermeneutics. Evanston, IL: Northwestern University Press. produced by social researchers and for what Truzzi, M. (1974) Verstehen: Subjective they ought to aim at. It raises questions about Understanding in the Social Sciences. any attempt to understand human behaviour Reading, MA: Addison–Wesley. in causal terms, at least as modelled on causal- Weinsheimer, J. (1985) Gadamer’s Herme- ity in the physical realm, and thereby about neutics: a Reading of ‘Truth and Method’. experimental method as the methodological New Haven, CT: Yale University Press. ideal. It emphasizes the importance of openness in seeking to understand other people’s beliefs and actions, and therefore distrusts those approaches that employ highly structured elicitation devices, whether question- HISTORICAL ANALYSIS naires or systematic observation schedules. Definition Evaluation A method that seeks to make sense of the past Hermeneutics has been subjected to consid- through the disciplined and systematic analysis erable criticism. Positivists have argued that, of the ‘traces’ it leaves behind. Such traces

134 H-Jupp-3373.qxd 3/8/2006 10:42 AM Page 135

HISTORICAL ANALYSIS

may be of many different kinds, ranging from than to the present. Where data that purport everyday ephemera, artefacts and visual to represent a particular time period do not images, to old buildings, archaeological sites themselves come from that time, then the or entire landscapes. The most widely used consequence is likely to be anachronistic historical traces, however, are written docu- rather than accurate historical analysis. The ments, whether of public or private origin. requirement for historical accuracy demands that historical sources should be ‘found’ by Distinctive Features researchers rather than ‘produced’ by them (in the way, for example, of data derived from Historical analysis is commonly used in social interviews, social surveys or observation). research as an introductory strategy for estab- Claims for the reliability of historical sources lishing a context or background against which depend on the fact that they are preserved in a substantive contemporary study may be set. open archives as unmodified traces and are A stronger conception of historical analysis available to multiple successive readings sees it as a pervasive and necessary technique which are therefore themselves open to confir- in its own right, without which no account of mation or challenge. phenomena in the present may be properly understood. In this more substantial form, Evaluation historical analysis is often combined with other methods to engage social research questions. The emphasis on contemporary sources History has the power to challenge domi- points to more general problems upon which nant assumptions because it records, if in all those undertaking historical analysis will much less detail, the activities of the over- need to take a position. These derive from the looked and the marginalized as well as elites. temporal divide between the past (the site com- For this reason, historical analysis has proved prising the object of study) and the present (the particularly valuable to those researching site occupied by the researcher). To what gender, ‘race’ and other oppressed groups, and degree can it be possible to understand life in for those interested in developing alternative the past from the perspective of the present? models of social change. The charting and To the extent that all individuals (including assessment of currents of continuity, as well as social researchers) are shaped by the social change, is a major concern for the historical and intellectual contexts of their own time, method, alongside other key conceptualiza- can researchers divest themselves of the lim- tions such as chronology, periodization, causa- iting biases and preconceptions of the present tion, context and influence. Historical research in order to engage the past openly? If such may be oriented towards thematic or compar- limitations cannot be overcome, does this ative questions and is also often concerned mean that historical analysis must rest upon with some of the same debates about the myth, prejudice and misrepresentation? relationship between structure and agency Historical researchers are divided in their which have characterized the social sciences. answers, which in each case offer different Crucially, however, historical analysis has gen- conceptualizations of the relation between erally defined itself in terms of its essentially the past and the present. One group argues diachronic concerns and its complex narrative that the influence of the present may be form, thereby distinguishing itself from the nullified by trained historians who possess generalizing ambitions of social theory. the analytical skills effectively to reconstruct The validity of historical data depends on the past. Such scholars adhere to the princi- the key requirement that such data must have ple of scientific historicism which requires been produced within the particular time that researchers immerse themselves in all of period that is being studied. They must, in the relevant historical data available to them, other words, be contemporaneous with that labouring day after day in archives and record period and continue to ‘belong’ to it, rather offices. In so doing, they seek to release the

135 H-Jupp-3373.qxd 3/8/2006 10:42 AM Page 136

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

true record of the past which is perceived as research, life history interviewing, oral history, immanent in the sources. By consulting all phenomenology, postmodernism, validity such data exhaustively, the claim is that the researcher will be able to transcend the constraints of the present and to know the past Key Readings ‘as it really was’. Tosh, J. (2000) The Pursuit of History, 3rd A second group holds to the same ambition edn. London: Longman. pp. 55–70. for achieving historical truth as the first. In Tuchman, G. (1998) ‘Historical social science: this case, however, there is a stronger recog- methodologies, methods and meanings’, nition of the temporal distance between past in Norman K. Denzin and Yvonna S. and present, between object and subject. It is Lincoln (eds), Strategies of Qualitative not difficult, for example, to show that the Inquiry. London: Sage. pp. 225–60. events of the same historical period may be quite differently represented in the analyses of historical researchers from different generations. For this approach, therefore, the present and the past can never be finally rec- HUNTING onciled. In consequence, the results of historical analysis are seen to have their truth (SNOOPING AND FISHING) claims about the past inescapably qualified by the exercise of interpretation on the part of Definition the researcher in the present. The act of inter- A form of statistical analysis that involves pretation itself is therefore raised to a cardinal searching through a data set to identify rela- position in historical epistemology. tionships between variables which are worthy A third, more recent group, drawing upon of further investigation. the assertions of postmodernism and post- structuralism, radically reverses the epistemological flow between past and present in Distinctive Features arguing that ‘all history is the history of the Selvin and Stuart (1966) provide a classifica- present’. Here, because it is claimed that there tion of strategies of survey data analysis can be no certain truths about a past which which has been influential in guiding social has vanished, historical analysis may be trans- researchers. The classification highlights dif- figured into an entirely discursive or rhetori- ferent types of interchange between research cal exercise. This approach sees little point in hypotheses and data analysis. ‘Snooping’ is the painstaking archival ‘fact-grubbing’ of the process of using a set of data to test a pre- traditional historicism and turns instead to designated set of hypotheses. ‘Fishing’ is the constructing expansive analyses based upon process of dredging data, usually using very small numbers of isolated, contradictory computer software such as SPSS, in order to or obscure historical traces. Such apparently decide which variables to include in a causal cavalier use of sources is deprecated by more model on the grounds that each accounts for traditional historical practitioners, evoking their a worthwhile proportion of the variance in a earlier critiques of social science history as dependent variable. These contrast with little more than ‘raiding the past’. ‘hunting’ which is a much more flexible strategy aimed at generating hypotheses from the data which can be subsequently tested. Philip Gardner Evaluation Associated Concepts: anti-racism research, critical research, diary, discourse analysis, Such a data-dredging strategy can be a useful document analysis, epistemology, feminist source of research hypotheses about the

136 H-Jupp-3373.qxd 3/8/2006 10:42 AM Page 137

HYPOTHESIS

relationship between variables, especially as a further assumption that social science should the full data set can be utilized and unexpected proceed by seeking to establish evidence of relationships may be indicated. However, statis- relationships between variables. This is tically, it is unwise to test an hypothesis on the sometimes referred to as variable analysis. same data that generated it. Hypotheses gen- There can be variations in the specificity erated by one set of data should be tested for and breadth of an hypothesis. Specificity significance on another set. What is more, it relates to the nature of the relationship should be recognized that where two or more between variables which is being hypothe- variables are found to be statistically related it sized; breadth relates to the coverage which does not necessarily indicate that they are the- is being hypothesized. With regard to speci- oretically related. Data dredging, hypothesis ficity, an hypothesis may simply assert that a formulation and testing should be accompa- relationship is one of symmetrical covari- nied by meaningful theorizing. ance; that is, changes in the values of one variable are associated with changes in the values of another, without making any sug- Victor Jupp gestion as to causal influence or causal direction. The assertion that there is a relationship Associated Concepts: causal model, correla- between crime levels and unemployment tion, hypothesis, hypothetico-deductive model, levels is an example of such an hypothesis. multivariate analysis, SPSS, variable analysis Being more specific, one may postulate the nature of the relationship in terms of values of the variables. A statement which hypothe- Key Readings sizes that there is a positive relationship Byrne, D. (2002) Interpreting Quantitative between the variables is suggesting that high Data. London: Sage. levels of crime are associated with high levels Field, A. (2000) Discovering Statistics Using of unemployment; and conversely hypothesiz- SPSS for Windows. London: Sage. ing about a negative relationship suggests low Selvin, H. C. and Stuart, A. (1966) ‘Data values on one variable as being associated dredging procedures in survey analysis’, with low values on the other. An inverse rela- The American Statistician, 20: 20–3. tionship infers that high values on one are associated with low values on another. Some hypotheses go further by specifying the extent to which changes in one variable HYPOTHESIS are associated with changes in another (whether positive, negative or inverse). Further, where the assertion is that the Definition changes in one variable, say crime levels, are An untested assertion about the relationship produced by changes in another variable, say between two or more variables. The validity unemployment levels, the hypothesis is of such an assertion is assessed by examining suggesting an asymmetrical, causal relation- the extent to which it is, or is not supported ship. In such cases changes in one variable by data generated by empirical inquiry. (the dependent variable) are said to be dependent on another variable (the independent or Distinctive Features causal variable). The breadth of an hypothesis, of whatever Hypothesis formation and testing are closely level of specificity, is dependent upon whether associated with the quantitative approach in the investigator is asserting that the relation- social research and with the viewpoint that ship between variables holds for particular abstract concepts can be operationalized – or groups of people, particular contexts, for a indicated – by measurable variables. There is particular society or cross-culturally.

137 H-Jupp-3373.qxd 3/8/2006 10:42 AM Page 138

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

Two broad strategies of hypothesis testing can similar forms of social action are likely to be be identified. Verification involves looking found amongst other groups and in other con- for data that support the hypothesis whereas texts. In this sense, they represent a form of falsification – which involves looking for data hypothesis but not in terms of a statement of that refute the hypothesis – is viewed as a relationships between variables. What is much more stringent test. more, hypothesis ‘testing’ in qualitative The specific process of hypothesis testing research is a continuous process, involving usually involves testing the null hypothesis the search for cases or contexts that do not that there is no relationship between the spec- square with the assertions being made, rather ified variables. A statistical test is carried out than a once-and-for-all event. This process of to seek to refute this null hypothesis. If this is analytical induction implies that generaliza- not achieved then it is said that we have tions are modified until no further discon- evidence to support the alternative hypothesis. firming evidence can be found. This is known This is the research hypothesis that is under as the point of theoretical saturation and is consideration, namely that there is a relation- the point at which the validity of the general- ship as hypothesized by the researcher. ization is at its greatest. There are risks in the testing of hypotheses. A researcher can make two types of errors. First, the researcher can accept an hypothesis Victor Jupp as being correct when it is actually false. Second, the researcher can reject an hypothe- Associated Concepts: analytic induction, sis when in fact it should be accepted. The for- causal model, falsification, grounded theory, mer is known as a Type 1 error and the latter hypothetico-deductive model, indicator, mul- is known as a Type 2 error. tivariate analysis, Type 1/Type 2 errors, vari- Hypotheses about the relationship between able analysis two variables, say unemployment levels and crime levels, represent simplifications of the Key Readings social world. Many more variables are likely Bryman, A. (2001) Social Research Methods. to influence crime levels, or indeed any depen- Oxford: Oxford University Press. dent variable. It is for this reason that more Moser, C. A. and Kalton, G. (1971) Survey sophisticated procedures have been developed Methods in Social Investigations. London: to examine a series of interlocking hypotheses Heinemann. at one and the same time. This is what is Pawson, R. (1989) A Measure for Measures: A known as causal modelling and involves sta- Manifesto for Empirical Sociology. London: tistical techniques which come under the gen- Routledge. eral heading of multivariate analysis.

Evaluation Although typically associated with the hypo- HYPOTHETICO-DEDUCTIVE thetico-deductive model and with quantita- MODEL tive research it should not be assumed that hypotheses are irrelevant in qualitative inquiry. Definition Much qualitative research is concerned with producing descriptions of social interactions A general model of science (Popper, 1934, and social contexts and also moving to a 1959; Hempel, 1970) in which science is higher theoretical level of generalizing about stated as involving the formulation of hypo- these. Such generalizations are grounded in theses and theories from which particular observations collected in specific contexts occurrences can be deduced and thus also but often seek to suggest that, for example, predicted and explained. As a model of

138 H-Jupp-3373.qxd 3/8/2006 10:42 AM Page 139

HYPOTHETICO-DEDUCTIVE MODEL

scientific discovery and explanation the While the hypothetico-deductive approach hypothetico-deductive method is advanced as to theory and testing, and the construction of an alternative to Baconian ‘inductive method’ systematic formal theories, has sometimes (Bacon 1561–1626) in which the simple accu- been advocated as a way forward for the mulation of instances gives rise to generaliza- social sciences (for example, in sociology, tions. The model is based on the idea that, Zetterberg’s formalization of Durkheim’s rather than the accumulation of facts, hypo- Division of Labour, 1893), it has more often theses are essential to science as the basis of been strongly opposed (for example, by those proposed generalizations and their empirical who advocate grounded theory). However, testing (cf. ‘falsification’). ‘ideal types’ (for example, ‘rational action’, ‘perfect competition’) as heuristic aids to Distinctive Features explanation or as parts of formal theories have the merit of allowing exploration of the In terms of the model, hypotheses and theories more complex factors involved in empirical are advanced and generalizations and predic- ‘departures’ from these ‘idealized’ forms. tions made on the basis of deductions from A further general criticism is that adher- these, with: (1) successful prediction being ence to the hypothetico-deductive model and taken as a test of the adequacy or plausibility of its emphasis on the formal features of expla- the hypothesis and theory, and (2) explanation nation has distracted attention from the being seen as achieved (at least provisionally) concrete practices and complex historical once successful predictions have been made. processes of science and the way in which At its fullest extent, proponents of the theories and broader ‘scientific paradigms’ hypothetico-deductive model also advance emerge, prosper and are ultimately replaced the view that, ideally, scientific knowledge (as especially discussed by Thomas Kuhn should be formulated as a network of deduc- (1970) in his seminal volume The Structure of tively interrelated propositions and theories. Scientific Revolutions). As Blaug (1980: 9) sees At its fullest extent, the hypothetico-account it, it is precisely the aim of the hypothetico- portrays scientific theories as involving logi- deductive or covering-law model ‘to tell it cally (and mathematically) interrelated theo- like it should be’ and not to ‘tell it like it is’. retical statements, bridging principles or It is this prescriptive, normative function of correspondence rules that link theories to the covering-law model of explanation that observations, which include empirical laws critics find so objectionable. The most gen- and, ideally, sense datum statements. eral criticism is that there is so much more going on in physical, let alone social science, Evaluation that is not captured or adequately considered by the hypothetico-deductive model. An emphasis on the importance of the More generally, it is apparent that no advancing and the testing of empirical single model of scientific explanation can be hypotheses means that the hypothetico- said to enjoy universal support. Originally, deductive method is often identical to the logical positivists – in association with the ‘covering-law model of explanation’ and ‘verification principle’ – sought an absolute related to ‘falsificationism’. In so far as the grounding for empirical laws in a universal hypothetico-deductive model also implies data language and logic but failed in this that it constitutes the method of science, it objective. It is also a problem that generaliza- meets similar objections to those arising in tions that appear to qualify as ‘covering laws’ connection with these approaches. For under the D–N schema sometimes seem to rounded discussions of the combined model lack explanatory power or causal force within particular social sciences, including its (causality is anyway a disputed concept for roots in ‘positivism’ and ‘mathematical logic’, positivists). There are also problems with see Blaug (1980) and Halfpenny (1982). alternative general accounts of science such

139 H-Jupp-3373.qxd 3/8/2006 10:42 AM Page 140

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

as Bhaskar’s ‘scientific realism’ for the Key Readings grounding (and the links with observations) Blaug, M. (1980) The Methodology of of its statements concerning the existence Economics or How Economists Explain. and action of underlying powers and mecha- Cambridge: Cambridge University Press. nisms can be indeterminate (for example, the Halfpenny, P. (1982) Positivism and Sociology: Marxian theory that social classes exist and Explaining Social Life. London: Allen & determine social forms), since choosing Unwin. between theories avowedly depends on Hempel, C. (1965) Aspects of Scientific judgements of the overall plausibility, fertility Explanation. New York: Free Press. and elegance of theories. Kuhn, T. (1970) The Structure of Scientific Revolutions, 2nd, enlarged edn. Chicago: University of Chicago Press. David Jary Popper, K. (1934, English trans.; 1959) The Logic of Scientific Discovery. London: Associated Concepts: analytical induction, Hutchinson. Zetterberg, H. (1961) On Theory and causality, covering-law model of explanation, Verification in Sociology, Totawa, NJ: falsification, grounded theory, hypothesis, Bedminster Press. ideal type, positivism

140 I-Jupp-3373.qxd 3/8/2006 5:45 PM Page 141

in violation of the criminal law will actually IDEAL TYPE conform to all of these attributes. Secondly, an ideal type is an analytical tool Definition and not an ethical ideal. In other words, there A conceptual description of any social are no value judgements attached to the phenomenon (for example, mental disorder, elements used to describe a particular phe- crime, religion, bureaucracy) that helps nomenon. It is not a moral statement about researchers understand the subject matter what ought to be the defining features of a as it exists in its abstract or pure form. The particular subject, simply a description of foundations of ideal-type analysis in sociol- what has been observed to be in existence in ogy were adopted from the work of the emi- the real world. Nor should it be viewed as nent scholar Max Weber who originally presenting the ‘average’ type of a particular borrowed the practice from economics. social phenomenon. Weber’s aim was to employ ideal-type Thirdly, any ideal-type formations must descriptions in order to investigate social dif- be ‘objectively possible’ in the sense that ferences. For example, he explored different although they will not be found to exactly types of religion by contrasting the ideal replicate reality, they are at the very least description of a ‘Protestant’ with that of the required to approximate what is going on ideal ‘Jew’, ‘Hindu’ and Buddhist’. around us. Moreover, they must also satisfy the condition of being ‘subjectively adequate’ in the sense that the elements defined as being Distinctive Features ideal must be understandable in terms of the There continues to exist considerable confu- subjective views of the individual. For exam- sion within the social sciences regarding the ple, establishing an ideal-type conceptualiza- nature and purpose of ideal-type construction of the causes of crime in terms of tions and their relationship to the real world. demonological explanations (demons, witch- First, while an ideal type is created from craft, spirits) would probably not satisfy these observations made in the real world, nothing conditions, as it would have no objective or in reality precisely fits that ideal-type con- subjective reality in terms of the popular dis- struction. For example, even though existing courses and experiences found within late- research evidence may be used to generate modern Western societies. an ideal-type description of the causes of Finally, confusion also exists over whether crime (for example, biological, psychological, or not these ideal-type models can be used as sociological, political), this framework only the basis for research. Weber was not clear on exists as an abstract, exaggerated ‘typical’ this point, as on the one hand he claimed that model to help us explain the behaviour of ideal-type concepts were not models to be criminals; it is unlikely that individuals found tested, but served a more specific purpose – as I-Jupp-3373.qxd 3/8/2006 5:45 PM Page 142

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

heuristic descriptions which help sharpen the Colombo, A., Bendelow, G., Fulford, B. and classification and comparison of social events Williams, S. (2003) ‘Evaluating the influ- across time and place – while on the other ence of implicit models of mental disor- hand he often implicitly used them himself der on processes of shared decision during his studies (Rogers, 1969). making within community-based multi- disciplinary teams’, International Journal of Evaluation Social Science and Medicine, 56: 1557–70. Freund, J. (1969) The Sociology of Max Weber. Several researchers have successfully demon- New York: Vintage Books. strated that there is a great deal to be gained Rogers, R. E. (1969) Max Weber’s Ideal Type from recognizing the close relationship Theory. New York: Philosophical Library. between ideal types and real world situations. Winch, P. (1958) The Idea of Social Science. In particular, the present author (Colombo, London: Routledge. 1997; Colombo et al., 2003) has managed to: (1) formulate from existing theoretical knowledge several ideal-type representa- IDIOGRAPHIC tions of a particular social issue, namely the treatment, management and care of Definition mentally disordered offenders; (2) employ these ideal-type ‘models of mental An approach or style within social research disorder’ as a baseline in order to evaluate that focuses on specific elements, individuals, current professional practice; and events, entities and situations, documents and (3) use the findings in order to redefine or works of culture or of art and concentrates on ‘improve’ the original ideal type models what is particular to these. This differs from (medical, social, cognitive-behavioural, research that highlights regularities and psychotherapeutic), and integrate them repeatable elements of form or behaviour as as the basis for training programmes on part of larger processes or patterns concerned inter-professional practice. with general laws and theories. The latter approach is known as nomothetic. It is important to note, however, that some sociologists would argue that employing ideal types in this way, that is, to generate a ‘mean- Distinctive Features ingful understanding’ of specific issues, is Idiographic (with an ‘i’) should not be confused inappropriate and that they should only be with the term ideographic (with an ‘e’) – which used in the development of general concepts refers to images of graphic symbols and theories (Winch, 1958). that represent things and ideas. Idiographic research is usually explained as distinct from nomothetic research. The idiographic/nomo- Anthony Colombo thetic opposition forms a significant distinction within social research. ‘Idio’ derives from Associated Concepts: covering-law model of Greek and is concerned with the individual, explanation, deviant case analysis, hypo- and ‘nomos’ refers to ‘law’ in the sense of order thetico-deductive model, sensitizing concepts, and consistency. Whereas idiographic refers to Verstehen specific, idiosyncratic cases, the particular circumstances that give rise to them and the occurrence of unique and non-recurrent Key Readings aspects, nomothetic approaches, on the other Colombo, A. (1997) Understanding Mentally hand, emphasize that the social world cannot be Disordered Offenders. Aldershot: Ashgate characterized as a set of random occurrences, Press. but as repeatable entities that give rise to

142 I-Jupp-3373.qxd 3/8/2006 5:45 PM Page 143

IMPRESSION MANAGEMENT

patterns, predictions and rules. For example, an Associated Concepts: autoethnography, idiographic research project centred on a strike covering-law model, ethnography, micro- or labour dispute might investigate the specific ethnography, nomothetic, qualitative research, nature of the dispute, the local circumstances quantitative research, thick description, that gave rise to it, and the groups, individuals, validity of generalization issues, demands and actions involved in creating, conducting and concluding the conflict. On Key Readings the other hand, a nomothetic approach will place the dispute as part of a series of repeatable Crotty, M. (1998) The Foundations of Social occurrences of disputes between employers and Research. London: Sage. workers that show persistent patterns within Wallerstein, I. (1987) ‘World systems analysis’, in A. Giddens and J. Turner (eds), Social wider historical trends and symptomatic fea- Theory Today. Cambridge: Polity Press. tures of the social and economic relations of Wallerstein, I. (1997) ‘Social science and the production. Interpretivist and qualitative tradi- quest for just society’, American Journal of tions of social inquiry are more readily identi- Sociology, 102 (5): 309–24. fied with idiographic research; nomothetic research is identified with quantitative approaches associated with the natural sciences.

Evaluation IMPRESSION MANAGEMENT Idiographic research typically provides information and data of value to an understanding Definition of the specific case in question. This is only valid for the investigated case and cannot be The conscious management of self-presenta- used to describe the set from which it is drawn. tion, image and role in order to gain research It offers little or no insight into social patterns access, maintain research relationships and and regularities. Nomothetic research, because facilitate the general progress of social of the nature of its inquiry, requires multiple research. cases in order to show patterns of repeatable elements but is incapable of dealing with the Distinctive Features specificity of any individual case. The difference between the two research Impression management is a key feature of approaches can be viewed in two ways. In social research. Creating a good impression what is often referred to as the ‘hard’ version, can aid research relationships and the collec- the idiographic and the nomothetic are tion of data. Alternatively, leaving a poor or deemed incompatible. For example, adherents bad impression can hinder all stages of the of idiographic research might emphasize the research process, for example, hindering primacy of specificity of the element of analy- access to settings, people or data, making sis and insist that generalizations built on two research relationships difficult to forge or or more cases are impossible. However, the sustain and jeopardizing the establishment of ‘soft’ version of the idiographic/nomothetic rapport and trust. divide recognizes that the two approaches are Social researchers usually pay particular different ways of coming at social reality, using attention to their self-presentation, or the man- different research methods, but that the two agement of their ‘personal front’ (Goffman, might be usefully used in conjunction 1959), during the early stages of research. (Wallerstein, 1987). This is akin to mixed- Creating a positive impression is often vital to method research and triangulation. the successful negotiation of access, and to the initial establishment of research relationships. The social researcher is tasked with creating Chris Wharton a working identity that is acceptable to the

143 I-Jupp-3373.qxd 3/8/2006 5:45 PM Page 144

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

setting or to potential informants. This can address the issue of power within social involve some very practical considerations, research. It also does not guarantee recipro- including according attention to physical cal research relationships. appearance. Dress is one of the ways in which Adopting different roles within research we create an impression of ourselves. Clothing settings can be conceptualized in terms of the and accessories can be used to fit into a performance of self. Goffman has argued that setting, show allegiance with informants, all social actors are tasked with the perfor- demonstrate deference or the seriousness with mance of self in everyday social life. The task which the research is being undertaken. for social researchers is to be reflexive about Knowing how to dress (for example to ‘dress this process. up’ or to ‘dress down’) is part of the theoretical work of understanding the setting or people being studied, and of knowing what kind Amanda Coffey of appearance will create the appropriate impression. Other practical considerations Associated Concepts: access, covert research, might include speech and demeanour, the use fieldwork, gatekeeper, reflexivity of props (there are some settings, for example, when a briefcase may create the wrong Key Readings impression) and personal habits (such as drinking alcohol). Coffey, A. (1999) The Ethnographic Self: Impression management is also about creat- Fieldwork and the Representation of ing and maintaining a role over the course of Identity. London: Sage. the research. Indeed, the social researcher may de Laine, M. (2000) Fieldwork, Participation adopt a number of different roles, at different and Practice: Ethics and Dilemmas in Qualitative Research. Thousand Oaks, CA: times and in relation to different social actors Sage. within the research setting (de Laine, 2000). Goffman, E. (1959) The Presentation of Self in Social researchers may have to respond to dif- Everyday Life. Garden City, NY: Anchor. ferent expectations that may be placed on Kondo, D. (1990) Crafting Selves: Power, them, and adapt their roles accordingly. Gender and Discourses of Identity in a Japanese Workplace. Chicago: University Evaluation of Chicago Press. Being sensitive to the research setting is good research practice. However, the active management of the self for the purposes of research might be considered to be unethical. INDICATOR Adopting a completely false persona, entering into non-genuine relationships or making Definition promises you have no intention of honouring raises difficult questions about deception, A measurable quantity which ‘stands in’ or exploitation and trust. substitutes, in some sense, for something less Self-disclosure can also be an issue for readily measurable. social researchers keen to create a good impression. How far should the researcher Distinctive Features go in sharing their lives and experiences with research participants in order to establish There are three related ways in which this rapport and ‘fit in’? Self-disclosure may facili- term is used. First, some people would use tate stronger and more genuine research ‘indicator’ for any situation where a chain of relationships. However, it does not necessarily logic and evidence is needed to reason from

144 I-Jupp-3373.qxd 3/8/2006 5:45 PM Page 145

INDICATOR

the measurements taken to the extent of the ones. All of these have demonstrable problems. criterion. In this sense thermometer readings Car ownership, for example, correlates well would be indicators of temperature, because with income, but only in urban areas; the what is actually measured is the height of, nature of rural life is different. Overcrowding say, mercury in a tube; inferring temperature does not work as a measure of affluence in any from this requires knowledge of physics. area with a high proportion of social housing, A less inclusive use of the term would because those who allocated the housing make encompass those ‘measures’ where there is a sure that overcrowding does not take place. high ‘face validity’ connection between the measure and the underlying criterion and/or Evaluation successful effort has been put into demonstrating the measuring instrument’s validity. One problem with this approach, despite its An example would be intelligence testing. conceptual clarity, is that it is very much a Here the underlying concept, the criterion, reflection of a positivistic worldview. It takes involves problem-solving ability and mental for granted, as unproblematic, the underlying agility in real-life settings, and most of the nature of what is being measured; the measuring instruments are of the ‘paper- problems are measurement problems. This and-pencil’ variety. However, the test items becomes a less attractive way of conceptual- are often themselves puzzles to be solved izing the research if one conceives of the indi- quickly, and a wide range of studies have cator and the criterion concept as aspects of demonstrated that the measuring instruments the same thing – not correlated, but in some do largely behave as an accurate indicator of sense formally identical or jointly produced – intelligence ought to behave. Another exam- or if the nature of the criterion might be ple might be indicators of social class, based specific to individuals or settings in ways that crudely upon current job (or recent job, or do not show an overall regularity, or if the usual job, or head of household’s job), which nature of what is being measured is in itself by no means mirrors precisely the complexi- problematic. In measuring material depriva- ties of the underlying concept. tion, for example, one might take income as A third use, the most common, would be a measure of poverty or use correlated indi- for criterion concepts which we are unable to cators such as car ownership or density of measure directly but with which some other occupation. However, some might argue that measurement correlates sufficiently highly to to ascertain whether people are materially be a useful predictor of them. Examples would deprived, as opposed to just poor, may require be various measures of affluence and material talking to individuals and/or observing their deprivation. It is often not possible to obtain living conditions and life chances in idio- reliable answers to questions about income. graphic rather than nomothetic ways. In British studies of health and deprivation it has been found that car ownership provides Roger Sapsford a reasonable proxy; it is not a measure of wealth – it is perfectly possible for poor people Associated Concepts: composite measure- to own cars – but there is a sufficient correla- ment, measurement, performance indicators, tion with income for it to be a reasonably valid positivism, projective technique, reliability, indicator of income. Extent of overcrowding – scaling, validity of measurement ratio of number of people in household to number of rooms in house – has been used as another indicator. Another occasionally used in Key Reading cross-national comparisons is extent of literacy, given that there is a tendency for those nations Sapsford, R. and Jupp, V. (1996) Data where literacy rates are high to be the wealthier Collection and Analysis. London: Sage.

145 I-Jupp-3373.qxd 3/8/2006 5:45 PM Page 146

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

all forms of rational inquiry involve a sharp INDUCTION distinction between generating and testing hypotheses. While inductive inference might Definition play a role in the former, he insisted that it A form of reasoning from statements about played no role in the latter. Moreover, observed cases to statements about other, hypotheses produced by induction were no unobserved, cases or – more usually – to a more likely to be true than those produced in general claim about most or all cases of the other ways, for example through dreams. same kind. It is contrasted with deduction (or However, by no means all twentieth-century the hypothetico-deductive model), which is philosophers were persuaded by this critique. reasoning from a general premise (for exam- A number of approaches within social ple, ‘all humans die’), plus a statement about science have adopted an inductive orientation, some particular case (‘Socrates is human’), in one sense or another. Forms of ‘data dredg- to a further conclusion about that case ing’ in quantitative analysis are one example. (‘Socrates is mortal’). The most important In the context of qualitative research, there is distinction between these two forms of rea- the ‘analytic induction’ developed by Znaniecki soning is that induction is ampliative: the and Lindesmith (see Hammersley, 1989) and conclusion is not already contained in the Glaser and Strauss’s ‘grounded theorizing’ (see meaning of the premises; whereas deduction Dey, 1999). The latter authors argued that can tell us no more than what is logically theory ‘grounded’ in – in other words, devel- implied by the premises. oped out of – close empirical study of the social world is more likely to be true than that which is produced by armchair theorists. They criti- Distinctive Features cize the tendency of much empirical research Some commentators regard ‘induction’ as at the time they were writing for ‘verification- central to most forms of inquiry, while others ism’, for setting out to test hypotheses derived deny that it is a valid form of reasoning. from speculative theory. Another form of qual- Within philosophy, for much of the second itative research which adopts an inductive half of the last millennium, a crude contrast orientation is conversation analysis (see can be drawn between those approaches Hutchby and Wooffitt, 1998: 94–5). which took deductive logic – or mathematical proof – as the model for scientific reasoning, Evaluation and those – from Bacon to Mill – which Several kinds of induction can be distinguished: treated induction as the core of scientific method. Early twentieth-century philosophy (1) Enumerative induction. This involves gen- of science, initially dominated by logical eralization from a sample whose features positivism, witnessed a shift from inductivism have been studied to a larger, finite toward the hypothetico-deductive method. population of cases many of which have This change in position was motivated by not been studied. This form of inference what came to be referred to as the problem of serves as a substitute for complete enu- induction: the failure of attempts to provide a meration. Within the social sciences, logical account of induction equivalent to that enumerative induction is fundamental to available for deduction, one which showed social survey research. how induction must be pursued if the conclu- (2) Hypothetical induction (also sometimes called sions reached are to be absolutely certain in ‘abduction’, ‘retroduction’, or ‘inference to the validity, and why they would be certain. best explanation’). Here, inference is from The most influential twentieth-century what is observed to what is taken to be opponent of induction was Karl Popper, who the most likely explanation for what has waged a war against inductivist accounts of been seen, where this explanation involves science throughout his long career. For him, a general theoretical principle (usually

146 I-Jupp-3373.qxd 3/8/2006 5:45 PM Page 147

INFERENTIAL STATISTICS

attributing causality) that would apply to Key Readings other cases. Dey, I. (1999) Grounding Grounded Theory. (3) Probative induction. This involves reason- San Diego, CA: Academic Press. ing from evidence collected for the pur- Hammersley, M. (1989) The Dilemma of pose of testing a hypothesis or theory to Qualitative Method. London: Routledge & a conclusion about its validity. Kegan Paul. (For a useful discussion of different kinds of Hutchby, I. and Wooffitt, R. (1998) induction, see Rescher, 1978.) Conversation Analysis: Principles, Practices It should be clear from the above discus- and Applications. Cambridge: Polity Press. sion that whether induction is a valid form of Rescher, N. (1978) Peirce’s Philosophy of Science reasoning depends on what sense is given to . South Bend, IN: University of Notre Dame Press. the term. It seems clear that there is no such thing as logical – that is, demonstrative – induction from knowledge of particulars to knowledge about universals. However, there are various forms of ampliative reason- INFERENTIAL STATISTICS ing that are legitimate in principle, though whether any particular inference is likely to Definition be true is of course a matter of judgement on the basis of the evidence available. These Statistics from which inferences are made forms of reasoning include all three types of about situations or social groupings that have induction identified above. not been observed directly; for example, It is also important to stress that these forms making inferences about a population based of induction serve different functions within on a sample drawn from that population. the process of inquiry; they are not competitors In contrast to descriptive statistics, inferential or substitutes. Part of the problem with argu- statistics are used to make generaliza- ments about induction, especially criticisms tions derived from estimates based on proba- of it, is that the term is often treated as if bility. It is therefore important that the it referred to a single form of reasoning, demon- sample on which inferences about a popula- strative in character, which constitutes the tion are based is drawn randomly from that whole of the research process. And, as Hume population. made clear, it is not difficult to show that no such logic of inquiry is available: to infer that Distinctive Features the sun will rise tomorrow solely from the fact that it rose this morning is of doubtful validity. Statistics can be either descriptive or inferen- However, combining the fact that the sun rose tial. Descriptive statistics simply state what this morning with knowledge of the laws of has been observed. Inferential statistics make planetary motion, the inference to what will inferences based on those observations. happen tomorrow is much stronger – though, Rowntree, along with other writers, places of course, neither those laws nor anything else much emphasis on that aspect of inferential guarantees that the sun will ‘rise’ tomorrow. statistics which involves ‘generalizing from a sample to make estimates and inferences about a wider population’ (Rowntree, 1981: Martyn Hammersley 19). However, Gayle places them in a wider theoretical context by saying that ‘Inferential Associated Concepts: analytic induction, statistics allow us to make some statistical conversation analysis, grounded theory, generalizations about an aspect of the social hunting (snooping and fishing), hypothetico- world’ (Gayle, 2000: 385). Inferential statis- deductive model, inferential statistics, valid- tics enable hypotheses to be tested and gen- ity of generalization eralizations to be made. These can include (i)

147 I-Jupp-3373.qxd 3/8/2006 5:45 PM Page 148

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

determining whether a sample came from a groups of offenders on other theoretically particular population, (ii) whether differences important variables. observed between groups are likely to be real differences, or are more likely to have Evaluation occurred by chance, and (iii) whether a relationship between two or more variables is sig- The results of surveys are based on a sample nificant. Inferential statistics also enable us drawn from a population. While the result to make predictions about what is likely to given by a random sample produces a reason- happen on the basis of what has been able estimate of what the population is like, it observed. Statistics such as chi-square, seldom produces exactly the same result as if Student’s t and Pearson’s product moment the entire population had been surveyed, and correlation coefficient are all inferential. different samples drawn from the same popu- Thus, a descriptive statistic may tell us lation are likely to give slightly different that ‘the mean age of a group of students who results. For example, if there is a school with answered a questionnaire was 19.7 years’. As 1000 pupils, and if one had enough resources long as the students who answered the ques- to interview every pupil (the population in tionnaire were a random sample of students this instance), you might find that the mean at their institution, it may then be said that age of all the pupils in the school is 15 years. the mean age of students at Institution X If, however, you only have enough time and is 19.7 years, within a margin of error – an money to interview 100 pupils, drawn at ran- inference. The qualification ‘within a margin dom, then the mean age of the sample might of error’ is important, because error is always be 14.5 years. If someone else were to draw liable to occur. In this instance calculations another random sample of 100 pupils, their are made to determine the standard error of sample might have a mean age of 15.5 years. the mean that enable us to say that there is These variations are known as sampling a 95 per cent probability that the mean age errors. The larger a sample is (that is, the of students at Institution X is, let us say, closer it gets to encompassing the whole pop- between 19.5 and 19.9 years. These are the ulation) the closer the result is likely to be to confidence intervals of the mean. So infer- the actual population figure. ences are made on the basis of clear rules and Assuming, as is usually the case, that the procedures about how they may be deduced. figure for the population (the mean age in the Inferential statistics enable us to test example above) is not known, and all one has hypotheses about population parameters. If is the sample result, then an inference is descriptive statistics based on a sample tell us made about the mean age of the population that, during a particular period, Court A sent based on the sample result, but one knows 12 per cent of the convicted offenders that it is unlikely to be exactly the same as the appearing before it to prison, whereas Court figure for the population. So the population B sent 18 per cent of similar offenders to figure is going to be an estimate. This esti- prison, these proportions can be compared so mate is based on probability. Probability as to tell us whether this difference could theory is therefore the basis for inferential have occurred by chance, or we can be rea- statistics, and an awareness of the role played sonably certain that Court B is more ‘puni- by probability is important. tive’ than Court A. Of course, the similarity of the offenders is crucial in drawing any conclusions: if the offenders at Court B have in Iain Crow general committed more serious offences this may explain the difference. So, in addition to Associated Concepts: chi-square test, corre- testing whether the difference is real or may lation, descriptive statistics, error, induction, have occurred by chance, we also need to test prediction, probability (random) sample, for the similarity or difference of the two validity of generalization

148 I-Jupp-3373.qxd 3/8/2006 5:45 PM Page 149

INFORMED CONSENT

Key Readings right whenever and for whatever reason they wish. Moreover, whilst the researched may Gayle, V. (2000) ‘Inferential statistics’, in agree to participate generally they should D. Burton (ed.), Research Training for Social nevertheless also feel free and be free to exer- Scientists. London: Sage. pp. 385–413. cise their powers of veto during the research Kranzler, G. and Moursund, J. (1995) ‘Statistics for the terrified’, is Introduction process and reject the use of specific data to Inferential Statistics. Englewood Cliffs, gathering devices such as tape recorders and NJ: Prentice Hall. ch. 7. video cameras. Rowntree, D. (1981) Statistics without Tears. For consent to be fully informed it is London: Penguin. incumbent upon the researcher to explain as fully as possible, and in terms meaningful to participants, the questions of what, who, why and how. That is, what the research is about, who is undertaking and financing it, why it is INFORMED CONSENT being undertaken, how it is to be promoted and how any research findings are to be dis- Definition seminated. None of these aspects of the An ethical principle implying a responsibil- research design are clear cut or straightfor- ity on the part of the social researcher to ward. Who for example are the participants strive to ensure that those involved as par- in the research? It is one of the responsibili- ticipants in research not only agree and con- ties of the researcher to determine who sent to participating in the research of their informed consent must be obtained from. own free choice, without being pressurized If access to research is gained through a ‘gate- or influenced, but that they are fully keeper’ for instance, informed consent informed about what it is they are consent- should be sought directly from the research ing to. participants to whom access is required. This may mean adhering to the principle of informed consent at several different levels Distinctive Features of access. Special care in this respect must Most professional and academic organiza- also be taken where research participants are tions have codes of conduct or codes of ethics vulnerable by virtue of factors such as age, to offer guidance to promote good practice social status, or powerlessness, or are ill or and high ethical standards in research. infirm or where proxies may need to be used The principle of informed consent ensures in order to gather data. researchers conduct themselves with honesty Informed consent places the onus on and integrity and with consideration and the researcher to decide what is meant by respect for the research subjects. This can anonymity and confidentiality whilst research be achieved by establishing guidelines, fol- participants should be informed about and lowing good practice, consultation and understand how far they will be afforded communication. Under such codes, informed anonymity and confidentiality. The researcher’s consent is one of the main responsibilities responsibility extends to a careful consideration that researchers have towards their research about making promises or unrealistic guaran- subjects, whose rights should be respected. tees that might later be difficult, impossible or This particular responsibility implies that tempting not to keep. researchers should base research, so far as possible, on the freely given informed con- Evaluation sent of those studied. Research participants should be made aware of their right to say no, Whilst informed consent ought to be inte- decline or refuse permission and to partici- gral to the research design this can prove pate. They should be able to exercise this difficult to adhere to in practice during

149 I-Jupp-3373.qxd 3/8/2006 5:45 PM Page 150

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

the conduct of social research. Informed or during the research, it might nevertheless consent might be seen as an ideal-typical still be obtained afterwards. principle to which all social research should aspire. In reality, it may be impossible to achieve consent that is fully informed and Pamela Davies in practice informed consent is never likely to be fully attained. Associated Concepts: access, confidentiality, The practice of social research is often ethics, experiment, fieldwork, gatekeeper, par- a process or activity that involves continuous ticipant observation, politics and social research decision making. As the research progresses it often creates a range of ethical choices Key Readings and dilemmas. In some research contexts Barnes, J. A. (1979) Who Should Know What? informed consent may need revisiting as the Harmondsworth: Penguin. research develops and evolves. The aims and Bryman, A. (2001) Social Research Methods. objectives of a research question may be Oxford: Oxford University Press. clearly stated at the outset and empirical or Bulmer, M. (ed.) (1982) Social Research experimental fieldwork will be conducted Ethics. London: Macmillan. according to an original research design. Bulmer, M. (2001) ‘The ethics of social Informed consent might be sought at an early research’, in N. Gilbert (ed.), Researching stage but research is often dynamic and Social Life. London: Sage. pp. 45–57. developmental and consequently the nature Homan, R. (1991) The Ethics of Social and use of the material obtained from the Research. London: Pearson Education. research is likely to alter. It may be necessary Humphreys, J. (1970) The Tea Room Trade: for the obtaining of informed consent to be Impersonal Sex in Public Places. Chicago: regarded, not as a one-off event at the outset, Aldine. but as a process, subject to renegotiation or Punch, M. (1994) ‘Politics and ethics in qual- reminder, especially during the course of itative research’, in N. K. Denzin and Y. S. research over prolonged periods of time. Lincoln (eds), Handbook of Qualitative Informed consent is an ethical principle that Research. London: Sage. pp. 83–97. may demand continuous dialogue and communication if participants are to fully understand what the research is about, who is doing and financing it, why it is being done INTELLECTUAL CRAFTSMANSHIP and how it will be used. Whilst the principle might be generally adhered to, at any stage of the research it Definition is questionable how far consent can ever A way of working, associated with the be informed, especially where research cuts American sociologist C. Wright Mills, which across cultural and linguistic divides. encompasses a distinctive methodological Similarly, it is questionable how far partici- approach, in terms of the way research ques- pants can ever fully understand what it is to tions should be conceptualized, but also which they are committing themselves and includes practical strategies regarding keep- what use will be made of the research of ing research files and presenting findings and which they are a part. conclusions. Finally, covert methods of social research do not adhere to the principle of informed Distinctive Features consent in any respect. This raises particular ethical issues. Where informed consent of ‘On Intellectual Craftsmanship’ is the title of those studied has not been obtained prior to the appendix to a highly influential book, The

150 I-Jupp-3373.qxd 3/8/2006 5:45 PM Page 151

INTELLECTUAL CRAFTSMANSHIP

Sociological Imagination, published in the late personal problems and public issues: for Mills 1950s by C. Wright Mills (who died prema- there should be a continuous interchange turely but who left an important legacy in the between personal experience and scholar- social sciences). His writings were very ship. This is a much wider interpretation of firmly in the critical tradition. reflexivity than is used by those who simply Intellectual craftsmanship is a way of refer to it in terms of reflecting on the process unlocking the sociological imagination, which of fieldwork. He also advocates a number of is characterized by seeking to locate and other reflective devices such as thinking explain personal troubles in terms of public about opposites or extremes from the phe- issues. For example, the personal and private nomena under investigation; deliberately experiences of unemployment need to be inverting the sense of proportion by imagin- understood in terms of structural changes in ing something as bigger or smaller than the economy which displace people’s jobs; the it really is; looking at phenomena from a dif- pain of marital separation needs to be ferent disciplinary viewpoint, for example explained in terms of the pressures on the that of a psychologist or historian; and imag- family system in modern industrial society. ining instances of the phenomena under con- In this way, the sociological imagination exam- sideration in different contexts. All such ines how individual biographies are shaped by devices are employed to stimulate the socio- social structure and how that structure changes logical imagination. over time. For Mills social science should con- A third feature of scholarship relates to cern itself with the relationships between indi- the presentation of findings and conclu- vidual biographies, social structure and history, sions drawn from them. This is what Mills and research questions should be framed in describes as ‘literary craftsmanship’. Such these terms. craftsmanship involves paying attention to Intellectual craftsmanship is the means of the interplay between reflexivity in the devel- posing and addressing such questions and has opment of a research file (the ‘context of dis- a number of features. First, it involves keeping covery’) and writing in the publication of a file or journal of research ideas and ques- books and articles (the ‘context of presentations. These are continuously under review tion’). The former involves thinking clearly and revision rather than in the form of a once- and presenting ideas to oneself so that one and-for-all research proposal (which Mills can eventually write clearly and intelligibly believed was likely to be a form of window in order to present ideas to a wider audience. dressing aimed at attracting funding). The sociological imagination, he argued, is released by Evaluation constantly revisiting and reorganizing the file, by subdividing existing plans and by throwing There is much to admire and respect in the together what, on the surface, appear to be writings of C. Wright Mills. He stood steadfastly unrelated facts and ideas. It is not released by against theorizing for its own sake and against adopting official or institutional definitions of empirical inquiry for its own sake. In doing so what is problematic and then perhaps engaging he left a legacy that emphasized the necessity in applied, policy-related contract research. for a close interchange between theory, method (However, asking why certain matters receive and personal and public problems. He was also the official seal of being ‘problematic’ at any against those who were so locked in the jargon given time, and with what consequence would of theory or the jargon of empirical inquiry that be an important question for the file.) they could not present ideas, findings and con- A second key feature of the intellectual clusions in an intelligible manner. endeavour is reflexivity, that is, developing Perhaps what has not stood the test of time is the contents of the file by systematically his use of the term intellectual craftsmanship, reflecting how life experiences shed light on which for some would seem inappropriate.

151 I-Jupp-3373.qxd 3/8/2006 5:45 PM Page 152

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

However, a critical thinker such as Mills, online journals). Carrying out secondary writing from the distant 1950s, would not research on the Internet requires choosing deliberately intend this to be sexist language. appropriate tools, for example, search engines, or the resources mentioned above, and using these to locate relevant material. The credibil- Victor Jupp ity of the information accessed must then be assessed, before compiling a bibliography or Associated Concepts: critical research, policy- literature review. related research, politics and social research Using the Internet to conduct primary reflexivity, writing research research is appealing due to the mass con- nectivity of millions of individuals worldwide Key Reading which the Internet affords. The size and diversity of this population is likely to Mills, C. Wright (1959) The Sociological continue to expand (Mann and Stewart, Imagination. Harmondsworth: Penguin 2000), thus increasing the possibilities for (especially the appendix ‘On Intellectual Craftsmanship’). Internet-mediated primary research. The most common approach to date is survey research. Experiments, observational studies, interviews and focus groups, and analysis of linguistic archives have also been conducted. INTERNET RESEARCH The stages involved in Internet primary research are much the same as in traditional Definition (non-Internet) modes. Extensive piloting Use of the Internet to conduct primary during the design phase is especially impor- research, as in recruiting participants, adminis- tant in Internet research due to the novelty tering materials and collecting data, or sec- of the procedures being used. Recruiting par- ondary research, as in accessing online sources ticipants can be done by posting requests to (such as library databases and online journals) newsgroups or individual mail accounts, or which report the results of primary research. by placing a study advertisement on the World Wide Web. A number of websites Distinctive Features dedicated to listing currently active online studies exist (for example, Psychological Secondary research conducted via the Internet Research on the Net: psych.hanover.edu/ creates possibilities for accessing a particu- research/exponnet.html) and can be a useful larly large and diverse body of information. place to advertise a study (though the possible Resources relevant to social research include bias in the type of participants accessed via electronic journals (for example, in the UK, such sites must be carefully considered). Sociological Research Online: www.socreson Administering study materials and procedures, line.org.uk), bibliographic databases (for exam- and collecting data, can be done via a range of ple, Bath Information and Data Services Internet technologies, for example email, FTP [BIDS]), library catalogues, websites containing (File Transfer Protocol), HTML (HyperText lists of subject-relevant links, online news- Mark-up Language, which allows interactive papers (for example, www.guardian.co.uk), forms to be placed on a web page), and Internet subject-based discussion groups, and electronic chat systems (further discussion of the tools texts available from individuals’ or institutional available is provided in Hewson et al., 2003). web pages (for example, bibliographies, lecture A third sense of the term Internet research notes, unpublished manuscripts). Some of refers to the study of the Internet itself (for fur- these resources can be accessed free of charge ther discussion see Jones, 1999). In the context (for example, online newspapers), others of social research, this can involve studying require a subscription fee (for example, most online communities and social structures, as

152 I-Jupp-3373.qxd 3/8/2006 5:45 PM Page 153

INTERPRETIVE REPERTOIRES

well as examining the effects of the Internet on Internet and traditional implementations of a people’s lives, behaviour and experience. study for example (see the chapters in Reips Research questions that have emerged include: and Bosnjak, 2001). A possible barrier in What are the characteristics of Internet users? Internet research is the level of technical Is ‘Internet addiction’ a real phenomenon? Do expertise required of the researcher, espe- people interact differently online (compared cially when more sophisticated implemen- with offline)? How has the Internet impacted tations are required. However, a number upon the experiences of marginalized or of tools exist (for example, SurveyWiz: psych. oppressed groups? What is the nature and fullerton.edu/mbirnbaum/programs/survey- extent of Internet crime? wiz.htm), designed to aid development of Internet studies with minimal levels of technical expertise. While Internet-mediated Evaluation primary research is still in its infancy, some Secondary research using the Internet can be encouraging results have been obtained to both time-saving and allow access to more date. Further implementation and testing of information than in traditional (for example, Internet research procedures is needed to library-based) modes. However, it is crucial to clarify the future scope and usefulness of the assess the relevance, quality, and reliability Internet as a tool for gathering primary data. of sources located. A number of strategies have been outlined to assess the quality of Claire Hewson information accessed via the Internet. These include checking author details and affiliation, checking the date of publication and Associated Concepts: access, confidentiality, checking whether other sources have been data archives, ethics, focus group, informed accurately represented (Hewson et al., 2003). consent, literature review, media analysis, Advantages of primary Internet research primary research, secondary analysis include cost- and time-effectiveness, access to a large and diverse participant pool, auto- Key Readings maticity of features such as response com- Hewson, C., Yule, P., Laurent, D. and Vogel, pleteness checking, and availability of data in C. (2003) Internet Research Methods: A a format ready for statistical analysis. These Practical Guide for the Social and Behavioural features could help enhance data validity, and Sciences. London: Sage. facilitate cross-cultural research, access to Jones, S. G. (ed.) (1999) Doing Internet special populations, and projects with limited Research: Critical Issues and Methods for time and funding. It has also been suggested Examining the Net. London: Sage. that the greater levels of interactivity possible Mann, C. and Stewart, F. (2000) Internet while maintaining complete anonymity, for Communication and Qualitative Research. example in an online interview, may help London: Sage. reduce bias due to biosocial attributes of the Reips, U.-D. and Bosnjak, M. (eds) (2001) experimenter or participant (Hewson et al., Dimensions of Internet Science. Pabst 2003). Unobtrusive observational research is Science Publishers: Lengerich, Germany. also made accessible due to the availability of, for example, archives of online discussion groups. Issues of concern in Internet primary research include sample bias, reduced levels INTERPRETIVE REPERTOIRES of researcher control and difficulty implementing ethical procedures such as obtaining Definition informed consent or ensuring confidentiality. However, a number of studies to date have A concept developed to aid the discourse shown Internet data to be valid, by comparing analysis of talk and texts and most commonly

153 I-Jupp-3373.qxd 3/8/2006 5:45 PM Page 154

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

used by social psychologists and other inconsistency of everyday discourse on discourse researchers to summarize relatively controversial topics such as race and racism. global patterns in people’s sense making – People’s discourse is better described as a particularly around controversial issues and patchwork (Billig, 1991) of inconsistent motifs. matters of public opinion. Interpretive reper- Much everyday discourse is organized around toires operate at a broad semantically based dilemmas and involves arguing and puzzling level. They are recognizable routines of over these (Billig et al., 1988). Arguments connected arguments, explanations, evalua- around meritocracy, for example (‘the best per- tions and descriptions which often depend on son should get the job’) are intertwined with familiar anecdotes, illustrations, tropes or equal opportunities discourse (‘those disadvan- clichés. Interpretive repertoires are the build- taged in the job market should be positively ing blocks through which people develop supported’). Reference to the practical consid- accounts and versions of significant events in erations (such as not being ‘able to turn the social interaction and through which they clock backwards’), for instance, give way to perform identities and social life. arguments of principle (‘social justice applies whatever the context’). The concept of inter- Distinctive Features pretive repertoire is an attempt to identify and summarize the regular elements in this shifting The term ‘interpretive repertoires’ was first mosaic. Interpretive repertoires are ‘what developed in 1985 by two sociologists of sci- everyone knows’. Indeed, the collectively ence, Nigel Gilbert and Michael Mulkay, shared social consensus behind an inter-pretive to describe patterns in the discourse of the sci- repertoire is often so established and familiar entists they were studying. Gilbert and Mulkay that only a fragment of the argumentative noted that scientists had two contradictory and chain needs to be formulated in talk to provide inconsistent but very regular and familiar ways an adequate basis for the participants to jointly of describing their scientific activity – an recognize and sometimes resist and challenge empiricist repertoire (a formal objectivist the version of the world that is developing. account of scientific discovery and progress) and a contingent repertoire (which stressed the Evaluation informal social factors in scientific work such as people’s networks and ambitions). Each reper- The term interpretive repertoire can be con- toire constructed a different social world, popu- trasted with Foucault’s much wider ranging lated that world with different kinds of notion of ‘discourses’. Both terms are con- characters (heroes and villains) and constructed cerned with the configurations of intelligibility, different teleological histories and causal stories with global forms of semantic order and with for the same events. These repertoires, sepa- ways in which discourse is constitutive, con- rately and in combination, were used to power- structing realities and truths. Interpretive ful rhetorical effect in different contexts. repertoires, however, are more fragmentary, This notion was taken up and further elabo- less ‘tectonic’ (see the debate between Parker, rated by Jonathan Potter and Margaret 1990 and Potter et al., 1990) and more con- Wetherell (1987; see also Wetherell and Potter, cerned with the performative aspects of every- 1992; 1998) in their attempt to develop forms day talk. The use of the word repertoire is of discourse analysis for social psychology (see apposite and a key notion here is the idea of an Edley, 2001, for an extended discussion of the interpretive repertoire as like a choreographed history and use of the term with examples). dance. People have available a set of familiar Whereas the social psychology of beliefs, cultural resources which they can use to make attitudes and public opinion had worked on sense, but these resources are highly flexible, the premise that people held consistent, stable they need to be done anew in each new and enduring attitudes, Potter and Wetherell discursive context, they are a set of steps which noted the pervasive variability, flexibility and can be put together creatively in new routines.

154 I-Jupp-3373.qxd 3/8/2006 5:45 PM Page 155

INTER-SUBJECTIVE UNDERSTANDING

There are some clear parallels and overlaps Wetherell, M. (1998) ‘Positioning and with Bakhtin’s notions of reported voice and interpretive repertoires: conversation analy- genre and with the emphasis in that tradition sis and post-structuralism in dialogue’, on the ways in which words move across con- Discourse and Society, 9: 387–412. texts from voice to voice. In recent years, the Wetherell, M. and Potter, J. (1992) Mapping concept has been used to signal a more critical the Language of Racism: Discourse and the form of discourse analysis in social psychology. Legitimisation of Exploitation. London and In contrast to conversation analysis, for New York: Harvester Wheatsheaf and instance, this work on interpretive repertoires Columbia University Press. Wetherell, M. and Potter, J. (1998) ‘Discourse has sought to make connections between analysis and the identification of interpretive patterns in talk and the broader social context, repertoires’, in C. Antaki (ed.), Analysing and the ways in which locally realized argu- Everyday Explanation. London: Sage. mentative threads implicate discursive history (Wetherell, 1998).

Margaret Wetherell INTER-SUBJECTIVE UNDERSTANDING Associated Concepts: constructionism, con- Definition versation analysis, deconstruction, discursive psychology, epistemology, ontology, realism, Understanding means having a sense of, grasp- relativism ing, or comprehending the meaning of something such as a text or human action. An understanding is inter-subjective when it is Key Readings accessible to two or more minds (subjectivities). Billig, M. (1991) Ideology and Opinions: Studies in Rhetorical Psychology. London: Distinctive Features Sage. Billig, M., Condor, S., Edwards, D., Gane, The notion occupies a prominent place in M., Middleton, D. and Radley, A. (1988) social scientific research in three wide-ranging Ideological Dilemmas: A Social Psychology ways. First, inter-subjective understanding is of Everyday Thinking. London: Sage. generally accepted as the basis for the objectiv- Edley, N. (2001) ‘Analysing masculinity: inter- ity of judgements and claims made in research. pretive repertoires, ideological dilemmas Many social researchers accept a fallibilist epis- and subject positions’, in M. Wetherell, temology – that is, they hold that there is no S. Taylor and S. J. Yates (eds), Discourse as absolutely secure foundation for knowledge; Data: A Guide for Analysis. London: Sage. hence all claims to know are always in princi- pp. 189–228. Gilbert, G. N. and Mulkay, M. (1985) Opening ple uncertain and corrigible. Fallibilists hold Pandora’s Box: A Sociological Analysis of that because knowledge claims never literally Scientist’s Discourse. Cambridge: Cambridge mirror or represent an independently given University Press. reality, objectivity cannot inhere in the results Parker, I. (1990) ‘Discourse: definitions and of inquiry (that is, in the claim itself). A knowl- contradictions’, Philosophical Psychology, edge claim cannot be objective by virtue of the 3: 189–204. fact that it expresses the way things ‘really are’ Potter, J. and Wetherell, M. (1987) Discourse because we can never know that for certain. and Social Psychology. London: Sage. Rather, judging whether a given knowledge Potter, J., Wetherell, M., Gill, R. and claim is objective is a matter of appraising Edwards, D. (1990) ‘Discourse: noun, verb the conduct or procedure of inquiry. Such a or social practice?’, Philosophical Psychology, process appeals to inter-subjectively shared stan- 3: 205–17. dards of scientific behaviour (that is, seeking

155 I-Jupp-3373.qxd 3/8/2006 5:45 PM Page 156

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

out evidence to disprove one’s claim, detaching know, and they focus instead on intra-personal sufficiently from one’s own preferred perspec- models in which knowing and understanding tives so they might be submitted to scrutiny, are hermeneutic, social, transactional and and so forth) as well as rules of evidence and inter-subjective accomplishments. reason that are accessible to public scrutiny and inter-subjective criticism (Fay, 1996). Evaluation Second, a somewhat different notion of inter-subjective understanding is a central Three distinct but related kinds of debates tenet of interpretive social theory. Owing, frame the issue of what inter-subjective under- in part, to the influence of phenomenology, standing is and how it happens. The first especially the work of Alfred Schutz (1899– addresses the question of whether understand- 1956), interpretivists of various schools of ing human beings is methodologically different thought (ethnomethodology, hermeneutics, than understanding the physical world; in social constructionism) hold that the lifeworld – other words, whether there is something the everyday world of objects, practices and unique or special about inter-subjective under- people in which we find ourselves – is inter- standing (or Verstehen) as opposed to causal subjectively (that is, socially) constituted. explanation (or Erklären). A second set of This kind of inter-subjectivity is not regarded debates centres on how inter-subjective under- as a contingent aspect of human existence but standing happens; how to provide a psycholog- rather a fundamental, essential (ontological) ically realistic account of the capacity to dimension of what it means to be human. interpret, predict and explain the thoughts and Inter-subjective meanings and norms expressed actions of others. In this controversy, some in languages and practices are ways in which scholars hold that understanding others is prin- we experience ourselves and the actions and cipally a matter of empathy, of putting our- utterances of others. These meanings are selves in their shoes, so to speak (simulation not the property of individual minds per se model). Other scholars hold that understanding but the common property of society and requires an interpretive reconstruction of the constitute the social matrix in which individ- basic beliefs, assumptions and practices of the ual actors find themselves. A significant con- other (hermeneutic model). Still others hold sequence that follows from this view is that that understanding and predicting the behav- the activities of knowing and understanding iour of others unfolds very similarly to the way are not a private affair of mind but rather are a natural scientist explains and predicts the located in the sphere of social activity (prac- occurrence of any event. tice, engagement, or relations with others and A third debate surrounds the claim that the artefacts of one’s world), discourse and inter-subjective understanding is the basis dialogue. for objectivity. The principal issues here are A third significant use of the term follows what constitutes the putatively agreed upon, from the foregoing insight. In social research inter-subjectively shared understandings of sci- concerned with the relationship of mind and entific method; just who is included in and culture (for example, socio-cultural psychology, excluded from the ‘community’ of scientific activity theory, practice theory), attention is inquirers that relies on such understandings; focused on the processes by which people and, are concepts of reason, rationality and come to know the thoughts, intentions, beliefs evidence transcendent or culturally mediated. and mental states of others and how they adjust accordingly (Bruner, 1990; Chaiklin and Thomas A. Schwandt Lave, 1996; Wertsch, 1991). These kinds of inquiries argue that inter-personal (that is, com- Associated Concepts: causality, construc- putational, private, individualistic, informa- tionism, epistemology, ethnomethodology, tion-processing) models are inadequate for falsification, hermeneutics, methodology, grasping what it means to understand or to ontology, phenomenology, realism, Verstehen

156 I-Jupp-3373.qxd 3/8/2006 5:45 PM Page 157

INTERVIEW

Key Readings Most commonly, interviews are conducted on a face-to-face basis and they can take a Bruner, J. (1990) Acts of Meaning. variety of forms. They can range from infor- Cambridge, MA: Harvard University mal, unstructured, naturalistic, in-depth dis- Press. cussions through to very structured formats Chaiklan, S. and Lave, J. (eds) (1996) Understanding Practice: Perspectives on with answers offered from a prescribed list Activity and Context. Cambridge: Cambridge in a questionnaire or standardized interview University Press. schedule. Questions asked as part of attitude Fay, B. (1996) Contemporary Philosophy of scaling are typical of the latter. Thus at Social Science. Oxford: Blackwell. one extreme interviews can be conducted as Kögler, H. H. and Steuber, K. R. (eds) (2000) purposeful conversations, as in ethnographic Empathy and Agency: The Problem of interviews (Burgess, 1984) and at the other Understanding in the Human Sciences. extreme they involve little interaction Boulder, CO: Westview. between the researcher and the researched. Schutz, A. (1967) Collected Papers I: The An example of the latter is Computer Assisted Problem of Social Reality (ed. M. Personal Interviewing (CAPI), where intervie- Natanson). The Hague: Martinus Nijhoff. wees enter responses into a lap-top computer Wertsch, J. V. (1991) Voices of the Mind: A by ‘self-keying’. Since 1994 this mode of inter- Sociocultural Approach to Mediated Action. viewing has been used in the British Crime Cambridge, MA: Harvard University Survey (BCS) for more sensitive topics within Press. the interview and many large-scale surveys now use this form of interview. In some contexts and for some purposes interviews may be conducted by telephone or by way of electronic communication such as INTERVIEW email or fax. Interviews of this nature are popular for reasons of cost-effectiveness and Definition for the speed of data collection. Telephone A method of data collection, information or interviews are routinely used for the conduct opinion gathering that specifically involves of opinion polls by market researchers. asking a series of questions. Typically, an interview represents a meeting or dialogue Evaluation between people where personal and social interaction occur. However, developments The advantages of interviews are that they in computer and information technology enable the interviewer to follow up and probe have resulted in other formats, for example, responses, motives and feelings and their Internet interviews. potential added value is that the recording of nonverbal communications, facial expres- Distinctive Features sions and gestures, for example, can enrich the qualitative aspects of the data. Interviews are typically associated with both The alternative types of interview are asso- quantitative and qualitative social research ciated with separate and distinct advantages and are often used alongside other methods. and disadvantages. Unstructured interviews Although interviews can be generally defined where the respondent talks freely around a and have some commonality of meaning, topic can produce rich grounded data but can they can vary enormously in terms of the be very time-consuming to analyse and the context or setting in which they are carried potential for bias on behalf of the interviewer out, the purpose they serve as well as how might be increased. The more guided or they are structured and conducted. This gives focused the interview, generally speaking, rise to many different types of interview. the less time-consuming and less problematic

157 I-Jupp-3373.qxd 3/8/2006 5:45 PM Page 158

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

is the analysis. This is due to the more and that which represents the ideal. There is standardized nature of the responses. often therefore a balance or trade-off between However, in opting for the latter form of the ideal form that the interview might take interview there is generally an increased like- and issues related to feasibility. lihood that the researcher might not be asking the most significant questions or unduly structuring responses. Pamela Davies Whilst interviews are often associated with qualitative research, they are not always Associated Concepts: CASIC, elite inter- adopted as the principle research method viewing, ethnographic interviewing, life his- despite their potential usefulness, on practical tory interviewing, narrative interviewing grounds. Some research subjects may fall within categories often classified as ‘hard to reach groups’, for example, the homeless or Key Readings those in prison. People belonging to different Burgess, R. G. (1984) In the Field: An social groups might also require particular or Introduction to Field Research. London: different interview techniques and skills that Allen & Unwin. might deter the use of this method. Thus the Keats, D. M. (2000) Interviewing: A Practical appropriate use of interviews is often compro- Guide for Students and Professionals. mised. In reality interviews might not be fea- Milton Keynes: Open University Press. sible or where interviews are part of the McCracken, G. (1988) The Long Interview. research design the precise nature of the inter- Beverly Hills, CA: Sage. view might be determined by striking a Mason, J. (1997) Qualitative Researching. balance between what is practical and feasible London: Sage.

158 L-Jupp-3373.qxd 3/8/2006 5:46 PM Page 159

The Biographical Interpretive Method LIFE HISTORY INTERVIEWING (BIM) is a sub-type that emerged from techniques developed to conduct interviews with Definition survivors of Nazi oppression. BIM has a ther- An approach that uses a form of individual apeutic element and orientation to handling interview directed to documenting the sensitive topics, exemplified by Hollway and respondent’s life, or an aspect of it that has Jefferson’s (1997) research exploring fear of developed over the life course. crime, where BIM was used to elicit stories non-directively in pursuit of subjects’ gestalt Distinctive Features (see Wengraf, 2001). A contrasting application is when detailed Plummer (2001) notes three main types of and extensive data are needed, like case his- ‘life stories’: the naturalistic (stories of a life tories or behavioural records, to enable told in a given culture, unshaped by research statistically based analysis. While still inten- intervention); the researched (specifically sive, the style is more structured. Freedman, gathered for research purposes); and the Thornton and Camburn (1988) offer a ‘life reflexive–recursive (life stories constructed in history calendar’ method for such applica- self-consciousness, being the product of post- tions. Bertaux and Kohli (1984) suggest a dis- modernity’s ‘crisis of representation’). tinction between interpretive and ‘scientific’ The method is long-established in social approaches to life histories. The former history, anthropology, sociology and psychol- focuses on meaning in individual lives, and ogy, and has also become associated with the latter seeks accurate descriptions of life feminist research (Devault, 1990). It is trajectories to uncover processes shaping pat- demanding on respondents and researchers, terns of social relations. as it seeks to go into depth and recovers expe- Cavan’s (1929) pioneering contribution riences across long periods. It can also emphasized the ‘genetic’ dimension of life uncover strong emotions. The interviews are histories, enabling researchers to see how normally time-intensive, sometimes taking attitudes and behaviour were formed. She several all-day sessions. Other than topical noted lack of consensus on analysing life his- breadth and extensive duration, characteris- tory interviews, an enduring concern. tics are shared with other semi- and non- Faraday and Plummer (1979) commend as standardized styles of interview (Atkinson, particularly applicable to life history data an 1998). There tends to be a stronger emphasis exploratory ‘AHFA (Ad Hoc Fumbling on non-directiveness than is usual in semi- Around) approach’ which is used to generate standardized interviewing, with researchers concepts. adopting the posture of the good listener, Bertaux (1995) takes a different tack, using seeking interviewees’ own interpretations of life history data to derive ‘social genealogies’ their experience. to research social mobility where long-term L-Jupp-3373.qxd 3/8/2006 5:46 PM Page 160

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

generational effects are important. The unit must acknowledge the situated character of of observation is not the individual but a set their interpretive work. While life history of life trajectories in kinship groups. Bertaux accounts are cultural constructions, aware- contrasts social genealogies and the more ness of this stimulates analytic depth. As to limited information derivable from statistical the relationship between life history and surveys. Frank’s (1979) phenomenological wider social forces, accounts based on more approach disbars such systematic applica- ‘objective’ data downplay the role of social tions. For Frank, the life history is a text actors’ agency in favour of structural forces, resulting from a collaboration involving the to which life history data is a corrective. consciousness of researcher as well as sub- There cannot be a ‘pure’ life history. Any life ject. We always approach life histories with history is a cultural construction. The point is the core question of how this person com- to weigh the data with this in mind (Yow, pares to ourselves. 1994). Ferrarotti (1983) offers an intermediate approach, where life history is ‘history from below’, a corrective to the elite versions domi- Nigel G. Fielding nating historical texts. The method connects individual biography with structural charac- Associated Concepts: critical ethnography, teristics of historical situations. Although tra- diary, ethics, ethnographic interviewing, fem- ditional biographic methods prefer supposedly inist research, historical analysis, idiographic, ‘objective’ secondary materials, Ferrarotti interview, narrative interviewing, oral history argues that the ‘explosive subjectivity’ of primary data should be prioritized. Like Bertaux, he suggests that every individual act is the Key Readings totalization of a social system, but commends Atkinson, R. (1998) The Life Story Interview. a ‘biography of primary groups’ because such London: Sage. groups are not reducible to networks of socio- Bertaux, D. (1995) ‘Social genealogies com- economic interactions. mented on and compared’, Current Sociology, 43 (2–3): 69–88. Bertaux, D. and Kohli, M. (1984) ‘The life Evaluation story approach’, Annual Review of Plummer (2001) details ethical issues relating Sociology, 10: 215–37. to the method. As he noted in Faraday and Blackman, M. B. (1992) ‘The afterlife of life Plummer (1979), there is concern about history’, Journal of Narrative and Life exploiting rapport for research purposes. For History, 2 (1): 1–9. example, an empathetic life history interview Bowes, A., Dar, N. and Sim, D. (1997) ‘Life histories in housing research’, Quality & may confirm the respondent in a deviant Quantity, 31 (2): 109–25. identity. Cavan, R. (1929) ‘Interviewing for life his- Much commented on is the idea that life tory material’, American Journal of history enables interviewing interviewees to Sociology, 15: 100–15. ‘speak through’ published accounts to their Devault, M. (1990) ‘Talking and listening from own community. For Blackman (1992), the women’s standpoint’, Social Problems, issues here are construction of the self 37 (1): 96–116. through narrative and the cross-cultural dia- Faraday, A. and Plummer, K. (1979) ‘Doing logue of narrator and interviewer. life histories’, Sociological Review, 27 (4): Bowes, Dar and Sim (1997) maintain that 773–98. the three main criticisms of the method are Ferrarotti, F. (1983) ‘Biography and the disguised strengths. Bias is the other side of social sciences’, Social Research, 50 (1): the coin of reflexivity, where researchers 57–80.

160 L-Jupp-3373.qxd 3/8/2006 5:46 PM Page 161

LIKERT SCALE

Frank, G. (1979) ‘Finding the common clear as possible. These statements are then denominator: a phenomenological critique administered to a sample of an appropriate of life history method’, Ethos, 7 (1): 68–94. population, as a part of the test construction. Freedman, D., Thornton, A. and Camburn, Once the scale has been checked for internal D. (1988) ‘The life history calendar: a consistency and reliability it can be applied to technique for collecting retrospective the research population. data’, Sociological Methodology, 18: 37–68. A Likert scales employs a graded response Hollway, W. and Jefferson, T. (1997) ‘Eliciting to each of the statements, and these are usu- narrative through the in-depth interview’, ally five in number: Strongly Agree (SA), Qualitative Inquiry, 3 (1): 53–70. Agree (A), Undecided (U), Disagree (D) and Plummer, K. (2001) ‘The call of life stories Strongly Disagree (SD). Each statement is in ethnographic research’, in P. Atkinson, then scored according to the meaning of the A. Coffey, S. Delamont, J. Lofland and L. Lofland (eds), Handbook of Ethnography. statement. Where the statement is in favour London: Sage. pp. 395–406. of the attitude in question, the scores will be Wengraf, T. (2001) Qualitative Research 5(SA), 4(A), 3(U), 2(D) and 1(SD); where it is Interviewing. London: Sage. against the attitude in question, the scores Yow, V. (1994) Recording Oral History. for each question will be 5(SD), 4(D), 3(U), London: Sage. 2(A), 1(SA).

Evaluation Likert attitudinal scales are a useful tech- LIKERT SCALE nique, since it is possible to make a research tool that is very attractive to participants, and Definition they can therefore become much more likely to be completed, improving response rates A Likert scale is a summated rating scale used and generalization reliability. for measuring attitudes. The method was As with all scales, the Likert scale poten- developed by Rensis Likert in 1932. tially suffers from the problems of acquies- cence (respondents acquiescing to that which Distinctive Features they believe they should) and social desirability (portraying themselves in what they The first stage in creating a Likert scale is the believe is a desirable light). production of a series of statements express- A good Likert scale is founded on the ing a favourable or an unfavourable attitude following: piloting the original statements towards the concept of interest. For example, on a large sample of participants representa- if the research is concerned with attitudes tive of the intended population; reversing the towards the police a typical statement might appropriate items on scoring; conducting a be ‘Police officers are not biased in carrying reliability analysis; adding more items if the out their duties’. One way of obtaining state- reliability analysis removes too much, and ments for scale construction is to use state- begin piloting again; checking validity to ensure ments from interview transcripts, where the that the scale is measuring the intended topic in question has been addressed. It is construct. also possible to obtain statements from local experts, to use open-ended questions in surveys, to look at magazines and newspapers Jeanette Garwood covering popular culture, or scour the relevant literature for appropriate comments that Associated Concepts: composite measure- can be made into statements from more ment, indicator, scaling, semantic differential academic sources. Statements should be as scale, summated rating scale, Thurstone scale

161 L-Jupp-3373.qxd 3/8/2006 5:46 PM Page 162

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

Key Readings reviewing the existing literature the researcher can observe how other investigators have stud- Anastasi, A. (1988) Psychological Testing, 6th ied the chosen social phenomena, and how edn. New York: Macmillan. valid these methodologies were. By reading Guilford, J. P. (1956) Psychometric Methods. previous studies, the researcher may be influ- New York: McGraw–Hill. Kline, P. (1993) The Handbook of Psycho- enced as to which is the most effective metho- logical Testing. London: Routledge. dological tool to employ. Third, at a more Robson, C. (2002) Real World Research, 2nd immediate level, undertaking a literature edn. Oxford: Blackwell. review will widen the researcher’s knowledge of the chosen area of research and grant a greater command of the research area and relevant issues. Such a command of the existing knowledge will therefore reinforce the validity LITERATURE REVIEW and findings and ‘convince the reader of the legitimacy of your assertions by providing suf- Definition ficient logical and empirical support along the way’ (Rudestam et al., 2001: 57–8). ‘A critical summary and assessment of the Carrying out a literature review is a dis- range of existing materials dealing with tinct skill in which the student must marshal knowledge and understanding in a given field material in a manner that is effective and … Its purpose is to locate the research pro- manageable; the literature review is a selec- ject, to form its context or background, and to tive process. The wealth of existing academic provide insights into previous work’ (Blaxter material is vast, and must therefore be et al., 1996: 110). broken down into only the information relevant to the specific study. Such an approach Distinctive Features has its roots within C. Wright Mills’s The Sociological Imagination (1959). As Mills Kumar (1996) argues that one of the essential states, the ‘craft’ of social thinking and social initial responsibilities of the researcher is to research requires the researcher to compile a locate and review the existing literature that file in which to ‘develop self-reflective habits’ pertains to the research topic. For all social and ‘build up the habit of writing’ (1959: researchers, ‘information search and informa- 197). The contents of this file, argues Mills, tion management skills are the fundamental include ‘ideas, personal notes, excerpts from building blocks of all research projects’ books, bibliographical items, (1959: 198). (Burton, 2000: 137). The literature review This is effectively a guide to the creation of therefore involves a detailed search through a a literature review. As Mills advises, as the range of resources such as books, journals, the sociological file grows, the researcher must Internet and electronic journals, abstracts and take detailed notes and keep accurate biblio- microfiche. Kumar argues that a literature graphical information. This is the foundation review has three functions. First, the literature of a successful literature review. review clarifies and focuses the research ques- The primary steps in undertaking a suc- tion. It helps the researcher understand the cessful literature search are to formulate rele- subject area more fully and identify the differ- vant search terms, or themes that may be used ent theoretical approaches that have previ- within electronic search engines or printed ously been applied. The literature review sources of information (Burton, 2000). For therefore enables the researcher to conceptu- instance, if investigating the impact and status alize the research question and point to gaps in of postmodernist thought, relevant search existing social research into the chosen area of terms might include postmodernism, post- study. Second, the literature review may structuralism, ontology, anti-foundationalism, improve the researcher’s methodology. By post-Fordism, hyperreality, globabalization,

162 L-Jupp-3373.qxd 3/8/2006 5:46 PM Page 163

LOG-LINEAR ANALYSIS

modernism, late capitalism, or discourses; or enable the researcher to identify gaps in the names of specific theorists such as existing knowledge and whether the chosen Baudrillard, Lyotard, Jameson and Foucault. research project has already been undertaken. Within such a search, certain articles and A comprehensive literature review will also books will be central to a study and will be ensure that the research project is suitably up- read thoroughly, but no researcher can, or to-date. If no adequate literature review is should attempt to read all of the potentially undertaken, this may mean that on completion relevant material. Burton (2000) suggests that the researcher discovers that he or she has effective search techniques of selected lit- simply repeated existing research and has erature should include surveying journal therefore not contributed anything of signifi- abstracts, the indexes of books, introductory cance to the academic community. paragraphs, concluding paragraphs, graphs, tables, diagrams and photographs, or section or paragraph headings. These will enable the Lee Barron researcher to move quickly through a range of literature examples. Additionally, as the Associated Concepts: exploratory research, process continues, researchers will develop intellectual craftsmanship, Internet research, the ability to scan articles or book chapters research design, writing research and take out the relevant data. In addition to the taking of detailed notes from some sources Key Readings of information, the use of photocopying (copy- Blaxter, L. Hughes, C. and Tight, M. (1996) right issues permitting) also enables rapid How to Research. Milton Keynes: Open consultation with numerous examples of University Press. potentially relevant literature. As the review Burton, D. (2000) ‘Using literature to sup- progresses, and as Mills notes, some informa- port research’, in D. Burton (ed.), tion will not be valid to the chosen area of Research Training for Social Scientists: A study, therefore the ability to quickly assess Handbook for Postgraduate Researchers. the validity of a book, article, chapter or graph London: Sage. is essential. The number of literature examples Hart, C. (1998) Doing a Literature Review. will of course be dependent upon the nature of London: Sage. the topic. If researching postmodernism, or Kumar, R. (1996) Research Methodology: A crime, then there will be a wealth of material Step-by-Step Guide for Beginners. London: uncovered. However, some research topics Sage. may yield very few existing examples within Mills, C. Wright. (1959) The Sociological the literature. If a literature review is properly Imagination. Harmondsworth: Penguin. carried out and few sources are unearthed, Rudestam, Kjell, Newton, E. and Rae, R. this may well underscore the originality of the (2001) Surviving Your Dissertation: A research. Comprehensive Guide to Content and Process, 2nd edn. London: Sage. Evaluation Failure to undertake a literature review prior to carrying out social research may have serious LOG-LINEAR ANALYSIS repercussions as the research progresses because the literature review serves to estab- Definition lish the originality of a chosen research project/ thesis. The literature review may support the A form of multivariate analysis used to research questions raised by the researcher, examine relationships between categorical and assist in establishing a valid, workable variables with a view to modelling these research framework. The literature review can relationships.

163 L-Jupp-3373.qxd 3/8/2006 5:46 PM Page 164

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

Distinctive Features LONGITUDINAL STUDY Log-linear analysis involves the construction of a multivariate contingency table in which Definition all the variables are treated as independent variables. The dependent variable is treated Any social or developmental research involv- as the number of cases found in each cell in ing collection of data from the same individ- the table. Analysis proceeds by examining the uals (or groups) across time. Observing degree of fit between the predicted or change in these individuals gives a better expected frequencies in cells and the actual basis for causal inference than a cross- frequencies. The closer the fit the better the sectional study, because of the temporal explanatory power of the theoretical model sequencing involved. In this sense the longi- which has been proposed. This is because the tudinal study (without manipulation of the model can accurately predict the values sample) is a form of ‘quasi-experimental found in the cells in the multivariate contin- design’. gency table. One variant is logit analysis, which facili- Distinctive Features tates an examination of relationships between Longitudinal studies can range from repeated two or more categorical independent vari- measures of a treatment group and a control ables and a dichotomous dependent variable group measured at two time points in an and to predict values of the dependent vari- experimental design, to a large-scale long- able from these independent variables. term birth cohort study, involving follow-ups of the same sample of individuals from birth Evaluation through to adult life. The distinctive feature Log-linear analysis is similar to the chi-square of such work is prospective data collection, test of independence between two variables that is to say, information is collected about which also examines the degree of fit individual sample members to assess current between predicted cell frequencies and behaviours, feelings and circumstances; observed cell frequencies. However, log-linear though retrospective information covering analysis is much more advanced because it the period between ‘sweeps’ (also called can deal with complete patterns of relation- ‘waves’) may also be collected. Intervals ships and interactions between several vari- between sweeps can range from daily contact ables simultaneously. with sample members to periods of several years. Such longitudinal data support analysis Victor Jupp concerned with predicting likely outcomes from earlier circumstances and experiences Associated Concepts: causal model, chi- and contribute to explaining outcomes in square test, discriminant function analysis, terms of situations earlier in life. Both general linear modelling, multivariate analy- approaches enable models to be estimated in sis, path analysis which statistical controls can be placed on potential biases due to ‘confounding’ variables and to ‘selection effects’. (Thus, in Key Readings concluding that a given experience such as Byrne, D. (2002) Interpreting Quantitative higher education leads to a given outcome, Data. London: Sage. we need to be sure that people who had the Sapsford, R. and Jupp, V. (1996) Data experience can be equated in terms of past Collection and Analysis. London: Sage. circumstances of experiences to those who Wright, D. B. (2002) First Steps in Statistics. have not had the experience.) Causal infer- London: Sage. ence is further strengthened by the use of

164 L-Jupp-3373.qxd 3/8/2006 5:46 PM Page 165

LONGITUDINAL STUDY

repeated measurements of key variables so of the population are a powerful tool in social that relationships can be ‘conditioned’ on and developmental science. They may be seen baseline measures. at one level as comprising a sample of life his- The major problem with prospective longi- tories offering enormous potential at individ- tudinal studies is attrition, that is to say exit ual and aggregate level for identifying and from the study of sample members brought modelling sequences of cause and effect rela- about by such factors as mobility, death, tionships in people’s lives. For them to be changing name through marriage, or refusal. effective, however, an infrastructure is needed However, one advantage of the longitudinal to support their continuation, and especially study is that in principle the distributions of the maintenance of contact with study mem- key variables in the original sample can be bers between surveys. In large and complex restored by differential weighting of the examples, data also need to be formatted and sample in later sweeps. The related problem documented in such a way as to supply easy is missing data, occurring at question level, access for users. The major large-scale national which can seriously reduce the size of the longitudinal studies are typically deposited in data set for analysis across different sweeps. data archives under such conditions. The huge Statistical methods of imputation are often investment of resources that has to go into a used to estimate missing values in these longitudinal study to ensure its continuation instances, but none is able fully to address means such studies are still relatively rare in the possible biases brought about by this social science. form of sample loss. The other problematic feature of longitudinal studies concerns measurement, including John Bynner its reliability and validity. Random measurement error can have escalating effects in lon- Associated Concepts: attrition, causal gitudinal analysis, with the possibility of model, causality, cohort study, cross-sectional drawing wrong inferences from estimated survey, panel study, prospective study, retro- relationships between variables. Validity spective study issues arise particularly in connection with memory loss of past events occurring in the intervals between sweeps and ‘fading rele- Key Readings vance’. The design of surveys early on in the Alwin, D. F. ‘Taking time seriously: studying series reflects scientific concerns prevailing at social change, social structure and human the time and is unlikely to anticipate the con- lives’, in P. Moen, G. M. Elder and K. cerns that may take over later. Lüscher (eds), Linking Lives and Contexts: Methods of dealing with these problems Perspectives on the Human Ecology of include maintenance of contact with cohort Development. Washington, DC: American members between surveys, tracing opera- Psychological Association. tions through national databases to locate Giele, J. Z. and Elder, G. H. (1998) Methods their current whereabouts, and expert advice of Life Course Research. London: Sage. on variable coverage to ensure the long-term Magnusson, D. and Bergman, L. (1990) Data value of the data collected in the study. Quality in Longitudinal Research. Cambridge: Cambridge University Press. Evaluation Magnusson, D., Bergmann, L., Rüdinger, G. and Törestad, B. (1994) Problems and Despite their flaws, longitudinal studies on a Methods in Longitudinal Research. large scale and from a representative sample Cambridge: Cambridge University Press.

165 M-Jupp-3373.qxd 3/8/2006 10:43 AM Page 166

interchangeably. In the past market research MARKETING RESEARCH has tended to refer to research into the characteristics of a single market, while marketing Definition research is directed at finding out about the The systematic collection, analysis and full range of an organization’s activities in dissemination of information about the exter- that market. These would normally include nal environment, consumers and customers analysing trends, marketing planning and to enable organizations to keep in touch with evaluation, as well as problem solving. the people who use their products or services Marketing research that supports the plan- and to make informed decisions about their ning function of an organization aims to short-and longer-term objectives. collect and analyse data on consumer trends and customers, on market size and on the Distinctive features channels of distribution in order to support key management decisions. Problem-solving Marketing research is a form of applied studies in marketing research will generally research with a practical purpose. It is a ‘per- focus on one or more aspects of the market- vasive activity’ (Churchill, 1999) that can ing of the product or service and whether it take many forms. It is used in the public needs changing, in order to meet the organi- sector and nonprofit organizations as well as zation’s objectives. It might involve a study of in business. Its main purpose is to help pricing, for example the impact of a change in managers to make better decisions that have prices on sales or on competitors. Other prob- an impact across the organization, not just in lem-solving studies might focus on the marketing. The particular focus of marketing design, use or packaging of the product, or research is on consumption and human and the effectiveness or otherwise of its distribu- organizational behaviour and it draws freely tion channels. Studies of the reactions of from academic disciplines in the social consumers to advertising campaigns or of the sciences, statistics, sociology, psychology and efficacy of particular promotions are also social anthropology. common. The information collected through marketing Three research designs are commonly research links the consuming public to man- distinguished for marketing research agers, who use it to identify marketing oppor- (Malhotra and Birks, 2003). These are tunities; to warn of problems; to evaluate exploratory, descriptive and causal research marketing activities; and to monitor the per- designs. Exploratory research designs enable formance in the market place of their own the development of theories that may be and other organizations. Marketing research tested later in bigger studies or over a longer is sometimes distinguished from market time period. An exploratory study might, for research, though the terms are also used example, be looking for possible reasons why M-Jupp-3373.qxd 3/8/2006 10:43 AM Page 167

MARKETING RESEARCH

sales appear to have peaked in a particular rather than in the past where too often it was area and then declined, or why a re-launched buried in voluminous printed reports. service has not yet reached its predicted level A large international marketing research of customer awareness. A series of focus industry has grown up, dominated by some groups, some depth interviews and the analysis very big companies. The size of some of the of secondary data might be sufficient to firms in the marketing research industry explore the issues and suggest possible deters many potential clients for fear of the avenues for further research. Alternatively, costs involved, and the belief that they will they may be sufficient in themselves to be sold dull, packaged solutions to marketing answer a research question. problems, which may not be what the client Descriptive research designs tend to be needs (McGivern, 2003). The sheer scale of quantitative and typically use surveys or the industry means that there is something of observations to profile consumers, collect a divide between the academic approach to data on the price and availability of products marketing research, with its stress on devel- in store, or to monitor customers’ exposure oping theory, and the practitioners, with their to a promotional campaign. Causal research stress on client-oriented solutions. designs are used when the organization While marketing research may be very commissioning the marketing research wants good at analysing trends that have already to identify the trigger for a change in sales or happened, it is not necessarily a good predic- in customer behaviour. Causal studies there- tor of the future, which is really what busi- fore investigate the relationship between two nesses want to know about. As a result, or more variables. They try, for example, to marketing research studies can help people measure the effectiveness of an advertising in organizations to make more informed deci- campaign in changing consumer behaviour, sions but it can never substitute for the making or the impact of a price change. Since the aim of careful judgements about the risks of action of the study is to determine the cause of or inaction. Many products and services have something, such studies are carried out in a been launched successfully by ignoring the relatively controlled environment that results of early marketing research. enables the researcher to manipulate one of the variables. The main research method used in these studies is the experiment. Teresa Smallbone

Evaluation Associated Concepts: applied research, auditing, causal model, experiment, Most organizations carry out some kinds of exploratory research, forecasting, modelling, marketing research but the key issue in many organizational research, secondary analysis, of them is how to capture and to use this infor- social survey mation effectively. Sales records, customer complaints, returned goods, personnel records and accounts data can all provide Key Readings valuable secondary data for marketing Churchill, G. A. (1999) Marketing Research research studies, if it is made available in the Methodological Foundation, 7th edn. Fort right form. Business decisions can be very Worth, TX: The Dryden Press. risky and the goal of good marketing research Malhotra, N. K. and Birks, D. F. (2003) is to supply accurate information to help Marketing Research: An Applied Approach, reduce this risk and to improve the quality of 2nd European edn. Harlow: Financial decision making. The use of information Times/Prentice Hall. technology is at last enabling marketing McGivern, Y. (2003) The Practice of Market research to be continuously updated and made and Social Research: An Introduction. available to managers on their computers, Harlow: Financial Times/Prentice Hall.

167 M-Jupp-3373.qxd 3/8/2006 10:43 AM Page 168

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

Nominal and ordinal measurement are MEASUREMENT non-metric; interval and ratio measurement are metric. Most sociological measurement Definition is non-metric, using numbers essentially as Any process by which a value is assigned to convenient ways of summarizing data where the level or state of some quality of an object the differences between observations are of study. This value is given numerical form, qualitative rather than expressed in metric and measurement therefore involves the terms. The level of measurement is of criticial expression of information in quantities rather importance for the use of different statistics than by verbal statement. in social research. Different statistical tests and techniques make different assumptions about the properties of the data being Distinctive Features analysed. Parametric statistics assume that A famous (though contested) dictum by Lord the data are metric; to use such statistics Kelvin, carved on the exterior of the Social on non-metric statistics is incorrect. Non- Science Building at the University of Chicago, parametric statistics may be used with scores maintains that ‘when you cannot measure … that are not exact in any sense, but are nomi- your knowledge is … meagre … and … unsatis- nal or ordinal. Data for most social variables factory’. Measurement is a most powerful (for example, religion, occupation, social means of reducing qualitative data to a more class) are non-metric. Hence non-parametric condensed form for manipulation, processing statistics are of particular importance in soci- and analysis. Measurement is the basic process ological research. that makes quantitative social research possible. Four levels of measurement are usually distinguished. First, nominal measurement, Evaluation which involves using numbers to represent The strengths of sociological measurement categories, and permit more efficient summa- are displayed in the achievements of quanti- rization of the data. An example would be: tative studies in fields such as social mobility, Woman = 1; Man = 2. Second, ordinal social stratification, status attainment, voting measurement, which in addition to having behaviour and medical sociology. The feasi- numbers represent categories, also permits bility of measurement has been the subject of the ranking of the categories in terms of the enduring controversy. Critics ask the funda- possession of some quality. An example mental question, ‘If social action is subjec- would be the scale in attitude research: tively meaningful, how can it be reduced to Strongly agree/Agree/Neither agree nor dis- quantities?’ In defence, proponents of quanti- agree/Disagree/Strongly disagree. Third, inter- tative research maintain that the measure- val measurement shares the characteristics of ment of objective characteristics is relatively ordinal measurement, and in addition, the straightforward, while the measurement of property of a unit (or scale) by which differ- subjective states (for example, using scales of ences in magnitude can be expressed pre- attitudes) is feasible and produces reliable cisely. Examples would be: a temperature results. scale in physics; number of years spent in full-time education. Fourth, ratio measurement shares the characteristics of interval Martin Bulmer measurement, plus the property that the scale has a true zero. Examples would be: a Associated Concepts: coding, composite temperature scale starting from absolute zero measurement, parametric tests, qualitative (−273 degrees centigrade); an income scale research, quantitative research, reliability, can also be treated as a ratio scale. scaling, validity of measurement

168 M-Jupp-3373.qxd 3/8/2006 10:43 AM Page 169

MEDIA ANALYSIS

Key Readings messages or signals over distance, it used wires supported by poles. Messages were sent via an Duncan, O. D. (1984) Notes on Social electrical connection and were decoded in Measurement: Historical and Critical. New much the same way as they were discharged. York: Russell Sage Foundation. While he was not the first writer to expose Lieberson, S. (1985) Making It Count: The Improvement of Social Theory and Research. this approach as naïve, Marshall McLuhan Berkeley, CA: University of California was its singlemost influential critic. His Press. Understanding Media, published in 1964, commanded almost biblical reverence. Written shortly after a decade that had witnessed the astonishingly rapid rise of television, McLuhan’s thesis became known through its MEDIA ANALYSIS central maxim ‘the medium is the message’. Television’s influence was due almost Definition wholly to its direct style of presentation; the The examination, interpretation and critique actual content of the programmes was irrele- of both the material content of the channels vant. It seems an elementary observation of media of communication and the struc- now, though, in the early 1960s it was icono- ture, composition and operations of corpora- clastic. Any medium, because it is an instru- tions that either own or control those media. ment, modifies our understanding of the Media, in this sense, refers to what used to be world and, by implication, our experience called the mass media, the means of commu- of it. We make sense of material in different nication of information to large numbers of ways, depending on whether we read news- people – television, radio and newspapers. It papers, listen to the radio, surf the Web or now encompasses multimedia, the electronic watch television. networks of communication made available McLuhan’s stress on the significance of the by the Internet. medium itself, the context in which it is sent and, importantly, the manner in which it is Distinctive Features received, effectively diverted analysts’ attention away from the sheer content of media and Media considered worthy of investigation towards the particular relationships human are typically those that have the potential for subjects had with the various media. While disseminating influential ideas, whether in content analysis of, for example, newspapers the form of information, entertainment, or proceeded inductively, accumulating large rhetoric (designed to persuade, or influence). banks of data that built into a picture of how These include: purveyors of words, such as media portray information about the world, books, newspapers (and analogous print they did not present conceptual analyses media) and radio; those concerned with the capable of informing critical evaluation. combination of words and images, especially Similar criticisms were raised about the television, film and advertising; and, to a school of thought originating in the research lesser extent, conduits of images, including of Paul Lazarsfeld and Robert Merton in the visual art forms. 1940s (most clearly summarized in 1948). Traditionally, the systematic study of Our reliance on mass media of communica- media was based on the premise that there tion was surrendering us to a form of passiv- were three components to communication: ity, argued Lazarsfeld and Merton. Their producers, receivers and, between them, research concluded that the mass media had the linking medium. Telegraphy was the a ‘narcotizing’ effect on us, ‘transforming the exemplar, literally as well as metaphorically. energies of men from active participation into An apparatus for transmitting electrically passive knowledge’.

169 M-Jupp-3373.qxd 3/8/2006 10:43 AM Page 170

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

While the critical sting in the tail was Contemporary media analysis encompasses salutary, the theoretical model on which it many features of earlier work, much of it was based was found faulty. Like recipients concerned with the quantitative exposition of of narcotics, consumers of print and auditory biases in media reportage. Textual analyses media were considered helpless recipients concentrate on the exploration of how we rather than active contributors. New and, read media. Perhaps the most neglected area occasionally hysterical, evaluations and prog- of media analysis is one of the most impor- nostications about its impact accompanied tant. The media’s close associations with television’s growth period in the 1950s. commerce have led to corporate realignments Research showed its effects were manifold. that have massively affected the content of Neil Postman’s Amusing Ourselves to Death media, the manner in which it is packaged to (New York, Viking Penguin, 1985) argued that cater for global audiences and the influence it the arrival of television constituted the begin- has on political regimes. ning of a new age in humanity. The break with previous forms of mass communication heralded a different type of relationship. Ellis Cashmore Humans engaged differently when staring at a screen than when reading a script or listening Associated Concepts: content analysis, to sounds. Enclosed in Postman’s analysis was cultural research, document analysis, netnog- a warning about the functions of television, or raphy, textual analysis, videography, visual more specifically function – there was only methods one: amusement. Television’s public service utility was strictly of secondary importance. Key Readings The title of Postman’s book served notice. Barker, C. (1999) Television, Globalization Critics not only of Postman but of the whole and Cultural Identities. Milton Keynes: doomwatch tradition insisted that media Open University Press. analyses had laboured with a flawed concep- Fiske, J. (1987) Television Culture. London: tion of the subject as having their critical Routledge. abilities blunted and their powers of reasoning Giles, D. C. (2002) Media Psychology. diminished by exposure to television. Viewers Hillsdale, NJ: Lawrence Erlbaum. merely succumbed to the overbearing power Lazarsfeld, P. and Merton, R. (1948) ‘Mass of the media, so that they became mere communication, popular taste, and orga- absorbers of messages rather than creators nized social action’, in L. Bryson (ed.), of them. The Communication of Ideas. New York: With the rise of cultural studies during the Free Press. 1980s, media analysts sought to restore some Miller, T. (ed.) (2002) Television Studies faculties to the human subject. Consistent London. London: British Film Institute. with the intellectual turn of the period, media analysis focused on how media are read by subjects. In other words, rather than explore the content, the channel and the impact, MESSY RESEARCH analysis centred on how a subject interprets, decodes and, generally, makes sense of the Definition media. In Television Culture, John Fiske’s interest lay not in the messages of the media them- Research that does not follow the tidy pattern selves, but in how subjects produce meanings generally recommended and which aims to from them. The content of the media is evaluate an intervention programme as a rendered text – it is comprehensible. But, the whole, as opposed to a specific focused social ways in which we comprehend is what intervention. Such research is sometimes also concerned Fiske and like-minded writers. called complex evaluation.

170 M-Jupp-3373.qxd 3/8/2006 10:43 AM Page 171

MESSY RESEARCH

Distinctive Features resettlement packages; there had been evidence that that such people were dispro- A ‘tidy’ design for an evaluation project would portionately represented among the homeless be one that demonstrated with as little ambi- and rough sleepers. For practical reasons the guity as possible that a given social interven- programme ran as a housing service for all tion did or did not have the desired effect. discharged people who were not rehoused Typical of this kind of design is the field exper- straightforwardly by the armed forces. This iment or quasi-experimental analysis – making gave a range of aims for the research, neces- an intervention in one area or group and sitating the use of different kinds of data not another (with the participants allocated collection. For example, statistical monitoring randomly to treatment or control group, in the of the clients was the first step – who they case of the field experiment) and comparing were (and specifically, whether they fell into the outcomes. Even here it would probably be the original target group) and whether they desirable to use a range of methods, to assess were found accommodation. In the process experiential side effects as well as its gross we were able to note (1) that a number of effect on output statistics, but in principle it clients referred themselves inappropriately is possible to ‘design out’ sources of error or (the agency dealt only with single personnel) control for them in analysis. but that these were consistently referred on Increasingly, however, researchers have to to a more appropriate agency, (2) that some employ a mix of methods in a comparatively solutions were regarded by the agency and unfocused way when carrying out evalua- the client as at best temporary but that these tions. This can be for a number of reasons. files remained open and work continued, and First, the programme itself may be complex (3) that where time was short before dis- and heterogeneous – funded social interven- charge the solution was less likely to be an tion now often involves a ‘basket’ of projects immediately satisfactory one. which bear only a loose thematic relationship Another obvious strand to the research to each other. Second, while it is possible to was to monitor the establishment and devel- assess the outputs and even the impact of opment of the agency, to report back to the particular projects, their contribution to the funders on how their money was being spent, overall programme ‘theme’ or ‘rhetoric’ may to identify possible areas for improvement be less easy to establish. Third, the audience and possibly to establish lessons for other for funded evaluations is generally heteroge- similar projects. neous, with each fraction of it interested in a Whether the ‘solutions’ were successful different kind of outcome and convinced by a was another strand. The method adopted, in different kind of evidence. Finally, social view of resources being limited, was to follow interventions often occur in areas that are up each case by telephone interview, to see targeted by a range of providers; some cities whether the arranged accommodation had may be in receipt of government, charitable been taken up and whether it was regarded or other funds for many more than one inter- as satisfactory in the long or short term. We vention project. It is therefore not at all easy – were able to gain further information from often not possible at all – to separate the the agency’s records, because part of their effects of one programme from those of mission was to follow up on placements another. which they regarded as at best a tempo- For example, the present author recently rary solution and to remain available to carried out an evaluation of a ‘rehousing’ discharged clients who needed further advice programme targeted in the first instance at or help some time after discharge. vulnerable young people discharged from Client satisfaction was another obvious the armed services before they had finished issue – with the service before discharge, their training (generally, having failed it) and with the accommodation provided and with before they became eligible for the full military the outcome of any further approaches to the

171 M-Jupp-3373.qxd 3/8/2006 10:43 AM Page 172

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

agency after discharge. This was tackled Key Readings by face-to-face interviews with a sample of Jupp, V., Davies, P. and Francis, P. (eds) people awaiting discharge and telephone (2000) Doing Criminological Research. interviews, over a period, with discharged London: Sage (especially chapters by Nick people. We also issued a brief ‘satisfaction Tilley and Iain Crow). questionnaire’ in an attempt to increase the Thomas, A., Chataway, J. and Wuyts, M. numerical base for judgement of satisfaction, (eds) (1998) Finding Out Fast: Investigative but the response rate was too low for this to Skills for Policy and Development. London: be useful. Sage. We also interviewed the staff of the agency (because an agency cannot be counted an unqualified success if its staff are not satisfied with it), key ‘players’ among the referring agencies and a small sample of local META-ANALYSIS authority departments which would otherwise have been housing these people. The Definition interviews gave insight into how the work of the agency fitted into the forces’ discharge The synthesizing of existing research studies processes and how it helped outside depart- related to a specific research hypothesis. ments to fulfil statutory requirements. Here the findings from the studies become Our interviews with housing and social the inputs to, and cases of the analysis, rather services departments, and the ‘literature’, than the resultant output. That is to say, they enabled us to make very rough estimates of become the data upon which the study is what the cost of discharging these people going to be based. Meta-analysis has been a homeless might have been, which gave us a major method of investigation in the area basis on which to talk about value for money of health care but the principles can be utilized and the wider concept of ‘best value’ (which in any area of research. The methodology takes into account human as well as financial can utilize both quantitative and qualitative costs). studies.

Evaluation Distinctive Features Messy research does not provide a clear-cut Glasziou et al. (2001) give an overview of how assessment of the outcome of one interven- to undertake a systematic review. They note tion as, say, in a field experiment or quasi- the following discrete steps: first, question experiment. However, as the above example formulation; second, finding studies; third, illustrates, it does permit more complex eval- appraisal and selection of studies; fourth, uation. By applying a wide range of methods, summary and synthesis of relevant studies; some well designed and some more ‘scrappy’ and finally, determining the applicability of and opportunistic, it was possible to produce results with regard to question formulation. information useful for all the different players It is important to ask the right question. involved. Perhaps this is more important in meta-analysis than in the more conventional forms of research. This is because the question one Roger Sapsford asks has implications for knowing which studies to include in the analysis, the accu- Associated Concepts: applied research, eval- racy of the analysis and the confidence in and uation research, field experiment, method- validity of the conclusions. This takes on ological pluralism, mixed-methods research, greater importance when the study has an policy-related research, quasi-experiment element of forecasting or prediction.

172 M-Jupp-3373.qxd 3/8/2006 10:43 AM Page 173

META-ANALYSIS

Finding the studies involves a search for can be further downgraded as a result of relevant literature and sources of data. In including poor-quality studies alongside high- recent years this has become less daunting quality studies. To a certain extent this may with the evolution of the World Wide Web, be nullified by choosing the correct filter online catalogues (OPACs), online databases rules to apply when thinking which studies to and CD-ROMs. include and which to exclude. Appraisal and selection of studies may not Thirdly, it is argued that the results that tend seem like a daunting task but the reality is to get published are those that have somewhat different. It is necessary to decide statistically significant findings. Non-significant which studies are going to be included and, results are not given such prominence but per- more importantly, why one is included and haps should be included in any meta-analysis. another is not. Clearly this relates directly with It may be argued that the goal of social the research question or hypothesis under research is equally met when a relationship is consideration. The inclusion of articles must be found not to exist. Such findings still add to based upon some systematic rules, which need our knowledge. The potential for bias resulting to be derived. Quality and bias are key issues from such omission can be reduced by looking in this area, as both have a direct effect on the at articles that have not been published. Due validity of the subsequent meta-analysis. care and attention must be exercised at this The summary and synthesis of relevant point to ensure that any such study is not studies tends to be based upon a quantitative inherently flawed in some way and that this approach, however, a qualitative methodology was the reason for non-publication. can yield important results. Combined tests Finally, regarding independency of results, it use a range of procedures (from simple is not unusual to see multiple results used from counts to summation tests based upon signif- the same piece of research. This may make the icance levels and/or weighted tests) to estab- results of the meta-analysis appear more valid lish the statistical significance of combining and generalizable than they actually are. the various studies. Such tests do not look at the strength of the relationship and therefore Paul E. Pye it is prudent to use the combined test in conjunction with measures of the size of the Associated Concepts: effect of combining studies. applied research, eval- With regard to determining the applicability uation research, good practice studies, of results questions need to be asked as to the methodological pluralism, mixed-methods validity of the synthesis, the degree of confi- research, policy-related research, validity dence in the results, generalizability, and prediction and forecasting. Key Readings Edgger, M., Davey Smith, G. and Altman, D. Evaluation G. (eds) (2001) Systematic Reviews in Health Care: Meta-analysis in Context, 2nd Four main criticisms can be levelled at meta- edn. London: BMJ Books. analysis. These relate, first, to the comparison Glass, G. V., McGaw, B. and Lee Smith, M. of studies. It can be argued that it is impossible (1981) Meta-analysis in social research. to compare studies since one is not compar- Beverly Hills, CA: Sage. ing like with like. Very few studies employ Glasziou, P., Irwig, L., Bain, C. and Colditz, exactly identical methodologies and measur- G. (2001) Systematic Reviews in Health ing systems. Therefore any comparison is not Care: A Practical Guide. Cambridge: logical or valid, and little confidence (if any) Cambridge University Press. can be placed in the resultant discussions. Hedges, L. V. and Olkin, I. (1985) Statistical A second issue relates to the quality of Methods for Meta-Analysis. New York: studies included. The results and discussions Academic Press.

173 M-Jupp-3373.qxd 3/8/2006 10:43 AM Page 174

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

and other quantitative methods. On the other METHODOLOGICAL PLURALISM hand, when more detail was required, or phenomena were complex, subtle, or unclear, Definition this was more suited to research by observa- An approach that advocates flexibility in the tion, less structured interviews, ethnographic selection of social research methods, based description and other qualitative styles. Not on the principle of choosing the most suitable every researcher would use every style of methods for the nature of the problem being research during their careers, nor should they researched. More generally, methodological be proficient in all research methods. The pluralism calls on the researcher to be tolerant plurality would be achieved in the total of other people’s preferred methods even research output of the discipline as a whole when they differ from one’s own. (Bell and Roberts, 1984).

Distinctive Features Evaluation From time to time, disagreements among Despite the common sense of methodological social scientists about which are the ‘best’ pluralism, more sociologists pay lip-service to social research methods become more vocal it than actively adopt it as a philosophy. The and indeed, confrontational. An example of main reason is that during their education this is the competition between older, ethno- and early careers, each researcher acquires a graphic research styles associated with soci- set of personal preferences for one type of ologists at the University of Chicago, and the social science over others. This is not just a (then newer) work based on social surveys question of technical skills, but rather an being promoted by Harvard and Columbia interest in certain topics and a philosophical after World War II. In the UK a conflict view of the social world and how it can be between quantitative methods and several analysed. newer forms of qualitative research took There are genuine differences between place during the 1970s. These disputes are schools of research, from those seeking to usually marked by antagonistic criticisms involve and empower the people being of published work, lengthy expositions in researched, through to those that regard defence of particular methods, and even ‘respondents’ merely as sources of informa- personal abuse. tion, and from those that see the social world Methodological pluralism, promoted by Bell as intricately interconnected and difficult and Newby (1977), rejected the idea that one to ‘know’, to those that concentrate on the type of methods was automatically better than generality of patterned associations between another. They argued that it was healthy for small sets of ‘variables’. While not always sociology to contain a number of different the- consciously returning to the complex social oretical perspectives, and that while each theories that underlie their positions, perspective tended to imply a given method of researchers read mainly a sub-set of the liter- research, each new research project should be ature written by like-minded colleagues, tackled on the basis of its own particular defining research problems in specific ways, features. The research methods selected for the and therefore carry out their research using a project should be the ones that best fitted the narrow repertoire of methods. characteristics of the phenomena being studied. In some cases, this results only in a rather For example, studies of the national rates at focused approach, without much concern for which something was occurring, or projects other approaches. In others, the intellectual dealing with simpler concepts that could be context of the research is strongly associated relatively easily measured, were better suited with a particular method: the context defines to social surveys, pre-coded questionnaires what is worth researching, how it should be

174 M-Jupp-3373.qxd 3/8/2006 10:43 AM Page 175

METHODOLOGY

researched and what order of interpretations Key Readings can be made. In its more extreme form this Bell, C. and Newby, H. (1977) Doing Socio- results in ritualized denunciation of alterna- logical Research. London: Allen & Unwin. tives often becoming a part of publication. Bell, C. and Roberts, H. (eds) (1984) Social Where some researchers adhere to a ‘stand- Researching. London: Routledge & Kegan point’ position, their intention is an explicitly Paul. ideological one which goes beyond just making Payne, G. and Payne, J. (2004) Key Concepts in new discoveries, to the promotion of the inter- Social Research. London: Sage. ests of one particular group. Challenging other Payne, G., Dingwall, R., Payne, J. and researchers’ methods is one way of undermin- Carter, M. (1981) Sociology and Social ing the position of rival interest groups (Payne Research. London: Routledge & Kegan and Payne, 2004: 89–93, 152–7). Paul. Such out-of-hand dismissal because of the Payne, G., Williams, M. and Chamberlain, S. type of methods a study has used is a different (2004) ‘Methodological pluralism in British matter from legitimately debating the compe- sociology’, Sociology, 38 (1): 153–63. tency of its research basis, when that is part of a general evaluation. However, it would be wrong to portray academic life as consisting solely of calm, rational, philosophical debate. Academics also compete for resources METHODOLOGY (research funding, access to journals, tenured posts, career promotion) in just as determined Definition a way as do people in other walks of life. The philosophical stance or worldview that Attacking the type of research methods used underlies and informs a style of research. by rivals is one weapon in the struggle between individuals, and institutions, for Distinctive Features supremacy. Not surprisingly, methodological plural- Methodology is the philosophy of methods. It ism’s failure to recognize these processes has encompasses, first, an epistemology – the meant that its call for toleration has largely ‘rules of truth’ for warranting the validity of gone unheeded. For example, in the UK the conclusions – and secondly, an ontology – sociology that has been published in recent establishing the ‘objects’ about which ques- years has depended heavily on a narrow range tions may validly be asked and conclusions of qualitative methods. A recent study of jour- may be drawn. nal papers found only a minority using quan- In virtually all research (but see below) the titative methods: only 2.6 per cent ‘involved epistemology is ‘rules of science’ – conclusions bivariate analysis and were written by soci- based on arguments cast in terms of informa- ologists at British universities, and only 8 tion that has been collected ‘transparently’ by (3.5%) involved multivariate techniques … known, and in principle, testable and repro- This can hardly be described as methodologi- ducible means. The ontologies of research, cal pluralism’ (Payne et al., 2004: 160). however, reflect the wordview or ‘map’ with which the researcher engages. One position, commonly labelled ‘posi- Geoff Payne tivistic’, takes for granted the nature of the real world and reserves all its doubts for the Associated Concepts: critical research, epis- nature of the evidence about it; problems temology, methodology, mixed-methods within positivism are problems of measure- research, positivism, qualitative research, ment – how to measure the real world most quantitative research, relativism accurately – rather than about its ontology.

175 M-Jupp-3373.qxd 3/8/2006 10:43 AM Page 176

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

In social and developmental psychology this rather than ‘objective’ factors. Ethnographic position is well illustrated by the nature/nurture research is concerned with what addiction debate, where the ‘objects’ are behaviours or means to those involved with it rather than character traits or attitudes – reductionist the number of addicts and with the social aspects of persons – and the physiological or processes undergone when close to death ‘personal history’ elements which give rise rather than with the number of deaths and to them. The question is whether you were born their causes. This position found a second nat- this kind of person, or made this way by social- ural home in social psychology and psy- ization or learning, or made this way by the chotherapy, in the humanistic concern with pressure of the immediate environment. how individuals experience their world and Explanations in terms of reasons or preferences how their understanding of their social world are ruled out as ill formed – the question determines its nature for them. It has also becomes why you are the sort of person who been influential in feminist research on gender has these reasons or preferences, and whether and family relations and in research on other it is because of biology, socialization/learning or disadvantage and prejudice experienced by the environment. What was radical about the other interactionally defined groups such as behaviourist revolution of the middle of the disabled people or minority ethnic groups. last century was its success in exiling one class ‘Qualitative’ research of this kind has of objects from valid use in this context – the developed into a third methodological class of hypothesized structures or processes position, generally characterized as ‘critical’ ‘in the head’. or ‘discursive’ or, more recently, ‘postmod- The equivalent in the earlier history of soci- ern’. Here the ontology still centres around ology was focus on ‘social facts’ by structural meanings, but meaning is differently defined functionalists. Though many of the social and understood – as something historically facts – suicide rates, crime statistics – are constructed within cultures rather than nego- now discredited as measures of anything use- tiated between individuals (though taken up, ful, this kind of explanation in terms of col- used and perhaps modified by individuals lective differences between areas or segments and groups for their own particular pur- of a social nexus and the underlying institu- poses). Even in psychology this has meant a tional/structural mechanism which produces move away from individual understanding to them is still a common form of argument in the materials, available to us in our (multiple) sociology. It was reinforced by the emphasis social worlds, out of which current under- in the more economistic kind of Marxist analy- standing is to be put together. sis on relations of production and economic/ structural power and by the analysis of patri- Evaluation archy, male power processes and family institutions produced by many feminists. The It has been taken for granted, above, that the ‘objects’ of positivistic sociology tend there- epistemological component is the same in all fore to be ‘facts’ about societies or social methodologies. All are concerned with estab- groups and factual descriptions of social insti- lishing ‘truth’ on the basis of transparent tutions and processes. evidence, and all would test and reject propo- Interactionist sociology, a reaction against sitions that were frequently contradicted by structural functionalism, came into social the evidence. If it follows from a given theo- research as ‘ethnography’ or ‘grounded theory’ retical line that such and such should be pos- and made a fundamental impact on its method- sible to the individual, for example, and ology. While the same emphasis is placed individuals most often prove to be unable to on a fundamentally scientific epistemology – do it, then we would question the soundness arguments based on transparent evidence – the of the theoretical line. ontology changed to accommodate a world- However, this line of logic fails when con- view in terms of meanings and understandings fronted with emancipatory theory – theory

176 M-Jupp-3373.qxd 3/8/2006 10:43 AM Page 177

MICROETHNOGRAPHY

that aims to change things. Here what would analysis may be coupled with ethnographic be being tested would not be that individuals methods such as informant interviews and do do such and such, but that they could do it participant observations, all in an effort to if they tried, at least some of the time. This better understand social organizations, prac- claim is not logically refuted by cases where tices and problems. the effect is not achieved, even if this appears to be the invariant outcome; it remains logi- Distinctive Features cally possible that the next individual will succeed. This is a problem for the social Multimedia technology has been an impetus sciences because a great deal of research and for microethnographic research because it theory is directed at changing things – facilitates close analysis of vocal and visible improving individuals’ capacity for dealing behaviour. This approach emerged in the with their lives, undoing power inequalities, 1960s and 1970s, when scholars, concerned at least making people aware of ways in about social inequality within public schools, which their world is currently being shaped used a new technology, called videotape, by factors that are outside their control while to record and analyse people’s ‘micro- they remain unaware of them. behaviours’ within particular classrooms. Researchers focused on the social interaction, rather than the individual. They showed how Roger Sapsford social realities and problems are formed and maintained turn by turn and moment to Associated Concepts: causality, critical moment, through forms of talk and nonver- research, emancipatory research, ethnography, bal behaviours that are the building blocks of inter-subjective understanding, ontology, posi- microcultures. Today, digital and computer tivism, postmodernism, Verstehen technologies support close analysis of video data. For example, Goodwin (1994) con- Key Readings ducted a frame-by-frame analysis of the courtroom activity that led to an acquittal of Hammersley, M. (ed.) (1993) Social Research: the LA policemen who beat Rodney King. Philosophy, Politics and Practice. London: The study showed how defence attorneys Sage (See articles in Part I: ‘Philosophy’). organized the jury’s perception of the video- Plantinga, A. (1993) Warrant: The Current taped beating so that King, rather than the Debate. Oxford: Oxford University Press. policemen, was seen as the aggressor. Willig, C. (2001) Introducing Qualitative Similarly, the author used computers to Research in Psychology: Adventures in analyse the talk and spatial manoeuvres of Theory and Method. Milton Keynes: Open police interrogators who had a reputation for University Press. getting confessions from suspects later found to be innocent. Contemporary microethnographic research represents a convergence of competencies MICROETHNOGRAPHY from various disciplines, including anthropology, psychology and sociology. Careful stud- Definition ies of visible (or nonverbal) behaviour have been influenced by context analysis, a Research that attends to big social issues method pioneered by psychologist Albert through careful examination of ‘small’ com- Scheflen, who closely examined people’s municative behaviours. Analysts study the use of space (relative location, distance, audible and visible details of human inter- orientation and posture) within social situa- action and activity, as these occur naturally tions. Context analysis has been further within specific contexts or institutions; micro- developed by Adam Kendon, who attends to

177 M-Jupp-3373.qxd 3/8/2006 10:43 AM Page 178

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

the sequential unfolding of human interaction at all points’ of social life. Patterns of through time and space. Examinations of talk interaction found within a courtroom have been heavily influenced by conversation might as easily be found in a boardroom analysis, a rigorously empirical method for or a beauty salon. However, site selec- studying the interactive structures and the tion is usually guided by a research social forms that talk embodies. Microethno- agenda. A concern for social stratification graphic studies regularly include conversation within public schools might guide a analytic arguments and analysis, but talk is not researcher to a classroom where children privileged at the expense of other symbol have a variety of socio-economic back- systems and communicative resources that grounds. Whether sites are selected for warrant close examination. basic research or to pursue a particular From a microethnographic perspective, research agenda, communicative phe- vocal and visible behaviours are best under- nomena are examined as inherently stood through their relationship to each other embedded within the local scene. and their social-material surround. This view (2) Collect naturally occurring data. Commu- contrasts with traditional social scientific nication is considered to be ‘naturally research that has artificially separated verbal occurring’ if it would have occurred and nonverbal messages, regarding them as whether or not it was observed or recorded independent rather than co-occurring and by the researcher. Participant observations, interrelated phenomena. When people ges- field notes and audio or video recordings ture with their hands, they usually talk to of everyday interaction are considered someone at the same time, coordinating their premium data for microethnographic visible and vocal behaviours to be understood work, but videotape has become a main- altogether. People occupy and move within stay because it helps analysts avoid an three-dimensional spaces that include physi- artificial separation of verbal and nonver- cal objects and structures, and movements bal channels, and it captures subtle details may be largely recognized and understood of interaction that analysts can review and through their relationship to the material others can verify. world within reach. Also, communication is a (3) Observe data carefully and repeatedly. process of interaction among participants Careful and repeated observation is the who co-author each other’s vocal and visible gristmill for empirical verification. A messages and meanings. Furthermore, talk videotaped moment may ‘jump out’ as and visible behaviours may be embedded obviously noteworthy, but it requires within extended processes or activities such repeated observation to be fully expli- that any particular behaviour is understood cated and understood. Through repeated through its relationship to the whole activity. observation, moments may become a Although microethnographic research focus of analysis as the features and pat- takes various forms, depending on the prefer- terns of interaction are noticed and ences of the analyst and the nature of the appreciated. Throughout this inductive project, the method typically involves the process, analysts’ eyes are unavoidably following steps (see LeBaron, 2004): informed by research agendas and litera- tures. Induction gives way to abduction (1) Select a research site. A site is a location as researchers look for specific kinds within time and space where people of phenomena related to their emerging associate and communicate to constitute claims and conclusions. In the end, something recognizably their own. Basic research claims must agree with what communication research, that is, discov- can be observed in the videotaped data. ering and documenting features of the (4) Digitize and transcribe key moments of interaction order – could be conducted interaction. When computers are used for almost anywhere because there is ‘order repeated observation, analysts can get

178 M-Jupp-3373.qxd 3/8/2006 10:43 AM Page 179

MIXED METHODS RESEARCH

different ‘views’ of an analytic object. participants’ experience during a police Digital video can be manipulated tempo- interrogation’, Human Studies, 20: 1–25. rally (for example, slowed down) and LeBaron, C. and Streeck, J. (2000) ‘Gesture, spatially (for example, zoomed in); dif- knowledge, and the world’, in D. McNeill ferent interactional moments are easily (ed.), Language and Gesture. Cambridge: juxtaposed on the computer screen for Cambridge University Press. comparison. New technology cannot LeBaron, C. (2004) ‘Considering the social replace the eyes and ears of a well- and material surround: toward microethno- trained analyst, but it can support smart graphic understandings of nonverbal communication’, in V. Manusov (ed.), The and rigorous research. Transcription Sourcebook of Nonverbal Measures: Going reduces interaction to a two-dimensional Beyond Words. Mahwah, NJ: Erlbaum. page, which highlights (precisely) struc- Streeck, J. (1983) Social Order in Child tural aspects of talk and its coordination Communication. A Study in Microethno- with nonverbal behaviours. graphy. Amsterdam: Benjamins. (5) Describe and report research findings. Microethnographic studies describe the details of visible and vocal behaviours that subjects performed for each other and thereby made available for analysis. MIXED METHODS RESEARCH Claims may appear on the same printed page as transcripts (featuring vocal behav- Definition iours) and frame grabs (showing nonverbal behaviour) taken from videotaped The combined use of both quantitative and data. Empirically grounded claims may qualitative methodologies within the same be supplemented by ethnographic insights study in order to address a single research and evidence from participant observa- question. tions, field notes, interviews and so forth. Although generalizability is not the sine Distinctive Features qua non of microethnographic work, researchers assume and argue that site- Mixed-methods research draws upon both specific findings have relevance beyond, quantitative and qualitative methodological and that particular patterns of behaviour approaches to answer a particular research resonate with larger social orders. question. For example, a researcher may start by conducting semi-structured interviews and then use the results from this phase to Curt Le Baron formulate specific closed-ended survey questions. This is an illustration of what has been Associated Concepts: conversation analysis, termed sequential mixed-methods research ethnography, ethnomethodology, induction, (Cresswell, 2003), whereby qualitative data transcription, videography collection and analysis is undertaken first, followed by quantitative data collection and analysis (or vice versa). Alternative strate- Key Readings gies within mixed-methods research have Jones, S. and LeBaron, C. (2002) ‘Research on been identified. Cresswell (2003) refers to the relationship between verbal and non- concurrent and transformational strategies. verbal communication: emerging integra- Concurrent techniques involve data collections’, Journal of Communication, 52 (3): tion using both quantitative and qualitative 499–521. approaches simultaneously, for example, LeBaron, C. and Streeck, J. (1997) ‘Space, administering a questionnaire which contains surveillance, and interactional framing of both closed-ended and open-ended questions.

179 M-Jupp-3373.qxd 3/8/2006 10:43 AM Page 180

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

Transformational techniques make use of advocates using whatever ‘works best’ in an overriding theoretical perspective (for any particular research context, and thus example, participatory research) to guide the opens the way for mixed-methods approaches. entire study design and to drive the choice of Pragmatism is a form of methodological methods used. pluralism. Within the three general mixed-methods research strategies outlined above, a number of different approaches can be taken. One ques- Evaluation tion concerns whether both approaches are to As with any approach, mixed-method research be given equal weight, or one approach given embodies a number of advantages and disad- priority, and this will depend on the individual vantages. Among the advantages is the poten- researcher’s goals and orientation as well as tial for gaining a fuller, richer and more the nature of the research question. Another complete understanding of a research question choice concerns the stage, or stages, at which by combining both quantitative and qualitative integration of the two approaches will occur. perspectives. This potential is one of the key In relation to this latter issue, Tashakkori and elements emphasized by pragmatists. Secondly, Teddlie (1998) have defined mixed methods the results from using one approach, strategy research as that which combines quantitative or method may help to guide and inform and qualitative methodologies in a specific another approach or method, as in the case aspect of a study (for example, data analysis) where the results of a semi-structured inter- and contrast this with mixed-model research view provide a useful basis for devising more which combines both approaches at all stages specific questions for a structured survey. in the research process (for example, concep- A potential disadvantage of mixed-methods tualization, data analysis, interpretation). approaches, however, is the often lengthy data Two concepts closely associated with collection and analysis phases required (espe- mixed-methods research are triangulation and cially in sequential designs) leading to heavy pragmatism. Triangulation refers to a research demands on both time and funding resources. strategy that involves approaching a research A further possible disadvantage is the demand question from two or more angles in order placed on the researcher to be expert in to converge and cross-validate findings from the use of both quantitative and qualitative a number of sources (for example, different approaches. Thirdly, the validity of mixed- data sources). A researcher may converge methods research has been called into question self-report data derived from interviews with in debates over the extent to which the under- observational data. In combining both quan- lying paradigms and methods of quantitative titative and qualitative approaches mixed- and qualitative research can be seen as com- methods research embodies the notion of patible. There have been advocates for both the triangulation. ‘compatibility’ and the ‘incompatibility’ theses. Pragmatism is the term given to a particular While these theoretical debates are ongoing, paradigm (or ‘worldview’) which has been a number of researchers across a range of identified as providing a rationale for mixed- disciplines in the social and behavioural sci- methods research. Pragmatism rejects the ences are utilizing and further refining mixed- traditional conception that the paradigms methods approaches. underlying quantitative and qualitative approaches (positivism and constructivism, respectively) are essentially incompatible and Claire Hewson in conflict. Instead, pragmatists argue that both quantitative and qualitative approaches Associated Concepts: messy research, have their own distinctive strengths and weak- methodological pluralism, paradigm, qualita- nesses and can be usefully combined to com- tive research, quantitative research, triangu- pliment one another. Essentially pragmatism lation, validity

180 M-Jupp-3373.qxd 3/8/2006 10:43 AM Page 181

MODELLING

Key Readings to estimate parameters and to define the behaviour pattern. Cresswell, J. W. (2003) Research Design: Modelling formalisms subsume approaches Qualitative, Quantitative and Mixed that support the description of continuous, Methods Approaches. London: Sage. discrete and hybrid systems. A continuous Greene, J. C., Caracelli, V. J. and Graham, W. F. (1989) ‘Toward a conceptual frame- system is one in which state variables change work for mixed-method evaluation designs’, continuously over time (Cellier, 1991). Educational Evaluation and Policy Analysis, Continuous systems are described by differ- 11 (3): 255–74. ential equations. Time base and state are rep- Tashakkori, A. and Teddlie, C. (1998) Mixed resented by real-valued variables. Graphical Methodology: Combining Qualitative and representations, for example block diagrams Quantitative Approaches. London: Sage. or systems dynamics diagrams, facilitate the Tashakkori, A. and Teddlie, C. (2003) development of models. Handbook of Mixed Methods in the Social In a discrete system only a finite number of and Behavioural Sciences. London: Sage. state changes occur during a finite time span (Banks et al., 2001). Discrete systems can be divided into those with a discrete and those with a continuous time base. In the former case the system develops step-wise. Examples MODELLING of formalisms that support the modelling of discrete time systems are cellular automata, Definition Petri Nets and difference equation systems. Formalisms for discrete event systems are The process of organizing knowledge about a based on a continuous representation of given system. The resulting model represents the time base where state changes, that is, a goal-directed simplification of the system events, can occur at any time. Examples for under study, which is achieved by abstraction this type of formalisms are DEVS, timed Petri and aggregation. Nets and queuing models. The state of discrete systems can be described by arbitrarily Distinctive Features scaled variables. Thus, discrete event models typically combine continuous and discrete, In the first step the boundary that separates that is, quantitative and qualitative variables, the system from its environment, its relevant whereas formalisms for continuous sys- components and their interrelations is identi- tem modelling are restricted to quantitative fied. This ‘system identification’ is followed variables. by the phase of model building, a term which Hybrid systems combine continuous and is sometimes used synonymously to model- discrete behaviour (Zeigler et al., 2000). Many ling but emphasizes more strongly the result continuous modelling formalisms allow of the modelling process, that is, the model. the integration of events, for example to Modelling is not constrained to the construc- model the switching from one continuous tion of formal models. However, in the behaviour to the next. Also discrete event for- context of this definition, we will focus on malisms like state charts can easily be models that describe dynamic systems in a extended to capture continuous aspects of a formal language and can be executed by system’s behaviour. computer simulation, that is, dynamic simu- Models may be further classified as being lation models. To develop a dynamic simula- deterministic or stochastic. Models that con- tion model, existing knowledge about the tain no random variables are classified as system and its underlying dynamics and being deterministic. Given a set of inputs the available data are evaluated to determine the simulation run will result in a unique set of component’s variables and their value ranges, outputs. A stochastic simulation model has

181 M-Jupp-3373.qxd 3/8/2006 10:43 AM Page 182

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

one or more random variables as inputs. The the set of experiments for which a model is resulting random outputs are estimates validated. This information should be kept for the true characteristics of the modelled with the model, because each simulation system (Banks et al., 2001). model is always related to the pair: system and Another important distinction is whether experiment. As models are influenced by the systems are described on macro, micro or modeller’s background and his/her perception on multiple levels of organization (Gilbert as well, this pair could easily be extended to an and Troitzsch, 2000). Macro models define n-tuple. one entity of interest. All variables and dynamics are attributed to it. Micro models comprise multiple entities with individual Adelinde Uhrmacher states, but with homogeneous state space and behaviour pattern. Since many phenomena Associated Concepts: causal model, econo- cannot be explained with reference to one metrics, experiment, simulation methods, organizational level only, in multilevel mod- validity els interdependencies between systems and subsystems and different subsystems can be Key Readings explicitly modelled and different abstraction Banks, J., Carson, J. S., Nelson, B. L. and levels can be combined. Both the network Nicol, D. M. (2001) Discrete Event System and hierarchical structure of systems are Simulation. New York: Prentice Hall. taken into account. Depending on the sup- Cellier, F. E. (1991) Continuous System ported flexibility of interaction and behaviour Modeling. New York: Springer. pattern of modelled entities, individual- Gilbert, N. and Troitzsch, K. G. (2000) based, object-oriented and agent-oriented Simulation for the Social Scientist. Milton approaches are distinguished in multi-level Keynes: Open University Press. modelling. Zeigler, B. P., Praehofer, H. and Kim, T. K. (2000) Theory of Modeling and Simulation. Evaluation New York: Academic Press. Modelling is a central activity in all scientific and engineering endeavours. While scientists in sociology and ecology are primarily interested in a better understanding of systems and MULTIPLE INDICATORS their dynamics, the purpose of model building in engineering is designing better systems. Definition The application areas, the objectives of the modelling are diverse, as are the modelling Observable variables used to measure or formalisms. As ‘the limits of my language represent some theoretical construct that is means the limits of my world’ (Ludwig not directly observable. If an individual score Wittgenstein, Tractatus 5.6), it is very impor- on a construct is inferred from more than tant to be aware about the many modelling one single datum point, the measurement formalisms available and to deliberately is referred to as being based on multiple choose among them. indicators. Simulation models are not constructed for a system per se, but for a system and an experi- Distinctive Features ment. Thus, a model of a system might be valid for one experiment and invalid for Some variables of interest to social scientists another. Bernard Zeigler introduced the con- (for example, demographics like gender or age) cept of experimental frame, which establishes can be measured satisfactorily by obtaining

182 M-Jupp-3373.qxd 3/8/2006 10:43 AM Page 183

MULTIPLE INDICATORS

data on a single indicator. However, virtually Evaluation all social research also involves the inves- The impact of increasing the number of tigation of theoretical concepts for which indicators on reliability and validity can be there is no simple one-to-one correspondence substantial. For example, Jaccard (1974) has to an item from the domain of observables. shown that correlations between a personal- Nevertheless, it is not an uncommon prac- ity scale and a behavioural criterion rise from tice to employ single indicators to measure an average of about 0.20 to 0.60 when multi- complex underlying constructs. ple instead of single acts are used as depen- Such practice may have serious negative dent variables; Rushton et al. (1983) consequences for the quality of measure- demonstrated largely the same effect across a ment, however. Shortcomings of single indi- wide range of fields of social research. cators take on two different forms, known as However, researchers have to consider a deficiency and contamination, respectively. trade-off between measurement quality and Single indicators are typically deficient in cost-effectiveness in practical applications. that they do not represent the domain of The functional relationship between number interest comprehensively, and they are conta- of indicators and measurement quality is minated by variables, which actually do not asymptotically increasing in the typical case belong to that domain. For instance, if the whereas that between number of indicators construct of ‘job performance as a salesper- and cost-effectiveness is typically linearly son’ is measured by the single item ‘volume decreasing. Thus, there is usually a point of sales per year’, that measure would be when adding new indicators no longer ‘pays deficient because many relevant aspects of off’. Furthermore, positive effects of increas- job performance are missing. At the same ing the number of indicators can only be time, the measure will be contaminated expected under the assumption that the new because sales volume does not only depend indicators have comparable individual prop- on the salesperson’s behaviour but also on erties as the ones already applied. external conditions, which lie outside his or her responsibility. As a consequence, the indicator will be both relatively unreliable (that Bernd Marcus is, lack stability in repeated measurement and show weak relationships to other indica- Associated Concepts: composite measure- tors of the same construct) and invalid ment, indicator, measurement, performance (weakly related to the entire underlying con- indicator, social indicators, reliability, validity struct domain). Moreover, single indicators of measurement do not permit empirical assessment of the internal structure of the construct domain, one important facet of construct validity. All of these flaws can be partially over- Key Readings come by employing multiple indicators Jaccard, J. J. (1974) ‘Predicting social behav- instead of just one: a more representative iour from personality traits’, Journal of (less deficient) sample of the construct’s Research in Personality, 7 (4): 358–67. observable manifestations can be obtained, Nunnally, J. C. and Bernstein, I. H. (1994) contaminations by irrelevant outside vari- Psychometric Theory, 3rd edn. New York: ables (measurement errors) tend to cancel McGraw–Hill. each other out in aggregates, and structural Rushton, J. P., Brainerd, C. J. and Pressley, M. analyses through procedures like psychomet- (1983) ‘Behavioural development and ric item analysis or factor analysis become construct validity: the principal of aggre- applicable. Thus, the use of multiple indica- gation’, Psychological Bulletin, 94 (1): tors is one key to sound measurement. 18–38.

183 M-Jupp-3373.qxd 3/8/2006 10:43 AM Page 184

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

multivariate modelling includes procedures MULTIVARIATE ANALYSIS that relate categorical variables to each other (the analysis of multi-dimensional tables Definition through log-linear and related procedures), categorical variables to continuous variables The set of statistical methods employed in the (multiple analysis of variance) and analyses of interpretation of data when more than one covariation among variables at all levels of variable property is measured for each case measurement (logistic regression). The objec- in the sample or population of interest. The tive of modelling is to establish a description methods explore covariation among variables – of the mechanism which generates the how the variables vary together – either in observed pattern of covariation. This is funda- order to establish relationships among vari- mental for variable-based causal reasoning ables or to allocate cases to categories. and consequent prediction. The other set of methods focuses on the Distinctive Features cases rather than the variables and uses variate information to allocate cases to pre- Multivariate methods require the construc- existing categories (discriminant analysis) or tion of a case-variable matrix in which the to construct a system of categories from rows of the matrix contain all measurements scratch and allocate cases to those categories (or recorded missing data) for each case and (cluster analysis). A generic term for such the columns contain all measurements for procedures is ‘numerical taxonomy’ – the use each variate aspect of those cases. Common of quantitative information in sorting things statistical packages (MINITAB, SPSS) lay out ‘into kinds’. the data spreadsheet in this way. Methods fall into two kinds. The first, and better known, Evaluation describe the relationships among variables through the construction of models. Most are The use of multivariate methods of any kind examples of the general linear model, in depends on the degree to which measure- which the relationships between variables ments produce a description of reality – the are assumed to be proportionate in that issue of operationalization. There are two changes in one or more variables produce aspects to this. The first is whether our ‘vari- changes in one or more other variables which ables’ correspond to anything at all in reality can be described by a linear equation or set or whether they are just our constructs with of simultaneous linear equations. Typically no real existence – the problem of reification. the form of the relationship is expressed by The second is whether we measured in such a coefficient which describes how one or a way that our measurements do describe the more dependent variables change consequent real variation – the problem of validity. The on changes in one or more independent vari- only way to resolve these problems is careful ables (regression coefficient or analogues) attention to measurement as a social process and the strength of the relationship is concerned with the quantitative description expressed through a coefficient which of reality. describes how much of the variation in one or Statistical modelling, which employs linear more dependent variables is known when techniques, faces the problem that social the variation in one or more independent change is generally nonlinear in form. In variables is known (correlation coefficient or other words, things do not change proportion- analogues). ately but rather what interests us is changes Usually, consideration of multivariate mod- of a kind which cannot be investigated elling begins with methods applicable to vari- through linear modelling. Numerical taxon- ables measured at the continuous level but omy procedures do not face this problem and

184 M-Jupp-3373.qxd 3/8/2006 10:43 AM Page 185

MULTIVARIATE ANALYSIS

are important exploratory procedures in social Key Readings investigation. Byrne, D. S. (2002) Interpreting Quantitative Data. London: Sage. David Byrne Everitt, B. (1993) Cluster Analysis. London: Arnold. Associated Concepts: analysis of variance Stevens, J. (1996) Applied Multivariate (ANOVA), causal model, cluster analysis, Statistics for the Social Sciences. Mahwah, correlation, covariance, discriminant func-tion NJ: Lawrence Erlbaum Associates. analysis, general linear modelling, log-linear analysis, prediction, regression, SPSS, validity of measurement, variable analysis

185 N-Jupp-3373.qxd 3/8/2006 5:47 PM Page 186

often linked to discipline. In social history NARRATIVE ANALYSIS and anthropology, narrative can refer to an entire life story, woven from the threads of Definition interviews, observation and documents. In In the human sciences the term refers to a socio-linguistics and other fields, the concept family of approaches to diverse kinds of of narrative is restricted, referring to brief, texts, which have in common a storied form. topically specific stories organized around As nations and governments construct pre- characters, setting and plot. In another tradi- ferred narratives about history, so do social tion (common in psychology and sociology), movements, organizations, scientists and personal narrative encompasses long sections other professionals, ethnic/racial groups, and of talk – extended accounts of lives in context individuals in stories of experience. What that develop over the course of single or mul- makes such diverse texts ‘narrative’ is tiple interviews. sequence and consequence: events are Investigators’ definitions of narrative lead to selected, organized, connected and evaluated different methods of analysis, but all require as meaningful for a particular audience. them to construct texts for further analysis, that Storytelling interprets the world and experi- is, select and organize documents, compose ence in it. Narratives are storied ways of field notes and/or choose sections of interview knowing and communicating. The focus here transcripts for close inspection. Narratives do is on oral narratives of personal experience. not speak for themselves or have unanalysed merit; they require interpretation when used as data in social research. Numerous typologies of Distinctive Features methods of narrative analysis exist. The one Research interest in narrative emerged from sketched here is a heuristic effort to describe several contemporary movements: the ‘narra- a range of contemporary approaches particu- tive turn’ in the human sciences away from larly suited to oral narratives of personal expe- positivist modes of inquiry and the master rience. The typology is not intended to be narratives of theory, for example Marxism; hierarchical or evaluative. Also, in practice, the ‘memoir boom’ in literature and popular different approaches can be combined; they are culture; identity politics in US, European and not mutually exclusive and, as with all typolo- transnational movements, such as emancipa- gies, boundaries are fuzzy. tion efforts of women, gays and lesbians, and other marginalized groups; and the burgeon- Thematic analysis Emphasis is on the con- ing therapeutic culture – exploration of per- tent of a text, ‘what’ is said more than ‘how’ it sonal life in therapies of various kinds. is said, the ‘told’ rather than the ‘telling’. A phi- Among investigators there is considerable losophy of language (albeit unacknowledged) variation in definitions of personal narrative, underpins the approach: language is a direct N-Jupp-3373.qxd 3/8/2006 5:47 PM Page 187

NARRATIVE ANALYSIS

and unambiguous route to meaning. As centered and temporally ordered stories – but grounded theorists do, investigators collect he retains the basic components of a narra- many stories and inductively create conceptual tive’s structure: the abstract (summary and/or groupings from the data. A typology of narra- point of the story); orientation (to time, place, tives (for example, narratives about experienc- characters and situation); complicating action ing illness) organized by theme is the typical (the event sequence, or plot, usually with a representational strategy, with case studies or crisis and turning point); evaluation (where vignettes providing illustration. the narrator steps back from the action to The thematic approach is useful for theo- comment on meaning and communicate rizing across a number of cases – finding emotion – the ‘soul’ of the narrative); resolu- common thematic elements across research tion (the outcome of the plot); and a coda participants and the events they report. (ending the story and bringing action back to A typology can be constructed to elaborate a the present). Not all stories contain all ele- developing theory. Because interest lies in the ments, and they can occur in varying content of speech, analysts interpret what is sequences. Labov’s microanalysis convinc- said by focusing on the meaning that any ingly shows how violent actions (in bars, on competent user of the language would find the street, etc.) are the outcome of speech in a story. Language is viewed as a resource, acts gone awry. From a small corpus of nar- not a topic of investigation. But does the ratives and prior work of Goffman, he devel- approach mimic objectivist modes of inquiry, ops a theory of the rules of requests, which suggesting themes are unmediated by the explains violent eruptions in various settings investigator’s theoretical perspective interests experienced by a diverse group of narrators. and mode of questioning? The contexts of Because structural approaches require an utterance – in the interview, in wider insti- examination of syntactic and prosodic fea- tutional and cultural discourses – are not usu- tures of talk, they are not suitable for large ally studied. When many narratives are numbers, but can be very useful for detailed grouped into a similar thematic category, case studies and comparison of several narra- readers must assume that everyone in the tive accounts. Microanalysis of a few cases group means the same thing by what they can build theories that relate language and say. What happens to ambiguities, ‘deviant’ meaning in ways that are missed when trans- responses that do not fit easily into a typo- parency is assumed, as in thematic analysis. logy, the unspoken? Investigators must decide, depending on the focus of a project, how much transcrip- Structural analysis Emphasis shifts to the tion detail is necessary. There is the danger telling, to the way a story is told. Although that interview excerpts can become unread- thematic content does not slip away, focus is able for those unfamiliar with socio-linguistics, equally on form – how a teller by selecting compromising communication across discipli- particular narrative devices makes a story nary boundaries. Like the thematic approach, persuasive. Unlike the thematic approach, strict application of the structural approach language is treated seriously – an object can de-contextualize narratives by ignoring for close investigation – over and beyond its historical, interactional and institutional fac- referential content. tors. Research settings and relationships con- Arguably the first method of narrative strain what can be narrated and shape the analysis, developed by William Labov and way a particular story develops. colleagues more than thirty years ago, this structural approach analyses the function of a Interactional analysis Here the emphasis clause in the overall narrative – the commu- is on the dialogic process between teller and nicative work it accomplishes. Labov (1982) listener. Narratives of experience are occa- later modified the approach to examine first sioned in particular settings, such as medical, person accounts of violence – brief, topically social service and court situations, where

187 N-Jupp-3373.qxd 3/8/2006 5:47 PM Page 188

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

storyteller and questioner jointly participate to narrative as praxis – a form of social action. in conversation. In interactional analysis, Consequently narrative researchers may attention to thematic content and narrative analyse different features: actors allowed on structure are not abandoned, but interest stage in an oral narrative (for example, char- shifts to storytelling as a process of co- acters and their positionings in a story, construction, where teller and listener create including narrator/protagonist); settings (the meaning collaboratively. Stories of personal conditions of performance, and setting of the experience, organized around the life-world story performed); the enactment of dialogue of the teller, may be inserted into question- between characters (reported speech); and and-answer exchanges. The approach requires audience response (the listener(s) who inter- transcripts that include all participants in the prets the drama as it unfolds, and the inter- conversation, and is strengthened when par- preter in later reading(s)). Performative alinguistic features of interaction are included analysis is emergent in narrative studies as well. (although the dramaturgic view originated Some research questions require interac- with Goffman), and researchers are experi- tional analysis. Jack Clark and Elliot Mishler menting with it, for example, in studies of (1992) sought to distinguish the features that identities – vested presentations of ‘self’ differentiated ‘attentive’ medical interviews (Riessman, 2003). from others. By analysing pauses, interrup- Kristin Langellier and Eric Peterson (2003) tions, topic chaining and other aspects of con- provide a compelling theory and many empir- versation, they show how medical interviews ical examples, ranging from detailed analyses can (and cannot) result in patient narratives of family (group) storytelling, and an illness that provide knowledge for accurate diagno- narrative told by a breast cancer survivor. sis and treatment. The authors analyse the positioning of story- An interactional approach is useful for teller, audience and characters in each per- studies of relationships between speakers in formance. They show how storytelling is a diverse field settings (courts of law, class- communicative practice that is embodied, sit- rooms, social service organizations, psycho- uated and material, discursive and open to therapy offices and the research interview legitimation and critique. itself). Like structural approaches, studies of The performative view is appropriate for interaction typically represent speech in all studies of communication practices, and for its complexity, not simply as a vehicle for detailed studies of identity construction – content. As in conversational analysis, tran- how narrators want to be known, and pre- scripts may be difficult for the uninitiated. cisely how they involve the audience in Pauses, disfluencies and other aspects of talk ‘doing’ their identities. The approach invites are typically included, but what cannot be study of how audiences are implicated in the represented in a transcript (unlike a video- art of narrative performance. Integrating the tape) is the unspoken. What happens to ges- visual (through filming and photography) ture, gaze and other displays that are enacted with the spoken narrative represents an inno- and embodied? vative contemporary turn.

Performative analysis Extending the Evaluation interactional approach, interest goes beyond the spoken word and, as the stage metaphor Analysis of narrative is no longer the province implies, storytelling is seen as performance – of literary study alone; it has penetrated all by a ‘self’ with a past – who involves, per- the human sciences, and the practising pro- suades and (perhaps) moves an audience fessions. There are several systematic meth- through language and gesture, ‘doing’ rather ods for data analysis, and each has its critics. than telling alone. Variation exists in the per- None of the approaches is appropriate for formative approach, ranging from dramaturgic studies of large numbers of nameless and

188 N-Jupp-3373.qxd 3/8/2006 5:47 PM Page 189

NARRATIVE INTERVIEWING

faceless subjects. Some modes of analysis are Labov, W. (1982) ‘Speech actions and reactions slow and painstaking, requiring attention to in personal narrative’, in D. Tannen (ed.), subtlety: nuances of speech, the organization Analyzing Discourse: Text and Talk. of a response, relations between researcher Washington, DC: Georgetown University and subject, social and historical contexts – Press. pp. 219–47. cultural narratives that make ‘personal’ stories Langellier, K. M. and Peterson, E. E. (2003) possible. In a recent reflexive turn, scholars Performing Narrative: The Communicative in autoethnography and other traditions are Practice of Storytelling. Philadelphia, PA: producing their own narratives, relating their Temple University Press. Riessman, C. K. (2002) ‘Doing justice: posi- biographies to research materials (Riessman, tioning the interpreter in narrative work’, 2002). in W. Patterson (ed.), Strategic Narrative: Narratives do not mirror, they refract the New Perspectives on the Power of Personal past. Imagination and strategic interests influ- and Cultural Storytelling. Lanham, MA and ence how storytellers choose to connect events Oxford: Lexington Books. pp. 195–216. and make them meaningful for others. Riessman, C. K. (2003) ‘Performing identities Narratives are useful in research precisely in illness narrative: masculinity and multi- because storytellers interpret the past rather ple sclerosis’, Qualitative Research, 3 (1). than reproduce it as it was. The ‘truths’ of narrative accounts are not in their faithful representations of a past world, but in the shifting connections they forge among past, present and future. They offer storytellers a way to re- NARRATIVE INTERVIEWING imagine lives (as narratives do for nations, organizations, ethnic/racial and other groups Definition forming collective identities). Building on A form of interviewing that involves the C. Wright Mills’s narrative analysis can forge generation of detailed ‘stories’ of experience, connections between personal biography and not generalized descriptions. Narratives come social structure – the personal and the political. in many forms, ranging from tightly bounded ones that recount specific past events (with Catherine Kohler Riessman clear beginnings, middles and ends), to narratives that traverse temporal and geographical space – biographical accounts that cover Associated Concepts: autoethnography, con- entire lives or careers. tent analysis, conversation analysis, discourse analysis, ethnographic interviewing, grounded theory, intellectual craftsmanship, interpre- Distinctive Features tive repertoires, microethnography, narrative The idea of narrative interviewing represents interviewing, textual analysis, transcription, a major shift in perspective in the human videography sciences about the research interview itself. The question and answer (stimulus/response) model gives way to viewing the interview as Key Readings a discursive accomplishment. Participants Clark, J. A. and Mishler, E. G. (1992) engage in an evolving conversation; narrator ‘Attending to patients’ stories: reframing and listener/questioner, collaboratively, pro- the clinical task’, Sociology of Health and duce and make meaning of events and experi- Illness, 14: 344–70. ences that the narrator reports (Mishler, 1986). Cortazzi, M. (2001) ‘Narrative analysis in The ‘facilitating’ interviewer and the vessel-like ethnography’, in P. Atkinson, A. Coffey, ‘respondent’ are replaced by two active partici- S. Delamont, J. Lofland and L. Lofland (eds), pants, who jointly produce meaning (Gubrium Handbook of Ethnography. London: Sage. and Holstein, 2002). Narrative interviewing has

189 N-Jupp-3373.qxd 3/8/2006 5:47 PM Page 190

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

more in common with contemporary ethno- participants – interviewee and interviewer graphy than with mainstream social science alike – to jointly construct narratives using interviewing practice that relies on discrete available cultural forms. Not all parents tell open-ended and/or fixed-response questions. stories to children on a routine basis and not When the interview is viewed as a conver- all cultures are orally based (in some groups, sation – a discourse between speakers – rules of course, stories are the primary way to com- of everyday conversation apply: turn-taking, municate about the past). Storytelling as a relevancy, entrance and exit talk to transition way of knowing and telling is differentially into, and return from, a story world. One story invoked by participants in research inter- can lead to another; as narrator and ques- views. Not all narratives are ‘stories’ in the tioner/listener negotiate spaces for these strict (socio-linguistic) sense of the term. extended turns, it helps to explore associations Sometimes it is next to impossible for a and meanings that might connect several participant to narrate experience in spoken stories. If we want to learn about experience in language alone. Wendy Luttrell (2003), working all its complexity, details count: specific inci- as an ethnographer in a classroom for pregnant dents, not general evaluations of experience. teenagers, mostly African Americans, expected Narrative accounts require longer turns at talk ‘stories’ from each girl about key events: than are typical in ‘natural’ conversation, learning of pregnancy, telling mothers and certainly in mainstream research practice. boyfriends, making the decision to keep the Opening up the research interview to baby, and other moments. She confronted extended narration by a research participant silence instead, only to discover a world of nar- requires investigators to give up some control. rative as she encouraged the girls’ artistic Although we have particular experiential paths productions and role plays. When she asked we want to cover, narrative interviewing them to discuss their art work, they performed means following participants down their trails. narratives about the key moments for each Genuine ‘discoveries’ about a phenomenon other – group storytelling. It is a limitation can come from power-sharing in interviews. for investigators to rely only on the texts con- Narratives can emerge at the most unex- structed from individual interviews – ‘holy pected times, even in answer to fixed-response transcripts’. Innovations among contemporary (Yes/No) questions (Riessman, 2002). But scholars include combining observation, sus- certain kinds of open-ended questions are tained relationships and conversations over more likely than others to provide narrative time with participants, even visual data with opportunities. Compare ‘When did X happen?’, narrative interviewing (for example, video tap- which requests a discrete piece of information, ing of participants’ environments, photographs with ‘Tell me what happened … and then what they take and their responses to photographs of happened?’, which asks for an extended others). account of some past time. Some investigators, In sum, narrative interviewing is not a set after introductions, invite a participant to ‘tell of ‘techniques’ nor is it necessarily ‘natural’. your story’ – how an illness began, for exam- If creatively used, it offers a way in some ple. But experience always exceeds its descrip- research situations for us to forge dialogic tion and narrativization; events may be relationships, and greater communicative fleetingly summarized and given little signifi- equality in social research. cance. Only with further questioning can participants recall the details, turning points and Catherine Kohler Riessman other shifts in cognition, emotion and action.

Evaluation Associated Concepts: discourse analysis, elite interviewing, ethnographic interviewing, In general, less structure in interview exploitative research, idiographic, interview, schedules gives greater control to research narrative analysis, videography

190 N-Jupp-3373.qxd 3/8/2006 5:47 PM Page 191

NATURALISTIC DATA

Key Readings proposed by Barker involved grossly defined behaviours and, in these pre-video days, typi- Gubrium, J. F. and Holstein, J. A. (2002) cally involved some kind of tick-box time sam- ‘From the individual interview to the inter- pling of occurrences. These researchers were view society,’ in J. F. Gubrium and J. A. hampered by shortcomings in theory, transcrip- Holdstein (eds), Handbook of Interview Research: Context and Method. Thousand tion, analysis and technology. Oaks, CA: Sage. The analysis of records of people interacting Luttrell, W. (2003) Pregnant Bodies, Fertile was stimulated most fundamentally by Minds: Gender, Race, and the Schooling of Harvey Sacks and the tradition of conversation Pregnant Teens. New York: Routledge. analysis which he and colleagues Gail Mishler, E. G. (1986) Research Interviewing: Jefferson and Emanuel Schegloff developed Context and Narrative. Cambridge, MA: (Sacks, 1992). This work started to exploit Harvard University Press. developments in recording technology (at first Riessman, C. K. (2002) ‘Positioning gender audio, more recently video). identity in narratives of infertility: South Advantages commonly offered for working Indian women’s lives in context’, in M. C. with naturalistic data include the following: Inhorn and F. van Balen (eds), Infertility Around the Globe: New Thinking on (1) It does not flood the research setting Childlessness, Gender, and Reproductive with the researcher’s own categories Technologies. Berkeley, CA: University of (embedded in questions, probes, stimuli, California Press. vignettes and so on). (2) It does not put people in the position of disinterested experts on their own and others’ practices, thoughts and so on, NATURALISTIC DATA encouraging them to provide normatively appropriate descriptions (as many interview and questionnaire studies do). Definition (3) It does not leave the researcher to make The form of records of human activities that a range of more or less problematic are neither elicited by nor affected by the inferences from the data collection arena actions of social researchers. to the topic. Claims about counselling, say, are not dependent on what counsel- Distinctive Features lors or clients say about counselling. (4) It opens up a wide variety of novel issues For much of the twentieth century social and concerns that are outside the prior scientists worked predominantly with materi- expectations embodied in question- als that involved self-reports (social surveys, naires, experimental formats, interview interviews, questionnaires) or measures of questions and so on. the effects of experimental manipulations. (5) It is a rich record of people living their Early critics of this work argued for more use lives, pursuing goals, managing institu- of unobtrusive measures (Webb et al. [1966] tional tasks and so on. (Modified from 1999) and systematic observation (Barker, Potter, 2002) 1968). Yet in practice the unobtrusive measures proposed mainly took the form of traces of In addition to these reasons for working with gross behaviour (worn flooring identifies popu- naturalistic data, researchers have stressed lar museum exhibits; contents of human its role in validating claims. The readers and garbage identifies consumption habits) with referees of an article can access transcripts of relatively little in the way of theory to the material and, increasingly, Web-based deal with the more complex records noted audio and video records. As Sacks put it, the (diaries, newspapers). The observation research goal is to develop analysis where ‘the reader

191 N-Jupp-3373.qxd 3/8/2006 5:47 PM Page 192

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

has as much information as the author, and two rather different kinds of social science can reproduce the analysis’ (1992, vol. I: 27). research going on simultaneously. Ethnographers in anthropology and sociology have often worked with materials collected Evaluation from participant or non-participant observation. Although these involve making records The significance of naturalistic data is contro- of people living their lives largely uninflu- versial because it is bound up with broader enced by social scientists, in practice the theoretical issues about the conduct of social actual data are typically field notes of some science. For some it undoubtedly pulls the kind. Such field notes will involve a range of researcher away from social scientists’ tradi- summarizing and categorizing practices that tional grand questions and deflects them onto may pre-impose a range of understandings on a concern with the trivial details of interaction. what is going on as well as making it particu- Sacks disparages an approach that assumes larly hard to recover the original organization ‘that you could tell right off whether some- of the interaction. thing was important [and thus should] start by The goal of generating naturalistic material looking at what kings did, or to look at votes, is easily distinguished from the goals of or revolutions’ (1992, vol. I: 28). He proposed hypothetico-deductive research using experi- instead the possibility that we might be able to ments, surveys and so on. In practice, of ‘read the world out of the phone conversation course, there are likely to be a variety of as well as we can read it out of anything else sources of ‘reactivity’ as participants have we’re doing’ (1992, vol. II: 548). filled in ethics permission forms, may be aware In recent years, the successes of conversa- of the recording, may have some idea of tion analytic studies of institutional interac- research questions and so on (Speer, 2002). tion (for example, Clayman and Heritage, Moreover, the process of transcription is 2002), of discursive psychological studies of embedded in a theoretical and analytic tradi- mind in action (for example, Edwards, 1997), tion that emphasizes some phenomena (for and video-based ethnography (Goodwin and example, overlap between speakers) and down- Goodwin, 1997) have moved the debate on. plays others (for example, regional accent). So The virtues of using naturalistic material although the data may not be ‘got up’ by the have been demonstrated and the onus is researcher, it is unwise to treat it as something increasingly on researchers to justify using independent of the research enterprise. The other kinds of data. Moreover, researchers term naturalistic (rather than natural) marks a are just starting to exploit the enormous caution about the status of such data (Potter, fluidity in working with audio and video that 2002). Nevertheless, every effort should be digital technology combined with cheap pow- made to eliminate or manage these problems. erful computing power has made available. Despite the strong contrast to be made by work using interviews, experiments and other Jonathan Potter forms of data elicitation, it is possible to study research practices as forms of natural interac- Associated Concepts: conversation analysis, tion in their own right. That is, rather than discursive psychology, participant observa- using a survey to discover something about a tion, transcription, unobtrusive measures, sample of people, the interaction involved in validity of measurement, videography administering that survey (including the situated identities of those involved, the practices of questioning and answering, the construc- Key Readings tion of social science categories such as ‘opin- Barker, R. G. (1968) Ecological Psychology: ions’ and so on) can be taken as a topic. Concepts and Methods for Studying the In effect, the research interaction has been Environment of Human Behaviour. Stanford, naturalized (see Maynard et al., 2002), with CA: Stanford University Press.

192 N-Jupp-3373.qxd 3/8/2006 5:47 PM Page 193

NETNOGRAPHY

Clayman, S. E. and Heritage, J. (2002) The These new formations, in their novel online News Interview: Journalists and Public context, required the development of new Figures on the Air. Cambridge: Cambridge research methodologies. University Press. At least four aspects of online, computer- Edwards, D. (1997) Discourse and Cognition. mediated or virtual, interaction and commu- London: Sage. nity formation are distinct from their Goodwin, C. and Goodwin, M. H. (1997) in-person, real life (RL), or face-to-face (F2F) ‘Contested vision: the discursive constitu- counterparts. First is the textual, nonphysical, tion of Rodney King’, in B-L. Gunnarsson, and social-cue-impoverished context of the P. Linell and B. Nordberg (eds), The online environment. Second is an unprece- Construction of Professional Discourse. dented new level of access to the heretofore London: Longman. pp. 292–316. unobservable behaviours of particular interact- Maynard, D. W., Houtkoop-Steenstra, H., Schaeffer, N. C. and van der Zouwen, J. ing peoples. Third, while traditional interac- (eds), Standardization and Tacit Knowledge: tions are ephemeral as they occur, online social Interaction and Practice in the Survey interactions are often automatically saved and Interview. New York: Wiley. archived, creating permanent records. Finally, Potter, J. (2002) ‘Two kinds of natural’, the social nature of the new medium is unclear Discourse Studies, 4: 539–42. as to whether it is a private or public space, or Sacks, H. (1992) Lectures on Conversation, some unique hybrid. Ethnography adapts com- Vols I and II (ed. G. Jefferson). Oxford: mon participant observation ethnographic Blackwell. procedures – such as making cultural entrée, Speer, S. (2002) ‘“Natural” and “contrived” gathering and analysing data, ensuring trust- data: a sustainable distinction?’, Discourse worthy interpretation, conducting member Studies, 4: 511–25. checks and conducting ethical research – to Webb, E. J., Campbell, D. T., Shwartz, R. D. these computer-mediated contingencies (see and Sechrest, L. (1996/1999) Unobtrusive Kozinets, 2002 for a detailed explication). Measures, 2nd edn. London: Sage. Given the wide range of choices of online communal form, including boards, rings, lists and dungeons (see Kozinets, 1999), researchers should spend time matching their research NETNOGRAPHY questions and interests to appropriate online fora, using the novel resources of online search engines such as Yahoo! and Google groups, Definition before initiating entrée. Before initiating con- Ethnography conducted on the Internet; a tact as a participant, or beginning formal data qualitative, interpretive research methodology collection, the distinctive characteristics of the that adapts the traditional, in-person ethno- online communities should be familiar to the graphic research techniques of anthropology netnographer. to the study of the online cultures and com- In a netnography, data take two forms: munities formed through computer-mediated data that the researcher directly copies from communications (CMC). the computer-mediated communications of online community members, and data that Distinctive Features the researcher inscribes. Reflective field notes, in which ethnographers record their Ethnography refers to in-person participative- observations, are a time-tested and recom- observational fieldwork among particular mended method. However, distinct from tra- social groups, and also to the representations ditional ethnography, rigorous netnographies based on such studies. In the past decade, the have been conducted using only observation mass adoption of networked personal com- and download, without the researcher writ- puting gave rise to new cultural formations. ing a single field note.

193 N-Jupp-3373.qxd 3/8/2006 5:47 PM Page 194

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

As with grounded theory, data collection Evaluation should continue as long as new insights on Compared to surveys, experiments, focus important topical areas are still being gener- groups and personal interviews, netnography ated. For purposes of precision, some netno- is a far less obtrusive method. It is conducted graphers may wish to closely track the amount using observations in a context that is not fab- of text collected and read, and the number ricated by the researcher. Netnography also is of distinct participants. Software solutions less costly and timelier than focus groups and expedite coding, content analysis, data linking, personal interviews. It is a naturalistic and data display and theory-building functions. unobtrusive technique – a nearly unprece- Contextualizing online data is necessary to its dented combination. analysis, and often proves to be challenging in The limitations of netnography draw from the social-cues-impoverished online context of its more narrow focus on online communities, netnography. Because netnography is based its inability to offer the full and rich detail of primarily upon the observation of textual dis- lived human experience, the need for course, ensuring trustworthy interpretations researcher interpretive skill, and the lack of requires a different approach than the balanc- informant identifiers present in the online ing of discourse and observed behaviour context that leads to difficulty generalizing that occurs during in-person ethnography. results to groups outside the online community Although the online landscape mediates social sample. Researchers wishing to generalize the representation and renders problematic the findings of a netnography of a particular online issue of informant identity, netnography seems group to other groups must therefore apply perfectly amenable to treating behaviour or the careful evaluations of similarity and consider social act as the ultimate unit of analysis, rather employing multiple methods for triangulation. than the individual person. Netnography is still a relatively new method, Research ethics may be one of the most and awaits further development and refine- important differences between traditional ment at the hands of a new generation of ethnography and netnography. Ethical concerns Internet-savvy ethnographic researchers. over netnography turn on contentious and still largely unsettled concerns about whether online forums are to be considered a private or a Robert V. Kozinets public site, and about what constitutes informed consent in cyberspace (see Paccagnella, 1997). Associated Concepts: access, confidentiality, In a major departure from traditional methods, covert research, ethics, ethnographic inter- netnography uses cultural information that is viewing, ethnography, impression manage- not given specifically, and in confidence, to ment, informed consent, Internet research, the researcher. The consumers who originally participant observation, unobtrusive measures created the data do not necessarily intend or welcome their use in research representations. Ethical procedures that involve full disclosure of the researcher’s presence, affiliations Key Readings and intentions, and extensive member check- Kozinets, R. V. (1998) ‘On Netnography: initial ing, are recommended. Gaining feedback reflections on consumer research investiga- on research from community members is tions of cyberculture’, in J. Alba and W. facilitated by the online medium. In addition, Hutchinson (eds), Advances in Consumer researchers should ensure the confidentiality Research, Volume 25. Provo, UT: Associa- and anonymity of informants. To show addition for Consumer Research. pp. 366–71. tional concern for ethics, researchers might Kozinets, R. V. (1999) ‘E-tribalized market- contact specific community members to obtain ing: the strategic implications of virtual their permission to directly quote specific post- communities of consumption’, European ings, or idiosyncratic stories, in the research. Management Journal, 17 (3): 252–64.

194 N-Jupp-3373.qxd 3/8/2006 5:47 PM Page 195

NOMOTHETIC

Kozinets, R. V. (2002) ‘The field behind the computer software, to develop various measures screen: using Netnography for marketing of density, strength, connectedness and also to research in online communities’, Journal examine structure in terms of cliques, blocks of Marketing Research, 39 (February): and the flow of information in terms of 61–72. ‘sources’ (initiators of information flows) and Paccagnella, L. (1997) ‘Getting the seats of ‘sinks’ (receivers of information). your pants dirty: strategies for ethnographic research on virtual communities’, Evaluation Journal of Computer-Mediated Communica- tions, 3 (June). Available online at http:// Network analysis is predominantly a form of ww.ascusc.org/jcmc/ analysis rather than a way of theorizing. There are different levels of analysis, from exploratory visual mapping to mathematical modelling. Its strength lies in the different ways in which it can be applied. For example, NETWORK ANALYSIS it can be used at different levels and with different units of analysis (individuals, small Definition groups, communities) and in different subject Analysis of any pattern of relations between areas. It can be quantitative or qualitative and individuals, groups, institutions or any other it can be used as a business tool or in inter- social collectivities. ventionist and policy-related ways, for example, in mapping AIDS and other sexually transmitted diseases. When used in the latter Distinctive Features way it is akin to snowball sampling. Network analysis has its roots in social anthropology, for example, in examining kinship net- Victor Jupp works and also trading networks (for example, Malinowski’s study of exchange relations Associated Concepts: community study amongst Trobriand Islanders). It has subse- method, snowball sampling, sociometry quently been used in social psychology to examine interpersonal relations in small groups. An example is Moreno’s development Key Readings of sociometry in which he introduced the Bott, E. (1957) Family and Social Network. idea of a social network as a diagram – a London: Tavistock. sociogram – in which points, representing Mitchell, J. C. (ed.) (1969) Social Networks in people, were connected by lines, representing Urban Situations. Manchester: Manchester social relationships. Sociometry can be used to University Press. study group formation and group dynamics, Moreno, J. L. (1953) Who Shall Survive? for example, the formation of cliques and the New York: Beacon. flow of information. It has also contributed to Scott, J. P. (1991) Social Network Analysis: the study of economic, political, criminal and A Handbook. London: Sage. familial networks within communities. An example of network analysis in relation to kinship is Elizabeth Bott’s study of kinship networks in terms of the degree of their con- NOMOTHETIC nectedness but without any element of measurement of such connectedness. A further Definition development of network analysis involved the application of sophisticated mathematical tech- Research into regularities and patterns in social niques, based on graph theory, and using behaviour and in social arrangements. This is

195 N-Jupp-3373.qxd 3/8/2006 5:47 PM Page 196

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

in contrast to idiographic research, which is under-structural and historical patterns. This concerned with the detailed examination of strand is likely to favour forms of triangula- specific instances, cases or contexts. tion, mixed or even messy research methods.

Distinctive Features Victor Jupp Typically, nomothetic inquiry is part of a wider endeavour of the search for universal Associated Concepts: causal model, covering- explanations or covering laws. In the ‘hard’ law model of explanation, epistemology, idio- sense such explanations or laws are deemed graphic, messy research, mixed-methods to be causal and deterministic, involving research, qualitative research, quantitative universal conditional statements such as ‘if a research, thick description, triangulation occurs then b’. In a ‘softer’ sense they can take the form of probabilistic statements such Key Readings as ‘if a occurs then there is a probability that b will follow’. Crotty, M. (1998) The Foundations of Social In its search for broad patterns and regular- Research. London: Sage. ities nomothetic inquiry usually involves large- Wallerstein, I. (1987) ‘World systems analy- scale, macro research. Also, it is usually – but sis’, in A. Giddens and J. Turner (eds), not always – quantitative research. This is in Social Theory Today. Cambridge: Polity Press. contrast with idiographic inquiry which is typically qualitative and concerned with the detailed analysis of specific cases, for example life histories, documents, events, institutions. Whereas the idiographic stands for the NON-PROBABILITY (NON-RANDOM) viewpoint that such specific cases can only SAMPLING be understood and explained in their particular context and time period, the nomothetic Definition stands for the search for patterns and regularities which are common to all such cases. Forms of sampling that do not adhere to prob- It would, for example, look for consistencies ability methods. Probability methods choose in the management structure of organiza- samples using random selection and every tions, from different sectors, which are member of the population has an equal undergoing rapid technological change. chance of selection. Some types of non- Further, it would aim to reach explanations, random sampling still aim to achieve a degree even laws, which ‘cover’ these consistencies. of representativeness without using random methods. Several different techniques are Evaluation associated with this approach, for example accidental or convenience sampling; snow- The idiographic–nomothetic divide is often ball sampling; volunteer sampling; quota presented as starkly as quantitative–qualita- sampling, and theoretical sampling. tive, micro–macro, thick description– thin description, on the grounds that they rep- Distinctive Features resent contrasting and incompatible epistemological and ontological positions. There is, Convenience samples are also known as acci- however, a ‘softer’, more pragmatic and less dental or opportunity samples. The problem adversarial strand in social research. This is with all of these types of samples is that there one that recognizes both the objective and sub- is no evidence that they are representative of jective aspects of social life and the need to the populations to which the researchers understand the specific as well as search for wish to generalize. This approach is often

196 N-Jupp-3373.qxd 3/8/2006 5:47 PM Page 197

NUD*IST

used when the researcher must make use of generalizability. In quota sampling, for example, available respondents or where no sampling the people who are interviewed are those who frame exists. A good example is provided by are on the streets and available at the time of my own experience of conducting evaluative fieldwork, and those who agree to form a research designed to explore the efficacy of volunteer sample might be those who have a community treatment programmes for sex particular interest in the topic or hold especially offenders. Here an accidental sample of those strong views about it. They are a self-selected men sentenced to treatment was used by sample and there is no way of knowing whether necessity (Davidson, 2001). TV, radio station, their views are representative. newspaper and magazine polls and question- Whilst it cannot be denied that such sam- naires are also an example of this type of pling techniques may not produce a represen- sampling, where respondents may be asked tative response, some of the most influential to ‘phone in’ or fill in a questionnaire. and interesting qualitative research has never- Quota sampling is very popular with theless been conducted on this basis. market researchers, first because instant opinions might be needed and probability sampling would take too much time, and sec- Julia Davidson ondly because it is less costly than probability sampling. A typical example is the market Associated Concepts: marketing research, researcher who stands in the high street, and opportunity sampling, probability (random) selects certain types of shoppers, often by age sampling, purposive sampling, quota sam- and gender. A quota number is set for each of pling, sampling, validity of generalization the categories of people required for the study. The problem with this method is that Key Readings bias can enter the process because the inter- Bryman, A. (2001) Social Research Methods. viewer has an element of choice in how to Oxford: Oxford University Press. select people to fill the quota. Snowballing, Davidson, J. (2001) The Context and Practice which is akin to a chain-letter, is used when of Community Treatment Programmes it is difficult to find a sample in any other for Child Sexual Abusers. PhD thesis way (there is no accessible sampling frame unpublished. for the population from which the sample is Kumar, R. (1999) Research Methodology: to be taken), and it is often used with difficult A Step by Step Guide for Beginners. London: to access groups. Here the researcher finds Sage. one respondent, who in turn introduces May, T. (1997) Social Research: Issues, another, and so on. This technique might be Methods and Process. Milton Keynes: used in research on deviant lifestyles, illegal Open University Press. drug users or any population with unknown characteristics. Theoretical sampling seeks to select respondents who will enhance the theoretical development of the research. Such sampling aims to identify cases which may NUD*IST challenge emergent concepts. Finally, as the term indicates, volunteer sampling relies on Definition individuals or groups of individuals volun- One of the pioneer software packages for teering to be the subjects of research. qualitative analysis. Designed to assist Evaluation researchers in handling Non-numerical Unstructured Data by Indexing, Searching Non-probability sampling has been criti- and Theorizing, it supports rapid coding and cized for its lack of representativeness and code-based access for small or large bodies of

197 N-Jupp-3373.qxd 3/8/2006 5:47 PM Page 198

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

qualitative data in academic, governmental • Automation of clerical tasks by command and other settings. files to handle large or repetitive processes. • Import and export of information or coding from and to statistical research Distinctive Features packages. NUD*IST provides symmetrical handling of data documents and ideas, with emphasis on Evaluation rapid access for interpretation, coding and searching. Experience with users led to Whilst QDA software grew more sophisti- development of a sister product for smaller cated, NUD*IST was developed as a respon- projects with more detailed analysis, NVivo. sive, stripped-down toolkit for rapid handling Distinctive features of the NUD*IST line of even very large data sets. Data imported or of software have been: pasted to NUD*IST appears in plain text, and is automatically formatted to a unit specified • Themes or categories that emerge in qual- by the researcher (line, sentence or para- itative analysis and may be used in coding graph). These constraints mean NUD*IST are presented as items (nodes) that can be can handle small or large quantities of data managed and explored in index systems as flexibly. Facilities for automating coding the understanding grows. and data handling support many types of • Symmetrical explorers show documents qualitative–quantitative linking, and han- and nodes. Selecting an item sets coding or dling of data that would not traditionally be searching processes to apply to that data called qualitative, such as the open-ended item. answers to surveys. • Coding of passages of text can be done Projects can be moved from NUD*IST for interactively, selecting text on the screen, or more subtle analysis to its QSR stable-mate, by typing in the unit number to be coded. NVivo, where data can be handled in rich • Viewing and reviewing coded data live, text, with character-level coding, linking, the researcher has immediate access to subtle ways of filtering and refining searches context and the ability to code on as inter- and modelling. Team or multi-site projects pretation of the coded material suggests can be merged with an accompanying pack- finer categories. age for combining multiple projects, compar- • Browsing the text of a document or all the ison between projects or combining work of text coded at a node moves the user between multiple researchers. linked windows on the textual data. • Records of the interpretative processes are in annotations and memos; document text, Lyn Richards coded. • Ability to edit coded data without invali- Associated Concepts: CAQDAS, coding, The dating coding. Ethnograph, QSR NVivo, qualitative research • Storing of information about cases, documents, respondents, sites etc. as coding, accessible to all search processes. Key Readings • Exploring patterns of coding and text via a Richards, L. (2002) Using N6 in Qualitative logically complete and linked set of search Research. Melbourne: QSR International. tools. Richards, T. (2002) ‘An intellectual history of • Iterative exploration, using the results of NUD*IST and NVivo’, International any search saved as coding, so another Journal of Social Research Methodology, question can be asked. 5 (3): 199–214.

198 O-Jupp-3373.qxd 3/8/2006 10:43 AM Page 199

1835 and 1935 governments in Europe and OFFICIAL STATISTICS the United States universally established official statistical departments and engaged in sys- Definitions tematic measurement of social conditions. Two definitions may be offered: Simultaneously the accountancy and related monitoring activities of governments gener- (1) Measurements produced by local or ated quantitative information describing the national states or international bodies operations of governments themselves. The which describe the general social condi- extension of the range of government activities tion with particular reference to change, to into both the economy and civil society gener- monitor the activities of state and related ated an ever-increasing number of data bodies, and indicate relative performance streams. State macro-economic management of providers of collective services. reinforced this tendency and the development (2) A system of collecting statistical data, of international bodies – the UN and its agen- gathering, storing and generalizing col- cies, the European Union and other global lected data; and announcing, providing regional blocks – generated new international and disseminating the results of statisti- data sets. Many data sets were intended for cal surveys as the official statistical data. the original purpose of statistics – to enable comparisons among states (including local Distinctive Features states) on a quantitative basis. There are two general sources of official Strictly speaking ‘official statistics’ is an oxy- statistics. The first is from the monitoring moron because the literal meaning of statistics processes of administration through bureau- is measurements made by the state describ- cratic data collection. Schools return data on ing state and society. The original charter of pupil numbers, exam successes and related the UK’s Royal Statistical Society stated that matters. These data are used to construct a its purpose was: set of data series, including league tables. The to collect, arrange, digest and publish facts, illus- other is from outreach investigative activities trating the condition and prospects of society in such as the UK decennial census but there its material, social, and moral relations; these are a series of other investigations, both into facts being for the most part arranged in tabular general social conditions and into specific forms and in accordance with the principles of areas of interest such as health. In either numerical method. case, users of the data sets, whether inspect- Such measurements by states, together with ing published tables or working with raw accounts of the state’s own activities in quan- secondary data sets, must pay particularly titative form, were the original official statis- careful attention to the operationalization of tics. Desrosières (1998) notes that between the variables measured. O-Jupp-3373.qxd 3/8/2006 10:43 AM Page 200

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

In the 1960s there was a general effort to performance data, for example on hospital move official statistics beyond being mere waiting lists, have become an important description towards becoming a systematic political issue to an unprecedented degree. and organized model of social processes at every level. This was the ‘social indicators’ Evaluation movement. At the simplest level a social indicator is simply a measure of something con- Statistics are always social constructs and this sidered to be socially significant as an index of is the origin of two related but ultimately dif- some wider social process or context – which ferent sets of criticisms of official statistics. can change. So the UK’s most accessible The first, which we might call radical statis- general official statistical publication is Social tics, is that being products of the state, official Trends in which a set of data series describe statistics represent the positions of powerful changes through time. However, the social groups in society which have an undue influ- indicators movement sought to move beyond ence on the ways in which measurements are description of isolated and distinctive trends both defined and constructed. In other towards the construction of a social informa- words, official statistics investigate things tion system on the lines of the economic infor- which state agencies and powerful interests mation system, which, using economic official want investigated and the actual measures statistics, underpins predictive models of the are distorted by political interests and pres- economic system. This project has not been sures. The issue of the integrity of official sta- successful but there has been a considerable tistics has been taken up in the UK through development of evaluative indicators as tools the appointment of a Statistics Commission for assessing the success of particular policies and is central to the UN Fundamental or projects. There has also been a development Principles of Official Statistics adopted in of overall indicators – for example of depriva- 1994 which focus particularly on official station at a ward level in the UK or the United tistics as part of the public information of a Nations Development Programme’s Human democratic society. This was the content of Development Index for nation-states. The pur- the mission statement of UK official statistics pose of such indices is comparison on a single in the 1960s and 1970s, although the Rayner measure which incorporates multiple dimen- reforms of the 1980s (see Levitas and Guy, sions within that measure. 1996) shifted the focus towards the state’s A recent and very significant extension of own requirements and the commercial provi- this has been the publication of data series sion of information on a market basis. Now describing the outputs of a range of public there is more emphasis on social statistics as sector institutions. The most important of sources of information for the consumers these in England and Wales are state schools of public services. The other set of criticisms and a wide range of performance data which of official statistics is the general rejection of are widely disseminated. Data of this kind quantified social measurement as reification have a novel purpose. Official statistics tradi- rather than in any way being a real version of tionally have been seen as sources of infor- how society works. This leads to a general mation for the state itself, at whatever level, dismissal of any social measurement includ- and as part of the information base that can ing official statistics whereas the radical sta- inform an active and collectively minded citi- tistics position argues instead with the nature zenry. Now institutional level data are pub- of the process and the role of interest and lished to inform individual consumption power in that process. choices, for example about choice of school The central point to recognize is that official for a child, and directed not at a socially con- statistics are constructs, not crude facts which scious citizenry but at an aggregate of individ- are given. However, the reasonable position ual consumers. This has meant that overall adopted by most social scientists is that they

200 O-Jupp-3373.qxd 3/8/2006 10:43 AM Page 201

ONE-SHOT DESIGN

are made out of something, not nothing, and selected may not be truly random. One-shot that provided we pay careful attention to the case study designs are favoured on grounds ways in which they are made, and in particular such as the proximity of the research site to the processes of operationalization, they can be the research team. The choice may also be of very considerable value to us. made when the current state of the knowledge base justifies a preliminary exploratory case study prior to a more fully developed theoret- David Byrne ically informed design to test a developed set of theoretical arguments, which may require a Associated Concepts: auditing, comparative longitudinal dimension. method, critical research, data, descriptive statistics, econometrics, forecasting, performance indicator, primary research, secondary Evaluation analysis, social indicators The findings from one-shot case study designs need to be treated with care. A one-shot case Key Readings study of the unemployed in London might find that 65 per cent were in poor health and con- Byrne, D. S. (2002) Interpreting Quantitative Data. London: Sage. clude that unemployment had a negative effect Desrosières, A. (1998) The Politics of on health status. Such a conclusion would not Large Numbers. Cambridge, MA: Harvard be warranted however for a number of reasons. University Press. First, it is assumed that the 65 per cent level of Levitas, R. and Guy, W. (1996) Interpreting poor health revealed is a consequence of the Official Statistics. London: Routledge. factor used to select the subject group. If older workers or those working in high health risk industries were more likely to be made unemployed these factors would ‘explain’ the 65 per cent finding. Second, the conclusion assumes ONE-SHOT DESIGN that the 65 per cent figure is indeed ‘high’ when similar levels might be found in a study Definition of the employed. Third, it assumes that the time order of the variables is that implied in A study design where a single group of indi- the causal inference, namely, that unemploy- viduals (or other interesting unit of analysis) ment is the cause of health and not vice versa. is selected for observation over a single, lim- Fourth, it neglects to seek out other relevant ited time period, usually because they have events such as the impact of unemployment experienced some factor taken as important on family relationships: unemployment may in shaping some outcome. It is akin to one-off have little effect on the health status of those cross-sectional design. whose domestic relationships remain intact. Fifth, asking the unemployed about their Distinctive Features health status may also lead them to overstate health problems. An example of one such design might be a one- One-shot case studies may redress these off survey of unemployed people in a specific inherent design limitations by the use of local area to assess their health status and sophisticated statistical manipulations of the the impact of unemployment on health. Thus data such as multivariate modelling. In the there is no attempt to approximate to the clas- above example, there could be a test to link sical experiment. There is no comparison the length of unemployment to the quality of group of employed people, there is no attempt health controlling for the other known and to study changes over time and the sample measured factors. However, the results of

201 O-Jupp-3373.qxd 3/8/2006 10:43 AM Page 202

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

such analysis could not be taken as valid for knowledge. However, ontological issues are subjects outside the group chosen or outside concerned with questions pertaining to the the locality of the chosen group. kinds of things that exist within society. Again using feminist thought as an example, ontological issues concern the idea of ‘being’, Aidan Kelly and would stress the existence of patriarchal social relations and the negative implications Associated Concepts: causality, cohort study, these have for women. The issue of ontology covariance, cross-sectional survey, longitudi- within social thought is a longstanding one, nal study, multivariate analysis, panel study, traceable, at least, to the philosophy of prospective study, quasi-experiment, research Aristotle. Within social research, however, design, sampling, social survey whilst epistemological issues concern the question of what is or should be regarded as Key Readings acceptable knowledge in a discipline such as sociology or philosophy, ontological issues are Campbell, D. T. and Stanley, J. C. (1966) concerned with the nature of social bodies or Experimental and Quasi-Experimental Designs entities. for Research. Chicago: Rand McNally. Marsh, C. (1982) The Survey Method: The Research traditions are similarly under- Contributions of Surveys to Sociological pinned by a distinctive epistemological and Analysis. London: Allen & Unwin. ontological nature. For instance, the ontologi- Sapsford, R. and Jupp, V. (1996) Data cal principles underpinning examples of quan- Collection and Analysis. London: Sage. titative research are typically held to belong to de Vaus, D. (2001) Research Design in Social that branch of social thought collectively Research. London: Sage. dubbed the ‘objectivist perspective’ such as positivism, within which, the ontological nature of society is based upon the view that society is a separate entity, existing quite dis- ONTOLOGY tinctly from the social actors that comprise a given society. Alternatively, the ontological outlook of qualitative research examines the Definition relationships between social actors and society A concept concerned with the existence of, drawing from interpretative sociology (for and relationship between different aspects of example, symbolic interactionism, phenome- society, such as social actors, cultural norms nology or ethnomethodology). The ontological and social structures. assumptions of sociologists studying the social world from this perspective are guided by the Distinctive Features desire to investigate the differing ways in which social actors are constantly interpreting Commentators on the process of social the social world from their own particular per- research generally concede that any instance spective. The ontological basis for qualitative, of social inquiry is based upon the dual interpretative sociologists is one in which social fundamental principles of epistemology and reality is seen ‘as a constantly shifting emergent ontology. Epistemology deals with the issue property of individual creation’ (Bryman, of knowledge, and specifically, who can be a 1988: 20). ‘knower’. For example, within feminist research, Sandra Harding (1987) argues that Evaluation women have been systematically excluded as ‘knowers’ from ‘masculine’ sociology and The relationship between ontological ques- therefore must establish a feminist ‘stand- tions and interpretative sociology is parti- point’, and legitimate female epistemological cularly insightful within the qualitative

202 O-Jupp-3373.qxd 3/8/2006 10:43 AM Page 203

OPERATIONAL RESEARCH MANAGEMENT

research tradition. Within qualitative types of ‘What happens when different kinds of world research there is the conspicuous acknowl- are placed in confrontation, or when bound- edgement that social reality is not a fixed set aries between worlds are violated?’ (McHale, of phenomena. This is intrinsic to the concept 1987: 10). Whereas the ‘typical’ modernist, of ontology, because, as Fay (1987) argues, the epistemological genre is detective fiction, way in which social research is carried out is McHale argues that the primary ontological intimately connected to the ways in which genre is science fiction, because this genre social researchers actually observe the world. habitually stages ‘close encounters’ between Social researchers are guided by whatever different worlds, placing them in confronta- knowledge perspective they adhere to and tion (alien invasions, time/dimension travel, this informs the concepts they subsequently the discovery of lost worlds). On a wider social develop. Therefore, feminist researchers will scale, theorists such as David Harvey, within see the world in ways that seek to uncover The Condition of Postmodernity (1989), argue and understand the existence and effects of that the ontological nature of postmodernist patriarchy. Alternatively, critical social fiction may indeed reflect distinct ‘postmod- researchers will seek to identify the relation- ernist’ trends apparent within ‘post-Fordist’ ship between social actors and social institu- Western societies, and describe the ontological tions in relation to power and class with a condition of the ‘postmodern subject’. view as to how such systems can be revealed and overcome. However, if the beliefs of researchers change over time, then so will the Lee Barron ontological (and epistemological) questions such researchers ask. Fay (1987) cites the Associated Concepts: critical research, epis- example of scientific knowledge, and the temology, feminist research, phenomenology, ways in which past explanations for concepts postmodernism, qualitative research, quanti- such as ‘nature’ have changed radically as tative research, realism new forms of thinking have displaced older models. This is analogous to Thomas Kuhn’s Key Readings notion of the ‘paradigm’, those systems of Bryman, A. (1988) Quantity and Quality in knowledge that can, and do give way to more Social Research. London: Routledge. sophisticated paradigms, which then trans- Fay, B. (1987) Critical Social Science. form the form and manner in which ontolog- Cambridge: Polity Press. ical and epistemological questions are asked. Harding, S. (ed.) (1987) Feminism and Metho- However, there are examples of research dology. Milton Keynes: Open University that actually takes the issue of ontology as the Press. central research question rather than episte- McHale, B. (1987) Postmodernist Fiction. mological concerns, such as McHale’s New York: Methuen. Postmodernist Fiction (1987). Based upon the textual analysis of novels, McHale argues that the primary concern of postmodernist fiction is the raising of questions of an ontological nature rather than epistemological questions. OPERATIONAL RESEARCH Returning to the definitions in which episte- MANAGEMENT mology equates with ‘knowing’, and ontology describes ‘being’, modernist fiction is charac- Definition terized by epistemological issues, raising questions such as: ‘What is there to be known?’ The scientific-based approach used tradition- ‘Who knows it?’ Postmodernist fiction is very ally by businesses (although recent studies different. Postmodernist fiction foregrounds using operational research management tech- questions such as: ‘Which world is this?’ niques may be found in health, education and

203 O-Jupp-3373.qxd 3/8/2006 10:43 AM Page 204

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

other public sectors) to improve their decision- along with other decision-making procedures making process regarding the allocation of such as cost-benefit analysis. Large busi- scare resources. The decision-making process nesses developed in the 1960s, in part due to is based upon the scientific school of thought the expansion in world trade. These busi- and consequently studies tend to be quantita- nesses had to make decisions regarding prod- tive in nature, and involve the analysis of ucts, new markets and investment in new probability and risk. plants and machinery, and they consequently adopted ORM techniques. Distinctive Features The methodology of ORM is one of completing a series of steps, these being: The overall aim of operational research man- • Identifying the nature of the problem. agement (ORM) is to provide a sound rational • Formulating the nature of the problem in basis for decision making. ORM recognizes, and relation to the various techniques available. is based upon the economic fact, that resources • Building the model from the available are limited, yet wants are unlimited. Given this data. situation, businesses are never in the position • Deriving a solution or solutions. where they can undertake all of their would-be • Evaluating the solution(s). future plans or investments. In other words, • Implementing the chosen solution. some form of choice has to be made. ORM is the mechanism for making that choice. In addi- This methodology is deemed to be the classi- tion, ORM makes the process more transpar- cal approach to ORM. There are many varia- ent, such that business can provide detailed tions on it, but the general pattern covers the analysis why one particular course of action has steps above. The techniques available in been favoured above all others. ORM are wide ranging, and large steps in the Traditionally ORM is thought of as only development in ORM techniques have been being used by large businesses, corporations taken with the introduction of the desktop or governments. However, the principles computer. The most commonly used tech- involved with ORM can be applied to very niques include: linear programming, critical simple and also to very complex business deci- path analysis, statistical modelling and multi- sions. Clearly there are links between ORM variate analysis, simulation and forecasting. and the discipline of economics. However, utilizing ORM in differing situations will lead to other disciplines being incorporated into the Evaluation decision-making process. For example, utiliz- Clearly the ORM specialist has decisions to ing ORM techniques in the building of a new make as to which of the above techniques space station will bring together engineers, should be utilized along with the validity of space experts, nuclear scientists, physicists, to the solution generated. From the tech- mention only a few. niques available an ORM specialist has to The history of ORM can be traced back have considerable knowledge of the area many hundreds of years. Elements of ORM under consideration in order to generate a can be found in the military campaigns of feasible solution. Napoleon. It is within this military tradition that we find ORM techniques being developed in World War I, and in particular during Paul E. Pye World War II. The logistics of large-scale troop and machine movements meant that Associated Concepts: applied research, audit- generals had to be aware of making a bad ing, decision-making research, econometrics, decision. Lack of ammunition and support forecasting, modelling, organizational research, does not win battles. After World War II, performance indicator, prediction, simulation ORM was further developed by the military methods

204 O-Jupp-3373.qxd 3/8/2006 10:43 AM Page 205

OPPORTUNITY SAMPLING

Key Readings practicalities of assessing reliability and validity of indirect measures of concepts. French, S., Hartley, R., Thomas, L. C. and White, D. J. (1986) Operational Research Techniques. London: Edward Arnold. Victor Jupp Hillier, F. (2000) Introduction to Operational Research. New York: McGraw–Hill. Associated Concepts: composite measure- Taha, H. A. (2002) Operations Research: An ment, measurement, multiple indicators, reli- Introduction. New York: Prentice Hall. ability, scaling, validity of measurement

Key Readings

OPERATIONALIZATION Blalock, H. M. Jnr (1968) ‘The measurement problem: a gap between the language of theory and research’, in H. M. Blalock Jnr Definition and A. B. Blalock (eds), Methodology in Social Research. New York: McGraw–Hill. The process of devising ways of measuring pp. 5–27. theoretical concepts. Moser, C. A. and Kalton, G. (1971) Survey Methods in Social Investigation. London: Distinctive Features Heinemann. Theoretical concepts are tools for organizing Smith, H. W. (1975) Strategies of Social Research. Englewood Cliffs, NJ: Prentice knowledge. They are not, however, directly Hall. observable and measurable. What researchers do is devise rules for recognizing instances or indicators of such concepts. For example, the theoretical concept ‘social class’ is typically operationalized and measured by asking OPPORTUNITY SAMPLING subjects questions about their occupation. There are established rules for assigning Definition occupations to social class categories. Such rules are sometimes referred to as correspon- A sampling tool to identify and gain access to dence rules because they link theoretical data sources. Opportunity sampling uses the concepts to observable instances which cor- knowledge and attributes of the researcher respond to them. Where a number of mea- to identify a sample, for example, using a sures or indicators are combined in the researcher’s local knowledge of an area on operationalization of a concept it is common which to base a study or using a researcher’s to refer to composite measurement or multi- past experiences to contact participants or ple indicators. gatekeepers.

Evaluation Distinctive Features There are issues regarding operationalization Opportunity sampling is often grouped linked to ontological and epistemological together with incidental types of sampling debates about the nature of social reality such as convenience sampling, volunteer and whether it can be measured. However, sampling and purposeful sampling. Opportu- in social research literature most of the dis- nity sampling is often viewed as the weakest cussion is cast in terms of levels of measure- form of sample selection. It is also regarded ment (nominal, ordinal, interval or ratio), by some as being less demanding on what items to include in a scale (such as a researchers, in terms of resources or exper- Likert or semantic differential scale) and the tise, than other methods of sampling.

205 O-Jupp-3373.qxd 3/8/2006 10:43 AM Page 206

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

Opportunity sampling is often employed perspective maintains that social research by social researchers studying covert or hard- and the methods it employs should strive to to-access groups of people, objects or events. be value-free and objective. From this posi- These groups are often vulnerable, hidden, tivist perspective opportunity sampling is fluid or uncooperative. The traditional for- weak on external validity as it is impossible mula used to identify a sample may not apply to generalize from the data it produces in these circumstances. Often groups such as because it is not representative of the social these are not collectively identified in society world in general. by any agency and so no sampling frame Some may see it as an ‘easy’ way to collect exists. In such cases opportunity sampling data, but it is important that the researcher may be the only viable technique from which be aware of its limitations and appropriate- to create a sample. For example, Goode used ness for the aims of the study. Therefore, the her own knowledge of the drug agencies in sampling technique employed must tally with the West Midlands of the UK to identify a the epistemological starting point of the suitable area to base her study of drug-using research. The researcher must clearly make mothers (Goode, 2000). the reader aware of the sampling techniques In some instances it may be appropriate to used so that the limitations and validity of the combine opportunity sampling with probabil- research conclusions are fully understood. ity sampling techniques. For example, a researcher aiming to study drug sellers in nightclubs may use her own knowledge or Angela Brady experience to identify a particular town or city where this activity takes place. She may Associated Concepts: covert research, proba- then use a random or systematic sample to bility (random) sampling, purposive sampling, choose which nightclubs to investigate. sampling, snowball sampling, validity of generalization Evaluation Key Readings For research that investigates ‘invisible’ issues in the social world, opportunity sam- Goode, S. (2000) ‘Researching a hard-to- pling has some advantages. First, and possi- access and vulnerable population: some bly most importantly, it can allow researchers considerations on researching drug and access to groups that would remain hidden if alcohol-using mothers’, Sociological Research they had relied on conventional sampling Online, 5 (1). techniques. Secondly, it allows the researcher Mason, J. (2002) ‘Sampling and selection in qualitative research’, in Qualitative to define who and where to study. In qualita- Researching, 2nd edn. London: Sage. ch. II tive research that aims to understand and Shipman, M. (1997) The Limitations of Social describe social issues this can be useful as Research, 4th edn. New York: Longman. non-responses are then limited and only relevant data are recorded. In such cases this can be cost- and time-efficient. Thirdly, it releases researchers from relying on structural institutions in society to identify sam- ORAL HISTORY ples and allows them to engage with the social world proactively. Definition However, this sampling technique also has disadvantages. Although some would argue A method that seeks to open novel routes for that all social research is inherently subjec- understanding the past, the relation of past to tive and based on the politics and values of present and the lives of others through time, researchers and institutions, the positivist by listening to the voices of individuals talking

206 O-Jupp-3373.qxd 3/8/2006 10:43 AM Page 207

ORAL HISTORY

extensively about the events and experiences elderly respondents. In practice, oral history through which they have lived. The characteris also defined by the capacity of audio and istic form through which oral history data are video recording technologies, the increasing gathered is the in-depth interview. sophistication of which continues to enhance the practical viability of oral history inter- Distinctive Features view and also to facilitate the more effective archiving and dissemination of oral history The expansion of oral history as an important data. Some researchers use the term ‘oral his- technique for modern social research dates tory’ as synonymous with ‘life history’ and, from the 1960s. The initial impetus and in their central methodological character- impact of this movement was symbolic as istics, the two are very close. Both utilize well as substantive. Not only did it open up in-depth unstructured or semi-structured sources of data of rare quality – revealing in interview formats; both recognize interview content, rich in detail, intimate in character, data as a production of dialogue in which the evocative in tone – which were extremely dif- interviewer is himself or herself an integral ficult to locate through more traditional part; both respect the intense ethical and per- documentary methods, it also promised to shift sonal implications of the interview relation- the locus of conventional historical inquiry to ship; both acknowledge the requirement for a more democratic and inclusive plane. Early strong conventions governing the ownership exercises in oral history were often marked and end use of data; both recognize the chal- by the impulse to rescue or recover the accu- lenges raised by attempts to engage with mulated experiences of marginalized or for- conventional criteria for sampling, reliability gotten groups, those for whom the surviving and generalizability concerns; above all, both documentary record was largely silent. More approaches are predicated upon the central- recently, the scope of the approach has ity of the individual life as the most powerful widened to generate new insights into profes- lens through which may be revealed the com- sional or establishment groups which were plex social contexts within which lives are themselves originally marginalized by the lived. Whilst the concerns of life history orig- initial trajectory of the new method. At the inate in illuminating the course of individual same time, oral history has undergone signif- lives, however, oral history projects often icant advances in methodological sophistica- start with a substantive historical question, tion. Particularly important have been new for which oral history interview presents concerns for understanding the importance itself as the most appropriate method. of social or collective memory, identity, myth and the self, together with the means by Evaluation which life narratives are constructed. In these respects, oral history research – in One of the major epistemological difficulties which the spoken word is privileged – has for oral history has been its necessary worked towards the refinement of its concep- reliance upon memory – a faculty not infre- tual repertoire. It has also contributed to a quently invested as ‘the enemy of history’. To constructive critique of conventional the extent that memory may be shown to be approaches for the analysis of documentary factually inaccurate, the reliability of oral his- sources – the written word – the majority of tory data are open to challenge. However, it which are marked by their radically different can be argued that the mixing of factual and character as inscribed and crafted texts fictive elements in oral history narratives directed towards unseen audiences. itself constitutes an important insight into the In its nature, the reach of oral history is ways in which memory operates. In this demarcated by the span of human memory, respect the notion of narrated stories which and this is one of the reasons why many may be ‘“right” if untrue’ indicates an impor- projects involve disproportionate numbers of tant area in which oral historians need to do

207 O-Jupp-3373.qxd 3/8/2006 10:43 AM Page 208

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

more theoretical work. Further advances may Key Readings also accrue from oral history research designs Dunaway, D. K. and Baum, W. K. (eds) in which relevant documentary (‘factual’) (1996) Oral History: An Interdisciplinary sources are systematically employed along- Anthology, 2nd edn. Walnut Creek, CA: side, or even within, the elaboration of AltaMira Press. respondents’ testimonies (which is akin to the Perks, R. and Thomson, A. (1998) The Oral notion of triangulation). This may further History Reader. London: Routledge. help to promote the reconciliation of the spoken with the written word for explaining and understanding the past. In particular, it may help to break down a common association of the document as concerned princi- ORGANIZATIONAL RESEARCH pally with ‘hard’ explanation and ‘truth’, and oral testimony principally with ‘soft’ under- Definition standing and ‘authenticity’. A further problem relates to the accessibil- The systematic collection, analysis and inter- ity of oral history data for third-party pretation of data with the clear purpose of researchers. The historical document in an finding things out about an organization or archive has a pristine and public quality in organizations. This can include analysing, that it can be consulted by a succession of reviewing, explaining, understanding, evalu- researchers who each share the same initial ating or describing some aspect or aspects of status in approaching it. This is not so in the an organization or organizations to answer a case for oral history data, where the two orig- question or number of questions. inal participants are necessarily privileged in their access to the full range of ostensive Distinctive Features reference within which the interview took place. The interview transcript is the docu- Organizations are created by groups in soci- mentary device by which the interview data ety in order to achieve specific purposes or are opened to other researchers, but such meet specific needs through planned and documents – ‘speech masquerading as text’ – coordinated activities. Given an almost infi- are often difficult to make full sense of nite variety of possible purposes and differing because of the absence of ostensive reference. needs, it is not surprising that organizations Whilst some have experimented with aug- differ in nature and type as well as their mented transcription protocols in the manner structure, management and the way they utilized by conversation analysts, such docu- function. Despite this, there are a number of ments are often difficult to read. The freeing of components that are common: human or oral history data from the bonds that unavoid- people resources, other non-human resources, ably join them to their original producers purpose, structure and planning and coordi- remains an important issue in achieving the nating of activities. It is from these compo- wider utilization of data that are often incom- nents and their interactions that the subject parable in their originality, depth and impact. matter of organizational research is drawn. Organizational research has a number of features, other than its subject matter, that Philip Gardner make it distinctive. These relate to the setting in which the research is conducted and the Associated Concepts: conversation analysis, way in which researchers can draw upon document analysis, ethnographic interviewing, knowledge developed across a range of social historical analysis, life history interviewing, science disciplines. naturalistic data, qualitative data archives, Many writers (for example, Bulmer, 1988; transcription, triangulation Saunders et al., 2003) have highlighted that

208 O-Jupp-3373.qxd 3/8/2006 10:43 AM Page 209

ORGANIZATIONAL RESEARCH

organizations, as research settings, possess a been argued that combining knowledge from number of distinctive but interrelated features. a range of disciplines allows such research to First, in order to undertake research gain new insights that could not be obtained within an organization, physical access must through all of these disciplines separately be negotiated and gained through an organi- (Tranfield and Starkey, 1998). zational gatekeeper. An organization may not be prepared to engage in such additional and Evaluation voluntary activities because of the time and resources involved. The research may be per- Organizational research is undertaken within ceived to be of little value, it may be felt that different philosophical approaches using a the research topic is too sensitive, or there wide variety of research strategies and data may be doubts about the researcher’s credi- collection techniques. Data can be collected bility and competence. and analysed both cross-sectionally and lon- Secondly, after physical access has been gitudinally. Levels of analysis might include gained, the researcher needs to gain access to the individual, work teams, departments and the data that she or he needs the intended divisions, or comparative studies between participants to provide. In granting such organizations. The balance of power between access the organization’s gatekeeper may the researcher and the organization means restrict the pool of participants or the data to that where choices of data collection strate- which the researcher is allowed access. gies and techniques have not been specified Subsequently research participants within clearly or agreed, the researcher’s indepen- the organization may also restrict cognitive dence may be compromised. Consequently access. Such restrictions can be minimized by there needs to be prior agreement between the the researcher establishing and maintaining researcher and the organization regarding the her or his credibility with those individuals precise data collection strategies and tech- who can provide the data. Consequently, niques that will be used as well as regarding whilst an organization has power both associated issues of confidentiality and initially to grant or refuse access and subse- anonymity. However, given the power rela- quently to restrict or withdraw it at any time, tionship between the researcher and the orga- individual data providers also have power to nization, organizational research is likely to prevent access. be what is feasible and practicable rather Thirdly, although an organization may than what is theoretically desirable. have granted a researcher access there may In recent years the purpose of organiza- be preconditions. For example, in exchange tional research has been subject to some for access, the organization may insist upon a debate, particularly within disciplines such as particular research method being used or business and management. Two interrelated their right to veto the reporting of sensitive or aspects of this debate that have received con- unfavourable findings. Alternatively, it may siderable attention are a requirement that request anonymity in the research report. such research should complete a virtuous The researcher may also face ethical dilem- circle between theory and practice (Tranfield mas, such as a request by the organization to and Starkey, 1998) and the need for the name participants who have been promised research to surmount a double hurdle of anonymity, or to present data in such a way being both theoretically and methodologi- that it breaks confidentiality. Such requests cally rigorous and of practical relevance must be resisted. (Hodgkinson et al., 2001). This implies that Finally, organizational research is often organizational research should enable the transdisciplinary, that is, it draws upon development of new ideas and relate them to knowledge from a range of social science dis- practice thereby allowing research on prac- ciplines such as business and management, tice to inform theory that could, in turn, geography, psychology and sociology. It has inform future practice. Building upon this, it

209 O-Jupp-3373.qxd 3/8/2006 10:43 AM Page 210

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

could be argued that organizational research Key Readings should engage with both theory and practice Bulmer, M. (1988) ‘Some reflections on and, perhaps, have an applied focus. However, research in organisations’, in A. Bryman it must be remembered that opinions differ (ed.), Doing Research in Organizations. regarding whether the practical implications London: Routledge. of research findings for organizations must Hodgkinson, G. P., Herriot, P. and Anderson, also be considered. Research undertaken N. (2001) ‘Re-aligning the stakeholders in purely to understand the processes of organi- management research: lessons from indus- zations and their outcomes is also organiza- trial, work and organizational psychology’, tional research. Like all social research, it British Journal of Management, 12, Special should be theoretically and methodologically Edition, pp. 41–8. rigorous. Saunders, M. N. K., Lewis, P. and Thornhill, A. (2003) Research Methods for Business Mark N. K. Saunders Students. Harlow: Financial Times Prentice Hall. Associated Concepts: access, applied Tranfield, D. and Starkey, K. (1998) ‘The nature, social organization and promotion research, confidentiality, decision-making of management research: towards policy’, research, ethics, gatekeeper, informed con- British Journal of Management, 9: 341–53. sent, operational research management

210 P-Jupp-3373.qxd 3/8/2006 5:48 PM Page 211

followed up over time, their composition is PANEL STUDY likely to change as people join or leave the household. In some cases there are ‘following Definition rules’ to pursue panel members into the new In some social science literature a panel study households they are entering, which means comprises any form of longitudinal inquiry, that at the individual level the panel continu- of which the longitudinal cohort study is a ally expands in size. special case. More generally, the panel study A major attraction of panel studies is the refers not to a single cohort followed up facility they offer to investigate relatively short- through time, but the monitoring over time of term social processes. Thus in the typical a cross-section of a population, that is to say, household panel study, surveys are carried out multiple cohorts of different ages. This usually every year or at the most two years, which involves replacement of panel members as enables fine-grained analysis of household they leave the study through no longer meeting dynamics to be undertaken. We can assess the age criteria or movement or refusal. Examples extent to which people are moving in and out include household panel studies, audience of poverty both at an individual and a house- research panels and product testing panels. hold level, and how that is being affected by other experiences in their lives, including the national policy framework. By way of exam- Distinctive Features ple, a whole set of panel studies concerned Cross-sectional sample surveys provide with income dynamics has been set up in dif- descriptive estimates of the parameters of a ferent countries, starting in 1965 with the US particular population. In repeated cross- Panel Study of Income Dynamics (PSID) con- sectional surveys, such estimates provide a ducted by the University of Michigan Institute means of gauging the extent to which society of Social Research. is changing – ‘net effects’. Panel studies extend Panel studies suffer from the problems the monitoring over time to the individual common to all longitudinal surveys of attri- sample members, enabling the investigation of tion of respondents – owing to non-response, the interactions of societal change with per- movement and refusal – and missing data at sonal characteristics in affecting individual the variable level. In so far as the panel dif- lives – ‘gross effects’. They are therefore pow- fers in its characteristics at any given time erful tools for gaining understanding of social from that of the population at large then processes and for policy evaluation. weighting can be introduced to reflect the One branch of panel studies uses the population more precisely. Statistical imputa- household rather than individual members as tion procedures are typically used to estimate the sampling unit, though all members may population values. become included in the survey when they To minimize respondent burden and reach a specified age. As households are thereby reduce data loss, measurement of P-Jupp-3373.qxd 3/8/2006 5:48 PM Page 212

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

some variables is rotated through the study at Dale, A. and Davies, R. B. (1994) Analysing more extended intervals rather than repeated Social and Political Change. London: Sage. in all surveys. ‘Panel maintenance’ proce- Magnusson, D., Bergmann, L., Rüdinger, G. dures involving communications with panel and Törestad, B. (1998) Problems and members between surveys and ‘refusal con- Methods in Longitudinal Research. version’ to encourage drop-outs to rejoin the Cambridge: Cambridge University Press. study are also in common use. Rose, D. (ed.) (1998) Researching Social Change: Household Panel Studies: Methods and Substance. London: UCL Press. Evaluation Scarborough, E. and Tanenbaum, E. (1998) Panel studies are effective instruments for Research Strategies in the Social Sciences. studying social dynamics under a number of Oxford: Oxford University Press. headings, including family, income, housing, education, relationships and so on. Short- term changes are well assessed by the studies, which are not subject to the memory PARADIGM effects that can bedevil longer-term cohort inquiries. However, the major weakness of panel studies is the limited numbers of Definition members in any one cohort. Thus for effec- A ‘cluster of beliefs and dictates which for tive analysis of the impact of social change in scientists in a particular discipline influence specific cohorts born at different times we what should be studied, how research should need to take age ranges from the population be done, how results should be interpreted, that can be as wide as ten years or more. This and so on’ (Bryman, 1988: 4). limits analysis capability in relation to the identification and follow-up of special population cohorts such as people with disabili- Distinctive Features ties, for example, though the way the panel The term is chiefly associated with Thomas study is designed can compensate for this Kuhn and his research concerning the history to a certain extent, for example, by ‘over- of science, discussed in the book The sampling’ households containing such indi- Structure of Scientific Revolutions (1962/1996). viduals. Another attraction of the panel study Kuhn’s argument is that researchers who is that its continual renewal through mem- share a commitment to a particular paradigm bers joining the panel does, in principle, gives are committed to the same rules and stan- it a representativeness of the current popula- dards for scientific practice. His approach has tion, which the cohort study cannot claim. been influential within the social sciences within which two broad research paradigms John Bynner are often identified – the quantitative and the qualitative. Each paradigm rests upon distinctive foundations and applies a specific Associated Concepts: attrition, cohort approach to researching the social world. For study, cross-sectional survey, longitudinal instance, quantitative research is a research study, one-shot design, quasi-experiment, strategy that emphasizes measurement in time series design the collection and analysis of data. The quantitative approach is underpinned by a distinctive epistemological and ontological nature. Key Readings The epistemological basis for quantita- Blossfeld, H-P. and Rohwere, G. (2002) tive research is typically, but not exclusively, Techniques of Event History Modelling. characterized as positivist. Positivism advo- London: Lawrence Erlbaum. cates the application of the methods of the

212 P-Jupp-3373.qxd 3/8/2006 5:48 PM Page 213

PARADIGM

natural sciences to the study of social reality. includes participant observation, unstructured/ Positivism is usually associated with the soci- semi-structured interviews, focus groups, case ologists Auguste Comte and Emile Durkheim. studies and secondary analysis of diaries, The general principles of the positivist para- letters, autobiographies, newspapers, photo- digm are that only phenomena, and therefore graphs and documentaries. knowledge, confirmed by the senses can genuinely be warranted as knowledge; that the Evaluation function of theory models is to generate Social researchers who identify themselves as hypotheses that can be tested and that will affiliates to either the quantitative or qualita- allow explanations of law to be considered. tive approaches strictly adhere to the appro- One would cite Comte’s ‘Law of the Three priate research methods that reflect the Stages’ and Durkheim’s social explanation as theoretical traditions that underscore them. to the causes of suicide as examples of this However, this does not mean that such posi- principle. Ultimately, according to the positions are permanently fixed, nor that one must tivist school, the scientific study of society utilize the seemingly appropriate methods. must be undertaken in a manner that is value- Within scientific research, paradigms do free that is predicated upon an objective indeed shift as new discoveries are made; but study of the social world. The positivist more vitally in relation to social research, com- approach rests upon an epistemological foun- bining elements of the quantitative and quali- dation, which emphasizes the principles of tative paradigms may enhance the validity of measurement, causality, generalization and research findings. Because different methods replication. These theoretical underpinnings can reinforce and complement each other, the therefore presume a certain category of theoretical value of a study can be enhanced by research methods, which are conducive to combining methods. This is called ‘triangula- an objective, positivistic approach such as tion’ and sometimes mixed methodology. surveys, questionnaires, structured interviews, experiments, quasi-experiments and official statistics, and content analysis of documents. Lee Barron The second dominant research paradigm is qualitative research. The qualitative research Associated Concepts: causality, covering strategy differs markedly from the quantita- law model, critical research, epistemology, tive approach and is underpinned by very dif- ethnography, hypothesis, hypothetico-deductive ferent epistemological and ontological model, inter-subjective understanding, mixed- foundations. Qualitative research is con- methods research, ontology, positivism, qual- cerned with subjective social issues and has itative research, quantitative research, its roots in the phenomenology of Edmund triangulation Husserl and Alfred Schutz; the understanding of social action, or Verstehen approach of Max Weber; and symbolic interactionism as for- Key Readings mulated by G. H. Mead and Herbert Blumer. Barnes, B. (1982) T. S. Kuhn and Social Science. Such approaches emphasize the need to Basingstoke: Macmillan. understand society as social actors perceive Bryman, A. (1988) Quantity and Quality in and interpret it, and interpretations of social Social Research. London: Allen & Unwin. phenomena can vary markedly according to Bryman, A. (2001) Social Research Methods. the standpoint of the social actor. In light of Oxford: Oxford University Press. such an epistemological grounding, qualita- Hughes, J. A. (1976) Sociological Analysis: tive researchers do not typically seek objec- Methods of Discovery. Middlesex: Nelson. tive appraisal of social phenomena, therefore Kuhn, S. (1962/1996) The Structure of the arsenal of research methods utilized dif- Scientific Revolutions, 3rd edn. Chicago: fers from the quantitative tradition and University of Chicago Press.

213 P-Jupp-3373.qxd 3/8/2006 5:48 PM Page 214

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

Non-parametric tests use either rank (ordinal) PARAMETRIC TESTS data or frequencies. If they are applied to sets of interval data, the data have to be reduced Definition to rank or nominal form before the test is Tests of statistical significance which are carried out. based on certain assumptions about the shape There are non-parametric equivalents for of the distribution of population scores, for the common parametric tests which compare example that it is normally distributed (or the scores of two or more groups to see bell-shaped). In statistical analysis, paramet- whether the differences between them are ric significance tests are only valid if such statistically significant and for tests that assumptions are met. If they are not, non- determine whether there is a relationship parametric tests can be used. between two or more sets of scores.

Distinctive Features Evaluation A parameter is a measure of an entire popula- Non-parametric tests can be used with skewed tion, such as the mean height of every man in data. They are not influenced by outliers, London. In statistical analysis, one practically which are extreme scores that can affect para- never has measurements from a whole popula- metric tests because they increase the variance tion and has to infer the characteristics of the of the data set. Non-parametric tests are less population from a sample of that population. sensitive than the parametric equivalents, and The sample measure, such as the mean height do not permit one to test the significance of of 500 randomly selected London men, is a interactions between independent variables. statistic. Parameters are real entities, statistics are estimates. Jeremy J. Foster In many statistical analyses, the researcher draws inferences about the parameters of the Associated Concepts: chi-squared test, population from the sample statistics. This descriptive statistics, inferential statistics, requires making assumptions about the shape measurement of the distribution of the population’s scores. Techniques that do not make these assump- Key Reading tions are called non-parametric (or distribution- Howell, D. C. (2002) Statistical Methods for free) techniques. Psychology, 5th edn. Pacific Grove, CA: The assumptions required by the paramet- Duxbury. ric tests, such as the t-test or the F test, are: (1) observations are drawn from normally distributed populations; (2) the populations have the same variance; (3) interval scale measure- PARTICIPANT OBSERVATION ment has been used. It is usually clear if interval measurement has not been used: the data is in the form of ranks or frequency Definition counts. To test whether the data is normally A qualitative method of social investigation, distributed, one can examine a plot of the dis- whereby the researcher participates in the tribution and its skewness and kurtosis. Tests everyday life of a social setting, and records such as Levene’s can be used to see whether their experiences and observations. sets of data have equal variances. The t-test and F test are robust, and can be used if the Distinctive Features first two assumptions are not valid so long as there is the same number of cases in the Participant observation is part of the reper- groups being compared. toire of qualitative methods. It is most often

214 P-Jupp-3373.qxd 3/8/2006 5:48 PM Page 215

PARTICIPANT OBSERVATION

associated with ethnography and also with social worlds. They can also be used to record the theoretical framework of symbolic inter- the personal feelings and experiences of the actionism. It is a method widely used within researcher. It is usual practice for the partici- social and cultural anthropology, as well as pant observer to keep a fieldwork diary as part within allied social science disciplines such of the research process. as sociology and education. Participant obser- Access is a particular issue for participant vation is often used in conjunction with qual- observation. While public settings may not itative interviewing. require any formal access negotiations, there The ‘field’ refers to the setting within are still issues of managing the researcher role which the participant observation is under- within the field. In private settings, as well as taken. The extent to which the researcher or gaining formal permission, there may well be ‘fieldworker’ fully participates in the field an extensive period of negotiation needed for may depend upon the nature of the setting, the participant observer to be accepted into the quality of the research relationships that the field. This may entail paying particular are formed and the opportunities for partici- attention to the management of the fieldwork pation. Participation and observation can be identity and the adoption of roles within the conceptualized as being at either end of a field. Participant observation usually involves continuum of research positions. Sometimes a prolonged period of fieldwork and can lead it is not possible or desirable to become a full to long-term relationships. participant. This may be due to demographic factors such as age or gender, or to factors Evaluation such as the particular activities or skills under observation. (While a social researcher Participant observation is a long-established may be able to be present in a hospital oper- method of social inquiry, and can provide rich ating theatre it is highly unlikely that they accounts of social life and social worlds. There will be able to participate fully in the surgical has been a recurrent concern about the posi- procedure, unless they happen to have tion of the researcher within the field, and the advanced medical training). Sometimes full ways in which this may alter the social reali- participation may be an essential element of ties under investigation. Does the researcher the fieldwork for example in the case of change the setting by being there? The covert research. However, all participant researchers should certainly be reflexive about observation relies upon some degree of active their presence within the setting, but in gen- engagement with the setting on the part of eral it is difficult to sustain the argument that the researcher. Hence there is an assumption the hospital ward, school classroom, factory that social scientific understandings of social production line or swimming pool would worlds can be enhanced by observing, expe- really be different if the researchers were not riencing and talking to others in the field. present. Specific encounters may not or will The main method for recording observa- not have occurred but the general processes, tions is through field notes. These are notes with which participant observation is primar- that are preferably taken in situ, and then ily concerned, will have continued without, as expanded upon after the fieldwork encounter. well as with, the researcher’s presence. However, sometimes it may be difficult to take It is possible for researchers engaged in par- notes during participant observation, and in ticipant observation to become full members of these circumstances field notes are con- the setting they are studying. This can lead to a structed retrospectively. Participant observa- blurring of the boundary between research and tion, therefore, relies upon the memory work the self, or between personal and professional of the researcher. This can be aided by photo- life. This can be considered as disadvantageous graphic, documentary or audio materials, or to the research, as overfamiliarity may make by artefacts collected in the field. Field notes it difficult to assume or resume the role of give detailed descriptive accounts of everyday observer. The setting may become so familiar

215 P-Jupp-3373.qxd 3/8/2006 5:48 PM Page 216

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

that it is difficult to get any analytical purchase Emerson, R. M., Fretz, R. I. and Shaw, L. L. on it. Of course, some researchers choose to (2001) ‘Participant observation and field- undertake participant observation in settings notes’, in P. Atkinson, A. Coffey, with which they are already familiar (for exam- S. Delamont, J. Lofland and L. Lofland ple a teacher-researcher observing a class- (eds), Handbook of Ethnography. London: room). Indeed there can be advantages to Sage. pp. 352–68. having some understanding of the processes, Hammersley, M. and Atkinson, P. (1995) language and everyday activities of the setting, Ethnography: Principles in Practice. prior to undertaking participant observation. London: Sage. Lofland, J. and Lofland, L. (1995) Analyzing The key issue is that researchers engaged Social Settings: A Guide to Qualitative in participant observation should always be Observation and Analysis, 3rd edn. reflective about their positioning within the Belmont, CA: Wadsworth. setting and how that is challenged or changed over the course of the research, as well as recognizing the experiences, knowledges and assumptions they bring to the field. There are also ethical issues associated with PARTICIPATORY ACTION RESEARCH participant observation. It is a method that has been used for covert research. Sometimes this Definition may be purposive (for example, in a setting where it is assumed that research access may One of the categories into which action be denied) or it may be unavoidable (for research can be divided. Participatory action example, in a public setting where there are research (PAR) consists in an approach that no obvious permissions to be sought). Covert includes both understanding a situation (cre- participant observation raises particular con- ating knowledge) as well as changing or act- cerns over the informed consent of research ing upon that situation – using participatory participants and the ways in which any data methods, that is, challenging the dichotomy generated can or should be used. Overt par- between researchers and researched. ticipant observation is not without its own ethical dilemmas: for example, the conscious Distinctive Features forging of relationships in the field and the extent to which these are, or can be genuine; Participatory action research has emerged as a or participating in or observing activities that response to conventional, elitist research are personally difficult – on moral, political which assumes that people are not capable or legal grounds. of doing research, which consequently seldom benefits directly those who are its object. Within PAR people themselves are Amanda Coffey researchers and the knowledge generated during the research process is used to create Associated Concepts: access, covert action or improvements that benefit them research, ethics, ethnography, fieldwork, directly. The role of the outsider professional gatekeeper, ideographic, impression manage- or researcher is not that of an expert but a ment, naturalistic data, qualitative research, mere facilitator of the research/action process. reflexivity, research bargain, validity of PAR has three major distinctive elements: measurement people, power and praxis. It is people-centred as it is informed by, and responds to, the needs of the people, usually oppressed or disenfran- Key Readings chised people. PAR is about power, since Bailey, C. A. (1998) A Guide to Field projects aim to promote empowerment of the Research. Thousand Oaks, CA: Pine Forge. research participants. This type of research

216 P-Jupp-3373.qxd 3/8/2006 5:48 PM Page 217

PATH ANALYSIS

challenges the power relation between Evaluation researcher and ‘objects’, since the action Participatory action research may be abused, researcher is a peer of other research partic- when researchers use its good reputation as ipants. Finally, it is about praxis, since partic- a methodology while simultaneously they ipatory action research recognizes the actually maintain conventional relations with inseparability of theory and practice (Sohng, the project participants, using them as mere 1995). puppets. Moreover, empowered communities The key methodological feature of partici- have started to question the need for PAR pro- patory research is dialogue. Ideally, dialogue jects’ facilitators. The critique being put for- within action or participatory research is ward is that PAR does not go far enough and empowering by bringing ‘isolated people that the communities themselves are able to together around common problems and develop their own theory and research bases. needs’ (Sohng, 1995: 6). Thus, participatory research methods vary greatly from the more conventional research techniques such as Ana Lopes interviewing, which gives primacy to the researcher and implies a one-way flow of Associated Concepts: action research, information. applied research, disability research, disaster Participatory methods have been devel- research, emancipatory research, politics and oped to challenge and counteract ‘the intellec- social research, practitioner research, tual colonialism of western social research research bargain into the third world development process’ (Sohng, 1995: 3). The aim of this approach is to involve disenfranchised people in the Key Readings research processes that will enable them to Freire, P. (1972) Pedagogy of the Oppressed. find solutions in their daily struggle for sur- Harmondsworth: Penguin. vival. Thus, participatory research is research Kemmis, S. and McTaggert, R. (1990) The as collective action in the struggle over power Action Research Planner. Geelong: Deakin and resources, where knowledge is identified University Press. as power (Sohng, 1995: 3). Lewin, K. (1946) ‘Action research and minor- PAR has its roots in the work of Paulo Freire ity problems’, Journal of Social Issues, 2 (4): (1972), whose educational work emphasizes 34–6. dialogue, informed action, educational activity Sohng, S. S. L. (1995) Participatory Research based on the lived experiences of the par- and Community Organizing. A working ticipants, community enhancement and paper presented at The New Social ‘consciencialization’ – a process by which indi- Movement and Community Organizing viduals deeply analyse their own realities. Conference, University of Washington, However, the term ‘action research’ was first Seattle, WA, 1–3 November 1995. http:// used by social psychologist Kurt Lewin (1946). www.interweb-tech.com/nsmnet/docs/ references_sohng.htm Lewin attempted to improve relations in industrial situations and minimize hostility between different racial groups in the United States of the 1940s. He described action research as a spiral of steps, each one comprising the stages PATH ANALYSIS of planning, acting, observing and reflecting. This ‘action research spiral’ is Lewin’s main Definition legacy to action researchers. Although many models and diagrams have been developed, all A variant of multivariate regression analysis in action research processes are cyclical rather which causal relations between several vari- than linear. ables are represented by a path or flow diagram

217 P-Jupp-3373.qxd 3/8/2006 5:48 PM Page 218

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

and path coefficients provide estimates of the and above these it is assumed that the model strength of relationship between two vari- is recursive, which means that the causal flow ables when all the other variables are held is in one direction. In other words, causal constant. direction is assumed to be asymmetrical. Further, path analysis does not by itself give Distinctive Features evidence of causal direction. This must be provided by the researcher. Also, more fun- Path analysis allows a researcher to examine damentally, there is an in-built assumption the causal effects of several variables simulta- that social phenomena can be explained in neously. In this way, it is used to examine the terms of causality. This is an assumption that degree of ‘fit’ between the model and the data. is challenged by researchers who come from Path coefficients provide estimates of the different methodological positions. strength of the relationships between pairs of variables within the model when all the other variables are held constant. The coefficients Victor Jupp are standardized regression scores and are expressed as z-scores. Residual values measure Associated Concepts: causality, causal the amount of variation in the dependent vari- model, correlation, econometrics, general able, which cannot be explained by the com- linear modelling, multivariate analyses, bination of independent variables. regression analysis, SPSS The causal model can be portrayed visually by a path or flow diagram, which shows Key Readings the ‘paths’ along which causal influences are said to flow. Lines between variables indicate Black, T. R. (2001) Understanding Social relationships and arrowheads indicate causal Science Research, 2nd edn. London: Sage. direction. Path coefficients are stipulated on Byrne, D. (2002) Interpreting Quantitative each line or ‘path’. Sometimes the strength of Data. London: Sage. relationship is also represented visually by Sapsford, R. and Jupp, V. (1996) Data the thickness of a line. Collection and Analysis. London: Sage. The nature of path analysis in relation to Wright, D. B. (2002) First Steps in Statistics. ordinary multiple regression is that it is pos- London: Sage. sible to portray direct and indirect pathways from independent variables to a dependent variable and also to calculate the direct and indirect effect of an independent variable on PERFORMANCE INDICATOR the dependent variable whilst other effects are held constant. Definition Evaluation An indirect (or ‘proxy’) measure that is used when it is not possible to derive a direct mea- As with other forms of causal modelling, path sure of the activity concerned. An example is analysis has the advantage of being able the use of performance indicators (PIs) of hos- to examine the effects of several variables pital waiting times, numbers of complaints and simultaneously. This allows such models to length of referral lists as indicators of the over- represent the complexity of the social world all quality of service provided by hospitals. in a way that is not possible with cross- tabular or three-way analysis. However, path Distinctive Features analysis requires that certain assumptions are made. Some are the assumptions of multiple The introduction of PIs in North America and regression such as linearity of variables. Over Europe coincided with the development of

218 P-Jupp-3373.qxd 3/8/2006 5:48 PM Page 219

PERFORMANCE INDICATOR

scientific management (or ‘Taylorism’), which citizens satisfied with the overall service required the application of efficiency mea- provided by their local authority. sures and cost–benefit analysis (Van den Berghe, 1997). The growth of the use of PIs in Evaluation areas such as the provision of education, health and policing from the late 1970s While PIs are a useful contribution to the onwards resulted from the increasing appli- assessment of how an organization meets its cation of managerial approaches in public declared objectives, their limitations have to service provision (Pollitt, 1993). be acknowledged. PIs do not explain why dif- There has been a major growth in the use ferences exist between organizations, but they of PIs as a means of assessing the perfor- do raise issues that need to be addressed – mance of public bodies. In the United they act as a ‘can opener’. For example, the Kingdom for example there are requirements average cost per pupil will not, on its own, on local authorities, the National Health mean that one school is more efficient than Service, Social Services departments and cen- another, for there may be many reasons why tral government departments to agree perfor- costs vary. But we need to consider why the mance targets and to publish performance differences exist. indicators. Economy, efficiency and effectiveness do PIs are often used as part of a more not always ‘march together’ (Pollitt, 1993: 59). comprehensive performance information The achievement of greater economy, for system which can be used to measure example, may have negative impacts progress toward targets, meet the require- on effectiveness and efficiency. This may ments of public accountability, compare the encourage the focus on certain indicators activities of a number of organizations, or rather than others, in order to minimize such promote performance improvement (Audit discrepancies. Another possible distorting Commission, 2000). influence is the pressure to opt for PIs for It is common practice to distinguish PIs aspects that are easy to measure. Linked to that measure aspects of economy, efficiency, this is another possibility that the bodies effectiveness and quality (HM Treasury, 1992). required to produce the PIs may, for admin- Economy refers to the extent to which costs of istrative reasons, seek to limit the number of inputs are minimized. An example perfor- measures they have to generate. A conse- mance indicator would be the costs of admin- quence of such restrictions or distortions may istration as a percentage of total costs. be the emergence of what could be termed Efficiency refers to the ratio between inputs ‘perverse incentives’. For example, a school and outputs. An example performance may attempt to restrict entry and encourage indicator would be, in relation to the work of poor achievers to leave in order to improve its the Employment Service, the average cost relative position in a table of performance involved in placing an unemployed person based on examination results. in a job. Effectiveness identifies the extent to It is generally recognized that PIs must be which a particular outcome achieves the orig- used in a wider performance management inal objective of the programme. Taking arrangement, which will also include, for another Employment Service illustration, an example, direct performance measures, inspec- example performance indicator would be the tions and audits, and research and evaluation numbers (or proportions) of people leaving studies. Some would argue (Ranson and the unemployment register. Quality indicators Stewart, 1994) that the possibility of achieving relate to the usefulness or value of a service an adequate performance measurement (they are sometimes seen as an element of system for public sector activities is fundamen- effectiveness). A performance indicator from tally limited because of such factors as the local government would be the percentage of inherent complexity of performance, the fact

219 P-Jupp-3373.qxd 3/8/2006 5:48 PM Page 220

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

that market ‘forces’ do not apply in many Alfred Schutz (1962). An associated school of instances, and that final judgements on out- psychology has been vigorously developed by comes have to be the subject of political a small group of enthusiasts, without attaining debate. widespread influence, though, on its more philosophical side, it has been advanced by Note Maurice Merleau-Ponty (1962).

The above material does not represent the views Distinctive Features of the Department of Education and Skills, UK. For Franz Brentano the study of the domain of conscious awareness was the basic field for Vince Keddie a scientific psychology. All human experience was characterized by intentionality. According Associated Concepts: auditing, evaluation to Brentano ([1874] 1973), ‘Every mental phe- research, good practice studies, indicator, nomenon is characterized by what the meta-analysis, policy-related research, social scholastics in the Middle Ages called the indicators intentional (and also mental) inexistence of an object, and what we could also, although Key Readings in not entirely unambiguous terms, call the Audit Commission (2000) On Target: The reference to a content, a direction upon an Practice of Performance Indicators. London: object’. Thoughts, perceptions and so on have The Audit Commission for Local this direction whether or not there really Authorities and the National Health exists such an object as that to which the Service in England and Wales. experience points. HM Treasury (1992) Executive Agencies: A After a brief but promising career as a math- Guide to Setting Targets and Measuring ematician, including serving as assistant to the Performance. London: HMSO. great Weierstrass in Berlin, Husserl turned Pollitt, C. (1993) Managerialism and the to philosophy as a student in Vienna with Public Services. Oxford: Blackwell. Brentano. He seems to have accepted the gen- Ranson, S. and Stewart, J. (1994) Management eral outlines of Brentano’s point of view while for the Public Domain: Enabling the Learning at the same time being dissatisfied with his Society. London: Macmillan. teacher’s account of the relation between act Van den Berghe, W. (1997) Indicators in and object. Consciousness is directed. It is as if Perspective: The Use of Quality Indicators in of an object, whatever the status of that object Vocational Education and Training. might be. Surely the task of the philosopher/ Thessaloniki, CEDEFOP – European psychologist is to get to the bottom of what that Centre for the Development of Vocational ‘as if of’ relation might be. Training. Kant had emphasized the role of concepts in the genesis of objects perceived, shaping an inchoate flux of sensation into a determinate perception. Husserl too realized that somehow PHENOMENOLOGY the mind brings it about that there are objects without creating them. How can we explore Definition what it is that is implicit in the structure of conscious experience that this should be possible? A school of philosophy developed in the first Sometimes it seems as if Husserl is necessarily quarter of the twentieth century by Edmund locked within his own subjectivity, as he turns Husserl (1859–1938). It has been influential his attention exclusively to his own experience. in sociology, particularly in the writings of Yet, he insists that he is not denying that a real

220 P-Jupp-3373.qxd 3/8/2006 5:48 PM Page 221

PHENOMENOLOGY

world exists. His project is only to identify the whatever their nature. However, we still lack marks in our experience that express that basic a method by which these ‘constituents’ of presupposition. Persons, in their fullness as mental acts could be brought to light. In our conscious beings, each with a unique point of everyday lives we just carry on in the ‘life view on the world, are as much givens as are world’ performing our ordinary tasks, and material things. Again, he bypasses the tradi- never pausing to ask ourselves about the tional problem of ‘other minds’ by insisting noemata that make our experiences as if that his project is only to extract and make of vegetables, guests and furniture. From the ‘visible’ the defining characteristics of our point of view of the sources of a psychology, experience as if of people and things. However, this step is the most important of all given his insistence on the root character of the Husserl’s innovations. domain of conscious experience it is not easy to Research into the constitution of experi- see how Husserl can consistently take such a ence consists in performing two kinds of stand. Acknowledging the natural attitude to reductions. An eidetic reduction is achieved experience as of an independent world, and as we turn our attention to the generic char- combining that with the revelations of the tran- acteristics or essences of whatever it is the scendental attitude leads to a conception of the experience points to. The transcendental life world, the world that human beings as cen- reduction or epoché is accomplished by turn- tres of consciousness actually inhabit. ing our attention to the noemata, noeses and The directedness of acts of consciousness hyle that are the structures of a certain act of towards objects is a cluster of structural prop- consciousness. We ‘bracket’ what we have erties of experience. This he calls the noema. hitherto taken for granted, that there is a per- Noematic structures are generic, and take son in the other armchair. Setting aside that specific form in particular thoughts and per- taken-for-granted aspect of experience as if of ceptions as noesis. Since noesis is particular, it brings the relevant structures of conscious- is a process in time, while the noema is time- ness to light. The method as a whole consists less. Like many another philosopher breaking in the phenomenological reduction in which free from the rigidity of Kant’s twelve a the transcendental and eidetic reductions are priori categories, Husserl makes a point of applied successively. emphasizing the possibility that many differ- Husserl himself achieved his greatest suc- ent noemata are compatible with any partic- cess with this research method in studying ular flux of sensations. the role of time in consciousness. There is a How can this line of analysis be made to fit back and forth between fulfilled and unfilled with Husserl’s insistence on the existential expectations and retrospective and prospec- fullness of the experience of things and tive aspects of my conscious experience of people? Noema can be thought of as clusters any current state of affairs. of anticipations of further experience. Suppose a hundred years ago someone saw Evaluation the moon as a sphere. The noema would have included the anticipated experience of the Phenomenology has been very much more curvature of its hidden side. influential on the continent of Europe than in The third element in the fullness of experi- the English-speaking world. It has played a ence is hyle, derived from the Greek word for major role in the development of ideas of ‘matter’. Not every logically possible noema is the social construction of reality (Berger and compatible with the conditions of experience. Luckmann, 1967). It was brought to North The flux of sensations is limited. America by Alfred Schutz in his influential Here then are the three ‘elements’ that studies (Schutz, 1962). The influence of are involved in the fullness of all experi- Husserlian ‘reductions’ can also be seen in the ence as if of the world and its inhabitants, techniques of ethnomethodology through

221 P-Jupp-3373.qxd 3/8/2006 5:48 PM Page 222

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

which the taken-for-granted world is made focus on the context in which photographs strange (Garfinkel, 1967). are produced, their visual content, their tangible material qualities and the context Rom Harré in which they are interpreted and made meaningful. Associated Concepts: discursive psychology, epistemology, ethnomethodology, hermeneu- Distinctive Features tics, methodology, ontology, positioning theory, Social researchers analyse photographs realism, relativism ‘found’ during their research and also produce Key Readings their own photographs during the research process. The use of photography might include Bell, D. (1990) Husserl: The Arguments of the the analysis of historical photographs from Philosophers. London: Routledge. public or private collections, the analysis of Berger, P. L. and Luckmann, T. (1967) The contemporary personal, published or exhib- Social Construction of Reality. London: ited photographs, the production of pho- Penguin Books. tographs during research as a visual record, Brentano, F. ([1874] 1973) Psychology from or in collaboration with informants as a way an Empirical Standpoint (trans. A. C. of learning about their lives, the study of Rancurrello, D. B. Terrell and L. L. processes of production and consumption of McAlister). London: Routledge & Kegan photographs, and the use of photographs to Paul. represent the findings of one’s work. Garfinkel, H. (1967) Studies in Ethnometho- Photographs can be analysed for both dology. Englewood Cliffs, NJ: Prentice Hall. the visual information they provide and the Gurwitsch, A. (1966) Studies in Pheno- menology and Psychology. Evanston, IL: meanings that are attached to them. The Northwestern University Press. visual content of photographs can provide Husserl, E. (1964) The Idea of Phenomenology information not revealed through verbal (trans. W. P. Alston and G. Nakhmikian). interviews or noted through observation. For The Hague: Nijhoff. example, who was at a particular event, or in Husserl, E. (1965) Phenomenology and the what order did people stand in a procession? Crisis of Philosophy (trans. Q. Lauer). Thus photographs (whether they belong to an New York: Harper and Row. informant’s own collection, are published, or Husserl, E. (1987) The Phenomenology of were taken by the researcher) can serve as Internal Time Consciousness. Ann Arbor, visual records and information sources. MI: University of Michigan Press. Nevertheless, social researchers do not Merleau-Ponty, M. (1962) The Pheno- usually simply study the visual content of a menology of Perception (trans. C. Smith). photograph in isolation. Instead they might London: Routledge & Kegan Paul. contextualize it by examining the social rela- Schutz, A. (1962) The Phenomenology of the tionships and cultural conventions that were Social World (trans. G. Walsh and F. involved in its production and the subjectivi- Lehnert). Evanston, IL: Northwestern ties by which it is interpreted. For instance, University Press. studies of people’s existing photographic collections often reveal the ideal configurations of relationships and experiences that exist in PHOTOGRAPHY their cultures. A good example of this is the study of family photography (see, for example, Definition Chalfen, 1987). Often family photographs represent the ‘ideal’ family, rather than the The analysis and production of photographs experienced relationships between family as part of social science research. This might members. Therefore it is often more interesting

222 P-Jupp-3373.qxd 3/8/2006 5:48 PM Page 223

PHOTOGRAPHY

not only to use photographic content as an and photographic books based on their information source, but to examine how research (see, for example, Kratz, 2002). and why photographs were taken and how Digital photography has impacted on the they are later used. In addition, social use of photography in research. First, the researchers are interested in how people new technologies invite innovations in field- interpret photographs. One method that work uses of photography, for example addresses this is ‘photo-elicitation’ by asking enabling researchers to show informants informants to comment on photographs in photographs in their camera screen. Second, interviews they are often inspired to provide online exchange and exhibition has changed knowledge and information that would not the way we can represent and share our pho- have emerged from verbal questioning tographic work with others (both viewers of (Harper, 2002). our academic work and informants whom we In addition to analysing photographic con- might invite to comment on the photographs tent and the narratives it both inspires and is by email as part of a photo-elicitation study). embedded in, social researchers are con- However, generally social science uses of dig- cerned with the material properties of pho- ital images employ new media to enhance the tographs. As material objects, photographs existing methods established for printed pho- are inscribed with additional cultural mean- tographs rather than signifying a radical ings and social uses. For example, researchers change of approach. might focus on a photograph’s smell or tex- ture, the writing on its back, how it is framed, Evaluation hung on a wall, or pasted in an album. All these additional material and sensory features Photography in some form or other is part of of photographs contribute to the meanings almost all cultures and societies. We use pho- they have for people and their significance tographs as part of our everyday communica- for researchers. tion and our extraordinary celebrations. As As well as studying other people’s pho- such it is appropriate that the production and tographs, social scientists take their own analysis of photographs form part of qualita- photographs or ask their informants to take tive research. photographs for them during fieldwork. However, social researchers’ uses of pho- A researcher’s own photographs represent tography should be theoretically informed. her or his own subjective view of the It is now generally accepted that notions of research context. However, the method of photographic realism are not appropriate ‘directed photography’ (see Banks, 2001), for understanding photography in social whereby informants tell researchers what to research. Instead photographs are seen as photograph, allows researchers to understand representations that are inevitably subjec- what is important to them. Alternatively, tive. Therefore it is necessary for researchers informants can be given cameras and briefed to be aware of their own intentions as both to document a particular aspect of their expe- photographers and analysts of images, and rience visually. By later discussing the pho- those of the people they photograph, and to tographs with the informant the researcher analyse how these subjectivities intersect to can combine visual and verbal methods to produce photographic meanings. Similarly, learn more about how the informant experi- it is not sufficient simply to analyse the ences her or his world. visual content of an image, without locating Finally, researchers increasingly use it within the social, cultural and personal photographs to represent the findings of their contexts that give it meaning and accounting work. This is still largely confined to illustra- for its materiality. tions in books. However, specialist journals Photography has a key role to play in such as Visual Studies include photo-essays social research. However, to avoid naïve and some researchers produce exhibitions realist approaches that have sometimes

223 P-Jupp-3373.qxd 3/8/2006 5:48 PM Page 224

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

characterized its use in the past, it is necessary example, argues that policy research differs to attend to the points made above. from theoretical research in three major respects. First, whereas theoretical research is mainly interested in explanations of social Sarah Pink phenomena and the construction of models and theories, policy-related research is primar- Associated Concepts: cultural research, ily concerned with finding answers to prob- ethnographic interview, ethnography, histori- lems which can then be turned into action or cal analysis, life history interviewing, natu- policy. Secondly, the audience for theoretical ralistic data, oral history, qualitative research, research tends to be other scholars working in triangulation, unobtrusive measures, visual the same discipline. The language of theoretical methods research, therefore, is specialized as it is only meant to be understood by members of a par- Key Readings ticular academic community. Policy-related research, in contrast, intends to reach a wider Banks, M. (2001) Visual Methods in Social audience, including decision makers and any- Research. London: Sage. one with an interest in the issue being Chalfen, R. (1987) Snapshot Versions of Life. Bowling Green State University: Popular researched. As such, the technical jargon Press. of academic disciplines is replaced by more Edwards, E. (1992) Anthropology and Photo- easily understood forms of expression. Thirdly, graphy. New Haven, CT: Yale University Hakim argues that whereas theoretical Press. research is geared towards reporting any statis- Harper, D. (2002) ‘Talking about pictures: a tical association between variables, policy case for photo-elicitation’, Visual Studies, research is more interested in the size and 17 (1): 13–26. strength of relationships. As she puts it: ‘results Kratz, C. (2002) The Ones that Are Wanted. [in policy research] must report large and Berkeley, CA: University of California Press. strong effects rather than small but statistically Pink, S. (2001) Doing Visual Ethnography: significant effects’ (Hakim, 2000: 7). Images, Media and Representation in Bechhofer and Paterson (2000), however, Research. London: Sage. disagree with Hakim’s distinction between theoretical and policy research. They argue that it forces too rigid a separation between the two. In practice, policy researchers do address theo- POLICY-RELATED RESEARCH retical issues and are interested in causal explanations even though these matters are not of primary concern in the short-term exigen- Definition cies of policy making. Theoretical researchers, Research which the various stakeholders in the for their part, can have an important role in the decision-making process make use of in the for- policy process. When decision makers engage mulation, implementation or evaluation of in major reviews of policy areas it is sometimes policy. The research may not always have been necessary to undertake exercises that map out meant to be of use in policy making, but when or model the salient variables, factors and it is so designed (or commissioned) we can processes, often in an abstract or conceptual speak of policy-related research. way. Such exercises can usefully draw on the work of theoretical researchers. Distinctive Features Evaluation There has been debate about the features, if any, that distinguish policy-related research A great deal has been written about how from other kinds of research. Hakim (2000), for research can relate to policy; much less about

224 P-Jupp-3373.qxd 3/8/2006 5:48 PM Page 225

POLICY-RELATED RESEARCH

how it does. In the former category, rational This could not be said of the New Labour or knowledge-driven models see research administration elected to power in the UK in contributing information that will directly 1997. In a departure from what was seen help policy makers choose between alternative as ideologically driven policy making in the courses of action. Critics see this as a misrep- 1980s and early 1990s, Prime Minister Blair resentation of the policy process: decisions, it and his colleagues committed themselves is argued, are rarely made through the rational to evidence-based policy, epitomized by the analysis of options against predetermined slogan ‘what matters is what works’ (Davies, goals. Instead, policies are politically negotiated et al., 2000). As a result, there have been vig- compromises, and problems are typically dealt orous attempts since then to commission with in a piecemeal or incremental fashion more policy-relevant research, and also to (Lindblom, 1980). According to this view, ensure that new policy initiatives are rigor- research can contribute to policy making in a ously piloted and evaluated before being fully variety of ways, and should be seen as one of implemented. many influences. Carol Weiss, who has made a significant contribution to discussion about the uses of research (Weiss, 1979, 1992), John Newton reminds us that researchers are one of many participants in the policy process, including Associated Concepts: action research, politicians, advisers, civil servants and journal- applied research, critical research, emancipa- ists. Any of these actors can use research – tory research, evaluation research, politics often selectively – to suit their own purposes, and social research for example to bolster their own case or deni- grate that of their opponents. The suggestion here is that research is mainly used indirectly Key Readings and that evidence of the direct impact of research is limited. Bechhofer, F. and Paterson, L. (2000) Several explanations have been put forward Principles of Research Design in the Social to account for the relatively low uptake of Sciences. London: Routledge. social science research as compared with, say, Davies, H., Nutley, S. and Smith, P. (eds) medical research. In some accounts the focus (2000) What Works? Evidence Based Policy is on research and researchers: either social and Practice in Public Services. Bristol: science research is not usable because it does Policy Press. not have the precision associated with medical Hakim, C. (2000) Research Design: Successful science, or social scientists are not inclined to Research Designs for Social and Economic be ‘useful’ and prefer to keep their distance Research, 2nd edn. London: Routledge. from policy makers (Payne and Cross, 1991). Leicester, G. (1999) ‘The seven enemies of Other accounts emphasize how difficult it is evidence-based policy’, Public Money and Management, 19 (1): 5–7. for research to permeate the policy process. Lindblom, C. (1980) The Policy Making This may be something to do with the nature Process. Englewood Cliffs, NJ: Prentice of policies themselves, that is, they are the Hall. product of a political process and as such rep- Payne, G. and Cross, M. (1991) Sociology in resent a compromise of what it is possible to Action: Applications and Opportunities for do rather than what is the rational or ‘best’ the 1990s. London: Macmillan. thing to do (Leicester, 1999). Policy makers, Weiss, C. (1979) ‘The many meanings of too, may be less than accommodating to the research utilization’, Public Administration results of research. The economist John Review, 39 (5): 426–31. Maynard Keynes, for example, is often quoted Weiss, C. (1992) Organizations for Policy as saying that there is nothing a government Analysis: Helping Government Think. hates more than to be well informed. Newbury Park, CA: Sage.

225 P-Jupp-3373.qxd 3/8/2006 5:48 PM Page 226

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

that research should be directed towards POLITICS AND SOCIAL RESEARCH achieving other goals: social improvement or social change of some kind, causing a Definition shift in public policy or professional prac- A narrow definition states that politics is tice, the emancipation of the oppressed or the pursuit or exercise of state power. the pursuit of social justice. Broader definitions portray politics as the (2) The sponsorship of research, including its exercise of any kind of power; or as action funding, and the implications for problem relying on value commitments or furthering setting and research design. There are those particular interests. Some conceptions of who insist that there should be a close, social research, whether as serving public and interactive, relationship between policy making or as the work of intellectuals researchers and those who sponsor their committed to a particular political party or research or are likely to use it in policy cause, make it political in the narrow sense. making and practice. Others argue that In the broader sense, most researchers would it is essential for researchers to remain acknowledge, or insist, that their work is independent of both sponsors and users political (and it is difficult to see how it could in how they carry out and report their not be in some senses). work. (3) The effects on inquiry of the prior attitudes, assumptions and commitments, etc. of the Distinctive Features researcher. Some declare that researchers The relationship between social research and cannot but be biased, in the sense that politics is complex. This is because each of they cannot avoid having prior attitudes, these terms can refer to a range of activities, assumptions and commitments which and because the relationships among these can will shape how they do their research. take a variety of forms. Furthermore, the issue Others, while recognizing the danger of involves judgements about how things are, political bias, insist that researchers can how they can be, and how they ought to be. and should seek to minimize this. What comes under the heading of (4) The social relationships between researchers ‘research’ varies in ways that have implica- and researched. There are those who tions for the relationship to politics. Some require that researchers include the peo- research is commissioned by and/or designed ple they are studying as partners in deci- to serve government departments, interest sion making about how the inquiry is to groups, commercial organizations or political be pursued, on the grounds that to do parties. More broadly, many researchers see research ‘on’ people reproduces an themselves as public intellectuals whose task unequal society. At the other extreme are is to facilitate democratic discussion of social those who insist that the people studied problems. However, much social research has must not be allowed to control the a more distant relationship to state policy, research, that the maximum possible being concerned with myriad features of power must be exercised by the human life – micro as well as macro, histori- researcher in order to extract the infor- cal as well as contemporary. mation required (notably in studying If we turn to the relationship between poli- dominant social groups). tics and social research, this has a number of (5) The implications and consequences of aspects about which contrasting views can be research findings for policy and practice. taken: Some believe that unless research has a progressive impact on human social (1) The immediate goal of research. Some life it is worthless, and that researchers argue that the only goal should be the are responsible for achieving this. Others production of knowledge. Others believe argue that, while researchers must

226 P-Jupp-3373.qxd 3/8/2006 5:48 PM Page 227

POLITICS AND SOCIAL RESEARCH

endeavour to produce knowledge that is any immediate goal other than the production relevant to matters of human concern, of knowledge (see Hammersley, 1995: ch. 6). there is no obligation or distinctive right Furthermore, it is important to recognize on their part to draw policy conclusions that researchers may take on their role or from their findings or to work for the may carry out a particular piece of research implementation of these. because they hope that this will contribute to achieving political goals, without conduct- Over the past several decades, broader ing inquiry in such a way as to serve those interpretations of ‘politics’ have become goals. more influential. There has also been a shift Of course, the relationship between poli- away from seeing research as an activity in its tics and social research is not determined own right, towards emphasizing the contribu- solely by the views, or even by the actions, of tion it can, and should, make to political researchers. It is also shaped by the environ- activity, in both broad and narrow terms. As ment in which they operate, and in particular a result, research has increasingly been the pressures they come under. These can regarded as properly having practical or polit- arise from commercial organizations and ical goals, in addition to (and sometimes political pressure groups as well as from gov- instead of) any commitment to producing ernments. Some funders place restrictions on knowledge. For some, this requires a greater the publication of findings, or they may role in the research process for sponsors, cancel funding where the research seems likely users and/or the people studied. Along likely to produce what they regard as bad with this, claims to value neutrality or objec- news (see Crossen, 1994; Norris, 1995; tivity have come to be dismissed by many as Pettigrew, 1994). There are even external pres- ideological, as simply serving the interests of sures towards the adoption of some research researchers (see Hammersley, 2000). methods rather than others. And published These arguments are sometimes used as research may be subjected to public critique a basis for insisting that all research should where it is deemed to have undesirable impli- be explicit about the political agenda it is cations. Political efforts to ‘manage’ knowl- serving; in other words that researchers must edge in this and other ways are likely to have make clear ‘whose side they are on’ or what an increasing impact on the pursuit of social they are working for. However, a few com- inquiry in the future. mentators have recognized that if some political goals are to be served effectively then researchers must be covert about their politi- Martyn Hammersley cal intentions, not least because otherwise they would be cut off from funds or from Associated Concepts: applied research, important sources of information. For this bias, critical research, emancipatory research, reason, claims to objectivity or value neutral- participatory research methods, policy-related ity may not be abandoned even by those who research, sensitive research, standpoint see their research as serving political goals, research but used instead as a conscious ideological strategy (Back and Solomos, 1993). Key Readings Evaluation Back, L. and Solomos, J. (1993) ‘Doing research, writing politics: the dilemmas of Given the complexity of the issue, important political intervention in research in distinctions need to be recognized. Inquiry racism’, Economy and Society, 22 (2): may inevitably be political in some senses – 178–99. for instance, in that all researchers have polit- Crossen, C. (1994) Tainted Truth: The Mani- ical attitudes – but need not be political in pulation of Fact in America. New York: others – for example in the sense of pursuing Simon and Schuster.

227 P-Jupp-3373.qxd 3/8/2006 5:48 PM Page 228

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

Hammersley, M. (1995) The Politics of Social positioning theorists is to regard social Research. London: Sage. episodes as displaying story lines, as if they Hammersley, M. (2000) Taking Sides in Social were the living-out of narrative conventions. Research: Essays on Partisanship and Bias. For example, an apparently mundane episode London: Routledge. might be conceived and lived by one of the Norris, N. (1995) ‘Contracts, control and participants as the undertaking of a heroic evaluation’, Journal of Education Policy, quest. Analysis of the expressive conventions 10 (3): 271–85. displayed in that actor’s performance reveals Pettigrew, M. (1994) ‘Coming to terms the story line being lived out. Many social with research: the contract business’, in episodes are carried through wholly by dis- D. Halpin and B. Troyna (eds), Researching cursive means, by things said; the story of Education Policy. London: Falmer. a conflict may be almost all there is. Only occasionally is blood let, are bombs thrown and houses stoned. Many conflicts are sustained by the adoption by the hostile parties POSITIONING THEORY of conflicting story lines, in the light of which incompatible and irresolvable contradictions Definition in meanings have become entrenched. The other major influence that has nour- Positioning theory involves the need to attend ished positioning theory has come from to local moral orders and the viewpoint that feminist studies. The psychology of interac- the local distribution of rights and duties to tions between women and men seems to determine different kinds of acts determines involve presuppositions about the rights each the way episodes unfold. sex allows the other as to the kinds of social acts that a man or a woman can be under- Distinctive Features stood to perform. Deborah Tannen’s (1990) study of how men and women interpret what The shift from a cause/effect paradigm to a each says in conversation highlighted persist- rule/meaning pattern for identifying and ing tendencies of mutual misunderstanding. explaining psychological phenomena has led Positioning theory offers a way of getting at to the development of new models of psycho- the underlying presumptions that sustain logical processes, particularly those based on such misunderstandings and the conflicts an analogy between cognition and ‘conversa- that stem from them. It seems that what peo- tion’. Standard methodologies like the ‘ques- ple are taken to mean by what they say and tionnaire method’ have been reinterpreted do is partly a matter of what the various as revealing rules of discourse. At the same people involved in a social episode believe time attention has shifted to the interpersonal that persons of this or that category are enti- nature of psychological phenomena, both in tled to say and do. Such entitlements are mature forms and in development (Vygostky, called ‘positions’. Children, for instance, are [1929] 1962). usually not accorded the same speaking The inspiration for the development of rights as adults. They are positioned as recip- positioning theory as an analytical scheme ients of disciplinary admonitions rather than to deal with complex patterns of discursive as sources of them. acts, be they in words or making use of some Social psychology seemed to stand in need other medium, came partly from longstand- of a new approach that combined attention to ing dissatisfaction with the essentially static the small-scale and very rapid dynamics of conception of the psychological bases of social interactions with insights into the pre- social interactions. This conception had taken sumptions about rights and duties that were root in the 1950s and 1960s tied in with the invoked, explicitly or implicitly, by the actors. concept of role. The alternative preferred by Everyone does not have equal access to the

228 P-Jupp-3373.qxd 3/8/2006 5:48 PM Page 229

POSITIVISM

local repertoire of meaningful actions. Some up to nation-states – whenever there is an members of a group are more advantageously issue of the distribution of rights and duties positioned than others. Some categories of among the participants. What people do is persons are accorded rights and duties distinct drawn from what they believe they may do, from those of others in the same episode. which is itself drawn from what they believe In many cases, people are satisfied with they can do. Positioning analysis allows the their rights. In other cases, the distribution of social psychologist to examine the relations rights and duties can be challenged (Davies between ‘does’, ‘may do’ and ‘can do’ in spe- and Harré, 1999). Revealing the subtle pat- cific social and cultural circumstances. terns of the distributions of rights to speak and act in certain ways can open up the possibility of their transformation. At the Rom Harré same time, analysis of patterns of meaningful actions in terms of story lines can bring to Associated Concepts: constructionism, con- light previously unnoticed presumptions versation analysis, discourse analysis, discur- about what is going on in an episode. sive psychology, epistemology, interpretive The phenomena to be identified and repertoires, phenomenology, realism, relativism classified in social interactions, hostile or friendly, are the repertoire of meanings avail- Key Readings able to the actors. How are activities of the ‘meaning-making’ and ‘episode-forwarding’ Davies, B. and Harré, R. (1999) ‘Positioning kinds distributed among participants? and personhood’, in R. Harré and L. van To find the right level of analysis, one must Langenhove (eds), Positioning Theory. be careful to maintain a distinction between Oxford: Blackwell. pp. 32–52. intended actions and the various acts that can Tannen, D. (1990) You Just Don’t Understand. be performed by carrying out those inten- New York: William Morrow. tions. Two actions that come under the same Vygtosky, L. S. ([1929] 1962) Thought and Language (trans. E. Hanfmann and G. material description – say, shaking hands Vakar). Cambridge, MA: MIT Press. with someone – may serve for the performance of different acts in differently developing episodes. One may be a greeting and another may be a way of offering congratula- tions. Positioning theory is concerned exclu- POSITIVISM sively with analysis at the level of acts; that is, of the meanings of actions. It is only rarely Definition necessary to take account of the vehicles by which acts are performed. One of two methodological approaches recog- The basic principle of analysis is that posi- nized by many textbooks on social research, tions, as open sets of rights and duties, speech the other being: interpretivism. Positivism is and other act categories and story lines are said to be the methodological underpinning of mutually determinative. Fixing one vertex of survey research and experimental approaches, the positioning triangle fixes the other two. though in describing it thus there is a confla- Successfully challenging any vertex redefines tion between scientific approaches to social both of the others. research in general and the particular position of positivism. Evaluation Distinctive Features Positioning analysis can be applied to episodes of different scales – from individual There have been several versions of posi- persons face to face, to rival scientific teams, tivism, and though all favour a scientific

229 P-Jupp-3373.qxd 3/8/2006 5:48 PM Page 230

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

approach to investigation they are not (3) Logical positivism. This is also known as synonymous with science and indeed in the logical empiricism. Logical positivism natural sciences positivism disappeared as was enormously influential on US social an important methodological position many science in the period of its great expan- decades ago (Philips, 1987). Nevertheless, it sion in the 1930s. Though it had its crit- was via the positivist movement that social ics, such as R. S. Lynd and C. Wright science became science and was the form in Mills, positivism continued to dominate which scientific method entered the study social science research up until the of the social world. There are three main 1960s in the United States and was the forms of positivism that have been historically principle methodological approach in important to social science, though arguably most other countries where social none is present to any great extent today. research was undertaken. Indeed, it has left a lasting legacy, in the United States, (1) Comtean. The term positivism was in the form of a strong quantitative tra- coined by Auguste Comte (1798–1857), dition. The term ‘logical’ was not much along with the term ‘sociology’. His use used in social science and the different of the word positivism implied that this historical strands of positivism became was a positive philosophy and little to do conflated, particularly by its critics. with later positivist philosophy of science influenced by empiricism. He did, Positivist social science was criticized for however, advocate the development of ignoring the importance of individual subjec- sociology as a science, but in the context tivity and the role of consciousness in shap- of a shift in societal belief systems from ing the social world, though the vigorous theological, to metaphysical and finally criticism of positivism from the 1960s needs scientific. This was his Law of the Three to be seen in the context of an anti-science Stages. Comte’s positivism has had little sentiment and a growth in the popularity of influence on social research and is only interpretivist methods in studies of the social of historical importance in sociology. world. Unlike the natural sciences, social (2) Durkheimian. Emile Durkheim’s science has proposed few alternative scien- (1858–1917) influence was altogether tific methodologies to positivism, the main greater and he was one of the pioneers of exception being the realist programme (see, the numeric analysis of large data sets in for example, Pawson, 1989; Byrne, 2002). his comparative international study of suicide (Durkheim, 1952). In recent Evaluation years the positivistic aspects of Durkheim’s work have been less accen- The complex history of positivism and its con- tuated and indeed his emphasis on flation with science generally have rendered ‘social facts’ as knowable objects places any simple definition difficult. However, the him closer to realism than positivism. Table compares some key characteristics of Nevertheless, he equally stressed the science with some key characteristics of importance of scientific detachment and positivism. Positivists would align themselves value freedom in sociological research, with the characteristics of both columns, an approach shared by the later logical whereas methodologists committed to scien- positivists. However, unlike the latter, tific method in social science need only Durkheim was a methodological holist align themselves with the items in the first arguing that social facts (for example, column. rules, laws, practices) could not be Some writers, for example Philips (1987), reduced to explanations at an individual have argued that positivism is dead. It is cer- level. tainly true that few social scientists would

230 P-Jupp-3373.qxd 3/8/2006 5:48 PM Page 231

POSTMODERNISM

Key characteristics of science and positivism

Scientific method Positivism

Explanatory Prioritizes observation Predictive Verificatory (procedures should show whether a statement Evidence-based is true of false) Seeks truth Value-free (moral values have no part in science) Objective Operationalist (scientists can only deal in those things Logical which are measurable) Parsimonious Numerative

Source: Williams, 2003

now claim to be positivists and the more radical representations of the characteristics in POSTMODERNISM the second column are now rare. What lives on, particularly in survey research, is more Definition properly referred to as an empiricist tradi- Any ‘ism’, is more difficult to summarize as a tion, which in its various forms may draw on single movement or concept than any other. some or none of the latter characteristics, but Indeed, its anti-foundationalist character would certainly identify with the characteris- would be quite at odds with such a summary tics in the first column. or classification. The best that can be done is to provide some clarification of what is meant Malcolm Williams by the term ‘postmodern’ and how it impacts upon social research. Associated Concepts: empiricism, experiment, methodology, ontology, phenomenol- Distinctive Features ogy, postmodernism, realism, relativism, social survey, subjectivity A plethora of labels exist to describe both a condition and a movement and some of the thinkers described as postmodernist reject Key Readings the appellation. Some will argue that there is a distinction between post-structuralism Byrne, D. (2002) Interpreting Quantitative and postmodernism whilst others see no Data. London: Sage. important distinctions, or at least see each Durkheim, E. (1952) Suicide. London: as having common intellectual roots in post Routledge & Kegan Paul. positivism and the advocacy of ‘the abandon- Pawson, R. (1989) A Measure for Measures: A Manifesto for Empirical Sociology. London: ment of the entire intellectual culture of Routledge. modernity’ (Delanty, 1997: 95). Though the Philips, D. (1987) Philosophy, Science and Social roots of postmodernism are complex intel- Inquiry: Contemporary Methodological lectually and historically there are some Controversies in Social Science and Related commonalities of emphasis. First, on the cen- Applied Fields of Research. Oxford: trality of language in the production of the Pergamon. individual and the social, and second, the Williams, M. (2003) Making Sense of Social claim that there has been a shift in the nature Research. London: Sage. of society from modernist, homogeneous and

231 P-Jupp-3373.qxd 3/8/2006 5:48 PM Page 232

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

production-led to postmodernist heterogeneous anti-foundationalist rhetoric of postmodernism. and knowledge-led. The linguistic turn, par- The implications for social research are pro- ticularly in the work of Jacques Derrida, is as found. Those who have described their much associated with post-structuralism as research as ‘postmodern’ have placed empha- postmodernism, but provides the basis of a sis on the textual characteristics of data and ‘method’, in the form of deconstruction, for their interpretation. Thus investigations have postmodernism. Derrida emphasized the used interpretive methods, but have denied importance of language and focused on the any claim to epistemological privilege as a role of discourse. Yet unlike the structuralists, result of research. Whilst non-postmodern he made no separation between ‘things’ and interpretivists might aim to clarify meaning or the language that describes them: instead, he produce valid representations of the people maintains difference is produced through the studied, this is not the case in a postmodern difference in significance speakers attach to approach. Some postmodern anthropologists words (Williams and May, 1996: 166). The have attempted to avoid producing any repre- social and cultural world can be read like a sentations at all, for (it is claimed) ‘the point text, but importantly no one reading can be of discourse is not how to make a better rep- privileged and all readings are equally valid, resentation, but how to avoid representation’ only the emphasis changes, according to (Tyler, 1986: 128), the argument being that the reading. Meaning is never fixed, nor read- any representation amounts to a privileging ily apparent. Though Derrida denies being of one reading of a text over another, when all a postmodernist, the relativization of knowl- are equally valid. edge, particularly scientific knowledge, as par- Paradoxically, postmodernists often advo- tial and local has been a key feature of cate the abandonment of political meta- postmodernism since. narratives but politicize the social research The writings of Jean François Lyotard and process through the use of deconstruction ‘to Jean Baudrillard strongly emphasize such lin- examine the presuppositions that are buried guistic anti-foundationalism, but also give in the texts produced in the course of every- prominence to social context – what Lyotard day activities’ (Williams and May, 1996: 169). termed ‘the postmodern condition’ in society It is held that the task of social research is the (Lyotard, 1979/1984). Lyotard’s view was that deconstruction of texts to expose values and the myths of truth in science and liberation interests embedded in them. In doing so they through science and scientific rationality had must avoid themselves producing new hege- become undermined since the early 1960s. monies of knowledge. What are termed ‘meta-narratives’ (in social science these might include Marxism and func- Evaluation tionalism for example), attempt to explain the social world through transcultural and trans- Whilst postmodernism was taken seriously historical truths. These, he maintains, obfus- in theoretical writings in social ‘science’, its cate and are ultimately to be seen as simply impact on research has been minimal, possi- discourses (albeit historically powerful at the bly because the activity of research is time), ‘texts’ that can be differently interpreted. inevitably imbued with notions of the search In the social sciences the denial of scientific for explanation and representation. The rationality as a basis for social investigation problem of postmodernist research was well shifts the emphasis of the study of the social summed up by Todd Jones when he said from the scientific to the cultural. Methodo- ‘Postmodern social scientists, I fear, success- logical objectives and values, originally derived fully avoid giving biased hegemonic views from natural science ‘discourse’, such as about the people they study at a cost of not explanation, generalization, validity and reli- giving us much information at all’ (1998: 58) ability, have no useful meaning within the and ‘I believe that spending a great deal of

232 P-Jupp-3373.qxd 3/8/2006 5:48 PM Page 233

PRACTITIONER RESEARCH

time and money studying a people in order ontology, positivism, realism, relativism, not to represent what they are like is a grave reliability, validity waste of resources’ (ibid). Even if one accepts the correctness of postmodernism’s Key Readings epistemological relativist stance, this does beg the question of whether there is then Byrne, D. (2002) Interpreting Quantitative any point to doing research. Furthermore, its Data. London: Sage. moral relativism would seem to forbid any Cilliers, P. (1998) Complexity and Post- modernism: Understanding Complex Systems. political or policy project and thus any London: Routledge. research that would provide data to support Delanty, G. (1997) Social Science: Beyond such projects. Constructivism and Realism. Milton Though the popularity of postmodern Keynes: Open University Press. approaches is in decline, in social science it Jones, T. (1998) ‘Interpretive social science and may leave some useful legacies. The emphasis the “native’s” point of view’, Philosophy of on language, though not new, did at least the Social Sciences, 28 (1): 32–68. remind social scientists of its importance in the Lyotard, J. (1979/1984) The Postmodern Con- creation and maintenance of social order and dition: A Report on Knowledge. Minneapolis: the multiple layers of meaning in language (or Minnesota University Press. ‘texts’ generally) which can produce variable Tyler, S. (1986) ‘Post modern ethnography: and unintended consequences in individual from document of the occult to occult doc- thought and action. Secondly, the emphasis on ument’, in J. Clifford and G. Marcus (eds), the heterogeneity of social and cultural forms Writing Culture: The Poetics and Politics has led to a sensitivity toward their complexity. of Ethnography. A School of American At least one writer (Paul Cilliers) has taken the Research Advanced Seminar. Berkeley, CA: relationship between postmodernism and com- University of California Press. plexity in the literal sense of the mathematical Williams, M. and May, T. (1996) Introduction complexity of systems, whereby ‘the interac- to Philosophy of Social Research. London: tion between the system and its environment, Routledge. are [sic] of such a nature that the system as a whole cannot be fully understood simply by analysing its components’. Complex relationships are nonlinear, they are not fixed, ‘but shift and change, often as a result of self organ- PRACTITIONER RESEARCH isation.’ (Cilliers, 1998: x–ix). Cilliers’s work is perhaps representative of Definition the view that whilst we can accept many of Research concerned with issues and prob- the arguments about a societal shift from the lems that arise in professional practice. It is modern to the postmodern, it does not follow conducted by practitioners and aims to bring that we should abandon the goals of explana- about change, or influence policy in the prac- tion or representation. Though it may well tice arena. Practitioner research provides a mean that we will need to develop new tools framework for formulating practice knowl- of investigation, or use our existing tools in edge and allows such knowledge to be dis- quite different ways (Byrne, 2002). seminated to other professionals.

Malcolm Williams Distinctive Features The development of practitioner research Associated Concepts: cultural research, has a number of roots. The establishment of deconstruction, discourse analysis, empiricism, action research, which itself takes a number of

233 P-Jupp-3373.qxd 3/8/2006 5:48 PM Page 234

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

forms, is significant, as too is the development the world and favours more emancipatory and of reflective practice. It is essentially an expres- participatory research practices (Fook, 1996). sion of the politics of researching professional Practitioner research is grounded in the practice, which reverberate through the control everyday world of practice and can reflect the of research, setting of research agendas and voices of stakeholders in that context. To be professional training. Through practitioner of value, research does not have to be large- research, practitioners become researchers into scale or ‘scientific’. Methods such as case their own practice and engage in a continuing study and life history are familiar territory process of professional development. for practitioners and have the potential to Some common concerns investigated by be more emancipatory than those methods researcher/practitioners include: deficiencies that do not give participants a voice. in services or resources, conflict between Formalizing practitioner knowledge through professional values and agency requirements, research enhances professional credibility and the improvement of practice. Practitioner and strengthens the influence of practice research provides a framework for formulat- knowledge in the policy-making arena. In ing practice knowledge and allows such addition, the products of this kind of research knowledge to be disseminated to other profes- are more accessible to practitioners because sionals. Many of the skills and competencies they are written by and for practitioners. associated with practice are transferable to the conduct of research. Practitioners are Evaluation trained to engage with clients in a respectful and meaningful way, and to encourage clients The participatory and emancipatory paradigms to express themselves. They collect informa- represent an overarching framework for tion and develop understandings that are then informing the conduct and evaluating the prod- disseminated to those concerned. Such activi- ucts of practitioner research. All research has ties are also central to the conduct of research. the potential to exploit and oppress participants A traditional, linear view of professional if attention is not paid to the power imbalances development is that theory informs practice, between researchers and researched. an idea that has led to the separation of Whilst practitioner knowledge is regarded practice and research both physically and as an asset in the conduct of practitioner intellectually. In recent years, this has been research, it can also be seen as a source of bias challenged by the notion of reflective practice, that could invalidate it. The researcher practi- which has been encompassed within and tioner role is potentially confusing for all those propagated through widespread developments involved in the process. For example, research in professional training. Reflective practice participants may find it difficult to distinguish is the capacity to think, talk or write about between the professional-as-researcher and a piece of practice with the intention to review the professional-as-practitioner and may be or ‘re-search’ a piece of practice for new puzzled by a professional’s unwillingness to meanings or perspectives on the situation. take up a practice issue during a research inter- Practitioner reflexivity is just as central to the view. Similarly, a practitioner may find it diffi- conduct of research as it is to practice. Indeed, cult to distinguish between questions that Schön suggests that practice and research are are meant to be about collecting data, and inextricably linked through the process of those that are aimed at promoting change. reflection. When someone reflects-in-action he Practitioner researchers may feel uneasy about or she becomes a researcher in the practice using practice knowledge and insight in their context (Schön, 1983). Adopting a reflective analysis of data, while respondents may well approach to research allows the practitioner to divulge information to a practitioner researcher, question the power relations of research. Such on the basis of a close, professional relation- an approach can affirm the importance of expe- ship, that they may not have given to a riential and interconnected ways of knowing ‘researcher’. Practitioner researchers need to

234 P-Jupp-3373.qxd 3/8/2006 5:48 PM Page 235

PREDICTION

examine their research practice in order to usually considered as the ultimate way to test identify whose interests and concerns are the explanatory power of a scientific theory. being served. While it is recognized that practi- In order to understand why social scientists tioner research should be designed to give a say attempt to make predictions, it is important to all participants and can involve practitioners to recognize from the outset that science trying to see issues from the viewpoints of assumes the world to be made up of a series of clients (Reed and Proctor, 1995), such a partic- cause-and-effect relationships. For example, if ipatory approach is at the discretion of the pro- Mr Smith jumped out of an aeroplane at 1,000 fessional, who is already in a position of power feet and realized half-way down that he had with clients. inadvertently forgotten his parachute, then we might reasonably conclude, on the basis of what we already know about gravity and solid Maureen Gillman and John Swain objects, that the effect of falling such a distance will be to cause our absent-minded Associated Concepts: action research, friend’s demise. However, no matter how applied research, disability research, emanci- often one might come across true instances patory research, good practice studies, partici- confirming this relationship, there always patory action research, policy-related research, exists the chance that the former will not cause politics and social research, reflexivity the latter. It’s all a matter of probability, a way of thinking which states that all cause-and- Key Readings effect relationships, even the most apparently obvious, must be viewed as a matter of chance Fook, J. (ed.) (1996) The Reflective Researcher. which have a higher or lower probability of London: Allen & Unwin. occurring (usually defined as a percentage Fuller, R. and Petch, A. (1995) Practitioner Research: The Reflexive Social Worker. score or a value between 0 and 1). After all, Milton Keynes: Open University Press. there are many ‘stranger than fiction’ accounts Reed, J. and Proctor, S. (eds) (1995) about people who defy the odds and survive Practitioner Research in Health Care. such horrific events. Scientists, therefore, are London: Chapman & Hall. currently working with an internal contradic- Schön, D. (1983) The Reflective Practitioner. tion by first assuming that there are such New York: Basic Books. things as cause-and-effect relationships that need to be determined, but then settling for the empirical reality that such relationships may not always exist. Thus, the act of making predictions PREDICTION acknowledges the existence of these chance fluctuations, even though it is currently Definition beyond our ability to understand why they A statement about what will be observed occur. Generally, there exist two procedures before the actual event, a foretelling of some for making predictive assessments. future happening. We are all in the business (1) the actuarial approach: a highly stan- of making predictions about something: dardized process which involves statistically tomorrow’s weather, this week’s lottery testing a pre-defined set of factors relevant to numbers, the outcome of sporting events, the a particular situation. For example, an actuar- likelihood of a terrorist attack. ial risk model designed to predict an individual’s likelihood of committing a criminal Distinctive Features offence will require: (a) precise data about the subject’s age, gender, number of youth cus- The ability to make accurate predictions tody sentences, total number of court appear- about the phenomena under consideration is ances, type of offence; (b) an established

235 P-Jupp-3373.qxd 3/8/2006 5:48 PM Page 236

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

(statistically validated) prediction formula, previous experience and ability to reason usually employing sophisticated multiple intuitively. regression techniques, into which the data are entered in order to calculate the subject’s Evaluation risk of committing another offence within a pre-defined time period (say the next two An import issue in the development of pre- years), and (c) an outcome in the form of a diction methods is whether the future behav- prediction scale result (Copas and Marshall, iour of an individual can be more accurately 1998). This result represents the subject’s measured through the use of either an essen- prediction score which will show that on the tially subjective understanding of the person basis of all the statistically relevant facts being assessed based on the clinical approach imputed into the formula, there exists an x or an objective evaluation of the subject using per cent chance that this person will repre-defined statistical scales. One solution offend within the next two years. If the value has been to look at developing combined of x were 75 per cent, then we would classify risk/needs assessment measures that attempt this person as a high risk. to integrate the respective strengths of both It is important to note, however, that measures (Colombo and Neary, 1998). the social scientist is not making a predictive Scholars have also noted society’s growing judgement directly about this particular person, preoccupation with employing increasingly even though it was this subject’s data that was sophisticated prediction technologies in order used to make the assessment. All that the sci- to identify, assess, prevent and manage risk, entist can say is that out of a group of 100 peo- especially regarding a myriad of risk issues ple with similar characteristics and histories within areas such as crime, medicine, personal of offending as this person, 75 will re-offend lifestyle, leisure pursuits, economics and the within the next two years. With prediction environment. According to Adams (1995), risk measures it is not possible to say whether the is now ‘big business’ and in fact constitutes the individual subject we are currently concerned ‘world’s largest industry’. The word ‘risk’ has about constitutes one of the 75 who will re- several literal origins that translate as ‘run offend or one of the 25 who will not. into danger’ or ‘take a chance’ and is a highly (2) The clinical approach: this contrasts complex phenomenon with philosophical, with actuarial prediction and represents a social, ethical and political consequences. For highly interactive process between the example, should we punish criminals on the person making the assessment and the risk basis of their past offences or their predicted subject. The clinical method involves risk of re-offending in the future? employing in-depth interviewing and observational skills in order to uncover informa- Anthony Colombo tion about the precipitating factors underlying each subject’s behaviour. The data relied Associated Concepts: causal model, causa- upon to make predictions differ according to tion, econometrics, estimation, forecasting, the circumstances of each case. Typically, the inferential statistics, regression analysis data are highly subjective as they are largely dependent upon the beliefs, values and normative judgements held by the person doing the predicting. The knowledge base of this Key Readings method appears to focus on the predictor’s Adams, J. (1995) Risk. London: University of ability to collect data and then employ their Central London Press. own impressions and ideas as a means of gen- Beck, U. (1992) Risk Society. London: Sage. erating information about a particular subject. Colombo, A. and Neary, M. (1998) ‘Square In turn, these impression/ideas are probably roots and algebra: understanding percep- formulated as a consequence of their training, tions of risk/needs assessment measures

236 P-Jupp-3373.qxd 3/8/2006 5:48 PM Page 237

PRIMARY RESEARCH

in probation practice’ Probation Journal, Examples of approaches associated with 45 (4): 213–19. quantitative research are experiments, struc- Copas, J. and Marshall, P. (1998) ‘The tured observation, survey, and statistical offender group reconvictions scale: a analysis of documents and texts. Experiments statistical reconviction score for use by examine behaviour in a controlled environ- probation officers’, Journal of Applied ment, typically a laboratory setting, and strive Statistics, 47 (1): 159–71. to determine causal relationships by precisely Kemshall, H. (2003) Understanding Risk in manipulating variables. Structured observa- Criminal Justice. Milton Keynes: Open tion involves observing behaviour in terms University Press. of predefined categories; for example, a researcher may observe the occurrence of different types of activity in a classroom. Surveys may be conducted using methods PRIMARY RESEARCH such as structured interviews or questionnaires, which use predetermined questions to Definition elicit self-report data on attitudes, behaviour and beliefs, in a format amenable to numeri- The generation of new data in order to cal analysis. Analysis of documents and texts address a specific research question, using in quantitative research (for example, content either direct methods such as interviews, or analysis) places emphasis on precise coding indirect methods such as observation. Data and statistical analysis. are collected specifically for the study at Examples of qualitative approaches hand, and have not previously been inter- include ethnography and participant observa- preted by a source other than the researcher. tion, semi-structured and unstructured interviews, focus groups and analysis of language. Distinctive Features Ethnography and participant observation involve the researcher becoming immersed Traditionally, primary research has been cate- in a group or community and observing gorized into quantitative and qualitative language and behaviour in order to gain approaches. These approaches can be seen as insight into people’s understandings. Semi- forming two distinctive clusters of research and unstructured interviewing emphasizes strategy, where research strategy refers to a letting the interviewee guide the nature of the ‘general orientation to the conduct of social topics and themes covered in the course of an research’ (Bryman, 2001: 20). The distinction interview, and avoids limiting the discussion is not always clear-cut, however, and some to what has been predetermined by the studies may share characteristics of both researcher. Focus groups are group inter- quantitative and qualitative approaches. Charac- views, facilitated by the researcher, but with teristics of quantitative approaches are a focus a focus on letting the group guide discussion on numerical data and statistical analysis, with topics. Analysis of language (for example, emphasis on producing objective, reliable, conversation analysis, discourse analysis) in valid data. Qualitative approaches are charac- qualitative research emphasizes exploring the terized by a focus on language (rather than meanings inherent in language use. numbers), and an emphasis on participants’ The process of carrying out primary interpretations and understandings of their research can be seen as involving a number social world. The two strategies are under- of stages, including: formulate a research pinned by different ontological and epistemo- question, select a research design, choose an logical assumptions, quantitative approaches appropriate method, administer the method being associated with objectivism and positi- to collect data, analyse the data, interpret the vism, and qualitative approaches with con- results in relation to the original research structionism and interpretivism. question. Emphasis on either of these stages

237 P-Jupp-3373.qxd 3/8/2006 5:48 PM Page 238

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

can vary depending on the type of research different and therefore cannot be united (for strategy and methods being adopted. example, Smith, 1983).

Evaluation Claire Hewson Primary research allows the social researcher to gather data appropriate and tailored to the Associated Concepts: content analysis, specific research question to be addressed. data, ethnography, experiment, focus group, However, it can be a time-consuming and methodological pluralism, mixed-methods costly process, especially in large-scale research, participant observation, qualitative survey research, for example. The concepts of research, quantitative research, reliability, validity and reliability have been considered social survey, validity fundamental in primary research, the former referring to whether the conclusions drawn Key Readings from a study are accurate and correct, and Bryman, A. (2001) Social Research Methods. the latter whether the results can be repli- Oxford: Oxford University Press. cated (for example, if people obtain vastly dif- Robson, C. (2001) Real World Research: A ferent scores every time they take an IQ test, Resource for Social Scientists and then the test is not reliable). In quantitative Practitioner-Researchers. Oxford: Blackwell. research these concepts have been consid- Smith, J. K. (1983) ‘Quantitative versus qual- ered crucial, and largely focus around the itative research: an attempt to clarify the idea of accuracy of measurement; however, issue’, Educational Research, 12: 6–13. qualitative approaches are less concerned with precise measurement and there has been some debate about the extent to which these concepts are applicable. Some authors have argued that alternative concepts should PROBABILITY (RANDOM) SAMPLING be applied in judging qualitative research, such as ‘credibility’ and ‘confirmability’. Definition Historically, social researchers have debated the relative merits of quantitative and Any method of sampling that uses some form qualitative approaches. Criticisms levelled of random selection, that is, one that will against quantitative approaches include a lack ensure that all units in the population have of ecological validity (behaviour in con- an equal probability or chance of being strained and artificial conditions is not repre- selected. Random selection is an assumption of sentative of behaviour in the real world), and probability theory and the ability to draw infer- the failure to recognize that, unlike molecules ences from samples to populations. Random and atoms, people interpret their social world sampling techniques include: the simple ran- and therefore the natural science model is dom sample; the stratified random sample; the inappropriate in social research. Qualitative systematic random sample; and multi-stage research has been criticized for being overly cluster samples. Probability sampling is most subjective and lacking replicability and closely associated with quantitative research, generalizability. Recently there has been and particularly with survey research. recognition that a mixed-method approach, which combines aspects of both qualitative Distinctive Features and quantitative research, may be useful. However, some authors have objected to the In simple random sampling all units within multi-method approach, arguing that the the sampling frame have an equal chance of underlying philosophies of quantitative and being selected. Computer software is often qualitative approaches are fundamentally used to generate random numbers. MS Excel,

238 P-Jupp-3373.qxd 3/8/2006 5:48 PM Page 239

PROCESS MAPPING

for example, has a random number generator quantitative research and is appropriate for facility. Simple random sampling is not the making inferences from samples to popula- most statistically efficient method of sam- tions (within calculable levels of probability). pling as there may not be good representation It is, however, inappropriate in research where of the subgroups in a population. If, for exam- populations are transient or difficult to ple, the views of a minority group are to be access, or where a complete list of the popu- represented in the research, it may be neces- lation (sampling frame) is not available. In sary to weight the sample in order to ensure such cases it is not possible to ensure that that a large enough proportion are included. every unit in the population will have an Therefore, stratified random sampling is equal chance of selection. often used (as simple random sampling does not necessarily mean that the sample selected will be representative of all the units within Julia Davidson a given population). Stratification enables the researcher to select specific categories, or Associated Concepts: cluster sampling, infer- strata, for inclusion within the sample. ential statistics, non-probability (non-random) However, stratified random sampling is only sampling, opportunity sampling, purposive feasible when the relevant information on the sampling, quota sampling, sampling, snowball stratifying criteria is available. sampling, social survey, volunteer sampling Systematic random sampling is also known as proportionate sampling and uses Key Readings a fixed sampling interval to determine the Bryman, A. (2001) Social Research Methods. sample. Here it is important to ensure that Oxford: Oxford University Press. the manner in which the list is constructed May, T. (2001) Social Research: Issues, does not introduce some form of bias. For Methods and Process, 3rd edn. Milton example, if a particular family name is very Keynes: Open University Press. common (in Wales, for example, you may Sapsford, R. and Jupp, V. (eds) (1996) Data frequently find the name Evans) then people Collection and Analysis. London: Sage. with that name might stand more chance of Silverman, D. (ed.) (1997) Qualitative being included. Care would need to be exer- Research: Theory, Method and Practice. cised if this were to lead to over- or under- London: Sage. representation. Multi-stage cluster sampling de Vaus, D. A. (1996) Surveys in Social can be used to avoid the wide dispersal Research, 4th edn. London: UCL Press. that simple random sampling or systematic random sampling could yield in some forms of research, such as a large, national survey. For example, at the first stage a random PROCESS MAPPING sample of cities might be selected; at the second stage a sample of districts in the chosen cities; at the final stage all individuals in the chosen Definition districts may be interviewed. In this way, field- A method of identifying the formal and infor- work is concentrated in particular districts, mal structures and processes within an with considerable savings of time and costs. agency or organization involved in the delivery of a particular function. Evaluation Distinctive Features Sampling decisions are an important part of the research design for both qualitative and Process mapping often forms an impor- quantitative approaches. Random or proba- tant feature of research techniques such bility sampling is considered most suited to as evaluative methods, since it is generally

239 P-Jupp-3373.qxd 3/8/2006 5:48 PM Page 240

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

implicit to an understanding of the mechanisms partner agencies in practice. It may also underpinning the programme or initiative to be shed light on organizational failings that have evaluated. It may therefore help to explain the impeded the delivery of a project. Process reasons why a particular project has either suc- mapping may also be essential to the dissem- ceeded or failed to meet its objectives. It may ination of good practice, since replication can also inform the research process in terms of be ineffective in organizations with differing suggesting the types of data that should be col- processes. However, the exposure of varying lected through the identification of the critical local practices may in itself compromise the stages of a project that require measurement. overall research process, since it suggests Process mapping may be a pre-requisite to pro- that, in terms of outputs and outcomes, one is jects where the organizational structure for the not comparing like with like. delivery of the project is not recorded or is Whilst, in some circumstances, variations unclear, or where national organizations have in project outcomes may be partially attribut- devolved responsibilities to regional areas. This able to the alternative processes between is a particular feature of, say, evaluations of organizations, this should not be overstated. police initiatives, where local practices within Discrepancies in findings may be attributable the forty-three police force areas in England to variables other than processes, such as and Wales vary widely. In such circumstances, the response that a particular initiative gener- process mapping provides a full picture of the ates or the context in which it operates, and operational context of a particular research accordingly process mapping should be used project. Therefore, mapping takes the form of as part of a wider evaluation of a project. an investigation, which may entail interviews with large numbers of personnel, and can be very labour-intensive. Helen Poole In agencies where processes vary, mapping can help to explain differences in perfor- Associated Concepts: applied research, mance or outcomes, thereby revealing the cost–benefit analysis, evaluation research, good comparative effectiveness of different organi- practice studies, organizational research, policy- zational structures. In the UK and elsewhere, related research the growth in multi-agency approaches to a variety of social interventions has increased Key Readings the need for process mapping, as an opera- Laycock, G. (2002) ‘Methodological issues in tional understanding of a partner’s systems working with policy advisers and practition- and organization may be fundamental to the ers’, in N. Tilley (ed.), Analysis for Crime development of future working practices and Prevention. Cullompton, Devon: Willan. for improving effectiveness. For example, Pawson, R. and Tilley, N. (1997) ‘An intro- criminal justice agencies in the UK are now duction to scientific realist evaluation’, required to work with a number of other in E. Chelimsky and W. R. Shadish (eds), service providers, such as health, education Evaluation for the Twenty-First Century. and social services. An understanding of how Thousand Oaks, CA: Sage. pp. 405–18. systems operate to allow this interaction in practice may be a fundamental component of evaluating their effectiveness in terms of reducing crime and reforming offenders. PROGRESSIVE FOCUSING Evaluation Definition Process mapping can perform a very important function in its own right since it can A term used in discussions of ethnography, provide research funders with an enhanced and sometimes of qualitative research gener- understanding of the operation of its own and ally. It represents the process of inquiry as

240 P-Jupp-3373.qxd 3/8/2006 5:48 PM Page 241

PROJECTIVE TECHNIQUE

involving a gradual clarification, and some- the unit of inquiry is the single investigation, times transformation, of the research prob- rather than a programme of studies addressing lem; and perhaps also a re-formulation of the a particular issue. Within such a programme, initial research questions into hypotheses different studies could have varying emphases that can be tested against data from the field. across the spectrum from exploration to hypothesis-testing. Distinctive Features A very different criticism of the idea of progressive focusing challenges the assumption that The need for progressive focusing reflects social inquiry can produce accounts that repre- the fact that ethnography is governed by an sent social phenomena, insisting that creating ‘inductive’ or ‘discovery’ orientation, rather illuminating or entertaining stories is its only than by a conception of inquiry which possible goal. Alternatively, there are those who requires specific hypotheses to be set up for would question whether ‘focus’ is required in testing at the start of the process. From this research, the proposal being that researchers point of view, inquiry is as much about dis- should produce writerly, rather than readerly, covering the right questions to ask as about texts: that is to say, texts that are open-ended or testing hypotheses. This orientation is cap- only loosely organized in internal structure, tured in Glaser and Strauss’s notion of thereby encouraging the reader to create his or grounded theorizing, which they specifically her own creative understanding from the mate- oppose to ‘verificationism’, emphasizing the rials provided. The model here is those forms value of developing theoretical ideas through of art and literature that subvert conventional the analysis of data (Glaser and Strauss, 1967). ways of seeing, or forms of knowledge. In the Underlying the concept of progressive context of research, such ideas challenge – focusing is an emphasis on the fact that any indeed, some would say that they completely formulation of initial research problems or undermine – the very concept of inquiry itself. questions involves presuppositions – about relevant social phenomena and how they can be known – and that these might turn out to Martyn Hammersley be false. Such a concern arises, in part, from an emphasis on the way in which human Associated Concepts: ethnography, grounded actions are culturally diverse, so that under- theory, hypothetico-deductive model, induc- standing them requires learning the relevant tion, research problems, sensitizing concepts, culture. However, the danger of setting out on theoretical sampling false premises is recognized even in the context of natural scientific research, where cul- Key Readings tural variation is not an issue. Glaser, B. G. and Strauss, A. (1967) The Evaluation Discovery of Grounded Theory. Chicago: Aldine. The concept of inquiry underlying progressive Hammersley, M. and Atkinson, P. (1995) focusing has been questioned from at least two Ethnography: Principles in Practice, 2nd directions. Some argue that the emphasis it edn. London: Routledge. gives to the exploration of ideas and of data means that insufficient time is allowed for the testing of hypotheses. A result of this is that ethnographers may engage simply in pattern- PROJECTIVE TECHNIQUE making or storytelling, without the necessary degree of attention to whether the patterns Definition or stories are sound representations of the phenomena being investigated. While there is Any technique in which the respondent is some truth in this argument, it assumes that given some task to complete that it is

241 P-Jupp-3373.qxd 3/8/2006 5:48 PM Page 242

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

assumed will tell the researcher something behind, to that person’s values, attitudes, about the respondent. The technique is perceptions or inner feelings. The risks, espe- regarded as a projection of the respondent’s cially with unstructured methods, are that perceptions and attitudes. inferences are dependent on the interpretations on the part of the researcher of a Distinctive Features respondent’s responses and traces and that different researchers would make different Projective techniques vary in the degree of inferences. their structure: for example, attitude scales whereby an individual’s score is used to make inferences about that person’s attitude Victor Jupp towards a particular topic tend to be highly structured in their format and in the response Associated Concepts: indicator, scaling, trace options available. Whilst this assists data measures, unobtrusive measures, validity analysis and facilitates comparison between subjects there are potential problems: for Key Readings example, that an inappropriate structure is Moser, C. A. and Kalton, G. (1971) Survey imposed or indeed that there is too much Methods in Social Investigation. London: structure. Also, there is the risk of acquies- Heinemann. cence effects (where a respondent acquiesces Oppenheim, A. N. (1966) Questionnaire with what he or she believes the researcher Design and Attitude Measurement. London: wants) and social desirability effects (where a Heinemann. respondent attempts to portray him/herself in a desirable light). For such reasons, less structured techniques that permit free-flow responses are sometimes used. Such techniques can vary in their degree of ambiguity, PROSPECTIVE STUDY with the most ambiguous stimuli used to explore an individual’s inner states or feel- Definition ings (sometimes in clinical or therapeutic context). They include abstract paintings, A study that follows cases forward in time, cartoons, single words and Rorschach ink- measuring attributes at multiple time points. blots. The sentence completion test, in which Change is measured by examining differences a respondent is given a number of sentence between each time point or study wave. beginnings to complete, is less structured. Unlike experimental designs, prospective Unobtrusive or trace measures and also designs do not include randomized control garbology can be viewed as forms of projective groups or experimental interventions. techniques in so far as they may tell the researcher something about individuals. For Distinctive Features example, how a person dresses, what music he or she listens to or what a person eats can be Social researchers employ four main types of used to make inferences about that person. prospective design. The first is the ongoing single cohort design. This type of study con- Evaluation sists of a single cohort that is tracked indefi- nitely. The cohort may be a birth cohort, as in The use of a projective technique in research the UK-based Millennium Cohort Study, an can be very stimulating and creative. educational cohort (the graduating class of However, the main problem is one of making 2000) or a marriage cohort (those marrying valid and reliable inferences from a person’s in 2000). A second type of prospective responses to stimuli, or traces which are left study tracks multiple cohorts. Investigators

242 P-Jupp-3373.qxd 3/8/2006 5:48 PM Page 243

PROSPECTIVE STUDY

using the multiple cohort data ideally use depends on the length of the gap between broad-based representative samples from waves and how much reliance must be which they may focus on single cohorts placed on recall data to establish the within the sample or construct and track sequence of events within the gap between multiple cohorts from the sample. The British study waves. Household Panel Survey and the US Panel Since prospective designs track the same Study of Income Dynamics are excellent cases over time they have a major advantage examples of this type of prospective study. over repeated cross-sectional studies which A third type of prospective study is the use the same measures at different points of rotating panel design consisting of a series of time but recruit a new sample each time. staggered, limited-life panels. This design is Tracking the same cases over time allows used in many national labour force surveys investigators to identify change at the indi- where a sample (panel) is interviewed each vidual level while repeated cross-sectional month for, say, six months and then leaves studies only allow the identification of the study. Each month a new, six-month change at the aggregate level (Rose, 1995). panel is added to replace one whose six- Finally, prospective designs avoid the prob- month involvement ends. lems of the reconstruction of the past that A fourth type of prospective design links are encountered in retrospective studies. data from administrative records and does Since information is, ideally, collected about not require interviewing people at each time contemporary events, behaviour and atti- point. An example of such a design might tudes the problems of memory, selective focus on the period for which lone mothers recall and the reconstruction of the past are receive government income support. Admin- avoided. istrative records are kept regarding when However, prospective studies can encounter recipients of income support begin receiving important problems. One of the major prob- payments, when they stop, the reason for lems encountered is attrition. Tracking people stopping (obtained job, change of relationship across time, especially if waves are frequent, status etc.), whether they return to the pay- can impose a considerable burden on respon- ment system and so forth. From these records dents which can cause them to drop out. This a data set can be constructed containing in turn can affect the representativeness of information regarding the length of time for the study. Simply keeping track of people can which lone mothers typically receive pay- be difficult – especially of the more mobile ments, the number of times they move off members of a population. The effort required payments, the reasons for which they leave to track respondents can add consider- the payment system and the number of times ably to the cost of prospective studies. they return. Representativeness can be undermined in Prospective studies can vary in a number other ways. If a sample is collected to repre- of other ways. These include how many sent a population at a particular point of time waves of data are collected, whether the and that sample is tracked over time the sam- study has a finite life or is ongoing and how ple can become unrepresentative if the nature long the gap between waves will be. of the wider population changes due to certain types of people leaving and other types joining Evaluation the population through outward and inward migration. The major strength of prospective studies is A prospective study is not appropriate if a that they can help establish the time quick answer is required. Change must be sequence of events and thus establish causal studied in real time rather than the ‘com- direction more effectively than can either pressed time’ that is used in retrospective cross-sectional or many retrospective studies. studies. Depending on the types of changes However, their capacity to do this effectively to be observed this may take many years.

243 P-Jupp-3373.qxd 3/8/2006 5:48 PM Page 244

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

A further difficulty with most prospective designs is that they lack a randomized control PURPOSIVE SAMPLING group, which means that uncontrolled and unmeasured factors may be responsible for Definition observed changes. While multivariate analy- A form of non-probability sampling in which sis can control for the influence of measured decisions concerning the individuals to be factors the problem of the influence of included in the sample are taken by the unmeasured factors remains. researcher, based upon a variety of criteria Studies of change are only as good as are which may include specialist knowledge of the measures being used at each time point. the research issue, or capacity and willing- Unless highly reliable measures are used it ness to participate in the research. can be difficult to distinguish between real change and apparent change due to measure- Distinctive Features ment error. This problem is compounded in studies where there is likely to be develop- Some types of research design necessitate mental change (for example, in studies of researchers taking a decision about the indi- children), where measurement instruments vidual participants who would be most likely need to be adapted between waves to ensure to contribute appropriate data, both in terms that they are age-appropriate and therefore of relevance and depth. For example, in life measure the underlying construct (for exam- history research, some potential participants ple, intelligence) reliably and validly. may be willing to be interviewed, but may not be able to provide sufficiently rich data. David de Vaus Researchers may have to select a purposive sample based on the participants’ oral skills, ability to describe and reflect upon aspects of Associated Concepts: attrition, causality, their lives, and experience of the specific cohort study, cross-sectional survey, experi- focus of the research. ment, longitudinal study, one-shot design, A case study design is another type of panel study, representativeness, retrospective research that often requires a purposive study, sampling, survey, validity sample. Imagine that a research team wishes to explore the types of academic support pro- Key Readings vided for students in a single high school. In selecting that school the researchers may Buck, N., Ermisch, J. and Jenkins, S. (1996) need to take a variety of factors into account. Choosing a Longitudinal Survey Design: The They may want a school in which academic Issues. Occasional Paper 96–1. London: support is sufficiently innovative to make the Institute for Social and Economic final research report of wide interest in the Research. profession. They will require a school in Duncan, G. and Kalton, G. (1987) ‘Issues of which the management are supportive of design and analysis of surveys across the research, and in that the teachers and time’, International Statistical Review, 55: students show a willingness to participate. 97–117. Menard, S. (1991) Longitudinal Research. The researchers may want a school that is Quantitative Applications in the Social exceptional in terms of overall academic per- Sciences Series, 76. London: Sage. formance, or that has an average level of Rose, D. (1995) ‘Household panel studies: attainment. Finally, they may prefer a school an overview’, Innovations, 8: 7–24. http:// that is reasonably accessible for members iserwww.essex.ac.uk/pubs/occpaps/index. of the research team. When all relevant php factors have been considered, the research de Vaus, D. (2001) Research Design in Social team will select the case study school, which Research. London: Sage. chs 7–9. will constitute the purposive sample. If it is

244 P-Jupp-3373.qxd 3/8/2006 5:48 PM Page 245

PURPOSIVE SAMPLING

appropriate, a purposive sample may be of the research, and the criteria used for combined with a probability sample. Once selecting the purposive sample. the high school has been selected, a random sample of teachers and students could be Paul Oliver selected from whom to collect data.

Evaluation Associated Concepts: case study method, cluster sampling, good practice studies, life The advantage of purposive sampling is that history interviewing, non-probability (non- the researcher can identify participants random) sampling, opportunity sampling, who are likely to provide data that are quota sampling, snowball sampling, validity detailed and relevant to the research question. However, in disseminating the findings, the researcher should make fully transparent Key Readings the criteria upon which the sampling process Krathwohl, D. R. (1998) Methods of was based. Educational and Social Science Research, The principal disadvantage of purposive 2nd edn. New York: Longman. sampling rests on the subjectivity of the May, T. (2001) Social Research: Issues, researcher’s decision making. This is a source Methods and Process, 3rd edn. Milton of potential bias, and a significant threat to Keynes: Open University Press. the validity of the research conclusions. Robson, C. (2002) Real World Research: A These effects may be reduced by trying to Resource for Social Scientists and ensure that there is an internal consistency Practitioner-Researchers, 2nd edn. Oxford: between the aims and epistemological basis Blackwell.

245 Q-Jupp-3373.qxd 3/8/2006 10:44 AM Page 246

(4) create nodes to represent ideas and QSR NVIVO hold coding, and manage these flexibly in an index system; Definition (5) store attributes of sites, data sources, A software package for qualitative researchers. people, organizations etc. and use these From the developers of the NUD*IST software, in searching and sorting data. Attri- NVivo provides a range of tools for handling butes of documents or nodes are man- data, ideas, information and theories built up aged as in a spreadsheet; their values from observations, interviews, document can be text, numeric or date/time; analysis, literature reviews and other qualita- (6) code any characters in the text at new or tive research processes. Like NUD*IST, NVivo existing categories, using drag and drop (for ‘NUD*IST Alive’) supports coding and or other rapid methods, whilst viewing retrieval of coded material, searching and theo- coding displays in margin stripes; rizing, combined with ability to annotate and (7) ask questions about any occurrences of edit documents. But NVivo is designed for words in text, or patterns of coding or methods requiring more flexible development attributes, and scope the question to of rich data in dynamic documents, and more exactly the data to be explored; com- subtle ways of linking data to ideas and show- bine such questions to explore data ing and reflecting on the results. iteratively: results are saved as more coding; Distinctive Features (8) display and explore relationships in live matrices of coding, attributes or NVivo pioneered rich text in QDA software, text, to investigate patterns, jumping to freely edited and coded, and the integrating of the text represented by any cell; coding with many ways of qualitative linking, (9) show, filter and assay the relationships shaping, searching and modelling. Users can: of data, using tools for seeing patterns, (1) create documents or import and freely integrated with tools for searching; edit them in a rich text editor and edit (10) draw models of emerging ideas, use them at any time without affecting cod- layers to show stages in the analysis or ing or pointers to other data; different interpretations, and jump to (2) include non-text data (e.g. video clips) the data represented by any item in the by hyperlinks to sound, image or any model; other files; and call up that data (links (11) using a separate program, Merge for appear live in coded text); NVivo, align projects in great detail for (3) manage documents and nodes flexi- thorough comparison of their emerging bly in sets that can be filtered and analyses, and optionally merge them explored; for a wider picture. Q-Jupp-3373.qxd 3/8/2006 10:44 AM Page 247

QUALITATIVE DATA ARCHIVES

Evaluation Distinctive Features

The design of NVivo responds to widespread Qualitative data archives are growing in experience with early code and retrieve soft- number around the world. There are two ware and to complaints that such software established qualitative archives at the present encouraged students to a linear approach to time, with more in progress. The aim of these data gathering, coding and reporting. Whilst archives is to enhance and assist social many researchers still use it mainly for code- research by obtaining, sharing and re-using based analysis, it offered many (for some, too qualitative data. many) ways of linking data and ideas and of Archives hold a vast quantity of data directing analysis accurately to the pertinent collected from various research projects. In data. In a second revision (2002), the options, most cases the archives also hold details of particularly for search and filter, were clari- the methodological approach of the research fied and streamlined. Used in graduate and project in order to contextualize the data set. undergraduate methods teaching, across an Usually an archive holds a diverse range of expanding range of disciplines, NVivo materials, which can include interview tran- became the first QDA software to be sup- scripts open questionnaires, newspapers, ported by teaching textbooks. magazines, diaries, letters, audio tapes, video footage, observation schedules and field notes from the researcher. Lyn Richards In some cases, an archive will allow original copies of data (such as hard copies of Associated Concepts: Atlas.ti, CAQDAS, The questionnaires or field notes) to be used by Ethnograph, NUD*IST, qualitative research new researchers. More commonly, data are archived by scanning the original material or Key Readings by transcribing interviews and field notes. Bazeley, P. and Richards, L. (2000) The NVivo A popular method of accessing such qualita- Qualitative Project Book. London: Sage. tive data is to link the archive to the Internet. Gibbs, G. (2002) Qualitative Data Analysis: This allows researchers worldwide access to Explorations with NVivo. Milton Keynes: qualitative data from different cultures and Open University Press. historical periods. Data can be extracted Morse, J. M. and Richards, L. (2002) Read Me using a search function that identifies the First for a User’s Guide to Qualitative Methods. most relevant material to the criteria entered Thousand Oaks, CA: Sage. by the user. Richards, L. (2002) Using NVivo2 in Qualitative Archived qualitative data can be manipu- Research. Melbourne: QSR International. lated in many ways. It can be sampled, and in most cases the archive will hold demographic data on the respondents from the original project. Therefore probability sampling tech- QUALITATIVE DATA ARCHIVES niques such as stratified sampling can be applied to the data in order to extract the most Definition appropriate cases. As the data are usually listed (by some criteria such as respondent A data archive is a tool used to preserve data sets number or case number), random and system- that are the products of research. In social atic methods of sampling can also be applied research, the majority of archives are used to by selecting cases from the list. In some preserve quantitative data in computer-readable instances it may be appropriate to apply non- format. A number of qualitative archives have probability sampling to the archived data. developed as qualitative methods have become When cases are hard to identify or are few in more popular and more sophisticated. number then using snowball or opportunity

247 Q-Jupp-3373.qxd 3/8/2006 10:44 AM Page 248

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

sampling can be useful. Each data set can be the extent to which it is ethical or valid to use used to link to similar data sets as archives a subject’s response to a particular question usually provide links to similar studies. or situation for a different purpose. An example of one such archive is the The data held by any archive must be evalu- Murray Research Centre archive at Radcliffe ated in terms of reliability, validity and authen- University in the United States. It holds ticity, as any other data would be. The research data from 1976 onwards and stores researcher must always address the way the some classic longitudinal research projects. data have been collected, for example whether The stock consists of video cassettes, audio informed consent was obtained or whether the tapes, transcripts from interviews and case research was covert. Researchers must consider histories. The archive can be searched using the suitability of the secondary use of data for the Internet and new researchers can down- the project in hand. Potential users of archived load data sets. data sets must bear in mind the theoretical The most advanced qualitative data archive approach taken by the original researcher and in the world is Qualidata held as part of the UK critically evaluate the original data, and the Data Archive at the University of Essex in the method in which they were collected, in terms UK. Qualidata has been acquiring data since of the perspective of the current research. 1994 and is the only stand-alone data archive of its kind. The archive is fully connected to the Internet and data can be searched, refined and Angela Brady downloaded by researchers. This archive aims to actively encourage Associated Concepts: data, ethics, Internet the secondary use of qualitative research data research, naturalistic data, opportunity sam- collected during ESRC (Economic and Social pling, purposive sampling, qualitative Research Council) funded projects. The ESRC research, sampling, secondary analysis stipulate that all projects they fund, which ethically qualify, must submit the data set Key Readings from the research to Qualidata. This archive is a pioneer in its field. Qualidata organizes ESDS website (December 2003) (Online): international conferences to promote qualita- www.esds.ac.uk/ tive data sharing and to encourage other IASSIST website (December 2003) (Online): countries to set up similar archives. It also http://datalib.library.ualberta.ca/iassist/ gives seminars on how to archive material to ICPSR (December 2003) Guide to Social Science Data Preparation and Archiving. social researchers around the world. ICPSR website (Online): www.icpsr.umich. Countries currently developing qualitative edu/access/dpm.html data archives are: Canada, Czech Republic, Ryssevik, J. and Musgrave, S. (2001) ‘The Denmark, Finland, France, Germany, social science dream machine: resource Netherlands, Slovenia and Switzerland. discovery, analysis, and delivery on the Web’, Social Science Computing Review, Evaluation 19 (2): 163–74. Qualitative data archives are a tool used to share and re-use social research data. Before any social researcher uses material from an archive they must address the ethical issues QUALITATIVE RESEARCH this creates. Perhaps most importantly, the prospective user must address the context in Definition which the data were collected and assess whether this would tally with the current Research that investigates aspects of social research project. Another important issue is life which are not amenable to quantitative

248 Q-Jupp-3373.qxd 3/8/2006 10:44 AM Page 249

QUALITATIVE RESEARCH

measurement. Associated with a variety of emphasis on depth and detail of understanding theoretical perspectives, qualitative research and interpretation, that it is often small-scale uses a range of methods to focus on the or micro-level. meanings and interpretation of social phenomena and social processes in the particular Evaluation contexts in which they occur. Qualitative research is sometimes seen as Distinctive Features lacking the rigour of quantitative research, producing ‘soft’ data that is subjective and Qualitative research is not a single set of theo- not easy to replicate, often based on small retical principles, a single research strategy or samples or case studies. There are difficulties a single method (Silverman, 1993). It developed in meeting the usual scientific criteria of in the nineteenth and twentieth centuries, validity, reliability and representativeness. across a range of disciplines, on varied and However, the extent to which these criteria, sometimes conflicting philosophical and theo- derived from quantitative research, are retical bases, including cultural anthropology, applicable to the nature of qualitative interpretive sociologies (such as symbolic research has been a matter of debate (see, for interactionism), phenomenology and, more example, Hammersley, 1992; Seale, 1999). recently, hermeneutics, critical theory, femi- Because it is predicated on traditions that nism, post-colonial theory, cultural studies, point to the contested nature of social reality post-structuralism and postmodernism. These and which are critical of the idea of a single diverse approaches inevitably give rise to sub- objective ‘truth’ about the world, qualitative stantial differences and disagreements about research fits well with critical perspectives, the nature of qualitative research, the role of the such as feminism, which aim to challenge the researcher, the use of various methods and the political assumptions embedded in social analysis of data. institutions and in the research process itself. However, qualitative research is often Although qualitative and quantitative based upon interpretivism, constructivism and research have traditionally been seen as inductivism. It is concerned to explore the sub- opposed, and there are differences in terms of jective meanings through which people inter- the underlying philosophical approaches, these pret the world, the different ways in which differences are not always clear-cut (Layder, reality is constructed (through language, 1993). For example, qualitative research is images and cultural artifacts) in particular increasingly used for theory testing as well as contexts. Social events and phenomena are theory generation (Silverman, 1993). Many understood from the perspective of the actors studies have used a combination of qualitative themselves, avoiding the imposition of the and quantitative methods. The value of quali- researcher’s own preconceptions and defini- tative research in a variety of settings, includ- tions. There is also often a concern with the ing market research and applied social exploration of change and flux in social rela- research, has increasingly been recognized. tionships in context and over time. The methods used in qualitative research, often in combination, are those which are Maggie Sumner open-ended (to explore participants’ interpretations) and which allow the collection of Associated Concepts: constructionism, criti- detailed information in a relatively close cal research, discourse analysis, epistemology, setting. These methods include depth inter- ethnography, ethnomethodology, feminist viewing, ethnography and participant obser- research, inter-subjective understanding, vation, case studies, life histories, discourse methodological pluralism, naturalistic data, analysis and conversational analysis. It is in participant observation, phenomenology, the nature of qualitative research, with its quantitative research, Verstehen

249 Q-Jupp-3373.qxd 3/8/2006 10:44 AM Page 250

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

Key Readings statistical tests to be undertaken. For example, descriptive statistics can be used to illustrate Denzin, N. K. and Lincoln, Y. S. (eds) (1994) and summarize findings, detect relationships Handbook of Qualitative Research. London: between variables, as in correlation coeffi- Sage. cient values, or inferential statistical analysis Hammersley, M. (1992) What’s Wrong With Ethnography. London: Routledge. can be undertaken to establish the effects of Layder, D. (1993) New Strategies in Social different interventions, as in analysis of vari- Research. Cambridge: Polity Press. ance, analysis of covariance and multivariate Seale, C. (1999) The Quality of Qualitative analysis of variance (Johnson and Wichern, Research. London: Sage. 1998). Interactions between variables can Silverman, D. (1993) Interpreting Qualitative also be investigated within experimental Data. London: Sage. designs and also as part of the analysis of data from surveys or secondary sources (Pilcher, 1990). Changes over time can be more easily QUANTITATIVE RESEARCH tracked using quantitative methods, as measures of the same properties can be taken at several points during an intervention. For example, if Definition a community project is intended to engage Research involving the collection of data in young people, comparing those inclined numerical form for quantitative analysis. The towards antisocial behaviour with those who numerical data can be durations, scores, are not, measures of shoplifting, juvenile counts of incidents, ratings, or scales. arrests, nuisance calls to the civil services, Quantitative data can be collected in either such as fire and ambulance, can be used controlled or naturalistic environments, in across the time of the intervention in order to laboratories or field studies, from special establish the efficacy of the project. It would populations or from samples of the general be very difficult to establish actual effects population. The defining factor is that numbers without numerical data. result from the process, whether the initial data collection produced numerical values, or Evaluation whether non-numerical values were subsequently converted to numbers as part of the Quantitative research has certain strengths and analysis process, as in content analysis. weaknesses as a methodology, and is mostly associated with the positivist tradition. Distinctive Features Quantitative research produces ‘facts’ about the world and behaviour, and these are viewed Quantitative research tends to be associated as adding to the sum of human knowledge. with the realist epistemology, the approach to That is, the data thus collected tend to be knowledge that maintains that the real world accepted as they stand, and as valid measures exists, is directly knowable (although not of the variables they purport to indicate. necessarily at this moment) and that the real Qualitative researchers tend to criticize these world causes our experiences. That is, real methods on the basis that most sources of data things exist, and these can be measured, and are not quite what they appear to be. They have numerical values assigned as an outcome do not pay attention to social meanings and measure, and these values are meaningful. the ways in which the world is socially con- These values can only be meaningful if structed. Also, from the viewpoint of critical researchers accept some of the criteria asso- researchers, the data are obtained using meth- ciated with the positivist standpoint. ods where the person or group under study are Gaining numerical materials facilitates the given no status, being subject to unequal power measurement of variables and also allows relations (Tavris, 1993).

250 Q-Jupp-3373.qxd 3/8/2006 10:44 AM Page 251

QUASI-EXPERIMENT

There is a considerable tension between the Pilcher, D. M. (1990) Data Analysis for the qualitative and quantitative methodologies, and Helping Professions: A Practical Guide. the researchers who deploy them. It has been Newbury Park, CA: Sage. suggested that the qualitative methodologies are Tavris, C. (1993) ‘The mismeasure of woman’, best used when an area is little known, and so Feminism and Psychology, 3 (2): 149–68. hypotheses cannot be generated for testing by those who support the hypothetico-deductive method. But this viewpoint negates the place of QUASI-EXPERIMENT qualitative research as a methodology in its own right. Strong qualitative methodologists suggest that in quantitative research, the positivist view Definition of facts leaves no place for participants as An experiment that attempts to test a hypoth- agents, and that many constructs do not exist esis about the effects of an intervention by except in the social world, and so cannot be methods other than those used in a ‘true investigated outside social interaction. Prag- experiment’, where the latter is deemed to matists suggest that quantitative methods, on require random allocation to experimental the other hand, should be best deployed when and control conditions. more is known, so that hypotheses and research questions can be formulated, and easily tested. Distinctive Features Researchers intending to use any methodology need to have very clear ideas about the ques- Quasi-experiments are conducted when true tions they need to address, and the most appro- experiments are either impracticable or impos- priate ways of investigating them. sible. They are often used in evaluation The greatest advantage of quantitative research. True experiments supposedly deal at research is the fact that the data obtained via a stroke with alternative explanations for any these methods can be subject to considerable change brought about by the introduction of an statistical analysis, can generalize beyond the experimental measure. They do so by ran- sample under investigation, allowing the test- domly allocating potential cases to experimen- ing of hypotheses, and the evaluation of the tal and control conditions. Any variation in efficacy of interventions in various area of change between the sets of cases may be interest, including social policy. In addition, explained by reference to the presence in the experimentation would have no meaning one but not the other of the experimental without quantitative research methods. measure. Random allocation ensures that the experimental and control groups are equivalent, maturation will affect both groups Jeanette Garwood equally, events unrelated to the experiment will also affect both groups equally and so on. Associated Concepts: correlation, data, Selection bias and alternative sources of descriptive statistics, experiment, hypothetico- change in the experimental group are dealt deductive model, indicator, inferential statis- with by the design of the study. However, prac- tics, measurement, methodology, positivism, tical problems for true experiments do arise, qualitative research, realism, Verstehen for example, where the unit of intervention is too large for sufficient numbers to be allocated to experimental or control conditions, where a Key Readings policy is being introduced which will affect the Cramer, D. (1994) Introducing Statistics for whole population, or where it is not possible to the Social Sciences. London: Routledge. contain the measures introduced to the experi- Johnson, R. A. and Wichern, D. W. (1998) mental group in ways that prevent them from Applied Multivariate Statistical Analysis, 4th reaching potential members of control groups. edn. Englewood Cliffs, NJ: Prentice Hall. Under these circumstances, something short of

251 Q-Jupp-3373.qxd 3/8/2006 10:44 AM Page 252

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

a true experiment has to be undertaken faut de Key Readings mieux. Campbell, D. and Russo, M. Jean (1999) True experiments become especially diffi- Social Experimentation. Thousand Oaks, cult where communities rather than individ- CA: Sage. uals are the units of analysis. Communities Campbell, D. and Stanley, J. (1966) often cannot be insulated from one another, Experimental and Quasi-Experimental random allocation is often impracticable and Designs for Research. Boston, MA: finding sufficient experimental sites for sta- Houghton Mifflin. tistical analysis is generally prohibitively Pawson, R. and Tilley, N. (1997) Realistic expensive. In these circumstances quasi- Evaluation. London: Sage. experimental studies are conducted. Evaluation Non-equivalent pre- and post-test designs are QUESTIONNAIRE often used as a next best choice where true experiments cannot be conducted. Here, one or two experimental sites are identified along- Definition side one or two nearby areas, with similar social A set of carefully designed questions given in and physical attributes, to act as controls. One exactly the same form to a group of people of the key difficulties here is that the compar- in order to collect data about some topic(s) in isons of experimental and control areas which the researcher is interested. assume no significant difference in terms of the responses either would have to the experi- Distinctive Features mental measure. In the absence of tested theory there is no basis for this assumption. There is often a great deal of confusion in the True experiments cannot be generalized use and the meaning of the word ‘question- beyond the populations from which random naire’. For instance, some researchers reserve assignment has been made, and cannot assure this term exclusively for self-administered or external validity. Results of quasi-experimental postal questionnaires, while others would studies using non-equivalent groups of the include interview schedules (personally design described here do not even enjoy administered face-to-face) under the general internal validity. Alternative, theory-based rubric of ‘questionnaire’. Examples of the approaches in evaluation research attempt to latter type of questionnaire are market make good shortfalls in experimental methods researchers stopping shoppers in the street, and quasi-experimental methods which depend or a researcher conducting a study into on non-equivalent pre- and post-test designs. the retail habits of a particular group. Donald Campbell, the most influential Administered, or postal questionnaires, have writer on experimental and quasi-experimental many advantages. Even though the designs for the social sciences, recognized researcher should consider postal and print- that the quasi-experimental tradition resem- ing costs, postal questionnaires tend to be bles the laboratory experiment of the natural much less expensive than travelling to far sciences more closely than ‘true experiments’ away places to conduct interviews. Collected involving random assignment to treatment data can be inputted into a computer either (Campbell and Russo, 1999)! automatically by the use of a scanning device or manually by using computer programs Nick Tilley such as SPSS. This saves the researcher a great deal of time and, with the advent of Associated Concepts: causality, evaluation new technology, allows the researcher to research, experiment, field experiment, establish correlations between different vari- messy research, validity ables. In addition, as the researcher has no

252 Q-Jupp-3373.qxd 3/8/2006 10:44 AM Page 253

QUOTA SAMPLING

direct contact with the respondent, the issue of interviewer bias tends not to arise. QUOTA SAMPLING On the other hand, many criticisms can be levelled against this type of research method, Definition the largest one being response rate. The A non-probability method of selecting res- intended respondent may regard postal ques- pondents for surveys. The interviewer begins tionnaires as something of a nuisance; there- with a matrix of the target population that is to fore, the researcher may find that he or she be represented and potential respondents are can only obtain a 25 per cent response rate, selected according to that matrix. Quota sam- which is not unusual in these types of study. pling is also known as a purposive sample or a Those who do take the time to complete the non-probability sample, whereby the objective questionnaire may therefore be unrepresen- is to select typical, or representative, subjects tative of the sample population. A more sub- and the skill and judgement of selectors is stantive criticism of questionnaires is that deliberately utilized (Abrahamson, 1983). they simply cannot capture the true extent of a social phenomenon. Reality, according to this argument, is much more complex than Distinctive Features the few variables that are found in a ques- Quota sampling allows the researcher to tionnaire. Interpretative scholars often level control variables without having a sampling this criticism. Finally, it is erroneous to frame. This method is often used for market assume that questionnaires are an ‘easy’ research because it does not require a list of research technique. In order to understand a potential respondents. The interviewer finds topic, and therefore be able to create a good respondents, usually in public areas, who fit questionnaire, the researcher needs to under- into the predetermined categories until the stand the subject matter. Therefore, before quotas are filled. To that end, quota sampling designing a questionnaire, the researcher is a convenient and inexpensive method of must be fully aware of all of the various research. If the interviewee is unavailable or debates that surround it. refuses to participate they can easily be replaced with another potential respondent Evaluation who meets the same criteria. Questionnaires provide an excellent means of collecting large-scale quantitative data. The Evaluation closed-question nature of this technique, Statisticians criticize quota sampling for its however, demonstrates that a researcher may methodological weakness. Although the inter- need to employ further methods, such as a viewer ‘randomly’ chooses respondents he semi-structured interview, to gain a full or she comes across on the street, quota understanding of a given issue. sampling cannot be considered a genuinely random method of sampling because not every Craig McLean member of the population has an equal chance of survey selection (for example, those who are Associated Concepts: interview, quantitative at work or at home). Therefore, the principles research, research design, social survey of statistical inference cannot be invoked. There are a number of factors that can result in research bias. First, interviewers may mis- judge a potential respondent’s characteristic, Key Reading such as their age. Secondly, the interviewer runs Oppenheim, A. N. (1992) Questionnaire the risk of subconsciously making a subjective Design and Attitude Measurement. London: judgement before approaching a potential Pinter. respondent. As a result, the interviewer may not

253 Q-Jupp-3373.qxd 3/8/2006 10:44 AM Page 254

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

approach those deemed unfriendly and runs the Associated Concepts: inferential statistics, risk of distorting the findings. This is also non-probability (non-random) sampling, known as systematic bias (Abrahamson, 1983). opportunity sampling, probability (random) Finally, quota sampling can never be truly rep- sampling, purposive sampling, sampling, resentative because certain factors may prevent social survey, volunteer sampling certain groups of people from being chosen for the research. For example, as noted above, market research conducted during the day may Key Readings over-represent housewives shopping in the city centre and under-represent office workers. Abrahamson, M. (1983) Social Research Methods. Englewood Cliffs, NJ: Prentice Despite these limitations, quota sampling Hall. continues to be used because there are circum- Gilbert, N. (2001) Researching Social Life, stances when random or stratified random 2nd edn. London: Sage. sampling is not possible. Moser, C. A. and Kalton, G. (1971) Survey Methods in Social Research. London: Rachael Beth Moss Heinemann.

254 R-Jupp-3373.qxd 3/8/2006 5:49 PM Page 255

synthesis of sensation and mental constructs, RANDOM SAMPLING for advocates of critical realism aspects of the real can only be ascertained through rational See Probability (Random) Sampling abstraction. Indeed, the social relations that exist for critical realists, such as ‘capitalism’, can never be observed in the strict sense (that is, we can only observe individual companies REACTIVITY and groups of workers in the context of markets but not capitalism, the bourgeoisie or the proletariat). See Validity of Measurement Distinctive Features For empiricists the real world is composed REALISM of discrete entities that are not intrinsically complex and which in combination produce Definition the complex empirical patterns we can identify through our sensations, impressions and Realism takes a number of forms depending perceptions. By taking these objects at face on how ‘the real’ is understood. All value, empiricists are able to identify a lim- approaches to knowledge that endorse real- ited number of them, observe and measure ism accept that a world exists that is in some their frequency and (subject to the screening respect independent of the knowing subject. out of extraneous variables) begin to identify This real external world may be a purely and predict clear relationships between them empirical reality so that what has not yet (for example, between independent variables been observed will at some point appear to and dependent variables). This combination us. Alternatively, it may be a reality that is of assumptions is often referred to as a closed concealed from us in some way beyond system. In social research it is important to appearances and independent of our knowl- focus three approaches to knowledge, each edge of the world. These different ontological with their own distinctive characterization of assumptions about what really exists have causality (empiricism, idealism and realism). produced corresponding epistemologies – the Empiricists accept that observed empirical first associated with empiricism and idealism regularities (constant conjunctions of events and the second with critical realism – that are whereby if x happens then y will follow) are nec- central to social research. Whereas for essary and sufficient for establishing a causal empiricism real objects can only ever be law. For idealists and realists, this fails to accessed through sense data and for idealists sustain the intelligibility of perception: that empirical reality can be understood through a sensations, impressions and perceptions are R-Jupp-3373.qxd 3/8/2006 5:49 PM Page 256

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

organized and make sense only through their realists often conflate these two different synthesis with prior mental constructs. This epistemologies to ‘the empirical real’. For real- rejects the view that characterizes some forms ists, such as Roy Bhaskar, one should start by of empiricism (especially positivist versions) distinguishing between the transitive dimen- that human beings are blank slates or empty sion (rival theories and approaches about the vessels passively accepting sensations. For world that imagine what the real is like, tran- idealists, the phenomenal world of appear- sitive objects) and the intransitive dimension ances is the product of both our experiences (what the world is like regardless of our cur- of the object of analysis (which as ‘things-in- rent state of knowledge). This is tied to the themselves’ or noumena are beyond cognitive view that objects are intrinsically complex faculties and thus unknowable) and the a with undiscovered capacities or susceptibili- priori conditions within which the mind ties that are only identified in certain condi- provides shape to this experiential content. For tions. Indeed, for critical realists experiments this reason, an empirical regularity is still nec- are only intelligible if the objects already have essary for establishing a causal law but it is not an unknown internal structure (such as a sufficient for two reasons: first, it needs to take capacity to produce certain effects prior to the account of the range of possible empirical fac- discovery and consequent knowledge of that tors that may produce the effect in question; real capacity) (Bhaskar, 1978). Their account of second, it needs to make sense in terms of the causality reflects these assumptions. The struc- existing body of knowledge and be relevant to tures of these objects can produce certain the values evident in the context. effects or events but do so when the conditions Given that empirical reality is seen by ideal- are right for that structure to generate a causal ists as too complex to fully reconstruct and that mechanism. These conditions include the there is no unmediated access to the ‘real’, presence of other structures, mechanisms and then this has generated a distinctive tradition events. If the conditions are not appropriate in social research that recognizes the limita- then we may not identify an effect in terms if x tions of its own theories, concepts and models then y, but for a realist this does not mean that of that world – that they should be seen as the structure does not exist. In terms of estab- necessary simplifications. For this reason, it is lishing a causal law, an empirical regularity is often labelled the ‘interpretive’ approach. For neither necessary nor sufficient (Smith, 1998). example, Max Weber developed the technique In social research, critical realists argue we of ideal types (such as habitual, emotional and have to take into account the ontological or rational action) as simplistic exaggerations in real differences between natural and social order to organize empirical evidence, while objects. Social structures do not exist indepen- phenomenologists such as Alfred Schütz advo- dently of the activities they govern, they do cated the use of typifications or types to estab- not exist independently of the agent’s concep- lish closer relations between the second order tions of what they are doing, they are time- concepts of the researcher and the first order space specific and only relatively enduring constructs of everyday life. Not all idealists are and they are reproduced, modified or trans- so humble; advocates of rational choice theory formed through the activities of agents. In and game theory tend to be so convinced that short, we should start with the assumption of rational and instrumental logic is the basis for an open rather than a closed system. In turn all human action in all situations they risk the activities of agents are shaped by the pre- imposing their own assumptions on their existing social structures they are thrown into objects of analysis. and have capacities to do things and suscepti- Critical realists, building on idealist bilities to other forces depending on their assumptions about the intelligibility of experi- position in the structures (Bhaskar, 1979; ence, challenge the assumption that ‘the real’ Sayer, 1992, 2000). Social structures are both is purely empirical (an assumption character- the medium and the outcome of the activities istic of empiricism and idealism). In so doing, of agents. These ‘real’ structures include

256 R-Jupp-3373.qxd 3/8/2006 5:49 PM Page 257

REALISM

relationships such as that between tenants and Karl Popper’s ‘objective third world’). Indeed, landlords, capitalists and workers or tutors and aspects of ‘the real’ are present in experience students (in each case they are mutually defin- and in the mental constructs we use to make ing). If over time and across different contexts sense of this. Recent attempts to portray ideal- the operation of these kinds of structures ist theorists as diverse as Max Weber, Alfred becomes deeply embedded then they can be Schütz and Niklas Luhmann as embryonic or seen as institutionalized, making change more proto realists merely highlight the problem. difficult compared to the more flexible struc- More symptomatic of the problems of critical tured relations, for example, between speakers realism is their tendency to draw upon their and listeners in a conversation. own cultural values and assumptions in the accounts of the real. Both idealist and discursive approaches to knowledge draw attention Evaluation (in their different ways) to the way that social Challenges to critical realism fall into three context is relevant to the kinds of knowledge main camps. The empiricist critique argues produced and how organized knowledge regu- that critical realism is a form of objectionable lates what can and cannot be thought and metaphysics constructing imaginary and ideo- spoken. As a result of ignoring the complexi- logically inspired visions of the social world as ties of representation and values, critical real- oppressive and exploitative. For critical realism has often reproduced the language of ists, such responses are conservative attempts pathologization that stigmatizes certain social to prop-up an unjust society. For realists, groups in much the same way as crude ver- explaining and understanding social structures sions of positivism (Smith, 2000). is the first step to identifying how to transform it so that unwanted forms of social determina- Mark J. Smith tion can be replaced by wanted ones. For this reason, critical realism is an emancipatory Associated Concepts: causality, empiricism, approach seeking social change though there epistemology, ontology, relativism are disagreements between critical realists as to whether the problem that needs changing is the social class system, patriarchy or other Key Readings forms of structured social division such as those based on racism and other forms of cul- Bhaskar, R. (1978) A Realist Theory of Science, tural difference. 2nd edn. Hemel Hempstead: Harvester The idealist critique highlights the difficul- Wheatsheaf. ties caused by the battlelines between critical Bhaskar, R. (1979) The Possibility of realism and empirical realism. Critical realists Naturalism: A Philosophical Critique of the argue that it is possible to accommodate the Contemporary Human Sciences. Hemel insights of empiricism and idealism by chang- Hempstead: Harvester Wheatsheaf. Blaikie, N. (1993) Approaches to Social Enquiry. ing one’s assumptions about what is real. At its Cambridge: Polity Press. most strident, realism assumes that empiri- Sayer, A. (1992) Method in Social Science: cism and idealism have the same flaw of a ‘flat A Realist Approach, 2nd edn. London: ontology’ (being unable to distinguish between Routledge. ‘sensations, impressions and perceptions’ and Sayer, A. (2000) Realism and Social Science. the ‘events and states of affairs’ that serve as London: Sage. our objects) and refusing to accept the exis- Smith, M. J. (1998) Social Science in Question: tence of a deep reality beyond observation. As Towards a Postdisciplinary Framework. indicated above, this is not an accurate London: Sage. description of idealist approaches nor of some Smith, M. J. (2000) Culture: Reinventing the epistemologies associated with empiricist Social Sciences. Milton Keynes: Open research such as falsificationism (for example, University Press.

257 R-Jupp-3373.qxd 3/8/2006 5:49 PM Page 258

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

detail about the research process – and REFLEXIVITY decisions taken within it – to allow readers to make some judgement by themselves. Definition On occasions reflexivity involves not just The process of monitoring and reflecting on all monitoring and assessing validity (to what aspects of a research project from the formula- extent are conclusions credible and plausible?), tion of research ideas through to the publication but also questions of ethics (has anyone been of findings and, where this occurs, their utiliza- harmed by the research?) and questions of tion. Sometimes the product of such monitoring politics (whose side am I on, if any?). Validity, and reflection is a reflexive account which is ethics and politics can impact one on another. published as part of the research report. For example, a decision not to publish interview material collected from corrupt policemen, for Distinctive Features fear of exposing them, will not result in a full and valid account of the realities of police sub- Although important in all areas of social cultures. Therefore, reflexivity often involves research, reflexivity has an especial role in the researcher in reaching a personal stand- ethnography, in which the researcher is close point in relation to trade-offs which often have to the subjects and to the data. What is more, to be made between validity, politics and ethics. the whole process of formulating ideas, collecting observations, analysing and reaching Evaluation conclusions is part of the role of the investigator vis-à-vis those who are the objects of In assessing the potential threats to validity, inquiry. It is for this reason that ethnographers reflexivity is concerned with the social pro- see reflexivity as part of research in its own duction of knowledge. It involves reflecting on right and not as a collection of afterthoughts the various social roles, interactions and on how a project has been accomplished. processes which resulted in the kinds of obser- At one level, a reflective account will be vations and conclusions which emerged. It is descriptive in terms of providing an account of, possible to consider reflexivity in a sense for example, how interviews were carried out, wider than that of monitoring and assessing what methods of recording data were used and the validity of a particular research project. so on. However, at another – much more This involves viewing critical reflection as a important – level a reflective account should be form of research in its own right, and as part evaluative in terms of providing some assess- of the school of critical thinking and theorizing ment of the likely validity of the conclusions in the social sciences in general. A critical that have been reached. This might, for exam- research agenda could include a consideration ple, involve a consideration of whether respon- of why particular (say, punitive) law and order dents selected for interview are typical of the discourses take the form that they do and group about which conclusions are to be made; come to be accepted as ‘truth’ when they do. whether there is a possibility that responses to What is more, it could consider the role of questions were the outcome of exaggeration or research in the production and dissemination even downright falsification; and whether the of such ‘truths’. This would involve reflecting method of recording data resulted in only a on the research enterprise as a whole and ask- partial or even distorted, account of reality. ing what gets studied, when, by whom, what These are what are known as ‘threats to valid- gets published, and with what effect? ity’. A researcher may not be able to anticipate and rule these out. However, he or she should be aware of them and provide some assess- Victor Jupp ment of their potential effects on the validity of conclusions. At a minimum, the expectation is Associated Concepts: critical research, disas- that the researcher will provide sufficient trous research, ethics, impression management,

258 R-Jupp-3373.qxd 3/8/2006 5:49 PM Page 259

REGRESSION ANALYSIS

politics and social research, research bargain, Y = a + bX validity Here Y is the value of the dependent variable, Key Readings X of the independent variable, a the value Y has when X is zero, and b the amount Y Findlay, L. and Gough, B. (eds) (2003) changes when X changes by one unit. We can Reflexivity: A Practical Guide for Researchers extend this simple bivariate relationship to in Health and Social Sciences. Oxford: become multivariate by adding in more inde- Blackwell. pendent variables thus: Mason, J. (2002) Qualitative Researching. London: Sage. + 1 1 + 2 2 3 3 + n n Steir, F. (ed.) (1991) Research and Reflexivity. Y = a b X b X + b X … b X London: Sage. There is usually a difference between the actual value of Y and the value predicted by the regression equation – the residual. In simple linear regression – linear because changes REGRESSION ANALYSIS in Y are proportionate to changes in X – the line fitted minimizes the sum of the square of Definition these residuals – squares because squaring eliminates negative numbers and some resid- A body of statistical techniques in which the uals will be positive and some negative. The form of the relationship between a dependent simplest way to see the relationship between variable and one or more independent vari- variables in the two or three variable case is ables is established so that knowledge of the to generate a scatter plot in two or three values of the independent variables enables dimensions. Standard statistical packages, prediction of the value of the dependent vari- such as SPSS, fit a line to the plot for the two able or likelihood of the occurrence of an variable case and the degree of spread of the event if the dependent variable is categorical. real values around the line can be inspected. The correlation coefficient is a measure of the Distinctive Features degree to which the real values of Y correspond to those predicted by the regression Regression analysis is a method by which equation. quantitative social science seeks to establish Logistic regression is a robust technique in how things are caused. The objectives are which the dependent variable can be categor- both scientific description and prediction. ical or ordinal. The independent variables If we know the form of the relationship can be at any level of measurement and at between things we have measured and know different levels of measurement. Logistic to be causal to something else, then we can regression can handle nonlinear relationships predict the value of the caused thing. For and gives us an indication of the likelihood of example, using logistic regression, knowledge the dependent variable having a particular of the blood levels of various hormones and value. of the results of an X-ray in a patient with prostate cancer can be used to predict if the Evaluation cancer has spread to the patient’s lymph nodes. Traditional regression analysis requires that a The relationship between independent and series of demanding conditions be met by the dependent variables is expressed through a data, especially when the data comes from a regression equation. In the simplest case of sample. However, the biggest drawback is the linear relationship between two continu- that traditional regression procedures assume ous variables this can be written as: that the relationships among variables are

259 R-Jupp-3373.qxd 3/8/2006 5:49 PM Page 260

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

linear. Linear relationships are not always, description can be true, no approach can be indeed perhaps not often, present in the privileged and that no assessment of the social world. The detection of interaction, of value of human conduct is valid. Another the relationships among variables depending case merely claims that all descriptions are on the specific values of those variables, and equally true, all approaches have the same changing as those values change, is an indi- standing and all considerations of value are cation of nonlinearity. Logistic regression equally valid (Harré and Krausz, 1996). makes fewer assumptions and can cope with nonlinear relationships because it is essen- Distinctive Features tially a predictive tool rather than an analytical method of establishing causal models. These two interpretations, which can be described as the strident case and the reasonable case respectively, assume that what is David Byrne said and acted upon in a particular location should be understood in its own terms and Associated Concepts: causal model, causal- according to the standards of that location. ity, correlation, discriminant function analy- On a note of caution, the strident case for sis, econometrics, forecasting, general linear relativism is more often deployed by those model, log-linear analysis, multivariate analy- who argue against it, such as critical realists. sis, prediction Rather like the label totalitarian, there are few who claim to be strident relativists. Most Key Readings who favour relativism simply seek to prob- lematize, deconstruct or place in question the Byrne, D. S. (2002) Interpreting Quantitative certainties and essences that generally serve Data. London: Sage. as starting points for analysis. In this way, the Menard, S. (2001) Applied Logistic Regression canonized points of reference such as ratio- Analysis. London: Sage. nality, instrumental or self-interested actions, Schroeder, L. D., Sjoquist, D. L. and Stephan, cognitive processes and various conceptions P. E. (1986) Understanding Regression of agency are seen as topics for research Analysis. London: Sage. rather than the unquestioned conceptual resources of social scientists. Sometimes relativism is seen as associated with social constructionism, especially the RELATIVISM claim that both ‘the natural’ and ‘the social’ (or for that matter ‘appearance’ and ‘reality’) Definition are linguistic concepts or discursive constructs. A strident case would claim that all There is considerable variety in the uses of categories and terms are discursively consti- the term relativism, but all endorse the view tuted and that ‘the real’ or ‘the deep’ do not that no single absolute principle or concept is exist independently of the way we construct adequate to explain and understand events them; in short, that they are just convenient and states of affairs in all times and places, fictions for those keen to believe in their exis- and that no single criterion holds for attribut- tence. The reasonable case for relativism ing value. By challenging the idea that uni- would simply state that since we can only versal statements can be established or that ever consider the existence of ‘the real’ universal entities exist, relativism under- through a specific discourse and that the way mines truth, objectivity and the necessity of human beings have viewed reality has under- foundations. Nevertheless, there are two gone significant changes then we should not quite different interpretations of what this place too much faith in the currently fashion- means. One case for relativism asserts that no able forms of imagining reality. As Foucault

260 R-Jupp-3373.qxd 3/8/2006 5:49 PM Page 261

RELATIVISM

argued, how can we have a history of truth if theoretical and empirical tests, weakening the truth has its own history (see Smith, 1998). scientific method. Feyerabend’s contribution Discourses are texts or writing that can be is his description of the accidental, messy and seen as sets of signifying practices and sys- discontinuous character of the history of tems of representation in order to constitute science and the problems of methodological a ‘reality’ to provide the conditions within monism (the idea that one scientific method which subjects experience the world (bearing fits all cases). In its place, he proposes the in mind that the conceptions of knowing sub- principle of proliferation (the ‘anything goes’ jects and objects of analysis are themselves approach), whereby a range of different theo- discursive categories rather than independent ries and research programmes (with different of the accepted procedures for knowing assumptions about the world and appropriate about the world). procedures for studying it) co-exist in order Relativism is often linked to Thomas to generate new knowledge. Stung by criti- Kuhn’s account of the scientific revolutions cisms that this led to an inability to make that take place as one paradigm replaces judgements on events such as the holocaust another. In this historical narrative of the (Gellner, 1979), he replied that his work, far sudden shifts between long periods of puzzle- from being relativist, actually contained a solving normal science (in which Kuhn was in normative project for methodological plural- fact much more interested) and the point at ism. He saw this as a means of defending the which a radically new account of reality came freedom of the individual against the homo- into existence has often been described as genization of opinion that characterized akin to a ‘Gestalt shift’ or religious conversion modern science. Sociologists of science have experience. Each paradigm has its own inter- similarly demonstrated how epistemologies nalized standards as to what knowledge is such as falsificationism are idealized recon- legitimate. According to Kuhn, paradigms are structions that have very little to do with incommensurable, that is, it is impossible to logics-in-use – what scientists actually do. use the criteria of one paradigm to judge the Relativism has been taken to be relevant in adequacy or truthfulness of another. three distinctive areas concerned with kinds Paradigms are thus mutually exclusive and of human judgement – epistemology (as with inhabitants of different paradigms can be said Feyerabend), ethics and aesthetics. The rea- to be living in ‘different worlds’ (Kuhn, 1970). sonable case for relativism draws our atten- Cultural relativism raises similar questions. tion to tacit assumptions social researchers If we use the standards established in one’s hold regarding what is true, good and beauti- own culture to make a judgement as to the ful as they engage in research. It also alerts us adequacy of the values and practices of to how easy it is to assert that the greatest another, then the likely outcome is to infer the benefit for the greatest number can be superiority of one’s own culture and the inferi- achieved by using scientific procedures for ority of other cultures. By privileging the taken- establishing the maximum benefits at mini- for-granted assumptions of what is considered mum costs. In social research where utili- to be true, good and beautiful in a specific tarian assumptions are embedded there is a society, so that these become the measure of constant play between the good and the value, one falls into the trap of ethnocentrism. true. Reasonable relativism can be seen in Another kind of relativism was developed phenomenological social science which rec- by the philosopher Paul Feyerabend, though ognizes how knowledge is situated in the it must be added that at times in his life he plausibility structures that exist in specific sought to establish a single criterion for com- social contexts. This means that the paring incompatible theories and research researcher should seek to find ways of draw- programmes. In addition, Feyerabend wor- ing upon the first order constructs of every- ried that relativism encouraged too great a day life in constructing their own second leniency in assessing the significance of order constructs and reap the benefit of being

261 R-Jupp-3373.qxd 3/8/2006 5:49 PM Page 262

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

intelligible to the people being studied – the Kuhn, T. ([1962] 1970) The Structure of postulate of adequacy (Schütz, 1953). Scientific Revolutions, 2nd edn. Chicago: University of Chicago Press. Evaluation Schütz, A. (1953) ‘Common-sense and scientific interpretation of human action’, Critics of relativism argue that without Philosophy and Phenomenological Research, assumptions that provide for absolute or more XIV (1): 1–38. specifically universal judgement, this under- Smith, M. J. (1998) Social Science in Question: mines the possibility of developing a common Towards a Postdisciplinary Framework. language for describing the world, a common London: Sage. theoretical and conceptual systems for analysing and evaluating social relations and a common basis for moral and aesthetic judgements. For defenders of the intellectual heritage of Western civilization, relativism RELIABILITY challenges the values and assumptions that have served well for generations. This is often Definition associated with the claim that relativism prevents us from possessing a moral compass, The extent to which a measuring instrument, leaving us in a position where it is impossible for example a test to measure intelligence, gives to make judgements as to the right course of consistent results. action or the good outcome on crucial matters such as genetic modification or events such as Distinctive Features genocide. For critics of existing social relations, such as Marxists, relativism undermines the There are three broad ways of assessing reli- belief that alternative forms of rationality and ability. First, test–retest reliability involves morality can be forged to replace the exploita- administering the same test to the same set of tive and oppressive social structures; that rela- respondents but on different occasions. tivism destroys the possibility of social change Correlation coefficients are calculated to and removes all hope from politics. All such compare the different data sets. The higher critics level their charges at the strident case the coefficient the more reliable the measur- for relativism and not the reasonable case. The ing instrument. Values of 0.7 are considered difficulty of contrasting universal judgement good. The problem with test–retest reliability with judgements that are relative to a time and is that individuals may become familiar with place is that even the statement that ‘all forms items and simply answer on the basis of their of knowledge are situated’ is still a kind of uni- memory of what they answered on the previ- versal statement. ous occasion. A second type is alternate-form reliability, which involves using differently worded Mark J. Smith items to measure the same variable. Items are either administered to the same sample at dif- Associated Concepts: causality, empiricism, ferent points in time or to two sub-samples at epistemology, ontology, realism. the same point in time (known as the split halves method). Correlation coefficients are calculated to compare the two data sets and Key Readings measure reliability. Where the split halves Gellner, E. (1979) Spectacles and Predica- method is used it is important to ensure that ments. Cambridge: Cambridge University the two sub-samples are as similar as possi- Press. ble. Whichever method is used the two tests Harré, R. and Krausz, M. (1996) Varieties of should measure the same concept despite Relativism. Oxford: Blackwell. wording changes.

262 R-Jupp-3373.qxd 3/8/2006 5:49 PM Page 263

REPERTORY GRID TECHNIQUE

Third, internal consistency reliability Kelly. His landmark publication was The involves using groups of items (as opposed to Psychology of Personal Constructs, which a single item) to measure different aspects of appeared in 1955. His work was part of a the same concept. Internal consistency is broad theoretical and research agenda which measured by Cronbach’s alpha, which is a was, and still is, concerned with the ways in measure of how well the different items com- which people make sense of their worlds. In plement each other in measuring the same theoretical terms it can be described as con- concept and form a single scale. structionist and in methodological terms it tends towards qualitative research strategies. Evaluation As such, Kelly was against psychological behaviourism, determinism and explanations Reliability assessments are essential, espe- in terms of causality. Instead, he emphasized cially in scaling, but they are not sufficient in the ways in which individuals are creatively themselves. It is also vital to assess validity of engaged in giving meaning to events, situa- measurement, that is, assessing the extent to tions and other people. They do this via a per- which a scale measures what it purports to. A sonal construct system which is not fixed; scale that is reliable but measures the wrong rather it constantly undergoes ‘testing’ and thing is not particularly helpful. modification as new events, situations and people are encountered. Kelly’s theoretical Victor Jupp ideas included ‘constructive alternativism’, that is, different individuals may create dif- Associated Concepts: composite measure- ferent and alternative construct systems in ment, correlation, indicator, measurement, relation to the same social situations. multiple indicators, scaling, social indicators, The repertory grid technique was devel- validity of measurement oped to elicit an individual’s personal construct system at any given point in time. The Key Readings grid is assumed to comprise two components: first, elements to which individuals relate, for Smith, H. W. (1975) Strategies of Social example situations, objects and other individ- Research: The Methodological Imagination. uals; second, constructs, which are ways of Englewood Cliffs, NJ: Prentice Hall. describing such elements. The personal con- Torgeson, W. S. (1958) Theory and Methods of Scaling. New York: Wiley. struct system comprises the array of constructs in relation to elements. Some constructs are peripheral and relatively unimportant whereas others are core, being central to the individual’s value system. Each REPERTORY GRID TECHNIQUE construct is assumed to have opposite poles because Kelly believed that we all make sense Definition of the world by looking for likenesses and dif- A technique that can be used to elicit the per- ferences. So, for example, we might experi- sonal constructs that, it is argued, individuals ence and describe situations as more or less use to interpret the social world and to form ‘supportive’ or ‘threatening’. judgements about appropriate forms of The specific application of repertory grids action. For this reason it is known as an typically involves a process of ‘triad elicitation’ elicitation technique. whereby an individual is asked to consider elements in groups of three drawn from a larger pool. He or she is asked to describe what fea- Distinctive Features tures (i.e. constructs) make two of these similar Repertory grids were developed in the 1950s to each other and different from the third. For by the American social psychologist George example, an individual may be presented with

263 R-Jupp-3373.qxd 3/8/2006 5:49 PM Page 264

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

pen portraits of three occupations (for example, In short, repertory grid techniques are a joiner, a plumber and a company director) in fruitful and versatile tools of social research order to elicit constructs relating to social class. which can embrace the micro and the macro, Once the first bipolar construct (say, income) the idiographic and the nomothetic, the qual- has been identified the individual is asked to itative and the quantitative. rate the three occupations on that construct. Those who are critical of repertory grid Then the process of eliciting a second construct and other techniques would argue that con- proceeds by presenting another triad of occu- structs are not context-free and also that, pations. The process continues until the although it is claimed that the respondents’ respondent cannot offer any further ways of own constructs are accessed, the research distinguishing between elements. This can be situation is artificial and they might be described as the point of theoretical saturation. putting forward ‘official’ rather than their Sometimes they are asked to offer new ways of own constructs. distinguishing elements or offer further constructs, without going through triad elicitation, in order to ensure that theoretical saturation Victor Jupp has indeed been reached. The outcome of the process is a completed Associated Concepts: constructionism, grid that describes the individual’s personal exploratory research, idiographic, nomo- construct system and comprises a set of inter- thetic, qualitative research, scaling, summated related bipolar constructs on which each ele- rating scale ment is assigned a rating. Computer software packages have been developed to facilitate Key Readings detailed analyses of patterns between con- Fransella, F. (1995) George Kelly. London: structs and elements. Sage. Evaluation Fransella, F. and Bannister, D. (1977) A Manual for Repertory Grid Technique. The repertory grid is not an ‘off the peg’ tool, London: Academic Press. as is usually the case with attitude scales or Kelly, G. A. ([1955] 1991) The Psychology personality tests. It does have a set of proce- of Personal Constructs, 2 vols. London: dures, for example triad elicitation, but it is Routledge. also a flexible methodology which fits with Kelly’s theoretical viewpoint that personal constructs are actively developed and revised. The repertory grid was originally devel- RESEARCH BARGAIN oped for use with individual clients in a clinical and therapeutic setting and in this sense was an idiographic rather than nomothetic Definition tool. However, Kelly was always of the belief The agreement made between a researcher that there could be a sharing of meanings and and a research participant such as a gate- this has led to the use of repertory grids with keeper or respondent regarding what the par- a small number of individuals as an ticipant can expect from the researcher in exploratory tool and a preliminary to the return for their cooperation with the research. development of more formal scales that can be administered to much larger samples of Distinctive Features people. For example, grids can be a forerun- ner to the development of semantic differen- Within social research, research bargains are tial scales for examining how criminal justice made throughout the research process personnel differentiate between ethnic between researchers and participants to help groups and with what consequences. ensure cooperation. They include bargains

264 R-Jupp-3373.qxd 3/8/2006 5:49 PM Page 265

RESEARCH DESIGN

reached with gatekeepers at the start of the researcher to know in advance what issues research in return for enabling access, as well will arise during the research, it is important as those made with respondents to secure to establish the extent to which the proposed their cooperation when collecting data, such research can be amended to agree a research as contributing to a focus group, respond- bargain. If the changes required interfere too ing to interview questions or completing a much with that which is deemed essential for questionnaire. the research, then the researcher may con- Although the bargain made by a sider it wrong to proceed (Horwood and researcher with a gatekeeper is likely to out- Moon, 2003). line the entire research process, it provides the gatekeeper with an opportunity to ensure that, in return for granting access, she or he Mark N. K. Saunders receives something useful from the research. This may take a variety of forms, including an Associated Concepts: access, confidentiality, opportunity to influence the research so that ethics, gatekeeper, informed consent, organi- it incorporates issues that she or he believes zational research are pertinent, preferential access to the research findings, and in extreme cases, the Key Readings right of veto. Whilst ethical considerations Bryman, A. (2001) Social Research Methods. and, in particular, issues relating to Oxford: Oxford University Press. anonymity and confidentiality normally form Horwood, J. and Moon, G. (2003) ‘Accessing part of the research bargain with the gate- the research setting: the politics of keeper, they tend to take on greater promi- research and the limits to enquiry’, Area, nence in the bargains made with individual 35 (1): 106–9. respondents. Such respondents usually agree to participate through a belief that the research will not cause them harm and is of value. Although often participants do not expect payment, this does not preclude RESEARCH DESIGN research bargains containing some form of incentive for the respondent to take part in Definition the research. These include donations to charity, payment or some form of gift. A design or strategy that justifies the logic, structure and the principles of the research Evaluation methodology and methods and how these relate to the research questions, hypothesis or Increasingly within social research, research proposition. bargains are being formalized as part of research governance through the use of Distinctive Features agreed research proposals and consent forms. This means that research bargains are less There are several reasons for devising a likely to be either implicit or vague than was research design. First, an effective research the case a few years ago. design will demonstrate that the research will Research bargains require careful thought produce valid and credible conclusions that as, although research bargains can help in flow logically from the evidence generated. securing research participants’ cooperation, Second, it sets out the research strategy they may place restrictions on the researcher. for the benefit of the audience, readership, For this reason some researchers argue that, funders, gatekeepers and those researched. other than for ethical considerations, Third, the process of devising it not only research bargains need to be treated with ensures that the research will be of value in caution. Although it is difficult for a terms of intellectual credibility, external

265 R-Jupp-3373.qxd 3/8/2006 5:49 PM Page 266

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

accountability, coherence and rigour, but that the best method for any research will usually it becomes a useful plan or schedule for the involve compromise or a delicate balance or researcher. blend of intellectual and practical concerns Specific elements of an effective research where, for example, precision may be tem- design cannot be prescribed and there is no pered by factors such as feasibility due to cost standard format, however some key compo- or other restrictions. nents are likely to include the following. Although many of the hallmarks of an First, a clear, researchable set of questions or effective research design outlined above hypothesis. This aspect of the research design apply to both qualitative and quantitative is perhaps most fundamental and the ques- research projects, there has been resistance, tion of what is to be done will require justifi- in particular from the qualitative research cation in terms of why it is being done and a tradition, to the need for a precise research carefully planned strategy of how it will be design. Some feel that it is not possible or carried out. Second, the research design will desirable to have a rigid research design as show how the research questions will be this militates against the fundamental ques- aligned to the relevant data sources and tions or problems they are researching and research methods, the selection of which against the need for continuous reformula- should be clearly justified taking account of tion of research questions and also for the issues surrounding validity and representa- process of exploration. tive methods of sampling, reliability and generalizability, comparative issues, ethical considerations (consent, anonymity, confi- Pamela Davies dentiality) and practicalities (such as resources, access and skills). Third, in addi- Associated Concepts: exploratory research, tion to selecting and justifying the data gatekeeper, intellectual craftsmanship, litera- collection methods, specifying the sampling ture review, messy research, methodology, criteria and timing of the research, the reliability, validity research design should identify how data processing will operate and how the analysis Key Readings will be conducted. Jupp, V., Davies, P. and Francis, P. (eds) (2000) Doing Criminological Research. Evaluation London: Sage. chs 1 and 2. To design an effective research project Mason, J. (1997) Qualitative Researching. London: Sage. requires skill. In principle it is straightfor- Sapsford, R. and Jupp, V. (eds) (1996) Data ward to address the questions of what is being Collection and Analysis. London: Sage. researched and to delineate why and how the research will be effected. In practice, these are often much bigger questions than they might at first appear. For example, fully addressing the fundamental question of what RESEARCH PROBLEMS the topic is, might include a consideration of ontological and epistemological positions for Definition some researchers. Whilst how research is executed is often a complex process involving A diverse range of pitfalls encountered dur- many decision-making stages where choices ing the research process. These can include are made under different and often difficult getting started, framing general and specific conditions and within different and perhaps research questions, and discovering that changing constraints. In reality the research one’s data do not address the research ques- design will often require flexibility. Selecting tions that have been set.

266 R-Jupp-3373.qxd 3/8/2006 5:49 PM Page 267

RESEARCH PROBLEMS

Distinctive Features income and the incidence of youth suicide? Or, what is the relationship between parental Perhaps one of the most difficult tasks that break-up and the incidence of youth suicide? students face in large research projects, par- Because the subject matter has been focused ticularly dissertations, is that of getting down to such a degree, data collection ques- started. The student may have already taken tions will become easier to generate. This is the decision to look at a general research important because the data collection ques- area, but then struggle to narrow this down to tions are the most specific questions the some manageable size. Hence, one of the researcher addresses. At this stage, one could main failings of undergraduate dissertations conduct a quantitative study into divorce and is that the student has been unable to specify suicide rates, or conduct interviews with precisely the main focus of his or her partic- bereaved parents, although ethical and sensi- ular dissertation. Instead of analysing the tive interviewing techniques would need to be internal dynamics of the Communist Party of considered in the latter case. East Germany in 1989, in effect trying to Punch’s framework is elegant because it evaluate how this precipitated the fall of the encourages the student to systematically Berlin Wall in that year, a poor dissertation by hone what may be a very abstract idea down contrast may study nothing more specific into a project that would be manageable. than the demise of euro-Communism in Nevertheless, research problems do not end 1989. The former project may be manage- there. As Jupp (2000) reminds us, the conclu- able; the latter would, at best, offer a superfi- sions of research will be credible and plausible cial overview of the subject matter. only to the extent to which the questions The process of getting started need not and problems they address are clearly formu- be too daunting, however. Provided that the lated and expressed and followed through in a student tries to be as systematic as possible, consistent manner during the inquiry. In effect, honing down a research proposal can be this refers to data collected from empirical greatly simplified. As Kenneth Punch (2001) research not matching the original research argues, research problems may be ameliorated questions. Another of the common failings of by following an inductive–deductive hierarchy. students writing dissertations is that although Following Punch’s model, we could begin at they may have collected a wide range of empir- the broadest possible degree (the research area) ical material, they have not been able to link it by choosing a very broad research theme. This back to the original objectives of their thesis. may be the fall of euro-Communism in 1989, On such occasions, it may be necessary to or youth suicide. At this level, any project revise completely, or in part, the original would be too broad and general. We would research design. Specifically, if empirical data then need to narrow down our focus, effec- indicated that rates of parental divorce had no tively choosing certain issues to study, whilst influence on youngsters taking their own life, ignoring others. This can be achieved by and that suicide among the young was caused constructing a research topic. Using the latter by drug abuse, then the original general and example, this could be either suicide rates specific research questions would need to among young groups (a quantitative approach) reflect this in some way. Tempting as it may or youth culture and the meaning of suicide be, a researcher should avoid writing a disser- (which would entail a qualitative strategy). tation that concentrates on one particular area, Focusing down again, the student would then only to completely disregard this area in his or frame one or several general research ques- her conclusion. Factors that suddenly appear tions, for example, what is the relationship in the conclusion as being the most important between family background and the incidence issue in youth suicide, for example, and that of youth suicide? Finally, specific research have not been discussed elsewhere in the questions are constructed, which may include: work, tend to weaken the overall argument of what is the relationship between family the author. Finally, the researcher should be

267 R-Jupp-3373.qxd 3/8/2006 5:49 PM Page 268

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

aware that not all interviews may be of equal to measure and understand change and to importance. Some interviewees give responses include a time dimension to the data that can that are irrelevant to the research on hand. At be used to identify causal factors contributing the time of writing up, the author should to any observed change. The capacity of a ret- always filter out data that are not relevant to rospective study to adequately detect change the research questions. and ascertain causes depends on how well the investigator can reconstruct the past from Evaluation the vantage point of the present.

Although the student or researcher may be Distinctive Features more interested in beginning their research as soon as possible, they should not underesti- The main purpose for collecting retrospective mate the benefits of good research design. data is that such data provide a means of Being careful to prioritize and narrow down measuring change for either descriptive or one’s research can ultimately save a great deal explanatory purposes. Retrospective studies of time and effort, and thereby eliminate many rely on recalling information about the past research problems. Yet even when one follows but vary in the extent to which they rely on a methodical research design, it is often the such recall. An oral history, for example, is case that research problems will arise. The totally reliant on people recalling the past. author should therefore be aware that he or Other retrospective studies involve collecting she may need to alter their research questions, information about the past and comparing and may have to prune the data that he or she that with contemporary information collected may have collected. In sum, research prob- from the same cases. An example of such a lems can be tackled by continually revising study might be one in which parents are and updating one’s research design. asked to describe their current marital satisfaction, say, five years after having children and then to recall their level of marital satis- Craig McLean faction just before they had children. More often than not, retrospective information is Associated Concepts: disastrous research, collected as part of a study that collects exploratory research, hypothesis, research contemporary data or even as part of the first design phase of a longer-term prospective study. Many studies that are primarily cross- Key Readings sectional (that is, collect data about the present) contain some retrospective questions to Jupp, V. (2000) ‘Formulating research prob- detect individual change. For example, lems’, in V. Jupp, P. Davies and P. Francis respondents are often asked about parental (eds), Doing Criminological Research. occupation when they were 14 years old, or London: Sage. pp. 13–28. to recall how often they attended church Punch, K. F. (2001) Developing Effective when they were 14. These retrospective Research Proposals. London: Sage. measures are then compared with contemporary circumstances to construct measures of the individual’s social mobility or religious change. RETROSPECTIVE STUDY Other studies collect information that enable the study of change at the aggregate Definition level. For example, some surveys ask about relationship histories that can provide insight A study that involves collecting data about into aggregate change over time. For example, past events. This design is mainly employed respondents may be asked to recall when

268 R-Jupp-3373.qxd 3/8/2006 5:49 PM Page 269

RETROSPECTIVE STUDY

they first married and then if they lived with reliably collect information about subjective their partner before they married. This states and even about less significant events enables the investigator to determine and more objective information. To use retro- whether the person’s first marriage was a spective methods to learn about feelings at direct marriage (no premarital cohabitation) some earlier time or to reliably recall past or an indirect marriage (included premarital opinions is fraught with dangers. The most cohabitation with the partner). By asking obvious problem is that of faulty memory. these questions of a range of age groups it Faulty memory has two elements. People may is possible to identify people who entered simply misremember. For example, to ask a marriages directly and indirectly in, say, the person how much they earned five years ago early 1970s, the late 1990s or some time in risks people simply inaccurately recalling between. Using retrospective data it is possi- their level of income. The other element is ble by collecting data at just one time point to that of ‘telescoping’ and ‘reverse telescoping’. plot changes in marriage patterns. What people recall as occurring five years ago Retrospective studies can be distinguished may have been just two years ago or may from cross-sectional and prospective designs. have been seven years ago. Prospective studies collect information about A further difficulty with recalling the past, the present time and then track cases forward especially when recalling more subjective in time and collect new, contemporary infor- states, is the tendency for people to recon- mation at each time point to construct struct the past in the light of present circum- measures of change. In the pure form, a stances. For example, retrospective studies of cross-sectional study collects information voting behaviour demonstrate that the way only about the present and cannot construct people recall their vote overestimates the measures of change at all. extent to which people voted for the winning party. A related shortcoming of retrospective Evaluation studies can be that of selective recall. For example, the recollection of the quality of a The main reason advanced for collecting ret- marriage after a divorce will often be differ- rospective information is that it provides a ent from that provided before the divorce. quick and efficient way of obtaining mea- Certain characteristics will be remembered sures of change and of constructing life histo- and seen as significant while others will be ries and avoids the expense and time of forgotten or re-interpreted. waiting for change to occur. Furthermore, Retrospective studies have a valuable role where data about past patterns simply were to play in social research. They are best not collected at the time (for example, levels suited to the construction of sequences of of premarital cohabitation), there is little significant events and in that respect can pro- alternative to using retrospective data to vide valuable insights into the sequence of explore patterns of change. events in individual lives or into historical Collecting information about significant changes at the aggregate level. However, past events can be quite reliable. For example, considerable care should be taken in using the date when a person first married, the date retrospective designs for studying subjective when they divorced or the date of birth of states and less memorable events. children can be reliably collected – especially from women. Major changes in jobs can be collected with a good degree of accuracy. As David de Vaus such, retrospective designs can be quite useful in constructing life histories about partic- Associated Concepts: causality, cross- ular types of past events. sectional surveys, life history interviewing, However, there are serious concerns longitudinal study, panel study, prospective about the ability of retrospective methods to study, sampling, social survey

269 R-Jupp-3373.qxd 3/8/2006 5:49 PM Page 270

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

Key Readings histories in longitudinal research’, in J. Z. Giele and G. H. Elder (eds), Methods of Dex, S. (1995) ‘The reliability of recall data: Life Course Research: Qualitative and a literature review’, Bulletin de Quantitative Approaches. Thousand Oaks, Methodologie Sociologique, 49: 58–89. CA: Sage. pp. 98–127. Freedman, D. S., Thornton, A., Camburn, D., Alwin, D. and Young-de Marco, L. (1988) ‘A life history calender: a technique for collecting retrospective data’, Sociological Methodology,18: 37–68. RISK RESEARCH Scott, J. and Alwin, D. (1998) ‘Retrospective versus prospective measurement of life See Prediction

270 S-Jupp-3373.qxd 3/8/2006 5:52 PM Page 271

sampling frame. A sampling frame can be any SAMPLING listing, for example, members of an organization, addresses (the post office address file) or Definition an electoral register. Techniques used to select groups from a The sampling unit is the unit of study. wider population. This is done because it is Whilst a unit is often an individual, it could not usually possible to include whole popula- also be groups of people. Advertisements, tions in research, for example as a result of newspapers or television programmes can time or financial constraints. Whom we study also be sampling units, on which content or and how we choose to study them is linked to other forms of textual analysis could be the theoretical context of the research and undertaken. The sample may be defined as the hypotheses or aims. Sampling theory is the segment of the population that is selected therefore based on the assumption that infer- for the research. ences can be made, or conclusions drawn Random or probability sampling refers to an about the population from which the sample approach where each element of the population is taken. Sampling is an important element in has an equal and known chance of being research planning and design. selected for inclusion in the sample. Non- random or non-probability sampling is an umbrella term that includes all forms of sam- Distinctive Features pling that do not adhere to probability methods. In sampling, the term ‘population’ has a very specific meaning and refers to the group of Evaluation people or other unit of analysis which is the focus of the study. The population as defined It is generally accepted that sample represen- will depend upon the research aims and the- tativeness is an important issue in social oretical context. research. If the views expressed by the sam- The sampling frame may be defined as the ple upon which research findings are based listing of all units in the working population do not reflect the views of the ‘population’, from which the sample will be selected. For the validity of the findings might be ques- example, if we were going to undertake a tioned. This is the issue of generalizability or survey of young people’s experiences of bully- external validity. Some researchers do argue, ing, we could decide to define the working pop- however, that this should not necessarily be ulation as all young people between the ages of the aim of sample selection in qualitative 12 and 15 attending schools in a particular city. research (Silverman, 1997), where the The schools would have lists of addresses of uniqueness of respondents’ experiences and registered pupils, although parental permission the depth of data provided are more impor- would be needed before they could be made tant. The following issues are identified as available. These lists would comprise the important in the literature on sample design S-Jupp-3373.qxd 3/8/2006 5:52 PM Page 272

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

(the caveat regarding qualitative research should be borne in mind). SCALING Sample response rates are important in survey research. Non-response introduces Definition error or bias. Non-response may relate to Methods used by researchers to quantify an entire survey or to particular items (item human psychological responses to stimuli. non-response). Non-response is important Many scaling methods have been developed because it affects the amount of data to capture feelings, judgements, opinions collected and the comprehensiveness of the and perceptions of stimuli. From such proce- data, in relation to the sample. It can also be dures, one can assess many human attributes, a source of bias in the evidence collected, for including attitudes, emotions, perceptions example when certain types of people may be and personality. They can also be used to cap- less likely to respond than others. This is not ture human judgements, such as in the evalu- solely a sampling problem and response rates ation of performance of athletes in Olympic can be affected by other factors such as the competition or employees in the workplace. method of data collection. Statistical methods may be used to correct for non-response. Distinctive Features Sample size is an important issue in random sampling, as this approach is based upon Every scaling procedure consists of two mathematical concepts such as probability the- components – stimuli and a response method. ory and upon inferential statistics. Some statis- The stimuli can be actual objects, but typically tical calculations can only be performed on they are written words or phrases that random samples and some require a minimum describe the stimuli or situation. Responses sample, size. Generally, larger samples provide can be spoken or written using a structured more accurate findings in quantitative research. procedure. Part of that procedure is the way in which the responses are quantified. Responses are made on a specified dimension, such as Julia Davidson evaluation or liking. For example, a researcher might wish to assess how much voters like Associated Concepts: bias, cluster sam- each candidate running for a particular office pling, inferential statistics, non-probability in an election. A scaling method could provide (non-random) sampling, opportunity sam- a score for each candidate that reflects how pling, panel study, probability (random) sam- much voters like him or her. pling, quota sampling, snowball sampling, Scaling methods were developed to assess social survey human responses at ordinal or ratio levels of measurement. However, although in theory these methods can achieve this goal, it is Key Readings quite difficult to demonstrate measurement level. Thus, measurement level with scaling Bryman, A. (2001) Social Research Methods. Oxford: Oxford University Press. methods is debatable, as many researchers May, T. (2001) Social Research: Issues, feel only ordinal measurement is achieved. Methods and Process, 3rd edn. Milton Nevertheless, these methods are widely used Keynes: Open University Press. in human assessment. Sapsford, R. and Jupp, V. (eds) (l996) Data One of the simplest methods is paired Collection and Analysis. London: Sage. comparisons. With this method stimuli are Silverman, D. (ed.) (1997) Qualitative presented in pairs, and a response dimension Research: Theory, Method and Practice. is defined. People are asked to indicate which London: Sage. stimulus in each pair is higher or lower on the de Vaus, D. A. (l996) Surveys in Social dimension of interest. For example, they can be Research, 4th edn. London: UCL Press. asked to indicate which of two candidates they

272 S-Jupp-3373.qxd 3/8/2006 5:52 PM Page 273

SCALING

like best. Mathematical procedures can be used ease in quantifying responses. The response to translate the pattern of responses into scores choices are numbered consecutively from for the stimuli. The paired comparison method lowest to highest, and overall stimulus values is one of the easiest for people to use, but unless are computed as the mean of all respondent there are very few stimuli the number of com- scores. It is also used to provide individual parisons is too large for most applications. respondent scores for each stimulus as opposed The rank order method is a good alternative to overall stimulus scores across all people. to paired comparisons when there are many The dimension underlying the rating scale stimuli. With this method respondents are can be almost anything, but most use one of a asked to place the stimuli in order along the very small number of possible options. Perhaps defined dimension. For example, respondents the most popular is agreement. Respondents might list candidates in order from least to indicate how much they agree or disagree most liked. Mathematical procedures can be which each item, (for example, ‘Disagree very used to derive scores for each candidate much’, ‘Disagree slightly’, ‘Agree slightly’, across the respondents. Scores with this pro- ‘Agree very much’). Other popular dimensions cedure are usually quite similar to those from are evaluation (for example, ‘Poor’, ‘Fair’, the paired comparison method. However, it is ‘Good’, ‘Excellent’), frequency (for example, somewhat more difficult for respondents to ‘Never’, ‘Seldom’, ‘Frequently’, ‘Often’), and rank order stimuli than paired comparisons. liking (see prior paragraph for example). Direct magnitude estimation is a method in Hundreds of instruments to assess specific which respondents are asked to provide a constructs have been developed using scaling score on a dimension for each stimulus. This methods. Most include multiple items to assess can be done most easily by having people each construct of interest, and yield individual choose a number from 1 to 100, where 1 scores for each respondent. Such scales are represents the lowest score on the dimension called summated rating scales when they use and 100 the highest. Respondents could be the rating scale format for responses. asked, for example, to choose a number for each candidate. Another variation is to use a Evaluation line to represent the dimension of interest. The extremes of the dimension are indicated, The scaling procedures mentioned so far have for example, ‘Dislike very much’ and ‘Like involved single dimensions on which stimuli very much’. The respondent is asked to make are scaled, for example, liking. Stimuli are a mark representing his or her position on the ordered along the single continuum from least dimension in question. A separate line can be to most liked. It is possible, however, to use used for each stimulus or one line can be more complex scaling methods in which multi- used for all stimuli. Direct magnitude estima- ple dimensions are dealt with simultaneously. tion has not been widely used, in part Multidimensional scaling procedures allow one because it is not clear that people are able to to identify two or more dimensions on which provide the precision of judgement necessary. stimuli can differ. Graphical representations of Rating scales ask people to select one the stimulus positions in multidimensional response choice from several that are ordered space can help interpret the ways in which along a dimension. The choices are defined individual stimuli differ from one another. by an anchor, or statement that describes its These procedures are quite complex, requiring position on the dimension. For example, for considerable statistical sophistication to use. assessing liking of election candidates the anchors might be, ‘Dislike very much’, ‘Dislike slightly’, ‘Like slightly’ and ‘Like very much’. Paul E. Spector Respondents give a rating for each candidate. Each stimulus is termed an item. This method Associate Concepts: composite measurement, is the most widely used, perhaps because of its indicator, Likert scale, semantic differential

273 S-Jupp-3373.qxd 3/8/2006 5:52 PM Page 274

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

scale technique, summated rating scale, by academics and social scientists are important Thurstone scale, validity of measurement resources for secondary analysis. Data sets can be continuous – that is, gathered on a number Key Readings of occasions over a period of time – or ad hoc – carried out just once in response to a specific Guilford, J. P. (1954) Psychometric Methods, research request (Procter, 1993). An important 2nd edn. New York: McGraw–Hill. example of a continuous data set is the UK gov- Torgerson, W. S. (1958) Theory and Methods ernment’s General Household Survey (GHS), of Scaling. New York: Wiley. carried out every year since 1971. An example of an ad hoc survey is the Women and Employ- ment survey carried out in 1984 by the UK government’s Office of Population Censuses and SECONDARY ANALYSIS Surveys (OPCS). Other data sets have been generated precisely for the purpose of providing a Definition secondary analysis resource; an example is the General Social Survey collected by US acade- The further analysis of an existing data set mics at the National Research Centre at with the aim of addressing a research ques- Chicago. While traditionally secondary analysis tion distinct from that for which the data set has been more widely used with quantitative was originally collected, and generating novel data sets, the analysis of qualitative data sets interpretations and conclusions. has recently received increasing attention. A collection of qualitative data sets – ‘Qualidata’ – Distinctive Features is now available at the UK data archive at Essex (www.qualidata.essex.ac.uk). The use of secondary analysis with large-scale Having accessed a suitable data set the next survey data, using statistical analysis tech- step is to prepare this for re-analysis, which may niques, is most common, and has been well involve recoding the original variables (which documented. A researcher may re-analyse a can be done using a statistical software package data set that he or she has collected earlier, or such as SPSS). Stewart (1984) has outlined a may use a data set collected by another number of questions that should be answered researcher, or team of researchers. The tech- prior to engaging in secondary analysis, includ- nique can be distinguished from meta-analysis, ing: What was the purpose of the original which attempts to summarize the already research? What sampling procedure was used? interpreted results of a number of primary What information was collected? Is the original research studies; in contrast, secondary analysis data set valid and reliable? attempts to re-interpret the original data set in relation to a new research question. Evaluation The first stage of secondary analysis is to obtain a suitable data set. Data sets are available Secondary analysis is an important technique from databases, which contain information for certain types of research, in particular about the context (time of collection, sampling research requiring large-scale and longitudi- techniques employed and so on) of each data set nal data sets. Using data that have been col- they offer. The data sets available are of high lected by a specialist team of experts not only quality, often the result of years of data collec- maximizes data quality, but also proves effi- tion by a team of highly qualified experts. cient in terms of time and cost. Large data The main UK database is the UK National sets often contain a wealth of information Data Archive, housed at Essex University that is not exhausted in relation to one particu- (www.data-archive.ac.uk). Most developed lar research question. Secondary analysis both countries have an equivalent national database. facilitates research where the data required Both government data sets, and those gathered would be too costly and time-consuming for

274 S-Jupp-3373.qxd 3/8/2006 5:52 PM Page 275

SELF-REPORT STUDY

the researcher to collect themselves, and can Hakim, C. (1982) Secondary Analysis in Social enable access to special populations due to the Research: A Guide to Data Sources and sheer size of the data sets often available Methods with Examples. London: Allen (Procter, 1993). Longitudinal studies are also and Unwin. facilitated through access to continuous data Procter, M. (1993) ‘Analysing other sets such as the GHS. Instead of having to plan researchers’ data’, in N. Gilbert (ed.), the collection of a primary data, the secondary Researching Social Life. London: Sage. analyst is freed up to focus on analysis and pp. 255–69. interpretation. Stewart, D. W. (1984) Secondary Research: Information Sources and Methods. Beverly A number of drawbacks emerge when eval- Hills, CA: Sage. uating secondary analysis, however. The researcher is not able to make decisions concerning the way the data is collected, and therefore must be careful to assess the original data set on a number of key aspects (such as those SELF-REPORT STUDY identified by Stewart, 1984) in order to decide whether it is suitable for the new research ques- Definition tion. There is always the danger in secondary research of tailoring the research question to fit A study in which respondents report their the data (Procter, 1993). Lack of contact with own behaviour. For example, in self-report the primary researchers can be a problem if fur- studies of criminal behaviour, respondents ther questions emerge about the context of data are asked to declare if they have committed collection which are not addressed in the anno- criminal offences. Self-report studies are tations provided in the relevant database. This often done for comparison with official issue of context of the original data collection records. procedures may be especially important in the re-analysis of qualitative data sets; indeed, the Distinctive Features question of whether it is possible for a secondary analyst to properly engage with qualita- A self-report study is used where the point of tive data which they were not involved in interest is in comparing official or organiza- collecting has been raised as an issue. The tional records with alternative measures of researcher engaging in secondary analysis must behaviour. Examples include research on also be aware of ethical requirements, and the physical and mental health, and voting extent to which the original ethical procedures behaviour. Self-report studies are particularly (such as obtaining informed consent) extend to prominent in research on crime. the new analysis being carried out. The validity of official criminal statistics is limited by omissions and bias; a self-report study offers the promise of data free of these Claire Hewson problems. For example, respondents may be asked if they have committed offences during Associated Concepts: data, data archives, a set period (often the past 12 months) and, ethics, informed consent, meta-analysis, offi- if they have, how often. Self-report studies cial statistics, primary research, qualitative are conducted using self-completion ques- data archives, SPSS, validity tionnaires or by interview (Jupp, 1996). The research instrument is usually a list of offences (and sometimes other behaviours). Key Readings Many self-report studies are conducted on Dale, A., Arber, S. and Procter, M. (1988) juveniles, both because delinquency is an Doing Secondary Analysis. London: Unwin assumed indicator of crime trends, and self- Hyman. report studies of adults can prompt demands

275 S-Jupp-3373.qxd 3/8/2006 5:52 PM Page 276

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

to disclose respondent identities to the impress. To assess this, several studies authorities. followed completion of the offence list by Empey and Erikson’s (1966) landmark telling respondents they would now take a study suggested that over 90 per cent of delin- polygraph test and inviting them to change quency went undetected. Males aged 15–17 any responses before doing so. Few took the were interviewed about 22 offence types. The opportunity. offence was defined, the respondent asked if However, this cannot detect respondents he had participated, and, if so, how often, who believe they are telling the truth but and whether he had been detected, arrested whose recall is faulty, a particular problem and appeared in court. As a validity check it where recall periods are long or the number of was noted that none of those with a court offences committed is large. Another validity record concealed it. check is to compare results with other sources. The sample of 180 boys admitted to enor- For example, Gold (1966) disclosed results to mous numbers of offences, including 24,199 informants like teachers who knew respon- thefts and 9026 assaults. They also admitted dents, asking if the findings rang true, and esti- numerous instances of ‘defying authority’ and mated that up to 72 per cent were truthful. under-age drinking. Note that the former is not This still only controls concealment, not exag- a crime, and the latter only a crime because of geration, and is only as valid as informants’ the respondent’s age; self-report studies have knowledge – after all, the method was devised been criticized for including behaviours that to compensate the failings of other sources. are not illegal. Empey and Erikson’s other Non-response bias is also a threat, because notable finding was the large proportion who refusers may be particularly active offenders. had committed offences. For example, 93 per Sampling schoolchildren introduces lower- cent reported committing petty theft and 59 class and ethnic minority bias because of their per cent, a serious theft or robbery. higher drop-out and truancy rates (Kleck, These were important findings as official 1982; Junger, 1989). Further, aggregating triv- statistics gave no hint of such widespread ial behaviours with offences overstates the delinquency. Subsequent studies tested if proportion involved in ‘delinquency’; for these findings applied generally; a particular example, Elliot and Ageton’s (1980) otherwise concern was the implication that class differ- sophisticated study included items like ‘hitch- ences were exaggerated. After reviewing 40 hiked’ and ‘been loud, rowdy or unruly in a self-report studies, Box (1980) concluded that public place’. working class juveniles were not dispropor- Offending is not the only field where prob- tionately delinquent. But self-report methods lems arise. Admitting involvement in socially also encountered methodological criticisms. disapproved behaviours is subject to cultural Researchers remain divided between those factors and can suggest false variance. who see self-report as flawed and those who Dohrenwend’s (1964) study of mental illness feel it is reliable. It remains much-used in in New York found that Puerto Ricans had a criminological and other research, particu- higher self-reported rate than Jewish- larly social aspects of health, for example, in Americans, Irish-Americans or African- AIDS research where respondents are asked Americans. However, when Dohrenwend to report ‘risky’ sexual behaviours. subjected the mental health inventory to a social desirability rating it emerged that Evaluation Puerto Ricans regarded the items as less undesirable than did the other groups and Self-report methods rely on respondents’ were thus more willing to admit to them. honesty. If they regard researchers with sus- Once methodological, measurement and picion as agents of authority they may down- selection biases are accommodated there is play their involvement. But they might also considerable congruence between self-report ‘talk up’ their involvement in an effort to and official statistics. Farrington’s (1989)

276 S-Jupp-3373.qxd 3/8/2006 5:52 PM Page 277

SEMANTIC DIFFERENTIAL SCALE

sample reported contacts with police accurately, and under 4 per cent who denied SEMANTIC DIFFERENTIAL SCALE committing particular offences had actually been convicted of such offences. Some thus Definition argue that official and self-report data are A method of attitude measurement in which complementary and that researchers can respondents rate the object of interest (for have considerable confidence in truthful example, the police) on several bipolar scales, answers being given (Tarling, 1993). the ends of which are defined by pairs of adjectives that are opposite in meaning (for Nigel G. Fielding example, honest/dishonest; hard/gentle).

Associated Concepts: bias, indicator, inter- Distinctive Features view, official statistics, social survey, triangu- The semantic differential rating scale was lation, validity developed by Osgood in the late 1950s. The technique is very adaptable and easy to administer. It is concerned with respondents’ Key Readings perceptions or attitudes in relation to the Box, S. (1980) Deviance, Reality and Society. object of the research. Respondents are pre- New York: Holt Saunders. sented with a number of items each of which Dohrenwend, B. (1964) ‘A use for leading is presented as opposites. For example, if the questions in research interviewing’, aim is to examine attitudes towards police offi- Human Organization, 23: 76–7. cers, bipolar items such as kind/cruel, power- Elliott, D. and Ageton, S. (1980) ‘Reconciling ful/not powerful, weak/strong, honest/ race and class differences in self-reported dishonest are offered. Such items can be and official estimates of delinquency’, developed from repertory grid techniques. American Sociological Review, 45: 95–116. Sometimes an assumption is made about Empey, E. and Erikson, M. (1966) ‘Hidden equality of intervals so that each item can be delinquency and social status’, Social scored 1–7 and by summation an overall score Forces, 44: 546–54. can be determined. In this way the image or Farrington, D. (1989) ‘Self reported and profile of almost anything can be obtained. official offending from adolescence to adulthood’, in M. Klein (ed.), Cross- national Research in Self-reported Crime Evaluation and Delinquency. Dordrecht: Kluwer. As with other such techniques, the semantic pp. 399–423. differential runs the risk of response sets, Gold, M. (1966) ‘Undetected delinquent behaviour’, Journal of Research in Crime which is a tendency to respond in a particu- and Delinquency, 3: 27–46. lar way irrespective of the context of the Kleck, G. (1982) ‘On the use of self-report questions. For example, there is a risk that data to determine the class distribution of respondents choose all the positive adjectives criminal and delinquent behaviour’, or all the negative adjectives. Therefore, it is American Sociological Review, 47: 427–33. necessary to vary the ends or poles of the Junger, M. (1989) ‘Discrepancies between items to counteract such response sets. police and self report data for Dutch ethnic minorities’, British Journal of Criminology, 29 (3): 273–84. Victor Jupp Jupp, V. (1996) Methods of Criminological Research. London: Routledge. Associated Concepts: repertory grid tech- Tarling, R. (1993) Analysing Offending. nique, scaling, summated rating scale, London: Her Majesty’s Stationery Office. Thurstone scale

277 S-Jupp-3373.qxd 3/8/2006 5:52 PM Page 278

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

Key Readings safeguards. This is best illustrated with an example from my own experience. Moser, C. A. and Kalton, G. (1971) Survey I conducted a longitudinal study with men Methods in Social Investigation. London: who had been convicted for committing sex- Heinemann. ual offences against children; the research Osgood, C. E., Suci, G. J. and Tannenbaum, P. H. (1957) The Measurement of Meaning. was conducted within the context of proba- Urbana, IL: University of Illinois Press. tion work and interviews held on probation service premises. There were several issues to consider at the outset: first, participation in all social research should be voluntary. The difficulty here was that potential respondents SEMIOTICS may have felt forced to participate as they were probationers, or may have felt that See Textual Analysis participation would have a positive effect upon their sentence. In order to overcome this the group were informed that participation was on an entirely voluntary basis and would not affect their sentence outcome. The SENSITIVE RESEARCH second more problematic issue was that of confidentiality. This is extremely important Definition given the nature of the research and the Sensitive research may be defined as that which researcher’s desire to protect the identity of addresses difficult areas such as sexual respondents, whilst encouraging full partici- deviance, or that which is conducted with vul- pation. Assurances regarding confidentiality nerable groups of respondents such as children. were compounded by the existence of a child protection policy within the organization. Distinctive Features The policy required that any information divulged by this group of offenders regarding Primary qualitative methods of data collec- the commission of un-convicted offences tion (depth interviewing and observation for against children be passed to the police. example) are frequently employed in sensi- Respondents were informed about this policy tive research, particularly work that seeks to before they agreed to participate in the study. explore feelings and beliefs amongst vulnera- Since the research aimed to explore previous ble or difficult to access groups. The intru- offending, this may have impacted upon the siveness of these techniques can increase the extent to which respondents were willing to sensitivity of such research and consequently participate fully in interviews. care must be taken in designing and undertaking this type of work. Reference to a code Evaluation of ethical conduct, such as that produced by the British Sociological Association, should The effects of such research upon researchers always be made. is an important but largely neglected area and Researchers should not be deterred from two issues were raised here: first the physical conducting sensitive social research as this is safety of researchers working with these offend- frequently the only way that policy makers and ers (some of whom had convictions for violent social scientists can learn about socially behaviour). It was agreed that one-to-one inter- excluded groups, for example, but plans should views with offenders would be conducted in a be made to safeguard both the researcher and room with an adjacent observational room the researched at the design stage. There may (using video equipment), observed by a second need to be a contractual agreement with the researcher. Fortunately, these facilities were funder regarding the nature of the planned available. This set-up allowed for observation

278 S-Jupp-3373.qxd 3/8/2006 5:52 PM Page 279

SENSITIZING CONCEPTS

of interviews, without the intrusive presence founder of symbolic interactionism. By of another researcher in the interview room. contrast with what he called ‘definitive’ Respondents’ written permission to be concepts, sensitizing concepts do not involve observed in this way was sought prior to inter- using ‘fixed and specific procedures’ to identify views and the presence of a video recorder a set of phenomena, but instead give ‘a general in the room did not appear to interfere with sense of reference and guidance in approach- the conversational flow of the interview. The ing empirical instances’. So, whereas definitive second point relates to the impact upon the concepts ‘provide prescriptions of what to see, researcher. Accounts of this type of offending sensitizing concepts merely suggest directions can be distressing and researchers are unable along which to look’ (Blumer, 1969: 148). to share such disturbing information with others not involved in the research, particularly Distinctive Features given confidentiality and anonymity issues. Blumer recognized that vagueness in theoret- Such graphic material will have an impact ical concepts could be a problem. However, upon the recipient, particularly in longitudinal he denied that this can be solved in the social work; there should be recognition of this and a sciences by developing ‘definitive’ concepts – strategy put in place during the research design the solution put forward by some influential phase. In this research regular de-briefing ses- US sociologists in the 1930s, 40s and 50s, sions were held between the researchers including advocates of operationism. directly involved in this work, and the sponsor Blumer’s championing of sensitizing concepts agreed to fund sessions with a therapist when reflected the pragmatist philosophy from requested. This is unusual but constitutes good which his own thinking about methodology research practice given the highly disturbing grew. Many pragmatists argued that concepts and sensitive nature of the work. are tools for use, and should be judged according to their instrumental value in Julia Davidson enabling people to understand, and especially to act upon, the world. Also crucial to Associated Concepts: ethics, informed con- Blumer’s argument is the idea that concepts sent, interview, research bargain, volunteer have to be developed and refined over the sampling course of inquiry, rather than set in stone at the beginning, if they are to capture the nature of phenomena. He insisted that begin- Key Readings ning research by stating precise definitions of Ethical Codes of Conduct: concepts, as advocated by many positivists, http://www.britsoc.co.uk British Sociological amounts to a spurious scientificity; and is Association likely to obstruct the task of gaining genuine http://www.the-sra.org.uk/ethicals.htm knowledge about the social world. Social Research Association http://www.britsoccrim.org/ethics.htm Evaluation British Society of Criminology There are suggestions within Blumer’s work of the more radical view that, in an important sense, vagueness may be an inevitable feature of social scientific concepts. One rea- SENSITIZING CONCEPTS son for this is the difficulty of capturing the diverse, processual and contingent nature of Definition social phenomena. Blumer argues that this seems to imply inevitable reliance on tacit A notion found in the methodological writings cultural knowledge, in other words on of the US sociologist Herbert Blumer, the Verstehen. Thus, he writes that only physical

279 S-Jupp-3373.qxd 3/8/2006 5:52 PM Page 280

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

objects can be ‘translated into a space–time Distinctive Features framework or brought inside of what George The understanding, analysis, and design of Mead has called the touch–sight field’, complex dynamic systems are at the core of whereas the observation of social phenomena simulation employed for scientific and engi- requires ‘a judgment based on sensing the neering research. Realistic animations are social relations of the situation in which the central for using simulation for entertain- behavior occurs and on applying some social ment, training and education. norm present in the experience of the Whereas modelling is primarily concerned observer’ (Blumer, 1969: 178). with the relation between reality and model, Blumer recognizes that most of the key simulation is concerned with the relation concepts of sociology fulfil an important sensi- between model and simulator (Zeigler et al., tizing role. However, he insists that they need 2000). The simulator is responsible for to be clarified, as far as this is possible, through executing the model. researchers enriching their first-hand experi- The simulation of a model comprises the ence of the social phenomena to which they following phases: simulation initialization, relate. This is to be done, above all, by what he simulation run and interpretation of results. calls ‘naturalistic research’, which is close to The simulation experiment has to take into what has come to be referred to as ethnogra- account the constraints that are inherent in a phy, in which empirical instances are subjected model. Simulation experiments are executed to ‘careful flexible scrutiny’ in their natural during the validation of the model but also context, ‘with an eye to disengaging the generic for predicting and analysing the behaviour of nature’ of the concept (Blumer, 1969: 44). the system under study. In the latter case experiments have to be consistent with the Martyn Hammersley set of experimental conditions for which the model has been validated, that is, the model’s Associated Concepts: ethnography, grounded experimental frame. The initialization of theory, positivism, research design, Verstehen the simulation includes the estimation of parameters and the identification of distribu- Key Readings tions for the inputs of the model. Beginning, Blumer, H. (1969) Symbolic Interactionism. completion and number of simulation runs Englewood Cliffs, NJ: Prentice Hall. are defined. Hammersley, M. (1989a) The Dilemma of Simulation is very similar to experimenta- Qualitative Method: Herbert Blumer and the tion in the laboratory: many simulation runs Chicago School. London: Routledge. are required for understanding a model’s Hammersley, M. (1989b) ‘The problem of behaviour. This is true not only for the simu- the concept: Herbert Blumer on the lation of stochastic models but also for the relationship between concepts and data’, simulation of deterministic models. Journal of Contemporary Ethnography, Verification and validation, although they 18 (2): 133–60. are conceptually distinct, are conducted simultaneously by the researcher. Verification involves assuring that the conceptual model is SIMULATION adequately represented in the computerized representation, and that the simulation execu- Definition tion is correct with respect to the computerized representation of the model. For validation An experiment performed on a model and the model and the behaviour produced during aimed at imitating the operation of systems simulation are compared to the real system and over time for the purpose of analysis or of its behaviour. Validation involves checking face creating virtual worlds. validity, that is, model and behaviour produced

280 S-Jupp-3373.qxd 3/8/2006 5:52 PM Page 281

SNOWBALL SAMPLING

during simulation should appear reasonable. Associated Concepts: econometrics, experi- Sensitivity analysis is also used to check a ment, modelling, validity model’s face validity. Model assumptions, structural and data assumptions have to be val- Key Readings idated the same as input–output transformations (Banks et al., 2001). Banks, J., Carson, J. S., Nelson, B. L. and Simulation employed for analysis and Nicol, D. M. (2001) Discrete Event System design requires that simulation runs are repro- Simulation. New York: Prentice Hall. Fujimoto, R. (2000) Parallel and Distributed ducible and correct (Zeigler et al., 2000). Simulation Systems. Chichester: Wiley. Numerical integration algorithms have to be Zeigler, B. P., Praehofer, H. and Kim, T. K. carefully selected to produce valid results (2000) Theory of Modeling and Simulation. during continuous simulation. In discrete event New York: Academic Press. simulation specific strategies are defined to resolve uniformly potential conflicts between events and to prevent causal errors. Virtual environments, in which simulation is used for entertainment, learning, or train- SNOWBALL SAMPLING ing, require a sufficiently realistic imitation of the system’s behaviour. Whereas the simu- Definition lation for analysis is executed as fast as possible, in virtual environments the progression A form of non-probability sampling in which the of simulation time is related to the progres- researcher begins by identifying an individual sion of wall clock time. In both cases an effi- perceived to be an appropriate respondent. This cient and effective simulation may require respondent is then asked to identify another the application of parallel, distributed simu- potential respondent. The process is repeated lation methods (Fujimoto, 2000). until the researcher has collected sufficient data. Sometimes called ‘chain letter’ sampling. Evaluation Distinctive Features Analytical techniques have generally a smaller domain of applicability than simula- Snowball sampling can be a useful technique in tion. However, analytical techniques are research concerned with behaviour that is more powerful, as they provide an under- socially unacceptable or involves criminal activ- standing of the model’s behaviour under arbi- ity. The nature of such activities may make it a trary experimental conditions. Thus, they virtually impossible task to identify all members should be applied where possible. However, of the research population; even identifying a often they are not applicable and simulation is few members can be very difficult. In the case the only technique available for the analysis of of research on, say, shoplifting or car theft, the the system’s behaviour. identification of a single willing respondent may This generality and ease of applicability be difficult. The first stage in the process usually are the strengths of simulation but ironically involves a purposive sampling decision to iden- its most serious drawbacks. All too fast, users tify one respondent who is willing to provide fall prey to the easiness of developing models data. Once the data collection has been com- and letting them run. Model and simulation pleted, the researcher asks the respondent to represent a system and its behaviour only nominate another person who may be willing to under certain constraints, which have to be provide the type of data required. The process carefully observed. continues until either the researcher fails to make any new contacts, or the new data do not appear to add anything substantial to existing Adelinde Uhrmacher understanding.

281 S-Jupp-3373.qxd 3/8/2006 5:52 PM Page 282

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

Snowball sampling can also be a relevant technique for groups of people who may feel SOCIAL INDICATORS lacking in confidence to participate in a research project. Such people could include Definition the homeless, alcoholics, or those who have Social indicators, often combined into suffered illness or an assault. In such cases, ‘indices’, are indirect empirical representa- they may have more confidence and be more tions used to define or refer to concepts when likely to participate if they are approached by no direct measurement is possible. They are a person with similar experiences. a special sub-set of the more general problem of operationalization, which is how to formu- Evaluation late precise definitions of social phenomena in empirical terms, ready for data collection. The advantage of snowball sampling is that it enables the researcher to identify Distinctive Features potential participants when it would otherwise be extremely difficult to do so. It is also Many social science topics can only be a sampling strategy that demonstrates a researched indirectly. Abstract concepts have sensitivity to potential participants, in that to be converted into separate, clearly speci- they are identified by people with a similar fied components that can be studied empiri- experience. cally. We cannot tell whether a ‘thing’ is The disadvantage of the approach is that it is present or absent, how often it occurs, in dependent upon each participant sufficiently what circumstances and what importance it understanding the nature of the research in has, without these intermediate constructions. order to be able to identify another suitable par- However, whereas some phenomena can be ticipant. The next nominated participant may relatively easy to operationalize, others are have a limited or biased understanding of the more difficult. research issue. In addition, the members of the In Western societies, classification by age snowball sample may have certain features in and gender can normally be achieved by common which are uncharacteristic of the straightforward measurement of years and research population as a whole. The very fact biological sex. Of course, the cut-off points of that they are all acquainted with each other is a age bands, or gender identification depend on source of potential bias. the research focus. Other topics are more problematic. To take the topics covered in just one issue of a recent journal, ‘refugee’, Paul Oliver ‘community’, ‘forced migration’, ‘social integration’, ‘gypsy’, ‘inner city schooling’ and Associated Concepts: non-probability ‘governance’ are all social phenomena requir- (non-random) sampling, opportunity sam- ing operationalization. If we consider the first pling, probability (random) sampling, purpo- of these, even the number of ‘refugees’ – a sive sampling, quota sampling, sampling politically important question – cannot be accurately counted without defining who is, and who is not, a refugee. Can a person with Key Readings a work permit also be a refugee? Can one Bryman, A. (2001) Social Research Methods. be a refugee in one’s own country, or only if Oxford: Oxford University Press. officially listed by a government or non- Knight, P. T. (2002) Small-scale Research: governmental agency? If refused such official Pragmatic Inquiry in Social Science and the recognition, does a person immediately cease Caring Professions. London: Sage. to be a refugee? Lee, R. M. (1993) Doing Research on Sensitive Concepts like ‘poverty’ and ‘social exclu- Topics. London: Sage. sion’ are even more complex. They are

282 S-Jupp-3373.qxd 3/8/2006 5:52 PM Page 283

SOCIAL INDICATORS

conventionally measured by social indicators, they ‘measured’ the indicator, the canary’s sets of variables which are known to be asso- health. ciated with the key concept. Local social In the social deprivation indicators listed deprivation has been identified using multiple- above, the first dozen or so variables are con- indicator combinations of scores for, among veniently available from the UK Census others, unemployment rates, overcrowding, (albeit likely to be out-of-date because the lack of housing amenities, renting, residential Census is taken once a decade), while the turnover, car ownership, occupational profile, remainder can be obtained from other pre- single parent and pensioner household struc- existing official statistics. They may not be tures, young children, qualification levels and the best possible indicators, but they have illness, together with death rates, derelict land, been chosen because the data are readily to benefit recipients and contents insurance pre- hand, in some cases for local authority areas, mium levels (Payne et al., 1996). and in others down to ward or enumeration No single indicator would be sufficient on its district. While ready-made social indicators own, nor do scores precisely measure the are therefore useful, this also should alert underlying concept (Carley, 1981). Alternative us to the fact that they may have been con- social indicators of deprivation (the ‘Carstairs structed for administrative convenience, or Index’, the ‘Jarman Index’, for example) have political expediency, rather than social scien- chosen different combinations to meet particu- tific precision. lar needs, like health resources allocation or However, indicators may change indepen- determining local government grants. Social dently of the original concept. Social class indicators of disadvantage can be seen as an is about social groupings developing from alternative to dependence on social class; some, ownership and control of production. like Townsend et al. (1992) seeking to expand Traditionally, this has been empirically repre- class’ explanatory powers, while others like sented by occupation. That worked tolerably Carr-Hill (1990) wishing to escape class alto- well while most workers were men in full- gether. Social indicators may be about measure- time paid employment, in an unchanging ment, but they are embedded in theoretical labour market consisting of large blocks of perspectives. similar occupations. However, the rise of Social indicators should achieve four goals. female employment, part-time working, self- They must properly reflect the essential employment, second jobs, early retirement, nature of the core concept, while also cover- high levels of unemployment, training- ing the whole of the concept (which is why delayed entry into the labour market and new they are usually based on multiple indica- types of occupation have all complicated the tors). They must exclude other non-relevant scene. Nor are life style or social identity as concepts. They must produce evidence in a neatly associated with occupation as they form appropriate to the level of study, either once were. It is impossible to tell if social as a quantified form of measurement or con- class per se has changed, independent of the tributing to a plausible narrative of meanings changes in the indicator. If we measure some- and interpretations. thing indirectly, this must always be a risk. This tendency for operational definitions Evaluation to freeze concepts in a fixed way is one of the main reasons why qualitative researchers Social indicators are sometimes chosen reject prior operationalization, preferring to because they are effective rather than exact. retain flexibility of response to fieldwork Miners used to take canaries down the pits: encounters. Pre-conceptualization cannot cover when the canary collapsed, it was time to the full complexity of a concept, and depen- escape, because the air was bad. The bird was dence on common sense is likely to result in more susceptible to carbon monoxide than the researcher’s own ideas being imposed at the men. The miners did not measure the gas; the expense of the informants’ subjective

283 S-Jupp-3373.qxd 3/8/2006 5:52 PM Page 284

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

meanings. Exploratory studies are a good way cases, columns representing variables and cells of discovering and clarifying the components that contain information about a case’s and variations in the empirical manifesta- attribute on the specific variable. tions of a concept. There are two broad types of social surveys – descriptive and explanatory. The descriptive survey seeks to describe the dis- Geoff Payne tribution of phenomena in a sample and population. The explanatory survey seeks to Associated Concepts: composite measure- explain relationships between variables – to ment, exploratory research, indicator, mea- explain why things are as they are. Many surement, multiple indicators, positivism, surveys fulfil both functions. realism, reliability, sensitizing concepts, Survey analysis is based on systematically validity of measurement comparing cases and examining variation and correlation between variables. Explanations Key Readings are sought by examining variation in the dependent variable (presumed effect) and Carley, M. (1981) Social Measurement and selecting an independent variable (presumed Social Indicators. London: Allen & Unwin. cause) that might be responsible for this vari- Carr-Hill, R. (1990) ‘The measurement of ation. Analysis involves observing whether inequalities in health’, Social Science and variation in the dependent variable (for exam- Medicine, 31 (3): 393–404. ple, income) is systematically linked to varia- Payne, J., Payne, G. and Hyde, M. (1996) tion in the independent variable (such as ‘The refuse of all classes’, Sociological gender). While any such covariation does not Research Online, 1 (1). demonstrate causal relationships, such Townsend, P., Davidson, N. and Whitehead, covariation is a prerequisite for causal rela- M. (1992) Inequalities in Health: The Black tionships. More complex multivariate data Report and the Health Divide (rev. edn). Harmondsworth: Penguin. analysis methods seek to untangle the complex relationships between the many factors that affect social behaviour and to control for the effects of extraneous variables that experiments achieve via random allocation to SOCIAL SURVEY groups. While survey research is often equated Definition with questionnaire-based studies there is no necessary relationship between the survey A method of social research with three defin- method and the particular techniques by ing characteristics – its type of content, its which data are collected. A critical and dis- form of the data and the method of analysis tinguishing characteristic of a survey is that it employed (Marsh, 1982). Its content is social, yields a structured data set that produces a the form of data is systematic, structured and variable-by-case data grid. The particular based around variables and the method of techniques by which the data are collected analysis relies on comparisons across groups. for this grid can vary widely.

Distinctive Features Evaluation The content of a social survey is social – it deals Social surveys can provide a relatively effi- with human behaviour, knowledge, attributes, cient method for collecting information from beliefs and attitudes. Surveys produce a struc- a large number of cases. The data grid that is tured data set in the form of a variable-by-case produced in a survey is well suited to statisti- grid. The grid consists of rows, representing cal analysis and thus is amenable to the

284 S-Jupp-3373.qxd 3/8/2006 5:52 PM Page 285

SOCIOMETRY

potential power of statistical methods. Where individuals in small group situations. a survey collects data from large probability Sociometry is a form of network analysis. samples the extent to which patterns in the sample are likely to hold in the population Distinctive Features at large can be estimated as such survey research results can be generalized to a Sociometry was developed in the 1930s by wider population with a known degree of the social psychologist Jacob Moreno. confidence. Moreno introduced the idea of a sociogram, There is general agreement that surveys which is a diagram representing the relation- provide an effective way of describing the ships between individuals. The sociogram, more objective characteristics of a popula- now often called a directed graph, is a dia- tion. However, there is less agreement about grammatic representation of a social net- whether survey research can produce data work, with individuals represented by dots or that enable the testing of causal relationships points and social relationships represented by and understanding of behaviour in terms of lines. Arrow heads can be added to the lines its subjective meaning to actors. Some critics to indicate whether a relationship is one-way argue that the focus on variables ignores the or reciprocal. Additionally, the strength of a context in which the behaviour occurs and relationship can be indicated by the thickness the intentions of actors. They argue that sur- of a line. veys rely on imposing external explanations The technique used to develop a socio- for behaviour and ignore the intentional and gram requires each member of a group to subjective component. In her book The indicate which other members he or she likes Survey Method Catherine Marsh reviewed most. A limit is put on the number of choices and countered these criticisms (Marsh, (usually three) each member may make. 1982). Once the choices have been collected from each member, the data are used to draw a sociogram, which can then be analysed to David de Vaus make inferences about the structure of a group, its integration and connectedness, the Associated Concepts: causality, correla- existence of subgroups and about whether tion, data, inferential statistics, multivariate there are ‘stars’ (members selected by a large analysis, sampling, variable analysis number of others) and ‘isolates’ (those not selected by anyone). Key Readings Evaluation Marsh, C. (1982) The Survey Method: The Contribution of Surveys to Sociological As with all such structured techniques, the Explanation. London: George Allen & wording of instructions is crucial to the valid- Unwin. ity of the inferences that are made. If the aim Rosenberg, M. (1968) The Logic of Survey is to find out which other individuals group Analysis. New York: Basic Books. members would most like to spend time with, they should not be asked which individuals they most admire. The people we admire are not necessarily the people with whom we would want to spend most time. SOCIOMETRY Victor Jupp Definition A theoretical and methodological approach Associated Concepts: community study which seeks to analyse relations between method, network analysis, snowball sampling

285 S-Jupp-3373.qxd 3/8/2006 5:52 PM Page 286

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

Key Readings whole urban systems. Observational data reveal the intensity and direction of flow of Bott, E. (1957) Family and Social Network. people through an area over time and this can London: Tavistock. then be plotted in map form and linked to Mitchell, J. C. (ed.) (1969) Social Networks in other geographical information systems (GIS) Urban Situations. Manchester: Manchester University Press. data to show, for example, why some areas Moreno, J. L. (1953) Who Shall Survive? flourish socially and economically while New York: Beacon. others decay, or why some areas suffer higher Scott, J. P. (1991) Social Network Analysis: A rates of crime than others. This type of analy- Handbook. London: Sage. sis purports to allow problems in layouts to be diagnosed, emphasizing the particular application of spatial research in the evaluation of design strategies in both new-build and remedial developments. SPATIAL RESEARCH This evaluative aspect of spatial research was evident in the research of Newman Definition (1973), which has inspired a great deal of interest in the relationship of the built envi- The exploration of physical and spatial fea- ronment and crime, but has also influenced tures of a particular environment which are public policy in the designing of spaces and associated with certain social outcomes, such places, such as housing areas, to prevent as social exclusion or crime, and are therefore crime. In his study of crime in high-rise hous- seen as having a contributory or perhaps ing in New York, Newman observed that causal effect on social processes or social predatory crimes of mugging, robbery and phenomena. rape tended to occur in spaces such as stair- wells, lift lobbies and deck accesses, which Distinctive Features were least inhabited, watched over or ‘owned’ by residents. As a result of Newman’s Spatial research has applications in subject research, many of the problems with the UK areas with a specific interest in the built envi- high-rise developments of the 1950s and ronment and its relationship with society, for 1960s, which became notorious for crime and example architecture and the way individual social and physical malaise, were attributed buildings and groups of buildings, conurba- to the way they had been designed. tions and even entire cities are designed to Moreover, Newman set a precedent for a reflect the culture of their users and inhabi- whole movement of research into the sorts of tants. Physical features are studied through spatial factors that were thought to be crim- various means, including observation of their inogenic, such as unlimited access through use and detailed analysis of spatial configura- housing areas, and which therefore should be tion in terms of other data, such as land use, designed out of schemes for new building movement patterns or crime figures. For development at the outset, and those thought example, the spatial patterning of offence and to prevent crime by invoking a sense of com- offender rates can be analysed by physically munity and territorial attitudes among resi- locating where offences occur relative to dents, such as inwardly focusing, defensible where offenders live. cul-de-sacs. Another example of spatial research is space syntax (Hillier and Hanson, 1983), an approach Evaluation that uses computer modelling of space to analyse factors such as patterns of human The fundamental question relating to spatial circulation in spaces ranging from individual research is to what extent the environment rooms in buildings to specific public spaces or shapes human behaviour and response as

286 S-Jupp-3373.qxd 3/8/2006 5:52 PM Page 287

SPSS

opposed to merely reflecting the social conditions and aspirations of its inhabitants. SPSS Although there is some scope for environmental or spatial determinism, the environment’s Definition influence on the social context is part of a One of the major computer packages for process of reciprocity (Hillier and Hanson, analysing quantitative data. 1984) in an extremely complex interrelation- ship. Attributing causal powers to space can Distinctive Features lead to the problem of over-determinism, which attributes human behaviour solely to SPSS stands for Statistical Package for the the influence of particular environmental con- Social Sciences, indicating its initial purpose; ditions and this leads to the questionable sug- more recently, however, it has been developed gestion that environmental modification will for business use. Since its first appearance it therefore effect human behavioural change, a has undergone considerable development and major criticism of Newman and his followers. new versions appear regularly, providing addi- For example, Coleman’s (1985) study of tional facilities. The most significant develop- deprived housing attempted to show that cer- ment was when the package became available tain housing design features led to social for desk-top computers. malaise by simply quantifying the simultane- SPSS comes as a series of modules; the ous occurrence of design features such as Base module is the minimum needed, and number of dwellings in a block and the num- Statistics, Advanced Statistics and other mod- ber of storeys, and malaise indicators such as ules are required for anything other than litter, graffiti, vandal damage and the number simple analytic procedures. of households with children in care. However, To use SPSS the researcher needs access to Coleman’s preoccupation with spatial indica- a computer that has the package installed and tors failed to account for the fact that social then has to create a data file containing the malaise could also have resulted from social figures to be analysed. This can be created in factors such as the concentration of problem a spreadsheet and imported into SPSS. The families through housing allocation policies. data file consists of a table in which the data for one case or respondent are entered in one row, with each column containing the data Graham Steventon for one variable. The analyses to be performed are selected from screen displays and Associated Concepts: causal model, GIS, the relevant commands can then be pasted prediction into a syntax or command file. Alternatively, the syntax file can be prepared on a word processor, but this requires considerable Key Readings familiarity with the SPSS command language. (It is not necessary to create a command file, Coleman, A. (1985) Utopia on Trial: Vision but there are huge benefits in doing so.) and Reality in Planned Housing. London: When the commands are run, the package Hilary Shipman. Hillier, B. and Hanson, J. (1984) The Social generates an output file containing the results Logic of Space. Cambridge: Cambridge of running those commands on the data in University Press. the data file. The contents can be transferred Newman, O. (1973) Defensible Space. to a word processor for further editing. London: Architectural Press. The analyses available depend on the Steventon, G. J. (1996) ‘Defensible space: a modules that have been installed, but include critical review of the theory and practice the common statistical tests of significance of a crime prevention strategy’, Urban as well as more complex procedures such as Design International, 1 (3): 235–45. factor analysis and multiple regression.

287 S-Jupp-3373.qxd 3/8/2006 5:52 PM Page 288

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

Evaluation share distinct experiences and that the ‘truth’ of that experience can be uncovered. The printout from SPSS often includes a lot of Researchers and theorists who support this material that is not needed, cannot be sup- approach include some feminists and some of pressed and has to be ignored. This means the those interested in the experience of disabil- user needs to know how to read the output to ity, race and ethnicity, and childhood. find the particular parts which are of interest. The SPSS manuals were exemplars of computer Distinctive Features manuals that required readers to be experts before they could understand them, and this Standpoint research is founded on founda- has led to a cottage industry in writing books to tionalist perspectives based on an insistence explain how to use the package (for example, that ‘truth exists independently of the Anthony, 1999; Foster, 2001; Kerr et al., 2002). knower’. This approach is political in that SPSS is used world wide and provides enor- experience is considered to be the starting mous data processing power. The dangers are point for any knowledge production. that users can access complex procedures Feminist Standpoint Epistemology, for while having little understanding of what they example, begins from the view that ‘masculine’ are doing and too easily trawl the data to see science is bad science because it excludes what might be thrown up. It is important that women’s experience and suggests the impor- the researcher should know what analyses are tance of developing a ‘successor science’ to wanted before using the package. existing dominant social science paradigms. Thus, Feminist Standpoint Epistemology starts from the position that the ‘personal is political’. Jeremy J. Foster Some suggest that this perspective is a development of Marxist ideas about the role of the Associated Concepts: analysis of variance proletariat and argue that women are an (ANOVA), causal model, correlation, discrimi- oppressed class and as such have the ability not nant function analysis, factor analysis, hunting only to understand their own experiences of (snooping and fishing), multivariate analysis, oppression but also to understand the experi- path analysis, quantitative research, regression ence of their oppressors. This implies that analysis, social survey research based on women’s experience provides a more valid basis for knowledge. It is not Key Readings just that the oppressed see more – their own Anthony, D. (1999) Understanding Advanced experience and that of the privileged – but also Statistics: A Guide for Nurses and Health that their knowledge emerges through the Care Professionals. Edinburgh: Churchill struggle against oppression: in this instance the Livingstone. struggle against men. Supporters of this Foster, J. J. (2001) Data Analysis using SPSS for approach suggest that objectivity is possible Windows versions 8 to 10. London: Sage. but that the critical scrutiny of all aspects of the Kerr, A. W., Hall, H. K. and Kozub, S. A. (2002) research process is necessary to achieve objec- Doing Statistics with SPSS. London: Sage. tivity. This presents a challenge to traditional notions of objectivity, which Harding (1993) argues are weak because the researchers’ own values, assumptions and so on are hidden. STANDPOINT RESEARCH Standpoint research is not the preserve of those concerned with women’s experience. Definition For example, some researchers working in the areas of disability, ethnicity, and race and An approach that starts with a focus on expe- childhood argue for a standpoint approach. rience, arguing that groups of individuals When searching for an epistemology based

288 S-Jupp-3373.qxd 3/8/2006 5:52 PM Page 289

STANDPOINT RESEARCH

on the experience of African American focus on biology that some supporters adopt. women, the values and ideas that African Some feminists leave traditional definitions of writers identify as being characteristically women unchallenged and resist instead the ‘black’ are often very similar to those claimed traditional value given to women and woman- by white feminist scholars as being charac- hood. However, many of the values ascribed teristically female. This suggests that the to women in this analysis originated in the material conditions of oppression can vary historical subordination of women – their dramatically and yet generate some unifor- association with nature, procreation, nurtu- mity in the epistemologies of subordinate rance and so on lies at the very heart of tra- groups (Hill Collins, 1989). Similarly, ditional sexist conceptions of womanhood. researchers working in the area of childhood Further, any standpoint position brings with have argued that as both women and children it the danger of viewing a group of people as are subject to patriarchy, those in power all the same. Not women, nor children, nor regard both groups as social problems and black people (and so on) are a homogeneous both groups find it hard to have their points group. We all occupy multiple, combined and of view heard and respected (Mayall, 2002). intersecting positions of gender, class, ethnicity, disability and sexuality. Evaluation Yet there are values of taking a standpoint approach. Feminist Standpoint Epistemology, One problem with standpoint research is that it for example, insists on viewing traditionally can imply that one group’s perspective is more defined female characteristics in a different real, more accurate and better than that of and much more positive way. In addition, others. But if we accept a position which taking a woman-centred approach, a child- implies that there is only one (real, accurate, centred approach, or an approach that fore- best) experience, this can only be built upon grounds the standpoint of people with the suppression of voices of persons with expe- disabilities when undertaking research pro- riences unlike those who are in a position vides a way of looking beyond individual to define. Further, the view that the more perspectives and enables us to challenge oppressed or more disadvantaged groups have stereotypical definitions. Furthermore, the greatest potential for knowledge implies Harding (1993) argues that the claim that the that the greater the oppression the broader or experience of an oppressed group provides a more inclusive one’s potential knowledge. This better starting point for thought is not about could lead to an unproductive discussion about arguing for one position but enables us to hierarchies of oppression: that is, those who are consider the significance of the experiences more oppressed (and how do we prove this any- of the powerful and the less powerful. way?) are potentially more knowledgeable. Recognizing the importance of difference Even if we find the most oppressed group of all, and yet acknowledging the significance of each how do we know that their way of seeing is the of the multiple identities that individuals ‘most true’: surely no one particular social loca- occupy, it is possible to argue for, standpoints tion has the complete access to truth. With all rather than a standpoint position. By doing this of this in mind, once we acknowledge the exis- we acknowledge that gender, age, disability and tence of several standpoints it becomes impos- so are significant identifiers of status and expe- sible to talk about ‘independent truth’ and rience whilst acknowledging that none of these ‘objectivity’ as a means of establishing superior is the only defining feature in anyone’s life. or ‘better’ knowledge because there will always be alternative knowledge claims arising from contextually grounded knowledge of different Gayle Letherby standpoints (Letherby, 2003). A further problem specifically in relation Associated Concepts: anti-racism research, to Feminist Standpoint Epistemology is the critical research, disability research,

289 S-Jupp-3373.qxd 3/8/2006 5:52 PM Page 290

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

emancipatory research, epistemology, population member has an equal chance of feminist research, relativism being selected. This is due to the method of selecting respondents. Key Readings Stratified random sampling has some advantages over simple random sampling, Harding, S. (1993) Rethinking standpoint where members of the population are num- epistemology: ‘what is strong objectiv- bered and then numbers chosen at random to ity?’, in L. Alcoff and E. Porter (eds), determine which individuals are included in Feminist Epistemologies. New York: the sample. There is no guarantee that a sim- Routledge. pp. 49–82. ple random sample will be representative of Hill Collins, P. (1989) ‘Black feminist the population – for example, from a popula- thought signs’, Journal of Women in Culture and Society, 14 (4): 745–73. tion where there are 500 men and 500 women, Letherby, G. (2003) Feminist Research in a sample of 100 may consist of 70 women and Theory and Practice. Milton Keynes: Open 30 men. University Press. In contrast, choosing a stratified random Mayall, B. (2002) Towards a Sociology for sample can ensure that the individuals Childhood. Oxford: Blackwell. selected are representative of the population in terms of certain variables. Using the above example again, an even gender balance in the sample of 100 could be ensured by taking a sample stratified by gender. The 1000 people STATISTICAL SIGNIFICANCE in the population would be listed with the 500 women first, followed by the 500 men. See Inferential Statistics Every tenth name would be chosen from the list, with a random form of selection used to determine whether the starting point was 1, 2, 3, 4, 5, 6, 7, 8, 9 or 10. If the randomly chosen starting point was 4 then the people STRATIFIED SAMPLING included in the sample would be those number 4, 14, 24, through to number 994. Definition This sample would include 50 women and 50 men. A method of sampling designed to ensure It is possible to stratify in terms of more that the sample has certain characteristics, than one variable. In the above example, the usually that it is representative of the popula- sample could have been stratified by gender tion on key variables. and employment status if the population was listed in the following order: women in Distinctive Features employment; women not in employment; men in employment; men not in employ- As with all sampling methods, stratified sam- ment. It is important to note that a stratified pling is used when there is insufficient time sample is still a random sample because the or resources to conduct a census (which col- researcher has no control over who is chosen. lects information from every member of a This method guarantees some characteristics population). One form of stratified sampling of the sample but not which individuals is quota sampling, where the population is appear in it. divided into strata and the number of sample members in each stratum is predetermined. Evaluation Quota samples can be representative of the population in terms of the stratifying vari- Stratified random sampling requires accurate ables but are not random because not every information about the key variables prior to

290 S-Jupp-3373.qxd 3/8/2006 5:52 PM Page 291

STRUCTURED OBSERVATION

the research being undertaken (for example, may entail just watching or listening. By the researcher would need to know which contrast, informal or casual observation is members of the population were in employ- unstructured, and may form the basis of ment and which were not). There is then future structured observation. Informal or considerable work involved in listing mem- casual observation methods are sometimes bers of the population in the correct order. seen as less objective than structured obser- However, once this task is completed, the vation, because the observer may be focusing sample can be selected quickly. on behaviours without a clear theoretical Stratified sampling cannot guarantee that framework, and may not be coding the very small groups are represented: in the behaviours in a reliable, that is, repeatable above example, if there were only four men way. The counter viewpoint is that a theoreti- who were not in employment, the likelihood cal framework can act as a strait-jacket that is that none of them would be included in the distorts reality. sample. It is also important to stratify by the most appropriate variable(s) for the research Distinctive Features study in question: in a study of attitudes to alcohol consumption, it may be more impor- Structured observation is used widely in the tant to have a sample that is representative in social sciences, from working on the behav- terms of age than one that is representative in iour of children and animals in the develop- terms of employment status. ment and use of play behaviours, to group decision making in adults, and the non-verbal behaviour of politicians (Gottman and Jamie Harding Bakeman, 1998; Martin and Bateson, 1993). Precisely because these kinds of behaviours Associated Concepts: bias, census, cluster can be subject to bias in recording, it is con- sampling, error, non-probability (non- sidered important to pre-code the behaviours random) sampling, probability (random) sam- to be observed. All the observers deployed in pling, quantitative research, quota sampling, the study will need to have a very clear under- sampling standing of the definitions of the behaviours to be coded, and to be sure of examples of those Key Readings behaviours. This ensures that the observations are valid, that is, that they measure what it is Mendenhall, W., Ott, L. and Scheaffer, R. L. intended to measure; and also that observation (1971) Elementary Survey Sampling. codings are reliable, that is, individuals coding Belmont, CA: Wadsworth Publishing. at different times, or different observers at the de Vaus, D. (2002) Surveys in Social Research, same time, record the same behaviours as 5th edn. London: Routledge. examples of the code in question. Where the behaviour being coded is laughter, or raising the right eyebrow, there tend to be few disagreements, but for other behaviours coding STRUCTURED OBSERVATION can be far more challenging. For example, when is an infant protruding its tongue, and when is it licking its lips? How still does the Definition tongue have to be, and how far from the lips, A systematic method of data collection, to be a tongue protrusion? This is an important where there is considerable pre-coding and question when infancy researchers are asking the observation takes the form of recording questions about infants’ abilities to mimic when, how often, or for how long the pre- adult facial gestures. coded behaviours occur. Observing usually Researchers using structured observation means watching and listening, although it techniques often prefer to have a permanent

291 S-Jupp-3373.qxd 3/8/2006 5:52 PM Page 292

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

or semi-permanent record of an observation relying on what people say they do. It allows session on film, videotape, or, now, digital researchers, under the appropriate condi- recording. This allows observers to play sam- tions, to find out what happens ‘in the real ples of behaviours over and over again, in world’. It is seen as an uncontaminated tech- order to agree that a particular behaviour has nique, provided that those being observed are occurred. This can be very time-consuming ceasing to be aware of this. and therefore it is necessary to take this into Apart from being very time consuming, account when planning and costing the inves- structured observation suffers from the prob- tigation or study. lem that it may exclude behaviours that could How do researchers measure behaviours? be of importance but are not catered for They may do this by counting and timing in the coding frame. Thus nothing new can be the incidence of a given behaviour. How reported in the findings, because by defini- researchers classify their observations will tion structured observation requires all influence how they can analyse their data behaviours of interest to be coded before later. There are four ways of classifying: nomi- observation begins. It is in this way that a nal or categorical measures, ordinal measures, theoretical framework can act as a distorting interval measures and ratio measures. An mechanism. There are other issues, for exam- example of nominal measures would be indi- ple, those being observed have no say in the cating qualitative differences, such as ‘walk- interpretation or coding of their behaviours. ing’ or ‘standing’; they would be counted as There is also the problem of inter-rater relia- incidents. Ordinal measures require that the bility: if different observers cannot agree on behaviour being observed can be classified in examples of behaviours, then what does this some sort of order, for example, level of posi- mean for research purposes: that the behav- tive facial expression, or level of aggression iours do not exist, or that some observers displayed; it too, illustrates qualitative differ- are much better at seeing examples than are ences. Interval measures in observation tend others? to be time duration, for example, length of time spent talking, two minutes being twice as long as one minute. Ratio measures are used Jeanette Garwood within observation when a time period is sampled, and the whole period is divided into the Associated Concepts: coding, interview, times in which the observed behaviours occur, measurement, participant observation, relia- or do not occur. bility, videography Narrative accounts of observations can also be produced, but these are more difficult to analyse, and tend to be analysed in a far Key Readings more qualitative fashion. A narrative record Gottman, J. M. and Bakeman, R. (1998) is similar to a transcript but differs in that it ‘Systematic observation techniques or requires more rigorous detail and informa- building life as an observational tion about the setting, time and participant researcher’, in Gilbert, D. T., Fiske, S. T. under study; it also requires that the observer and Gardner, L. (eds), Handbook of Social be uninvolved in the action, which is not Psychology, 4th edn. Oxford: Oxford the case when interviews are undertaken University Press. (Wilkinson, 1995). Martin, P. and Bateson, P. (1993) Measuring Behaviour: An Introductory Guide, 2nd edn. Evaluation Cambridge: Cambridge University Press. Wilkinson, J. (1995) ‘Direct observation’, Observation has the advantage of being in G. M. Breakwell, S. Hammond and direct, and of allowing researchers to see C. Fife-Schaw (eds), Research Methods in what happens at first hand, rather than Psychology. London: Sage. pp. 224–38.

292 S-Jupp-3373.qxd 3/8/2006 5:52 PM Page 293

SUBJECTIVITY

The clear lines of development in the SUBJECTIVITY understanding of subjectivity, from its Greek foundations forward, were broken in the Definition second half of the twentieth century as a The inner state of the self constituted by result of the debates that arose from develop- thinking, experience, emotion, belief, inten- ments in the philosophy of mind, and from tionality, self-awareness and the awareness of programmes of neuropsychological empirical others. research into the nature of consciousness. Gilbert Ryle’s timely argument was that sub- Distinctive Features jectivity, understood particularly as self- awareness, is an illusion, since the observing Speculation by philosophers on the nature of self aware of the experiencing self is not itself consciousness has a history stretching back at observable. His work suggested that the broad least to Aristotle (Plato did not really recog- understanding of subjectivity that had been nize a distinction between subjective and developed over centuries was basically in- objective), who thought that the forms of our coherent, and others, like Daniel Dennett, consciousness are derived in some way from came along to develop the basic theme. The the objective world, but also recognized the neuroscientific research programme had uniqueness of human consciousness to be come to be, at least in part, about demystify- able to receive or even create in some way ing subjectivity by re-describing states of these forms and use them for the develop- mind in terms of physically recordable states ment of scientific and moral understanding. of the central nervous system. Thomas The coming importance of the inner self is Nagel’s famous discussion of how we cannot clear in the fourth-century Christian teach- know what it is like to be a bat because we do ings of St Augustine, whose recognition, not have access to a bat’s subjective experi- recorded in Book 7 of his Confessions, that the ence, was responded to by Paul Churchland, a cause of evil resides in free will, occurs just die-hard physicalist, who pointed out that all prior to his conversion from Manichaeism to that means is that we lack the appropriate Christianity. Descartes increased the weight synaptic connections. Probably the clearest to be placed on inner life, in his First statement from a physicalist position, and by Meditation of 1641, when he declared cogito no means the entirely reductive one that ergo sum at the end of an inquiry that might be feared, is from David Chalmers, who declared the only certainty available to thinks it now possible that we are moving to humankind was to be found in subjective a genuinely scientific understanding of con- experience. A century and a half later, Kant’s scious experience, and regards the primary argument – summarized in the Prolegomena – empirical data for this science as first person that the idea of pure subjectivity makes no accounts of subjective states, for example sense, reaffirmed that subjective experience of hunger, memory, anger, the perception of has to be of something, and must be, if it is to colour, and so on. For Chalmers, a science of go beyond mere perception, framed by a gen- consciousness will not reduce first-person eral structure of understanding. From this data to statements about physical systems like point, the scope of subjectivity was effec- brain states, but it will explore the connec- tively widened to include the structures of tions. A scientific theory of subjectivity understanding, and this is the key move that would, he thinks, explain the relationships enables us to begin to understand the com- between the subjective and the physical. plexities of the philosophies of Verstehen Within the domain of late twentieth- through nineteenth-century German think- century social theory, the key discourses con- ing, and Edmund Husserl’s notion of tran- nected to subjectivity connect to the topics of scendental subjectivity as a basic component identity and the self. There is a structural reluc- of twentieth-century phenomenology. tance here even to conceive of a non-social

293 S-Jupp-3373.qxd 3/8/2006 5:52 PM Page 294

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

subjective condition. Some historic examples translation into the outer context of the are to be found in Durkheim’s notions of research, may become a complex area requiring anomie and egoism, and in the Hegelian careful analysis and public response – again notion of alienation, but these are far from leading us back to the paradox of subjectivity perfect examples of subjective conditions and social research, that it should not figure, uninflected by society. In general the debates but that it must be ventilated. on self and identity within sociology are concerned with the social component of subjective Evaluation experience and self-understanding. Freudian developments here are not an exception to The topic of subjectivity is controversial. Two this. From a strict sociological point of view thousand years of philosophical development the removal of the social from the subjective may seem to be coming to an end with the leaves behind an empty chasm. neuropsychologists’ promise of a science of The place of subjectivity with respect to consciousness, but, as can be seen with the sociological method is much more important. fundamental importance of questions of sub- The general issue concerns the relevance of jectivity in qualitative research, the notion of the researcher’s subjective position. In the fic- subjectivity is multi-faceted. The theoretical tional ideal situation the researcher should engagement of sociology with the concept have no presence in the research findings. of subjectivity is underdeveloped, but this is However, the increasingly strong position of compensated for by the increasingly well qualitative research in sociology means that highlighted issues around subjectivity and some account of the researcher’s assumptions, qualitative research. preferences, experiences and decisions during the research process may be indispensable. If Roy Boyne this can be provided in a succinct manner, then it can enable the meaning and perfor- Associated Concepts: constructionism, mance of the research to come into focus. epistemology, inter-subjective understanding, Breaking down the process into stages shows methodology, ontology, phenomenology, posi- the importance of taking account of subjective tivism, realism, Verstehen variables, especially in qualitative research. The opening stage requires choice of research topic, and this emerges out of an engagement Key Readings between research possibilities and discipline, Boyne, R. (2001) Subject, Society and Culture. beliefs, value commitments. The complex London: Sage. process leading to choice segues into questions Breuer, F. and Mruck, F. (2003) Subjectivity of method and data collection and it would be and Reflexivity in Qualitative Research. mistaken to think that subjective factors play Deutsche Forschungsgemeinschaft. no part here. Most importantly, the decision Shear, J. (ed.) (1997) Explaining Consciousness. about the form of data and their collection will Cambridge, MA: MIT Press. substantially determine the form of the inter- Taylor, C. (1989) Sources of the Self. actions that will take place between the Cambridge: Cambridge University Press. researcher and his or her subjects. Following the research engagement, the production of research outcomes will fix the range of subjective and intersubjective meanings that the SUMMATED RATING SCALE whole process will come to have for a range of related actors including the researcher, Definition research subjects and sponsoring organization. At each stage of the research, the question of One of the most frequently used methods the inner experience of the researcher, and its for assessment of people’s characteristics,

294 S-Jupp-3373.qxd 3/8/2006 5:52 PM Page 295

SUMMATED RATING SCALE

especially attitudes, in the social sciences. choices, the anchors for the response choices, Hundreds of summated rating scales have the underlying continuum for the rating been developed for the measurement of atti- scales (for example, agreement), and the tudes, beliefs, emotions, feelings, perceptions number of items and number of subscales and personality. They are widely used in both must be decided. An initial pool of items is social science applications and research. One written that is larger in number than is of the most popular is the Likert scale. needed for the final scale, because some items might not survive the item analysis conducted in the next step. Distinctive Features A good item is a concise and unambiguous There are four characteristics of a summated statement that reflects the intended construct. rating scale. First, the scale must contain two There are four rules to good item writing. First, or more items. Second, the format of each each item should express a single idea. item must be a rating scale. Third, the scale Multiple ideas cause confusion as the respon- must be designed to assess an underlying dent might agree with one part and disagree construct that is quantifiable. Fourth, the with the other. Second, avoid colloquialisms items do not have correct answers. and jargon. This is particularly important if the Summated rating scales are preferred to scale is to be used with multiple populations. single rating scales because they tend to have An expression that is clear to people in one better reliability. country, for example, might not be to someone As with a single rating scale, each item in in another, even if they all speak the same lan- a summated rating scale yields a numerical guage. Furthermore, word meanings can score. These are combined by adding the change over time. Third, consider the reading items to give a total scale score. There are level of potential respondents. Fourth, avoid four steps to developing a summated rating the use of negatives to reverse the meaning of scale. First, the construct is defined as care- an item. For example, the item ‘I do not like to fully as possible. Second, an initial pool of eat in fast food restaurants’ can be misread to items is written. Third, the scale is adminis- mean the opposite if the respondent fails to see tered to a development sample so that an the word ‘not’. item analysis can be conducted. Fourth, the Once the items are written, the scale scale is validated by conducting research to should be administered to a developmental support the interpretation of what scores on sample of at least 100–200 people. This step’s the scale represent. purpose is to select items that are related to Defining the construct carefully and com- one another so that the scale has internal con- pletely is a vital first step in scale construction. sistency and reliability. We assume that an It forms the foundation upon which subse- internally consistent scale has items that quent activities are based. This step is a con- assess the same construct. A statistical proce- ceptual/theoretical task of delineating exactly dure termed an item analysis will help select the nature of the construct of interest. A clear the best items from an internal consistency and specific definition is necessary in order for standpoint. Each item is correlated with a items to be written that will assess what is score computed as the sum of all other items intended. A clear definition can be derived minus the one being evaluated. This item- from prior research and theory in the domain remainder coefficient is used to select items of interest. Part of this effort should include for the final scale. The most frequently cho- the formulation of hypotheses about how the sen measure of internal consistency is coeffi- construct should relate to other constructs. cient alpha. A standard of 0.70 or higher for Once the nature of the construct is clearly acceptable reliability is widely used delineated, an initial test version of the scale (Nunnally, 1978). can be written. This requires first choosing a The final step in scale development is to format for the scale. The number of response collect construct validity evidence to support

295 S-Jupp-3373.qxd 3/8/2006 5:52 PM Page 296

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

the interpretation of the scale’s meaning. rarely have access to representative samples. This step is the most difficult as many social International norms are available for some science constructs are abstract and hypo- scales, but translation equivalence problems thetical with no objective manifestation. and cultural differences in language interpre- Validation relies on testing hypotheses about tation make comparisons difficult. how the scale should relate to other variables. With many scales, particularly those that assess attitudes, we often rely on face validity Paul E. Spector assuming that respondents interpret items in the same way as the scale developers. Validity Associated Concepts: composite measure- is not as much a characteristic of a scale as it ment, indicator, Likert scale, reliability, is the interpretation of the meaning of a scale. scaling, semantic differential technique, Thus, a scale can be found useful as a mea- Thurstone scale, validity of measurement sure of one construct but not another. Key Readings Evaluation DeVellis, R. F. (1991) Scale Development: For many scales it is important to collect nor- Theory and Applications. Newbury Park, mative data from various populations. CA: Sage. Ideally, the scale would be administered to Nunnally, J. C. (1978) Psychometric Theory, representative samples to yield reasonable 2nd edn. New York: McGraw–Hill. estimates of population mean scores. More Spector, P. E. (1992) Summated Rating Scale commonly, convenience samples are used for Construction: An Introduction. Newbury norms as researchers who develop scales Park, CA: Sage.

296 T-Jupp-3373.qxd 3/8/2006 10:45 AM Page 297

codes into meaningful sentences, so all other TEXTUAL ANALYSIS cultural signs – including those that are linguistically formulated – must be seen as Definition organized by their specific codes into narra- In social research the term refers to a method tive accounts and forms of discourse. From of analysing the contents of documents that this standpoint, the interpretation of the text uses qualitative procedures for assessing the contained within a document involves a significance of particular ideas or meanings decoding of the signs that were encoded in its in the document. It contrasts with content construction. analysis. The most influential semiotic writer in recent years has been Barthes, and his partic- Distinctive Features ular methods have been those that are most widely used in textual analysis. Barthes dis- Textual analysis is rooted in the hermeneutic tinguished the denotative meaning of a sign tradition of textual interpretation and has from its connotative meaning. The former most recently been organized around the par- refers to the meaning that a sign has by virtue ticular procedures implied by work in semi- of describing and interpreting the world by otics. Hermeneutic philosophy stresses that pointing to something and labelling it: the the interpretation of a text must always sign denotes a cultural object. The connota- be undertaken from the reader’s particular tive meaning of a sign consists of the range of standpoint. The inference of meaning is pos- associated meanings that it triggers in the sible only by relating the text to some other mind of the reader or audience: any particu- frame of reference and entering into a dia- lar denotation also has certain connotations. logue with the text. We must comprehend a A message that goes beyond mere denotation text by understanding the frame of reference and is structured around connotative mean- from which it was produced, and the ings to convey a particular message is, for researcher’s own frame of reference becomes Barthes, a ‘myth’, a term that he uses in pref- the springboard from which this becomes erence to the Marxian term ‘ideology’. possible. Semiotics (or semiology) adds to this A myth is a form of communication in which the claim that this dialogue involves a ‘decod- cultural meanings are structured in such a ing’ of the categories used by those who way as to convey a particular message to produced the text. those who see, hear, or read it. It is ‘distorted’ As formulated by Saussure, semiotics is by the demands of the underlying power rela- the study of cultural signs in terms of the cul- tions. Barthes’s analysis of myth is extended tural codes through which they are orga- to non-linguistic forms of discourse, such as nized. Linguistic signs (letters and words) are paintings, films, photographs and objects organized by grammatical and semantic of consumption. Any cultural object that T-Jupp-3373.qxd 3/8/2006 10:45 AM Page 298

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

conveys a message can, therefore, be subjected Barthes, R. ([1961–73] 1977) Images–Music– to a textual analysis. Text. London: Fontana. A principal task of textual analysis, there- Mannheim, K. ([1929] 1936) Ideology and fore, is to uncover power relations through dis- Utopia. London: Routledge & Kegan Paul. closing these distortions, demonstrating Rose, G. (2001) ‘Semiology’, in Visual whether any particular myth reflects the posi- Methodologies. London: Sage. tion of the oppressor or the oppressed. In this respect, Barthes’s work has a similar intention to that of Mannheim, also rooted in hermeneutics, whose sociology of knowledge contrasted THEORETICAL SAMPLING the ideological knowledge of the powerful and the utopian knowledge of the powerless. Definition Evaluation Theoretical sampling is a central part of the grounded theorizing advocated by Glaser and Because of its sensitivity to cultural context, Strauss (1967). Grounded theorizing is usu- textual analysis overcomes many of the limi- ally regarded as a form of qualitative inquiry, tations of content analysis. However, it does though in fact its originators saw it as applic- involve distinct problems of its own. Any able to quantitative data too. In this context, reading of a text reflects the standpoint of the theoretical sampling is tied to the purpose of reader, and semiotics has provided no clear generating and developing theoretical ideas, basis on which the validity of a particular rather than being aimed either at producing reading can be demonstrated. Barthes’s solu- findings that are representative of a population was to stress the criterion of coherence: tion or at testing hypotheses. if a reading of a text produces a coherent and systematic account, then it is, other things being equal, to be preferred. It is not clear, Distinctive Features however, why coherence should serve as such This form of sampling does not take place at a criterion. Indeed, Derrida and other post- a single point in the inquiry process but is structuralist writers have stressed that texts a recurrent feature: at various times the tend to be incoherent, to a greater or lesser researcher must ask what settings, events, extent, and they see the interpretative task as people etc. it would be worthwhile investi- involving the uncovering of such inconsisten- gating next in order to develop aspects of the cies and incoherencies. emerging theory. In this way, theoretical sampling is guided by, and helps to generate, the John Scott ‘theoretical sensitivity’ that is necessary in grounded theorizing, and indeed in qualitative and ethnographic work generally (Glaser, Associated Concepts: content analysis, cul- 1978; Strauss and Corbin, 1998). tural research, discourse analysis, document There is no fixed terminus to the sampling analysis, epistemology, hermeneutics, histori- of cases for theoretical purposes. Glaser and cal analysis, inter-subjective understanding, Strauss suggest that it should be stopped narrative analysis when what they call ‘theoretical saturation’ is reached; in other words, when looking at new cases is no longer contributing to devel- Key Readings opment of the theory. Judgement is involved Barthes, R. ([1957] 1973) Mythologies. here, but only in the same way that those London: Paladin. engaged in hypothesis-testing must decide Barthes, R. ([1964] 1967) Elements of Semiology. when to stop their work and conclude that a London: Cape. hypothesis is either true or false.

298 T-Jupp-3373.qxd 3/8/2006 10:45 AM Page 299

THEORETICAL SAMPLING

Theoretical sampling is a form of the them as practically distinct even if they can be comparative method, aimed at discovering sig- distinguished analytically. What is true is that nificant dimensions of variation in categories of there is usually a shift over the course of much social phenomena and their relations with one inquiry from a preoccupation with theory another. At different times it will be designed development towards an emphasis on testing either to minimize or to maximize the differ- out ideas; though a concern with testing is not ences on particular theoretical dimensions that absent at the start, neither does interest in have emerged as important in the course of theory development usually completely disap- inquiry. Minimization of differences allows pear even by the end. For this and other reasons, core features of theoretical categories to it is important to recognize that theoretical emerge more clearly. By contrast, maximizing sampling is not incompatible with other kinds the contrast on a particular dimension displays of sampling, including that concerned with the heterogeneity within the theoretical cate- subjecting theoretical ideas to test; and repre- gory, but also enables researchers to recognize sentative sampling may play an important part what is common across key categories. even in qualitative forms of inquiry (see Schofield, 1990). Evaluation Martyn Hammersley It is indisputable that there are useful forms of sampling that are different in character and purpose from representative sampling based Associated Concepts: analytic induction, on statistical theory. And the idea that sampling constant comparative method, grounded can be carried out in such a way as to develop theory, hunting (snooping and fishing), hypo- theory is surely a sound one. However, the con- thetico-deductive model, induction, probabil- cept of theoretical sampling is associated with ity (random) sampling, progressive focusing, a view about the nature of inquiry that raises a sampling number of questions (see Dey, 1999). One of these concerns the relationship between developing and testing theories, an issue over which Key Readings the originators of grounded theorizing later disagreed (see Glaser, 1992). Sometimes in The Dey, I. (1999) Grounding Grounded Theory: Discovery of Grounded Theory Glaser and Guidelines for Qualitative Inquiry. San Strauss suggest that theoretical sampling is Diego, CA: Academic Press. designed to test as well as to develop hypothe- Glaser, B. L. (1978) Theoretical Sensitivity. ses. At other points, they insist that a concern Mill Valley, CA: Sociology Press. with hypothesis testing, or ‘verification’, has no Glaser, B. L. (1992) Emergence versus Forcing: place in their approach or that it is a stage of Basics of Grounded Theory Analysis. Mill inquiry that can come after grounded theoriz- Valley, CA: Sociology Press. Glaser, B. L. and Strauss, A. L. (1967) The ing. Furthermore, theoretical sampling is very Discovery of Grounded Theory. Chicago: different from selecting a case, or cases, so as Aldine. to subject some theoretical idea to the sharpest Schofield, J. W. (1990) ‘Increasing the gener- possible test – what is sometimes referred to as alizability of qualitative research’, in critical case analysis. Within grounded theoriz- E. W. Eisner and A. Peshkin (eds), Qualitative ing discrepant evidence is treated as indicating Inquiry in Education: The Continuing a need further to elaborate the theoretical cate- Debate. New York: Teachers College gories, rather than as necessitating abandon- Press. pp. 200–25. ment of a theoretical idea in favour of one of its Strauss, A. and Corbin, J. (1998) Basics competitors. of Qualitative Research: Techniques and The development and testing of theories are Procedures for Developing Grounded Theory closely related activities and it is wrong to see 2nd edn. Thousand Oaks, CA: Sage.

299 T-Jupp-3373.qxd 3/8/2006 10:45 AM Page 300

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

Geertz contrasts his own approach with THICK DESCRIPTION Lévi-Strauss), and interpretation becomes thick description that must be deeply embedded in Definition the contextual richness of social life. The notion of ‘thick description’ is outlined by Clifford Geertz in the Interpretation of Culture, Evaluation which draws on the work of the philosopher Geertz is at his best when he employs thick Gilbert Ryle, and especially his winking anal- description to interpret ethnographic particu- ogy. Ryle directs us to distinguish between a lars; unlike the cognitivist approach which he boy whose eyelid twitches, a boy who is wink- sees as reductionistic and spuriously formal- ing and a boy who is parodying another boy’s istic, he finds these particularities in the rich- clumsy attempt at a wink. Drawing on these ness of real people in real life: a funeral, a seemingly silly examples, Ryle shows that these cockfight, a sheep theft. His thick descrip- similar behaviours were different because tions are not disembodied and decontextual- winks and fake-winks, parodies of winks, ized myths or customs, but humans engaging rehearsals of parodies of winks and so on com- in symbolic action. municate meaning while the twitch does not. His critics argue that it is easy to write a The point is that even such a seemingly simple text that is thick, but it is very difficult to form of communication – the rapid contraction write a text that is valid. The problem has not of an eyelid – produces multiple possibilities; escaped Geertz (1973). He admits that it is unravelling and identifying those contexts and difficult to fathom ‘what our informants are meanings requires ‘thick description’. up to and what it all means’. Moreover, they According to Geertz (1973), Ryle’s example pre- ask, is not thick description simply detailed sents an image of ‘the sort of piled-up structures documentation of events and circumstances? of inference and implication through which an As such it is, and for long has been, the cor- ethnographer is continually trying to pick his nerstone of much anthropology whatever way’. Raw field notes are readable, but to its theoretical framework. In spite of these understand them requires thick description. and other critiques, Geertz’s use of thick descriptions, his evocation of the metaphor of Distinctive Features the interpretation of reading culture as a text, ‘has led to the present dominant interest One avenue of approach to culture, related to within interpretive anthropology about how but distinct from both the American cognitivist interpretations are constructed by the anthro- and continental structuralist approaches, has pologist, who works in turn from the inter- been to treat cultures as systems of shared pretations of his informants’ (Marcus and symbols and meanings. Geertz has been one Fischer, 1986: 26). That key point has brought of the most notable pioneers of this view. about a sea change in the discipline. Borrowing from Ricoeur a broader sense of ‘text’, Geertz has treated a culture as ‘an assem- blage of texts’. Paul Valentine

Doing ethnography is like trying to read (in the Associated Concepts: cultural research, epis- sense of ‘construct a reading of’) a manuscript – temology, ethnography, idiographic, ontology, foreign, faded, full of ellipses, incoherencies, subjectivity suspicious emendations, and tendentious commentaries, but written not in conventionalized graphs of sound but in transient examples of Key Readings shaped behavior. (1973: 10) Geertz, C. (1973) The Interpretation of Anthropology thus becomes a matter of Culture: Selected Essays. New York: Basic interpretation, not decipherment (in this, Books.

300 T-Jupp-3373.qxd 3/8/2006 10:45 AM Page 301

TIME SERIES DESIGN

Inglis, F. (2000) Clifford Geertz: Culture, simply have to indicate with which statements Custom and Ethics. Cambridge: Polity they agree. The criticism of the use of the Press. scale is that the characteristics of the judges Marcus, G. and Fischer, M. (1986) Anthropo- and the underlying dimensions of their atti- logy as Cultural Critique: An Experimental tudes may be significantly different from Movement in the Human Sciences. Chicago: those of the respondents and yet the former University of Chicago Press. determine the construction of the scale Shankman, P. (1984) ‘The thick and the thin: which is used to measure the attitudes of the on the interpretive theoretical program respondents. of Clifford Geertz’, Current Anthropology, 25 (3): 261–80. Victor Jupp

Associated Concepts: composite measure- THURSTONE SCALE ment, indicator, Likert scale, scaling, semantic differential technique, summated rating scale Definition Key Readings An attitude scale with equal-appearing intervals. It was developed by Thurstone in 1928 Anastasi, A. (1988) Psychological Testing, 6th to measure attitudes towards the church in edn. New York: Macmillan. the United States. Guildford, J. P. (1956) Psychometric Methods. New York: McGraw–Hill. Distinctive Features Schuman, H. and Presser, S. (1996) Questions and Answers in Attitude Surveys. Thurstone scaling starts by generating an item Thousand Oaks, CA: Sage. pool comprising approximately one hundred Thurstone, L. L. (1928) ‘Attitudes can be statements, for example in terms of measured’, American Journal of Sociology, Thurstone’s research ‘I think that the church 33: 529–54. represents everything that is good in society’. A group of ‘judges’ (not in a judicial sense) is selected on grounds of representativeness and typicality. Each judge is asked to sort the state- TIME SERIES DESIGN ments into a set number of piles (for example seven) according to their favourableness to the Definition attitude in question (for example, attitude towards the church). The piles are sorted as if A research design in which measurements of they are equally spaced, say from 1 to 7. The the same variables are taken at different researcher retains those statements on which points in time, often with a view to studying the judges agree regarding the allocation to social trends. For this reason such designs are piles. These are then embodied in a question- sometimes also known as trend designs and naire, in random order, and respondents are are distinguishable from ‘one shot’ cross- asked to agree or disagree with them. The scale sectional designs in which measurements are score for any individual is the average of all of taken only once. the statements with which he or she agrees. Distinctive Features Evaluation Time series designs can be used in conjunction The advantage of a Thurstone scale is that the with official data, for example by plotting procedure is simple insofar as it requires no crime rates for the same area but for different scoring or judging of distances. Respondents point in time (monthly, quarterly, annually).

301 T-Jupp-3373.qxd 3/8/2006 10:45 AM Page 302

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

This acts as a basis for making statements victim surveys such as the BCS are descriptive about trends in levels of crime. It is possible to studies that measure crime at particular points plot the trends for different variables at one in time, and look for trends, but have no great and the same time with a view to making claims to providing explanations. There is the inferences about their relationship, for exam- potential, as with all trend designs, to theorize ple, to map unemployment rates against crime about why changes in trends have occurred, rates for England and Wales, 1990–2000, to but surveys do not by themselves provide suf- consider whether changes in one rate coin- ficient evidence as to causality. cide with changes in the other. Although such The value of time series surveys is that they analysis may provide evidence of association, can be used to map changes and trends in soci- that is, that changes in crime rates are corre- ety (or a sub-section of it). However, as indi- lated with changes in unemployment rates, it cated earlier, they are based on equivalent not does not by itself provide a sufficient basis for identical samples, in terms of including the making inferences about causality (that is, that same individuals. Therefore they should not be changes in crime rates are caused by changes used as a basis for making inferences about in unemployment rates). individual development (for example, develop- The term time series is sometimes also used ment of criminal careers). For this some variant to refer to a form of survey design in which of a longitudinal cohort survey is required. equivalent samples of the population are taken at different points of time and data collected about them, usually by interviewing sample Victor Jupp members. The samples are equivalent because they are collected by the same principles and Associated Concepts: causality, cohort study, using the same criteria but that does not cross-sectional survey, longitudinal study, mean that the same individuals are selected one-shot design, panel study (although in statistical terms that is a theoretical possibility). The British Crime Survey, car- Key Readings ried out under the auspices of the Home Goldstein, H. (1979) The Design and Analysis Office, is an example of a time series sample of Longitudinal Studies: Their Role in the survey. The first BCS was carried out in 1982 Measurement of Change. New York: and has been repeated at regular (usually Academic Press. 4-year) intervals. Although there are variations, Magnussen, D. (ed.) (1991) Problems and at each point in time sample members are Methods in Longitudinal Research: Stability asked to specify the type of crime and also and Change. Cambridge: Cambridge asked to indicate whether the crime was University Press. reported to the police. Adjustments are made to Magnussen, D. and Bergman, L. R. eds (1990) the number of crimes reported by sample Data Quality in Longitudinal Research. members to facilitate comparisons with crime Cambridge: Cambridge University Press. officially recorded in the publication Criminal de Vaus, D. (2001) Research Design in Social Statistics. In this way an estimate of the ‘dark Research. London: Sage. figure’ of unrecorded crime is obtained. The BCS also provides evidence of trends in crime levels and because of recognized deficiencies in official recorded data it is TRACE MEASURES usually viewed as the more reliable indicator. Definition Evaluation Alternative methods by which social activities Time series designs are predominantly descrip- or phenomena can be examined and quanti- tive rather than explanatory. For example, fied in non-reactive or unobtrusive ways.

302 T-Jupp-3373.qxd 3/8/2006 10:45 AM Page 303

TRACE MEASURES

Distinctive Features have suggested ‘sneaky measurement of crime and disorder’ and have discussed several exam- Trace measures are usually physical traces of ples from this area of social research. Some of past behaviour often synonymous with unob- their suggestions as regards modern mea- trusive measures and non-reactive measures, sures or traces of crime and disorder repre- all of which refer to alternative and often sent the realities of the street, store and club. imaginative units of measurement. According Thus signs of disorder and incivility might to Webb et al. (1966), who first documented include an Alcopop bottle in a town centre the use of unobtrusive measures and non- litter bin, shards of glass outside a specific reactive research in the social sciences, these nightclub and used hypodermic syringes in a forms of data amount to proxies which ‘are public park. not specifically produced for the purposes of comparison and inference, but are available Evaluation to be exploited opportunistically by the alert investigator’ (Webb et al., 1966: 36). As they are not specifically produced as Trace measures are therefore approximations social scientific data, there are several key to knowledge and might include hidden advantages of physical evidence data. They observation, contrived observation, trace are not only relatively easy methods of deter- analysis and secondary records. mining long-term change at low cost but their Distinctively, trace measures, as vestiges or very inconspicuousness as units of measure- remaining marks of the presence or existence of ment ensures they are relatively free from something, are often cheap and simple indices bias; in other words they are ‘free of reactive or signs that can be used as ways of closely measurement effects’ (Webb et al., 2000: 50). monitoring social activities. There are two Furthermore, gathering data and evidence by main types of physical traces of past behaviour, means of physical traces often involves imag- these being natural erosion measures and inative and unconventional approaches to natural accretion measures. Where physical research and can make available data in con- traces manifest themselves as wear and tear of tent areas where verbal reports are otherwise materials, these are erosion measures. Where unavailable, invisible or unreliable. Thus they are deposits of materials such as debris, they can be an asset in terms of accessing inscriptions, marks, remnants or other indica- some hard-to-reach populations or groups. tors, these are natural accretion measures. Whilst there are clear advantages to the use Unobtrusive measures, or trace measures, of trace measures in the social sciences there can be valuable sources of data and evidence are perhaps greater advantages in employing in the social sciences. No matter how sophis- them alongside, or in consort with, more tradi- ticated and expensive the methods of mea- tional methods. In most cases, trace measures surements adopted by the researcher, they are simply signs or indices of activity or behav- are often criticized for being insufficiently iours, approximations to knowledge that sensitive to measure with enough precision cannot usually be taken alone as evidence of an and detail, especially in terms of time, place actual state of affairs or as absolute evidence. and circumstance. Take, for example, the con- In some contexts, and when used alone with- text of crime and disorder. Despite the ‘data out triangulation of methods or combining explosion’ in terms of information derived with supplementary obtrusive and reactive from both police crime figures and victimiza- techniques, physical evidence data may remain tion surveys, some critics claim that specific largely anecdotal. There are other specific local patterns, trends and demographics of reasons to be cautious of the use of trace mea- crime and disorder remain insufficiently sures, including the problem of selective sur- detailed for focused action to be based upon vival of materials over time and across space them (Garwood et al., 2002). In order to affecting the ability to generalize findings as generate this level of detail, these authors well as the risk of there being a high ‘dross

303 T-Jupp-3373.qxd 3/8/2006 10:45 AM Page 304

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

rate’. Finally, for some there remain significant The most influential technical system for ethical problems with the use of trace mea- social (rather than linguistic) research has been sures. These ethical dilemmas are similar to developed by Gail Jefferson for Conversation those that arise when conducting covert or hid- Analysis (CA). It is designed to support the den and unobtrusive research more generally, examination of talk as social interaction; to namely collecting data about individuals with- include the features that participants demon- out their knowledge and consent. strably use and treat as significant; and to display the grounds for analytical claims. The system shows the precise onset and Pamela Davies ending of overlapping talk, timed pauses, and intonational features including emphasis, Associated Concepts: covert research, indi- elongation, cut-offs and pitch movement. cator, spatial research, triangulation, unob- Here is a short segment of a telephone con- trusive measures, validity of measurement versation transcribed by Jefferson:

Key Readings Emma: uh HONEY I’LL PRAY FOR EV’RYBODY= Lottie: [=Alri:ght,] Garwood, J., Rogerson, M. and Pease, K. Emma: [=I:- I ] don’t kno:w,hh this: [uh w]orld (2002) ‘Sneaky measurement of crime and goes o::n= disorder’, in V. Jupp, P. Davies and Lottie: [Yeh.] P. Francis (eds), Doing Criminological Emma: =we have to keep ↓goin’ do[n’t we.] Research. London: Sage. pp. 157–67. Lottie: [Ye:ah, ] Webb, E. J., Campbell, D. T., Schwartz, (.) R. D. and Sechrest, L. (1966) Unobtrusive Lottie: [U h h u h ] Measures. Chicago: Rand McNally. Emma: [D’you feel h]a:ppy toda:y? Webb, E. J., Campbell, D. T., Schwartz, R. (0.4) ↑ D. and Sechrest, L. (2000) Unobtrusive Lottie: Ye:ah. Measures, rev. edn. London: Sage. Emma: Good.

Square brackets mark overlaps; colons signal elongation; underlining shows emphasis; TRANSCRIPTION and the arrows mark pitch rise and fall. Familiar punctuation marks are re-defined to signify intonation or speech delivery rather Definition than grammar. For example, periods (full stops) The rendering of recorded talk into a stan- signal a falling, ‘terminal contour’, and ques- dard written form. tion marks signal a rising ‘questioning’ contour, irrespective of grammatical criteria such as Distinctive Features sentence boundaries and interrogative syntax. There are many other features in the system. Transcription of recorded talk is done to facili- The initial difficulty of reading CA tran- tate analysis, where the choice of one system scripts is largely due to our familiarity with over another, or of a level of detail within conventional written text. But that familiarity a system, depends on the researcher’s analytic should not be mistaken for neutrality; ordi- perspective. In social research, talk is often nary typescript is not a straightforward way transcribed non-technically, often by secretarial of representing talk. Rather, it transforms a assistants, so that researchers can work through lived medium of social interaction into a form materials such as interviews, looking for ‘con- that was neither invented nor designed tent’ (ideas, arguments, etc.) of various kinds. for transcription. There is a large literature Technical transcripts include additional linguis- on the history, social functions, uses and tic intonational and interactional features. effects of written text that warns against its

304 T-Jupp-3373.qxd 3/8/2006 10:45 AM Page 305

TRIANGULATION

ostensible neutrality. Jefferson’s system Key Readings encourages and permits analysis of talk as Atkinson, J. M. and Heritage, J. (1999) situated social action rather than, say, a ‘Jefferson’s transcript notation’, in decontextualized record of someone’s ideas. A. Jaworski and N. Coupland, (eds), The Discourse Reader. London: Routledge. Evaluation pp. 158–66. Producing technical transcripts is time- Edwards, J. A. and Lampert, M. D. (eds) (1993) Talking Data: Transcription and consuming, and requires practice and train- Coding in Discourse Research. Hillsdale, ing. It also goes with an approach to language NJ: Lawrence Erlbaum. materials, including interviews, that many Goodwin, C. (2000) ‘Gesture, aphasia and researchers may not wish to adopt, which interaction’, in D. McNeill (ed.), Language treats them as interaction prior to any status and Gesture. Cambridge: Cambridge they might have as collections of facts, infor- University Press. pp. 84–98. mation and opinions. One general advantage Have, P. (1999) Doing Conversation Analysis: is the attention that has to be paid to talk’s A Practical Guide. London: Sage. details; conventional transcripts are often Psathas, G. and Anderson, T. (1990) ‘The surprisingly inaccurate. “practices” of transcription in conversa- Although Jefferson’s system reflects a keen tion analysis’, Semiotica, 78: 75–99. interest in the precision of conversational turn-taking, it is wrong to imagine that the system is limited to that interest. Primarily, it renders talk as social action. Features such as pauses, overlaps and detailed self-corrections TRIANGULATION display the status of talk’s content as sensitive to interactional considerations and to speak- Definition ers’ concurrent activities. It is not a matter of In social research, the term is used to refer to choosing whether to use a transcription sys- the observation of the research issue from (at tem or not. Conventional typescript imposes least) two different points. This understand- alterations and omissions that are easily over- ing of the term is used in qualitative as well looked, but that transform talk into a form as quantitative research and in the context of that distracts attention from its status as social combining both. activity. No transcription system is perfect or all- purpose. But alterations have to be slow and Distinctive Features gradual, because standardization is crucial. Triangulation is most often equated with apply- Contemporary work (for example, Goodwin, ing different methodological approaches. 2000) is extending the system for the kinds Earlier discussions about non-reactive mea- of non-linguistic features, captured in video surement and of the multi-trait–multi-method recordings, that were less significant in the matrix were the starting points for developing telephone conversations that provided much this idea of triangulation. Initially, triangulation of the material for the system’s initial was understood as a validation strategy. But in development. a broader understanding, four different forms are distinguished. Triangulation of data com- Derek Edwards bines data drawn from different sources and at different times, in different places or from dif- Associated Concepts: conversation analysis, ferent people. Investigator triangulation is char- discursive psychology, epistemology, ethno- acterized by the use of different observers or methodology, narrative analysis, naturalistic interviewers, to balance out the subjective data influences of individuals. Triangulation of

305 T-Jupp-3373.qxd 3/8/2006 10:45 AM Page 306

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

theories means to approach data from different Evaluation theoretical angles, which are used side by side Triangulation has been criticized for several to assess their usefulness. Most often, triangu- reasons. The idea to use it as a simple form of lation is seen as methodological triangulation, validation has been questioned, as was the either ‘within-method’ or ‘between-method’. mere and simple combination of methods The intention here was originally to maximize without taking differences in theoretical the validity of research by playing the methods backgrounds and ways of constructing the off against each other. issue under study into account. Examples of within-method triangulation are The idea of a ‘systematic triangulation of the use of different sub-scales within a ques- perspectives’ (Flick, 1992) tries to address tionnaire or the episodic interview (Flick, 2002). both points. Here different research perspec- Here, some research issue is explored by means tives within qualitative research are com- of invitations to narrate, focusing on experi- bined with one another in a targeted way, to ences in concrete situations. These narratives complement their strong points and to illus- are combined with questions that focus more trate their respective limitations. So the com- on definitions and general answers. Thus an bination of methods is contextualized in the attempt is made in such an interview systemat- combination of theoretical and research ically to unite the methodological approaches of perspectives. the semi-structured interview and the narrative, Triangulation is currently also being used using their respective strengths. in the debate about the relationship between Between-method triangulation is most qualitative and quantitative research. If the strongly associated with the keyword triangu- concept of triangulation is taken seriously, it lation. For example, combination of different is characteristic of all of these variants that (qualitative or quantitative) procedures is pro- they see the procedures they combine as posed in order to transcend the boundaries of being of equal value and that they do not both methodological approaches. Moreover, begin by regarding one procedure as central this makes it possible to capture different and the others as preliminary or illustrative. aspects of the research issue – such as con- We therefore have three modes of application crete professional routines in observations for triangulation: as a validation strategy, as and practitioners’ knowledge of their own an approach to the generalization of discover- modes of action and routines via interviews. ies, and as a route to additional knowledge. The most consistent variant is to apply the triangulated methods to the same cases: the persons being observed in a particular field are (all) Uwe Flick interviewed. This makes a case-related analysis of both types of data possible. In addition, com- Associated Concepts: analytic induction, parisons can also be undertaken at a higher data, methodology, mixed-methods research, level: patterns that emerge from analysing one qualitative research, trace measures, validity type of data can be set against patterns in the other type of data. However, this increases the load for an individual participant in a study Key Readings and the danger of dropouts rises markedly. Denzin, N. K. (1978/1989) The Research Act: Everyone who refuses either to provide an A Theoretical Introduction to Sociological interview or to be observed is ‘lost’ to the entire Methods. New York: McGraw–Hill (2nd investigation. In such cases, triangulation edn)/Englewood Cliffs, NJ: Prentice Hall should be implemented at the level of data sets. (3rd edn). Each method is applied independently of any Flick, Uwe (1992) ‘Triangulation revisited– other. Both data sets resulting from this are strategy of or alternative to validation of analysed first. Triangulation then relates the qualitative data’, Journal for the Theory of results of both analyses and to each other. Social Behavior, 22: 175–97.

306 T-Jupp-3373.qxd 3/8/2006 10:45 AM Page 307

TYPE 1 AND TYPE 2 ERRORS

Flick, Uwe (2002) An Introduction to underpinnings of the social sciences and Qualitative Research, 2nd edn. London: Sage. recognition that, under certain conditions, Flick, Uwe (2004) ‘Triangulation’, in U. Flick, false-positive findings can be dangerous. E. von Kardorff and I. Steinke (eds), Therefore, focus is placed upon the null Qualitative Research – Paradigms, Theories, hypothesis, which typically assumes that Methods, Practice and Contexts. London: there is no relationship between two vari- Sage. ables. The counter to the null hypothesis is the research hypothesis, which has some degree of theoretical justification in suggesting a relationship between two variables. In TYPE 1 AND TYPE 2 ERRORS order to minimize the probability of committing a Type 1 error, we retain the null hypothesis unless the significance test of a Definition relationship is typically at the level of 0.05 or The category of error associated with the below. This results in a probability of 95 per incorrect retention of a hypothesis. Error cent that the observed relationship between results when inference testing of a sample two variables represents a real relationship in population suggests the retention of the the universe, and not a function of chance in hypothesis, when the unobserved true value the sample population. Over time, a social present in the universe from which the sample scientist employing a 0.05 significance level was drawn would call for the rejection of the for the rejection of the null hypothesis will hypothesis. A Type 1 error occurs when the commit a Type 1 error 5 per cent of the time. null hypothesis is falsely rejected, while a Type 2 error occurs when the research hypoth- Evaluation esis is falsely rejected (thus wrongly retaining the null hypothesis). Essentially, Type 1 errors The probabilities of committing a Type 1 and are ‘false-positive’ findings, while Type 2 Type 2 error are inversely related, although the errors are ‘false-negative’ findings. exact probability of committing a Type 2 error is unknowable. The more conservative the Distinctive Features approach in deciding whether or not to reject the null hypothesis necessarily results in a The logic behind Type 1 and 2 errors rests on likely higher probability of committing a Type the underlying nature of random probability 2 error. While the level of significance used to sampling. As a sample is at best representative determine retention or rejection of the null is of the larger universe under study, the decision necessarily arbitrary, convention in the social whether or not to retain a hypothesis is based sciences suggests a minimum level of signifi- on an estimation known as a test of significance appropriate for the rejection of the null cance. As sampling makes it impossible to hypothesis at 0.05. It is also common to report know the precise nature of the universe popu- significance levels of 0.01 and 0.001, which lation under study, significance testing is further reduces the probability of committing a unable to offer absolute confirmation that a Type 1 error to 1 per cent or 0.1 per cent, hypothesis should be retained or rejected, but respectively, with the concomitant increased rather only a probability that we have commit- probability of resulting in a Type 2 error. ted an error in retention or rejection. A signifi- A common example used to illustrate the cance test results in the probability that an difference between Type 1 and 2 errors is the association observed in the data results from presumption of innocence in criminal trials. chance in the sample rather than a revealed A defendant in a criminal trial enjoys the pre- true relationship in the universe population. sumption of innocence, while the state prose- Hypothesis testing is an inherently conserva- cutor has the responsibility to prove ‘beyond tive practice. This results from the philosophical a shadow of doubt’ that the defendant is

307 T-Jupp-3373.qxd 3/8/2006 10:45 AM Page 308

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

guilty. If an innocent citizen is found guilty of Key Readings a crime, a Type 1 false-positive error is com- David, M. and Sutton, C. (2004) Social mitted. If a guilty person is found innocent, a Research: The Basics. London: Sage. Type 2 error is committed. Nachmias, C. F. and Nachmias, D. (1996) Research Methods in the Social Sciences, 5th David Brockington edn. New York: St Martin’s Press.

Associated Concepts: error, falsification, hypothesis, inferential statistics, probability (random) sampling

308 U-Jupp-3373.qxd 3/8/2006 10:45 AM Page 309

risk. An example is that of an investigation of UNOBTRUSIVE MEASURES the levels and patterns of drug use in the community. Police data are severely limited Definition as this constitutes arrest data, and arrests are Measures that concentrate on the traces of subject to police activity and ‘special’ opera- human action, on the evidence of human tions. Therefore the data will misrepresent activity left in the environment. Traces need certain types of people or drug types that are not be incidental, they can be intentionally less likely to come to police attention. Data created, such as graffiti. from drug treatment agencies represent only those whose use has become problematic and have accessed services. Further, it is widely Distinctive Features documented now that few drug users can Although some disciplines are reliant on obtain services, or even try to obtain them, these methods, such as archaeology, in the even when needed. A survey of drug users social sciences the methods were not given a would be subject to response bias with framework until 1966 when Webb, Campbell, respondents either playing down or exag- Schwartz and Sechrest (1966) published gerating their substance use, according to Unobtrusive Measures: Nonreactive Research in their own motives. Unobtrusive measures of the Social Sciences (updated in 1981 as Non- drug use, including disposed or discarded reactive Measures in the Social Sciences). drug paraphernalia and hepatitis ‘C’ rates, Unobtrusive measures can be used within can identify patterns, which may confirm or an experimental framework, for example, complement trends in administrative statis- within the ‘lost letters’ paradigm. Merritt and tics, while at the same time avoiding the Fowler (1948) wanted to know about public problems associated with reactivity, invasion honesty, and to test this they put letters in of privacy and ethical principles. envelopes with addresses, and in some they It is possible to use other kinds of unobtru- placed lead weights the size of 50 cents sive measures, such as minutes of public meet- pieces. The independent measures were the ings, or management committee meetings in numbers of letters of each kind mailed back. schools or other public organizations. Unobtrusive measures can also be used to Although this is commonly called archival or investigate much more sensitive matters, document-based research, it is also a type of such as drug use of certain kinds. The goal in unobtrusive measure. The main principle all of these cases has been to study behav- behind the method is that the data should iours that are very likely to be subject to already be in existence, or very easy to obtain. response bias under more direct forms of Our personal experience is that asking public measurement, for example direct question- authorities for copies of information they ing, or where researchers might be put at already hold can make information suddenly U-Jupp-3373.qxd 3/8/2006 10:45 AM Page 310

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

very sensitive. Therefore it is wise to find girls’ dormitory may have less to do with unobtrusive measures that researchers can eating problems or worries about weight con- observe or collect for themselves. trol, and more to do with sales promotions The kinds of unobtrusive data suggested and free gifts. It is important to eliminate the here are all amenable to quantitative analysis, unlikely explanation, before assuming that and with care, inferential statistics. unobtrusive measures are signs of particular kinds of behaviour. Finally, although unob- Evaluation trusive measures research may avoid certain ethical problems it can also raise others. For Unobtrusive measurement has several crucial example, analysing the contents of residents, advantages over the alternatives. It is normally garbage cans may indicate forms of lifestyle cheap, it is not liable to response bias and is but can also contravene the principle of seldom attached to any single individual, informed consent. reducing ethical problems associated with invasion of privacy. Such data cannot always be conclusive when taken alone, but will be Jeanette Garwood and Michelle Rogerson suggestive and are best gathered using the principle of triangulation, where several Associated Concepts: content analysis, kinds of data are collected in order to address diary, Internet research, naturalistic data, the same question. photography, trace measures, triangulation, The advantages include the non-reactivity validy of measurement of the methods: since the performers are not aware of being part of an investigation, they do not behave as if they are being observed. Key Readings They provide cross-validation of mainstream methods. They encourage researchers to look Lee, R. M. (2000) Unobtrusive Measures in Social Research. Milton Keynes: Open at their research questions afresh, and consider University Press. new ways of investigation. Therefore they pro- Merritt, C. B. and Fowler, R. G. (1948) ‘The vide a spur to ingenuity and creativity in pecuniary honesty of the public at large’, research. Journal of Abnormal and Social Psychology, The disadvantages of the methods, some 43: 90–3. suggest, include the lack of identity of the Webb, E. J., Campbell, D. T., Schwartz, R. person providing the data. Also, there are D. and Sechrest, L. (1966) Unobtrusive problems of making inferences about the Measures: Non-reactive Research in the level of incidence. For example, one person Social Sciences. Chicago: Rand McNally. may be responsible for most of the graffiti in Webb, E. J., Campbell, D. T., Schwartz, R. an area. Also, the apparent level of cause and D., Sechrest, L. and Grove, J. B. (1981) effect may be mediated by other influences. Non-reactive Measures in the Social For example, consumption of diet foods in a Sciences. Boston, MA: Houghton Mifflin.

310 V-Jupp-3373.qxd 3/8/2006 5:53 PM Page 311

time periods, cultural, social and political VALIDITY frameworks), including phenomenology, constructionism, hermeneutics, critical research Definition and postmodernism. The extent to which conclusions drawn from research provide an accurate description of Victor Jupp what happened or a correct explanation of what happens and why. Associated Concepts: realism, relativism, reliability, validity of explanation, validity of Distinctive Features generalization, validity of measurement The assessment of the overall validity of conclusions drawn from a research project can Key Readings be made by addressing three aspects. First, validity of measurement involves asking Bryman, A. (2004) Social Research Methods, whether a research instrument, for example a 2nd edn. Oxford: Oxford University Press. questionnaire, measures what it purports to. Smith, H. W. (1975) Strategies of Social Second, validity of explanation (sometimes Research: The Methodological Imagination. also known as internal validity) involves ask- Englewood Cliffs, NJ: Prentice Hall. ing whether the explanations and conclusions derived from research are the correct ones for the specific subjects or contexts that have been studied. Third, validity of generalization VALIDITY OF EXPLANATION (sometimes referred to as external validity) involves asking whether the conclusions Definition drawn from a particular study can be generalized to other people (population validity) The extent to which an explanation of how and other contexts (ecological validity). and why some social phenomenon occurs is the correct one. It is sometimes known as Evaluation internal validity.

Validity is associated with the notions of Distinctive Features truth and of realism (the philosophical viewpoint that there is some social reality which Most social research involves studying a can be studied, described and explained). sample of cases (for example, individuals) However, validity does not sit easily within in specific contexts, with a view to generaliz- forms of relativism (the philosophical viewing beyond these. Such generalization is only point that knowledge and truth are relative to sound if the conclusions drawn from the V-Jupp-3373.qxd 3/8/2006 5:53 PM Page 312

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

study of particular cases in specific contexts qualitative researchers are also concerned can be shown to be internally valid. In experi- with establishing validity. This is not in terms mental, survey and other quantitative, of controlling or ruling out alternative vari- variable-based research this usually means ables but in terms of providing assurances assembling evidence that no variables other that the account that has been put forward than those under consideration could have (say, of interactions in the classroom) is the produced the outcome that has been observed. correct one. This is done by analytic induc- In experimental research this is done by ensur- tion (searching for conclusions that do not fit ing that control and experimental groups are the conclusions) and reflexivity (reflecting on matched on key variables – which may offer the possible effects of the researcher on the alternative explanations – prior to adminis- conclusions put forward). tering the ‘treatment’ to one group and not to the other (the control group). Any subsequent and significant differences between the Victor Jupp groups can be attributed to the treatment variable. In this way, the threats to the inter- Associated Concepts: analytic induction, nal validity of the explanation which may causality, experiment, field experiment, qual- have come from the ‘matching’ variables have itative research, reflexivity, social survey been ruled out. The direct manipulation of variables to Key Readings reduce threats to validity is only feasible in Bryman, A. (2004) Social Research Methods, controlled laboratory conditions. For this reason 2nd edn. Oxford: Oxford University Press. survey-based research relies on statistical meth- Hague, J. and Meeker, B. F. (1988) Social ods to rule out possible alternative explanatory Causality. London: Unwin Hyman. variables. Statistical techniques allow the data McKim, V. R. and Turner, S. P. (eds) (1997) analyst to hold some variables constant whilst Causality in Crisis: Statistical Methods and manipulating others to assess whether any vari- the Search for Causal Knowledge in the ance in the dependent variable can be attrib- Social Sciences. Notre Dame, IN: uted to such manipulation. Multivariate University of Notre Dame Press. analysis encompasses such techniques. Smith, H. W. (1975) Strategies of Social In addition to demonstrating that changes Research: The Methodological Imagination. in one (dependent) variable are associated Englewood Cliffs, NJ: Prentice Hall. with changes in other (independent) variables and also that other variables have been ruled out (either by experimental design or statistical control), valid conclusions – especially in VALIDITY OF GENERALIZATION relation to causality – require evidence of temporal sequence. This is evidence that the Definition changes in the dependent variable only occurred after changes in the independent The extent to which information from a sam- variables. Outside of the laboratory it is often ple gives us information about a population, difficult to obtain direct evidence of temporal or the extent to which information about one sequence. For this reason, the explanatory setting tells us about others (which may be of validity of surveys tends to be weaker than more interest to us). that for experimental research. Distinctive Features Evaluation Two varieties of the term can be noted. First, The notion of explanatory or internal validity population validity is the extent to which sam- is most usually associated with quantitative ple distributions mirror those of the popu- research and the search for causes. However, lation which the sample is supposed to

312 V-Jupp-3373.qxd 3/8/2006 5:53 PM Page 313

VALIDITY OF GENERALIZATION

represent. Given perfectly conducted random interpretation of the data can be generalized sampling, probability theory permits us to esti- beyond the immediate research sample or mate population parameters from sample setting to give information about a wider pop- readings with a calculable error margin known ulation. The term is used most commonly in as sampling error – if 30 per cent of the sam- qualitative research. ple believe something or have a certain char- Theoretical generalizability is generally acteristic, our best estimate is that this will be established in the first instance by looking true of 30 per cent of the population, plus or at other, similar settings or samples that differ minus a calculable error margin, in a calcula- in some key respect, to see whether the orig- ble proportion of the samples that could possi- inal theorized interpretation will also fit the bly have been drawn. To the extent that the new data. Researchers may check, for exam- sample is less than perfect (non-sampling ple, that accounts of doctors’ decision making error), this estimate is distorted. A common in terms of UK gender stereotypes based on source of non-sampling error, for example, is a sample from the north of England can be the fact that a proportion of most preselected reproduced in samples of southerners or random samples will decline to take part, or people from Scotland or Wales, and whether we will fail to locate them, thereby changing conclusions based on predominantly working- the nature of the sample in ways that cannot class areas hold true for doctors with middle- be predicted reliably. class practices. A next step might be to see In ethnographic and other qualitative whether it is just doctors’ decision making research, where samples are small, this kind of which is being described, by carrying out generalizability is seldom achievable. In these similar work with school teachers, perhaps, cases the researcher will generally attempt to or social workers; the results might mean establish population validity by arguing the typ- that theory needs to be recast in terms of the icality of the setting or institution that is being decision making of ‘social professionals’ in studied, or assessing the extent to which it may general. Finally, we might look for situations not be typical, and/or demonstrating that the where the model should definitely not fit; in cases investigated at least cover the range of this example, if we are talking about doctors’ what might have been expected, with no group or professionals’ decision making rather than obviously excluded from the study. generally shared cultural stereotypes, then Second, ecological validity is the extent to data from the doctors’ patients should not be which results from one setting can safely be gen- describable in the same terms. eralized to others. More specifically, the concept is often used to criticize the more structured Roger Sapsford varieties of research which attempt to generalize from the artificial social situations of the survey or the social laboratory experiment to what peo- Associated Concepts: inferential statistics, ple do or say or believe in their ordinary lives, non-probability (non-random) sampling, pro- outside the research situation. What is at stake bability (random) sampling, sampling, valid- here is naturalism – the need to study everyday ity of explanation, validity of measurement life ‘undisturbed’ and free from the artefacts introduced by research structures and proce- Key Readings dures. We should note that this is seldom entirely achievable even in qualitative research; Sapsford, R. and Jupp, V. (eds) (1996) Data virtually all research imposes some degree of Collection and Analysis. London: Sage. chs 1 structure or change on the natural circumstance. and 2. Schofield, J. W. (1989) ‘Increasing the gener- Evaluation alisability of qualitative research’, in M. Hammersley (ed.) (1993), Social Research: A third sense of the term is theoretical Philosophy, Politics and Practice. London: generalizability, the extent to which the Sage.

313 V-Jupp-3373.qxd 3/8/2006 5:53 PM Page 314

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

children gives much the same results as the VALIDITY OF MEASUREMENT existing form. It does not itself solve the problem of validity unless the validity of the com- Definition parison test can be demonstrated. The extent to which an indicator or variable Predictive validity is quite a strong form adequately measures the theoretical concept of warranty – demonstrating that your test can it purports to measure. predict performance on something agreed to be an expression of the construct to be mea- Distinctive Features sured. Tests of initial ability, for example, are validated by their prediction of eventual per- The major requirement of measurement is formance. For example, the use of car owner- validity. For research to be conducted, theo- ship in an area as a measure of affluence can retical concepts (and research questions and be validated against average income in the hypotheses of which they are part) must be area (and is used when employing census operationalized, that is, cast in terms of infor- data, for example, where car ownership is mation that can be collected and manipulated measured but income is not). A personality in the public domain. The theoretical concept scale of Machiavellianism (the tendency to may be social class or social capital or socia- achieve one’s goals by manipulating other bility, but the operationalized indicator or people) was validated against performance variable will be about current job or extent to in a game where manipulation brought which people are engaged in vertical social results. networks or their scores on a personality All of these may be seen as aspects of questionnaire. The validity of this opera- the broader concept of construct validation – tionalization is the extent to which the fitness demonstration of the extent to which the of the operationalized indicator to measure underlying concept (‘construct’) coherently the theoretical concept can be justified and fulfils what is demanded of it, by demonstra- demonstrated. tion that measures can be devised which do The main ways of warranting validity are: so. Within construct validation we speak of face Validity – that ‘on the face of it’ the test two varieties. First, convergent validation is measures what it says it does; concurrent valid- where tests perform as theory demands of ity – that it gives the same answers as an them. Many of the senses of validity above already validated measure; predictive or crite- are concerned with demonstrating this – that rion- related validity – that the test or measure- a trait measurement, for example, is stable ment predicts an outcome or correctly over time, performs similarly to other identifies group membership. Face validity is accepted measures and correctly predicts obviously the weakest form of warrant – ‘well, behaviour. Unidimensionality (generally it looks as though this ought to measure that’. demonstrated by factor analysis) would be It may be sufficient in some circumstances – it another aspect of convergent validation – may be a sufficient warrant for the validity of demonstrating that the test measured a single a test of arithmetic that the items are patently trait or characteristic and was not contami- arithmetical problems. Even then, doubts may nated by confounded variables. Second, dis- be cast; arguably, timed tests measure people’s criminant validation, equally important, is to tolerance of stress as well as their arithmetical demonstrate that a test does not do what it ability, in that those who can tolerate stress will is supposed not to do. A test of arithmetical tend to do better at them. ability, for example, should not also measure Concurrent validity is useful when pro- anxiety, and it should give high scores only to posing an alternative measure because the those who subsequently perform well at accepted one is not appropriate – showing, arithmetic. A test that also picked out people for example, that a short form of a test or one good at English or Art might more properly rewritten in language appropriate to young be considered a test of general ability.

314 V-Jupp-3373.qxd 3/8/2006 5:53 PM Page 315

VALUE-FREE RESEARCH

Evaluation Moser, C. A. and Kalton, G. (1971) Survey Methods in Social Investigation. London: In the main, validity of operationalization Heinemann. and measurement relates to quantitative Smith, H. W. (1975) Strategies of Social Research. research. Qualitative research does not oper- Englewood Cliffs, NJ: Prentice Hall. ationalize but attempts to work holistically. However, it still faces problems of validity in trying to demonstrate that the patterns it records are typical of the persons or settings and not products of the research situation. In VALUE-FREE RESEARCH other words, it needs to consider reactivity – personal reactivity (reaction to the person, Definition role and actions of the researcher) and procedural reactivity (reaction to the way in which Research that has been produced by a com- the research was conducted). One strategy pletely impartial and dispassionate researcher. for overcoming these kinds of doubts is The proposition is that if a researcher can con- triangulation – using more than one person or duct a study shorn of his or her own particular more than one procedure (for example, com- beliefs, values, prejudices and opinions, this paring oral interviews with text produced by impartiality will presumably be reflected in the or about the informants) to look for conver- end product of the research. gence. However, the major tool is reflexivity – careful and practised sensitivity about the Distinctive Features extent to which personal characteristics and The proposition that a researcher can truly behaviour, the accidents of the research set- be ‘value-free’ is hotly contested in the social ting, the relationships involved and the biases sciences. On one side of the equation, one and preconceptions of the researcher may could argue that a researcher may indeed be have an effect on the nature of the data pro- ‘value-free’. This is most likely in studies that duced. The aim is to produce a research incorporate large-scale statistical analysis in report that is transparent about how data are which a respondent is only able to provide collected and interpretations made, so that answers in a closed setting. Examples of this the reader can form his or her own judge- can be found in market research, or opinion ment of their validity. This kind of sensitivity polling. Value-free research is therefore more is also present in the best quantitative work. indicative of a quantitative or positivist research methodology. Defenders of the Roger Sapsford approach sometimes argue that researchers can be just as impartial as a scientist observing a laboratory experiment or some particu- Associated Concepts: indicator, measure- lar phenomenon in the natural world. ment, reflexivity, reliability, social indicators, On the other hand, however, a large num- triangulation, validity of explanation, validity ber of academics dispute that research can be of generalization value-free. The main thrust of this line of thinking is that whenever a researcher analyses a given social area, he or she will always Key Readings bring his or her own belief system into the Blalock, H. M. Jr (1968) ‘The measurement research. In addition, and by implication, problem: a gap between the language of when a researcher decides to analyse a spe- theory and research’, in H. M. Blalock Jr cific area, they will have done so because and A. B. Blalock (eds), Methodology in they have an interest in the subject matter. Social Research. New York: McGraw– And where an interest exists, the researcher Hill. pp. 5–27. will necessarily have opinions and values.

315 V-Jupp-3373.qxd 3/8/2006 5:53 PM Page 316

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

From the outset, therefore, the researcher Phillips, D. C. (1987) Philosophy, Science and simply cannot be value-free. This is largely Social Inquiry. Oxford: Pergamon Press. recognized in many forms of interpretative Scheffler, I. (1982) Science and Subjectivity, and qualitative research, particularly those 2nd edn. Indianapolis: Hackett. that involve interaction with the respondent. Furthermore, quantitative research can be criticized for the type of questions that are put to respondents. In a closed setting, the VARIABLE ANALYSIS researcher is only asking questions on areas that he or she has already deemed fit for inclusion. Hence, even though the questions Definition may be neutral and value-free, the process of Variables are units of data that can change framing the questions has been affected by between different cases. The different values value judgements. that a variable can take affect the type of analy- A counter-argument is that a researcher sis that is possible. Variables can be analysed may stumble into an issue area by chance (for on their own (univariate analysis), with one example, becoming a research associate in an other variable (bivariate analysis) or with a already established project). In such circum- number of others (multivariate analysis). stances, the researcher may not have any previous knowledge, or values for that matter, in Distinctive Features the subject. One could then plausibly assume that the researcher is indeed value-free. Variable analysis is a key characteristic of Those coming from an interpretative stand- quantitative research. While qualitative point, however, would assert that as the analysis tends to be concerned more with research progresses, the researcher will nec- themes, interpretation and the use of language, essarily build up their own value-system, quantitative analysis examines variables, thereby prejudicing their work. which are typically collected through surveys, although sometimes by other methods Evaluation such as observation. Variables can represent a characteristic (for example, a person’s age), For some, value-free research is a goal that all factual matters (the number of times they researchers should aim for. Nevertheless, an have been a victim of crime) or opinions (the increasing number of researchers now recog- punishment for criminals that they think is nize that their own value systems will influ- most appropriate). The types of analysis that ence their fieldwork to a greater or lesser can be undertaken depend on the nature of degree. For those who come from a critical, the variable, so it is necessary to consider emancipatory or standpoint research tradition, some distinctions. value freedom is not even a desirable goal. Categorical variables – here any number that is applied to a value has no meaning, other Craig McLean than as a label. For example, it is common to code male as 1 and female as 2, but there is no reason why this should not be 1 for female Associated Concepts: anti-racism research, and 2 for male. critical research, emancipatory research, femi- Ordinal variables place data in a particular nist research, relativism, standpoint research order – for example a respondent may be asked to number different leisure activities Key Readings from the one that they most like to do to the one that they least like to do. Keat, R. and Urry, J. (1982) Social Theory as Cardinal variables represent a ‘real’ value, Science. London: Routledge & Kegan Paul. for example someone’s height in centimetres

316 V-Jupp-3373.qxd 3/8/2006 5:53 PM Page 317

VERSTEHEN

or the number of children that they have. to have mental health problems, but is this Cardinal variables are sometimes further because the onset of mental illness makes it divided between discrete variables (which can more difficult to remain in accommodation, only have a limited number of values, for or because the experience of homelessness example number of children) and continuous is likely to be damaging to mental health? variables (where it is not possible to specify Second, failure to take account of other vari- all the possible values, for example height). ables: for example, a programme of teaching They can also be divided between interval in schools about the value of exercise may and ratio variables. be followed by an increase in the number of The type of univariate analysis that can be children taking part in sport, but it is possi- undertaken with categorical variables is ble that the children were more influenced restricted to identifying the frequency with by watching the Olympic Games than by the which values occur or calculating the percent- teaching. age of responses falling into one or more categories. In contrast, ordinal variables and cardinal variables can be discussed in terms of Jamie Harding measures of central tendency – the mean, median and mode – and measures of disper- Associated Concepts: chi-square test, coding, sion, such as the range and standard deviation. data, descriptive statistics, factor analysis, Bivariate and multivariate analyses for any inferential statistics, measurement, multivari- type of data seek to establish relationships ate analysis, parametric tests, probability between variables, for example whether high (random) sampling income tends to be associated with voting for a particular political party. When a random Key Readings (or probability) sample has been taken, infer- Fielding, J. and Gilbert, N. (2000) Under- ential statistical tests can be used to measure standing Social Statistics. London: Sage. the likelihood that a relationship between de Vaus, D. (2002) Analysing Social Science variables that appears in a sample will be Data. London: Sage. reproduced in the population. Examples of such inferential tests are the chi-square test of independence – which examines the relationship between two categorical variables – and the t-test, which can VERSTEHEN be used where there is one cardinal variable and one dichotomous categorical variable Definition (that is, one that can take two possible values, Verstehen is a German word meaning ‘under- such as gender). standing’. Verstehen sociology emphasizes the necessity of understanding the meaning of Evaluation human action.

Bivariate and multivariate analysis are often Distinctive Features used to establish that one variable (the independent variable) has an effect on another Verstehen sociology is usually associated with (the dependent variable). However, care the work of Max Weber (1864–1920), but he must be taken not to draw incorrect conclu- was not the first to use the term, which sions about causation. Common difficulties derives from the tradition of hermeneutics in include the following. First, confusion over nineteenth-century German thought. Writers the dependent and the independent vari- in this tradition, such as Wilhelm Dilthey ables: for example, homeless people are (1833–1911), argued that the subject matter more likely than the population as a whole of the human sciences differs from that of the

317 V-Jupp-3373.qxd 3/8/2006 5:53 PM Page 318

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

natural sciences since human action is Evaluation meaningful and purposive. Social events, there- Many aspects of Weber’s work remain fore, are contingent upon subjective inter- controversial (see, for example, Bauman, pretations, motivations and choices. The 1978), including his emphasis on rationality methods of the natural sciences could not in human action, the extent to which subjec- deal effectively with these dimensions and it tive meanings can be determined objectively was argued that a special method of under- and proved to have played a causal role in standing, Verstehen, is needed to interpret the social events. Issues about the relationship subjective and cultural aspects of social phe- between structure and agency remain a cen- nomena. Drawing upon this tradition and tral concern of social theory and Weber’s applying it to social science, Weber was criti- work, although criticized, remains relevant. cal of positivism, and also of Marxism, which Weber’s work on Verstehen is an important proposed the existence of general laws, the strand of interpretivism in social science and discovery of which would provide a causal social research in itself as well as informing explanation of human action and social the development of phenomenology and eth- events. In contrast, Weber emphasized the nomethodology and symbolic interactionism. need to understand the values that shape people’s actions, and to understand these values in a non-judgemental (‘value-neutral’) Maggie Sumner way (Weber, 1949). However, he did not consider that social Associated Concepts: hermeneutics, inter- science should be concerned only with the subjective understanding, methodology, tex- interpretation of subjective worlds. For tual analysis Weber, Verstehen includes explanation, an understanding of motive and its relationship Key Readings to action. He maintained that subjective meanings could be studied objectively in a Bauman, Z. (1978) Hermeneutics and Social scientific manner, through the analysis of Science: Approaches to Understanding. ideal types, particularly of human rationality. London: Hutchison. He also argued that social science should look Parkin, F. (1982) Max Weber. London: for widespread regularities in subjective Tavistock. values over periods of time: such regularities Swingewood, A. (1991) A Short History of become social facts and amenable to scien- Sociological Thought, (2nd edn.) London: tific analysis and empirical proof. A scientific Macmillan. analysis should also demonstrate a ‘fit’ Weber, M. (1930) The Protestant Ethic and between the subjective choices typically the Spirit of Capitalism. London: Allen & made by individuals and the actual occur- Unwin. rence of phenomena. (Weber’s own work on Weber, M. (1949) The Methodology of the Social Sciences. New York: Free Press. the relationship between protestantism and the development of capitalism (Weber, 1930) illustrates his approach.) Thus, Weber did not see subjective interpretation and causal explanation as opposed: rather he argued that VIDEOGRAPHY in social science, any adequate explanation of phenomena must include an understandable Definition account of the subjective meanings of action. That is, Verstehen, the rational understanding A form of visual anthropology encompassing of the subjective world-views of the actors the collection, analysis and presentation of involved in social events, is an essential com- visual data; more specifically, an audiovisu- ponent of explanation. ally based ethnography that is the product of

318 V-Jupp-3373.qxd 3/8/2006 5:53 PM Page 319

VIDEOGRAPHY

a participant-observational research method videographic research techniques offer a that records interviews and observations of middle ground between unobtrusive researcher particular peoples, groups and their cultural observation and autovideography, in which artefacts, utilizes them as data, edits them researcher and informants jointly negotiate the into a format for presentation, and represents edited end product. As well, by interviewing it in the form of a film. and observing other culture members, informants can leverage their own status as cultural Distinctive Features insiders in ways that would be difficult and time-consuming for ethnographers. Videography collects a distinctive form of Retrospective applications are also possi- data: audiovisual (‘AV’) cultural data. It uses ble, using, for instance, home movies which distinct means to capture and analyse it: may be combined with subsequent elicitation video cameras, film and other AV recording of verbal informant data. The film footage, in devices. It produces its research representa- this case, acts as a projective stimulus for tion in a unique form: as an edited film. As eliciting other videographic data. Other film digital technology has advanced over the past archives, such as those of commercial and decade, the expenses associated with produc- promotional films, are also often used in con- ing and distributing broadcast quality video junction with retrospective videographic have plummeted. These developments techniques. Unobtrusive observations – which have led to a blossoming of possibilities for sometimes use concealed video equipment – almost any ethnographic researcher to enter also fit into the toolkit of videographic the domain of filmmaking once reserved for researchers. The use of concealed camera documentary journalists and only the most techniques raises the importance of research well financed social scientists. ethics protocols in videographic research. The most basic use of video, and still the Once videographic data are collected, a most common, is the videotaped individual variety of analytic tools may be marshalled or group interview, conducted either in a to code and categorize them. Videographic research facility or in a field setting. analysis follows the basic principles of Videotaped interviews offer the powerful interpretive analysis, from grounded theory- advantage of capturing body language, prox- building to hermeneutic cycling. Video analy- emics, kinesics and other temporal-spatial sis can happen holistically, or occur through dimensions of human behaviour and social a very formalized frame-by-frame coding and meaning. In this use, the researcher is inter- classification procedure. Once the AV data ested in capturing information about what are analysed, the edited production of the the cultural informants say about what they film is undertaken. do. The second most common use of video There are two basic formats for disseminat- in research is to record naturalistic observa- ing video material. The first we term local tions. In recording naturalistic observations, access, and this would include fixed, physical, the videographer is more interested in cap- media such as CD-ROM, DVD and videotape. turing what people do rather than what they The second is a more broadcast format, which say about what they do. we term distributed access, and includes tele- Variants of the videography based on natu- vision broadcasts and Internet streaming. ralistic observation are common. One involved Local access media usually offer higher qual- the attempt to understand the culture mem- ity, better control of access to material and its ber’s viewpoint by having them directly pace, order and sometimes language, view and capture it on videotape or film. This autobio- content control. Web-based access can offer graphical and autovideographical technique has some of these options as well, in addition to the advantage over researcher-conducted obser- updating and linking to related material. vation of being less intrusive and less directed However, due to technological limitation on by researcher motives and needs. Collaborative bandwidth there are currently compromises

319 V-Jupp-3373.qxd 3/8/2006 5:53 PM Page 320

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

in quality (in terms of lower frame rates, naturalistic data, photography, qualitative lower resolution, and smaller screen size) in data archives, validity of measurement, streamed video material. visual methods

Evaluation Key Readings Videography offers many compelling advan- Banks, M. (2001) Visual Methods in Social tages over other forms of ethnographic Research. London: Sage. research, which is why is has become an Belk, R. W. (1998) ‘Multimedia consumer increasingly frequent accompaniment to tra- research’, in B. Stern (ed.), Representation ditional ethnography. By collecting AV data, in Consumer Research. London: Routledge. researchers hope to capture more of the sub- pp. 308–38. tle temporal, social and spatial dimensions of Evans, J. and Hall, S. (eds) (1999) Visual human behaviour, including the intricacies of Culture: The Reader. London: Sage. human body language. The resulting research Pink, S. (2001) Doing Visual Ethnography. representation on film can be resonant, emo- London: Sage. tional, vibrant and humanizing. It can give audiences a sense that they have actually experienced some of the cultural events being represented, deepening human understand- VISUAL METHODS ing, and making the unfamiliar seem familiar. There are some important drawbacks to Definition the videographic method as well. First, filmmaking involves technique, and in a world The use of visual images and technologies filled with experienced screen watchers, such as video, film, photography, art, draw- audience expectations tend to be quite high. ing and sculpture in qualitative social Talking head video, for example, fails to take research to both produce and represent full advantage of the medium. Second, the knowledge. This includes using the visual as camera can prove an unwelcome hindrance a documenting tool to produce visual records, to the formation of interviewer–interviewee in interviews to elicit comments from infor- rapport. Third, ethnographic film tends to be mants, in participant observation to research read by audiences using the same suspension ways of seeing and understanding, analysing of disbelief they learn to apply to commercial visual and material culture and using visual films and television. Prevalent techniques media to represent the findings of such such as rich imagery, special effects, emo- research. tional swells of music, cuts and other editing flourishes tend to engage viewers, but poten- Distinctive Features tially manipulate their emotions and exacer- Because visually we can communicate bate the impression of realism. Even more knowledge, experiences and ideas in ways than other research representations, video- that we cannot using only written or spoken graphy must deal with crisis of representa- words, the visual is gaining increasing impor- tion issues in the many decisions involved in tance as a social research method. A range creating the research product, and realize of visual methods have been developed for that videographic representation is capable of social research, including visual documenta- telling many ‘truths’. tion, photo elicitation, photographic and video interviewing, asking informants to pro- Robert V. Kozinets and Russell W. Belk duce images, participant observation with a camera, ethnographic film making and most Associated Concepts: coding, ethics, ethno- recently ethnographic hypermedia. We may graphy, grounded theory, idiographic, interview, classify images used in visual research into

320 V-Jupp-3373.qxd 3/8/2006 5:53 PM Page 321

VISUAL METHODS

two categories. First, images produced by the hypermedia project produced for the Internet, researcher, and second, images found or CD-ROM or DVD. produced by informants. Images produced by informants are also Photography, video or drawings produced used in social research. Social researchers by the researcher might be used to visually analyse other people’s existing image collec- document a place or event, providing a visual tions and images they have invited infor- record of that context as seen by the mants to produce for the research. For researcher. Such images may also have further example, studies of family or personal photo- uses. First they might be used to establish rela- graphy might focus on individuals’ existing tionships with informants, for example offer- photographic collections. These teach us ing photographs one has taken of people to much about ideal representations of family them as gifts. Second, they can be used in and other relationships and offer insights into photo or video elicitation, a method that the social history of family and personal life. involves the researcher showing informant(s) Researchers also might analyse people’s art images to elicit their response. It has been collections, the images they display on their found that showing people images is likely to walls and their home movies. evoke a whole range of responses and infor- Visual researchers also ask informants to mation that it would not occur to them to offer produce images for them to analyse. For in a verbal interview (see Harper, 2002). Other example, an informant might photograph, types of images are also used in elicitation video or draw aspects of her or his own interviews, for example images from infor- everyday life, keeping a photographic or mants’ own photography, drawing or video video diary to hand in to the researcher for collections, or other ‘found’ images. analysis. In a method called ‘image work’, Visual methods often involve collabora- that has parallels with some types of therapy tively producing images with informants, for (Edgar, 2004), people are asked to make draw- example in video or photographic interviews. ings that express their feelings about particu- Here researchers will interview a person lar issues or experiences. These can then be using a video camera or stills camera to discussed by informants and analysed by the record the visual aspects of that interview. researcher. Such methods allow people to This allows informants to explicitly use the represent to researchers ideas, knowledge visual to communicate during the interview, and experiences that cannot be represented and provides researchers with audiovisual using solely the written or spoken word. records of that encounter. Recording video, taking photographs or making drawings as Evaluation part of a participant observation project can also help researchers understand or ‘see’ Until recently there was much resistance to their informants’ points of view, or ‘ways of the use of the visual in social research. It was seeing’. For example in Spain I photographed often considered to produce data, or research the bullfight and asked my informants to materials that were too subjective. However, comment on my photographs as a means of if we accept that subjectivity is an essential learning about the way they saw the bullfight element of any qualitative social research and an attempt to photograph the event the project then it becomes clear that visual way they saw it. Grasseni similarly used materials are just as valid as written or verbal video in Italy to learn to see cattle from the data. Understanding this subjectivity is key to perspective of her local cattle-breeding infor- analysing visual images. Therefore when mants (Grasseni, 2004). Finally, images pro- social researchers analyse images they focus duced by researchers are used to represent on three main issues: the processes by which the findings of social research. This might be the images were produced, the content of the in the form of a documentary film, a photo- images, and the context in which they are graphic or art exhibition or book, a multimedia viewed and understood.

321 V-Jupp-3373.qxd 3/8/2006 5:53 PM Page 322

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

This emphasis on subjectivity is linked to Pink, S. (2001) Doing Visual Ethnography: the insistence amongst visual researchers on Images, Media and Representation in reflexivity. Visual researchers need to be Research. London: Sage. aware of how their own subjective feelings and Pink, S., Kurti, L. and Afonso, A. I. (eds) assumptions influence the images they pro- (2004) Working Images. London: Routledge. duce and their understandings of other people’s Rose, G. (2001) Visual Methodologies. London: images. To be fully reflexive researchers should Sage. reveal these assumptions in the texts they produce about their research. Visual methods are never purely visual. They usually combine the visual with words, VOLUNTEER SAMPLING whether this is in the form of asking questions or conversing with informants, or in written texts. The visual is also never totally isolated Definition from other sensory experiences, such as touch, A form of case selection which is purposive smell, and sound. For example, videos are not rather than based on the principles of random purely visual, but audiovisual, and printed or probability sampling. It usually involves photographs are also material objects that one individuals who agree to participate in touches, feels and talks about. research, sometimes for payment. However, it Using visual methods is clearly beneficial can involve other units of analysis, for example to social researchers; it adds a new dimension when organizational heads volunteer their to qualitative research by introducing the institutions as sites for research. possibility of knowledge and experience that cannot be communicated verbally or in Distinctive Features written words. The basic principle of random or probability sampling is that every case in the population Sarah Pink being studied has an equal and non-zero chance of being selected as a sample member. If ran- Associated Concepts: ethnography, mixed dom selection is used it is possible to use prob- methods research, photography, projective ability theory to make statistical inferences techniques, qualitative research, reflexivity, about the population from a sample selected subjectivity, unobtrusive measures from it. For example, it is possible to infer the extent to which an estimate of average age collected by surveying a sample is likely to be close Key Readings to the true figure for the population as a whole. Sometimes it is not possible or desirable to Visual Methods in Social Banks, M. (2001) use random or probability techniques, in Research. London: Sage. which case the sample members are chosen Edgar, I. (2004) ‘Imagework in ethno- for a particular purpose. The reasons may be graphic research’, in S. Pink, L. Kurti and that a sampling frame (a complete list of A. I. Afonso (eds), Working Images. London: Routledge. everyone in the population) prevents random Grasseni, C. (2004) ‘Video and ethnographic selection; there are not sufficient resources or knowledge: skilled vision in the practice time to contact all those selected for a ran- of breeding’, in S. Pink, L. Kurti and dom sample; the topic of research is so sensi- A. I. Afonso (eds), Working Images. London: tive that few members of a sample would be Routledge. willing to be sources of data. Harper, D. (2002) ‘Talking about pictures: a Volunteer sampling is a form of purposive/ case for photo-elicitation’, Visual Studies, non-random sampling for all such reasons. It is 17 (1): 13–26. used especially in sensitive research when it is

322 V-Jupp-3373.qxd 3/8/2006 5:53 PM Page 323

VOLUNTEER SAMPLING

necessary to rely on those who are willing to establish the extent to which those who answer requests to provide data. For example, volunteer to participate are typical of the research conducted in the UK by the London group to which findings are to be generalized. Metropolitan Police in 2004 contacted impris- For this reason it is important to be cautious oned serial killers to construct a pool of volun- about the claims that are made about and teer subjects. Detailed interviews were from such research. conducted in order to study the psychological It is also important to pay attention to the profile of such killers with a view to identify- ‘research bargain’. It is possible that those ing potential offenders in the future. who volunteer to take part in research do so Volunteers are also used as subjects in in order to gain some benefit to themselves. It psychological experiments in which some are is important for a researcher to consider allocated to a control group and others to an whether such benefits can, or indeed should experimental group (which is given some be met. stimulus or ‘treatment’ to see if being a member of the latter produces some predicted effect). Victor Jupp Similar principles can apply in programme evaluation research whereby individuals Associated Concepts: ethnographic inter- volunteer to be the recipients of some policy viewing, focus group, non-probability (non- initiative, the efficacy of which is subse- random) sampling, opportunity sampling, quently evaluated. Finally, focus groups, purposive sampling, research bargain, sensi- which are used in a wide range of marketing tive research, snowball sampling and social research, are often constructed from individuals who volunteer to take part. Key Reading Evaluation Moser, C. A. and Kalton, G. (1971) Survey A key problem with volunteer sampling is Methods in Social Investigation. London: representativeness insofar as it is difficult to Heinemann.

323 W-Jupp-3373.qxd 3/8/2006 10:46 AM Page 324

using a specific style of referencing (such as WRITING RESEARCH the Harvard style). The structure generally follows a logical Definition sequence that provides the reader with infor- The process and techniques of communicat- mation about what was being studied, how ing significant information about a piece of it was studied and what was found as a research. Writing research serves two func- result of the investigation. Conventionally, tions, both of which make it an essential the research is reported in line with the fol- ingredient of the overall research process. lowing headings. First, writing research enables findings to become publicly available and thus allows Title This identifies the topic and scope of the research. other researchers to use the findings Abstract A brief overview of the research and build upon them in the furtherance of summarizing its principal aims, knowledge. Second, writing research allows methods and conclusions. the methods of investigation to become Key words A small number (4–10) of open to scrutiny by peers. On the basis of words that facilitate biblio- written accounts of how the research was graphic searches. conducted and how the results were Acknowledgements A statement noting the help obtained other researchers can evaluate provided by others in conduct- the research and judge the quality of the ing the research or producing findings. its findings. Aims A broad overview of what the research was attempting to Distinctive Features achieve. Introduction A relatively short section that Despite the variety of disciplines and explains the background to approaches that characterize the social sci- the research. ences there are certain conventions associ- Literature review A critical evaluation of previous ated with writing research that are widely studies whose purpose is to accepted. These conventions cover both the demonstrate how the research style and the structure of the writing. The will enhance what is already style of writing research tends to be formal. known about the particular subject. This involves writing in the ‘third person’ Research and adopting a style that is succinct, direct questions A list of specific points to be and clear. It also involves a meticulous atten- investigated. These can take tion to citing the sources from which ideas the form of hypotheses, and findings have been derived, and doing so propositions, or questions. W-Jupp-3373.qxd 3/8/2006 10:46 AM Page 325

WRITING RESEARCH

Methods A factual statement of how the research Evaluation was conducted. The design of the research, the methods of data collection The growth of qualitative research and the and the techniques of data analysis need influence of the postmodernist critique have to be described. They also need to be effectively removed any illusion that writing justified as appropriate in relation to the research questions and to broader social, research simply involves ‘reporting what ethical and resource issues. actually happened’. Instead, it is recognized Findings A section presenting the main findings that writing research involves producing from the investigation. The emphasis an ‘account’ of the research. Such accounts is on describing what was found with- are, in the first instance, edited highlights of out delving into possible explanations what happened. Inevitably, writing research of the findings or considerations of entails reducing the huge amounts of mater- what their implications might be. ial researchers accumulate as a record of Discussion A section that attempts to analyse the what was done, why it was done and what findings and to explain what factors outcomes resulted. The editing process calls lay behind the findings. Conclusions A relatively brief, but very important, on researchers to prioritize certain things in section that considers the implications terms of their significance for the investiga- of findings from the research. tion and possibly omit altogether some other Appendices Technical and other supporting mate- aspects of the research process from their rial that readers might need to refer to report. The extent of reduction will vary but whose presence within the body depending on whether the research is writ- of the text would unduly disrupt the ten up in the format of a journal article, PhD flow of the discussion. thesis or practitioner report, but the parame- References A list of the published works cited in ters of space will guarantee that the the report. The details allow readers to researcher could never provide a full and search for the original sources. exhaustive account of the research. Accounts of research also call for creative writing skills Additionally, there are aspects of research that on the part of the researcher. As Van sometimes get subsumed under other head- Maanen (1988) has shown, writing research ings but which can also appear as separate and entails the use of ‘rhetorical devices’ through distinct sections in their own right. Three which the researcher aims to persuade the important examples are: reader of the genuineness and authoritative- ness of the account. The style and amount of detail provided in an account of research Outcomes A section that specifies the ‘tangible’ end-products of the investigation. These will vary, as well, depending on the expertise can take the form of contributions to the of the audience and the extent to which the stock of knowledge, developments of researcher can presume the readers have theory, suggestions for future research, prior knowledge of the subject matter and practical recommendations or guidelines methods of research. The skill of writing for good practice. research, then, resides in knowing what ele- Limitations An acknowledgement of what can be ments of the investigation to focus upon and justifiably concluded on the basis of being able to present them in a way that the methods, findings and analysis adequately and honestly describes the and, importantly, the limits to what investigation and also justifies the approach can be deduced. Ethics A consideration of the ethical issues and decisions in terms of expected standards involved in the topic of research, the of good practice among peers, colleagues, methods of data collection and fellow researchers and others such as policy analysis, and the dissemination of makers and the public who might feel findings. the need to assess the worth of the research.

325 W-Jupp-3373.qxd 3/8/2006 10:46 AM Page 326

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

The skill of writing research is to match the Key Readings need for transparency with the expecta- Atkinson, P. (1990) The Ethnographic tions of the audience within the confines Imagination: Textual Constructions of Reality. of the space allowed by the format of the London: Routledge. publication. Denscombe, M. (2002) Ground Rules for Good Research. Milton Keynes: Open University Press. Martyn Denscombe Hammersley, M. (1990) Reading Ethnographic Research: ACritical Guide. London: Longman. Tales of the Field: On Associated Concepts: ethics, literature Van Maanen, J. (1988) Writing Ethnography. Chicago: University review, postmodernism, reflexivity, research of Chicago Press. design, validity

326 Subject-Index-Jupp-3373.qxd 3/8/2006 5:55 PM Page 327

Subject Index

access 1–2, 36–7, 126, 265 CASI 22 context analysis 177–8 elite interviewing 85 CASIC 21–2 context–mechanism–outcome organizational research 209 categorical (nominal) variables pattern configurations participant observation 215 168, 292, 316, 317 (CMOCs) 106 accidental sampling 196–7 CATI 21 continuous systems 181 action research 2–3 causal marketing research continuous variables 57, 274, additivity 22 166, 167 275, 317 AHFA (Ad Hoc Fumbling causal models 22–3, 218 convenience sampling 196–7 Around) approach 159 causality 23–5, 44, 137, 164–5 convergent validation 314 alternate-form reliability 262 census 25–6 conversation analysis analysis of covariance chi-square test 26–8, 46, 58 42–3, 67–8, 104, 146, (ANCOVA) 5, 46–7, 127 Chicago School 32, 120, 174 178, 191, 304–5 analysis of variance (ANOVA) cluster analysis 28–9, 184 correlation 43–5, 262 3–4, 127 cluster sampling 10, 29, 239 ecological fallacy 82–3 analytic induction 5, 21, 146 coding 30, 37 indicators 145 applied research 8–10, 45 software 11, 19, 98, 99, correspondence rules 205 archives, data 58–60, 197–8, 246–7 cost-benefit analysis 45–6 247–8, 274 cohort study 30–2, 242–3 covariance 5, 46–7, 127 area sampling 10 Columbia School 91 covering-law model of Atlas.ti 10–12 communality 114–15 explanation 47–8, 139 attrition 12–13, 32, 165, community study methods 32–3 covert research 48–50, 97, 211, 243 comparative method 33–4 102, 150, 216, 304 audiovisual data 319–20 constant 33, 37–8, 131–2 critical ethnography 50–1 auditing 13–15 theoretical sampling 299 critical rationalism 115 autoethnography 15–16 complex evaluation 170 critical realism 93, 255, 256, 257 composite measurement 34–5 critical research 51–2 best practice 130–1 concurrent techniques 179 Cronbach’s alpha 263 bias 12–13, 17–18, 95–6, concurrent validity 314 cross-sectional studies 268, 269 275, 276 confidentiality 35–7, 97, 278 cross-sectional surveys Biographical Interpretive confirmatory research 111, 114 53–4, 211 Method (BIM) 159 consent, informed 97, 149–50 cultural relativism 261 bivariate analysis 316, 317 constant comparative method cultural research 54–5 block sampling 10 33, 37–8, 131–2 construct validation 314 dangerous fieldwork CAPI 21, 157 constructionism 38–40 49, 56–7 CAQDAS 19–20 constructive alternativism 263 data 57–8 see also software constructivism 38, 39, 93, 94, 105 archives 58–60, 247–8, 274 cardinal variables 316–17 constructs 200, 263–4 attrition 12–13 case studies 20–1, 72, content analysis 33, 40–1, dredging 136–7, 146 201, 244–5 72, 80 naturalistic 191–3 Subject-Index-Jupp-3373.qxd 3/8/2006 5:55 PM Page 328

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

decision-making research ethics 96–8, 278 fallibilists 155 60–2, 204 access 1–2 falsification 91, 92–3, deconstruction 62–4 anti-racism research 7 115–16, 139 deduction see hypothetico- archives 248 feasibility studies 110 deductive model covert research 49, 102 feminist research 116–18, 228 deductive–nomological disaster research 73 feminist standpoint explanation 47–8 experiments 107 epistemology 288, 289 demography 65–6 exploitative research 108–9 field experiments 118–19, 171 descriptive statistics informed consent 149–50 field notes 120, 215 66–7, 147–8 life history interviews 160 fieldwork 56–7, 119–21 descriptive studies 53, 166, netnography 194 fishing 136–7 167, 284 organizational research 209 focus groups 121–3, 237 deterministic models 181 participant observation 216 forecasting 123–4 deviant case analysis 3–4, 67–8 trace measures 303 Frankfurt School 51 diaries 68–70 unobtrusive measures 310 digital photography 223 The Ethnograph garbology 125, 242 direct magnitude (version 5) 98–9 gatekeepers 1, 49, 97, estimation 273 ethnographic 126, 264–5 directed photography 223 interviewing 99–101 general linear modelling disability research 70–1 ethnography 101–3, 176, 192 5, 127–9, 184 disaster research 71–3 autoethnography 15–16 GIS (geographical information disastrous research 73–4 covert research 49 systems) 129–30 discourse analysis 74–6, 153–5 critical 50–1 good practice studies 130–1 discrete systems 181 microethnography 177–9 grounded theory 37, 92, 111, discrete variables 57, 317 netnography 193–5 131–2, 146, 241, 298–9 discriminant function analysis primary research 237 76–7, 184 progressive focusing 240–1 hermeneutics 133–4, 297, 298 discriminant validation 314 validity 313 historical analysis 33, 134–6 discrimination, racial 6 videography 318–20 historiographic methods 72 discursive psychology 77–9 visual methods 320–1 hunting 136–7 document analysis 40–1, ethnomethodology 103–4 hybrid systems 181 79–81, 297–8 evaluation research 2, 104–6 hypotheses 115–16, 137–8 drawings 320–1 evidence-based policy hypothetical induction 146–7 111, 112, 225 hypothetico-deductive ecological fallacy 45, 82–3 evolution, anti-racism research 6 model 138–40, 146 ecological validity 311, 313 experimentalists 105, 106 econometrics 83–4 experiments 106–8, 237 ideal type 141–2 economics, applied research 9 field 118–19, 171 idealism 93, 255–6, 257 eidetic reduction 221 laboratory 107, 118–19 ideographic 142 elite interviewing 85–6 quasi 33, 107, 164, idiographic research 142–3, 196 elite selection 87–8 171, 251–2 image work 321 emancipatory research explanatory surveys 284 impression management 143–4 70, 71, 88–90 exploitative research 108–10 indicators 144–5 emancipatory theory 176–7 exploratory factor analysis 114 multiple 182–3 empiricism 90–2, 93, 117, exploratory research 20, 21, performance 218–20 255–6, 257 110–11, 166–7 social 200, 282–4 enumerative induction 146 extending social induction 139, 146–7 epistemology 92–4, 175, 176, research 111–13 analytic 3–4 202, 288–9 external validity see validity inferential statistics 66, Erklären 156 of generalization 147–9, 317 error 94–5, 148 informed consent 97, 149–50 bias 18 F test 46, 214 institutionalized racism 7, 8 Type 1 and 2 5, 138, 307–8 face validity 145, 314 intellectual estimation 95–6 factor analysis 114–15 craftsmanship 150–2

328 Subject-Index-Jupp-3373.qxd 3/8/2006 5:55 PM Page 329

SUBJECT INDEX

inter-subjective meta-narratives 232 oral history 206–8, 268 understanding 155–7 metadata 60 ordinal variables 168, 292, interactional analysis 187–8 methodological pluralism 174–5 316, 317 interactionism 176 methodology 175–7 organizational research 208–10 internal consistency micro models 182 overlay mapping 129 reliability 263 microanalysis 187 overt research 102, 216 internal validity see validity microethnography 177–9 of explanation mixed methods research paired comparisons 272–3 Internet 171, 179–81, 238 panel studies 211–12, 243 netnography 193–5 mixed-model research 180 paradigms 212–13, 261 research 152–3 modelling 181–2, 184, 280 parametric tests 168, 214 interpretive approach 256 multi-stage sampling 10 participant observation 48–9, interpretive repertoires 153–5 multiple indicators 182–3 101–2, 122, 214–16, 237 interval estimation 96 multiple regression 22, 23, 96 participatory action research interval measures 168, 292 multivariate analysis 66, 76, 32, 216–17 interviews 121–2, 157–8 163–4, 184–5, 316, 317 participatory research computer assisted 21–2 myths 297–8 70, 71, 90 elite 85–6 path analysis 22, 23, 217–18 ethnographic 99–101 narrative analysis 186–9 performance auditing 13, 14 life history 100, 159–61 narrative interviewing 189–91 performance indicators 218–20 narrative 189–91 natural accretion performative analysis 188 primary research 237 measures 303 personal constructs 263–4 natural erosion measures 303 phenomenology 156, laboratory experiments naturalistic data 191–3 213, 220–2 107, 118–19 netnography 193–5 photo-elicitation 223 least squares estimation 95, 96 network analysis 195 photography 222–4, 320–1 life history interviewing New Public Management pilot studies 110 100, 159–61 (NPM) 13 pluralism, life tables 65 nominal (categorical) measures methodological 174–5 Likert scale 58, 161–2, 295 168, 292, 316, 317 policy-related research 224–5 literature reviews 162–3, 324 nomothetic approaches politics and social log-linear analysis 163–4 142–3, 195–6 research 226–8 logical positivism 90, 91, 230 non-parametric tests 168, 214 population validity longitudinal studies 164–5 non-probability sampling 196–7 311, 312–13 attrition 12 purposive 244–5 positioning theory 228–9 cohort study 31 quota sampling 196, 197, positivism 175–6, 202, panel studies 211–12, 243 253–4, 290 229–31, 250, 251 secondary analysis 274, 275 snowball sampling 88, 196, bias 17, 18 time series 31–2, 301–2 197, 281–2 logical 90, 91, 230 volunteer sampling 196, opportunity sampling 206 macro models 182 197, 322–3 paradigms 212–13 mapping 66–7 NUD*IST 197–8 postmodernism 94, 118, 176, marketing research 166–7 numerical taxonomy 184–5 203, 231–3 maximum likelihood NVivo 198, 246–7 practitioner research 233–5 estimation 96 pragmatism 180, 279 mean square error 95 objectivity 17–18, 117, prediction 235–7 measurement 168–9 118, 156 predictive validity 314 bias 17 official statistics 199–201 preference 43 composite 34–5 one-shot design 201–2 prejudice, racial 6, 7 validity 74, 311, 314–15 ontology 92, 175, 176, 202–3 primary research 237–8 media analysis 169–70 operational research principal components messy research 170–2 management 203–5 analysis 114 meta-analysis 172–3, 274 operationalization 205 privacy 97 meta-evaluation 13 opportunity sampling 205–6 probability 91, 148, 235

329 Subject-Index-Jupp-3373.qxd 3/8/2006 5:55 PM Page 330

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

probability (random) risk research see prediction social representations theory 39 sampling 29, 95, 206, Royal Statistical Society 199 social surveys 91, 92, 284–5 238–9, 271, 322 sociograms 285 inferential statistics 147, 148 sample size 12–13, 44, 148, 272 sociology stratified 290–1 sampling 271–2 applied research 9–10 probative induction 147 see also non-probability of scientific knowledge process mapping 239–40 sampling; probability (SSK) 39 programme auditing 13 (random) sampling sociometry 195, 285–6 progressive focusing 240–1 accidental 196–7 software projective technique 241–2 area 10 Atlas.ti 10–12 proliferation principle 261 bias 17 CAQDAS 19–20 proportional hazards model 65 block 10 The Ethnograph proportionate sampling 239 cluster 10, 29, 239 (version 5) 98–9 prospective studies 242–4 convenience 196–7 psychology, applied research 9 multi-stage 10 NUD*IST 197–8 purposive sampling 244–5 opportunity 205–6 NVivo 198, 246–7 proportionate 239 SPSS 287–8 QSR NVivo 198, 246–7 purposive 244–5 space syntax 286 Qualidata 59, 248, 274 quota 196, 197, 253–4, 290 spatial research 286–7 qualitative data archives 247–8 random 29 SPSS 287–8 qualitative research snowball 88, 196, 197, 281–2 standard error 95–6 237–8, 248–50, 251 stratified 290–1 standpoint feminism 117–18 quantitative research theoretical 132, 196, standpoint research 288–90 237–8, 250–1 197, 298–9 stereotyping, racial 6 quasi-experiments 33, 107, 164, volunteer 196, 197, 322–3 stochastic models 181–2 171, 251–2 scaling 272–4 stratification 239 questionnaires 21–2, 252–3 Likert scale 58, 161–2, 295 stratified sampling 290–1 quota sampling 196, 197, semantic differential structural analysis 187 253–4, 290 scale 277–8 structuralism 93 summated 294–6 structured observation racism, anti-racism research 6–8 Thurstone scale 301 237, 291–2 random sampling see scientific method 90–1 subjectivity 293–4 probability sampling scripts 77–8 summated rating scales 294–6 randomized controlled trials secondary analysis 274–5 surveys (RCTs) 105 self-disclosure 144 CASIC 21–2 rank order method 273 self-report studies 275–7 census 25–6 ratio measures 168, 292 semantic differential cluster sampling 29 rationalism 93 scale 277–8 coding 30 Rayner reforms 200 semiology, documents 80 comparative method 33 reactivity see validity semiotics see textual analysis cross-sectional 53–4 of measurement sensitive research 278–9 disaster research 72 realism 48, 93, 105–6, 140, sensitivity analysis 281 primary research 237 250, 255–7 sensitizing concepts 279–80 sampling 271–2 reflective practice 234 sequential mixed methods social 91, 92, 284–5 reflexivity 151, 258–9 research 179 survival analysis 65 regression analysis 259–60 significance see symbolic interactionism 100, 213 relativism 260–2 inferential statistics reliability 165, 262–3 signs, textual analysis 297 t-test 4, 46, 214 repeated measures 164, 165 simulation 181–2, 280–1 telegraphy 169 repertory grid technique 263–4 snooping 136–7 television 169–70 research bargains 264–5 snowball sampling 88, 196, test–retest reliability 262 research design 265–6 197, 281–2 texts 300 research problems 266–8 social constructionism 38, 39 textual analysis 297–8 response sets 277 social constructivism 39 thematic analysis 186–7 retrospective studies 268–70 social indicators 200, 282–4 theoretical generalizability 313

330 Subject-Index-Jupp-3373.qxd 3/8/2006 5:55 PM Page 331

SUBJECT INDEX

theoretical sampling 132, 196, univariate analysis 66, 316, 317 value-free research 315–16 197, 298–9 unobtrusive measures 191, 242, variable analysis theory-driven evaluation 105 303, 309–10 91, 316–17 thick description 120, 300–1 variance Thurstone scale 301 validity 311 ANOVA 5, 127 time series design 31–2, 301–2 applied research 9 estimation 95 trace measures 125, 302–4 disastrous research 73–4 verification 91, 115, 138, transcendental reduction 221 of explanation 105, 108, 139, 280, 299 transcription 304–5 311–12 Verstehen 93, 133–4, 156, 213, transformational techniques face 145, 314 279, 293, 317–18 179, 180 of generalization 105, 311, video, visual methods 320–1 trend designs 301 312–13 videography 318–20 triangulation 180, 305–7, 315 induction 147 violence, racial 6 turn-taking 42–3 longitudinal studies 165 visual methods 320–2 Type 1 and 2 errors of measurement 74, volunteer sampling 5, 138, 307–8 311, 314–15 196, 197, 322–3 multivariate analysis 184 UK data archive (UKDA) 59 reflexivity 258 Ward’s algorithm 28 UK Economic and Social Data self-report studies 275, 276 writing research 324–6 Service (ESDS) 59 simulation 280–1 unidimensionality 314 summated rating scales 296 Yule’s Q 46

331 Author-Index-Jupp-3373.qxd 3/8/2006 5:54 PM Page 332

Author Index

Abrahamson, M. 253, 254 Blackman, M.B. 160 Chaiklin, S. 156 Adams, J. 236 Blair, T. 225 Chalfen, R. 222 Adorno, 51 Blalock, H. 91 Chalmers, D. 293 Ageton, S. 276 Blank, K. 61 Charemza, W. 84 Albrow, M. 9 Blaug, M. 116, 139 Childress, J.F. 35 Altheide, D.L. 17 Blaxter, L. 162 Churchill, G.A. 166 Anthony, D. 288 Blee, K.M. 7 Churchland, P. 293 Ashmore, M. 39 Blumer, H. 213, 279–80 Cilliers, P. 233 Atkinson, P. 1, 102 Bobo, L. 7 Clark, J. 188 Atkinson, R. 159 Bochner, A. 15 Clayman, S.E. 68, 192 Austin, 63 Bogen, D. 104 Clifford, J. 16 Bosnjak, M. 153 Coffey, A. 2, 144, 216 Back, L. 227 Bott, E. 195 Coleman, A. 287 Bacon, 139 Bowes, A. 160 Collins, P. 89 Bakeman, R. 291 Box, S. 276 Colombo, A. 142, 236 Bakhtin, 155 Boyne, R. 294 Comaroff, J. 50 Banks, J. 181, 182, 281 Brady, A. 206, 248 Comte, A. 213, 230 Banks, M. 223 Brady, H.E. 36 Copas, J. 236 Banton, M. 6 Brentano, F. 220 Corbin, J. 37, 298 Barker, R.G. 191 Brockington, D. 47, 308 Corti, L. 60 Barnes, C. 71 Bruner, J. 39, 156 Cresswell, J.W. 179 Barron, L. 163, 203, 213 Bryman, A. 20, 92, 97, 126, 202, Cross, M. 10, 225 Barthes, R. 297–8 212, 237 Crossen, C. 227 Bateson, P. 291 Bulmer, M. 22, 30, 168, 208 Crow, I. 45, 83, 148 Baudrillard, J. 232 Burgess, R.G. 157 Bauman, Z. 318 Burton, D. 162, 163 Dahl, R.A. 88 Beauchamp, T.L. 35 Bynner, J. 32, 53, 165, 212 Dar, N. 160 Bechhofer, F. 224 Byrne, D. 185, 201, 230, Dartington, Y. 112 Becker, H.S. 24, 52 233, 260 Darwin, C. 6 Begg, I. 9 Bywaters, P. 111, 112 Davidson, J. 49, 197, 239, Behar, R. 100 272, 279 Belk, R.W. 320 Cain, M. 94, 118 Davies, B. 229 Bell, C. 174 Camburn, D. 159 Davies, H.T. 131, 225 Berger, P.L. 38, 221 Campbell, D. 252, 309 Davies, P. 111, 150, 158, Berkeley, 90 Capaldi, D. 12 266, 304 Bernhardsen, T. 129 Carley, M. 283 Davies, R. 31 Bertaux, D. 159–60 Carmichael, S. 7 de Laine, M. 1, 144 Bhaskar, R. 48, 93, 140, 256 Carr-Hill, R. 283 de Vaus, D. 108, 244, 269, 285 Bickman, L. 8, 9 Cashmore, E. 170 Deadman, D.F. 84 Billig, M. 77, 78, 154 Cavan, R. 159 Deegan, M. 120 Birks, D.F. 166 Cellier, F.E. 181 Delanty, G. 231 Author-Index-Jupp-3373.qxd 3/8/2006 5:54 PM Page 333

AUTHOR INDEX

Demers, M.N. 129 Gellner, E. 261 Hill Collins, P. 289 Dennett, D. 293 Gergen, K.J. 38, 39 Hillier, B. 286, 287 Denscombe, M. 1, 326 Ghiselli, E.E. 35 Hinde, A. 65 Derrida, J. 62, 63–4, 232, 298 Giddens, A. 89 Hobbs, D. 57, 102, 120, 121 Descartes, R. 293 Gilbert, N. 154, 182 Hodgkinson, G.P. 209 Desrosières, A. 199 Gillies, D. 90 Holdaway, S. 49 Devault, M. 159 Gillman, M. 235 Hollway, W. 159 Dey, I. 146, 299 Gitlin, A.D. 18 Holstein, J.A. 100, 189 Dilthey, W. 133–4, 317–18 Glaser, B.G. 37, 102, 110, 111, Holsti, O.R. 41 Dohrenwend, B. 276 131–2, 146, 241, 298, 299 Hooghe, L. 61 Dovidio, J.F. 7 Glasziou, P. 172 Horkheimer, M. 51 Duelli Klein, R. 117 Goffman, E. 42, 143, 188 Horwood, J. 265 Durkheim, E. 139, 213, 230, 294 Gold, M. 276 Hudson, B. 52 Gold, R. 102 Hume, 23, 90, 147 Edgar, I. 321 Goode, S. 206 Humphreys, J. 97 Edley, N. 154 Goodwin, C. 177, 192, 305 Humphreys, L. 102 Edwards, D. 39, 68, 77, 192, 305 Goodwin, M.H. 192 Hunter, 87–8 Elliot, D. 276 Gottman, J.M. 291 Husserl, E. 213, 220–1, 293 Ellis, C. 15 Gouldner, A. 52 Hutchby, I. 146 Emerson, R. 120 Grasseni, C. 321 Empey, E. 276 Gray, A. 54 Israel, M. 37 Erikson, M. 276 Greco, L. 43 Evelyn, J. 69 Greenwood, D.J. 2 Jaccard, J.J. 183 Everitt, B.S. 28 Guba, E.G. 13 Jary, D. 48, 116, 140 Gubrium, J.F. 100, 189 Jefferson, G. 42, 43, Faraday, A. 159, 160 Gundlach, J. 83 191, 304–5 Farrington, D. 82, 83, 276–7 Guy, W. 200 Jefferson, T. 159 Fay, B. 156, 203 Johnson, J. 17, 100 Ferrarotti, F. 160 Haas, E. 61 Johnson, P. 55 Feyerabend, P. 116, 261 Habermas, J. 64 Johnson, R.A. 250 Fielding, N.G. 101, 160, 277 Hakim, C. 20, 130, 224 Jones, S.G. 152 Fine, M. 50 Halfpenny, P. 139 Jones, T. 232–3 Fischer, M. 300 Hall, S. 80 Junger, M. 276 Fiske, J. 170 Halpern, E.S. 13 Jupp, V. 10, 12, 23, 33, 34, Flick, U. 4, 37, 38, 306 Hamilton, C.V. 7 67, 74, 77, 125, 137, 138, Fook, J. 234 Hammersley, M. 1, 18, 24, 52, 152, 164, 195, 196, 205, Foster, J.J. 27, 115, 214, 288 104, 134, 146, 147, 227, 218, 242, 258, 263, 264, Foucault, M. 75, 93, 154, 260–1 241, 249, 280, 299 267, 275, 277, 285, 301, Fowler, R.G. 309 Hanson, J. 286, 287 302, 311, 312, 323 Frank, G. 160 Harding, J. 26, 132, 291, 317 Freedman, D. 159 Harding, S. 17, 93, 117, 202, 289 Kant, I. 220, 221, 293 Freeman, D. 100 Harper, D. 223, 321 Keddie, V. 21, 131, 220 Freire, P. 217 Harré, R. 222, 229, 260 Kelly, A. 29, 202 French, S. 71 Harvey, D. 203 Kelly, G. 38, 263–4 Fujimoto, R. 281 Harvey, L. 52 Kelly, L. 12–13, 89 Hegel, G. 294 Kendon, A. 177–8 Gadamer, H.-G. 134 Hempel, C. 47, 48, 138 Kerr, A.W. 288 Gaertner, S.L. 7 Henry, S.G.B. 9 King, R. 177 Gardner, P. 136, 208 Hepburn, A. 40, 64, 78 Kirk, R.E. 5 Garfinkel, H. 42, 67, 103, Heritage, J. 192 Kleck, G. 276 104, 222 Hess, D. 120 Kohli, M. 159 Garwood, J. 58, 161, 251, 292, Hewson, C. 152, 153, 180, Kozinets, R.V. 193, 194, 320 303, 310 238, 275 Kratz, C. 223 Gayle, V. 147 Heyl, B. 99, 100–1 Krausz, M. 260 Geertz, C. 120, 300 Heywood, I. 129 Kuhn, T. 139, 203, 212, 261

333 Author-Index-Jupp-3373.qxd 3/8/2006 5:54 PM Page 334

THE SAGE DICTIONARY OF SOCIAL RESEARCH METHODS

Kumar, R. 162 Marks, G. 61 Patterson, G.R. 12 Kvale, S. 17, 100, 101 Marsh, C. 57, 284, 285 Pawson, R. 91, 230 Marshall, P. 236 Payne, G. 10, 32, 175, 225, 284 Labov, W. 187 Martin, P. 291 Payne, J. 175, 283 Lakatos, I. 91 May, T. 232 Pearson, G. 49 Lampard, R. 66 Mayall, B. 289 Pedhazur, E.J. 128 Langeheine, R. 28 Mayhew, H. 120 Pepys, S. 69 Langellier, K. 188 Maynard, D. 68, 192 Peters, R.S. 24 Lasswell, H. 40 Maynard Keynes, J. 225 Peterson, E. 188 Lather, P. 17 Mead, G.H. 213, 280 Pettigrew, M. 227 Lave, J. 156 Merleau-Ponty, M. 220 Philips, D. 91, 230 Law, I. 7 Merritt, C.B. 309 Piaget, J. 38 Lawrence, S. 7 Merton, R. 169 Picou, J.S. 36 Layder, D. 249 Michels, 87 Pierson, P. 61 Lazarsfeld, P. 169 Miles, S. 55 Pilcher, D.M. 250 LeBaron, C. 178, 179 Mills, C. Wright 87, 110, 150–2, Pink, S. 224, 322 Lee, R.M. 1 162, 189, 230 Plato 93, 293 Lehmann, T. 52 Mills, S. 75 Plummer, K. 159, 160 Leicester, G. 225 Mishler, E. 100, 101, 188, 189 Pollitt, C. 219 Letherby, G. 90, 109, 111, Mitchell, J.C. 67 Poole, H. 46, 240 112, 289 Moon, G. 265 Popkin, S. 37 Lévi-Strauss, 93, 300 Moravcsik, A. 61 Popper, K. 47, 91, 93, 115, 116, Levin, M. 2 Moreno, J. 195, 285 138, 146, 257 Levitas, R. 200 Morgan, D.L. 122 Postman, N. 170 Lewin, K. 217 Mosca, 87 Potter, J. 7, 39, 74–5, 77, 78, Likert, R. 161 Moscovici, S. 39 154, 191, 192 Lincoln, Y.S. 13 Moss, R.B. 254 Power, M. 14 Lindblom, C. 225 Moyser, G. 86, 87, 88 Pratt, M.L. 15 Lindesmith, 146 Mulkay, M. 154 Presthus, 88 Locke, 90 Muncie, J. 52, 76 Prior, L. 79, 80 Lopes, A. 217 Musgrave, A. 91 Procter, M. 274, 275 Lowman, J. 37 Psathas, G. 68 Luckmann, T. 38, 221 Nagel, T. 293 Punch, K. 267 Luhmann, N. 257 Nancy, M. 236 Putnam, R.D. 87 Lynch, M. 104 Nelder, J.A. 127–8 Pye, P.E. 84, 124, 173, 204 Lynd, R.S. 230 Neter, J. 128 Lyotard, J.F. 232 Newby, H. 174 Ragin, C.C. 24 Newman, O. 286, 287 Ranson, S. 219 McConahay, J.B. 6 Newton, J. 3, 10, 225 Reason, P. 3 McCullagh, P. 127–8 Norris, N. 227 Reed-Danahay, D. 15, 16 McGivern, Y. 167 Nunnally, J.C. 295 Reinharz, S. 101, 109 McHale, B. 203 Reips, U.-D. 153 McLaren, C. 268 Oakley, A. 91, 109 Renner, K.-H. 29 McLaren, P. 50 Oliver, P. 245, 282 Rescher, N. 147 McLaughlin, E. 8 Orr, L. 107 Richards, L. 19, 198, 247 McLaughlin, R.H. 36 Osgood, C.E. 277 Ricoeur, 300 McLean, C. 62, 253, 316 Riessman, C.K. 189, 190 McLuhan, M. 169 Paccagnella, L. 194 Riis, J. 120 Macpherson, W. 7 Page, E.C. 130 Roberts, H. 174 Malhotra, N.K. 166 Palys, T. 37 Roberts, J. 32 Malinowski, B. 99, 195 Pareto, 87 Robinson, W.S. 82 Mann, C. 152 Park, R. 120 Robson, C. 58 Marcus, B. 35, 183 Parker, 154 Rog, D.J. 8, 9 Marcus, G. 300 Parsons, S. 32 Rogers, R.E. 142 Marcuse, 51 Paterson, L. 224 Rogerson, M. 310

334 Author-Index-Jupp-3373.qxd 3/8/2006 5:54 PM Page 335

AUTHOR INDEX

Rose, D. 243 Sohng, S.S.L. 217 Uhrmacher, A. 182, 281 Rossi, P.H. 9 Solomos, J. 227 Rost, J. 28 Spector, P.E. 273, 296 Valentine, P. 300 Rowland, D.T. 65 Speer, S. 192 Van den Berghe, W. 219 Rowntree, D. 147 Spradley, J. 100 Van Maanen, J. 325 Rudestam, K. 162 Stacey, J. 89, 109 Vygotsky, L.S. 39, 228 Rushton, J.P. 183 Stack, S. 83 Russo, M. 252 Stallings, R.A. 73 Wagstaffe, M. 87 Rutherford, A. 5, 128 Stanley, L. 89, 118 Wallerstein, I. 143 Ryle, G. 293, 300 Starkey, K. 209 Warren, J.W. 7 Staufenbiel, T. 95, 96 Waterman, H. 2, 3 Sacks, H. 42, 43, 104, 191–2 Stebbins, R.A. 110 Wax, M. 120 Sapsford, R. 12, 145, 172, 177, Steffens, L. 120 Webb, E.J. 191, 303, 309 313, 315 Steventon, G. 130, 287 Weber, M. 93, 133, 141–2, 213, Saunders, M.N.K. 126, 208, Stewart, D.W. 274, 275 256, 257, 317–18 210, 265 Stewart, F. 152 Weiss, C. 225 Saussure, H. 297 Stewart, J. 219 Weiss, L. 50 Sayer, A. 256 Stewart, K. 50 Wengraf, T. 159 Scarce, R. 37 Strauss, A.L. 37, 102, 110, Wertsch, J.V. 156 Scheflen, A. 177 111, 131–2, 146, 241, Wetherell, M. 6, 7, 74–5, 77, Schegloff, E.A. 42, 43, 68, 191 298, 299 154, 155 Schofield, J.W. 299 Stuart, A. 136 Wharton, C. 81, 143 Schofield, R.S. 65 Suck, R. 95, 96 Whyte, W. 49 Schön, D. 234 Sumner, M. 13, 94, 98, Wichern, D.W. 250 Schutz, A. 156, 213, 220, 221, 118, 249, 318 Wilkinson, J. 292 256, 257, 262 Swain, J. 71, 235 Williams, M. 92, 231, Schuurman, N. 129, 130 232, 233 Schwandt, T.A. 13, 14, Tannen, D. 228 Willis, P. 50 51, 156 Tarling, R. 277 Winch, P. 142 Schwartz, R. 309 Tashakkori, A. 180 Wise, S. 89, 118 Scott, D. 112 Taussig, M. 50 Wittgenstein, L. 182 Scott, J. 41, 69, 80, 87, 298 Tavris, C. 250 Wojciech, W. 84 Seale, C. 249 Te Molder, H. 77–8 Wolman, H. 130 Searle, 63 Teddlie, C. 180 Wooffitt, R. 146 Sears, D. 6, 7 Thomas, B. 55 Wright, G. von 48 Sechrest, L. 309 Thornton, A. 159 Wrigley, E.A. 65 Seidel, J. 98 Thurstone, L.L. 301 Selvin, H.C. 136 Tilley, N. 106, 119, 252 Yin, R.K. 20–1 Silver, C. 12, 99 Townsend, P. 283 Young, T.R. 52 Silverman, D. 249, 271 Tranfield, D. 209 Yow, V. 1 6 0 Sim, D. 160 Troitzsch, K.G. 182 Smallbone, T. 167 Trussell, J. 65 Zeigler, B.P. 181, 182, 280, 281 Smith, J.K. 238 Twine, F.W. 7 Zetterberg, H. 139 Smith, M.J. 256, 257, 261, 262 Tyler, S. 232 Znaniecki, F. 4, 146

335