New Directions for L2 Research

NEW DIRECTIONS FOR L2 RESEARCH

Edited by

SARAH RANSDELL

Florida Atlantic University, USA

MARIE-LAURE BARBIER

University of Lyon, France

TABLE OF CONTENTS

PREFACE v Gert Rijlaarsdam

AN INTRODUCTION TO NEW DIRECTIONS 1 FOR RESEARCH IN L2 WRITING Sarah Ransdell & Marie-Laure Barbier

CRITICAL EXAMINATION OF L2 WRITING 11 PROCESS RESEARCH Julio Roca De Larios, Liz Murphy, & Javier Marin

BUILDING AN EMPIRICALLY-BASED MODEL 49 OF EFL LEARNERS’ WRITING PROCESSES Miyuki Sasaki

THE RELATIONSHIPS BETWEEN BILINGUAL 81 CHILDREN’S READING AND WRITING IN THEIR TWO LANGUAGES Aydin Durgunoğlu, Montserrat Mir, & Sofia Ariño-Martin

LINGUISTIC KNOWLEDGE, METACOGNITIVE KNOWLEDGE 101 AND RETRIEVAL SPEED IN L1, L2, AND EFL WRITING: A structural equation modelling approach Rob Schoonen, Amos van Gelderen, Kees de Glopper, Jan Hulstijn, Patrick Snellings, Annegien Simis, & Marie Stevenson

EARLY EXPOSURE TO AN L2 PREDICTS GOOD L1 123 AS WELL AS GOOD L2 WRITING Rosario Arecco & Sarah Ransdell

THE EFFECTS OF TRAINING A GOOD WORKING 133 MEMORY STRATEGY ON L1 and L2 WRITING Sarah Ransdell, Beverly Lavelle, & Michael Levy

A COMPARISON BETWEEN NOTETAKING IN L1 145 AND L2 BY UNDERGRADUATE STUDENTS Martine Faraco, Marie-Laure Barbier & Annie Piolat

AN INTRODUCTION TO NEW DIRECTIONS FOR RESEARCH IN L2 WRITING iii

COLLABORATIVE WRITING IN L2: 169 THE EFFECT OF GROUP INTERACTION ON TEXT QUALITY Folkert Kuiken & Ineke Vedder

INVESTIGATING LEARNERS’ GOALS 179 IN THE CONTEXT OF ADULT SECOND-LANGUAGE WRITING Alister Cumming, Michael Busch, & Ally Zhou

WHEN AND WHY TALKING CAN MAKE WRITING HARDER 179 Margaret Franken & Stephen Haslett

A PROBLEM-POSING APPROACH TO USING 179 NATIVE LANGUAGE WRITING IN ENGLISH LITERACY INSTRUCTION Elizabeth Quintero

REFERENCES 179

AUTHOR INDEX 179

SUBJECT INDEX 179

LIST OF CONTRIBUTORS 179

PREFACE

GERT RIJLAARSDAM

University of Amsterdam & Utrecht University, the Netherlands

Multilingualism is becoming the default in our global world. The present-day global citizens use different languages in different situations. Apart from their mother tongue, they learn languages that give them access to other regions, nations, and worlds. In all countries of the European Union, for instance, at least one foreign language is mandatory in secondary schools. Most students are taught English as a foreign language, the lingua franca in Europe. In large parts of the USA, students move from Spanish to English schooling. In parts of Canada, bilingual education is standard. In Catalonia (Spain) children learn Catalonian and Spanish, in Hong Kong English and Chinese. The smaller the world becomes, the more languages are used and learned. For writing process research, this development into multilingualism entails at least two challenges. First of all, studying the relation between writing in L1 and L2 provides an opportunity for collaborative studies, in different language settings. Second, the issue of generalization of findings comes to the fore. It becomes evident now that we have unjustly neglected this issue in writing process research. We for- got to ask whether it is feasible to talk about ‘writing processes’ in general, without referring to the language of the written texts, and without taking into account the educational and linguistic culture in which these texts originate. If it is true that writing processes are – to some extent – linguistically and culturally bound, then the implication is that our L1 (and L2!) process studies have a limited scope. Strangely enough, the issue of linguistically and culturally bound writing processes has been disregarded for a long time. However, now our alarm bells are starting to ring, as soon as we consider the variable of ‘text quality’. For how do we define quality of text?1 In Europe, the definition of a good argumentative text is deeply embedded in the various cultures. In short, and at the risk of overgeneralizing: the German argumentative text is a philosophical personal essay, the French argumenta-

1 Note that the IEA study in the 80s in which the quality of writing performance in several countries was compared, met large rating problems to reach a satisfactory reliable and valid level for an international report.

VI RIJLAARSDAM tive text is defined by logical rationalism, and the British argumentative text is an empirical deductive text. These different versions of what constitutes a ‘good’ text are easily recognizable, even in contributions to European international journals.

Let’s have a look ahead. In the near future, all students in European secondary education are taught at least one foreign language, in most cases two languages. One of these languages will be a variety of English. Via this variety, students will be able to communicate with other persons from other linguistic regions. As a consequence, all kinds of ‘Englishes’, ‘Spanishes’, ‘Chineses’ will come into existence. The master- ing of at least one international language will be decisive in the near future for ob- taining interesting jobs. As a result of this language movement, most students not only will learn to speak ‘English’, but also to write in ‘English’. And at this point writing researchers meet again. Here we have something in common, something that binds us, irrespective of the country we live in. How do students in France, Italy, Spain, Germany, and the Netherlands connect their L1-writing-processes to their L2- writing-processes? Are there linguistico-cultural particularities, to what extent is the acquisition process a general cognitive process, independent from the particular L1 and the cultural schools of thought about what defines a ‘good text’?

With the growing awareness that the global will be a multilingual one, in part domi- nated by varieties of English (or Spanish? Or Chinese?), we also stay acutely aware of the particularities of the various L1-situations and the limitations of generalization. As researchers, we seize the opportunity to develop research programs on writing processes in L1 and L2, to be carried out in different language environments/settings. These studies may reveal much about the extent in which L1- processes are influenced by linguistico-cultural factors, and at the same time, about the different ways in which students in various countries cope with L2 writing. This 11th volume in the series Studies of Writing provide us with research para- digms and findings from various regions, showing us that in different countries similar but different linguistic situations have been tackled. Readers will find a variety of research designs and techniques to study the relation between writing in different languages. I hope this volume will inspire many readers to study the L1 and L2 relationship in their particular environment.

DEDICATION

The editors of this volume would like to dedicate the book to their beloved children. Sarah Ransdell would like to dedicate the volume to Kirsten Elna Laursen and Erik Ejvind Laursen. Marie-Laure Barbier would like to dedicate the volume to Isia Mokhtar.

AN INTRODUCTION TO NEW DIRECTIONS FOR RESEARCH IN L2 WRITING

SARAH RANSDELL* & MARIE-LAURE BARBIER**

*Florida Atlantic University, USA, **University of Lyon, France

Abstract. The introduction gives an overview of the divergent research represented in the eleven chapters of this volume, including: A comparison of models of L1 and L2 writing; the parallel development of reading and writing skill; the impact of specific techniques to train L2 writing skill; note-taking and goal formation in L2 writing; metalinguistic awareness; peer interaction; and a problem-solving method for teaching L2. Psycholinguistic, linguistic, and pedagogically, based research findings in each chapter add to our theoretical understanding of the subject and provide implications generated by the research. This chapter discusses the complementarity of these research approaches, followed by a preview of the rest of the chapters, and finally presents two important questions for studying L2 writing. A critical review of new research is presented in Chapter 1. Then chapters 2, 3, 4, and 5 focus mainly on modeling the correlations between L1 writing skill components and L2 skill development and transfer. Chapters 6 through 11 discuss writers’ general cognitive abilities, resources, and goals in L1 and L2. These chapters include empirical research relevant to both educational applications and theoretical advances.

Keywords: bilingualism, cognitive processing, education, English as a second language, foreign language writing, linguistics, psycholinguistics, research methods in psychology, second language (L2) writing, think-aloud protocols.

1 APPROACHES TO L2 WRITING STUDY

1.1 Pedagogical, linguistic and psycholinguistic as complementary approaches Second language (L2) writing encompasses a remarkably complex and variable set of behaviors. L2 writing research substantially benefits from the wide range of complementary approaches that have been used to study it. The majority of studies have concentrated on pedagogical issues in learning to write in L2 (e.g., Silva & Matsuda, 2001). Other prominent descriptive approaches focus on linguistic features of written texts in L2 within an educational setting (Genesee, 1994). Psycholinguis-

S. Ransdell & M.-L. Barbier (2002). An introduction to new directions for research in L2 writing. In G. Rijlaarsdam (Series ed.) & S. Ransdell & M.-L. Barbier (Volume eds.), Studies in Writing, Volume 11: New Directions for Research in L2 Writing, 1 – 10 . © 2002. Kluwer Academic Publishers. Printed in the Netherlands.

2 RANSDELL & BARBIER tic approaches focus more specifically on the cognitive processes involved in L2 writing activity. Together, pedagogical, linguistic, and psycholinguistic approaches give us a more complete understanding of L2 writing than any single approach can. The purpose of this book is to present new research directions in psycholinguistic, pedagogical, and linguistic accounts of L2 writing. Case studies and other ecologically-rich methods are primary research procedures used in both pedagogical and linguistic research. Behaviors are studied in broad contexts. Social and emotional factors including motivation, goals, and social interactions are given special consideration. Accordingly, pedagogical and linguistic approaches generally take into account the variety among L2 writer/participants, i.e., specific second languages, learning environments, and participants’ interest and motivation to write in L2. Also, groups of L2 writers studied are comprised of participants who may be of different ages and have various educational experiences. Be- cause the writer’s interest in L2 writing is often needed to succeed in an academic context where the foreign language is a necessity, this is a primary consideration attended to in these perspectives (Leki, 2000). Researchers in the pedagogical and linguistic perspectives devote less attention to what goes on cognitively as particular learners formulate text in L2. More attention is given to devising ways to promote progress in L2 writing. These approaches are essential because they place the findings from experimental studies in psycholinguistics within a broader L2 writing context. Factors relative to both the learners, and social settings where the learning takes place, are important additions to traditional psycholinguistic accounts conducted in laboratory settings. The psycholinguistic approach uses the experimental method. Experimental methodology is the only way to identify the causal determinants of written production in L2 with reliability. But, when evaluating such research, it is necessary to take into account the fact that experimental studies are limited in as much as they typically evaluate participating L2 writers who have the same common native language and general educational experiences and achievements. L2 writer/participants are also likely to share such personal characteristics such as age and socio-economic backgrounds. Often the chosen L2 writers represent a group assumed to be sufficiently homogeneous in order to make generalizations from large-scale experimental studies. However, within these studies, particular characteristics of individuals are sometimes deemed to be relatively unimportant. Those who evaluate L2 writers’ production are routinely given only incomplete details about the individuals who participate in the research project. The study of bilingualism itself has been concerned with the same issues of identifying the type of bilingual individuals to which results can generalize (i.e., de Groot & Kroll, 1997). Despite potential limits to generalization, the experimental method is essential for the disconfirmation of theories and models. The studies and findings in this volume are designed to take advantage of the strengths of the experimental method most typical of a psycholinguistic approach, but also the pedagogical and ecologically-valid characteristics of the other approaches to L2 writing that are available. NEW DIRECTIONS FOR RESEARCH IN L2 WRITING 3

1.2 Current issues in L2 writing studies From the 1980’s, case studies and experimental studies have been complementary approaches used in L2 writing research to describe the specific skills required for L2 writing and to determine the role of background knowledge transfer from L1 to L2 writing processes. These original psycholinguistic studies were inspired essentially from writing models elaborated within L1 studies (Hayes & Flower, 1980; Levy & Ransdell, 1996; Scardamalia & Bereiter, 1991) whose purpose was to identify the factors determining monolingual learners’ writing performances. Two of these factors have been mainly discussed in the L2 literature: the impact of linguistic knowledge in L2 (Yau, 1991; Sasaki & Hirose, 1996) and general writing expertise (Cumming, 1989; Jones & Tetroe, 1987). In fact, the acquisition of L2 writing ability means much more than just the appropriation of new graphic codes. Written language in L2 is a specific mode of communication that requires new skills and may even lead to a fundamental reorganization of communicative competence. L2 writing requires a sufficient level of lexical, syntactic and spelling knowledge in the target language in order to express ideas in a correct linguistic form. But it also neces- sitates activation and specific control of writing processes, i.e., planning, transcription, and reviewing during the production, in order to achieve writing goals, as in L1. The focus of research was therefore oriented on high-level processes such as planning and revising, as well as metalinguistic knowledge. Metalinguistic knowledge transcends knowledge about language as ‘meaning’ extending to knowledge about language as ‘form’ separable from its meaningfulness. In this perspective, researchers have turned attention to writing strategies according to writing purpose, expertise, and temporal signatures, and to transfer phenomenon between L1 and L2. This leads today to a better understanding of L1 and L2 learners’ writing processes and difficulties. By describing the processes involved in L2 writing settings, researchers were also interested in learning more about the activation and control of writing processes in real time. Beginning in the 1980’s, keystroke-capturing software, graphic tablets, and the use of think-aloud protocols, have been used to extract temporal data and to address questions of the temporal dynamics of cognitive processing. Keystroke- capturing software, i.e. FauxWord (Levy & Ransdell, 1996), allow one to replay the characters typed at the point-of-utterance. Graphic tablets allow for point-of- utterance of handwriting transcriptions rather than keyboarding (i.e. Barbier, 1996). Think-aloud protocols involve asking the writer to generate thoughts aloud that come to mind during the primary task of writing without evaluating one’s processing (Ericsson & Simon, 1993; Silva, 1993). All three of these online process methods allow one to determine the most likely cognitive processes taking place at any given instant. The methods have only recently been applied to L2 writing. From the systematic comparison of writer’s functioning in L1 and in L2, it is generally agreed upon that adult writers refer to their conceptual and discursive knowledge acquired in L1 during their production in L2 (Cumming, 1990; Victori & Lockhart, 1995). Cognitive abilities in L1 and in L2 are intrinsically tied to one another, and if writers are skilled in their native language, they should be able to use general strategies that allow them to control the impact of writing task demands. 4 RANSDELL & BARBIER

These strategies should be particularly stable from one writing session to another (Levy & Ransdell, 1996), and should transfer to the L2 writing situation. Cummins’ (1980) notion of common underlying proficiency and interdependence between languages would suggest this type of transfer. But the question remains to identify the conditions under which writers can use these strategies during their production in a foreign language, despite the limitations imposed by their relative lack of linguistic knowledge (Barbier, 1998a, 1998b; Broekkamp & van den Bergh, 1996; Whalen & Ménard, 1995). Several studies suggest that high-level writing processes, such as attention and memory, can be activated when writing in a foreign language, even though they are probably not adequately co-ordinated with low-level writing processes, such as typing and spelling (Barbier, 1996; Jones & Tetroe, 1987). Moreover, this ability to treat high-level writing aspects would rely on a sufficient level of the writing expertise acquired beforehand in L1 (Cumming, 1989; Zamel, 1983). But the capacity to use effective writing strategies depends on one’s level of linguistic knowledge in the target language, as it is suggested by studies presented in the first chapters of this volume. Many researchers have suggested that a writer’s lexical and syntactic competencies in L2 constitute the principal factors that determine written performances observed (Arndt, 1987; Kobayashi & Rinnert, 1992; Pennington & So, 1993; Sasaki & Hirose, 1996). Successful L2 writing is dependent on the writers’ attention to and control of the linguistic as well as the pragmatic dimensions of the text to produce. In recent years, data concerning the cognitive demands of writing processes in L2 have emerged from studies using both on-line data recording and a dual-task paradigm. The dual-task paradigm involves asking participants to carry out two nearly simultaneous tasks to determine their relative demands. Many chapters in this volume represent this type of analysis. These methods illustrate how general cognitive abilities, such as memory strategy, transfer modalities, and goal setting are involved within L2 writing processes and performances. As it is treated especially in the latter chapters of this volume, pedagogical and linguistic perspectives complement the psycholinguistic perspective. Information obtained from educational settings embedded within an experimental tradition is particularly valuable for studying how adult writers manage their writing processes according to their available cognitive resources. This type of ‘cross-pollination’ has been especially fruitful in the area of L1 writing acquisition. It should best inform models of L2 writing.

2 THE CHAPTERS OF THIS VOLUME While there has been a resurgence of interest in the educational aspects of teaching L2 (e.g. issue 10 of the journal Learning and Instruction) there are relatively few published studies specifically geared to better understanding L2 writing in the psycholinguistic tradition. In the psycholinguistic tradition, researchers have focused on general cognitive processing issues directly related to L1 writing research. These studies have therefore been less likely to investigate such topics as the social context within which L2 writing is learned, and motivation and goals for acquiring a new written language. But the inherent strength of psycholinguistics research is that it NEW DIRECTIONS FOR RESEARCH IN L2 WRITING 5 has always combined converging evidence from psychological experiments with case studies, linguistic interviews, field studies, and correlational and other structural analyses of individual differences. The goal of all of these sources of evidence is to describe, predict, and ultimately explain L2 writing: the correlations between L1 writing skill components and L2 skill development and transfer, but also L2 writers’ general cognitive abilities, resources, and goals. Though all of the chapters in the volume contribute in a number of ways to the discussion of these major issues in the field of L2 writing, each has its own range of topics as well several new and unique themes. What follows is an overview of each chapter in the volume. After this introductory chapter, Roca de Larios, Murphy, and Marin present a critical examination of L2 writing process-oriented research. The authors review a large number of psycholinguistic studies of L1 and L2 writing and highlight implications for further research directions. This chapter is intended as a critical analysis of the most relevant recent research into the cognitive processes underlying L2 composition. After a brief discussion of the most relevant methodological aspects of the research, its main theoretical foci are isolated. Each of these foci, in turn, has allowed for the derivation of a number of research areas under which the studies have been grouped. These include a comparison of skilled and unskilled L2 writers; the development of L2 writing skill; the comparison of L1 and L2 writing processes; and the relationship between general writing ability and L2 proficiency. The findings within each category are presented in a thematic fashion. And, finally, a number of limitations and suggestions for future research are discussed. In the next chapter, Sasaki begins with a discussion of building empirical models of L2 learners’ writing processes. Her research exemplifies the focus of the book in its use of converging measures of writing performance. An analysis of the written text is supplemented by measures of temporal behaviors during writing and think- aloud protocols. Sasaki focuses on the writing behaviors of L2 learners with different levels of L2 writing proficiency and the changes in learners’ writing processes over time. Using multiple data sources, including their written texts, videotaped writing behaviors, and stimulated recall protocols, this study investigated the cognitive processes of Japanese L2 writing experts and novices while writing an argumentative exposition in English. Based on the results, some preliminary models are presented to represent the characteristics of the participants’ writing processes according to their writing expertise. Sasaki insists on the necessity for this kind of study to be complemented by qualitative studies that examine the details of L2 writers’ individual writing processes. Such detailed qualitative studies would be useful for filling in gaps in L2 writing activity models. Next, Durgunoglu, Mir, and Arino-Marti discuss the joint development of reading and writing in L2. Again, a wide variety of measures are collected including those mentioned in Sasaki’s chapter as well as word recognition proficiency, knowledge of syntax, spelling, phonology, and vocabulary and oral proficiency. Dur- gunoglu et al. discuss the relationship between progress in writing activities and the development of linguistic competencies in L2. A cognitive analysis of the writing of fourth-grade Spanish-English bilingual children lead the authors to discuss the links between language, reading and writing variables in the two languages. Eight tests were given at the beginning of the school year to evaluate, in each language, word 6 RANSDELL & BARBIER recognition, spelling, oral proficiency, and syntax. These results were correlated with writing tasks given at the beginning and near the end of the year. According to the results obtained, word recognition and spelling proficiencies were correlated both within- and across-languages, and they were correlated with the form ratings of the writing samples. They did not, however, correlate with oral proficiency measures. A qualitative analysis of the writing samples yielded cross-language transfer effects (in both directions) at the level of graphophonic, syntactic and vocabulary knowledge as well as in terms of story structure. Schoonen, Van Gelderen, De Glopper, Hulstijn, Snellings, Simis, and Stevenson compare an L1 model of higher and lower-order skills to the same outcomes in L2 and L3, with special emphasis on structural differences in the relationships among measures. Schoonen et al. study writing activity according to the linguistic knowledge of students in their second or third language. This study presents an analysis of the Dutch-as-a-second-language (DSL) and the English-as-a-foreign-language (EFL) writing proficiency of secondary school students who learned Dutch as a second language (DSL), compared to students for whom Dutch is their first language (DL1). Their point of departure is a regression model in which writing proficiency is decomposed into constituent subskills: linguistic knowledge, metacognitive knowledge, and speed or fluency of processing linguistic knowledge. Furthermore, these authors investigate whether there are structural differences in EFL writing of these DSL students compared to their DL1 peers. Their hypothesis is that both groups will differ in their Dutch writing, but not as much in their EFL writing. Using structural equation modeling, they test whether the DL1 writing model holds for the DSL writers, and whether the EFL writing model will be the same across the two groups with respect to the explanatory value of the constituent skills. Faraco, Barbier and Piolat consider notetaking activity in L2 as a way to understand the L2 learner’s cognitive resources and strategies within a linguistic task that places constraints on working memory. The notetaker needs to store what is heard in order to transcribe it, while at the same time having to understand the speaker, transform and write down the information. Then, the processing of information and the allocation of cognitive resources require that the notetaker perform complex cognitive activities nearly simultaneously. To identify performances and strategies in such a situation, a comparison is undertaken between native notetakers in French and non-native notetakers more or less fluent in French. The performances and strategies developed by these students during their notetaking are analyzed from three main sources of information. First, a real-time analysis of the relations between the teacher's discourse and note taking by the student is made. Notetaking activity is recorded with a graphic tablet. Second, an analysis is made of the notes produced. And third, an analysis of the comprehension and memorization of the information is made. Ransdell, Lavelle, and Levy reveal training effects whereby a good memory strategy during writing is found to improve both L1 and L2 composition skills. Ransdell et al. analyze the effects of training a good working memory strategy on L1 and L2, and the extent to which monolingual and bilingual writers can flexibly allocate novel working memory resource strategies during a writing task. Within L1, on-line analysis of word-processed records and verbal protocols of skillful college NEW DIRECTIONS FOR RESEARCH IN L2 WRITING 7 writers shows that they have a distinctly different ‘writing signature’ compared to less skillful writers. Skillful writers move continuously between writing subprocesses, planning, typing, and revising throughout their essays while less skilled plan, then type, and finally at the end, revise (Levy & Ransdell, 1995). Better working memory control would promote continuous subprocess navigation and so such control may also help account for more successful writing. There is some evidence that fluency in an L2 can also facilitate the co-ordination of multiple cognitive processes like those involved in written composition (Ransdell, Arecco, & Levy, 2001). The central question addressed by the present study is the extent to which monolingual and bilingual writers can flexibly allocate novel working memory resource strategies during a writing task. Advanced and beginning college students who are either bilingual or monolingual attempted to employ a ‘step-by-step’ (poor) and an ‘all-at- once’ (good) writing strategy. The logic of the research was to try to impose the type of writing signature revealed by more or less skillful writers in previous research. Arecco and Ransdell compare the ability of L2 writers with relatively early and late L2 exposure to maintain L1 writing quality and fluency while also engaged in a competing memory task. Bilinguals are thought to have advantages over monolinguals in many cognitive tasks, and highly proficient bilinguals have been shown to perform better than less proficient bilinguals in the use of memory strategies to solve problems in aptitude and intelligence tests (Sasaki, 1993). This study investigated Spanish-English bilinguals, Polish-English bilinguals and Estonian-English bilinguals who were asked to write five ten-minute essays in L1 and L2 on randomly assigned topics. The first two control essays were written in L1 and in L2 as single tasks. The third and fourth essays were written in L1 and L2 with a secondary task of irrelevant L1 or L2 speech in the background. The final essays were written in L1 with a secondary task of a concurrent 6-digit load in L1. The participants were divided into two groups, bilinguals who had learned an L2 before the age of 11 (early L2 exposure group), and bilinguals who had learned an L2 after the age of 11(late L2 exposure group). Results suggest that early L2 exposure is associated with an increased ability to allocate strategies during a cognitive demanding task such as writing. Cumming, Busch, and Zhou supplement the experimental studies reported in earlier chapters by including an in-depth case study investigating goal setting and achievement in L2 writing. Cumming et al. investigate learners' goals in the context of adults learning to write in English in a university and other academic contexts. These authors analyze interview data from a case study of six adults from various language backgrounds preparing themselves for studies at a university in Canada. They suggest numerous ways in which students’ self-identified goals can specify and trace student achievement in L2 writing. Goal analysis can help identify aspects of writing achievement and processes of learning that otherwise may be obscured in experimental or interpretive research that uses global measures of writing improvement. The case studies also reveal multiple facets of learners’ goals for writing improvement. Multiple aspects of learning relevant to L2 writing are revealed including learners’ efforts to improve their abilities to function in specific situations and purposes and with diverse text. They conclude the chapter by suggesting how these broad, situated conceptualizations could help to refine process-product models of 8 RANSDELL & BARBIER teaching and learning related to L2 writing (cf. Cumming & Riazi, 2000), and they propose several ways of conducting future research that could do so. Kuiken and Vedder highlight the positive aspects of interaction among writers on metalinguistic awareness and text quality in L2. These authors focus on collaborative writing in L2 and how the interaction processes among writers can lead to differences in L2 writing quality. The authors suggest that collaborative language production may prompt learners to deepen their awareness of linguistic rules. This may trigger cognitive processes that may both generate new linguistic knowledge and consolidate existing knowledge. Furthermore, ‘metatalk’ about language in a communicative context may help learners to understand the relation between form and meaning, and positively affect the acquisition of L2 knowledge. In this perspective, the research focus is both on the product of collaborative writing (the linguistic quality of the text) and on the interaction process which has led to the production of this text. Subjects are learners of Dutch, English and Italian as a second language. Data concern several dependent variables: The syntactic and lexical quality of the text and how it is affected by the degree in which learners are encouraged to interact with each other and the kind of metacognitive, linguistic and interaction strategies learners make use of. A further qualitative analysis of the interaction between learners shows how reflection and discussions on language forms, content and the writing process itself result in more noticing and, as a consequence, better knowledge of certain grammatical and lexical forms. Franken and Haslett reveal both the facilitating and potentially distracting aspects of speaking with a peer before L2 composition. Franken and Haslett explore the different effects of speaking on written argumentative texts in L2. They explore the various roles that speaking can play in affecting the quality of argument texts written by second language learners in a New Zealand high school context. A quasi- experimental study tested the effects of a sequence of activities that were experienced in two ways. One way was for students to engage in talk with a peer before and during writing. The other way was for students to work in a solitary way on prewriting activities and the writing task. Quantitative data analysis compared the writing scores gained by the students on two sets of variables: One to indicate the quality of the writing in textual terms and the other to indicate the quality of the writing in terms of specific features of argument. The research points to the fact that the opportunity to speak with a partner has differential effects. Positive effects for opportunity to talk were seen in the quantity of topic-related material, but only when students wrote texts that appeared to require more content and domain-specific knowledge. However the opportunity to work with a peer did not appear to alleviate textualization demands. While it may facilitate referential development, it appears potentially to impose a cognitive load for students attempting to address textualization demands. In fact, working in a solitary way rather than with a partner resulted in significantly better mean scores for linguistic accuracy and complexity. Quintero analyses the evaluation of a training method for L1 and L2 composition and literacy. This author illustrates a problem-posing approach for teaching English as a second language in the areas of both literacy and composition. Three case study examples are discussed and a preliminary analysis is made using a type of critical theory. This critical theory analysis led to placing the findings into three categories: NEW DIRECTIONS FOR RESEARCH IN L2 WRITING 9 expression, interpretation, and transformation. On the basis of her analyses, Quintero espouses L2 instruction through the use of L1 writing. Native language writing is shown to provide literacy lessons embedded within many contexts. These lessons, in turn, complement the traditional training of specific writing skills in L2.

3 TWO QUESTIONS FOR RESEARCH IN SECOND L2 WRITING Psycholinguistic research on L2 writing has been focused on ‘writing as process’ (for a review see Roca de Larios et al., Chapter 1, this volume), but cognitive mechanisms might be analyzed through other aspects such as ‘writing as socializing’ and ‘writing as product’, considering individual and social factors; cf. Hayes, 1996). The studies in this volume offer some concordant data about writers functioning in L2 by thoroughly taking into account more systematically the diversity of writer’s knowledge and contexts of production during the writing task in L2. comprehension of cognitive processes is facilitated by recording the temporal The two questions that are addressed in this book relate to the control of knowledge and of writing processes, as well as questions about the cognitive resources allocated in a complex task such as writing in a foreign language. The first question is how cognitive demands on writing processes are made in L2. Further analyses about the temporal features of text production are required. The dynamics of production in real time. In other words, mental activities are tracked as they unfold. Proc- essing time allocation and analyses of pauses during language production are then possible (Barbier, 1998a; Faraco et al., this volume). Other on-line measures used within L2 writing research are reaction time, (Schoonen et al.; Sasaki, this volume); and secondary tasks, (Ransdell et al.; Arecco & Ransdell, this volume). With these kinds of methods, it is possible to evaluate resource allocation to writing processes in L2, as recently found within studies of written production in L1 (Levy & Rans- dell, 1996; Piolat & Pélissier, 1998). The second question is how L2 writing, L2 acquisition, and L1 knowledge interact. Recent studies emphasize the flexibility of the models on L2 writing activity, recognizing and illustrating that all parts of the writing process loop backwards and forwards through various subsystems. For example, Manchon, Roca de Larios, and Murphy (2000) and Zimmerman (2000) attempted to establish basic parameters within which individuals might be placed along a continuum of L2 writing expertise. Consequences of such insights are likely to move us away form a simplistic pedagogy of asking non-proficient writers merely to imitate proficient ones (Leki, 2000). To recognize the various subsystems involved within L2 writing activity, more studies must come together in describing the formulation process (see Roca de Larios, this volume), but not as opposed to planning and revision. The use of L1 in L2 writing for a variety of formulating operations is particularly analyzed not only in terms of transfer processes, but also as a process directly involved in L2 writing. Manchon, Roca de Larios, and Murphy (2000) find that backwards operations in L1 are used to translate the L2 text back into (mental) text that is more resonant with meaning for the writer. Also, backwards operations in L1 may elaborate on (not just reiterate) the 10 RANSDELL & BARBIER already written text. Such a role of the L1 may be particularly significant in L2 writing since the already written text is constrained by the writer’s L2 proficiency. The integration of knowledge across L1 and L2 is also facilitated by comparing L2 writing to oral and reading proficiency as in the chapter in this volume by Dur- gunoglu et al. Whereas research in L1 writing has typically lagged behind that of L1 reading, L2 writing may be ahead of the curve relative to L2 reading research. L2 writing research will continue to prosper as long as complementary approaches to research are discussed among colleagues who study L2 writing from within different fields and types of training. Among those complementary approaches to the psycholinguistic tradition found in this volume in the first seven chapters are those of Cumming et al., Kuiken and Vedder, Franken and Haslett, and Quintero.

ACKNOWLEDGMENTS The authors would like to Gert Rijlaarsdam, Richard Coff, Päivi Tynjälä, and an anonymous reviewer for their comments on earlier drafts of this chapter. We would also like to thank Christiane Roll at Kluwer Academic Publishers for her encour- agement of this project. Finally, we would like to thank each of our mentors, Mi- chael Levy (Ransdell) and Annie Piolat (Barbier). The authors may be contacted at the following email addresses for further correspondence, [email protected] and [email protected].

A CRITICAL EXAMINATION OF L2 WRITING PROCESS RESEARCH

JULIO ROCA DE LARIOS, LIZ MURPHY, & JAVIER MARÍN

Universidad de Murcia (Spain)

Abstract. The present chapter is intended as a critical analysis of the most relevant recent research into the cognitive processes underlying second language written composition. After an introduction of the research domain, a number of relevant methodological aspects are briefly discussed. These include the data collection procedures used, the assessment of writers’ command of the second language, the evaluation of written products, the context of the research, the type and number of participants involved, the type of tasks used, and the way reliability has been reported in the different studies. The substantive part of the research has been analyzed by isolating its main theoretical frames. Each of these frames has allowed us to derive a number of research sub-domains under which the studies have been grouped: the comparison of skilled and unskilled L2 writers, the development of L2 writing skill, the comparison of L1 and L2 writing processes, and the relationship between writing ability and L2 proficiency. A systematic analysis of the findings within each category has led us to identify a number of areas in need of further research: the notion of L2 writing skill, the formulation process, the temporal character of composition, the cognitive mechanisms involved in the transfer of writing abilities across languages, and the situated nature of L2 writing.

Key words: L1 and L2 writing processes, writing skill, transfer of writing skills, skilled vs. unskilled writers; relationship writing skill and L2 proficiency.

1 INTRODUCTION Second language writing research has shifted its focus of concern among the three basic elements that comprise the writing activity: the socio-cultural context where the writer writes and learns to write, the text the writer produces, and the acts of thinking the writer engages in to produce such text (Cumming, 1998). These three theoretical orientations are usually referred to as writing as socializing, writing as product, and writing as process. The original drive behind the process movement was pedagogical in orientation: it was believed that for teaching procedures to be

J. Roca De Larios, L. Murphy & J. Marín. (2002). A critical examination of L2 writing process research. In G. Rijlaarsdam (Series ed.) & S. Ransdell & M. L. Barbier (Volume eds.), Studies in Writing, Volume 11: New Directions for Research in L2 Writing, 11 – 47. © 2002 Kluwer Academic Publishers. Printed in the Netherlands. 12 ROCA DE LARIOS, MURPHY & MARÍN effective, they should be based on a theory that accounted for what writers actually do in the process of composing a text. In this regard, the word process was understood in two different ways that correspond to two different trends within the process movement: the expressivist and the cognitivist (Faigley, 1986). For the former, writing was seen as the process that allowed an expression of the self provided the student was encouraged to write freely on those topics that matter to him/her. As such, it was basically a paradigm of L1 writing that had little effect upon L2 writing research (for further information, see Johns, 1990; Schereiner, 1997; Timbur, 1994). For the cognitivists, the word ’process’ referred to the mental operations writers use when they are trying to generate, express and refine ideas in order to produce a text: this is the conception of process that we shall use to guide the review of the literature that follows. From this perspective, L2 writing research seems to have evolved into a research domain addressing a homogeneous set of problems which can be seen as the attempt to answer certain questions: (1) Do the reported findings on L1 writing processes apply equally to L2 composing processes? (2) If writing in a non- native language imposes extra constraints on learners, do these constraints create additional competing demands for attention? Do they also influence the type of problems attended to and the strategies used to solve them? (Manchón, 1997). From an applied perspective, answering these questions has been regarded as an essential step for teaching methods to be based on a sound theory that may help practitioners avoid the risks of blindly transposing recommendations for the teaching of L1 writing to the L2 situation (Silva, 1993; Zamel, 1983). Individual studies within the domain have produced increasing amounts of data in response to those questions. Given their cumulative nature, summarizing them is a necessary precondition for indicating what remains to be done. In this respect, substantial accounts have been undertaken within the L2 writing process literature (Cumming, 1998, in press; Grabe and Kaplan, 1996; Krapels, 1990; Manchón, 1997, in press; Mc Donough, 1999; Silva, 1989, 1993) which, to a greater or a lesser extent, have adopted a critical stance as to the studies’ methods, reporting and interpretation practices. The present review is an attempt to follow this critical tradition by approaching the domain covered by the research in a theoretically-motivated manner which allows us to present the research findings, discuss limitations and suggest areas for future research in a systematic fashion. For this review, all relevant studies that were located, both published and unpublished in the last twenty years, were included for analysis (for a similar approach, see Silva, 1993). The total number of studies analyzed was sixty-five (see appendix). In reviewing these studies, we have identified critical aspects of the research domain by isolating the theoretical frames used and the different research areas derived from their explicit or implicit application. These will be presented, together with an account of the procedures used for their analysis, in section three, after a brief discussion of some methodological aspects of the studies that may affect the comparability of findings (section two). We will proceed in the next four sections with the presentation and discussion of the main findings within each research area, and conclude the study (section eight) with a discussion of the main limitations identified and an examination of some of the main problems we believe L2 writing process research should address in the future. A CRITICAL EXAMINATION OF L2 WRITING PROCESS RESEARCH 13

2 METHODOLOGICAL CONSIDERATIONS ON THE STUDIES REVIEWED Variations in the methodology, purpose and quality of the studies reviewed impose some constraints on the interpretation of their results. Although a deep methodological analysis of the studies is beyond the scope of this chapter, the following considerations should be borne in mind when reading the sections that follow. First, data have been collected using a variety of techniques ranging, among others, from dual- task procedures, direct observation, analysis of written texts and computerized track- ing devices to introspective methods, each of them with its own pros and cons in the potential to access and interpret data. Dual-task procedures deliberately seek to in- terfere with the train of thought by asking participants to carry out the writing task while simultaneously listening to irrelevant speech or recalling series of random digits (Ransdell & Levy, 1998; Ransdell, Arecco & Levy, 2001). Direct observation, text analysis and analysis via computers, although minimally disruptive of the writing process and adequately informative in some respects, fail to capture some of the covert process of writers at work. They cannot, for instance, provide information about how ideas are generated or what alternatives are evaluated. In contrast, intro- spection procedures (either on-line or retrospective) are reputed to provide a better picture of what goes on in the mind when tackling a task, although there are drawbacks associated with them, too. For one thing, on-line techniques such as the think- aloud method have raised doubts as to whether they may distort writers’ normal composing processes or whether the process of composing aloud may in fact be distinct from that of composing silently (Bosher, 1998; Sasaki, this volume; see also Smagorinsky, 1994, for a review). On the other hand, both immediate retrospective techniques, such as stimulated recall (Bosher, 1998; Sasaki, 2000, this volume), and delayed retrospection techniques, in the form of questionnaires (Sasaki & Hirose, 1996) or interviews (Porte, 1997; Silva, 1992), rather than allowing access to actual mental processes, run the risk of showing ‘the way participants situate themselves vis-á-vis a particular question or the person asking it’ (Block, 2000: 760). In addition, most studies using on-line or retrospective techniques have not reported the segmentation criteria for the analysis of the resulting protocols, leaving the units of analysis in need of clarification (for exceptions, see Cumming, 1989; Manchón, Roca de Larios & Murphy, 2000; Porte, 1997; Roca de Larios, Marín & Murphy, 2001; Roca de Larios, Murphy & Manchón, 1999; Uzawa, 1996). Second, although the participants’ command of the second language has generally been reported (for exceptions, see Friedlander, 1990; Lay, 1982, 1988; Moragne e Silva, 1988; Ting, 1996), this variable has been assessed in a variety of ways which include institutional status, in-house assessments and standardized tests, each procedure representing different assumptions of what constitutes proficiency (Tho- mas, 1994). Many studies assumed that being registered for a certain academic course presupposed a specific level of L2 proficiency. The explanation usually given for using this type of institutionally-based assignment is the degree of verbal ability (Arndt, 1987; Fagan & Hayden, 1988; Tarone, Downing, Cohen, Gillette, Murie, & Dailey, 1993; Zamel, 1983) associated in principle with the course in question. This association is usually based on teachers’ impressions (Fagan & Hayden, 1988; Miller, 2000; Pennington & So, 1993; Swain & Lapkin, 1995; Uzawa, 1996) or 14 ROCA DE LARIOS, MURPHY & MARÍN exam systems (Chenoweth & Hayes, 2001; Henry, 1996; Valdés et al., 1992). The procedure is thus highly economical since it requires no extra work on the part of the researcher or the participants other than registering the position of the latter in each course. Yet, standards in academic institutions are far from homogeneous and even among individuals in the same group there may be large disparities in L2 proficiency level. A second procedure used to assess participants’ proficiency involves locally designed tests. This kind of in-house assessment, in contrast to the previous one, allows all participants within the sample to be tested uniformly and guarantees some kind of internal consistency. A typical example of an operational definition runs like this: ‘We based level of linguistic competency on a comprehensive four- skills (reading, writing, speaking, and listening) test administered at the end of the students’ second year of French studies’ (Whalen & Ménard, 1995: 386). Although most studies using this procedure have isolated L2 proficiency as a variable in its own right, on some occasions the assessment of writers’ L2 proficiency has been intermingled with that of their L2 writing ability (Hall, 1990), or has been pulled out of the grammar section of an otherwise writing-oriented test (Berman, 1994). In one particular case, the administration of this assessment type was supplemented with participants’ self-evaluation of linguistic abilities (Faraco, Barbier & Piolat, this volume). Being locally developed, however, the results of in-house assessment cannot be extended in principle to learners outside the sample. In addition, together with institutional status, this type of assessment tends to establish arbitrary cut-off points between proficiency levels that may affect the comparability of findings. Finally, standardized tests represent multi-trait assessment procedures that are used to identify stages of overall language development, of which writing may be one part. These types of tests are taken to be one of the most objective ways to assess L2 proficiency because their content and validity are subject to on-going scrutiny, which helps to legitimate the generalizability of research results (Thomas, 1994). The tests used in the studies reviewed include the TOEFL (Carson & Kuehn, 1992; Devine, Railey & Boshoff, 1993; Hall, 1991; Kasper, 1997; Qi, 1998), the CELT (Kobayashi & Rinnert, 1992; Sasaki, 2000; Sasaki & Hirose, 1996), the MTELP (Raimes, 1985, 1987; Silva, 1992), the MELAB (Bosher, 1998), the IELTS (Shaw & Ting Kun Liu, 1998), the OPT (Manchón et al., 2000; Roca de Larios et al., 1999; Roca de Larios et al., 2001), the SLEP (Sasaki, this volume), the Cambridge FCE (Victori, 1995), and Carroll’s (1980) scale (Smith, 1994). One particular study (Jones & Tetroe, 1987) made use of both the Carroll and the TOEFL tests at different moments in their longitudinal design. In other cases (Ransdell & Levy, 1998; Ransdell et al. 2001) the TOEFL was used as the benchmark for participants’ self-ratings of proficiency in English, and a further study (Akyel, 1994) made use of the TOEFL inter- changeably with a locally designed test considered to be equivalent to the MTELP. Although in many studies informants were partitioned following the specifications of the test used, in some cases students were selected on the basis of further composition scores. One of the main drawbacks of standardized measures is that they are only available for English, leaving researchers with no other option but to rely on more local solutions when they have to assess writers’ command of other languages. Another drawback is that they are essentially psychometric in orientation, which prevents them from assessing many aspects of communicative competence. A CRITICAL EXAMINATION OF L2 WRITING PROCESS RESEARCH 15

Third, variability is also observed in the way compositions have been evaluated. A few studies, in spite of claiming that their participants’ written products had been analyzed, failed to report the instrument used (Zamel, 1983; Uzawa & Cumming, 1989). In contrast, many studies reported having used the ESL composition profile (Jacobs, Zingraf, Wormuth, Hartfiel & Hughey, 1981) either in its original version or modified to suit their research purposes (Bosher, 1998; Cumming, 1989; Cum- ming, Rebuffot & Ledwell, 1989; Pennington & So, 1993; Sasaki, 2000, this volume; Sasaki & Hirose, 1996; Uzawa, 1996). The TWE (Carson, Carrell, Silberstein, Kroll, & Kuehn, 1990; Silva, 1992), the SSQS (Ransdell & Levy, 1998; Ransdell et al., 2001) and the HKCEE scoring procedure (Sengupta, 2000) were other standardized tests used less frequently. Other studies fell back on in-house instruments of either a holistic (Akyel, 1994; Friedlander, 1990; Kasper, 1997; Raimes, 1985; Skib- niewski, 1988; Victori, 1995) or an analytical nature (Berman, 1994; Hall, 1991; Jones &Tetroe, 1987; Koda, 1993; Qi, 1998; Smith, 1994; Tarone et al., 1993) or a mixture of both (Kobayashi & Rinnert, 1992). Special mention should be made of those studies which, deriving information on composing processes from the analysis of written texts, designed purpose-built text assessment categories with an emphasis on discourse aspects (Carson et al., 1990; Carson & Kuehn, 1992), form and function (Cohen & Brooks-Carson, 2001), and conceptual and linguistic complexity (Yau, 1991), or have developed measures of overall quality, accuracy and fluency comparable to the ACTFL guidelines (Henry, 1996; Valdés et al., 1992; Way, Joiner & Seaman, 2000). The use of such a variety of instruments indicates that there is no single standard of evaluation of writing quality for writing studies. This lack of ho- mogeneity is further complicated if we bear in mind three considerations: (1) some of the rating scales used were originally intended for the placement or admission of learners into academic institutions, not for achievement purposes; (2) the common practice of rating students’ compositions holistically has tended to conflate different dimensions of written productions into a single score, thus hiding the very elements that should be highlighted; and (3) the stance of the evaluator has a decisive influence on the score produced (see 8. 1. below). Unfortunately, although some work has been done on the objective indicators that should correspond to levels of L2 writing ability, there is still no consensus on the matter (Cumming & Riazi, 2000). Fourth, the contexts of most studies fall into two distinct categories: second language and foreign language, although one study (Ransdell et al., 2001) exceptionally included participants from both contexts. The difference is important because the presentation of self as conforming to social values is stronger in a second language than in a foreign language situation (Widdowson, 1983). In addition, each situation offers learners qualitatively different kinds of interaction, input and affective relations, as well as quantitatively different proportions of them, each of which may dramatically affect the acquisition and uses of the target language (Ellis, 1994). Ap- proximately half of the studies reviewed here involve writing in a foreign language context and this tendency has gradually increased in the last ten years. Most studies in this context were devoted to English, but there were also several studies focused on French (Chenoweth & Hayes, 2001; Cohene & Brooks-Carson, 2001; Faraco et al, this volume; Way et al., 2000), Russian (Henry, 1996), German (Chenoweth & Hayes, 2001; Thorson, 2000), Arabic (Khaldieh, 2000), Chinese (Bell, 1995), Span- 16 ROCA DE LARIOS, MURPHY & MARÍN ish (Valdés et al., 1992) and Japanese (Uzawa & Cumming, 1989; Pennington & So, 1993) as foreign languages. English has also been the target language of most studies in the second language context although writing in French by Anglophones is the focus of three studies with adults (Cumming et al., 1989; Whalen & Ménard, 1995), one with children (Fagan & Hayden, 1988), and one with adolescents (Swain & Lapkin, 1995). The increasing awareness of the specifics of these two contexts has given rise to recent calls (Henry, 1996; Reichelt, 1999; Sasaki, 2000; Way et al., 2000) for the recognition of the unique situation of the foreign language writer in a move away from excessive reliance on conceptions of writing skill solely derived from either L1 or ESL writing research (Roca de Larios & Murphy, 2001). Fifth, studies do not evenly report what might be salient features of participants, particularly their writing ability in the native language. In addition, factors such as age and sample size show high variability. Although plenty of information is presented about participants’ writing ability in the second language, very few studies report their L1 writing ability. When reported, researchers have assessed this variable on the basis of students’ L1 compositions (Berman, 1994; Lay, 1988), students’ self-ratings (Cumming et al., 1989), or a combination of both procedures (Cum- ming, 1989; Sasaki & Hirose, 1996). Alternatively, they have relied on the standards demanded in the institution writers are enrolled in (Henry, 1996; Miller, 2000; Valdés et al., 1992) or the academic status and writing experience of the participants (Sasaki, 2000, this volume). In only one case has writing skill been assessed from a genuine process perspective (Pennington & So, 1993). Many studies do not report the age of their informants (Chenoweth & Hayes, 2001; Jones & Tetroe, 1987; Raimes, 1985, 1987; Smith, 1994; Uzawa, 1996; Uzawa & Cumming, 1989; Zamel, 1983; Zimmermann, 2000), but it can nevertheless be inferred through the academic courses taken. In fact, the studies that do provide the age of their writers and their year of study at university show that they are generally in their late teens, twenties or even thirties: 18 to 21 (Sasaki & Hirose, 1996; Victori, 1995), 19 to 25 (Roca de Larios et al., 1999), early twenties (Manchón et al., 2000), 18 to 19 (Porte, 1997), 19 to 23 (Miller, 2000; Uzawa, 1996), 22 to 25 (Pennington & So, 1993), 19 to 37 (Hall, 1991), or 21 to 39 (Ransdell et al., 2001). It may accordingly be concluded that the vast majority of participants in the studies reviewed are university undergraduates or postgraduates who happen to be, on many occasions, the researchers’ own current or former students, a situation that may have prejudicial effects on task completion (Silva, 1989; but see Zamel, 1983). A few studies, however, have included university professors (Bell, 1995; Sasaki, 2000, this volume), school children (Edelsky, 1982; Fagan & Hayden, 1988) and high school students (Berman, 1994; Roca de Larios et al., 2001; Sengupta, 2000; Swain & Lapkin, 1995; Tarone et al., 1993; Yau, 1991; Way et al., 2000) as their only group or in combination with other age groups. The number of informants ranges from one (Bell, 1995; Moragne e Silva, 1988; Qi, 1998; Ting, 1996) to 330 (Way et al., 2000). Generally, studies using questionnaires or analysis of written texts as their main data collection procedures involve a range from 28 to 330 informants (Akyel, 1994; Berman, 1994; Car- son et al., 1990; Carson & Kuehn, 1992; Cohen & Brooks-Carson, 2001; Fried- lander, 1990; Henry, 1996; Kobayashi & Rinnert, 1992; Sasaki & Hirose, 1996; Sengupta, 2000; Shaw & Ting Kun Liu, 1998; Valdés et al., 1992; Yau, 1991). In A CRITICAL EXAMINATION OF L2 WRITING PROCESS RESEARCH 17 contrast, small-scale studies, involving between one and ten informants, use think- aloud, stimulated recall, observation, and self-report as main data collection techniques (Arndt, 1987; Bell, 1995; Bosher, 1998; Brooks, 1985; Fagan & Hayden, 1988; Gaskill, 1986; Hall, 1990; Hall, 1991; Jones & Tetroe, 1987; Lay, 1982, 1988; Manchón et al., 2000; Moragne e Silva, 1988; Pennington & So, 1993; Qi, 1998; Raimes, 1985, 1987; Skibniweski, 1988; Smith, 1994; Swain & Lapkin, 1995; Ting, 1996; Uzawa & Cumming, 1989; Victori, 1995; Zamel, 1983). In between, there are a number of studies which, in spite of the labor intensive handling of the data collection techniques used (think-aloud, stimulated recall, self-report, semi-structured interviews, computer analysis of the composing process, and use of secondary tasks) have nevertheless managed to include a reasonably large number of informants - between 12 and 43- in their research designs (Chenoweth & Hayes, 2001; Cumming, 1989; Cumming et al., 1989; Faraco et al., this volume; Khaldieh, 2000; Miller, 2000; Porte, 1997; Ransdell & Levy, 1998; Ransdell et al., 2001; Roca de Larios et al., 1999; Roca de Larios et al., 2001; Sasaki, this volume; Silva, 1992; Thorson, 2000; Whalen & Ménard, 1995). Sixth, there is considerable debate about what constitutes a representative sam- pling of second language writing, whether brief tasks or students’ written samples collected during a period of time (Hamp-Lyons, 1990; Henry, 1996; Raimes, 1998). Although in a few cases researchers collected a number of course-related assignments on the assumption that this procedure would be in consonance with ordinary class writing (Edelsky, 1982; Zamel, 1983), the vast majority of studies, in line with the problem-solving approach followed (Pozo, 1989), opted for short time- compressed compositions. But even so, conspicuous differences can be traced between some studies in the amount of time given to participants for the completion of tasks, the topics and text types set, the explicitation of audience, and the possibility of using external aids. Although there were cases where the allocation of time for tasks was not reported (Chenoweth & Hayes, 2001; Friedlander, 1990; Lay, 1982; Miller, 2000; Qi, 1998; Sengupta, 2000; Uzawa & Cumming, 1989), in those studies where the information was given the time ranged from a maximum of no limit at all in class (Zamel, 1983; Edelsky, 1982; Fagan & Hayden, 1988) or at home (Kaldieh, 2000; Ting, 1996) to a minimum of ten minutes (Henry, 1996; Ransdell & Levy, 1998; Ransdell et al., 2001), most studies allowing between half an hour and two hours. Alternatively, some authors opted for fixing a space and/or word limit (Pen- nington & So, 1993; Skibniewski, 1988), or a combination of time and space limits (Whalen & Ménard, 1995; Yau, 1991). Usually, if a set of tasks was fixed for completion, equal amounts of time tended to be allocated to each task individually considered, although a few studies gave more time to translation tasks over composition proper (Kobayashi & Rinnert, 1992; Uzawa, 1996; but see Cohen & Brooks-Carson, 2001), to less proficient writers over more proficient ones (Valdés et al., 1992), to certain text types over others (Thorson, 2000), to examination over non-examination situations (Hall, 1991), or allocated uneven periods of time according to the language of the composition (Moragne e Silva, 1988). Given the amount and diversity of studies, the range of topics covered is very wide. These include personal matters or experiences (Chenoweth & Hayes, 2001; Henry, 1996; Jones & Tetroe, 1987; Manchón et al., 2000; Pennington & So, 1993; Porte, 1996; Qi, 1998; Raimes, 1985; 18 ROCA DE LARIOS, MURPHY & MARÍN

Tarone et al., 1993; Valdés et al., 1992); young people’s (Berman, 1994; Way et al., 2000), and university students’ concerns (Bosher, 1998; Hall, 1990; Skibniewski, 1988; Ransdell & Levy, 1998; Ransdell et al., 2001; Whalen & Ménard, 1995); social (Akyel, 1994; Cumming, 1989; Hall, 1991; Jones & Tetroe, 1987; Khaldieh, 2000; Porte, 1996; Qi, 1998; Sasaki & Hirose, 1996; Sengupta, 2000; educational (Manchón et al., 2000; Moragne e Silva, 1988; Raimes, 1987; Roca de Larios et al., 1999; Roca de Larios et al., 2001; Sasaki, 2000), financial (Moragne e Silva, 1988), ethical (Akyel, 1994; Cohen & Brooks-Carson, 2001), environmental (Shaw & Ting Kun Liu, 1998; Swain & Lapkin, 1995), and second language acquisition (Miller, 2000) issues; political (Cumming et al., 1989) and historical events (Uzawa & Cumming, 1989; Thorson, 2000); jobs (Carson et al., 1990; Carson & Kuehn, 1992); comparisons of different kinds (Hall, 1991; Kobayashi & Rinnert, 1992, 1998); cultural (Akyel, 1994; Friedlander, 1990; Jones & Tetroe, 1987; Raimes, 1987; Thorson, 2000) and linguistic differences (Silva, 1992); literary criticism (Faraco et al., this volume); or issues of a more local character (Sasaki, 2000; Smith, 1994). Generally, the topics were set by the researchers but in a few studies writers were given the opportunity to decide their own topics (Edelsky, 1982; Fagan & Hayden, 1988) in combination with other set topics (Brooks, 1985; Lay, 1982) or by choos- ing them from a list (Hall, 1990; Hall, 1991; Kobayashi & Rinnert, 1992; Yau, 1991). Only on one occasion were prewriting procedures used to control for topic familiarity (Yau, 1991). In some cases writers had to react to a text previously read (Bosher, 1998; Devine et al., 1993), manipulate a specially-prepared fragment (Ko- bayashi & Rinnert, 1998), translate L1 texts (Qi, 1998; Smith, 1994; Uzawa, 1996) or their own L1 productions (Cohen & Brooks-Carson, 2001; Kobayashi & Rinnert, 1992) into the L2, write on previously presented data (Uzawa & Cumming, 1989), accommodate their composition to a final sentence previously given (Jones & Tetroe, 1987), summarize texts (Cumming, 1989; Cumming et al., 1989) or take notes from an audio-taped lecture (Faraco et al., this volume). In one particular case, the topic was the content of a film (Zimmermann, 2000). Variability was also visible in the handling of possible repetitions of a topic in those studies where more of one task was used. Some authors advocated the use of different topics across tasks (Moragne e Silva, 1988; Ransdell & Levy, 1998; Ransdell et al., 2001; Whalen & Ménard, 1995), others opted for mirror topics (Arndt, 1987; Carson et al., 1990) while others used the same topic for both the L1 and L2 task (Sasaki & Hirose, 1996) or for the pre-test and post-test conditions (Sengupta, 2000; Shaw & Ting Kun Liu, 1998). Usually the tasks were counterbalanced (but see Roca de Larios et al., 2001; Whalen & Ménard, 1995) as a way of controlling for learning effects, although in one case the order of tasks was decided as a function of knowledge level demands (Qi, 1998) and in another case writing tasks preceded reading tasks so that the latter could not provide models for the former (Carson et al., 1990). The genre most commonly used in the assignments was the argumentative type, probably as it was thought to be the one that better promoted critical thinking and textual organization, while expository, narrative and contrast and comparison text types followed in frequency. Some studies, in search of the different cognitive and linguistic demands generated by different genres or precisely as a result of those assumed differences, used various text types in combination (Brooks, 1985; Cumming, 1989; Lay, 1982; A CRITICAL EXAMINATION OF L2 WRITING PROCESS RESEARCH 19

Manchón et al., 2000; Miller, 2000; Porte, 1996; Qi, 1998; Raimes,1987; Roca de Larios et al., 1999; Way et al., 2000). As with topics, text types were generally fixed by the researchers, although on two occasions writers were given the chance to choose (Bosher, 1998; Fagan & Hayden, 1988). Although approximately half of the studies reviewed did not specify the audience in their task prompts, those in which it was mentioned asked participants to think of their own peers (Akyel, 1994; Uzawa & Cumming, 1989), teachers (Hall, 1990), administrators (Friedlander, 1990; Sen- gupta, 2000; Whalen & Ménard, 1995) or pen friends (Thorson, 2000; Way et al., 2000) as possible readers, or else, encouraged them to imagine that their compositions would be useful to magazine writers or official agencies (Akyel, 1994) or, alternatively, that they would be published in university magazines or newspapers (Arndt, 1987; Swain & Lapkin, 1995; Thorson, 2000), readers’ opinion columns in newspapers (Sasaki, 2000), academic journals (Silva, 1992), and high school bulle- tins (Skibniweski, 1988). Interest in audience was especially apparent in some studies. Hall (1990), for example, discussed the writer/audience relationship with each individual writer before the task was approached, Raimes (1987) controlled for a possible audience effect, and Way et al. (2000) used three different wordings of the same letter to a pen friend to look at possible communicative and interactive effects. Finally, the possibility of using aids while composing was also differentially handled. Although many studies did not report on this variable, some allowed participants to use dictionaries, reference materials or glossaries (Brooks, 1985; Carson et al., 1990; Chenoweth & Hayes, 2001; Pennington & So, 1993; Thorson, 2000; Uzawa, 1996; Valdés et al., 1992; Victori, 1995; Whalen & Ménard, 1995). Others, however, did not allow such use (Carson & Kuehn, 1992; Carson et al., 1990; Sasaki & Hirose, 1996) on the grounds that the completion of the task should resemble examination conditions (Smith, 1994), or else because the researchers were interested in the way writers dealt with linguistic problems (Roca de Larios et al., 1999; Roca de Larios et al., 2001; Swain & Lapkin, 1995) or had detected problems in the use of dictionaries by some participants (Sasaki, 2000). Seventh, an important issue affecting the rigor of the design and meaning of findings (Silva, 1989) is to what extent the instruments used in the different studies may be regarded as reliable. Some researchers have failed to report reliability estimates whether they analyzed composition processes through observation (Zamel, 1983), self-report (Brooks, 1985), think-aloud and stimulated recall protocols coded by means of schemes adopted from previous studies (Arndt, 1987; Bosher, 1998; Moragne e Silva, 1988) or specially designed for the study in question (Jones & Tetroe, 1987; Lay, 1988; Manchón et al., 2000; Qi, 1998; Roca et al., 1999; Smith, 1994), or whether they examined transfer processes (Valdés et al., 1990) and the ability to handle concepts across languages (Yau, 1991) through the analysis of written texts. Other researchers did not control for reliability either because the supposedly objective nature of the assessment procedure used released them from that need (Miller, 2000; Shaw & Ting Kun Liu, 1998; Thorson, 2000) or because such control was not a requisite of the research design (Bell, 1995; Ting, 1996). In contrast, many studies to a greater or a lesser extent showed some concern with the reliability of their instruments. Interrater agreement was thus reported for the analysis of composition processes studies which used a wide range of techniques including observa- 20 ROCA DE LARIOS, MURPHY & MARÍN tion (Fagan & Hayden, 1988), retrospective questionnaires (Porte, 1997), think- aloud protocols (Chenoweth & Hayes, 2001; Roca de Larios et al., 2001; Swain & Lapkin, 1995; Whalen & Ménard, 1995), self-reports (Kasper, 1997; Silva, 1992), as well as through a variety of instruments of text analysis intended to capture the transfer of composing processes (Carson et al., 1990; Carson & Kuehn, 1992; Henry, 1996), the efficiency of translation strategies (Cohen & Brooks-Carson, 2001; Kobayashi & Rinnert, 1992), the effects of task prompts on accuracy, syntactic complexity and fluency (Way et al., 2000), the learning of organization (Berman, 1994) and revision skills (Kobayashi & Rinnert, 1998), the development of L2 writing ability (Tarone et al., 1993), and the effect of the language of planning (Akyel, 1994; Friedlander, 1990), of revision instruction (Sengupta, 2000), and of dual-task procedures (Ransdell & Levy, 1998; Ransdell et al., 2001) on text quality. Finally, there is a set of studies which, looking at the nature of composing processes and the quality of written products in conjunction, reported measures of reliability for the assessment of both variables (Sasaki, this volume), some of them even providing interrater and intrarater percentages of agreement (Cumming, 1989; Cumming et al., 1989; Uzawa, 1996). Other studies within this group did not show this balance as they reported reliability estimates either for the evaluation of the product (Devine et al., 1993; Gaskill, 1986; Khaldieh, 2000; Pennington & So, 1993; Skibniweski, 1988; Victori, 1995) or the process (Raimes, 1985, 1987) but not for both.

3 THE REVIEW OF THE LITERATURE The following steps were taken for this literature review. First, studies were read and information was identified about purposes, participants, procedures, instruments, variables, topics investigated and main findings. Second, the information obtained was scanned to see if any patterns or themes could be drawn from the research issues addressed. This procedure allowed us to tentatively establish the theoretical frames under which a number of research areas could be identified. Third, different themes were discerned within each area by searching for similarities and differences between studies and their results. This involved a detailed analysis of the original papers so as to confirm or contradict the emerging hypotheses. Whenever a hypothesis was confirmed (for example, studies suggesting cross-linguistic similarities were of a qualitative nature while those suggesting differences between L1 and L2 writing tended to use quantitative analyses) at least some counter-explanation was considered (e.g. the case of a particular study which, in spite of being qualitative, may have found a number of differences). The scope of the different research areas was then refined with the result that many studies were found to fall into more than one area. Fourth, separate notes were written for the methodological problems of the studies as they emerged. Finally, after the analysis of the areas had been carried out, we wrote about each and considered the results as a whole. Any discrepancies were resolved by rereading the original studies. The reliability criteria for these analytical procedures were based on those commonly used in qualitative research (Moss, 1994). They include the extent of the researcher's knowledge on the range of work on the topic, the existence of multiple and varied sources of evidence (the studies A CRITICAL EXAMINATION OF L2 WRITING PROCESS RESEARCH 21 themselves in this particular case), and the transparency of the different pieces of evidence leading to the interpretation suggested. We have tried to meet each of these criteria as fully as possible, although we are aware of the shortcomings involved in narrative reviews such as the one attempted here (see Norris & Ortega, 2000, for details). In what follows, first the theoretical frames used by the studies will be briefly discussed and then an account of the results reported within each of the different research areas identified will be presented.

3.1 Theoretical frames used in L2 writing process-oriented research In general, studies have been based on two distinct theoretical perspectives: L1 writing theories or models and theoretical constructs in the field of second language acquisition research related to literacy transfer. The writing model most widely used, either explicitly or implicitly, is that of Flower and Hayes (1981). In fact, this model has fixed the vocabulary most commonly used by researchers in dealing with the composing process -especially the distinction between planning, formulation and revision, and their corresponding subprocesses- and is also the one with the highest degree of generality, as well as the most frequently cited (Fayol, 1991). In line with their assumption of composing as a goal-directed activity, Flower and Hayes con- tend that the key feature of writing is the careful consideration of the readers, the effect the writer wants to have on them, how the writer wants to present him/herself through the text, and the creation of a coherent network of ideas. Although this approach was originally intended to describe how mature writers' conceptualize the task and set up different goals in order to carry it out successfully, the Flower and Hayes model can actually accommodate other (lower) levels of efficiency within the general structure by allowing for descriptions of different levels of skill within the different components (Grabe & Kaplan, 1996). Fewer studies have made use of Bereiter and Scardamalia's (1987) knowledge- telling and knowledge-transforming dichotomy. Here writers are not seen as performing the same process with different degrees of efficiency, as would be the case with Flower and Hayes' model, but as carrying out two qualitatively different kinds of process. Less skilled writers are viewed as keeping the task relatively uncompli- cated, and relying on strategies which, while not involving a great deal of cognitive effort, allow them to produce compositions that make sense, stay on topic and somehow meet the structural requirements of the text type concerned. In contrast, expert writers are described as setting constraints for text construction via a dialectical process of attempting a search in two psychological problem-spaces: a content space and a rhetorical space. Decisions arrived at in the content space (for instance, that the basis for success in education lies more in the child's upbringing than in school training) are passed on to the rhetorical space as goals (for example, the need to produce the appropriate rhetorical devices to convince the reader). Conversely, decisions in the rhetorical space (the need to produce a transition between two topics, for example) are transferred to the content space as problems, questions or insights (for instance, the recognition of the relationship between previously unrelated 22 ROCA DE LARIOS, MURPHY & MARÍN ideas), this return trip being absent among novice writers (Scardamalia & Bereiter, 1991). The possibility of writing skills being transferred from one language to another has been analyzed in many studies. A theoretical construct widely used for this purpose has been Cummins’ (1980) notion of common underlying proficiency (CUP), which is probably the best-known theoretical account of how languages are related. Basically, it holds that there is a common set of abilities underlying both first and second language performance. When a literacy operation or procedure, such as planning in writing, has been acquired in a language it does not have to be re-acquired in a second: it will always be available, regardless of the language involved. This tenet is expressed through the ‘Interdependence Hypothesis’, which claims that L2 literacy is partially dependent on the literacy already developed in the L1 when exposure to the L2 begins. A related approach adopted by some studies through the ‘Thresh- old Hypothesis’ holds that, in order to write in an L2, a certain threshold or level of L2 proficiency must first be achieved if L1 skills are to be transferred. Within this hypothesis, L2 language proficiency is seen as a limiting factor that must be distinguished from literacy skills.

3.2 The research areas The theoretical frames presented above have prompted a number of studies which we have grouped into several macro-areas of research. The theoretical assumptions involved in L1 writing models have given rise to a great deal of research (1) comparing L2 skilled and unskilled writers, or (2) approaching writing skill as a developmental path L2 writers are supposed to go along. In the first case, the notion of expert or skilled L1 writing performance, as captured in controlled tasks, has been used as the benchmark against which L2 writers' composing behaviors have been assessed. In contrast to the essentially dual nature of the skilled/unskilled distinction, the second set of studies has looked at L2 writing skill as a continuum of abilities which evolve at different rates. In turn, Cummins’ (1980) Interdependence Hypothe- sis has been used as a framework for a variety of studies interested in analyzing the transfer of writing abilities across languages by comparing L1 and L2 composing behaviors. Finally, the study of literacy skills and language skills in combination has delineated the next area of concern in L2 composition studies: the analysis of the relationship between writing ability and L2 proficiency, inspired by Cummins’ (1980) Threshold Hypothesis. In each of the following sections, the main findings within each research area will be briefly presented and, when possible, the contradictions found will be discussed.

4 THE SKILLED-UNSKILLED DISTINCTION IN ITS APPLICATION TO L2 COMPOSITION RESEARCH As suggested above, L2 writing-process research has usually followed approaches and designs originally conceived for the study of L1 writing processes (Krapels, 1990). As these approaches have used the notion of writing expertise as one of their A CRITICAL EXAMINATION OF L2 WRITING PROCESS RESEARCH 23 theoretical foundations (Grabe & Kaplan, 1996; Hillocks, 1986), it is only logical that L2 composition researchers incorporated this notion into their writing research endeavor. This has involved the analysis of the different approaches shown by experts and novices in the following aspects of writing: task representation, the writer’s approach to planning, formulation and revision, and the relationship between metacognitive knowledge and writing skill. A number of components – sense of audience, awareness of the recursive nature of writing and management of discourse levels – which form part of the writers’ representation of the writing task, have been found to involve differences of skill. While skilled L2 writers seemed to understand the importance of taking into account the reader’s expectations and handle their composing process accordingly (Brooks, 1985; Hall, 1990; Zamel, 1983), their unskilled counterparts did not use these expectations either to generate or review their texts even when the wording of the task prompt explicitly specified purpose and audience (Arndt, 1987; Raimes, 1985, 1987). Over and above the issue of skill a number of factors have been adduced to account for these differences. On the one hand, it has been suggested that for students to perceive a writing task as having a real purpose and audience, ‘more is needed than just surface wording of the topic’ (Raimes, 1987: 461), as the intellec- tual effort involved in thinking both about the substantive content of a text and its linguistic components while composing may not be activated when writing is conceived of as a mere practice of language forms rather than as the conveyance of genuine information to others (Cumming, 1990). On the other hand, task time seems to play a role. Raimes (1985), for example, speculated that the fact that Zamel’s participants had been given unlimited time for course-related topics might have encouraged a discoursal awareness which her own participants were far from showing when tackling their own time-compressed unprepared topic. Similarly, time limitations have been reported as affecting unskilled L2 writers’ behaviors by limiting the scope of their revisions to superficial changes (Uzawa, 1996) or else by giving rise to anxiety, which often led to doubts about whether to correct their texts or not (Porte, 1996; 1997). Some authors have looked at recursiveness as a specific characteristic of skill in writing, understood as flexibility in the management of the composing process according to task-specific sense of priorities (Moragne & Silva, 1988; Zamel, 1983), as the use of frequent alternatives between the intensive (writing/editing) and the reflective (pausing/reading) modes (Pennington & So, 1993), or as a balanced interaction between planning, rehearsing, backtracking, revising, editing, etc. (Silva, 1989; Raimes, 1987; Victori, 1995). Unskilled writers, in contrast, seem to limit themselves to adding more ideas to those already written (Raimes, 1987; Skib- niewski, 1988; Victori, 1995), as reflected in their piecemeal use of pauses and re- readings (Zamel, 1983) as almost the only means at their disposal to progress through the text, with the result that they tend to fail in the effective coding of ideas (Pennington & So, 1993). Some consistent findings have also been reported regarding the management of discourse levels. As compared to their less skilled counterparts, skilled writers seem to be able to pay attention to gist, discourse organization and the choice of words and phrases in an integrated way (Cumming, 1989); summarize information by inte- 24 ROCA DE LARIOS, MURPHY & MARÍN grating situational, propositional and verbatim representations of text (Cumming et al., 1989); use automatized higher level processing which allows them to concentrate on linguistic forms without losing sight of their predefined pragmatic and textual representations of the message (Whalen & Ménard, 1995); and conjugate their mental representations at different levels of complexity by establishing links between the pieces of information drawn from their reading, experience and observation (Brooks, 1985). In contrast with those findings, Uzawa (1996) found that her Japanese EFL writers, characterized as unskilled or knowledge-tellers according to the Bereiter and Scardamalia (1987) model, paid attention to various levels simultaneously. A possible explanation for this lack of consistency may lie in the way the idea of level is understood in the different studies. While Uzawa uses a rather structural notion of level (word, phrase, sentence, etc.), the notions used by Cumming (1989) and Whalen & Ménard (1995), are more complex. They are based on the idea of mental representation as used, respectively, in the literature on problem-solving (Scar- damalia & Paris, 1985) and discourse analysis (de Beaugrande & Dressler, 1981). A number of different behaviors and procedures shaping the planning process have also been reported. Skilled writers, as a group, have been found to construct more planning episodes (Cumming, 1989; Raimes, 1987, Sasaki, this volume; Skib- niewski, 1988; Victori, 1995) and of a more flexible nature (Sasaki, this volume; Victori, 1995), follow a goal-to-text approach to composition (Moragne e Silva, 1988), and approach the planning task in a distinctive, hierarchical manner (Skib- niewski, 1988). On an individual basis, these writers have been found to employ two different but equally efficient procedures: advanced and emergent planning (Cum- ming, 1989; Sasaki, this volume; Victori, 1995). The process followed by advanced planners would be frame-driven (overall frames are initially delineated and then filled with relevant propositions) while that of emergent planners could be characterized as proposition-driven (propositions are the starting point and larger frames are constructed as writing progresses). From a socio- cognitive perspective, both approaches could each be viewed as different accounts of how discovery takes place through writing and of how writers manage social goals. Advanced planning would be a clear representative of the classical position, which claims that discovery depends on the adaptation of the writer's thought to rhetorical goals. It would allow the writer to control the way his/her ideas are presented in public but has the drawback of prematurely narrowing down the writer’s emergent conceptualization of the topic by prematurely imposing order on thought. Emergent planning, in turn, could be considered a subcategory of the romantic position, one of whose basic tenets is that discovery is the result of a spontaneous dialectic between the writer's internal dispo- sition toward the topic and the emerging text. This approach would enable the writer to develop his/her conception of the topic but at the expense of revising extensively to conform to textual constraints (Galbraith, 1996; Galbraith & Rijlaarsdam, 1999). As for unskilled writers, these have been found to usually reduce their source of ideas to the limited basis of their personal experience (Brooks, 1985), generate content by means of topic association rather than by analytical association (Cumming, 1989; Uzawa, 1996), operate in a sequential rather than in a hierarchical fashion (Sasaki, 2000; Skibniewski, 1988), and stick to fixed, fruitless, organizational for- A CRITICAL EXAMINATION OF L2 WRITING PROCESS RESEARCH 25 mulae such as the typical ‘introduction, body, conclusion’, which offer only a re- mote possibility that these parts will hang together (Brooks, 1985; Raimes, 1987; Uzawa, 1996). However, the consistency of these dual findings is questioned to some extent by Sasaki and Hirose’s (1996) findings. While they found that attention to overall organization was a reliable discriminating factor between their skilled and less skilled writers, the planning of content did not show any significant differences between them. Sasaki and Hirose speculate that this latter finding was probably due to the similar amount of writing instruction both groups had received (very little in L1 and in L2) and their similar attitudes towards L1 writing for academic and personal purposes. This explanation is in fact confirmed by Sasaki (this volume) who found significant differences in planning between real experts and novices (the latter being similar to Sasaki and Hirose’s (1996) unskilled and skilled writers in terms of instruction). The relationship between writing skill and formulation has mainly been examined from the perspective of fluency, a construct alternatively operationalized as 1) the number of words written, 2) the number and length of pauses, or 3) the interruption of transcription by other processes. As for the first two perspectives (length of texts and pausing) results are contradictory. Some studies report that skilled writers write longer texts than their unskilled counterparts, who have been found to keep their texts brief out of disinterest or fear that these would be full of errors (Brooks, 1985) or, alternatively, because they reported making use of more mental translation (Sasaki & Hirose; 1996; Sasaki, this volume). Other studies, however, have found that both groups of writers wrote a similar number of words, suggesting, as opposed to the previous studies, that the possibility of making errors would not be stigmatizing for L2 writers (Raimes, 1987; Victori, 1995). Equally ambiguous are the results of those studies in which fluency has been seen in terms of the number and length of pauses. While Sasaki and Hirose's (1996) skilled and unskilled writers similarly reported that they stopped on encountering a problem, Victori (1995) found that her skilled writers paused longer and wrote sentences in shorter chunks than her unskilled counterparts. However, more consistent results have been reported in relation to the interruption of the linearization process by other processes such as evaluating, reviewing, resourcing, editing and idea generation. The larger the number and the wider the variety of the interspersed processes, the more skilled the writing behavior is taken to be, as it is a direct reflection of the problem-solving approach to composition adopted by the writer (Cumming, 1989; Victori, 1995). Some authors have looked, though not in depth, at a number of specific formulation strategies used by skilled writers. Raimes (1987), for instance, emphasized the relevant role played by rehearsing -the trying out of ideas or linguistic expressions before being rejected or accepted and written down. Although no differences were reported by the author in the comparative use of this strategy, a detailed examination of the table of occurrences presented in the study indicates that increase in the use of rehearsal may be associated with an increase in writing ability. In a similar vein, it has been found that lexical and syntactic searches involving upgrading (Smith, 1994), ideational/textual concerns (Roca et al., 1999) and rhetorical refining (Sasaki, this volume) are essential characteristics of the writing process of skilled writers. 26 ROCA DE LARIOS, MURPHY & MARÍN

The relationship between revision and writing skill has also been addressed. From this perspective, it seems that skilled L2 writers distinguish between revising and editing purposes (Brooks, 1985) and tend to postpone grammatical and lexical concerns until the final stages of the composition (Zamel, 1983), although some of them may opt for a more on-line approach to revision (Victori, 1995). Similarly, these writers reread their texts for multiple purposes (Victori, 1995), are not scared of rereading because they have a greater awareness of stylistic options (Raimes, 1987), conceive of drafting as the making of substantial changes (Zamel, 1983), and use the addition of ideas as one of their recurrent revision patterns (Victori, 1995). In contrast, unskilled writers tend to be continuously interrupted by editing problems that bear little relation to substantive meaning (Zamel, 1983), and are not so inclined to reread their own texts for fear of coming across linguistic mistakes beyond their reach (Uzawa, 1996; Victori, 1995), a direct reflection of the fact that revising only means grammatical correction for them (Uzawa, 1996). The outcome is that drafting is seen by these writers as a mere cosmetic operation despite their own admission that at times major revisions are required (Zamel, 1983). However, this alleged in- competence of unskilled writers to revise for meaning has been challenged as there seem to be grounds to suggest that their seemingly low-level revising behavior may be determined to some extent by past learning experiences and received feedback which have narrowed down their concerns to surface considerations to please the teacher (Porte, 1996; Zamel, 1983). Ideas about themselves as writers (personal knowledge), about the task of writing in general (task knowledge) and about the strategies appropriate for successful task completion (strategic knowledge), in interaction with the specific demands of a given task, all converge to make up the writer's metaknowledge (Devine et al., 1993; Flavell, 1985). In this respect, consistent differences in skill have been reported for the task and strategic components. Skilled writers seem to be prepared to take risks in the construction of complex sentences (Khaldieh, 2000), emphasize the importance of fluency and clarity (Kasper, 1997), and have a broader conception of what writing entails since they tend to adopt a more flexible attitude towards the use of superstructures, paragraphing, and the addition and reformulation of extant ideas (Victori, 1995) as well as towards the reconstruction of mental frameworks to accommodate those changes (Zamel, 1983). Less skilled writers, in contrast, tend to see writing mainly as a grammatically-driven juxtaposition of sentences (Kasper, 1997; Victori, 1995) rather than the construction of a whole discourse (Zamel, 1983), and seem to be aware of their own shortcomings regarding the generation and organization of ideas and their ability to decide on the accuracy of sentences or even the right meaning of words in the dictionary (Victori, 1995). This may lead them to resist making a conscious effort to search for appropriate language structures and use newly learnt items without any inhibition (Khaldieh, 2000). As for personal knowledge, results are not so conclusive. Victori (1995) reported that all her writers, regardless of their composing skill, exhibited the same motivation, the same writing experience, and the same self-concept as L2 writers. In contrast, other studies have shown that while skilled writers identify with the written text and gain personal satisfaction through writing (Brooks, 1985), less skilled writers tend to adopt a negative attitude towards the writing task which may ultimately lead to anxiety and frustra- A CRITICAL EXAMINATION OF L2 WRITING PROCESS RESEARCH 27 tion (Brooks, 1985; Khaldieh, 2000). This lack of consistency in findings may partly be accounted for, however, if attention is paid to the participants’ different literacy experiences as reported in the studies. Victori suggested that the similar limited opportunities for writing in the L2 might have led both groups of writers to develop similar attitudes towards writing in English, and might have stopped them constructing a full representation of what EFL writing proficiency is. Brooks reported that the more proficient writers, unlike the less skilled group, had had extensive experience as readers and writers in their own language. They knew what they could do and were clearly gauging their progress in English writing by the sense of what they could do in their L1, hesitant though they might feel about the linguistic aspects of the L2. Finally, Khaldieh suggested that his less proficient writers’ deficiency in L2 language ability might have made them feel defeated by the task, an attitude which, in turn, hindered their ability to function in the L2: a cause and effect relationship between L2 proficiency and anxiety. Summing up, the studies reviewed suggest that there are a number of behaviors characteristic of skilled L2 writers: (1) the ability to manage complex mental representations; (2) the ability to construct rhetorical and organizational goals and hold them in mind in the course of composing; (3) the efficient use of problem-solving procedures in order to formulate their texts; (4) the ability to distinguish between editing and revision as two different operations distributed in different stages of the composition process, although there may be individual differences in this respect; and (5) the adoption of a flexible attitude toward the use of rhetorical devices. Yet the equation of writing skill with length of text produced and number and length of pauses is questionable.

5 DEVELOPMENTALLY-ORIENTED STUDIES The importance of the notion of development in L2 writing becomes clear if writing expertise is viewed not as a dichotomous fixed set of competencies – the picture the research above tends to portray for some people (Porte, 1996) – but rather as an evolving continuum of abilities where some aspects are more developed than others due to the influence of factors such as cognitive style, writing experience or personality traits. Given that the vast majority of studies looking at L2 writing from this perspective have focused on adult learners, there has been little opportunity to elicit dramatic changes over time. Some studies, however, have documented changes in writing skill over a time span ranging from one semester to three years. The analyses have mainly focused on the development of certain discourse skills, the components of metacognitive knowledge, the composing of different text types, and the use of L1 in the production and rescanning of L2 written texts. Some discourse-related skills, such as certain conventions of written academic English (Shaw & Ting Kun Liu, 1998) or certain organizational skills (Berman, 1994) seem to develop with relative ease in a fairly short period of time provided that some training is provided. In the same vein, Sasaki (2000, this volume) reported that, after two semesters of process writing instruction, writers increased their planning time, and Chenoweth and Hayes (2001) found that their fifth semester students 28 ROCA DE LARIOS, MURPHY & MARÍN of French and German as a FL showed a significant increase over third semester students in a number of measures including number of words written per minute (but see Sasaki, this volume), string length, and increase in the number of words accepted and written down. Also, explicit instruction in revision seems to be useful in raising writers’ awareness of the importance of factors other than accuracy (i.e., reader needs) in improving the quality of texts (Sengupta, 2000), a finding in close correspondence with the fact that the ability to detect and correct problems of global coherence seems to develop faster than the abilities needed to tackle those of a more local nature (Kobayashi & Rinnert, 1998). In all the cases above, however, the development of accuracy and complexity in the use of the L2 appears to be less ame- nable to explicit instruction and is probably more dependent on the acquisition of higher levels of L2 proficiency. This is to some extent corroborated by Swain and Lapkin’s (1995) analysis of the composing behavior of a group of Canadian grade 8 students. The most proficient in French produced more than twice the number of language-related think-aloud episodes and paid more attention to grammar than the least proficient ones, in spite of the fact that all had been all exposed to French since childhood as part of an early immersion program. Further research has also suggested that the development of the discourse-related skills reported above is more arduous in non-cognate languages, where the difficulty of the L2 may force the writer to approach the composition as a mere collection of loose sentences (Henry, 1996). The developmental consistency of the above findings has been contradicted, however, by a cross-sectional study dealing with both secondary and university students. Tarone, et al. (1993) reported that from grade 8 to University level, the written products of a number of South East Asian ESL students showed a striking lack of development on a variety of linguistic and textual dimensions. Age of arrival and age of entry into the school system were suggested by the authors as factors that could account for these surprising results: the 8th graders were the youngest on arrival in the USA and, as a result, were the only participants who had entered the school system at the pre-school stage. As it stands, however, the study has a serious methodological flaw. For cross-sectional studies to be valid, subjects must all have the same starting point. Otherwise, one is not measuring the different stages of a supposedly similar developmental processes but, instead, totally different processes. Unfortunately, the starting point (age of arrival, grade of entry and length of residence) in this study was not maintained constant for all subjects and, as a result, what was intended to be a cross-sectional study cannot genuinely be regarded as such. Developmental traces have also been identified in L2 writers’ metacognitive knowledge. It seems that not all its components are equally stable and, as a result, equally prone to development. Kasper (1997) found that the personal and task components of the metacognitive knowledge of a number of ESL college students did not change as they progressed from the intermediate to the advanced level of L2 proficiency over two semesters. However, a significant increase in the strategic component (planning, monitoring and evaluating) was reported. In addition, this component turned out to be a good predictor for the acquisition of the personal and task components. A CRITICAL EXAMINATION OF L2 WRITING PROCESS RESEARCH 29

The demands involved in the handling of different text types have also been considered from a developmental perspective but results are not clear cut. Way et al. (2000) found that, among novice FL high school students of French, the accuracy and fluency of their compositions were reliably higher for descriptive than for expository texts. This suggests that at initial stages the capabilities of L2 writers seem to be limited by the task type being used. In fact, the difficulty of some genres over others has been appreciated in that the completion of argumentative relative to narrative tasks seems to involve more instances of rescanning the already written text and of rereading the prompt (Raimes, 1987), more problem-solving episodes (Cumming, 1989), as well as more advanced planning, longer time spent on writing and more language switches (Qi, 1998). Also, the linguistic and rhetorical demands of narration seem to be higher than those involved in description (Koda, 1993). But, in contrast with these findings, lack of differences across text types have also been reported. Miller (2000), for example, found a striking similarity in pausing behavior, within a group of L1 and ESL writers, when tackling a descriptive and an interpretative academic essay that were supposed to involve different rhetorical patterns. The result was interpreted either as a lack of sensitivity to the different demands of the tasks or as unwillingness or inability to react to the directives suggested in the prompt. Strangely enough, the participants recognized in post-writing interviews that each task involved different discourse demands, which suggests that there was a gap between their declarative and procedural knowledge or that factors such as interest or motivation might also be playing a role in this context. Along the same lines, Roca et al. (1999) reported that the use of restructuring, understood as a formulation strategy involving either the ideational/textual upgrading of the upcoming text or serving a compensatory purpose, was qualitatively similar regardless of text type demands (argumentative or narrative). The use of the L1 for the production and rescanning of L2 texts has also been looked at from a developmental perspective. Yet results are controversial regarding this issue in connection both with writers’ proficiency level or writing skill and with their degree of familiarity with certain topics. On the one hand, it seems that using the L1 may be beneficial for the exploration of ideas among those writers whose L2 resources are so limited as to impede discovery of meaning, the recall of past experiences, or the access to greater lexical variety through the second language alone (Kobayashi & Rinnert, 1992; Lay, 1982, 1988; Uzawa & Cumming, 1989), although an excessive dependence on the L1 may prevent them from becoming progressively aware of the L2 audience's typical expectations (Kobayashi & Rinnert, 1992). On the other hand, it seems that the use of the L1 at higher levels of L2 proficiency is not at odds with the production of more sophisticated writing (Cumming, 1989; Lay, 1988), or with the efficient encoding of ideas, development of thoughts, verification of lexical choices when the writer must face high level task demands (Qi, 1998). In addition, L1 use to retrace the already written text in more elaborate ways than just rereading it word by word is an ability that might be associated with already developed forms of writing skill (Manchón et al., 2000). Inconclusive results have also been obtained in the analysis of the relationship between L1 use and topic familiarity. While Friedlander (1990) found that, among Chinese ESL students, the use of topic-related language to plan content resulted in 30 ROCA DE LARIOS, MURPHY & MARÍN better planning and better compositions written in L2 on two culture-specific topics (Chinese and American), Akyel (1994) could not replicate these findings. Her informants, Turkish EFL students at two different levels of L2 proficiency, wrote plans in L1 and L2 and compositions in L2 on three topics, one of which was culturally related to the L1, another to the L2, while the third was neutral. In the higher proficiency group, the language of planning made no difference to scores on the plan or the essay in any of the three topics. With respect to the lower group, planning in English turned out to have a better effect than planning in Turkish on the quality of the plan for both the L1 topic and the L2 topic, although no significant effect was recorded on the quality of the compositions. In other words, matching the culture- bound topic to the appropriate language did not seem beneficial for these writers. Akyel tried to account for this difference in results by appealing to educational/cultural, textual and task-related factors. For one thing, the different educational and cultural background of the participants in both studies might have given rise to different attitudes towards the use of the L1 in planning and composing processes. In addition, the fact that, unlike those in Friedlander’s study, her participants were required to use the same rhetorical pattern for the three tasks and to compose within a limited period of time should be regarded as additional reasons for the reported differences. In spite of the contradictions reported, the findings above have broadly differentiated between skills reported to be more conducive to being acquired and those whose acquisition is more arduous or needs further experience or instruction. This respect, writing skills should be seen as following a course of development similar to that proposed for other contexts involving skill acquisition. In this framework of development, known as ‘hierarchical skill integration’ (Bereiter, 1980) writers cannot integrate all the skills at once because of limitations on the capacity of working memory and thus they integrate whatever skills they can cope with at any given moment. As the use of these new skills becomes fluent or, at least, not so demanding of cognitive resources, writers will go on to integrate subsequent skills. As suggested by Valdés et al. (1992), the importance of this field of inquiry from a practical perspective derives from the fact that existing rating scales, such as the ACTFL, the TWE and the like, reflect a view of writing that is not based on a coherent view of writing development.

6 L1 AND L2 WRITING: SIMILARITIES AND DIFFERENCES Our purpose in this section is to review the relevant literature in order to assess the extent to which writers' acquisition of L2 writing competence is related to extant L1 writing abilities. The studies reviewed have been arranged in two main groups: intra-subject and inter-subject comparisons. In the first case, researchers have analyzed the writing processes of the same individuals while performing a writing task in their L1 and L2 respectively. The inter-subject approach uses data from L2 writers' composing processes and compares them either with those of native speakers performing the same task, or with the findings of L1 writing process-oriented research (see Silva, 1993). According to Cumming (1994), it is immediately apparent A CRITICAL EXAMINATION OF L2 WRITING PROCESS RESEARCH 31 that the design of the former approach is more valid than that of the latter. When the performance of L2 writers is compared with that of native speakers, it is almost impossible to tell if the groups compared are truly equivalent or have been affected by variables beyond control.

6.1 Intra-subject studies: Similarities Intra-subject similarities have been reported for a number of different behaviors including overall writing patterns, planning, and revision processes. L2 writers have been reported to use similar patterns in both languages in terms of their holistic approach to the task (Arndt, 1987; Edelsky, 1982), organization of goals (Skibniewski, 1988), representation of the rhetorical problem (Moragne e Silva, 1988), use of problem-solving strategies (Cumming et al., 1989), alternation of writing/pausing patterns (Pennington & So, 1993), linear or recursive way of composing (Thorson, 2000), repertoire of strategies used, and percentage of composing time devoted to formulation (Roca de Larios et al., 2001). Of special interest in this context is the study by Uzawa and Cumming (1989). These authors reported that their writers individually approached the L2 task in the form of a dilemma or as a kind of ‘mental dialectic’ which involved either keeping up the standard of their mother tongue writing by means of rehearsing, organizing information in the L1, etc., or else lowering the standard by reducing the amount of information to be covered, downgrading the syntax and the vocabulary or avoiding audience concerns. One of the merits of this distinction is that it has helped us to see the issue of similarities across languages not only as a question of automatic transfer but also as a deliberate pragmatic choice. In other words, because different strategies may come into play in different contexts depending on different purposes and tasks, Uzawa and Cumming opened up the possibility of including variables other than L1 writing, such as task demands, writers' motivation and purposes in the study of composition transfer. Valdés et al. (1992) provided support to this view when they claimed that the transfer of L1 organization skills shown by their writers (American University students of Spanish as a FL) would probably not have been so automatic if a different type of task, demanding more cultural authenticity, had been set. In that case, Valdés et al. speculated, the students would have had to ‘restructure’ the voice and style acquired in their L1 to accommodate it to the new demands of the L2, which would have involved much more than a simple transfer of writing skills. From the point of view of planning, both children and adults have been found to activate similar procedures across languages. Children have shown similarities in a number of planning behaviors including topic selection time, prior awareness of text ending, amount of text to be written, use of title, choice of genre, and audience awareness (Fagan & Hayden, 1988). Adults, in turn, have been found to activate abstract goal-setting tendencies (Jones & Tetroe, 1987; Whalen & Ménard, 1995), ideational content (Gaskill, 1986; Sasaki & Hirose, 1996; Uzawa, 1996) and attention to overall organization before and while writing (Sasaki & Hirose, 1996) similarly in both languages. The fact that Sasaki and Hirose’s writers were able to focus on organization while writing their texts runs counter to one of Jones and Tetroe's 32 ROCA DE LARIOS, MURPHY & MARÍN

(1987) lesser-known reported constraints of the L2 on planning: the limitations it imposes on writers’ ability to keep track of plans or to add plans on-line while writing. This lack of consistency might have been a result of the data collection procedures used (the retrospection technique used by Sasaki and Hirose might have led subjects to say what was expected of them) or, alternatively, of the participants' differences in writing ability. Similarities have also been found for revision behaviors concerning both the levels at which changes were made and the stages of the composition at which revisions concentrated. Meaning and grammatical/mechanical revisions have been found to predominate in both languages, with writers engaging these procedures mostly in the actual drafting of texts rather than between drafts (Gaskill, 1986; Hall 1990, 1987). For these authors, this set of similarities may be indicative of an underlying system of revision which might initially be shaped in the L1 and then transferred to the L2, although they also leave open the possibility that revision knowledge may be gained through the L2. In that case, transfer should be conceived of as a bi-directional and interactive process. The vast majority of findings reported above lend support to Cummins’ (1980) Interdependence Hypothesis and are a confirmation both of Jones and Tetroe's (1987) assumption that composing similarities across languages should be expected basically at a qualitative or substantive level and of Cumming et al.'s (1989) similar claim that empirical studies which have found similarities between L1 and L2 literate performance are those which have assessed reading and writing behaviors holistically, primarily tracing overall processes.

6.2 Intra-subject studies: Differences A number of differences between L1 and L2 writing have also been documented in the literature. The vast majority of these differences are closely related to students' perceptions of what L1 and L2 writing entails and to the use of planning, formulation and revision processes. The way cross-linguistic differences are determined by educational and cultural values has also been investigated. Students' perceptions and thinking have confirmed the existence of differences between their own L1 writing ability and their developing L2 composing ability. In Silva (1992) thirteen international students from eight different countries with six different native languages were asked about the differences they perceived. On average, more than four differences were reported per subject. They fell into fifteen categories which ranged from the need for more time to understand the L2 writing assignment to the consideration of L2 writing as more form-focused, laborious and time-consuming. Differences between L1 and L2 planning have also been documented. From a textual point of view, there seems to be a decrease in the number of goals generated in the L2 (Skibniewski, 1988) as L2 writers tend to concentrate their attention on the morpho-syntactic and lexical levels at the expense of the rhetorical and textual dimensions of the composition process (Whalen & Ménard, 1995). It also seems that the number of ideas planned that are actually incorporated into the text tends to be A CRITICAL EXAMINATION OF L2 WRITING PROCESS RESEARCH 33 lower in the L2 task in spite of their qualitative similarities across languages as assessed in terms of levels of abstraction (Jones & Tetroe, 1987). The constraints imposed by the target language may lead writers to envisage that their command of L2 expression is not as wide as in their L1, leading them to ‘reduce’ their plans accordingly. Nevertheless, it has also been reported that on some occasions a higher level of attention was paid to the generation of ideas (Moragne e Silva, 1988) or the writing of notes (Hall, 1990) in the L2 condition. Although apparently contradicting Jones and Tetroe’s (1987) results, these findings may also be taken to be manifesta- tions of the constraints imposed by the L2 on the writing process. In fact, in neither case did the work on idea generation result in the production of material to be incorporated into the text, but rather in the clarification of content or in rehearsal. This was not the case, however, with the behavior of some of the writers analyzed by Gaskill (1986). Due to their knowledge of the topic, some of them planned and incorporated more ideas in the L2 than in the L1 task, which seems to suggest that simply claiming that the L2 constrains planning may be a sweeping generalization if other factors are not taken into account. The transition from planning to formulation processes also shows some revealing differences. Difficulties have been attributed to the translation into the L2 of goals initially generated in the L1 (Moragne e Silva, 1988; Whalen & Ménard, 1995), the lack of L2 appropriate rhetorical schemata to convert goals into textually appropriate prose (Moragne e Silva, 1988; Sasaki & Hirose, 1996), the retention of goals in working memory when formulation is in progress (Moragne e Silva, 1988), and the inability to handle long stretches of L2 text so that rhetorical procedures already at the writer's disposal can be embodied (Henry, 1996; Ting, 1996). From the perspective of formulation proper, differences are visible above all in handwriting and in fluency. In handwriting, the possibility of alternating the cursive and the manuscript seems to develop earlier in the L1 than in the L2 among children (Edelsky, 1982). As for fluency, L2 transcription processes appear to be generally harder to sustain and seem to be less efficient in terms of the number of words written (Chenoweth & Hayes, 2001; Ransdell & Levy, 1998; Ransdell et al., 2001; Sa- saki & Hirose, 1996; Silva, 1993; Thorson, 2000). They are also more fragmented, as attested by the higher number of pauses (Ransdell et al., 2001), interruptions of the actual transcription process by other processes and by the lower number of words produced between pauses (Chenoweth & Hayes, 2001; Fagan & Hayden, 1988; Krings, 1994; Roca de Larios et al., 2001; Whalen & Ménard, 1995). In addition, interruptions are more likely to occur at clause boundaries when writing in the L1, their location in the L2 being more erratic (Ransdell & Levy, 1998). These features of L2 transcription processes have obvious consequences for the fragmentation of thought. Second language writers run the risk of overloading their working memory with linguistic concerns so that the coding of ideas and all related issues - association, elaboration, chunking- are constrained (Kobayashi & Rinnert, 1992; Pennington & So, 1993; Yau, 1991). In close connection with these findings, dual task experiments have shown that L2 formulation processes seem to require central working memory resources since texts produced in L2 have been found to be of in- ferior quality to those produced in L1 even when the process of L1 production is accompanied by irrelevant speech or by a 6-digit load (Ransdell & Levy, 1998; 34 ROCA DE LARIOS, MURPHY & MARÍN

Ransdell et al., 2001). This finding fits in well with experimental studies that have found indications of performance loss on picture and number naming, grammaticality judgements, and integration of semantic and syntactic information when the subjects approach these cognitive operations in their L2 (see Jeffery, 1996). Available evidence, however, seems to indicate that some of the above tendencies -especially those related to number of words written, number of pauses, pause location and dis- ruption of text quality- do not apply to bilinguals who have shifted their dominant language or who find school writing in English more familiar than in their native language (Ransdell et al., 2001). Differences have also been reported for revision. There are generally more revision episodes in the L2 (but see Fagan & Hayden, 1988), seemingly because in this condition writers have to intensify their efforts to make their texts reflect their writing goals (Hall, 1990) or because of the writer's lack of familiarity with the L2 material (Arndt, 1987; Thorson, 2000). This fact is closely related to the writers' emphasis on linguistic revisions of the morphemic type at the expense of lexical, textual and pragmatic considerations (Whalen & Ménard, 1995), an indication of the stronger saliency effect (Bereiter & Scardamalia, 1987) that the L2 imposes on writers. In this respect, Moragne e Silva (1988) noted that lack of concern over low-level problems was detrimental for her participant as many errors went unnoticed during reviewing. This behavior might be an indication that the writer’s level of frustration was lower in the L2 and, in this sense, similar in nature to the lack of obsession with error reported for L2 writers as compared to L1 writers (see 6. 3. 2 below) or, alternatively, a consequence of how the writer perceived the context of production: he knew his compositions would not be graded. Still in relation to cross-linguistic differences, reference can also be found in some studies to a number of social and contextual factors. These include the quantity and quality of previous literacy experiences, the role played by genre, and the hid- den L1 literacy assumptions that writers carry over to the acquisition of L2 literacy. From the first perspective, differences in patterns of literacy transfer have been attributed to the differential contact with L1 texts via which writers may have developed strategies applicable to the L2 condition (Bosher, 1998; Cumming, 1989; Cumming et al., 1989). For example, among one group of allegedly unskilled writers, only those who had read or written extensively in their mother tongue were able to carry over the capacities thus acquired to the L2 English tasks (Brooks, 1985). Carson et al. (1990), analyzing written products, found a significant correlation between the L1 and L2 texts produced by their Japanese ESL students in contrast to the non-significant correlation shown by the texts produced by their Chinese ESL students. According to the authors, the explanation may lie in cultural differences and in attrition processes among the Chinese writers who, though more proficient in the second language, had not used their L1 writing skills for a long time. Previous literacy experiences may also become an important factor mediating the transfer of composition skills where the use of the mother tongue is concerned. Cohen and Brooks-Carson (2001) found that an intermediate group of Spanish-English bilinguals studying French as a foreign language in an American University reported thinking or translating mentally into English rather than Spanish while writing a composition in French. This was unexpected given the proximity between Spanish A CRITICAL EXAMINATION OF L2 WRITING PROCESS RESEARCH 35 and French. The authors attributed this finding to the fact that American foreign language textbooks use English as the language for explanations. Moreover, these bilinguals were used to functioning in an English language University. In other cases, the differences between L1 and L2 writing processes have been accounted for in terms of genre. Thorson (2000) reported that, while a letter and an article prompted different composing processes in L1 writing, their effect on FL writing remained unclear. Yet the study in which social and cultural influences on transfer were brought out most clearly was probably Bell’s (1995). Through her own initiation into Chinese literacy, the author uncovered the unconscious assumptions about L1 literacy that one brings to bear on the process of acquiring literacy in a distant language and analyzed how these might influence the development of new L2 skills. The study makes it clear that acquiring a new literacy, such as Chinese, does not mean acquiring just a new linguistic code, but basically a new set of values. Bell described herself as ‘trying to move from being a left brain learner to a greater reliance on the use of the right brain’ (Bell, 1995: 702). A difference is thus posited between learning to transcribe English thinking via Chinese characters, which is undoubtedly possible, and developing Chinese literacy, which involves being aware of a whole set of unspoken assumptions. For the most part the differences reported above tend to be associated with the extent to which particular processes are used, and, from a cognitive angle, they reflect the constraining effect imposed on them by the L2. From a socio-cognitive perspective, it is stressed that that becoming literate in a second language involves something more than a mere ‘technological’ transfer of skills determined solely by proficiency. It should rather be seen as a far more complex process in which a great number of educational, social and cultural factors are involved.

6.3 Inter-subject studies As suggested above, the comparison of L1 and L2 writers' performance uses data from L2 writers' composing processes and compares them with those of native speakers performing the same task or with the findings from L1 writing process- oriented research. As very little information related to transfer can legitimately be inferred by comparing subjects whose personal and contextual variables may be so different, it is only logical that most studies have focused on the variable which lends itself most easily to comparison: the informants' writing ability. From this perspective, research has shown that, as in the case of intra-subject comparison, both similarities and differences can be found between L1 and L2 writers.

6.3.1 Similarities between L1 and L2 writers Because similarities between L1 and L2 writers’ composition processes have generally been reported in connection with their level of skill, there is a great deal of over- lapping with the findings reported above about the skilled-unskilled distinction. Thus, these similarities will be only briefly presented here. 36 ROCA DE LARIOS, MURPHY & MARÍN

Both efficient and inefficient composition strategies have been found to be used by L1 and L2 writers in similar ways. Among the former, the following have been emphasized from an overall perspective: (1) the flexible construction of working goals (Sasaki, this volume); (2) the conception of written composition as a complex endeavor which calls for the division and prioritizing of tasks and subtasks, the management of which is made possible by the efficient use of the recursive nature of writing (Zamel, 1983); (3) the use of problem-solving procedures operating on complex mental representations involving different discourse levels (Cumming, 1989); and (4) the interactive approach to text construction, as shown in the appropriate balance of writing processes set in motion (Lay, 1982; Roca de Larios et al., 2001; Victori, 1995), and in the time and mental effort invested in getting the message across (Victori, 1995). In addition, some strategies such as word choice (Cumming, 1989; Lay, 1982; Sasaki, this volume; Victori, 1995), going back and forth in the text (Lay, 1982; Raimes, 1987) and the ability to cover and structure major conceptual units when taking notes (Faraco et al., this volume) have also been reported. Conversely, a less efficient use of strategies has been reported for writers' (1) lower effort and commitment to the task and less interaction between strategies because of their smaller number and variety (Victori, 1995); (2) over-dependence on the already written text to move forward (Raimes, 1985; Roca de Larios et al., 2001; Uzawa, 1996; Victori, 1995); (3) rigid and formulaic conception of planning (Raimes, 1985, 1987; Uzawa, 1996); (4) tendency to edit their texts from the beginning to the end of the process (Zamel, 1983); (5) reluctance to reread their texts for fear of not being able to handle grammar mistakes (Brooks, 1985; Uzawa, 1996); (6) tendency to revise their texts at the surface level (Raimes, 1985); and (7) lack of concern with audience demands (Raimes, 1985, 1987). We have already reported findings relative to the stability of some composing procedures across languages in the studies reviewed in 6. 1 above. An additional issue is the consistency with which L2 writers engage in the different composing processes across time and tasks. In this respect, Jones and Tetroe (1987) found that abstract planners maintained their level of abstraction across task type (conventional vs. ending-sentence compositions) and writing sessions across time (November, February and April), while Hall (1991), analyzing whether writers behaved differently while composing under examination and non-examination conditions, found that participants, ESL graduate and undergraduate students, allocated their time to prewriting, inscribing, and post-drafting activities in similar ways in the two situations. All these studies confirm similar results in L1 writing research regarding the stability of writers’ profiles (Levy & Ransdell, 1996; Torrance, Thomas & Robin- son, 1999, 2000).

6.3.2 Differences between L1 and L2 writers The main differences between the composing behavior of L1 and L2 writers are of both a qualitative and quantitative nature. Qualitatively speaking, the ESL students' mental models of writing have been found to favor either grammar or communication as opposed to those of their L1 counterparts, which are more integrative in na- A CRITICAL EXAMINATION OF L2 WRITING PROCESS RESEARCH 37 ture (Devine et al., 1993). Multilingual and bilingual writers on the other hand, when writing in their L1, seem to be less prone than monolingual writers of comparable verbal and non-verbal skills to having their fluency and text quality disrupted by secondary tasks such as irrelevant speech and a 6-digit load, an indication that training and experience in a second language may provide writers with sufficient working memory flexibility to be used in dual task situations to improve performance (Ransdell et al., 2001). In addition, and contrary to current L1 composition models (Bereiter and Scardamalia, 1987; Flower & Hayes, 1981), planning before writing has not been found to be a clear stage in the composition process of some skilled EFL writers (Smith, 1994). As an alternative, Smith suggested that formulation should be taken to be the main composition process in L2 writing, since the interaction between lexical retrieval and syntactic processing was found to be one of her FL writers' major concerns. Similarly, in contrast to most findings in L1 writing research where scanning the already written text has been associated with revision, re-readings were reported to be used mainly to move forward through the text (Raimes, 1985; Raimes, 1987) because L2 writers did not appear to be so obsessed with error and editing as their L1 counterparts (Raimes, 1985; Porte, 1997). Finally, under similar contextual conditions, L2 writers seem to be more committed to the completion of the task than basic L1 writers (Raimes, 1985; 1987). From the quantitative angle, a number of important differences related to various processes have been reported. First, L2 writers tend to pause longer at word, clause and sentence completion levels, which suggests particular pressures of lower levels of word formation and of planning within the sentence that are not apparent to the same extent in L1 writers (Miller, 2000). L2 students were also found to rehearse a great deal. The main use of this strategy for the production of ideas was interpreted as an indication that concern with meaning was more important for ESL writers than the accuracy of their linguistic expression (Raimes, 1987). In fact, one of the procedures used to compensate for the lack of resonance that the L2 material might have for them (Leki, 1992) involved the use of their L1 words to prompt their thinking and narrow down the intended meaning (Lay, 1982, 1988; Cumming, 1989; Fried- lander, 1990; Qi, 1998). Second, L2 writers, probably as a result of their imperfect command of the L2, wrote fewer drafts and edited less than their L1 counterparts (Raimes, 1985). They also read whole drafts less (Raimes, 1985), although in this case the author suggested that the difference might be attributed to the nature of the narrative task used in her study. This text type, unlike the argumentative ones used in comparable L1 studies, might have made it unnecessary for the participants to reread chunks so as to establish the course of the plot. In this sense, Raimes’ explanation is coincidental with research into blind L1 writing (Britton et al., 1975) where participants reported being less concerned about not being able to reread the already written text in narrative writing than when engaged in more complex tasks. Yet, in contrast to these findings, Manchón et al., (2000) found that recourse to backtracking was slightly greater in a narrative than in an argumentative task, probably as a result of topic familiarity of the latter. Finally, L2 writers surprisingly show a tendency to convey the intended meaning more efficiently by changing not only words -the characteristic ‘thesaurus philosophy’ usually associated with unskilled L1 writ- 38 ROCA DE LARIOS, MURPHY & MARÍN ers (Sommers, 1980)- but also clauses and phrases in the final versions of their texts (Porte, 1996). A number of reasons have been suggested for the differences between second language and L1 writers (Grabe & Kaplan, 1996; Leki, 1992) in terms of uses of, and needs for writing. For one thing, the L2 writing population cannot be regarded as homogeneous at all. Its broadest distinction between ESL and EFL learners cov- ers a wide array of subgroups which may need writing proficiency to varying degrees and for various purposes. On the other hand, there are a number of variables which apply to mother tongue writers but which may not affect L2 writers' performance and vice-versa. Linguistically speaking, for example, both groups have very different strategies available to them. L1 writers tend to use top-down processing strategies while L2 writers rely more on bottom-up procedures, as they tend to be less capable of going beyond the written text and seeing beyond it. From the personal angle, for the L1 writer it is their very sense of identity which is at stake since self-perception is very much linked to one's native language, while for L2 writers failure is stigmatizing but only to the extent that failure in a foreign language can be. Also, writing may have a considerable impact on and be crucial to the future careers of L1 writers while its effect may be fairly negligible for the professional lives of ESL students. Finally, both groups may also differ in the role writing plays in their respective native cultures. In the countries of many ESL students writing is highly valued both from the cultural and the social points of view. In contrast, some basic writers have seldom been taught any respect for their linguistic traditions or dialects (Leki, 1992).

7 THE RELATIONSHIP BETWEEN WRITING ABILITY AND L2 PROFICIENCY The relationship between writing ability and L2 proficiency cannot be adequately understood if the notion of modularity, which underlies the whole discussion, is not put into perspective. Current models of language production (Dell, 1986; Levelt, 1989) although in disagreement about the precise form, agree on a number of characteristics that seem to be essential for the functioning of the system. One of them is the decomposition of the complete mechanism of language production into a number of subsystems or modules, each of which is associated with a knowledge base which contains its characteristic information. The relationship between these modules has traditionally been viewed in terms of either autonomy or interaction (Belinchón, Riviére, & Igoa, 1992). Within the first conception, the modules are seen as autonomous subsystems in charge of specific tasks and with access to a limited amount of information. They are also taken to be hierarchically organized so that each one can only receive information from the specific module(s) it is directly connected to since no feedback loops are envisioned from lower to higher levels. In contrast, the interactional conception, while accepting the idea of modules, rejects the notion of processing autonomy: interaction and mutual influence among the different modules of the system is posited as its essential characteristic. Accordingly, information can be forwarded or fed back from any module to any other module in A CRITICAL EXAMINATION OF L2 WRITING PROCESS RESEARCH 39 the system irrespective of their position in the hierarchy. For example, the availability of lexical representations may determine that a specific idea be conceptualized: a feedback loop may be sent from the lexical module to the module in charge of the conceptualization allowing for the concept to be produced or, otherwise, avoided. Writing, like any other language production process, involves the use of different modules with their corresponding knowledge bases: knowledge of the topic, the audience, and the composing process (usually operationalized as writing ability) as well as linguistic knowledge (usually operationalized as language proficiency). But in contrast to the idealization of modules used to account for L1 production processes, the characteristic of L2 writing is that the knowledge bases dealing with language -especially the lexical and the syntactic ones- are limited. Hence the relevance of the study of this relationship, which to a greater or lesser extent, underlies most studies of L2 composition processes. The studies investigating the relationship between writing skill and L2 proficiency have produced mixed results. Thus, it is not possible to make a simple statement about the overall issue. Generally, studies may be grouped into two tendencies: those which have gathered evidence in favor of the independence of both constructs, and those which have empirically supported their interaction. In both cases, their explanatory power has been directly related to the number of subjects analyzed, the statistical procedures used, and the way variables have been controlled and data collected. Although all these issues should be critically examined, a detailed analysis of these elements of the research design is beyond the scope of this study. We will limit our methodological concerns to the way writing ability has been handled. Generally, those studies presenting empirical evidence in favor of the independence of writing ability and L2 proficiency have operationalized the ability to compose as related to the use of strategies and their corresponding representational levels. A review of the literature shows that the following strategies do not seem to be constrained by the writer’s level of L2 proficiency: (1) using heuristic searches to evaluate and solve problems comprising gist, language, and discourse organization in co-ordination (Cumming, 1989; Cumming et al., 1989); (2) risk taking, revising at all levels, having a sense of audience (Brooks, 1985; Zamel, 1983), reasoning at higher levels of generalization, and relying on visual memory rather than auditory feedback to evaluate written texts (Brooks, 1985); (3) setting goals, organizing ideas and expressing them coherently (Smith, 1994); and (4) planning, monitoring and evaluating performance (Victori, 1995). However, this black and white panorama has been shaded in, on an individual basis, by Bosher (1998) who found that South Asian students who had been placed in the same instructional program and had similar overall L2 proficiency scores and similar writing sample scores on entry did not show the same equivalence in writing processes. The main conclusion to be drawn from this study is that the relationship between writing ability and L2 proficiency should not be viewed as an all-or-nothing business but rather as a continuum. From the point of view of the interaction of both constructs, it has been found that the L2 may impose constraints when writing ability is understood in two specific ways. First, the production and co-ordination of complex ideas may be limited because the expression of these is usually associated with the lexical and syntactic repertoire of the writer (Yau, 1991). Second, the implementation of strategies in- 40 ROCA DE LARIOS, MURPHY & MARÍN volved in register variation and the use of certain rhetorical patterns may be constrained as both depend on the writer's ability to handle long stretches of text (Ting, 1996). There are also a number of studies which, in spite of sharing their interest in the relationship between writing ability and L2 proficiency with other research concerns, have nevertheless produced a number of interesting results about the interaction of both constructs. It seems that the fewer cognates the languages have in common, the more interaction between writing ability and L2 proficiency is to be expected (Henry, 1996). Another finding is that, at similar levels of writing ability, an increase in L2 proficiency seems to require from writers more than a simple application of their existing L1 abilities: apparently, as suggested in 6. 1 above, some restructuring of their writing procedures is required to accommodate them to the specific demands and stylistic conventions of the L2 task (Valdés et al., 1992). Also, at similar levels of note-taking ability, L2 proficiency seems to play a part in the misin- terpretation of the original material (Faraco et al., this volume). Finally, L2 proficiency seemingly has greater explanatory power to account for L2 writing performance than writing ability (understood as scores on L1 writing composition) when the subjects under consideration are EFL students (Sasaki & Hirose, 1996). The reason seems to be that these subjects usually acquire their L1 writing ability and their L2 proficiency through formal education, with the result that both abilities develop relatively evenly because they are both related to aptitude for academic achievement. This would be a typical example of how, in an EFL context, writers still in the process of learning the L2 would also be developing their cognitive/academic proficiency (Cummins, 1980).

8 LIMITATIONS AND SUGGESTIONS FOR FUTURE RESEARCH Apart from the contradictions in the findings of some studies within and across research areas (some of which have been discussed above), from a substantive point of view the most important limitations of the studies we have reviewed concern problems related to the conceptualization of writing skill in L2 writing, the scant attention paid to formulation behaviors, the neglect of the intrinsically temporal character of the composition process, the little specificity of the underlying cognitive mechanisms that may facilitate or hinder the transfer of writing skills across languages, and the lack of attention paid to the contextual parameters involved in L2 writing processes. Each of these areas will be discussed in turn.

8.1 Problems related to the notion of skill in L2 writing The application of the skilled/unskilled distinction to L2 composition research has been laden with problems because it was not clear from the outset what being a skilled second language writer meant. Raimes (1985), for example, recommended caution with the meaning of the term ‘unskilled’, because she found that, unlike unskilled L1 writers, her unskilled L2 writers did not form a homogeneous group. Nu- merous individual differences were observed and neither the length of residence in A CRITICAL EXAMINATION OF L2 WRITING PROCESS RESEARCH 41 the L2 country nor the level of L2 proficiency were found to be reliable predictors of specific composing behaviors. Raimes suggested that the notion of skill in L2 writing might be best captured as a composite of variables, in each of which writers could be judged as more or less skilled: language proficiency, product quality, self- evaluation of L1 and L2 writing ability, knowledge of writing demands, writing background and teaching experience. Against this background, three different limitations have been pointed out concerning the use of the skilled/unskilled distinction in L2 composition research. The first criticism relates to the lack of clear criteria for assessing skill in writing in general and in L2 writing in particular, a context where the L2 proficiency/writing ability dichotomy is so conspicuous. Although classifying informants as more or less skilled according to the quality of their finished compositions is an ‘unspoken assumption of much writing process research’ (Pennington & So, 1993: 58), this type of assessment either takes the writer’s L2 language proficiency into account or takes for granted a direct relationship between the composition process and the written product. The first issue is so thorny that, although researchers like Zamel (1983) would argue that writing ability, as determined through the writer’s approach to the task, is the decisive factor in L2 composition, she herself judged her subjects as skilled or unskilled using a holistic assessment of their papers rather than an analysis of their composition processes and thus inevitably took their L2 language proficiency into account. In addition, research has shown that the measures used to assess the quality of compositions are far from stable. They seem to vary as a function of the writing context (Hall, 1991) or of the raters’ preference for accuracy or amount of information conveyed (Henry, 1996), their cultural values and/or previous experience (Kobayashi & Rinnert, 1996; Rinnert & Kobayashi, 2001), or the purpose of the course they are teaching (Cumming, 2001). As for the second issue, results from research about the relationship between efficient writing strategies and high ratings of written products are contradictory (Pennington & So, 1993; Raimes, 1987). Dif- ferent explanations have been put forward for this state of affairs (Pennington & So, 1993). For one thing, the idea of a poor written product as the outcome of what can be regarded as a skilled composition process is logically possible, especially when the writer’s command of the L2 is limited (Raimes, 1987). In addition, individual differences among writers' composition processes are so wide (Arndt, 1987; Smith, 1994) that strong generalizations about necessary connections between process and product are hard to make. Finally, it is posited that the different ways in which the notion of process has been operationalized may also have a bearing on the difficulties of generalizing results across studies (Pennington & So, 1993). Thus, it may be argued that in any comparison among studies a distinction must be made between evaluating writing skill through process or product (Moragne e Silva, 1988). The second argument against the skilled-unskilled dichotomy is that the distinction between the two becomes blurred on many occasions, especially when writers have to tackle certain task-types that do not demand much problem-solving (e.g. a basic narrative). Students are often asked to do assignments based on their personal experiences, feelings or opinions (Stotsky, 1995) in the belief that writing about what they know will increase their degree of involvement and motivate them to work harder on their composing process. The effect, however, is usually the oppo- 42 ROCA DE LARIOS, MURPHY & MARÍN site as the easy access to information that is readily available and already organized in one’s memory reduces the need for the use of heuristics (Graham & Harris, 1997). As a result, many efficient student writers convert the writing task -whatever it is- into the task of telling what they know, which might appear to be behavior typical of immature writers. In these cases the topic and the already written text alone are sufficient to prompt, via spreading activation, the retrieval of related ideas which are usually incorporated into the emerging text. Little explicit, analytic attention is paid to rhetorical concerns, audience needs, organization, and so on. Thus, apparently ‘skilled’ writers may behave in an unskilled way when they perceive that that behavior is enough to meet the requirements of the task. Intimately linked to those ideas, the next criticism has derived from the assumption that the terms ‘skilled’ and ‘unskilled’ are relative to the domain they are applied to or the discourse community the writers belong to. Gosden (1996), for example, has claimed that the type of skill required to tackle the time-compressed essay typical of L1 and L2 composition research -where the audience is normally reduced to the teacher and the dominant purpose is for students to show their knowledge- is very different from the abilities involved in the writing up of scientific research for publication. In the latter case, the inquiry would be more focused on clarifying the socio-rhetorical demands faced by writers than on describing the ‘moment-by- moment cognitive demands made on L2 writers facing much simpler tasks’ (Gos- den, 1996: 113). The main conclusion drawn from this approach would be that writing ability is a very complex construct that entails ‘a host of social and cognitive dimensions that may operate differently in different contexts, a wide range of interrelated language abilities, and, perhaps multiple literacies’ (Witte, Nakadate, & Cherry, 1992: 41). It is thus necessary, in order to define what is meant by skill in writing, for the concept to be situated within the context from which the type of task selected to elicit skilled performance has been taken (see also 8. 5.).

8.2 The need to study formulation as a process in its own right As is only to be expected given the writing models underlying most research, the analysis of the vast majority of findings reported above reflects a conception of writing as a controlled, rational process where writing skill is a question of working out what one wants to achieve with the text and then deciding how to do it step by step. It is, thus, a top-down account of writing whereby thoughts are separated from their translation into words rather than seen as part of the process of text production. The consequence has been that, as shown in the review above, formulation has been seen mainly as fluency in spite of the scattered observations on lexical searches and reformulation procedures (Raimes, 1987; Smith, 1994). As a result, the interface between thinking and language -the scenario where real formulation processes occur- has received little attention (Roca de Larios, 1999; Roca de Larios et al., 1999; Zimmermann, 2000). There are, however, at least four areas of research which might benefit from a consideration of formulation as a process in its own right. First, in developmental studies, although planning and revision have not been extensively analyzed, at least they have been the main focus of attention in some of the A CRITICAL EXAMINATION OF L2 WRITING PROCESS RESEARCH 43 studies investigating the paths of development followed by the different components of L2 writing ability. An example would be the development of global versus local concerns. The data seem to suggest that global aspects of planning (Berman, 1994), and the ability to tackle global aspects of revision (Kobayashi & Rinnert, 1998) develop with more ease and in a relatively shorter period of time than more local concerns, such as the ability to handle accuracy and local coherence problems. A useful way of confirming and even supplementing the present findings would involve extending similar developmental analyses to formulation, the locus where, by definition, global and local concerns meet. Second, in the studies reported above three different conceptualizations of the notion of level have been used: (1) abstraction level, as used to describe the degree of detachment of planning operations from actual text (Jones & Tetroe, 1987; Whalen & Ménard, 1995); (2) level as a synonym of linguistic unit (word, sentence, etc.), as used in revision studies (Gaskill, 1986; Hall, 1990); (3) discourse level, understood as specific rhetorical, textual or linguistic representations writers operate on when planning, revising or evaluating their texts (Whalen & Ménard, 1995). An analysis of the processes to which these three notions of level have been applied reveals that they have only been used in connection with planning and revision. In other words, the notion of level has been largely absent from formulation processes (though for exceptions, see Smith, 1994). Yet we think that looking at the mental representations writers are constructing when formulating their texts would be useful at least in two areas. On the one hand, it would help to broaden the conceptualization of formulation so that it is no longer seen as mere transcription, analyzable only from the perspective of fluency. On the other hand, it would probably help us to clarify some aspects of the complex relationship between planning and formulation processes. As shown above (Moragne e Silva, 1988; Sasaki & Hirose, 1996; Whalen & Ménard, 1995), research so far has mainly looked at the issue from the perspective of planning limitations, i.e. by attesting to the difficulties involved in the formulation of higher goals when linguistic resources are limited. The interaction between both processes has, however, been disregarded up to now. The consideration of formulation as a process whereby the writer can manage mental representations which are – to a greater or lesser extent – related to those constructed in planning opens up the possibility of looking at both processes in interaction. Third, within the very consideration of formulation in terms of fluency, there is a contradiction in the way the interruption of transcription by other processes has been interpreted. For Victori (1995), this decrease in fluency was the price skilled writers have to pay for the planning, evaluation and reprocessing of their ideas on-line. In- terruptions were thus seen as a reflection of writing ability. For Raimes (1985), in contrast, the cognitive status of these ‘interruptions’ was rather unclear. Research which focuses on this balance between fluent transcription and the range of processes occurring in interruptions would probably help clarify the issue. Fourth, as compared to other studies, Cumming's (1989) control of the independent variables used and the large number of subjects involved for a think-aloud study (which allowed the author to use statistical procedures which would be impossible in ordinary case studies) suggest that the validity of his findings regarding the relationship between writing ability and L2 proficiency is almost indisputable. Nevertheless, 44 ROCA DE LARIOS, MURPHY & MARÍN there are a number of considerations related to his own research that make one wonder whether the independence of writing ability and L2 proficiency can be so firmly assumed. The coding scheme used by Cumming only analysed those segments of the protocols above and beyond the written text that were regarded as decision-making statements. This means that, in line with problem-solving models of composition (Bereiter & Scardamalia, 1987; Flower & Hayes, 1981), Cumming focused on higher-order thinking processes from the start, leaving aside all the phases of rereading, writing and rehearsing of phrases before writing (Smith, 1994). In addition, the independence of writing ability from L2 proficiency was not found to be equally distributed among the three groups of writers that Cumming studied: it was clear for the expert and basic writers but not for the group of average writers. As Cumming himself speculates, the level of L2 proficiency favored the average group particularly by enhancing their writing ability and indicating that some sort of interaction was taking place. This panorama, together with the findings reported above about the interaction of both constructs, indicates that the relationship between writing ability and L2 proficiency still remains an open question. A possible contribution to the ongoing discussion would involve the analysis of formulation processes (explicitly disregarded by Cumming) as the scenario where, according to research into the temporary store of lexical items in working memory (see Cook, 1991; Kern, 1994), the constraints imposed by the L2 might in principle be more visible.

8.3 The temporal character of composition The review has also shown that very little attention has been paid to the way writers handle the different writing procedures within the course of text production itself, that is, the monitoring component in the Flower and Hayes (1981) model (Rijlaars- dam & van der Bergh, 1996). A serious consideration of the temporal character of composition might serve to address at least four limitations of the studies reviewed. First, the research above has reported efficient uses of recursiveness in composing but mainly in an impressionistic way (Raimes, 1987; Victori, 1995; Zamel, 1983). The only study in which data have been presented in a quantitative form is that of Pennington & So (1993) but no mention is made there of the mental representations the processes are being applied to. A useful alternative would involve the analysis of the temporal distribution of processes in conjunction with the purposes these processes are meant to serve throughout the composition. Second, Raimes (1985, 1987) found that ESL writers, in contrast to their L1 counterparts, showed a greater commitment to the task as measured by overall time spent on composing. Yet time on task may be regarded as a rather crude way of assessing commitment. As suggested by some of the revision studies discussed above (Hall, 1990; Gaskill, 1986), it seems that certain moments in the composition process are better suited to the occurrence of certain behaviors than others. The consequence is that, besides total composition time, more detailed analyses are required that specify how this time is distributed among the different processes in the course of the composition. Third, widely held views in traditional composition models (Bereiter & Scardamalia, 1987; Flower & Hayes, 1981) see planning as an essential stage in the performance of expert writers. A CRITICAL EXAMINATION OF L2 WRITING PROCESS RESEARCH 45

This contrasts, however, with the more flexible attitudes toward the distribution of that process shown by some of the subjects analyzed by Cumming (1989), Smith (1994) and Victori (1995). A possible way of disentangling this contradiction might involve measuring planning in terms of time and analyzing its distribution throughout the composition process across a wide range of different proficiency levels and tasks. Fourth, very little is still known about the variables affecting the functioning of ‘stable’ composition profiles. As shown above, descriptions of both individual writing profiles (Arndt, 1987; Edelsky, 1982; Moragne e Silva, 1988; Uzawa & Cumming, 1989) and overall group approaches to composing (Cumming et al. 1989) have been found to be equivalent across languages. For the most part, however, these descriptions have been based on a consideration of writing processes as static entities. Further analyses are needed to account for these profiles from the point of view of the co-ordination of composing procedures throughout the composition process (Roca de Larios, 1999).

8.4 The cognitive mechanisms of transfer A final limitation refers to the fact that Cummins (1980) used age (maturational development) and the relationships between cognitively demanding tasks in the first and second language as the only evidence to support his Interdependence Hypothe- sis – this second procedure being followed by many second language process research studies, as seen above. Yet no elaboration of the model has been provided with respect to the nature of the underlying cognitive mechanisms that may facilitate that interdependence of skills across languages (Francis, 2000). As a result, the hypothesis remains too general, as does the research derived from it. However, some findings reported above relative to differences of processing in L1 and L2 writing as well as to the interdependence of writing ability and L2 proficiency have shown indications that the trade-off between storage and processing demands (see Lea & Levy, 1999) have a bearing on writing in a second language. Thus the hypothesis should be refined and expanded by looking at L2 writing ability in terms of working memory capacity, processing speed, memory span, ability to analyze language, etc. (see Ransdell & Levy, 1998, and Ransdell et al., 2001, for examples). The analysis in terms of these cognitive processes might help us to clarify what is involved in the notion of common processes underlying L1 and L2 literacy, and would also extend the scope of the relationship between writing ability and L2 proficiency to more specific cognitive areas of concern.

8.5 The situated nature of L2 writing As suggested in the introduction to this chapter, the study of L2 writing processes has been mostly approached from a cognitive perspective. This means that the role played by contextual and social factors in composing has not been given enough attention within these field of inquiry. In fact, one of the central questions in composition research is whether the theoretical discourses underlying the cognitive and the social conceptualizations of writing are compatible. From the cognitive perspective, 46 ROCA DE LARIOS, MURPHY & MARÍN composing is arguably conceived of as a problem-solving task and emphasis is placed on the complex, recursive and individual nature of the writing process, independent of cultural and social influences. Social constructionists, in contrast, do not typically see writing as consisting of invisible processes occurring in the writer’s head but rather as a situated activity (Brown, Collins & Duguid, 1989) that can only occur within a specific context and for a specific audience. However, the dichotomous picture that arises in this description does not do justice to the complexity of writing (for a review of research on L2 writing from this perspective, see Cumming, in press). As shown by recent research in L1 writing (Carter,1990; Flower, 1994; Nystrand, 1989; Pittard, 1999; Witte, 1992), the analysis of composing processes in isolation from the contexts in which they occur may turn these processes into mean- ingless patterns of behavior since the writing task and the writing response to it are framed by social relationships and purposes operating in specific writing situations. In consonance with this tenet, the review above has shown that, although not at the forefront of findings but relegated to the sidelines, some hints can be found in a number of studies showing the situated nature of some L2 composing processes. In fact, while some studies have stressed the importance of time restrictions as condi- tioning the writers’ approach to the task (Raimes, 1985; Uzawa, 1996), others have noted that the way writers perceive task demands and purposes may have an important influence on how they engage in either compensatory or achievement production procedures (Uzawa and Cumming, 1989) or in the way they approach the processes of planning (Cumming, 1989; Smith, 1994) and revision (Moragne e Silva, 1988; Porte, 1996; 1997). Other studies have documented that writers’ perception of genre demands may give rise to differences in the predominance of some composing processes across tasks (Cumming, 1989; Manchón et al., 2000; Raimes, 1987; Qi, 1998), or, alternatively, that failure to perceive these different demands may prompt the unexpected activation of similar processes (Miller, 2000; Thorson, 2000). Still other studies have emphasized the influence of previous literacy experiences (1) in the way certain processes such as planning (Sasaki & Hirose, 1996; Sasaki, this volume) or revising (Porte, 1996; Zamel, 1983) are activated, (2) in writers’ self- concept as L2 writers (Brooks, 1985; Victori, 1995), their attitudes towards the use of L1 in planning and composing (Akyel, 1994; Cohen & Brooks-Carson, 2001), their unconscious assumptions about L1 literacy brought to bear on the process of composing in a distant language (Bell, 1995), their transfer of composing processes across languages (Bosher, 1998; Brooks, 1985; Carson et al., 1990; Cumming, 1989; Cohen & Brooks-Carson, 2001), or (3) in the greater explanatory power of L2 proficiency over writing ability to account for L2 writing performance (Sasaki & Hirose, 1996). Future research will have to advance towards a further understanding of how writers, as individuals shaped by and operating within a social environment, interpret and construct the writing task (Flower, 1994). In this respect, a promising area of inquiry would be the attempt to show that individual writing is also dialogic in nature (Hermans, 1996; Holquist, 1990; Kramsch, 2000), a task that would entail the consideration of the problem-space, the unit of analysis explicitly or implicitly used in most cognitively-oriented studies of L2 composing processes, as the locus of the writer’s internal dialogue (Roca de Larios & Murphy, 2001). In this new interpreta- A CRITICAL EXAMINATION OF L2 WRITING PROCESS RESEARCH 47 tion it would have to be shown that the composing task does not consist only of recurrent processes (planning, rereading, inscribing, evaluating, etc.) but also that the conceptual, linguistic and rhetorical choices involved are just a reflection of the ideational, interpersonal and textual positions arising from the writer’s experience in participating in genres and discourses (Ivanic & Camps, 2001; Prior, 2001). To be in consonance with this approach, the notion of context used might be characterized in terms of the different conceptualizations proposed by researchers working within the interactional view of writing (Chin, 1994; Nystrand, 1989; Thompson, 2001). We hope that these suggestions will help broaden our understanding of the way L2 writers manage composing processes as a function of the representations and approaches to the task they have developed in particular social environments.

ACKNOWLEDGEMENTS Some parts of the present chapter are based on the doctoral dissertation of the first author (Roca de Larios, 1999). He is very grateful to Rosa Manchón for her invalu- able guidance in this research. The authors also thank the editors and four anonymous reviewers for their useful and valuable comments.

APPENDIX The following studies have been reviewed (in chronological order): Edelsky (1982); Lay (1982); Zamel (1983); Brooks (1985); Raimes (1985); Gaskill (1986); Arndt (1987); Hall (1987); Raimes (1987); Jones & Tetroe (1987); Fagan & Hayden (1988); Lay (1988); Skibniewski (1988); Uzawa & Cumming. (1989); Moragne e Silva (1988); Cumming (1989); Cumming, Rebuffot, & Ledwell (1989); Carson, Carrell, Silberstein, Kroll, & Kuehn (1990); Friedlander (1990); Hall (1990); Hall (1991); Yau (1991); Kobayashi & Rinnert (1992); Carson & Kuehn, (1992); Silva (1992); Valdés, Haro, & Echevarriarza (1992); Pennington & So (1993); Tarone, Downing, Cohen, Gillette, Murie, & Dailey (1993); Devine, Railey, & Boshoff, (1993); Bosher (1998); Akyel (1994); Smith (1994); Berman (1994); Bell (1995); Swain & Lapkin (1995); Whalen, & Ménard (1995); Victori (1995); Porte (1996); Sasaki & Hirose (1996); Uzawa (1996); Henry (1996); Ting (1996); Porte (1997); Kasper, (1997); Shaw & Ting Kun Liu (1998); Kobayashi & Rinnert (1998); Bosher (1998); Ransdell & Levy (1998); Ransdell, Arecco & Levy, 2001); Qi, (1998); Roca de Larios, Murphy & Manchón (1999); Manchón, Roca de Larios, & Murphy (2000); Zimmermann (2000); Sasaki (2000); Khaldieh (2000); Miller (2000); Thorson (2000); Way, Joiner & Seaman (2000); Chenoweth & Hayes (2001); Cohen & Brooks-Carson (2001); Ransdell, Arecco, & Levy (2001), Roca de Larios, Marín, & Murphy (2001); Sasaki (this volume); Faraco, Barbier & Piolat (this volume).

BUILDING AN EMPIRICALLY-BASED MODEL OF EFL LEARNERS’ WRITING PROCESSES

MIYUKI SASAKI

Nagoya Gakuin University, Japan

Abstract. This chapter investigated Japanese learners’ processes of English expository writing using multiple data sources including their written texts, videotaped writing behaviors, and stimulated recall protocols. Two groups of Japanese EFL writers (12 experts and 22 novices) were compared both cross- sectionally and longitudinally. The study tested the following eight hypotheses formulated as a result of a pilot study (Sasaki, 2000): (1) EFL writing experts write longer texts at greater speed than EFL writing novices; (2) after two semesters of process writing instruction, neither the quantity nor the speed of the novices’ writing improves; (3) the experts spend a longer time before starting to write than the novices; (4) after the instruction, the novices spend a longer time before starting to write; (5) while writing, the experts stop to reread or refine their expressions more often than the novices, whereas the novices stop to make local plans or translate their ideas into L2 more often than the experts; (6) after the instruction, the novices stop to reread more often while making fewer local plans; however, they still have to stop to translate as often as before; (7) the experts tend to plan a detailed overall organization, whereas the novices tend to make a less detailed plan; (8) after the instruction, the novices learn to do global planning, but it is qualitatively different from the experts’ global planning. The obtained results are presented as flow- chart diagrams that represent the writing processes of the different groups of EFL learners.

Keywords: Japanese EFL learners, empirical model of writing processes, expert writers, fluency, global planning, local planning, novice writers, pausing behaviors, plan monitoring, stimulated recall protocols, writing expertise, writing strategies.

1 INTRODUCTION Composing process has been a major focus of L2 writing research for the past several decades (Cumming, 1998; Krapels, 1990; Silva, 1993). Basically following the designs of first language (L1) composition studies, researchers have investigated various aspects of L2 writing processes for different groups of participants. Of particular interest to the present study were those studies that examined writers’ micro- level cognitive processes while writing. Zamel (1983), in one of the earliest studies,

M. Sasaki (2002). Building an Empirically-Based Model of EFL Learners’ Writing Processes. In G. Rijlaarsdam (Series ed.) & S. Ransdell & M. Barbier (Volume eds.), Studies in Writing, Volume 11: New Directions for Research in L2 Writing, 49 – 80. © 2002 Kluwer Academic Publishers. Printed in the Netherlands. 50 SASAKI analyzed detailed observational data collected while six ‘advanced’ ESL (English as a second language, i.e., learning English in an English-speaking environment) students were completing ‘a course-related writing task’ (Zamel, 1983: 169). Among these six students, four were identified as ‘skilled’ and two as ‘unskilled’ based on experienced readers’ ‘holistic assessments’ (p. 172) of their writings, and these students spent four to eighteen hours writing several drafts. Zamel did not use the participants’ ‘think-aloud’ data while writing in spite of the fact that they were ‘used in most process studies’ (Zamel, 1983: 169), because ‘there is some doubt about the extent to which verbalizing aloud one’s thoughts while writing simulates the real composing situation’ (Zamel, 1983: 169). In contrast, Raimes (1985), adopting the methods employed in L1 process writing studies (e.g., Hayes & Flower, 1983; Swarts, Flower, & Hayes, 1984), examined concurrent think-aloud data collected from eight unskilled (determined by holistic measures of their essays) ESL students who wrote a narrative within a 65-minute class period. A much larger scale study was Cumming’s (1989) investigation of 23 French-speaking college students’ Eng- lish writing processes using their written texts and think-aloud data. The study was notable because (1) it applied multivariate statistical analyses, which was made possible by the relatively large sample size, (2) it compared students’ writing processes for three different tasks (letter writing, summary, and argumentation, one to three hours each), and (3) it introduced controlled variables of L1 writing expertise and L2 writing proficiency. Using the participants’ decision statements in the think-aloud data, Cumming focused on four aspects of writing the students attended to while writing (language use, discourse organization, gist, and procedure for writing) and five categories of problem-solving behaviors (heuristic searches with and without resolution, problem resolution, problem identification, and knowledge telling). More recently, Bosher (1998), using a modified version of Cumming’s (1989) coding systems, examined the L2 writing processes of three Southeast Asian ESL college students with different educational backgrounds. Bosher’s study was unique in that she used, as alternative data to think-aloud protocols, stimulated retrospective protocols collected from the participants who recalled their writing processes while watching their own videotaped writing behaviors. Finally, Roca de Larios, Murphy, and Man- chon (1999) again using the participants’ think-aloud protocol data while writing, conducted two studies investigating the ‘restructuring’ process where the writers seek ‘an alternative syntactic plan once the writer predicts, anticipates, or realizes that the original plan is not going to be satisfactory for a variety of linguistic, ideational or textual reasons’ (Roca de Larios et al., 1999: 16). Unlike the other studies described above, Roca de Larios et al.’s study concentrated on the particular writing strategy of restructuring that had ‘received very little attention in research on composing’ (Roca de Larios et al., 1999: 16). These previous studies that examined part of or the entire process of L2 writing commonly found that (1) skilled L2 writers were similar to their L1 counterparts in that they tended to plan more, revise more at the discourse level, and spend more time exploring the most appropriate ways to solve the given task (e.g., Cumming, 1989; Raimes, 1987; Roca de Larios, Murphy, & Manchon, 1999; Zamel, 1982, 1983); (2) unskilled L2 writers were similar to their L1 counterparts in that they tended to plan less and revise more at the word and phrase level (e.g., Raimes, 1985, EFL LEARNERS’ WRITING PROCESSES 51

1987; Roca de Larios, Murphy, & Manchon, 1999; Zamel, 1983), but they were different from their L1 counterparts in that they were relatively less concerned about surface level revisions (e.g., Raimes, 1985, 1987) and in that they showed more commitment to the given assignment (e.g., Raimes, 1985, 1987); (3) there appears to be a ‘writing expertise’ which is independent of L2 proficiency, affecting L2 writing (e.g., Bosher, 1998; Cumming, 1989; Raimes, 1985, 1987); (4) students’ attention patterns and problem-solving behaviors while writing differed according to their L1 writing expertise and the type of tasks they were involved in (e.g., Cumming, 1989). Because researchers have realized that L2 writers’ strategies are similar to those used for L1 writing, many studies from the late 1980’s on have also compared the same participants’ L1 and L2 writing processes. It should be noted that here again think-aloud protocol data were the main sources of analysis for most studies. In these studies, the participants’ L2 is mostly English (but see Cumming, Rebuffot, & Ledwell, 1989 and Whalen & Ménard, 1995 for exceptions), but their L1s greatly vary. For example, Jones and Tetroe (1987) compared six college-level Spanish- speaking ESL students’ planning behaviors while these students wrote two English and one Spanish descriptive expositions. Arndt (1987) compared six Chinese postgraduate EFL (English as a foreign language, i.e., learning English in a non English- speaking environment) students’ processes of writing expositions in L1 and L2 (completed within one hour each). Similarly, Skibniewski (1988) compared three college-level Polish EFL students’ processes of writing expository essays in L1 and L2. In contrast to Jones and Tetroe’s or Arndt’s study, Skibniewski could compare the differential effects of writing expertise on the three students’ L1 and L2 writing processes because they had distinctly different writing skills both in L1 and L2 (i.e., skilled, average, and unskilled). Similarly, Cumming, Rebuffot, and Ledwell (1989) compared the summary writing processes in English and French of 14 English- speaking college students with different writing expertise. Using Cumming’s (1989) coding scheme, Cumming et al. specifically focused on the participants’ problem- solving behaviors. Finally, two more recent and larger scale studies were completed by Whalen and Ménard (1995) and Uzawa (1996). Whalen and Ménard analyzed 12 English speaking participants’ planning, evaluation, and revision strategies at three different levels of discourse (pragmatic, textual, and linguistic) while writing argumentative texts in their L1 and L2 (French) within a maximum of two hours for each. On the other hand, Uzawa compared 22 Japanese ESL students’ processes of writing first drafts of descriptive expositions in Japanese (30 minutes) and in English (one hour), as well as their processes of translating a magazine article from Japanese into English (one hour). In addition to comparing overall characteristics of each writing process, Uzawa compared attention patterns employed for the three types of writing. Admitting the noticeable individual differences among the participants reported by some of these comparative studies (e.g., Arndt, 1987), we can also conclude that they have commonly found that (1) L1 and L2 writing strategies, whether the writers were skilled or unskilled, were basically similar, which indicates that L1 writing strategies can be transferred to L2 writing (e.g., Arndt, 1987; Cumming, Rebuffot, & Ledwell, 1989; Jones & Tetroe, 1987; Moragne e Silva, 1988; Skibniewski, 1988; Uzawa, 1996; Whalen & Ménard, 1995); (2) compared with their L1 writing proc- 52 SASAKI esses, students’ L2 writing processes, especially the higher-order cognitive operations, were negatively affected by their limited L2 proficiency (e.g., Moragne e Silva, 1988; Whalen & Ménard, 1995); and (3) the quality of written L2 texts is more strongly associated with the quality of the students’ L1/L2 writing strategies rather than with their L2 proficiency (e.g., Cumming, Rebuffot, & Ledwell, 1989; Jones & Tetroe, 1987). Although these previous studies provided insight into L2 learners’ writing processes, their designs were not without limitations. First, they investigated mainly ESL learners whose educational backgrounds were typically heterogeneous, and whose L2 proficiency was high enough so that they could receive their education in their L2. Even when EFL learners were examined, their L2 proficiency tended to be high (e.g., Arndt, 1987; Skibniewski, 1988). Second, even though some studies included ‘skilled’ versus ‘unskilled’ contrasts (mainly among student writers), virtually no studies have included a ‘novice’ versus ‘expert’ contrast where ‘experts’ were those who used L2 writing for professional purposes.1 Furthermore, many previous studies have employed cross-sectional designs only, and thus lacked developmental perspectives. Including multiple perspectives where novice writers are compared with expert writers as representatives of their ultimate possible goals of achievement, or where the novice writers are compared before and after a certain period of writing instruction with other intervening variables controlled, is crucial for building a more comprehensive and dynamic model of L2 writing processes. Another limitation of the previous studies of L2 writing processes is their almost exclusive use of think-aloud protocols as the main data source (but see the above description of Zamel, 1983 and Bosher, 1998 as exceptions). Although collecting concurrent verbal reports is an effective way to obtain real-time data on the participants’ writing processes (Ericsson & Simon, 1993), it entails various inherent problems (Smagorinsky, 1994), some of which are especially relevant to the present study. First, it is very difficult for some potential participants to produce ‘think- aloud’ data while writing in L2. It appears even more difficult when they are asked to speak in their L2 (e.g., Raimes, 1985, 1987) because many L2 writers often think in their L1 while writing (e.g., Cumming, 1989; Cumming, Rebuffot, & Ledwell, 1989; Uzawa, 1996). Moreover, even when participants were allowed to speak in any language they wished, some expressed difficulty with the task. For example, Whalen and Ménard (1995), who seem to have allowed the participants to choose the language they spoke in, admitted that ten potential participants (compared to the 12 who actually produced the data for the study) could not perform this difficult task, and thus were excluded from the study. Finally, even if researchers can manage to obtain analyzable data from participants (see Hayes & Flower, 1980, characteriz- ing the nature of analyzing protocol data as ‘following the tracks of a porpoise,’ p.

1 Although several studies such as Cumming (1989) included writers with professional experience, they were experts in L1 writing rather than L2 writing. I believe that research into the differences between the writing processes of novice and expert L2 writers is necessary to build a comprehensive model of L2 writing processes because experts’ writing ability represent an ultimate goal (and also an ultimate achievement limit) that any L2 learners with similar backgrounds can accomplish (Grabe & Kaplan, 1996). EFL LEARNERS’ WRITING PROCESSES 53

9), there is always the danger of ‘reactivity’. Previous empirical studies (e.g., Janssen, van Waes, & van den Bergh, 1996; Stratman & Hamp-Lyons, 1994) have reported that the think-aloud condition appeared to have significantly affected the quality and content of the participants’ cognitive activities while writing. With these methodological limitations in mind, I conducted a precursor of the present study as a pilot study (Sasaki, 2000). It investigated the writing processes of three types of L2 writers (professional, and more- and less-skilled) with similar cultural and educational backgrounds, both cross-sectionally and longitudinally (i.e., developmentally),2 using multiple data sources collected through a less disruptive method than the think-aloud technique.3 The method was similar to the one used in Bosher (1998) in that the participants produced recall protocols while watching their video-taped writing behaviors, but it was different from Bosher’s method in that the participants could choose the language(s) in which they produced the protocols, and in that the data were coded by a coding scheme specifically developed for this type of data (Anzai & Uchida, 1981). The pilot study was also motivated by the results of two preceding product- oriented studies (Hirose & Sasaki, 2000; Sasaki & Hirose, 1996). Sasaki and Hirose cross-sectionally investigated factors that could explain Japanese EFL students’ English writing ability. We found that the participants’ L2 proficiency, L1 writing ability, and metaknowledge of L2 expository writing (e.g., how to achieve unity and coherence in a paragraph) significantly explained the students’ L2 writing ability variance (52% of the variance was accounted for by L2 proficiency, 18% by L1 writing ability, and 11% by metaknowledge). We also found that good writers were significantly different from weak writers in terms of their attention to overall organization while writing in L1 and L2, their writing fluency in L1 and L2, their confidence in L2 writing for academic purposes, and their experiences of regularly writing more than one paragraph in L2 in high school. Based on these results, Hirose and Sasaki further examined the teachability of two of these explanatory factors, metaknowledge of L2 writing and regular L2 writing experience. The results indicated that teaching the metaknowledge to the students over 12 weeks significantly improved their metaknowledge, but not their L2 writing ability in general. In contrast, the instruction of metaknowledge combined with regular journal writing sig-

2 In the present study, I used the term ‘longitudinal’ as synonymous with ‘developmental.’ When I classified studies, I followed Isaac and Michael’s (1981:42) definition of ‘developmental’: To investigate patterns and sequences of growth and/or change as a function of time. In the pilot study, Sasaki (2000), I investigated the changes in eight student writers’ writing process during six months of process writing instruction, and thus I called the study ‘longitudinal.’ 3 Although the stimulated recall protocol method employed both for Sasaki (2000) and the present study is obviously less disruptive than the think-aloud method (all the participants in both studies could successfully complete the task), it might arguably have entailed some reactivity problems such as the possibility that the participants had been affected by the existence of the video-camera(s). Moreover, it is also true that the recall protocol method can only induce what the participants can recall, or what they think they were thinking about at the point of time in question. Unlike the think-aloud data, what the participants recall may not be a faithful reproduction of what they were thinking about at that particular moment. 54 SASAKI nificantly improved mechanical aspects (e.g., spelling, punctuation, capitalization), but not the overall quality of their L2 writing. When I conducted the pilot study (Sasaki, 2000), very few studies had investigated Japanese EFL learners’ L2 writing processes before, and thus, the study inevitably became exploratory in nature: I tried to select as the targets of analysis as many aspects as possible of the participants’ writing behavior (e.g., time spent before starting to write, pausing behaviors, writing strategies) that seemed to be important for building an empirical model of their writing processes. At that point, which aspects of those writing behavior would characterize the writing processes of the three different groups of participants was not yet clear. Furthermore, because the pilot study used relatively small samples (four experts, four more-skilled students, and four less-skilled students), the significance of the findings could not be tested by statistical procedures. In order to claim generalizability of the results obtained in the pilot study, hypotheses had to be formed based on the results and then tested with samples that would be large enough for statistical procedures to be legitimately applied (the relationship between the pilot study and present study is in a sense similar to that between exploratory and confirmatory factor analyses, where confirmatory factor analysis tests the relationships among variables that were constructed as a result of exploratory factor analysis; see Bollen, 1989). The present study thus replicates the general design of the pilot study by investigating the writing processes of Japanese EFL learners both cross-sectionally and longitudinally. The data analyzed in the present study included the participants’ written texts, their pausing behaviors while writing, stimulated recall protocols, and analytic scores given to the written texts. The present study tests the following eight hypotheses formulated as a result of the pilot study. These hypotheses reflect the differences found among the participant groups in the pilot study:4 1) EFL writing experts write longer texts at greater speed than EFL writing novices. 2) After two semesters of process writing instruction, neither the quantity nor the speed of the novices’ writing improves. 3) The experts spend a longer time before starting to write than the novices. 4) After the instruction, the novices spend a longer time before starting to write. 5) While writing, the experts stop to reread or refine their expressions more often than the novices, whereas the novices stop to make local plans or translate from L1 to L2 more often than the experts (see Appendix for the definitions of Re- reading, Rhetorical Refining, Local Planning, and Translating from L1 to L2). 6) After the instruction, the novices stop to reread more often while making fewer local plans. However, they still have to stop to translate from L1 to L2 as often as before. 7) The experts tend to plan a detailed overall organization (i.e., Global Planning), whereas the novices tend to make a less detailed plan (i.e., Thematic Planning).

4 Unlike the pilot study, the novices were not divided into the more- and less-skilled writers in the present study in order to make the research design less complex for the application of statistical procedures. EFL LEARNERS’ WRITING PROCESSES 55

8) After two semesters of process writing instruction, the novices learn to do global planning, but it is qualitatively different from the experts’ global planning. Hypotheses 1, 3, 5, and 7 were concerned with the cross-sectional differences between EFL writing experts’ writing processes and those of EFL writing novices whereas Hypotheses 2, 4, 6, and 8 were concerned with the effects of a process- writing instruction on the novices’ writing processes. Hypotheses 1 and 2 were related to the variable of ‘writing fluency’. This variable had been chosen as a target of investigation in the pilot study (Sasaki, 2000) because the good and weak writers in Sasaki and Hirose (1996), one of the product-oriented studies that motivated the pilot study, were significantly different in terms of writing fluency. Hypotheses 3 and 4 were related to writers’ planning behavior, which has been investigated in many previous studies of both L1 and L2 writing processes (e.g., Cumming, 1989; Raimes, 1987; Roca de Larios, Murphy, & Manchon, 1999; Zamel, 1982, 1983). The four strategies referred to in Hypotheses 5 and 6 (Rereading, Rhetorical Refin- ing, Local Planning and Translating from L1 to L2) were the ones the experts used differently from the novices among the 21 strategies investigated in the pilot study. Finally, Hypotheses 7 and 8 were concerned with the issue of writing expertise whose existence has been hypothesized in some previous studies (e.g., Bosher, 1998; Cumming, 1989; Raimes, 1985, 1987). In the pilot study, I speculated that particular characteristics of the experts’ global planning were similar to those of what Flower, Schriver, Carey, Haas, and Hayes (1992) called ‘constructive planning,’ which was exclusively used by L1 writing experts after long years of training and experiences.

2 METHOD 2.1 Participants Two groups of Japanese EFL learners (a total of 34) participated in the present study: an expert writer group (n = 12) and a novice writer group (n = 22). The experts (10 men and 2 women) were operationally defined as Japanese native speakers who had learned English mainly through formal education in Japan until they gradu- ated from universities, and who now regularly write academic papers in English at least once a year. They had studied English (both inside and outside the classroom) for an average of 23.5 years (SD = 6.0 years). Although I didn’t specifically measure their English proficiency for the present study because of time constraints,5 I assumed that their English proficiency was high (six of them reported that their best TOEFL [Test of English as a Foreign Language produced by Educational Testing Service, Princeton, NJ] scores were over 600). They were four applied linguists, two linguists, one communication researcher, three economists, one engineer, and one chemist, with a mean age of 36.8 years. Although they had spent an average of 4 years in English-speaking countries (seven of them had M.A.s, and three of them

5 The time taken for the composition writing session and stimulated recall (2.5 to 3 hours) was the maximum I could ask the busy experts to spare for the present study. 56 SASAKI had Ph.D.s from universities in English speaking countries), most of their current life is anchored in Japan, a non-English speaking environment. The novices were all 18 year-old college freshmen (4 men and 18 women), ma- joring in British and American studies.6 Candidates for those novices were randomly selected from a sample of 75 students (three classes combined), and asked to participate in the present study. Those who agreed participated in the present study. They were paid for their work. They had studied English for six years through formal education in Japan. None of them had been abroad longer than one month. They were judged to be ‘novice L2 writers’ because the results of a background questionnaire (see Appendix A of Sasaki & Hirose, 1996) indicated that they had received little L2 writing instruction, including instruction on matters such as ‘organizing a paragraph centered on one main idea’ or ‘developing a paragraph so that the readers can follow it easily.’ The mean total score on the Secondary Level English Profi- ciency (SLEP) Test (Educational Testing Service, 1993) was 40.68 (SD = 4.98), which indicates that these students’ English proficiency ranged from low- to mid- intermediate. The writing processes of the 22 novice writers were also compared before and after two semesters (a total of six months interrupted midway by a two-month sum- mer vacation) of process writing instruction. Although the instruction was basically intended to employ a ‘process approach’ (Silva, 1990: 15), it also incorporated some ‘current-traditional’ (Silva, 1990: 13) aspects (i.e., teaching students representative prescriptive writing patterns) because metaknowledge (e.g., the meaning of topic sentence, unity, coherence, and the organization of English expository writing) that could be acquired through such a current-traditional approach had been found to be significant in explaining L2 writing quality of a similar group of students (Sasaki & Hirose, 1996). The 22 novice students in the present study received instruction on paragraph writing in English with 53 other students once a week for 90 minutes. These students did not take any other English writing classes during these two semesters. The instructor (the researcher) taught them process writing strategies such as planning and revising, based on Bereiter and Scardamalia’s (1987) ideas of ‘Promot- ing the development of mature composing strategies’ (Bereiter & Scardamalia: 245), using Hashiuchi’s (1995) Paragraph Raitingu Nyuumon [Introduction to Paragraph Writing], a composition textbook with special emphasis on process writing. Thus, in the first class, the students were told that writing is an interactive process between what they write and what they want to write, and that such a process is cyclical, starting with planning and followed by writing and revising. Furthermore, in each chapter of the textbook (the class covered nine chapters in total), the students first learned rhetorical patterns such as comparison, classification, and expressing opinions, and then were instructed to write a similar paragraph themselves. Before they started to write, they discussed the following points with their instructor and peers:

6 The novices’ mean age was significantly different from that of the experts’ (t = 11.26, df = 32, p = .000). Therefore, it is possible that these two groups may differ not only in writing expertise but also in cognitive abilities. The relationship among age, cognitive abilities, and writing expertise, however, remains to be studied in future studies. EFL LEARNERS’ WRITING PROCESSES 57

1) What is the purpose of this writing? 2) Who are the readers of this writing? 3) What content should I include to make it effective? 4) How should I express the content to make it effective?

After they wrote the first drafts, most drafts were read by their peers or the instructor. When time allowed, these readers gave written comments on which points of the writing were interesting or appealing and which were difficult to understand. After receiving these comments, the students were asked to revise their writing with the ideas of unity and cohesion in mind. A total of nine texts were written during the class, but none of them were graded (the students were graded on the basis of their participation in the class and the scores of the mid-term and final examinations, which included writing compositions).

2.2 Data collection procedure For the present study I followed a slightly revised version of the technique employed in the pilot study (Sasaki, 2000) as explained below. The technique was originally developed by Anzai & Uchida (1981) for investigating Japanese children’s L1 writing process as an alternative data collection method to the concurrent think-aloud procedure,7 but I found it can also be successfully used for investigating Japanese EFL students’ writing processes. The L2 experts and the 22 novice students wrote an argumentative composition according to Prompt 1: There has been a heated discussion about the issue of school uniforms in the readers’ column in an English newspaper. Some people think that wearing high school uniforms is a good custom, whereas others believe high school students should be given an opportunity to choose what they wear. Now the editor of the newspaper is calling for readers’ opinions. Suppose you are writing for the readers’ opinion column. Take one of the positions described above, and write your opinion within 30 minutes. (Original in Japa- nese, translated by the author) The 22 novice writers then wrote according to Prompt 2 after the two-semester instruction period ended: There has been a heated discussion about the custom of non-Christian Japanese cele- brating Christmas in the readers’ column in an English newspaper. Some people think that it is a good custom, whereas others believe we should abandon such a custom. Sup- pose you are writing for the readers’ opinion column. Take one of the positions described above, and write your opinion within 30 minutes. (Original in Japanese, translated by the author)

7 Researchers in the field of L1 Japanese writing have also found that it was difficult to collect think-aloud data. For example, Uchida (1989) reported that only two out of ten potential participants produced analyzable think-aloud data in a study investigating children’s revising processes in L1 Japanese (see also Uchida, 1986 for discussion of this issue). 58 SASAKI

One week after the novices wrote for Prompt 2, I had to ask two of them to write for Prompt 3 because the video camera broke when they were writing for Prompt 2. There has been a heated discussion in an English newspaper about introducing English as an elementary school subject in Japan. Some people think that English should be taught at the elementary school level, whereas others believe that it is too early. Take one of the positions described above, and write your opinion within 30 minutes. (Origi- nal in Japanese, translated by the author) For these two students, the data for Prompt 3 were used for subsequent analyses. Prompts 1 to 3 were similar in the sense that they dealt with controversial issues that the students were familiar with. Prompts 1 and 2 were also used in the pilot study (see Sasaki, 2000 for the reasons why I selected these argumentative topics).8

All compositions were scored by two EFL writing specialists independent of the present study, according to Jacobs et al.’s (1981) English Composition Profile. In order not to make the raters aware of which topic was written first, all compositions were rated on one occasion after having been completely mingled with no indication of when they were written, or which participant groups they belonged to. The interrater correlation (Pearson correlation coefficient) for the content subscore was .93;

8 If these two prompts had been alternated with half of the eight students before and after the instruction (i.e., half of them receiving Prompt 1 before the instruction and Prompt 2 after the instruction with the other half receiving Prompt 2 before the instruction and Prompt 1 after the instruction), I could have avoided possible topic effects on the students’ composition scores and their use of writing strategies. However, if I had alternated the prompts for the novice I group, I would also have had to alternate the prompts for the expert group for a fairer comparison. In such a case, I would have had to consider the possible effects of these two different topics on the participants’ use of writing strategies, especially when the experts’ writing strategies were compared with those of the novices for one composition written on only one occasion. Previous studies (e.g., Carter, 1990; Cumming, 1989; Flower, Schriver, Carey, Haas, & Hayes, 1992) suggest that writers may change their writing strategy use according to different topics. Thus, I decided not to alternate the prompts for the novices. On the other hand, I used similar but different prompts for novices I and II because I was afraid that maturation effects caused by giving the same prompt before and after the instruction (e.g., the novices might have thought about the topic over the two semesters, and thus produced much better quality compositions than they might have if they had not written for the same prompt before the instruction, or they would not plan for the second time because they had already thought about the issue for the first time) might be stronger than possible topic effects (especially when Prompts 1 to 3 were intended to induce very similar argumentative writing, see Sasaki, 2000; Sasaki & Hirose, 1996). I had to make this compromise that might have introduced topic effects for the novices I and II comparison because I wanted to compare the experts and novices on the most equal basis possible while still comparing the same novices before and after the instruction. In other words, I wanted to incorporate into a single study both a cross-sectional design and a longitudinal design using the same participants. This was a real dilemma. But I concluded that topic effects, if they existed at all, would be larger for the expert-novice (inter-group) comparison rather than for the novices I and II (intra-group) comparison. Of course, however, I was aware of the fact that comparing novices I and II could be problematic because of possible topic effects (see the Results and Dis- cussion Section). EFL LEARNERS’ WRITING PROCESSES 59 the organization subscore, .86; the vocabulary subscore, .93; the language use subscore, .93; the mechanics subscore, .59; and the total score, .97. I judged that these correlations were acceptable for this sample population size. To collect the data, I asked the participants to come individually to a room and write the compositions in a quiet atmosphere. I asked them to finish writing within about 30 minutes but told them they could take more time if they wanted to. Conse- quently, some of them took longer than 30 minutes to finish, but all of them finished writing within one hour. The relatively short time allocation was chosen for the purpose of making the following video-watching session manageably short (approximately two hours). As in the pilot study, the participants were not allowed to use dictionaries. Before the participants started to write, I obtained permission to videotape them while they were writing. I then began to videotape them with one camera focused on their hand/pencil movement, and another focused on their entire writing behavior including their eye/head movements. Unlike the pilot study, I used two cameras to obtain clearer and more holistic views of which part of the text the participants were working on and how they were writing. As in Anzai and Uchida (1981) and in the pilot study, I asked the participants how much planning and what kind of plan they had made before starting to write just after they started to write the first word, when their memory was still fresh. The participants were not told that they would be questioned right after they started writing. When they answered the questions, they were told that their answering time would not be included in the 30 minutes allocated for writing. After the first question session, I let them continue writing without interruption until they finished. Immediately after they finished writing, each participant and I together watched the participant’s writing process on videotape on two monitors, which simultaneously were showing the participant’s hand/pencil movement and physical writing behavior. Every time the participants stopped writing for a period longer than three seconds on the videotape,9 I asked them to explain, either in Japanese or English, what they had been thinking about. No leading question (e.g., ‘Did you think such- and-such?’) were used to avoid biasing the students’ answers (Ericsson & Simon, 1993). This continued until they had finished the entire process of writing on the tape. All participants gave their accounts in Japanese except for some English expressions that they used or considered for use in the compositions (see the examples below). The videotaped behaviors projected through two monitors were clear enough to aid the participants’ recall. The participants’ accounts were all tape- recorded and subsequently transcribed. From these writing and recall sessions, I obtained and analyzed three types of data: (1) written texts and drafts, (2) time the participants spent before starting to write and time they spent writing the whole composition, and (3) participants’ retrospective accounts, while watching their video-taped writing performance, of what they were thinking about when they stopped writing.

9 Anzai and Uchida (1981) used pauses longer than two seconds instead of three. Longer pauses were used in the present study because it was concluded that three seconds were the shortest possible pauses that could be handled based on the pilot study results. 60 SASAKI

2.3 Coding of the protocol data In the pilot study, I used a revised version of Anzai & Uchida’s (1981) coding system. Their coding system was developed in a careful and well-designed empirical study, and successfully used to investigate the participants’ Japanese L1 writing strategies. Following Anzai and Uchida, I defined the term ‘strategy’ in the present study as a writer’s mental behavior employed to achieve a certain goal in the ‘ill- structured problem-solving’ (Anzai & Uchida: 46) activity of writing. Based on An- zai and Uchida’s coding system I encoded the pilot study participants’ retrospective protocol data into 21 strategies (Global Planning, Thematic Planning, Local Plan- ning, Organizing, Conclusion Planning, Plan Retrieving, Information Retrieving, Generating Naturally Generated Ideas, Generating Description-Generated Ideas, Verbalizing a Proposition, Rhetorical Refining, Mechanical Refining, Sense of Readers, Translating L1 to L2, Rereading, L2 Proficiency Evaluation, Local Text Evaluation, General Text Evaluation, Resting, Questioning, and Impossible to cate- gorize; see Appendix C of Sasaki, 2000 for more details). For the present study, I also encoded the data using the same system, but mainly focused on the six strategies of Global Planning, Local Planning, Thematic Planning, Rereading, Rhetorical Refining, and Translating from L1 to L2, for the purpose of testing Hypotheses 5 to 8, which emerged from the pilot study. As exemplified in the Appendix, the participants’ protocol accounts were encoded according to the definition of each writing strategy. One chunk of those accounts that consisted of several propositions was judged to be one category if it as a whole matched the definition of a single category.10 Before I coded the data, a subset of the data randomly selected from six out of the 56 protocols (11% of the total sample population) was coded by another trained applied linguist using the same categories. Cohen’s Kappa value, which excluded the chance value from the two coders’ agreement rate, was calculated for the 348 decisions on strategy classifications (Takeuchi, 1989). The Kappa value was 0.84 with a 95% confidence interval of .79 to .88. Having judged that this agreement rate was acceptable, I coded the rest of the data myself.

2.4 Determining the participants’ writing styles I also classified the participants’ writing styles according to a revised version of An- zai and Uchida’s (1981) writing style classification formulae based on the results of their analysis of 40 Japanese children’s use of L1 writing strategies described above. I revised the formulae so that they could properly identify the qualitative differences found in the pilot study between the experts and novices (this part of the analysis was related to Hypotheses 7 and 8). In the pilot study, the experts’ global planning appeared to include not only a detailed planning of overall organization, but also the

10 For example, in the Appendix, the example of Rhetorical Refining consisted of one phrase and four clauses, but it was treated as one example (or a chunk) of Rhetorical Refining in the analysis. In contrast, the example of Rereading consisted of two clauses, and it was also treated as one example of Rereading. EFL LEARNERS’ WRITING PROCESSES 61 behavior of seeking the most effective solution based on the assessment of the given working environment (i.e., writing an argumentative composition within 30 minutes to one hour) whereas the novices’ global planning included only planning of overall organization. Any previous classification of planning could not be used here because both the experts’ and the novices’ behaviors were partially similar to some of the planning categories proposed in previous studies, but not completely the same (e.g., the experts’ and the novices’ behaviors appeared to include both what Hayes & Nash, 1996: 44, called ‘non-content planning’ and ‘content planning’). The formulae revised for the present study determined each participant’s writing style according to his/her use of the writing strategies of Global Planning, Thematic Planning, and Lo- cal Planning as follows (see Appendix for definitions and examples of these strategies).

Type Description

A Write according to detailed global planning of both the content of the text and the most effective way to express it based on the assessment of the given task. B Write according to detailed global planning of the content of the text. C Write according to rough global planning (thematic planning) and consecutive local planning.

Type A is similar to Bereiter and Scardamalia’s (1987) ‘knowledge transforming model’ (Bereiter & Scardamalia, 1987: 10) in that the writers try to express their ideas in the best possible way according to their goal setting, whereas Type C is partially similar to Bereiter and Scardamalia’s ‘knowledge telling model’ (Bereiter & Scardamalia: 10) in that they make many local plans, but different from the model in that the writer makes a global plan, although it is not very detailed. Classifying the participants’ behaviors according to these formulae revealed the qualitative differences among the different groups’ writing patterns that might have been missed if these groups were compared only in terms of quantifiable variables (e.g., time spent before starting to write; frequencies of use of the writing strategies)

3 RESULTS AND DISCUSSION I used SPSS Version 6.1 (SPSS Incorporated, 1994) to perform all statistical analyses. An alpha level of .05 was used for the statistical tests unless it was necessary to adjust it for multiple comparisons.

3.1 Composition scores Although the participants’ composition scores were not directly related to the hypotheses to be tested in the present study, I will present them as sources of information that will be helpful for understanding the results related to the hypotheses. Table 1 shows the means and standard deviations for the composition scores (according to Jacobs et al.’s 1981, ESL Composition Profile) for the experts and the novices I and 62 SASAKI

II (before and after the two semesters of instruction). As I stated in Note 8, there might have been differential topic effects on the quality of novices I and II’s compositions because they wrote for different prompts. Thus, correlations between the quality of novices’ I and II’s compositions were investigated as a possible manifestation of topic effects. The correlations were .53* for the content subscore, .26 for the organization subscore, .53* for the vocabulary subscore, .44* for the language use subscore, .13 for the mechanics subscore, and .51* for the total score (*p<.05). These correlation values mean that those novices who scored high for Prompt 1 tended to score high for Prompt 2 after the instruction in terms of content, vocabulary, and language use in spite of the topic differences. It can be speculated that the organization and mechanics subscores did not correlate highly with each other because these aspects of compositions were deeply related to the content of the instruction (recall the Participants Section). These results suggest that the topic effects, if they existed at all, might not have been strong enough to affect the novice students’ writing ability in a drastically different manner. Nonetheless, to make the analysis on a fairer ground, the total and the content scores, which can be considered to have been the most susceptible to topic effects, were excluded from the comparison between these two groups (see also Hirose & Sasaki, 2000 for this issue). The experts’ total score and the five subscores were all significantly higher than those of novices I even after the alpha level was adjusted to .008 by a Bonferroni correction for multiple comparison (Tabachnick & Fidell, 1996). Similarly, novices I’s subscores for organization, vocabulary, language use, and mechanics were significantly higher than those of novices II after the alpha level was adjusted to .0125 by a Bonferroni correction. Although there might have been possible topic effects intervening in the novices I and II comparison, it appears that the novices’ composition scores did improve as a whole after the two semesters of process instruction.

Table 1: Effects of expertise (experts versus novices I) and writing instruction (novices I and II)a on writing scores.

Experts Novices I Novices II Expertise Instruction Variable n = 12 n = 22 n = 22 df = 32 df = 21

Max M sd M sd M sd t t

Total 200 187.8 7.1 128.0 19.2 148 14.7 13.1 -5.3 Content 60 56.9 1.9 40.3 5.8 46.0 4.2 12.4 -5.2 Organization 40 37.0 1.8 25.9 4.3 30.9 3.0 10.6 -3.8 Vocabulary 40 37.7 1.8 25.8 4.1 28.8 3.7 11.8 -4.7 Lang. Use 50 46.4 2.4 28.1 5.4 33.6 4.9 13.7 -4.6 Mechanics 10 9.8 0.4 7.9 0.7 8.7 0.6 8.9 -5.4

Underlined scores are statistically significant at p<.008 (adjusted by a Bonferroni correction for multiple comparison). a These scores are sums of the two raters’ scores. b Novices I and II are the same students before and after the two semesters of instruction. EFL LEARNERS’ WRITING PROCESSES 63

The mean scores of the experts and the novices were similar to those of the pilot study, which suggests that the sample populations used in these two studies were similar in terms of their writing ability. The only difference between the present study results and those of the pilot study was that unlike in the pilot study, there were significant differences in the organization, vocabulary, language use, and mechanics subscores between novices I and novices II. In the pilot study, these subscores also appeared to be different between novices I and II, but the differences were not statistically significant. Although the novice group might have improved their writing ability as a result of other factors (e.g., other English classes) because there was no control group with which to compare the true effect of the instruction, the results of the present study at least suggest that process writing instruction like that given in the present study has the potential to improve students’ L2 writing (see also previous studies where instruction improved at least some aspects of learners’ writing ability; e.g., Hirose & Sasaki, 2000; Shaw & Ting-Kun Liu, 1998).

3.2 Fluency The participants’ writing fluency was measured in terms of the two indices of mean total number of words written in the text (quantity) and mean number of words written per minute (speed; see Table 2). The t-test results indicate that the experts wrote significantly longer texts, and wrote significantly faster than the novices (t = 5.52, p =.000 for the quantity, t = 3.45, p = .002 for the speed; the alpha level was adjusted to .025 by a Bonferroni correction for multiple comparison). It is not always true that more skilled writers tend to write longer texts than their less skilled counterparts (see Reid, 1990; Silva, 1993), but the experts in the present study did write longer texts as well as faster than the novices. Thus, Hypothesis 1 presented in the Intro- duction Section was confirmed: The experts were significantly more fluent than the novices. As was noted in the pilot study, and as we will see in the Writing Strategies section below, the fluency difference between the experts and the novices appears to have been partially caused by the fact that the novices had to stop to translate their ideas into English more often than the experts, which was probably caused by the L2 proficiency difference between these two groups. In contrast with such a difference between the experts and the novices, the differences between novices I and II were not statistically significant (t = -.93, p =.37 for the quantity, t = 1.54, p =.14 for the speed by matched t-tests; the alpha level was adjusted to .025 by a Bonferroni correction). This might be because the prompts given to novices I and II were different, but considering the fact that the total scores given to their compositions were significantly correlated (r = .51; see Section 3.1), we can also suspect that the novices did not improve writing fluency as a result of the two semesters of process writing instruction. It appears that Hypothesis 2 concerning the novice writers’ improvement in terms of fluency was also confirmed. This result thus suggests that process writing instruction of two semesters may not help students improve their writing fluency. 64 SASAKI

Table 2: Effects of expertise (experts versus novices I) and writing instruction (novices I and II)a on fluency (number of words, and words per minute: wpm).

Experts Novices I Novices II Expertise Instruction Variable n = 12 n = 22 n = 22 df = 32 df = 21

M sd M sd M sd t t

Words Written 243.6 88.41 95.7 38.49 103.6 25.76 5.52 -.93 WPM 6.04 2.23 3.61 1.81 3.10 1.03 3.45 1.54

Underlined scores are statistically significant at p<.025 (adjusted by a Bonferroni correction for multiple comparison). a Novices I and II are the same students before and after the two semesters of instruction.

3.3 Time spent before starting to write Table 3 presents the mean ratios of the time the participants spent before starting to write to the total time they spent writing the whole composition. Because the total time spent by each participant varied, such a ratio rather than the time itself was used for comparison. For the first composition, the experts on average spent 23% of the total time for the initial planning, whereas the novices spent only 9%. After the instruction, the novices spent 20% of the total writing time for their initial planning. Because ratios of such time values tend to have skewed distributions, and because these values can acquire more normal distributions under logarithmic transformation (Iwahara, 1997), the normality of the distributions of these time ratios and those of their corresponding logarithms were compared by the Shapiro-Wilks normality test (SPSS Incorporated, 1993). Because the values after logarithmic transformation were judged to have more normal distributions,11 these values were used for subsequent t-tests. The t-test results show that there was a significant difference between the experts and novices I, and between novices I and II (see Table 3). Thus, the third hypothesis regarding the difference between the experts’ and novices’ planning time was confirmed. Although there might be topic effects involved again, the fourth hypothesis regarding the differences between novices I and II appears to have been also confirmed. The experts spent a longer time before starting to write than the novices, and the novices learned to spend a longer time before starting to write after the instruction. Recent research (e.g., Hayes & Nash, 1996) indicates that planning time may not be as strongly related to L1 writing ability as was once believed, but it appears that in this particular case of L2 writing, planning time is related to L2 writing expertise as many previous studies have reported (e.g., Bereiter

11 One of these time-ratio values before logarithmic transformation had a distribution that significantly deviated from a normal one, but after the transformation none of the values had distributions that were significantly different from a normal distribution. EFL LEARNERS’ WRITING PROCESSES 65

& Scardamalia, 1987; Krapels, 1990). Moreover, the present results suggest that instruction may also influence students’ planning time.

Table 3: Effects of expertise (experts versus novices I) and writing instruction (novices I and II)a on starting time (proportion).

Experts Novices I Novices II Expertise Instruction Variable n = 12 N = 22 n = 22 (df = 32) (df = 21

M sd M sd M sd t t

Starting time .23 .14 .09 .19 .20 .15 4.13 -3.07

Underlined scores are statistically significant at p<.01. a The t values were calculated after the ratios were transformed to their logarithms to normalize their distributions.

3.4 Writing strategies Table 4 shows the mean total numbers of the four strategies (token frequency) used by the participants in the present study (ranging from 63.5 to 87.8). These strategies are related to Hypotheses 5 and 6. The numbers presented in Table 4 were generally larger than those used by the participants in the pilot study (ranging from 32 to 36.25). This might be partly because the experts in the present study spent a longer time (M time = 41 minutes 10 seconds) completing the compositions than their counterparts in the pilot study (M time = 31 minutes 11 seconds). However, this explanation does not apply to the novices because the novices in the present study spent a shorter time (M time = 29 minutes 26 seconds) completing the compositions than their counterparts in the pilot study (M time = 35 minutes 22 seconds). Furthermore, unlike the findings in the pilot study, the experts in the present study appear to have used more strategies than the novices, whereas the novices appear to have used more strategies after the instruction than before (in the pilot study, the experts and novices used similar numbers of strategies whereas novices II used only half as many strategies as novices I). Because the total numbers of strategies used by each group thus varied, the mean ratios of individual strategy tokens (i.e., frequencies of the strategies used by the participants; if the same strategy was used three times, it was counted as three instead of one) used by the members of each group out of the total number of strategy tokens used by them were employed for comparison.12 As in the case of the time spent before starting to write, such ratio values tend to have skewed distributions, and these values can acquire more normal distributions by logarithmic transforma-

12 For example, Expert 2 reread 11 times while writing the composition. Because he used a total of 45 tokens of strategies for completing the composition, 11 out of 45, i.e., 0.24 was used as his value for the Rereading strategy. After the individual calculations were completed, those values were averaged for each group for each strategy. 66 SASAKI tion (Iwahara, 1997). Consequently, the normality of the distributions of these time ratios and those of their corresponding logarithms were compared by the Shapiro- Wilks distribution-normality test (SPSS Incorporated, 1993). Based on the judgment that the values after logarithmic transformation were more normally distributed, these values were used for subsequent t-tests.13 Because t-tests were applied to the same sample population four times, the alpha level was adjusted to .0125 by a Bon- ferroni correction for multiple comparisons. The t-test results show that, unlike the findings in the pilot study, there was no difference between the experts and novices I, or between novices I and II, in terms of Rereading. That is, no differences were found in the ratios of rereading among the three groups. Because the three groups used different total numbers of strategies, this does not mean that they used similar numbers of rereading strategies. Rather, it implies that all groups tended to pay a similar amount of attention to rereading. Whether ‘rereading’ is related to the quality of writing has remained an unsolved issue according to the results of previous L1/L2 writing studies. For example, Levy and Ransdell (1995) reported that L1 English writers of better compositions tended to allocate more time for ‘reviewing’ (Levy & Ransdell: 773) than the other writers. In contrast, Anzai and Uchida (1981) reported in their cross-sectional study of L1 Japanese writers that the participating elementary school children’s reading behavior increased in terms of frequency from the second grade to the fifth, but decreased in the sixth grade. In contrast, Breetvelt, Van den Bergh, and Rijlaarsdam (1996) introduced the idea of ‘time-dependent changes’ (Breetvelt et al., 1996: 19) into their investigation of cognitive activities such as rereading while writing in L1 Dutch. Breetvelt et al. reported that a positive correlation between rereading and text quality ‘increases during the first half hour, and slowly decreases afterwards’ (Breetvelt et al., 1996: 16). As for L2 writers, Zamel (1983) reported that ‘All of the students, their writing skill notwithstanding, reread’ (Zamel, 1983: 173). However, the more skilled writers tended to reread various aspects of the text ranging from the local text just written to entire paragraphs whereas the least skilled writer reread shorter ‘chunks of discourse’ (Zamel, 1983: 173). Furthermore, Cumming (1989) found that some of the less skilled L2 writers ‘made conscious choices not to monitor their writing, by resolving not to edit or proofread their written compositions’ (Cumming, 1989: 114), whereas some of the more skilled writers ‘reviewed their previous text, rereading it every few minutes’ (Cumming, 1989: 115) when they were concentrat- ing on how to express their ideas. It appears that both more skilled and less skilled writers tend to reread while writing but that they often reread different parts of the text for different purposes at different times. The three groups in the present study might have been differentiated if ‘what,’ ‘why,’ and ‘when’ they reread had been investigated. Such an investigation should be conducted in future studies.

13 The values for the experts’ Local Planning, novices I’s Rhetorical Refining, and novices II’s Local Planning and Rhetorical Refining had distributions that significantly deviated from that of a normal distribution. After the logarithmic transformation, none of the values had such abnormal distributions. EFL LEARNERS’ WRITING PROCESSES 67

Table 4: Effects of expertise (experts versus novices I) and writing instruction (novices I and II)a on writing strategies: Mean ratios of the four writing strategies out of the total number of strategies used.

Experts Novices I Novices II Expertise Instruction n = 12 n = 22 n = 22 (df = 32) (df = 21) Strategy

M sd M sd M sd t t

Rereading .14 .07 .15 .06 .14 .05 -.09 .021 Rhetorical refining .19 .13 .03 .03 .03 .04 5.71 -.15 Local planning .04 .04 .11 .07 .04 .05 -3.67 5.59 Translating L1 to .32 .18 .53 .14 .53 .12 -2.59 .02 L2 Total # of strate- 87.8 28.4 63.5 17.2 71.2 14.1 gies

Underlined scores are statistically significant at p<.0125. a The total numbers of strategies means the total number of tokens of the strategies used. Thus, if the same strategy was used three times, it was counted as three instead of one.

In contrast with the participants’ rereading behavior, there were significant differences between the experts and novices I in terms of Rhetorical Refining and Local Planning. First of all, the experts used Rhetorical Refining significantly more often than the novices. In Rhetorical Refining, writers try to choose the most appropriate L2 expressions to convey their meanings, as in Example 1 (the participants’ accounts in all examples below were originally given in Japanese except for the Eng- lish words in quotation marks, and the accounts were translated into English by the author): Example 1:

I was wondering which expression to choose, ‘my feeling was not important’ or ‘my feeling did not matter.’

(Expert 11) Because the writers must be proficient enough in English to refine their expressions in such a way, this strategy is probably related to the experts’ high English proficiency. Roca de Larios et al. (1999) reported a similar phenomenon where the more advanced group of L2 writers spent more time on ‘ideational/textual restructuring’ (Roca de Larios et al., 1999: 33), including the writers’ search for stylistically better expressions, than the less proficient group, who spent more time seeking ways to compensate for their limited L2 proficiency. Furthermore, such a characteristic of advanced writers may be related to the common core of composing competence shared by L1 and L2. For example, reporting that L2 writers with L1 writing expertise also demonstrated a similar behavior of paying special attention to word/phrase 68 SASAKI choice when they wrote in L2, Cumming (1989) speculated that such a behavior may be a manifestation of what Gardner (1983) called ‘linguistic intelligence’ (Cumming, 1989: 73) that transcends L1 and L2 differences. As for Local Planning, the novices stopped for local planning significantly more often than the experts. This behavior is related to the writing style differences between these two groups (see the Writing Style section below). Furthermore, the novices stopped to translate their ideas into L2 more often than the experts (t = -2.59, p = .024, ns.): About half of the novices’ strategies were devoted to translating generated ideas into English whereas only one-third of the experts’ strategies involved translation. Although the difference between the two groups was not statistically significant, we may at least say that the novices had a tendency to stop for translation more often than the experts because the obtained p value was close to the determined alpha level (.0125) for statistical significance, and because applying a Bon- ferroni correction tends to become too conservative as the number of comparisons increases (Nagata & Yoshida, 1997). As was speculated in the pilot study, this result is probably related to the difference in L2 proficiency between the experts and the novices. Furthermore, it can be speculated that the novices’ spending so much time on translation would be a factor limiting their fluency. A similar relationship among fluency, L2 proficiency, and mental translation was also reported in Sasaki and Hi- rose (1996) based on a sample population of comparable EFL backgrounds. In contrast, after the two semesters of instruction, novices II were still not able to rhetorically refine their expressions very often, and they still had to stop to translate often. However, they made significantly fewer local plans while writing. Al- though we have to take possible topic effects into consideration here again, we could hypothesize that the novice students would use similar strategies for the similar argumentative topics given the significantly high correlations between the quality of novices I and II’s compositions (recall Section 3.1). Thus, Hypotheses 5 and 6 were largely confirmed, except for the case of rereading and translating from L1 to L2. The experts stopped to refine their expressions more often than the novices, whereas the novices stopped to make more local plans. Furthermore, although the difference was not significant, the novices tended to stop to translate from L1 to L2 more often than the experts. After two semesters of instruction, the novices made fewer local plans. However, they still had to stop to translate as often as before. Here again, it is suggested that students can change their use of writing strategies as a result of instruction, which accords with the results of previous L1 and L2 studies (e.g., Anzai & Uchida, 1981; Bereiter & Scardamalia, 1987; Kobayashi & Rinnert, 1998; Cohen, Weaver, & Li, 1998).

3.5 Writing styles Table 5 presents the relative distribution of different writing styles among the three groups. As explained in the Method section, these three writing styles are related to the three strategies of Global Planning, Local Planning, and Thematic Planning. Most of the experts were classified as Type A (Write according to detailed global planning of both the content of the text and the most effective way to express it EFL LEARNERS’ WRITING PROCESSES 69 based on the assessment of the given task) or Type B (Write according to detailed global planning of the content of the text), and most of the novices were classified as Type C (Write according to rough global planning and consecutive local plans). Af- ter the two semesters of instruction, most novices changed to Type B. Although here again we must consider possible topic effects for the novices I and II comparison, we can assume that the novices would be likely to use similar writing styles for the similar argumentative topics (Prompts 1 to 3), especially when novices I and II’s composition quality was significantly correlated (recall Section 3.1). The results of a chi-square test and McNemar’s test indicated that there was a significant difference between the experts and the novices, and novices I and II (χ2 = 22.33, p < .001 for experts/novices; z for McNemar’s test = 3.90, p < .001 for novices I/II). Thus, Hypotheses 7 and 8 were both confirmed: The experts tended to plan detailed overall organization whereas the novices tended to make a less detailed plan; after the two semesters of instruction, the novices learned to do global planning of the content of the text, but their global planning was different from that of the experts in that the novices did not plan how to implement the content in the most effective way. It is interesting that the results related to Hypothesis 8 concerning the effects of instruction concur with those of previous developmental product-oriented studies in that instruction can direct learners’ attention to more global aspects of writing, although the aspect studied varied (e.g., organization skills for Berman, 1994; revision skills for Kobayashi & Rinnert, 1998). On the other hand, the present results also suggest that such shift of the learners’ attention patterns over a relatively short period of instruction time is probably related to Type B writing style, but not to Type A writing style, which seems not to have been acquired by the present novice participants after two semesters.

Table 5: Relative distribution of different writing styles (percentages).

Experts Novices I Novices II Style14 n = 12 n = 22 n = 22

Type A 58.3 0 0 Type B 33.3 13.6 90.9 Type C 8.3 86.4 9.1

χ2 = 22.33 (p < .001) for Experts/Novices I; z for McNemar’s test = 3.90 (p<.001) for Nov- ices I/II.

14 Styles: A: Write according to detailed global planning of both the content of the text and the most effective way to express it (cf. Hayes & Nash’s, 1996 content and non-content planning); B: Write according to detailed global planning of the content of the text; C: Write according to rough global planning (i.e., thematic planning) and consecutive local planning (see Appendix for the descriptions of the different types of planning). 70 SASAKI

As in the pilot study, the experts’ global planning can be characterized by two types of behaviors that were absent from the novices’ global planning. First, as can be seen in Example 2, the expert writers assessed the characteristics of the given task at a relatively early stage of planning. Example 2:

Because this is a letter to a readers’ column in a newspaper, I was thinking about using my own experiences [to support the idea that students should not be forced to wear uniforms]15.

(Expert 11) Secondly, the expert writers sought the most appealing way of presenting the text to the probable readers while searching though available resources in long term memory as in Example 3: Example 3:

Just presenting several reasons to support my opinion here would not be so interesting to the readers. But my directly saying that school uniforms should be abolished would not be appealing, either. So, I wondered if there are any other angles to look at the same thing. Then, although it is not entirely a new perspective, it occurred to me that the voices of the students who are actually forced to wear uniforms would be the most convincing to the readers.

(Expert 9) In contrast to these behaviors observed in the experts’ global planning, none of the novices showed such characteristics, even after the two semesters of instruction. They simply planned what they were going to write for overall organization as in Example 4 (see Flower et al., 1992 for a similar finding): Example 4 (Novice 6):

Researcher Writer

What were you going to write before you started to Because I agree with the opinion [introducing Eng- write? lish as an elementary school subject], I was going to write it. Have you decided what you are going to write in Yes. the end? What are you going to write? I am planning to conclude the composition by writing that we should use more and more English these days. I see. Have you decided what you are going to Yes16 (showing the researcher the draft he had write in the middle? written).

15 The insertions in brackets are supplementary explanations given by the author. 16 The draft contained the following seven Japanese sentences [translated by the researcher]: English should be taught in elementary schools. We should make children get used to English from the time they are small. We should make children get familiarized with Eng- lish. We should teach the joy of learning English --> They can continue to study English when EFL LEARNERS’ WRITING PROCESSES 71

Have you? Oh, this is your draft? Yes. Have you decided everything you are going to Yes. write in the middle?

Can you briefly explain what you are planning to I thought of two reasons, so I was planning to write? elaborate each reason by added one or two sentences. I was planning to write these two reasons so that they would lead to the concluding sentence. I see. Did you think about any other thing? No.

Regarding the experts’ global planning, it is also noteworthy that, unlike the findings in the pilot study, five out of the 12 experts (41.7%) were not concerned about ‘the most effective way to express’ the content they had planned in the beginning, and thus did not belong to Type A. One of them did not even do any global planning before starting to write. He simply planned he was going to write in the beginning of the composition (i.e., local planning), and started to write. However, a closer look at their accounts of what they did after starting to write reveals that they also made great efforts to assess the characteristics of the given task and shape the text to meet the probable readers’ needs and purposes. The experts’ frequent use of rhetorical refinement strategies (see Table 4) is one such example. In addition, in their local planning, these experts tried to search for the best way to express their ideas according to their assessment of the characteristics of the given task as in Example 5: Example 5:

[Here] I was thinking very hard about what I should write next (Researcher: You mean what you were going to write?) I wondered what kind of people would support the idea ‘that the uniformity and discipline should be maintained in school.’ (Uh, huh.) That means, I thought very hard here, well, that this idea will be convenient for the people who are on the side of educational administration. (Uh, huh.) So, I wondered if I should say something like the Ministry of Education or the Board of Education, but I don’t know how I can express such ideas...

(Expert 6 who was categorized as belonging to Type B) None of the novices exhibited such behaviors while writing even after the two semesters of instruction. Cumming (1989) reported similar phenomena where the L1 writing experts organized their ideas ‘through advance planning or emergent planning (emphasis added)’ (Cumming, 1989: 114). That is, some experts in Cumming’s study planned the overall organization of the composition before starting to write just like many of the experts in the present study, whereas the others ‘planned each composition as it was emergent on the page’ (Cumming, 1989: 115) while paying attention to the overall organization that was emerging. In contrast with the novice writers, however, they never lost control over their writing or got stuck. They kept writing by ‘amend- ing, adding to, and reordering their phrases frequently, reconsidering how best to

they become high school students --> English will be needed in the future international society. Therefore we should study English more. 72 SASAKI state what they were most recently deciding they wanted to state’ (Cumming, 1989: 116). We can find some similarity between these emergent planners and those experts in the present study who did not do elaborate pre-writing planning. It can thus be speculated that although efforts to find the most effective way to express their ideas were often observed in the experts’ pre-writing planning in the present study, such effort or behavior itself, rather than the pre-writing global planning, might be at the innermost core of writing expertise. Moreover, if I am allowed to speculate further, such behavioral characteristics of the expert writers might be a manifestation of a more general cognitive ability, or what Bachman (1990) called ‘strategic competence’ (Bachman, 1990: 98), which enables a language user whatever the skill they may be using, to assess the context, plan an appropriate action, and actually execute it in the most effective way. Here writing expertise might in fact be a part of strategic competence rather than an independent ability. Finally, one last qualitative finding related to the participants’ writing style was that, as in the pilot study, both global and local planning appeared to guide or monitor the participants’ writing process, whether they were the experts or the novices. I did not use ‘(Global or Local) Plan Monitoring’ as an independent writing strategy category in the present study because the pilot study results revealed that the phenomenon of plan monitoring was observed in many identified strategies as a kind of ‘secondary’ activity engaged in by the participants. See Examples 6 through 8 (see also Appendix C of Sasaki, 2000 for the details of these strategies): Example 6 (‘Local planning’ monitoring of Novice 7’s writing, which was manifested in her ‘Translating from L1 to L2’)

Here I wanted to write [in English] ‘you can wear uniforms only when you are students’

Researcher: Yes.

But I couldn’t think of an English word to express the meaning of ‘to wear.’

Researcher: Is that why you stopped writing here?

Yes, I was trying to find the right word.

In Example 6, what Novice 7 was actually doing is trying to translate her idea ‘you can wear uniforms only when you are students,’ but by doing so, she was simultaneously implementing the local plan she had previously made. In other words, that local plan was guiding her translating activity here. In contrast, in Example 7, Novice 17 originally planned to say ‘Students can wear whatever they want to at school,’ and retrieved that idea from his rough draft in Japanese. When it came to the time he actually put that idea on the paper, however, he wanted to revise it by adding the clause ‘Some people think that’ before expressing it in English. Here, we can see that Novice 17’s local plan was giving a general direction to his writing by having been retrieved although it was implemented in a more elaborated way. EFL LEARNERS’ WRITING PROCESSES 73

Example 7 (‘Local Planning’ monitoring of Novice 17’s writing, which was manifested in his ‘Plan Retrieving’)

Researcher: Were you reading the top part of your rough draft written in Japanese?

Yes, I looked at the original sentence (in the draft) ‘Students can wear whatever they want to at school [originally said in Japanese]’ and I was trying to change the sentence to ‘Some people think that students can wear whatever they want to wear (originally said in Japanese).’ Finally, in Example 8, Expert 6 was wondering if he should make the first paragraph longer, but realized that doing so might ruin the overall coherence that had been planned in the beginning. Here we can see that Expert 6 was monitoring his writing process by referring to his global plan. Example 8 (‘Global Planning’ monitoring of Expert 6’s writing, which was manifested in his ‘Rereading’)

I was reading the end of this paragraph, and I was wondering if I should end this paragraph like this.

Researcher: I see.

I felt I should write more [according to my plan], but I thought, ‘Wait a second, it would be disastrous [the whole composition would lose balance] if the first paragraph gets longer than this!’

Researcher: Hmmm. I see.

So, I was thinking about concluding this paragraph with this last sentence. A similar phenomenon of using plans to monitor or guide the writing process has been reported in studies of both English and Japanese L1 writing (Anzai & Uchida, 1981; Faigley, Cherry, Jolliffe, & Skinner, 1985; Uchida, 1990).

3.6 Summary of the characteristics of the participants’ writing processes The findings of the present study can be summarized as follows. 1) The experts wrote longer texts at greater speed than the novices. 2) After two semesters of process writing instruction, the novices did not improve their writing fluency. However, their writing ability seems to have improved. 3) While writing, the experts stopped to refine their expression more often than the novices, whereas the novices stopped to make local plans more often. The novices also tended to stop to translate from L1 to L2 more often than the experts. 4) After two semesters of instruction, the novices made fewer local plans than before, but they stopped to translate from L1 to L2 as often as before. 5) The experts spent a longer time before starting to write, planning both detailed overall content of the text and the most effective way to express the content, whereas the novices spent a shorter time making a less detailed plan. 74 SASAKI

6) After two semesters of instruction, the novices learned to spend a longer time before starting to write, planning detailed overall content of the text. 7) The experts’ global planning and subsequent writing was based on their assessment of the characteristics of the given task for successfully achieving the task. Such behavior appears to be a manifestation of a core aspect of writing expertise that cannot be acquired over a short period of time. 8) Both ‘Global Planning’ and ‘Local Planning’ guided the participants’ writing processes, whether they were experts or novices.

Figure 1 presents summary illustrations of these results. These flow-chart illustrations for the three types of EFL writers (experts, novices I, and novices II) are not presented as faithful records of these writers’ micro-level writing processes (cf. Zimmerman’s 2000 ‘formulating model’), but as illustrations of the findings and speculations that resulted from the present study. In other words, these models (if I can call them ‘models’) are crude, but representative of the differences found among the three different groups in the present study. The illustrations presented in Figure 1 do not include the 15 other writing strategies (e.g., Conclusion Planning, Generating Naturally Generated Ideas, Mechanical Refining, Local Text Evaluation) that were included in the pilot study because the differences among the participants in terms of these 15 strategies were not tested in the present study. Nevertheless, I believe that the Figure 1 illustration will help the readers to better understand the differences among the different groups of writers, which were supported (if not confirmed) by the results of the present study. In Figure 1, experts spend a significantly longer time for pre-writing planning, where they plan both the overall content of the text and the way to express it in the most effective way based on assessment of the characteristics of the given task. This behavior is a manifestation of writing expertise, which may transcend both L1 and L2 writing, and which may be part of what Bachman (1990) called ‘strategic competence.’ After they complete the global planning, they usually make several local plans to achieve the global plan. When they make these local plans, they also tend to search for the most effective way to express their ideas. This is also considered to be a manifestation of writing expertise. After making local plans, experts spend time refining their expressions before producing the following text output, which subsequently leads to retrieval of the local plan for the following content. Such a procedure may be interrupted by occasional translation of the generated ideas from L1 to L2. This process continues until the writers feel that they have finished writing (whether or not they are satisfied with the whole written product), with the original global planning guiding the entire writing process (i.e., they use their global plans to monitor their writing process). EFL LEARNERS’ WRITING PROCESSES 75

Novices II Experts Novices I Strategic Competence Thematic Assessment of the Planning characteristics of the (Rough Global Global given task Planning) Planning of ? the content of Efforts to Monitor the text find the most Global effective way Planning of to express both the content Local ideas of the text and Planninng the most effective way to Monitor Monitor Monitor express it

Composing Translation Local Competence Monitor from L1 to Planning L2 Monitor Monitor Local Planning Text Output Translation from L1 to L2 Translation Fluency from L1 to L2 Local L2 Planning Proficiency Rhetorical Refining Text Output Translation from L1 to L2 Text Output Local Planning

Local Text Output Planning

Translation Translation Local from L1 to from L1 to L2 Planning L2

Rhetorical Refining Translation from L1 to L2 Text Output (Improved)

Text Output

These processes continue for each group of writers until the writers feel that they have finished writing, whether or not they are satisfied with what they have written.

Figure 1. Flow-chart summary illustration of the present study results. See note next page. 76 SASAKI

Note. In Figure 1, each writing strategy is enclosed in a square. The unidirectional arrows indicate one-way flow from one writing process to the next. The length of the straight lines with arrows representatively suggest the relative length of time which the process enclosed in the upper square might take: the longer the arrows, the longer it might take to complete the process. The strategies written in boldface are the ones used often, and the strategies written in standard typeface are the ones used less often.

In contrast, novices I without any process writing training, make a thematic plan or a rough global plan followed by a local plan. In order to realize this local plan, the novices have to spend time translating the generated ideas into L2 because of their low L2 proficiency. After managing to produce one piece of L2 text, they make a second local plan to be translated into L2. Probably because so much of their attention is paid to the translating activity, they usually cannot spend much time refining their expressions. They are satisfied as long as they can translate what they want to express. Although the original thematic plan guides their writing processes, they have to make a relatively detailed local plan every time they run out of related ideas about the topic, which prevents them from writing longer and faster. This continues until they feel that they have nothing more to say. Finally, novices II after two semesters of process instruction have learned to make a detailed global plan of what they should say about the given topic. Thus, they no longer have to stop to make so many local plans every time they finish expressing an idea about the topic. Probably because the original global plan orchestrates their writing process more effectively than before the instruction, they can now produce better-quality writing with improved coherence and organization. However, the period of two semesters is not long enough to improve their L2 proficiency, and they still have to spend much time on translation. They still cannot refine their expressions very often, and they still cannot write faster or produce longer texts. In order to attain the writing expertise exemplified in the experts’ flow-chart, the novices will need ‘consistent practice in a variety of similar contexts to the point of proceduralization or automaticity’ (Grabe & Kaplan, 1996: 129) through many years of experience. They may also lack some strategic competence or ‘linguistic intelligence’ (Gardner, 1983: 73), which most likely cannot be acquired through instruction. The extent to which such writing expertise can be taught is an issue that definitely should be investigated in future studies.

4 CONCLUSION Using multiple data sources, this chapter investigated the cognitive processes of Japanese EFL writing experts and novices while writing an argumentative exposition in English. Although the novices’ within-subject comparison might have been affected by possible topic effects, the study is still significant in that it incorporated both cross-sectional and longitudinal investigation of sample sizes that were large enough to allow the hypotheses to be tested by statistical procedures. However, the results of the present study cover only part of the complex mechanism of L2 writing EFL LEARNERS’ WRITING PROCESSES 77 processes (see, for example, Grabe & Kaplan’s, 1996: 226, model of writing as communicative language use as a more comprehensive model). Moreover, the obtained representative patterns of different types of EFL writers’ writing processes are still preliminary findings, and thus are subject to revision and elaboration in light of the results of future studies. First, the present study should be complemented by qualitative studies that examine the details of EFL writers’ individual writing processes, which were not captured in this empirical confirmatory study. For example, how each participant within a particular group in the present study used different types of strategies (a total of 21 types) should be individually followed, and compared with the ways the other participants used these strategies in terms of content and order. Adopting the notion of ‘writing signatures,’ suggested by Levy and Ransdell (Levy & Ransdell, 1996: 158) to refer to the writing processes that are characteristic of individual L1 writers, should be useful in such investigation. Furthermore, when such individual comparisons are made, it might also be possible to investigate the changes in the writers’ psychological state in terms of new knowledge formation (e.g., whether the writers feel that they have created new ideas as a result of their writing) as advocated by Galbraith (1999). Detailed qualitative studies exemplified by Levy and Ransdell and by Galbraith would be useful for filling the gaps in the models presented in Figure 1. Secondly, future studies should investigate affective and emotional factors that were not treated in the present study. Affective factors such as motivation or attitude have not been given much attention in previous writing models, but have begun to be recognized as crucial for understanding writing behavior (e.g., Hayes, 1996). In Sasaki and Hirose (1996), a product-oriented study that targeted a similar sample population of Japanese EFL writers, we also found that confidence in L2 writing for academic purposes was one of the factors that distinguished the good and weak writers. Adding the results of future studies that examine how such affective factors influence the actual writing processes would make the target writing process models more comprehensible. Finally, the model should also go beyond the cognitive domains over more con- textualized domains in order to make it more realistic. If we assume that writing cannot be conducted in a social vacuum and that writing is ‘social construction’ (Cumming, 1998: 61), we cannot ignore the social/cultural contexts where the writing takes place. If we hope to extend the scope of our research in that direction, an ethnographic approach such as the one advocated by Grabe and Kaplan (1996) might be a good place to start. Thus, we should ask ‘the basic question: Who writes what to whom, for what purpose, why, when, where, and how?’ (Grabe & Kaplan, 1996: 203). For example, the present study can be replicated with different types of writing under different conditions. Using other topics and types of tasks, such as writing a complaint letter or a project report17 to be completed within a longer span of time with the help of dictionaries, may reveal somewhat different processing phases. The study can also be replicated with other sample populations that have different L1 or L2 and varied educational/cultural backgrounds. When we conduct

17 These examples were identified by Tannenbaum, Rosenfeld, and Breyer (1996) as writing tasks that might be required in the field of English for international communication. 78 SASAKI such studies, we should be careful to ‘situate writing socially as part of their explanations’ (Grabe & Kaplan, 1996: 215) while also paying attention to the writer’s own (i.e., emic) perspective (Ramanathan & Atkinson, 1999). Incorporating the results of these studies will surely contribute to enriching the content and usefulness of the comprehensive L2 writing process model we can eventually hope to build.

ACKNOWLEDGEMENTS An earlier version of this chapter was presented at the second Symposium on Sec- ond Language Writing at Purdue University, West Lafayette, Indiana, on September 16, 2000. I would like to thank Yasuko Nogami and Toru Kinoshita for statistical advice, Ryu Itoh for coding part of the data, Junko Hayashi and Ryu Itoh for scoring the compositions, and Carol Rinnert and anonymous reviewers for their valuable comments and suggestions. Needless to say, the positions taken and any errors that may remain are solely my responsibility. The preparation of this manuscript has been aided by a research grant for the 1999 and 2000 academic years from Nagoya Gakuin University. EFL LEARNERS’ WRITING PROCESSES 79

APPENDIX

SIX ENCODED STRATEGIES FOR PARTICIPANTS’ RETROSPECTIVE AC- COUNTS OF THEIR WRITING PROCESSES IN THE PRESENT STUDY These six strategies are among the 21 strategies used for the pilot study. Only six of these 21 strategies were investigated in the present study in order to test the eight hypotheses that were formulated as a result of the pilot study. The participants’ accounts in all examples were originally given in Japanese except for the English words in quotation marks.

1 GLOBAL PLANNING Definition: Detailed planning of overall organization Example: I am planning to write that there are several merits for students to have uniforms, and that the rule of wearing uniforms reflects the fact that people should obey rules of the society they are in. I am for the opinion that students should wear uniforms in the sense that by wearing them students will learn that it is natural for the school to have some rules because it is the epitome of a real society. (Re- searcher: Have you decided what you are going to write in the end?) In the end, I was thinking about summarizing the second half of what I said in the beginning. (Have you decided what you are going to write in the middle?) Oh yes, in the middle, I will talk about the purposes of schools. There are usually two purposes for schools. (Yes.) Well, to study and to learn how one should act in the real society in the epitome of a society. A school is a little society, you know. It is a commonplace people share. So, by writing these purposes, from the viewpoint of these purposes, to study and to learn how to act in a common place, for the second purpose, in a common place, there are always some rules, and wearing uniforms is a typical example of such rules. So, I start with the purposes of schools, and according to the purposes, it is OK to have a rule of wearing uniforms. That will be my supporting sentences. Then next, I thought how this issue of whether or not students wear uniforms should be discussed, there must be various ways to look at this issue, but I will try to present my own opinion about the merits and demerits [of wearing uniforms], ah, for example, [wearing uniforms is] economical, and the students don’t have to think about their own fashion. I was thinking about presenting some merits. As for the demerits, [wearing uniforms] is not individualistic, emphasizing the standardized uniformity. (Yes.) These things have been said about uniforms, so I can discuss such demerits. And after that, although [wearing uniforms] is actually non-individualistic, but if the students want to emphasize their own individuality, they can do it by modifying their uniforms in spite of the existence of the uniforms. (Yes.) So, the demerits are not so strong. (Yes.) I was trying to write something like that. That will be short, but after that, I was thinking about summarizing the content before the end. (Expert 7). 80 SASAKI

2 THEMATIC PLANNING Definition: Less detailed planning of overall organization. Example: I was going to write that senior and junior high schools should have uniforms. (I see. Have you decided what you are going to write at the end?) Not particularly. (Not yet?) No. (Have you decided what you are going to write in the middle?) Not yet. (Novice 7).

3 LOCAL PLANNING Definition: Planning what to write next. Example: In the beginning, I wondered what I should write here. (Novice 2).

4 RHETORICAL REFINING Definition: Refining the rhetorical aspect(s) of an expression Example: ... the idea of ‘both sides of the coin,’ (Hmmm.) I first wondered if I should make it ‘either side’ (of the coin), then I thought it would be OK to make it ‘both sides of the coin’ after all, or something like that ... (Expert 5)

5 TRANSLATING FROM L1 TO L2 Definition: Translating the generated idea into English. Example: I was wondering how I could express in English the idea that students should wear uniforms. (Novice 5)

6 REREADING Definition: Rereading the already produced sentence. Example: I was reading the sentence I just wrote. (Novice 19)

THE RELATIONSHIPS BETWEEN BILINGUAL CHILDREN’S READING AND WRITING IN THEIR TWO LANGUAGES

AYDIN YÜCESAN DURGUNOĞLU*, MONTSERRAT MIR**, SOFIA ARIÑO-MARTI***

*University of Minnesota Duluth US), **Illinois State University USA, *** Universitat Ro- vira i Virgili, Spain)

Abstract. In this explorative study, we discuss the links between language, reading and writing variables in the two languages of 26 fourth-grade Spanish-English bilingual children who have just been transitioned to all – English classrooms. Word recognition and spelling proficiencies were correlated both within- and across-languages, but they did not correlate with oral proficiency measures. The form ratings of the writing samples correlated with word recognition and spelling measures, but the content rating of writing samples did not correlate with oral proficiency measures. A qualitative analysis of the writing samples yielded cross-language transfer effects (in both directions) at the level of graphophonic, syntactic and vocabulary knowledge as well as in terms of story structure.

Keywords: bilingualism, reading, writing, cross-language transfer, literacy development.

1 INTRODUCTION Language and literacy (reading and writing) development are closely intertwined processes In addition, there is a strong link between reading and writing development. Traditionally literacy instruction in elementary grades had focused on developing reading proficiency and delaying writing practices until later. Thanks to the emergent literacy research (see Garton & Pratt, 1989 for a review), we now know that writing is an important part of literacy development (e.g., Clay, 1988). Writing and reading facilitate each other’s development because as Fitzgerald and Shanahan (2000) summarized in their review, reading and writing are ‘constellations of cognitive processes that depend on knowledge representations at various linguistic levels’ (p. 40). In the last 30 years, the cognitive processes of reading development have

Y. Durgunoğlu, M. Mir, & S. Ariño-Marti (2002). The relationships between bilingual childrens’ reading and writing in their two languages. In: G. Rijlaarsdam (Series ed.) & S. Ransdell & M. Barbier (Volume eds.), Studies in Writing: Volume 11: New Directions for Research in L2 Writing, 81 – 100. © 2002. Kluwer Academic Publishers. Printed in the Netherlands. DURGUNOĞLU, MIR & ARIÑO-MARTÍ 82 been identified with considerable success (see Adams, 1990 for a review). However, the cognitive processes of writing development in young children have not been studied as thoroughly (for exceptions, see Abbott & Berninger, 1993; Bereiter & Scardamalia, 1987). In reading and writing, identical or similar knowledge representations, cognitive processes and contexts are likely to be used since both tasks involve representing the oral language symbolically. Many contemporary models of literacy development recognize two semi- autonomous cognitive components of skilled reading: (a) linguistic competence as operationalized by listening comprehension and (b) decoding. Thus, to become literate, children need to understand both the spoken language and how it is represented in its written form (Gough & Tunmer, 1986; Hoover & Tunmer, 1993; Juel, Griffith, & Gough, 1986). These processes constitute the ‘building blocks’ of reading development (Durgunoğlu & Öney, 1999; Öney & Durgunoğlu, 1997). Difficulties in these processes, especially in decoding lead to slow or inaccurate word recognition which in turn results in comprehension failure (Shankweiler, 1989). These two cognitive processes – understanding the spoken language and knowing how it is represented in its written form – can also be assumed to operate in writing development. These cognitive processes operating in both reading and writing are described in more detail in a recent model by Fitzgerald and Shanahan (2000). Fitzgerald and Shanahan (2000) highlight four categories of knowledge shared between reading and writing. The first category, metaknowledge includes factors such as knowing about functions and purposes of reading and writing, as well as monitoring one’s meaning making. A second category is prior knowledge, domain knowledge and vocabulary that are used in both tasks. A third category is knowledge about text attributes. These include a) graphophonics: knowledge about graphemes, phonemes and how they match, including factors such as letter knowledge, phonological awareness that affect word recognition and spelling (Ehri, 1997), b) syntax: rules of grammar and punctuation and c) text format: knowing the attributes of larger units of text, such as the knowledge of story grammar or text organization. Fi- nally the fourth category of shared information is procedural knowledge, knowing how to access and use the information in the previous categories. Of course as literacy develops, the impact of these shared knowledge categories evolve and change as well. For instance, in younger students, the graphophonic knowledge has a great impact in both reading and writing development. As students reach middle school years, other variables, for example, understanding complex morphology, knowing about text format as well as prior knowledge and vocabulary gain prominence and the impact of the now-relatively-automatic graphophonic processes are reduced. In this chapter our focus is on the third category described above because the role of these factors in the development of reading has been studied closely (Adams, 1990) and our goal is to investigate how the same factors play a role in the development of writing. Writing development can be analyzed using different criteria. In this chapter we are focusing on two criteria: the form ratings including mechanics, spelling, syntactic structure, and content ratings including the quality of content, such as the richness of information, development of a story line in writing a story, or considering the recipient’s perspective in writing a letter. BILINGUAL CHILDREN’S WRITING 83

Most of the research on reading and writing development has been conducted on monolingual students. However, around the world as well as in the United States, a very common phenomenon is for individuals to speak more than one language. In this chapter we will call these individuals ‘bilingual’ for ease of discussion. How- ever, we do acknowledge that individuals may have varying levels of proficiency in their first (L1) and second (L2) languages. In addition, they may have completed varying degrees of formal instruction in L1 and L2. When it comes to literacy development of bilinguals, we are now beginning to understand their reading development (see August & Hakuta, 1997, for an overview), but research on bilingual writing development is still not very extensive. Most of the research on writing development of bilinguals is conducted on adult EFL learners with well-established first languages and who are learning a second language by choice, for example for academic or job purposes. However, as Roca de Larios, Murphy and Marin summarized in their chapter, ESL and EFL learners have many different characteristics. Hence, there is a gap in our knowledge when it comes to young ESL students who are becoming literate in their L1, L2 or both. Given that reading and writing share a large number of knowledge representations at different linguistic levels, the reading and writing development of bilingual children becomes even more interesting. An important educational issue is how reading and writing proficiencies develop in the two languages of a bilingual child and affect each other across languages, namely cross-language transfer. To state it more specifically, if a child has developed an understanding of a process in L1, will that insight be also available in L2?

2 THE CURRENT STUDY In this chapter, we discuss an exploratory study on language and literacy development of a group of fourth grade children who have just been transitioned to English instruction after being in a Spanish-English bilingual education program in the previous years. Spanish-speaking students constitute the largest group of language- minority students in the United States (73%, August & Hakuta, 1997), so it is important to understand their literacy development. We selected students who were transitioned to all-English classrooms, because after receiving instruction in both Spanish and English, they were assumed to have enough English proficiency. This way, we could assess the students’ proficiency in both of their languages. We measured several different types of literacy and language proficiencies, especially knowledge about text attributes described above as the third category of shared knowledge between reading and writing. We were interested in the following questions:

When it comes to basic reading processes, how are the task performances related within and across languages? As discussed in the introduction, understanding the oral language and how it is represented in print are the two basic processes of reading development. Our goal was to assess decoding knowledge in both languages through word recognition and spelling tests and also assess linguistic knowledge in DURGUNOĞLU, MIR & ARIÑO-MARTÍ 84 both languages through measures of syntax and oral proficiency. Correlating these measures within and across languages gives an indication of overlaps in the reading processes within and across-languages.

How are reading and writing measures related both within and across languages? The quality of writing can be assessed using both form and content ratings as discussed above. The question is how these writing measures are related to basic reading processes in both languages. Graphophonic knowledge is assumed to be shared across reading and writing. Hence it can be expected that word recognition and spelling measures should be related. In addition, the form ratings of the writing samples are expected to be related to independent measures of word recognition and spelling. Syntactic knowledge is also assumed to be shared between reading and writing. Hence, syntactic knowledge is also expected to be correlated with the form ratings of writing samples. Content ratings reflect the richness of content and these ratings are expected to correlate with linguistic proficiency as assessed by oral proficiency. If the insights gained in one language transfer to the second language, then the predicted correlations are expected to occur not only within – but also between- languages.

How do the writing samples reflect the different levels of linguistic knowledge in both Spanish and in English? The writing samples can be analyzed qualitatively to identify instances of the influence of one language on the other, at the level of spelling, vocabulary as well as syntax. In addition, knowledge about text format--for example, how a story is structured or how a letter is written--can operate across the two languages of a child. Hence we can expect that children who have high quality writings in one language, can also show the same pattern in the other language.

2.1 Method 2.1.1 Participants Participants were 26 fourth-grade students (8 boys and 18 girls) from a school in a suburb of a large midwestern city in the United States. All students had Spanish- speaking parents (18 from Mexico, and the rest from El Salvador, Puerto Rico and Colombia), with blue-collar jobs either in the service industry or in the factories around town. As summarized in Table 1, when asked about how much Spanish they knew, the majority said ‘a lot’. Except for one child who reported using only Eng- lish, all children used Spanish (solely or in combination with English) when they spoke to their parents. However they were more likely to use English with friends and teachers. Except for the five who had attended kindergarten in Mexico, all children were attending school in the USA and they had been in bilingual education classrooms. However, currently they were in all-English classrooms. In these classrooms, although several teachers understood and spoke Spanish, they only used English. In their self-reports, the majority of the children rated their reading proficiency in both Spanish and English as ‘good’ or ‘very good’.

BILINGUAL CHILDREN’S WRITING 85

Table 1. Self-reports of children on language proficiency: Number of children who chose that response.

1. How much Spanish do you know?

A lot Some A little bit None 19 5 2 0

2. Which language do you use to speak to …

Spanish English Both Parents 15 1 10 Friends 1 15 10 Teachers 1 21 4

3. How well do you read in……

Very good Good Average Poor Spanish 7 9 8 2 English 1 15 10 0

2.1.2 Materials and procedure The following eight tests were given at the beginning of the school year. Children were tested individually (except where noted) by testers fluent in both English and Spanish. English tests were given in one day, and Spanish in another day.

English word recognition (maximum score possible:126): Two different tasks were used to determine the level of English word recognition: The first was the word identification subtest of Woodcock. In this open-ended task of 106 words, children attempt to read progressively more difficult words and can continue until they make 6 consecutive errors. The second task included 20 exception words (e.g., ocean, is- land) which cannot be pronounced using spelling-sound correspondences (Adams, & Huggins, 1985). The two word recognition measures were highly correlated (r = .84), hence in the following analyses, the two tasks were combined into a single word recognition measure. The score was the total number correct in both tasks.

English spelling (maximum score possible: 161): In this task, the students were given a sheet of paper with numbers on it and asked to spell the 15 words that the experimenters pronounced (hat, dress, lost, duck, date, cheat, rice, shop, huge, mail, elephant, blank, spin, food, nothing). The experimenters first pronounced the word, then used it in a sentence and finally repeated the word. When scoring, both the number of correct letters irrespective of location, and then the location of the sounds were considered. The letters representing initial and final sounds received 2 points, and the sounds in the middle received 1 point. For example, the word ‘shop’ spelled DURGUNOĞLU, MIR & ARIÑO-MARTÍ 86 as shop received 9 points: 4 correct letters+ 2 /sh/ + 1 /o/ + 2 /p/ = 9. The same word spelled as sop received 6 points: 3 correct letters + 0 /s/ +1 /o/ + 2 /p/ = 6.

English oral proficiency (max: 65): Four subtests of the LAS (Language Assessment Scales) test (Duncan & DeAvila, 1987) were used. In the minimal pairs section, the students listened to 30 pairs of words and identified pairs that were dissimilar (e.g., them-them or pet-pat). In the oral vocabulary section, the experimenters pointed to 20 different objects in the picture and the students named those objects (e.g., apple). In the listening comprehension test, students listened to 10 different sentences and selected the picture that represented each sentence (e.g., The cat jumped and the dog sat). In the story retelling, the students listened to a short story and then retold it. The retellings were transcribed and scored according to a 5-point scale using the guidelines in the test manual.

English syntax (max: 60). In this test, the students listened to a sentence read by the experimenter and corrected the syntactic error in the sentence (e.g., Raul has three cat). The sentences included errors of different types such as tense, inflection, word order (Johnson & Newport, 1989). Each correct answer was given 3 points. If the students corrected the error, but made a mistake in another part of the sentence, they got 2 points. Finally, if they identified the error, but did not correct it accurately, they got 1 point.

Spanish word recognition (max: 20): In this test students read 20 common words with varying degrees of morphological complexity (lee, charlar, trabajoso). The number correct was the score.

Spanish spelling (max: 155): The students spelled 14 Spanish words (pelo, nado, bola, sapo, tela, beso, bajo, camión, peine, queso, pronto, circo, baño, calle). The procedure and the scoring system were identical to that for English words.

Spanish oral proficiency (max: 65). Four subsections of the Spanish LAS test (Dun- can & De Avila, 1987) were used. The subsections, testing and scoring procedures were identical to the English test, but the test was not a direct translation.

Spanish syntax (max: 60). This test was parallel to the English version, but not a direct translation. The testing and scoring procedures were identical to those for the English version.

Writing tasks. The students wrote four pieces of text, two in November and two in May, as a group. Two genres of writing were used, a story and a letter. In Novem- ber, they wrote a story in English and a letter in Spanish. In May, they wrote a letter in English, and a story in Spanish. For the stories, the students were given a photograph and asked to describe the events in the photograph. For the English story writing, ‘the old lady and the tomatoes’ picture (Chall, Jacobs & Baldwin, 1990) was used. The black and white photograph (hererafter called ‘lady with tomatoes’) depicted an old lady with three tomatoes in her hand, and a puzzled expression on her BILINGUAL CHILDREN’S WRITING 87 face. For the Spanish story writing, the students saw the color photograph of a boy whose face had turned green, and there were three adults staring at him with various expressions of surprise on their faces (called ‘green face’). For the letter writing tasks, students were given the name and characteristics of the recipient and the topic for the letter. For the Spanish task, children wrote a letter to a girl named María who lived in Mexico, and described their school to her. For the English letter task, they wrote a letter to the school librarian recommending their favorite book to be purchased for the library. To prevent practice effects, we did not give the same topic to the same child twice. However, it was an oversight on our part not to counterbalance the writing topics across languages. The writing samples were scored using both qualitative and quantitative measures. Quantitative measures included, number of words, sentences, and percent of misspellings. The qualitative measures included separate ratings of form and content. The form and content ratings were based on the work of Bruce (1983); Chall, et al, (1990), and Myers (1980). Three judges rated each writing sample. The average rating for each sample was computed by taking the agreed rating of at least two judges. Cases when all three judges disagreed were resolved by discussion. The percent agreement of at least two judges ranged from 84 –100% for the four writing samples. Table 2 summarizes the guidelines for form and content scoring.

Table 2. The guidelines for scoring the writing samples.

Content scoring for letters 1 No coherence, one sentence or two disconnected sentences 2 Flat list of facts or details 3 Interesting varied presentation, enumeration of facts, some explicit connection between facts 4 An interesting, varied presentation, developed logically, example, detail, cause and effect present

Content scoring for stories 1 Narrative theme not developed, no cohesive storyline 2 Cohesive storyline present, but no attention to supporting detail such as character development and setting 3 Clear development of theme or storyline with some attention given to supporting detail such as characters or setting 4 Clear development of an imaginative theme or storyline, great attention given to supporting detail such as characters or setting

Form scoring for both tasks 1 Severe sentence structure problems (fragmentation, grammatical, mechanical problems) 2 Sentence structure problems, other grammatical errors 3 Few isolated errors in mechanics, grammar largely correct 4 No errors or 1-2 errors

DURGUNOĞLU, MIR & ARIÑO-MARTÍ 88

2.2 Results and discussion 2.2.1 Quantitative analyses The performance levels on language, word recognition and spelling measures are summarized in Table 3. The students performed quite well on the Spanish tasks. On parallel, standardized LAS tests, Spanish and English oral proficiencies were at comparably high levels (88%). The students also spelled words quite accurately in both English and Spanish. However, since the spelling tests were not standardized tests, but rather developed by us, these possible ceiling effects may be an instrument artifact. Among all tasks, recognizing English exception words proved to be the most difficult.

Table 3. Means, standard deviations, percent accuracy and range of the language and literacy tasks.

M sd Percent Range

English Woodcock word recognition 61.50 9.9 42 – 86 Exception word recognition 11.23 2.7 56 6 – 18 Spelling 148.50 9.1 92 125 – 160 Syntax 38.40 14.6 64 12 – 60 Oral proficiency 56.92 2.7 87.5 51 – 61 Spanish Word recognition 16.50 3.9 82.5 6 – 20 Spelling 148.54 6.4 96 133 – 155 Syntax 44.89 11.3 75 9.5 – 58 Oral proficiency 57.13 3.1 88 52 – 63

Table 4 summarizes the correlations among these measures. First considering within-language results, not surprisingly, English word recognition and spelling tasks were correlated (.63). Also Spanish word recognition and spelling tasks were correlated (.67). This pattern supports the assumption of a graphophonic knowledge base shared by both reading and writing, especially for word recognition and spelling. Replicating our previous findings (Durgunoğlu, Nagy & Hancin-Bhatt, 1993), the oral proficiency measures were not correlated with word recognition and spelling measures in either English or in Spanish. When correlations across the languages are considered (in italics in Table 4), English and Spanish syntactic measures were correlated (.44), as were the spelling (.48) tasks. There was no significant correlation between Spanish and English oral proficiency measures. These results indicate that at the graphophonic level, there are some commonalities across the two alphabetic languages. However, this common BILINGUAL CHILDREN’S WRITING 89 factor is not related to oral proficiency, but to the decoding processes. The significant correlation between English and Spanish syntax tests may be due to a common metalinguistic awareness used in both tasks. Children who could analyze and correct the syntactic structure of a sentence accurately in one language were more likely to show this analytic accuracy in the other language as well.

Table 4. Correlations among the English and Spanish literacy and language measures.

2. 3. 4. 5. 6. 7. 8.

1. English word recognition .63 .33 .03 .36 .44 .05 .40 2. English spelling -.04 -.14 .35 .48 -.16 .35 3. English syntax .11 -.25 -.12 .44 .06 4. English oral proficiency -.18 -.14 -.01 -.04 5. Spanish word recognition .67 .41 .32 6. Spanish spelling .20 .27 7. Spanish syntax .13 8. Spanish oral proficiency

Underlined correlations are statistically significant at p< .05

The next question is how these language and literacy variables are related to the writing measures. In Table 5, a descriptive overview of the writing samples is given. Children tended to write more in Spanish, possibly because it’s their home language. Overall for both English and Spanish, their stories were longer than their letters. Contrary to what we found on the single word spelling tests, the misspellings were quite common in the writing samples. Of course, it is not surprising to see more spelling errors in the writing samples because first of all, the students were using a wider variety of words and also they were focusing on the content more than on the mechanics while composing.

Table 5. Means and standard deviation of the quantitative measures of the writing samples.

Number of words Number of sentences Percent misspellings

M Sd M sd M sd

English story* 59.04 24.4 8.33 3.7 18 0.1 Spanish story 73.68 31.0 9.24 4.4 22 0.2 English letter 46.52 11.6 6.72 1.5 15 0.0 Spanish letter 60.84 22.4 9.80 4.2 16 0.1

* English story: Lady and the tomatoes; Spanish story: Green face; English letter: Letter to the librarian; Spanish letter: Letter to María. DURGUNOĞLU, MIR & ARIÑO-MARTÍ 90

When the form and content ratings of the 4 samples were correlated with each other, there were not many significant correlations, except for the following: As would be expected, the form and content ratings for each sample were correlated. (.45 for the English story, .37 for the English letter, .42 for the Spanish letter, and .43 for the Spanish story). This pattern is not surprising, because samples with better form ratings are easier to understand and can be rated more highly in terms of content. More interestingly, the form ratings were correlated across languages. The correlations between Spanish and English form ratings were .56 for the November samples, and .37 (marginally significant) for the May samples. Because the form ratings evaluated the general mechanics of the samples, such as spelling and syntax, these correlations across languages provide more support for the common decoding processes discussed above for word recognition and spelling measures (see also the discussion about Table 6, further below). There was a significant correlation between the content ratings of the English and Spanish stories (.53) even though the stories were written at different times (English story was written in November, and Spanish story in May), and they were about different topics. This correlation suggests that the general understanding of a story’s structure and its contents seems to transfer across languages. Children who are good storytellers in one language also show this trend in the other language. This pattern will be discussed further with qualitative data. However, the letters in Span- ish and English did not show such a correlation, possibly because the topics (the letters to María and to the librarian) as well as the communicative goals required different kinds of detail. The librarian letter required giving a book title and explaining the rationale for the selection. The María letter, on the other hand, involved providing detailed descriptions of a school day, but did not require any supporting arguments. It is also possible that topic familiarity was a factor. Describing the school is a more familiar and easier task than recommending a book, especially for children who do not like to read.

Table 6. The correlations between the writing evaluations and other measures.

English Spanish Lady & Tomatoes Librarian letter María letter Green face Language and literacy measures Form Content Form Content Form Content Form Content

1. English word recognition .60 .25 .25 .06 .60 .23 .37 .11 2. English spelling .57 .37 .03 .10 .50 .26 .48 .05 3. English syntax .51 .24 -.04 -.35 .04 -.23 -.29 .04 4. English oral proficiency .04 -.11 -.22 .16 -.10 -.12 -.29 -.22 5. Spanish word recognition .26 .04 .15 .27 .71 .41 .51 -.03 6. Spanish spelling .38 .23 .25 .08 .78 .39 .71 .20 7. Spanish syntax .25 .17 .02 -.07 .38 -.05 .00 .01 8. Spanish oral proficiency .20 .14 -.22 -.21 .47 .48 .07 -.08

Underlined correlations are statistically significant at p < .05. BILINGUAL CHILDREN’S WRITING 91

In Table 6, the form and content ratings of the writing samples are correlated with language and literacy measures. The form ratings were – for the most part – consistent with word recognition and spelling measures, especially within the same language, reflecting the operation of the first building block, decoding, discussed in the introduction. However, contrary to our predictions, the content ratings did not really correlate with any language or literacy measures, including oral proficiency. Of course, it is also possible that our content ratings or the oral proficiency measures were not sensitive enough to detect this link. The sample size was too small to conduct multiple regression analyses, but we could still do qualitative analyses on the writing samples to which we turn next.

2.2.2 Qualitative explorations of content In the qualitative analyses, we will first provide some samples to illustrate the differences in the overall quality of writings and highlight the factors that make some samples more mature than others (see the Roca de Larios et al chapter for a more detailed review of developmental studies analyzing differences between mature and less mature writers). In this section of the chapter we give quite a few examples because when averaged across the children, and only expressed as numbers, the rich information in the writing samples cannot always be conveyed. We also provide illustrations of transfer across the writings in the two languages. When judged holistically, the maturity of the writings showed great variation. For example, after viewing the ‘lady and the tomatoes’ picture, some children described what the woman was doing or thinking in a straightforward manner. In contrast, some children provided a storyline with some suspense and a happy or sad ending. The form of the writing sample (e.g., sentence structure, spelling) did not always predict the richness of content. Some children spelled and used grammar well, but their story was a listing of facts (see Lorena’s writing below, Sample 1). Some children, on the other hand, had quite a bit of trouble with the mechanics, but provided a very rich storyline (see Elizabeth’s writing below, Sample 2). Example 1 Maybe the woman thin[think] that somebody is coming or something wrong. Or the woman is jest [just] looking around. Or somebody stold [stole] something. But I think that something broke or something feld [fell] down. Lorena

Example 2 One day a old lare [lady] went to the market. And she bat [bought] some tamatoes[ tomatoes] and She sar [said] I nat [not] fer [fair]. Way [why] tho [do] day [they] cast [cost] to mach [much] cer [said] the old lare. And then the man cer well if you don’t liket [like it] then lelts [let’s] go to kocrt [court] he cer the man. And they left because they har [had] to go to korct. DURGUNOĞLU, MIR & ARIÑO-MARTÍ 92

And then the judge cer that the old lare was enoccent [innocent]. And then the lare lived [lived] hapali [happily] evar [ever] after. Elizabeth The letter to the librarian provided a lot of interesting data on children’s tastes. Some children provided a book title, but did not give a rationale for selecting that book (see Sample 3). On the other hand, some children provided very rich descriptions of the books they like and why they liked them. Example 3 Dear Liberyyen [librarian] I think you sud [should] have thes [this] book becus[because] it is good. I love thes [this] book and I hope you get it. Elena

Example 4 Dear Librarian, I want a book of Butterflys [butterflies] and pleas[please] let it be when they are cadapelers [caterpillars]. And when you send it pleas [please] send it to me and put my name !!Thank you!! your helper, Esmeralda Olivo started the letter by giving the reason for the book suggestion: Example 5 Dere [dear] libaryens [librarians] I have notece [noticed] that you nid [need] my help to pikc [pick] a bast [best] book. My best book is romple stilski [Rumpelstiltskin] it is abot [about] a Millers doghter [Miller’s daughter] she does note [does not] know how to so [sew] and a little man that helps her sow fore [for] her fert [favorite?] baby. Olivo Pablo first discussed his favorite book genres, and then gave specific titles: Example 6 Dear Librarian, My faveit [favorite] book are mystory [mystery] and funny story. I like does [those] becuse [because] the funny’s book make my laugh rely [really] rely hard and I like the mystory because it is very excyting [exciting] the funny books tity [title] is 101 presnt jouch [present joke] book. And the mystory book is the kaun of baske vally [Hound of the Baskervilles]. I tink [think] you sould [should] have the book because all of the kid like it. BILINGUAL CHILDREN’S WRITING 93

Your truly Pablo Anayaeli provided a strong argument for selecting this book not only for her classmates, but also for younger grades. Example 7 Dear Librarians, I want you to buy a book that is named Say Cheese and Die. I want you to buy that book because is enteresting[interesting] and its has adventures. And its good for the forth [fourth] grade and eys [easy]. And if the kids that are loer [lower] then forth grade. There teache [teacher] could read it. ATT Anayeli

Lysett’s letter poignantly expressed her interest in languages and bilingualism: Example 8 Dear librarian, I want you to put Happy Birday [Birthday] Gramie [Grammy]. The author is Susan Pearson. Bicos [because] the book is about one girl that gaves [gives] a present to his grandfather. I like that book bicouse [because] it’s like a happy and sad. Bicouse her grandfather talks a other languesh [language]. And a little english. Your friend, Lysett

The following two letters in Spanish also show the differences in the writing quality. Ivan’s letter (Sample 9) addresses the recipient, introduces himself, provides a lot of detail. Elena’s letter (Sample 10), on the other hand, is a random listing of events, with no concern for communicating with the recipient.

Example 9 Hola Maria, soy Iván tego [tengo] 9 años Hello Maria, I am Ivan I am 9 years old vivo en wuakegan [Waukegan], mi escuela se I live in Waukegan, mi school is llama East school estoy en cuarto called East school I am in fourth grado y he aprendido ingles. grade and I have learned English. Porque tengo un buen maestro. Because I have a good teacher. Vamos a jugar al patio y tenemos we go outside to play and have un maesro de Educación física y a Physical Education teacher and hacemos ejercicio y hay niños we do exercise and there are kids de otros países. Y tengo muchos from other countries. And I have lots amigos como tu. of friends like you.

Ivan DURGUNOĞLU, MIR & ARIÑO-MARTÍ 94

Example 10 Lo que llo [yo] ago [hago] en la escuela es What I do in school is llo [yo] guego [juego] afuera. Las [los] libros en la I play outside. The books in the school escuela And we learn more in the school y a se [así] aprendo mas [más] en la escuela, and play gym in the school. llo [yo] gueg [juego] gem [gym] en la escuela. Elena

The differences in the overall quality of writing can also be observed in the following two Spanish stories: Maria’s story (Sample 11) does not have a very clear description of events or a discussion of a problem. Contrast that with Viviana’s story (Sample 12) that has the setting, problem and its resolution. There are also detailed descriptions of characters and snippets of conversation. Example 11

Las personas andan a The people are merando [mirando] el niño porque looking at the boy because el niño se endava [andaba] the boy was lavado [lavando] los dites [dientes] washing his teeth y las personas estan van and the people are go algers [alegres] y el neno [niño] happy and the boy no se los lava por muchas [mucho] tempo [tiempo] does not wash them for a long time. Maria

Example 12 Una mañana Arturo fue a One morning Arturo went to lavarse los dientes in [y] se los lavo y brush his teeth and he brushed them and despue [después] iso [hizo ] gargajos con la pasta y then he gargled with the paste and se puso vien [bien] verde y se miro en el he turned very green and he looked himself at the espejo y grito. Mama papa abuelo y mirror and screamed. Mother, father grandfather todos vinieron coriendo [corriendo] y se quedaron and viendo muy estraño [extraño] a Arturo y everybody came running and looked ls [la] mama de Arturo dijo hijo pero very puzzled at Arturo y que es esto! Y Arturo le conto toda his mother said son but la istoria [historia] y dijo su papa te asponido [has what is this! And Arturo told all ponido – actually, puesto] the story and said his father you have turned vien [bien ] verde como un sapo y despues very green like a toad and then fueron al doctor y dijo el doctor they went to the doctor and said the doctor que la pasta estava [estaba] podria [podrida] y que that the paste was rotten and that lo enferma y que esto solo quitara it made him sick and that this will only go away cuando se tome su medisina [medicina] y ona when he takes his medicine and one [una] morning Arturo got up and looked himself mañana Arturo selevanto [se levantó] y semiro [se at the mirror and was cured and screamed. miró] Mother father grandfather and everybody came en el espejo y estava [estaba] sano y grito. and smiled. Mama papa abuelo y todos vinieron y sesorierno [se sonrieron].

Viviana BILINGUAL CHILDREN’S WRITING 95

2.2.3 Cross-language transfer The influence of the children’s two languages in their writing was examined using the shared knowledge between reading and writing as the framework. According to the Fitzgerald-Shanahan model, text attributes, namely, graphophonic and syntactic knowledge as well as awareness of text structure are shared between reading and writing. Another shared knowledge is meaning, especially vocabulary. While looking at transfer between Spanish and English we examined these factors separately.

Graphophonic knowledge. The most dominant strategy that we observed in the children’s writing was the influence of Spanish on English spellings and vice versa. For English spelling, the most common strategy was to use spelling-sound correspondences systematically and spell the words as they were heard, hence transferring a strategy that is quite effective for the more transparent Spanish orthography to Eng- lish spellings. English speaking children also make regularization errors (Treiman, 1993), but they also use analogies to exception words in their spelling (Goswami, 1991). Spanish-speaking children in our sample were more likely to make regularization errors. As Elena put it simply ‘[It is easier to read and write in Spanish because] When you read or write some thing, you just sound it out. When you do it English, you have to remember’. (In fact Elena was speaking from experience. Since she could only read 6 of the 20 items on the exception words test, she was aware that some English words cannot be pronounced using spelling-sound correspondences). In the samples below, note how the words ‘witch, wardrobe, adventures, that, favorite, read, buy’ are spelled. Example 13 you shold [should] get wide side school because its reel [real] funny the Lion the Wich and the worldrol [The Lion, the Witch and the Wardrobe] those two books funny and good the Lion the Wich and the worldrol is about some kids thet [that] get in a worldrol and they have a lot of adengers [adventures] Susana

Example 14 My favret [favorite] book is Enormous egg and I like it Bcuase [because] it is interesting to look at. And it’s fun to rid [read] it But you don’t have it. Allot’s of [a lot of] people By [buy] the Book you should have it Ana As seen in the case of the word ‘read’ spelled as rid (Sample 14) and also ‘need’ spelled as nid (Sample 5), the children’s English spellings reflected the influence of Spanish, since Spanish does not distinguish between ‘short i’ and the ‘long i’. Because the children tended to spell the words as they heard them, they omitted the silent letters in both English and Spanish words. Ago for ‘hago’, asemos for ‘hacemos’ aser for ‘hacer’, stor for ‘store’ and hom for ‘home’ were quite common. DURGUNOĞLU, MIR & ARIÑO-MARTÍ 96

(See some examples in writing samples 10, 16 and 17). However, omitting the silent letters such as the ‘t’ in ‘witch’ (sample 13) is also common in monolingual English spellers too. Children also used common English consonant clusters when spelling Spanish words scuela for ‘escuela’, sta for esta, stoy for ‘estoy’, spero for ‘espero’ studios for ‘estudios’, different for ‘diferent’ (see some examples in Sample 15 below). Finally children interchanged sounds between English and Spanish for example i and y. In both Spanish and English spellings, the phoneme /i/ which is sometimes spelled as ‘y’ was interchanged with i or e. For example, ‘happily’ spelled as hapali or ‘y’ [and] spelled as i, ‘mi’ spelled as my, ‘lady’ spelled as late (for examples see Samples 15, 16, 17.) Example 151

Hello Maria how are you, but I am fine look my Hola Maria como sta [esta] pero school is called East School yo stoy [estoy] bien mira my [mi] my teacher is named scuela [escuela] se llama East School Mr. Star and my uncle his mi maestro se llama teacher is named Mrs. Star, Mr. Star y mi tio su the principal is named Mrs. maetra [maestra] se llama Mrs. Star. Von and also she is good la directora se llama Mrs. and if we behave well Von y tambien es buena she gives us achts [?] and on y si nos partamos bien Fridays she announces the ella nos de achts [???] i [y] los winners. venes [viernes] ella anunsia [anuncia] alos [a los] Attentively ganadores Patricia Atenta ment [atentamente] Patricia Example 16 The old late [lady] is thinking how much des [these] the tometos [tomatoes] cas [cost] do you pey [pay] them at. And she saw a man that worki [working] ther [there] and he told her wer [where] do you pay them at and how mock [much] do they cast [cost]. Then she boght [bought] them and went hom [home] Olivo

Example 17 Maria en los Estados Unidos, la escuela Maria in the United States, the school es grande y es muy divertido [divertida] porke is big and is a lot of fun because [porque] they can read a lot and sometimes they take us pueden ler [leer] much [mucho] y abeses [a veces] outside only when we behave nos sacan well. I like more to do para fuera nomas cuando nos partamos stories because then I can write bien. A my [mi] me gusta mas aser [hacer] a lot. When I get istorias [historias] porque asy [asi] puedo escribir home I do homework and go

1 The school and teacher names have been changed in this sample. BILINGUAL CHILDREN’S WRITING 97 mucho. Cuando yo llego a my [mi] to study. casa yo ago [hago] la tarea y me pongo e estudiar Sandra

Syntactic knowledge. In Spanish, the direct object can be attached to the verb in structures with an infinitive (Vamos a comprarlo: We are going to buy it), a gerund form (Estamos haciéndolo: We are doing it) or imperative (Hazlo: Do it). In other verb structures the direct object needs to precede the verb (Lo hago: I do it; La he pintado: I have painted it, etc.) Some children transferred this knowledge to their English writing. In producing the morphology of English, the Spanish structure was used as a guide, leading to amalgamated words such as getit (get it) liket (like it) or baiden (buy them) or stoldan (stole them) as in Sample 18 below. Of course, one can also argue that there are instances when words are joined because their sounds seem to blend in spoken language, such as alat [a lot] in Sample 18. Example 18 The woman can baid [buy] the tomatos [tomatoes] or not baiden [buy them] or stoldan [stole them] or if she baidam [buy them] she can do food or if she stoldan [stole them] she can get in big trouble beacise [because] she stoldan dey [they] cand [can] give a tiket [ticket] from 50 dolar [dollars] and 100 dolar or put dan [put them] in gel [jail]. And she wel [will] get in alat [a lot] troabe [trouble] Patricia Another syntactic transfer was when the subject was omitted in English sentences reflecting the Spanish structure. For example, in the letter to the librarian, Anayeli wrote ‘because is interesting’ (sample #7 above).

Vocabulary: Of course lexical borrowings and transformations are an example of the dynamic forces of language when two cultures come in contact. Some of those words become an accepted new word such as vamos al lonche for going to ‘lunch’. Because they had mostly attended schools in the USA, the children had the school concepts in English. Therefore, they used the English versions or variations of these words in their Spanish letters when they were describing their schools, for example, pricenpal [principal] instead of ‘directora’, gym for ‘gimnasio’ subjectos [subjects] for ‘materias, temas’. In the sample below, in addition to the borrowings from Eng- lish, notice how the child includes the pronoun ‘yo’ [I] as required in the English syntax, but not in Spanish. Example 19 en mi escuela, salimos a fuera [afuera] y in my school, we go outside vamos a baño siempre. y yo and go to the bathroom always. and I DURGUNOĞLU, MIR & ARIÑO-MARTÍ 98 estudio Mathematicas [matematicas] y sciencias study Mathematics and Sciences [ciencias] and many things. and during the y muchas cosas. y durante el day I work and go outside. and we play dia yo trabajo y voy a fuera [afuera]. y jugamos a game of mathematics. un juego de mathematicas Raymundo

Text format. As discussed before, there was a significant correlation between the content ratings of English and Spanish stories. Children who used effective storytel- ling techniques in one language tended to transfer that knowledge to the other language. Below are examples from a student with relatively weak story writing skills and two others with strong story writing skills. (While analyzing the richness of a story, problems with misspellings or sentence structure were ignored.) Elena had minimalistic stories with very little detail and character development in both her English and Spanish stories. Example 20 She wint [went] to the stor [store] and bote [bought] some of tomeroes [tomatoes]. She wint to buy food. She wint to her hous [house]. Elena

Example 21

El nino [niño] es ta [esta] en el bano [baño] se The boy is in the bathroom drinks a green juice toma un jugo verde y el se pone and he turns verde su mama se [su] papa y se abelito [abuelito] green his mother his father and his grandfather all todos se van al go the banio [baño] y se alegran y todo paso con el jugo bathroom and are happy and everything happened vere [verde] with the green juice el nino [niño] se puso banco [blanco] otra ves [vez] the boy turned white again. In contrast, Viviana expressed a dilemma faced by the main character in both of her stories and described how the conflict was resolved. (Viviana’s other story, ‘green face’ was discussed earlier, as Sample 12). Example 22 One day a old woman went to the store to get some tomatoes and wen [when] she got to the store she soy [saw] the tomatoes and they were only one bag and she get the bag with the tomatoes and she soy [saw] they were soy [so] dorti [dirty] and she pot [put] them away and she said her salfe [herself] if I dont’ get does [those] tomatoes she wont do the saled [salad] put [but] if she get the tomatoes the people will said the she get dorti things to eat well she let [left] the tomatoes and BILINGUAL CHILDREN’S WRITING 99

she went to the other store and she bot [bought] some tomatoes and she live happy for fevere [forever] the End Viviana Anayeli also developed the stories by posing a dilemma and describing its resolution in both English and Spanish. She included details about characters and their actions (e.g., suspects speaking in ‘a soft voice’ in Sample 23) and the names of the characters in Sample 24). She also had parts of some standard story elements such as ‘Un dia [one day]’ or ‘once upon a’ in her writing. Example 23 One upon a ther [there] was a old woman that she went to the stor [store]. And she boit [bought] some tomatos [tomatoes]. And she saw one boy that was saing [saying] to a nather [another] boys put dis[this] in your jacket he sad [said] in a soft voice. And the old woman was tinking [thinking] that thae [these] boys were robing [robbing] same ting [something]. then the woman tall the police and the police put them in JELL [jail] Anayeli

Example 24 Un dia un nino [niño] que se yamava [llamaba] One day a boy that was called Jose put on a Jose sepuso [se puso] una formula that his friend gave him. Jose drank it formula que le dio su amigo. Jose se la tomo and turned into a green monster. When y se convirtio en un montro [monstruo] verde. his grandfather that was called Arturo saw him he Cuando called su abuelo que se yamava [llamaba] arturo lo vio el Delia who was his mother and Delia called yomo [llamó] Rutido who was his father. When they a Delia quin [quien] era su mama y Delia yamo looked at him they got scared. The family called [llamó] the police. In a few months an author a Rutido quien era su papa. Cuando ellos lo goes to the house of Jose and told him if miraron se esperanton. La familia yamaron [lla- he could participate in a movie. Jose started maron] to be famous. a la policia. Dentro de unos meses un autor yamo a la casa de Jose y le dijo que si podia partipisa [participar] en una pelicula. Jose empeso [empezó] a ser muy famoso.

3 SUMMARY In this chapter, we discussed the links between language, reading and writing variables in the two languages of fourth-grade bilingual children who have just been transitioned to all–English classrooms. The results will be summarized using the three questions posed in the beginning as the framework. The first question was about the basic processes of reading, how they are related within and across languages. Word recognition and spelling measures did not corre- DURGUNOĞLU, MIR & ARIÑO-MARTÍ 100 late with oral proficiency measures suggesting that the decoding processes are relatively independent of the oral linguistic proficiency. In addition, spelling tasks were also correlated across the languages, implying that with two alphabetic languages, there are some commonalities in the children’s spelling proficiency. The syntactic analysis seemed to have some commonalities across languages, as English and Spanish syntax performances were also correlated. The second question was about how reading and writing measures are related both within and across languages. Replicating previous research, word recognition and spelling proficiencies were correlated within-languages, reflecting the common graphophonic processes between reading and writing tasks. Further evidence for this common process was the correlations of the form ratings of writing samples with the word recognition and spelling measures. The same patterns also held across languages, although they were smaller in magnitude. Across languages, form ratings were correlated, indicating that students who knew the spelling and mechanics of writing in one language also tended to do better in the other language. The content ratings of the writing samples were predicted to correlate with oral proficiency, but this was not supported. One possibility is that content ratings and/or oral proficiency measures were not sensitive enough. However, when only the stories are considered, English and Spanish content ratings were correlated, suggesting that the understanding of text format, story structure can transfer across languages. Children who had this insight in one language used it in their stories in the other language as well. In the future, counterbalancing the writing tasks and topics more fully can yield clearer analyses of these patterns. Finally, the qualitative analysis of the writing samples yielded cross-language transfer effects between Spanish and English (in both directions) at the level of graphophonic, syntactic and vocabulary knowledge as well as in terms of story structure.

LINGUISTIC KNOWLEDGE, METACOGNITIVE KNOWLEDGE AND RETRIEVAL SPEED IN L1, L2 AND EFL WRITING

A structural equation modelling approach

ROB SCHOONEN, AMOS VAN GELDEREN, KEES DE GLOP- PER, JAN HULSTIJN, PATRICK SNELLINGS, ANNEGIEN SIMIS & MARIE STEVENSON

University of Amsterdam, the Netherlands

Abstract. This chapter presents an analysis of the Dutch and English writing proficiency of students who learned Dutch as a second language, compared to students for whom Dutch is the first language. Partici- pants are 13/14 year-old secondary school students. The point of departure in the analyses is a regression model in which writing proficiency is decomposed into three types of constituent subskills: linguistic knowledge (vocabulary, grammar, and orthographic knowledge), speed or fluency of processing linguistic knowledge (word retrieval and sentence building), and metacognitive knowledge of both text characteristics and reading and writing strategies. This study investigates whether there are structural differences in both the Dutch writing and the English writing of the non-native speakers of Dutch compared to their native Dutch-speaking peers. Our hypothesis is that the two groups will differ in their Dutch writing, but not so much in their English writing. Using structural equation modelling, we firstly test whether the writing model for the native Dutch writers, with respect to the explanatory value of the constituent skills, holds for the non-native Dutch writers as well. Secondly, we test whether the English writing ‘decomposition’ is the same across the two student groups. Despite the differences in writing proficiency, no essential differences between the two groups were found in the pattern of weights of either Dutch or English writing proficiency on the three types of predictor variables (language knowledge, retrieval speed, and metacognitive knowledge).

Keywords: L1 writing ability, L2 /FL writing ability, language abilities, structural equation modelling

R. Schoonen, A. van Gelderen, K. de Glopper, J. Hulstijn, P. Snellings, A. Simis, & M. Ste- venson (2002). Linguistic knowledge, metacognitive knowledge and retrieval speed in L1, L2, and EFL writing. In: G. Rijlaarsdam (Series ed.) & S. Ransdell & M. Barbier (Volume eds.), Studies in Writing: Volume 11: New Directions for Research in L2 Writing, 101 – 122. © 2002. Kluwer Academic Publishers. Printed in the Netherlands. 102 SCHOONEN, ET AL.

1 INTRODUCTION Almost everywhere in the world, school education at elementary and secondary level is provided in the designated official language of the country or region. For large numbers of students, this is their native, or first, language – the language with which they grew up at home before they went to school. For a considerable number of students, however, the language of schooling is not their first language. In the Netherlands, as in most countries of the European Union, the school population consists of a majority of students for whom the national language is their native language, plus a considerable minority of students who are non-native speakers of the national language. Many of these students either migrated with their parents to the Netherlands at an early age, or were born in the Netherlands, but come from migrant families. In the remainder of this chapter, we will refer to the former group of students with the term ‘native speakers of Dutch’ or ‘native Dutch speaking’ (NS), and to the latter group with the term ‘non-native speakers of Dutch’ or ‘non-native Dutch speaking’ (NNS).1 Obviously, exposure to oral, colloquial Dutch differs between the two groups, whereas exposure to written Dutch may differ far less. In the case of writing, differences in writing experience between NS and NNS students may be smaller, as young people tend to write, when they write at all, mainly in the context of school assignments and relatively little in other contexts. However, as writing also calls upon language knowledge acquired orally (see below), the aforementioned differences in oral language proficiency may also show up in writing. Research in the Netherlands and many similar European countries has shown that NNS students lag behind their NS peers in school success (Vallen & Stijnen, 1991). One of the main causes for this difference in educational performance probably is a poorer command of Dutch, especially in the breadth and depth of their vocabulary knowledge (Appel & Vermeer, 1998; Verhallen & Schoonen, 1993, respectively). Little is known, however, whether and, if so, to what extent NNS students fall behind their NS peers in their Dutch writing ability. A recent report of a national assessment at the end of primary education showed mixed results. Of the sixteen assignments that were administered, two showed large differences between NS and NNS writers. Five assignments showed differences of medium effect size, six showed small effects and three showed no differences (Sijtstra, 1997). In learning a foreign language, the NS students probably experience the same difficulties as the NNS students, because the foreign language is ‘new’ to both groups. By foreign languages we mean languages that are taught in the educational system as true non-indigenous, foreign languages. For example, English, French and German are the three compulsory foreign languages in Dutch secondary schools. There is indeed some evidence that suggests that NNS and NS students may not differ in their performance in foreign languages (Sanders & Meijers, 1995).

1 We use these not entirely adequate terms for want of better ones. We will not refer to these students as majority or minority students for two reasons. First, these latter terms refer to numbers rather than to linguistic background. Second, in many urban schools, the majority of the students consists of non-native speakers of Dutch. KNOWLEDGE AND SPEED IN L1, L2 AND EFL WRITING 103

The present study compares the writing performance in Dutch and English of 13- /14-year old NS and NNS students in terms of a number of types of knowledge and skills. The aim of the study is to investigate the relative weights of these knowledge and skill components in the following three constellations: (1) in Dutch writing when Dutch is a native, and hence relatively strong language, (2) in Dutch writing when Dutch is a non-native, and hence potentially weaker language, and (3) in Eng- lish, the language to which both NS and NNS students have been exposed during the same relatively short period in their school careers. The knowledge and skill components in our investigation fall into three categories: (1) knowledge of language (vocabulary, grammar, and orthographic knowledge), (2) the speed with which language knowledge can be retrieved (word retrieval and sentence building), and (3) metacognitive knowledge of text characteristics and reading and writing strategies (in Dutch and English as a foreign language). In the following subsections, we will address the componential nature of writing abilities in general and of writing skills in L1 and L2 in particular. We will round off the introduction with our research questions.

1.1 The componential nature of writing ability in general Writing calls upon several language skills, as well as upon more general (meta) cognitive skills. To our knowledge, a generally accepted model for writing ‘subskills’ does not exist (see, however, Abbott & Berninger (1993) and Grabe & Kaplan (1996) for an attempt to arrive at such a model). Most writing models are process models (Flower & Hayes, 1980; 1983; Hayes, 1996; Kellogg, 1996) or developmental models (Bereiter & Scardamalia, 1987). But which components might make up the construct of writing ability? First of all, writers need to have some degree of vocabulary knowledge of the language in which they are writing (cf. Grabe & Kaplan, 1996). Lexical knowledge or vocabulary size may well influence the quality of the text to be written. Measures of lexical richness of texts correlate substantially with holistic ratings by readers of the text (Engber, 1995). In a study by Laufer and Nation (1995) it was shown that vocabulary size, use of words of different frequency bands (Lexical Frequency Pro- file) and composition rating are highly intercorrelated. The writer’s ideas are not just expressed in single words, but need to be cast in grammatical structures. Therefore, the writer needs to have some grammatical knowledge available (cf. Grabe & Kaplan, 1996). Furthermore, writing, compared to speaking, also requires knowledge of the orthography of the language, i.e., spelling (Abbott & Berninger, 1993). We will refer to knowledge of vocabulary, grammar and orthography as language or linguistic knowledge. Writing is a laborious, demanding task as writers must be familiar with the organization of texts, the intended readers, and the way texts function in a community (Torrance & Jeffery, 1999). Furthermore, writers must find ways to deal with all these constraints simultaneously (‘juggling with constraints’, cf. Flower & Hayes (1980)). Schoonen and De Glopper (1996) showed that proficient (L1)writers have more (declarative) knowledge about writing than poorer writers, and that they have a 104 SCHOONEN, ET AL. different perception of what is important for a text to be adequate; the proficient writers focused more on text organization compared to the poor writers who focused on mechanics and layout. Victori (1999) showed that successful and unsuccessful EFL-writers could be distinguished by their metacognitive knowledge in each of three domains: knowledge one holds about oneself as cognitive processor, task knowledge and strategy knowledge. Thus, it also seems to be important to have sufficient (metacognitive) knowledge about texts and about strategies to deal adequately with texts. We will use ‘metacognitive knowledge’ to refer to knowledge about text characteristics and (reading and) writing processes and strategies. However, having this metacognitive and linguistic knowledge does not guarantee proficient writing. Writers must also be able to access and activate these knowledge resources more or less simultaneously, which will overburden their short-term working memory. Therefore, to reach a certain level of proficiency, writers must find ways to deal with these multiple sources of knowledge within their cognitive constraints (cf. McCutchen, 2000). In recent years, writing research has included research into the role of cognitive capacity and working memory (cf. Benton, Kraft, Glover & Plake, 1984; Fayol 1999; Kellogg, 1999; Lea & Levy, 1999; Levy & Marek, 1999; McCutchen, Covill, Hoyne, & Mildes, 1994; McCutchen, 1996 and Ransdell & Levy, 1999), and from this research it can be inferred that it is not enough to have language knowledge available while writing; the writer most also be able to apply this knowledge fluently (see for an extensive discussion McCutchen, 1996). Measures of working memory capacity and resource accessibility are related to writing fluency and writing quality, and there is plenty of evidence that constraining working memory capacity during writing by requiring writers to carry out a simultaneous secondary task affects writing fluency and (to a lesser extent) writing quality. This cognitive psychological research with experimental tasks, e.g. reading and writing span measures, has been confined to small-scale experimental studies. In (large scale) correlational research such measurements have been largely ignored (for understandable reasons). However, fluent access to words and phrases or grammatical structures in memory2 may lower the cognitive processing load for a writer and may thus enhance the writing process and possibly the quality of the written text (cf. Cumming, 2001; Grabe & Kaplan, 1996; McCutchen, 1996; Pennin- groth & Rosenberg, 1995). So, one may assume that fluent lexical retrieval and sentence structuring are also constituent subskills of writing. The ‘model’ of writing proficiency which we present here is a simplification of a very complex construct (see for more extensive analyses of writing and theories about writing, Grabe & Kaplan, 1996; Grabe, 2001), but it addresses three different components of knowledge and skills which we think are relevant for writing proficiency: linguistic knowledge (vocabulary, grammar, and orthographic knowledge), metacognitive knowledge (of text characteristics and reading and writing strategies), and the speed with which linguistic knowledge can be retrieved (word retrieval and sentence building).

2 How this reduction of cognitive load exactly is achieved remains to be investigated. See McCutchen (2000) for a discussion about the roles of short-term working memory and long- term working memory (cf. Kintsch 1998) in writing. KNOWLEDGE AND SPEED IN L1, L2 AND EFL WRITING 105

1.2 The differential influences of components for NS and NNS writers For L1 writers, the availability of words and grammatical structures may be fairly highly automatized and therefore less discriminative between good and poor writers (see Stanovich (1991) for similar findings with word recognition in reading). Thus, for the NS students in our study, we expect a less substantial weight of the fluency measures (lexical retrieval and sentence building) than for language knowledge components in their Dutch L1 writing. It is not well known whether the NNS writers have had less exposure to written Dutch and less experience in writing Dutch during their school careers, but they have had less exposure to (oral) Dutch than their NS peers. Thus, when writing in Dutch, our NNS students may have greater problems than their NS peers in retrieving lexical and grammatical information from memory, which may cause ‘cognitive overload’ during writing at the expense of the quality of the text. This may be reflected in a larger contribution of the fluency measures to the quality of their writing. In contrast, writing in English is writing in a foreign language for both the NS and the NNS writers. Therefore, one might expect few differences between the two groups, i.e. English writing is expected to appeal to the components of knowledge and fluency in a similar pattern for both groups. Both groups are expected to suffer from similar problems and shortcomings with respect to their linguistic EFL knowledge (vocabulary, grammar and orthography) and fluency or accessibility of this linguistic knowledge (word retrieval and sentence building). Individual differences in these subskills should relate to individual differences in English writing proficiency. For all students in our study we expect that they are more involved in the ‘lower level’ problems of quick retrieval of lexical, grammatical and orthographic knowledge when writing in their weaker language English than when writing in their stronger language Dutch. When engaged too much in knowledge retrieval processes, they may not have sufficient cognitive resources left to deal with ‘higher level’ aspects of writing. To describe similar phenomena in L2 reading, a linguistic threshold hypothesis was formulated by Alderson (1984) among others: ‘Poor foreign language reading is due to reading strategies in the first language not being employed in the foreign language, due to inadequate knowledge of the foreign language. Good first-language readers will read well in the foreign language once they have passed a threshold of foreign language ability’ (Alderson, 1984: 4). Our expectation is that a similar kind of threshold may apply to writing. Some empirical evidence for this expectation was provided by Jones and Tetroe (1987), who studied the interaction between composition skills and (second) language competence in what they call a ‘natural experiment’: writing in a second language. The study focused on planning strategies of six native Spanish adult ESL writers. Par- ticipants wrote essays in Spanish L1 and English L2 while thinking aloud. In addition, they performed (thinking aloud) a more constrained paragraph writing task in which the final sentence was given. This more constrained writing task was intended to affect a writer’s planning behaviour. The study’s basic finding was that the normal planning processes in Spanish L1 and English L2 were generally speaking the same. In the more constrained experimental task there was an increase in the amount of planning behaviour in comparison to the more natural composition task. The au- 106 SCHOONEN, ET AL. thors studied to what extent writers kept to their plans and the constraints of the tasks. The ‘success rate’ was found to be much lower in L2 than in L1 writing. This led the researchers to conclude that not only were the plans in L2 not as fully developed as in L1, but writers also had more difficulty in keeping track of their plan. ‘We can conclude that there is some decrease in performance simply due to the fact that it is in a second language, that working in an unfamiliar language does take up cognitive capacity that would be used for other tasks, such as monitoring and revising the plan, in first-language composing’. (o.c.: 53) Sasaki and Hirose (1996) developed a model for writing in English as a foreign language (EFL), based on their research into the EFL writing of Japanese students. In their study, there appeared to be an interplay between EFL proficiency, L1 writing ability, strategic knowledge and writing experience/education. The dependent variables in this study were the scores for a writing assignment, performed once in L1 and once in FL. EFL proficiency turned out to be the major predictor of EFL writing, with minor roles for metacognitive knowledge and L1 writing ability. (Total R² = 54.5%, unique contributions: 32.6, 0.3 and 1.5% variance explained, respectively).3 From their qualitative think-aloud data Sasaki and Hirose learned that good (EFL) writers paid more attention to the overall organization of the text than poor writers. Furthermore, good writers seemed to be more fluent (i.e., they wrote longer texts, did less mental translation from L1 to L2, and did not pause longer while writing). Sasaki and Hirose propose a threshold as an explanation for the relatively small contribution of L1 writing. ‘Probably the weak writers were tied up with word- or sentence-level processing (in other words, the ‘what next strategy’ cited in Bereiter & Scardamalia, 1987), and could not afford to think about overall organization much. ’ (o.c.: 158). In sum, the kind of writing processes that L1 and L2 writers use, can be considered to be basically the same (see also Cumming, 2001; Silva, 1993; Uzawa, 1996), but the effectiveness of the writing processes may differ (Silva, 1993) and the distribution of cognitive resources over these processes may differ between L1 and L2 writers (Cumming, 2001), as well as the level (‘pragmatic’, ‘textual’ or ‘linguistic’) on which the processes are focused (cf. Whalen & Ménard, 1995).

1.3 Research questions The above-mentioned studies of Jones and Tetroe (1987) and of Sasaki and Hirose (1996) focused on the role of lower level skills in interaction with higher level planning behaviour in L2 writing compared to L1 writing, but they did not take into account the speed or ease with which linguistic knowledge can be accessed during writing in L1 or L2. However, fluent (i.e. effortless) access of linguistic knowledge might be relevant for proficient writing; it may decrease the burden on short-term working memory and save cognitive capacity for higher-level behaviours necessary in writing, such as text organizing. Our study is the first one, to our knowledge, to combine measures of speed of access to linguistic knowledge with linguistic knowl-

3 In their pilot study, however, the authors came to a somewhat different result. KNOWLEDGE AND SPEED IN L1, L2 AND EFL WRITING 107 edge and metacognitive knowledge scores. It investigates the differential weights of three types of components in the prediction of writing performance in Dutch and English for NS and NNS Dutch school children: (1) language knowledge (vocabulary, grammar, and orthography), (2) speed of knowledge retrieval (word retrieval and sentence building), and (3) metacognitive knowledge of text characteristics and reading and writing strategies. We expect the NNS students to have a lower level of Dutch linguistic knowledge (i.e. vocabulary and grammar) than their NS peers due to their less extensive exposure to Dutch in general. Similar differences can be expected with respect to the fluency in lexical retrieval and the availability of grammatical structures. As exposure to written texts and experience in Dutch composition writing may not be substantially different for the two students groups (because both activities are induced by the educational curriculum, which is essentially identical to both groups), we may not find group differences in their orthographic knowledge and their metacognitive knowledge of texts characteristics and reading and writing strategies. However, there may be differences in the quality of their Dutch compositions to the extent that these are induced by differences in their vocabulary and grammatical knowledge, and fluency of lexical retrieval and the availability of structures. With respect to the ‘decomposition’ of writing proficiency, we expect that the weights of the contributions of the three types of predictor variables will differ in that the contribution of the knowledge and speed variables will be smaller for the NS students than for the NNS students, since we may expect that the Dutch students have surpassed a certain level of linguistic knowledge and especially fluency beyond which individual differences matter less to an adequate writing performance. Thus, these variables will be less discriminative among the Dutch students. As for the English linguistic knowledge, fluency and writing, we expect no differences between the two groups, because both groups have had more or less the same (kind of) exposure to and experience with oral and written English. As a consequence, the contribution of the knowledge and speed variables to the explanation of English writing proficiency will be similar for both student groups. In the present study we aim to explore these hypotheses, i.e. that English writing calls upon the same subskills for the NS and the NNS group of writers, and that Dutch writing may be different across the two groups.

2 STUDY 2.1 Participants Nearly 400 Dutch grade 8 students (i.e., in their second year of secondary education) participated in a longitudinal study (NELSON) into the development of reading and writing proficiency in Dutch and in EFL. For this study eight secondary schools from the western part of the Netherlands were recruited for participation. It was known that at least four of these had a substantial population of NNS students. Stu- dents were sampled from all tracks in the Dutch educational system, from vocational to pre-academic. These students had received on average 3.5 years of education in English as a foreign language; 1.5 year at the level of secondary education and two 108 SCHOONEN, ET AL. years at the level of primary education. In primary education attention is only paid to basic oral communication skills. Students’ language development will be traced in three successive years (1999- 2001). In this chapter, we present data from the first round of measurement, collected in 1999. We confine ourselves to students who participated in all tests relevant for the following analyses and who identified themselves as speakers of Dutch as a first language (N = 281) or as speakers of Dutch as a second language (N = 57). Students who reported speaking a language other than (or in addition to) Dutch with their parents, and who acquired a language different from Dutch as their first language were considered non-native Dutch. A greater part of this NNS group comes from either a Turkish, Moroccan or Surinam background. Students who could not unambiguously be identified as NS or NNS were excluded from these analyses (N = 59).

2.2 Procedure Trained test assistants collected the data during school hours. Tests were assigned in a quasi-random order, taking into account both the availability of the laptop computers that were used in the data collection of the speed tests and the timetable of the schools.

2.3 Instruments Our hypotheses concern the following variables: Writing Proficiency, Vocabulary Knowledge, Grammatical Knowledge, Orthographic Knowledge, Metacognitive Knowledge, Speed of Word Retrieval and Speed of Sentence Building. Except for Metacognitive Knowledge, all tests had an English and a Dutch version. However, these tests cannot be considered equivalent across languages, as the difficulty level was adjusted to the expected level of the students. Below we briefly describe the format of each test and the number of items. All tests were piloted to assess the quality of the instructions, the time limits and the quality of the items. Where necessary the tests were adjusted. The resulting tests are described below.

2.3.1 Writing proficiency Students wrote three texts in each language. Van den Bergh, De Glopper & Schoonen (1988) have estimated that depending on the kind of assignment, the grade level and the text characteristics to be rated, at least two or three writing tasks should be assigned. In many cases, more than five need to be assigned to attain a satisfactory score reliability (i.e., .80). However, considering the large test battery the students had to complete, three writing assignments per language was the maximum number that was feasible. Text types were held constant across the two languages: a personal (descriptive) letter, an argumentative letter and a descriptive text. The writing prompts were KNOWLEDGE AND SPEED IN L1, L2 AND EFL WRITING 109 matched as far as possible. In the writing assignment the rhetorical situation and intended audience were specified. The Dutch writing prompts were: • Write to a friend who emigrated about what you plan to do when he visits you and stays at your place (personal descriptive letter); • Write to a television network to complain about their pulling the plug on your favourite soap (argumentative letter); • Describe your Dutch language lessons for a school magazine (descriptive essay). The English writing prompts were: • Write to an English pen pal who is moving to your home town (personal descriptive letter); • Write to an English music magazine to complain about their ignoring your favourite pop group (argumentative letter); • Describe your English language lessons for an English school magazine (descriptive essay). Three (different) panels of two independent raters were involved in the rating of the texts. Each panel rated an assignment type, i.e. both the Dutch texts and the English ‘counterparts’. The raters rated according to a primary trait instruction, i.e. they had to give a general rating of the quality of the text considering its primary trait, e.g. ‘how well or convincing is this text as a letter to a television network to complain about their pulling the plug on one’s favourite soap’. Raters were instructed to concentrate on content and organization, and to consider linguistic errors only as far as they might hinder the persuasiveness of the text. Besides these general instructions the raters were provided with benchmark texts as a reference. Five benchmarks were selected in a separate analysis according to a scaling procedure (cf. Blok, 1986) and represented a very weak (10th percentile), a weak (25th), an average (50th = score of 100), a good (75th) and a very good text (90th). Panel reliability (k = 2, Cronbach’s alpha) ranged from .76-.82 (Dutch), and from .87 to .90 (English). The students’ score for Dutch writing consisted of the sum of the three panel means (the Dutch personal letter, argumentative letter and description), so the average score for Dutch is about 300. This procedure was also applied to English writing.

2.3.2 Vocabulary knowledge The vocabulary tests consist of 75 Dutch and 65 English multiple-choice questions. Nouns, verbs, adjectives and adverbs were included in the tests. In the Dutch test we followed the format of Hazenberg and Hulstijn (1996): each item consists of a neutral carrier sentence with a target word in bold print and the students had to choose between four synonyms or paraphrases of the stimulus word. In the English test there were four Dutch translations of the stimulus word. The Dutch word selection was based on Hazenberg and Hulstijn. The English word selection was based on four frequency bands of the COBUILD: the first 2000 words, 2000-3000, 3000- 4000, and 4000-5000. Dutch-English cognates were avoided. The words were not specifically related to the topics of the writing assignments. 110 SCHOONEN, ET AL.

2.3.3 Grammatical knowledge Students had to fill in the correct form of verbs, adjectives, anaphora, comparatives and articles or had to use the correct word order in sentences, taking into account the correct form of number, time, aspect and agreement. The Dutch test consisted of 69 items. The English test consisted of 80 items.

2.3.4 Orthographic knowledge In the Orthographic Knowledge tests students had to choose between two or three concurrent options for the spelling of the missing part of a word. The words were presented in a sentence that strongly pointed to the intended word. The Dutch test consisted of 100 items covering well known spelling problems (single or double consonants, homophones, etc.), including the spelling of conjugated verbs. The Eng- lish test consisted of 89 items which tested well-known spelling problems in English (Castley, 1998). For some items the students had to fill in either of two options (e.g. single or double consonant) (i.e., multiple choice), while for other items they had to decide which consonant(s) or vowel(s) needed to be inserted (i.e., fill-in-the-blanks). In the English test a Dutch translation of the intended word was provided for this latter kind of items, to ensure that the students knew which word they were required to spell.

2.3.5 Metacognitive knowledge Metacognitive Knowledge was measured by means of a questionnaire consisting of 80 statements about text characteristics, and reading and writing strategies. Students had to tick whether they agreed or disagreed with each statement. All the statements were in Dutch. Statements concerned texts, and reading and writing in general. For example (literal translation from Dutch, the intended answer in italics): The order in which you present the information in your text is usually not relevant. agree/don’t agree

It is wise to keep the organization of the text in mind while writing. agree/don’t agree Some statements applied specifically to English texts, and reading and writing strategies. So far no attempt has been made to split the questionnaire into a reading and a writing part, or a Dutch and an EFL part. The questionnaire as a whole proved to be internally consistent (see below).

2.3.6 Speed of word retrieval Students were presented with 39 (Dutch) or 38 (English) pictures on a laptop screen and were requested to type the first letter of the noun referring to the person or object depicted. The answers and reaction times (RT) were registered. RTs were cor- KNOWLEDGE AND SPEED IN L1, L2 AND EFL WRITING 111 rected for typing fluency. Typing fluency was assessed separately using a task in which the students had to type a letter as quickly as possible after it appeared on the screen (see Note 4). The testing format was the same for Dutch and English and each test started with 10 trial items. Only test items with sufficient hit rates (see scoring) were used to compute a test score (i.e. mean RT). For Dutch, 38 items met this criterion, for English, 18.

2.3.7 Speed of sentence building Students were presented with the beginning of a sentence (43 in Dutch and 44 in English). They then had to choose which of two constituents should be the next to continue the sentence by pressing a key. Each test started with 10 trial items. Only test items with sufficient hit rates (see scoring) were used to compute a test score (i.e. mean RT). For Dutch, 32 items met this criterion, and for English 24 met the criterion.

2.3.8 Scoring Skipped items in the knowledge tests were scored as ‘incorrect’. Missing more than half of the items led to a missing value on the test as a whole. In the speed tests two scores are relevant: RT and accuracy. Since we are solely interested in speed of processing and want to avoid interference from linguistic knowledge, we only used the items with sufficiently high hit rates. For the Word Retrieval task, in which students had to type the first letter of a word, there was a negligible chance for guessing the right answer correctly (1 out 26). The minimum hit rate was set at .75 correct. For the Sentence Building task, in which students had 50% chance of guessing the correct answer, the minimum hit rate was set at .875 (.75 + (.25/2)). Items with hit rates lower than the specified minimum were removed from the data set. Mean RTs for Word Retrieval and Sentence Building were computed across the hits. RTs of (incidental) wrong answers and missing RTs were replaced by estimates according to the EM-algorithm in SPSS (cf. Acock, 1997). Missing test scores (e.g. due to ab- sence of students on one of the testing days) were also estimated according to this procedure.

2.4 Analyses Means and standard deviations were computed for all tests. All tests, except for Writing Proficiency, were split into two parallel parts (according to the Gulliksen procedure). With these two observed (parallel) test scores latent variables could be estimated to perform structural equation modelling (SEM) in LISREL (Jöreskog & Sörbom, 1996a). Reliabilities of the sums of the two test halves and the three writing assignments respectively were estimated within the SEM analyses, i.e. the models 1 to be described in Table 3 and 5, according to the procedure described in Fleishman & Benson (1987). Variables were normalized in PRELIS (Jöreskog & Sörbom, 1996b) and the covariance matrixes of the observed (parallel parts of the) tests were 112 SCHOONEN, ET AL. computed in each group (NS and NNS) as input for the SEM. The hypotheses were tested in multi-group analyses (Jöreskog & Sörbom, 1996a). Analyses were conducted separately for Dutch writing and English writing. In each case, we first tested whether we can assume a similar measurement model in both groups, i.e. whether the relationship between the observed test scores and the postulated latent variables (factors) is equal (models (1)-(3)). This is a necessary precondition to be sure that we measure the same thing in both groups. Next, we tested whether the relationship between the (latent) predictor variables and the (latent) dependent variable writing proficiency is equal across both groups (models (4)-(6)). These hypotheses can be tested by comparing the fit of a model in which the aforementioned equalities exist and that of a model in which they are not postulated. When there is no substantial difference in fit, one may assume that the equality restrictions are valid. The fit of a model to the data can be expressed in several different ways. First of all, the fit can be tested statistically by means of a chi-square test. In the case of a statistically significant chi-square, the model should be rejected. However, with large data sets chi-square is very ‘sensitive’, in the sense that minor deviations from the model lead to a significant chi-square and thus to rejection of the model. There- fore, more descriptive measures of fit should be reported as well, for example the chi-square/df ratio (which preferably is lower than 2), the root mean square error of approximation (RMSEA, which preferably is lower than .05), and the non-normed fit index (NNFI, which should be in the upper .90’s) (Raykov & Marcoulides, 2000). The difference in fit between nested models (such as models (1)-(6)) can be tested statistically by taking the differences in their chi-squares and degrees of freedom. When these new chi-square and df are statistically significant then one can assume that the models differ in fit.

3 RESULTS 3.1 Descriptives Table 1 and 2 contain the descriptive and psychometric statistics per variable respectively, both for Dutch and EFL, and for the NS and NNS students. Table 1 shows that the language proficiency tests are of average difficulty and that there is a reasonable amount of variance in the scores. The L1 and the EFL tests cannot be compared because of different scales (i.e. number of items), however the speed tests use the same reaction time scale of milliseconds. According to these speed tests, it seems that EFL is far less ‘automatized’ than Dutch, both in the NS and NNS group. But we still must be cautious about this comparison, since no attempt was made to equate the Dutch and English stimuli. Furthermore, it seems that, in general, students differ more in their EFL proficiencies than in their Dutch proficiencies (cf. the ratio’s of standard deviations and the corresponding means). KNOWLEDGE AND SPEED IN L1, L2 AND EFL WRITING 113

Table 1. Descriptive statistics for Dutch and English tests: Mean and standard deviation for the NS students and the NNS students. Max. is the maximum possible score for respectively the Dutch and the English tests. NNS = 281, NNNS = 57. Note: Different maximum scores prevent comparison of the means across the languages.

Dutch English

NS NNS NS NNS

Variables M sd M sd M sd M sd

Writing Proficiency (scale mean is 300) 298.8 19.5 293.1 16.1 295.7 24.7 287.8 24.3 Vocabulary Knowledge (max 75; 65) 51.0 9.7 40.0 9.2 51.8 9.8 51.3 10.1 Grammatical Knowledge (max 69; 80) 49.6 8.7 43.5 8.4 43.7 16.2 40.8 15.4 Orthographic Knowledge (max 100; 89) 70.8 11.2 67.4 10.3 48.8 13.1 48.8 12.0 Metacognitive. Knowledge* (max 80) 61.6 7.3 60.6 6.0 Speed Word Retrieval4 (RTs in ms) 3652 574 4171 618 4411 849 5063 782 Speed Sentence Building (RTs in ms) 3876 636 4130 488 4689 928 4879 831

* Test is not language specific.

Note that NNS students lag behind their NS peers in Dutch language knowledge and fluency. As predicted, there are substantial and significant differences in two of the three knowledge tests (Vocabulary and Grammar) and the two speed tests. Differ- ences between the NS students and the NNS students in tests of written language (Orthographic Knowledge, Metacognitive Knowledge and Writing Proficiency) are rather small, though statistically significant in the case of Orthographic Knowledge and Writing Proficiency. In the Introduction, we expressed our uncertainty as to whether to expect a difference in writing because NS and NNS students may have had the same amount of experience in writing, but differ in their oral language skills. English is a foreign language for both groups, and differences in EFL, although they do exist, are rather small (statistically significant in the case of Writing Profi- ciency and Word Retrieval, but this latter difference concerns the uncorrected Word Retrieval RTs and therefore may be attributed to differences in typing speed) 5. Table 2 presents some reliability estimates for our measures. We estimated the reliability of the total scores of the tests as a whole, and of the sum of the three writing assignment scores based on the regressions of the observed scores on the latent

4 The Word Retrieval response times were corrected for typing speed in the subsequent SEM analyses: Typing speed was partialled out and the residuals were used as indicators of Word Retrieval speed. In both groups the mean residual is (by definition) 0. 5 These findings with observed differences are largely confirmed by an analysis of the means on the latent variables in the subsequent SEM analyses. The two groups differed significantly in their Dutch Vocabulary and Grammatical Knowledge, and their Sentence Building speed, but also in Orthographic Knowledge; for English there is a (just) significant difference in Writing Proficiency (t=1,96). Word retrieval differences could not be tested since we used residuals in our SEM, which are by definition 0 in each group (see note 4). 114 SCHOONEN, ET AL. variables and the error variances as depicted in Figure 1 (cf. Fleishman & Benson, 1987). Reliabilities were estimated under model 1 of Table 3 (Dutch) and Table 5 (English). Table 2 shows that, in general, all tests show satisfactory reliability estimates. Most estimates are in the .80’s and .90’s. The estimates for Dutch Writing are rather low in both groups, which is possibly not very surprising given the known effect of topic and assignment in writing assessment (Cooper, 1984, Schoonen & De Glopper, 1999). The estimate for Writing English is satisfactory. The reliability estimates for Speed of Word Retrieval are also not very high. These measures are based on residuals of RTs in the word retrieval tasks with the Typing Speed partialled out, and are therefore probably less reliable (see also Note 4). In general, the English tests show a somewhat higher reliability than the Dutch tests, but this may be due to the larger amount of variance in the English tests. At this point it is important to stress that in the subsequent SEM analyses we will be dealing with the latent variables from which measurement error is partialled out and in that sense should not influence the correlations or regression weights in the structural part of the model.

Table 2. Reliability estimates for Dutch and English tests for the NS and NNS group: NNS = 281, NNNS = 57.

Dutch English NS NNS NS NNS

Writing .63 .70 .78 .84 Vocabulary Knowledge .89 .80 .90 .92 Grammatical Knowledge .87 .78 .95 .96 Orthographic Knowledge .88 .88 .92 .92 Metacognitive knowledge* .81 .78 Speed of Word Retrieval .79 .72 .74 .68 Speed of Sentence Building .95 .91 .92 .92

*Test is not language specific.

3.2 Writing in Dutch In the Introduction, we pointed out that although the quality of the Dutch texts produced by NS and NNS students may not differ substantially, we expect the process of writing to differ, in that the NNS students may suffer more than their NS peers from cognitive overload due to a smaller knowledge base (Vocabulary, Grammar, and as it turned out Orthographic Knowledge) and a slower processing rate (Speed of Word Retrieval and Sentence Building). Thus, for the Dutch data, we expect a difference in the weights of the predictor variables on the dependent variable (Writ- ing Proficiency) between the two student groups. To test this hypothesis we will perform a multiple group analysis in LISREL. As was mentioned, to make a ‘fair’ KNOWLEDGE AND SPEED IN L1, L2 AND EFL WRITING 115 and valid comparison between the two regression models in both groups, it is important to be sure that the observed test scores show identical relations with the latent variables they measure, i.e. the ‘measurement model’ should be similar (cf. Figure 1). We will then test whether the structural model (i.e. the relationships between the latent variables) is equal in both groups. When we have selected a model that describes our data best, we will give an account of the regression weights (and the corresponding correlations) as estimated in the selected model.

Metacognitive Knowledge 1

Metacognitive Knowledge 2 Meta cognitive Know ledge

Vocabulary Knowledge 1 Task 1 Vocab ulary Vocabulary Know ledge Knowledge 2

Grammatical Knowledge 1 Task 2 Gra m matical Writing Grammatical Knowledge Proficiency Knowledge 2

Orthographic Knowledge 1 Task 3 Orth ographic Knowledge Orthographic Knowledge 2

Word Word Retrieval 1 Retrieval Word Retrieval 2 Sentence Sentence Building 1 Building Sentence Building 2

Figure 1. Basic model fitted in both groups, NS and NNS; ovals refer to latent variables, squares are observed variables, i.e. the split parts of the tests and the writing assignments. 116 SCHOONEN, ET AL.

3.2.1 Model fit: Writing in Dutch First, the model as depicted in Figure 1 was fitted (simultaneously) to the covariance matrix of the Dutch variables in both groups, NS and NNS writers. No restrictions with respect to the equality of the regressions in the measurement model were implemented (model (1)), and second, we fitted a model (model (2)) with the restriction of equal regressions across the groups of the observed variables on the corresponding latent variable. This implies that the tests (and assignments) measure the same amount of ‘true’ score variance, but still may differ in the amount of error variance they measure, and thus in reliability. Thirdly, we also imposed an additional restriction of equal measurement errors in both groups (i.e. ‘external’ arrows to the observed variables); this is model (3). In this model it is assumed that tests are equally reliable in both groups (i.e. the same of true score variance and error variance). The fit indices can be found in Table 3, as well as the difference in fit between subsequent models. It shows that going from model (1) to model (2) the fit hardly decreases, going from model (2) to model (3) the fit decreases slightly, but significantly (χ2 = 33.8, df = 15, p = .004). A closer look at the data showed that two error terms differed between the two groups. In the subsequent analyses these two error terms are allowed to differ across the two groups (model (3')). In general, we can conclude that the measurement models in both groups are similar, and this war- rants further analyses of the structural model, i.e. the relations between the latent variables.

Table 3. Model Fit Indices for Writing in Dutch: Chi-square (χ2), the degrees of freedom (df), Root Mean Square Error of Approximation (RMSEA) and Non-Normed Fit Index (NNFI); NNS = 281, NNNS = 57. Subsequent models compared.

Model fit indices Comparison

Model restrictions χ2 df RMSEA NNFI χ2 df

1. No across groups restrictions 184.1 138 .038 .97 n.a. n.a. 2. Regressions on latent variables equal 192.9 146 .036 .98 8.8 8 3. Measurement error terms equal 226.7 161 .052 .97 33.8 15 3’. Measurement error terms equal, except for two 206.3 159 .035 .98 13.4 13 4. Structural regressions equal 219.5 165 .038 .98 13.3 6 5. Correlations of the predictors equal 250.0 186 .040 .97 30.5 21 6. Amount of unexplained variance equal 250.1 187 .039 .97 .0 1

Underlined χ2 are statistically significant at p = .05.

In the next step, we tested whether the regressions of the latent variable Writing in Dutch on the latent predictor variables can be considered equal in both groups. Our expectation is that this is not the case. A model with such an equality constraint (model (4)) fits the data slightly worse than the model with no such constraint KNOWLEDGE AND SPEED IN L1, L2 AND EFL WRITING 117

(model (3')). Although this difference in fit is statistically significant (χ2 = 13.3, df = 6, p = .039), it is rather small. Considering the other fit indices (RMSEA and NNFI) of the two models, (3') and (4), the fit has not or hardly deteriorated. To further examine to what extent the interrelations between the latent variables is equal between the two groups, we fitted two additional models (with additional constraints) to the data. First, we tested whether we can assume that the intercorrelations of the latent predictor variables are equal across the two groups (model 5), and secondly we tested whether the amount of variance that remains unexplained in the dependent latent variable Writing in Dutch is the same across the two groups (model 6). Altogether, it turns out that – contrary to our expectations – there are no substantial differences in the structural model for the relations between the NS and the NNS group. Although model (6) should be rejected statistically speaking (χ2 = 250.1, df = 187, p = .001), it proves to be a fairly good description of the data, i.e. the χ2 /df ratio is lower than 2 (1.3), the RMSEA is lower than .05 (.039) and the NNFI is in the upper nineties (.97); all criteria for good model fit (cf. Raykov & Marcoulides, 2000)

3.2.2 The regression of writing in Dutch on subskills In Table 4 we present the actual (standardized) regression weights for Writing in Dutch, as estimated in model 6. For a full appreciation of the relationship of the independent and dependent (latent) variables the bivariate correlations are also included in this table.

Table 4. Correlations (r) and standardized regression weights (ß) for Writing in Dutch and the independent variables, and the percentage of variance explained as estimated under model 6; NNS = 281, NNNS = 57.

Vocabulary Grammatical Orthographic Metacognitive Word Sentence Percentage Knowledge Knowledge Knowledge Knowledge Retrieval Building explained

r .45 .66 .54 .61 -.19 -.47 ß -.25 .45 .10 .38 -.02 -.09 51

Underlined regression weights are statistically significant at p =.05.

Table 4 shows that all independent variables, except for Speed of Word Retrieval, correlate substantially with Writing in Dutch; the two negative correlations for the speed measures are as expected since these are RT measures. The regression weights, however, show a quite different pattern. Vocabulary Knowledge’s regression weight has become negative; this must be due to a suppressor effect (cf. Guil- ford, 1973), i.e. Vocabulary Knowledge correlates relatively low with Writing in Dutch and relatively high with the other two major predictors Metacognitive and Grammatical Knowledge (r > .70). 118 SCHOONEN, ET AL.

The speed variables have no unique contribution to the prediction of Writing in Dutch. Recall that this pattern is equal for both groups, so even for the NNS students the speed or fluency measures have no unique contribution to the prediction of Writ- ing Proficiency as rated in this study. Grammatical and Metacognitive Knowledge have the largest contributions to the prediction of Writing in Dutch. Altogether the variables explain 51% of the variance in Dutch Writing Proficiency. A closer look at the results shows that the aforementioned small decrease in fit, going from model (3') to (4) was mainly caused by the regression of Writing on Metacognitive Knowledge and Orthographic Knowledge; these regression weights tend to be slightly higher in the NS group than in the NNS group.

3.3 Writing in English as a Foreign Language We followed the same procedures in analysing Writing in EFL. In EFL writing, we expect no major differences between the NS and NNS students.

3.3.1 Model fit: Writing in EFL

Table 5. Model fit indices for Writing in EFL: Chi-square (χ2) and the degrees of freedom (df), Root Mean Square Error of Approximation (RMSEA) and Non-Normed Fit Index (NNFI); NNS = 281, NNNS = 57. Subsequent models compared.

Model fit indices Comparison

Model restrictions χ2 df RMSEA NNFI χ2 df

1. No across groups restrictions 195.5 138 .043 .98 n.a. n.a. 2. Regressions on latent variables equal 204.8 146 .041 .98 9.3 8 3. Measurement error terms equal 223.8 161 .043 .98 18.9 15 4. Structural regressions equal 225.7 167 .040 .98 1.9 6 5. Correlations of the predictors equal 250.9 188 .042 .98 25.2 21 6. Amount of unexplained variance equal 253.0 189 .043 .98 2.1 1

Underlined χ2 are statistically significant at p = .05.

Without loss of model fit we can assume that the measurement model for EFL writing is identical in the two groups (compare model (2) with model (1), and model (3) with model (2)). The same goes for the structural regression model: the regression weights, correlations and the amount of unexplained variance can be considered identical in the two groups, as well. In this respect our expectation of no between group differences in EFL writing is confirmed. Compared to the former analyses, imposing equality constraints on the English data of both groups seems to affect the fit of the model slightly less than in the case of the Dutch data; one might conclude KNOWLEDGE AND SPEED IN L1, L2 AND EFL WRITING 119

that the relations between the variables concerning Writing in EFL are more alike in the two groups than the relations are between the variables concerning Writing in Dutch. This is as expected. However, the differences between the groups in Dutch writing are still very small.

3.3.2 The regression of writing in EFL on the subskills The actual (standardized) regression weights, with the corresponding bivariate correlations, for Writing in EFL as a dependent variable and the English predictor variables as estimated under model 6, are presented in Table 6.

Table 6. Correlations (r) and standardized regression weights (ß) for Writing in EFL and the independent variables, and the percentage of variance explained, as estimated under model 6; NNS = 281, NNNS = 57.

Vocabulary Grammatical Orthographic Metacognitive Word Sentence Precentage Knowledge Knowledge Knowledge Knowledge Retrieval Building explained

r .57 .81 .80 .72 -.37 -.59 ß -.02 .39 .27 .22 -.03 -.09 75

Underlined regression weights are statistically significant at p = .05.

Table 6 shows that in EFL all independent variables correlate substantially with Writing, and higher than the corresponding variables did with Writing in Dutch. The regression weights show more or less the same pattern as in the analysis of Writing in Dutch. However, Orthographic Knowledge now makes a statistically significant contribution to Writing in English, but did not so to Writing in Dutch. The regression weight of Vocabulary Knowledge can in this case be considered 0. Again, the speed variables have no unique contribution to the prediction of Writing in EFL; Speed of Sentence Building just misses statistical significance (t = -1.86). Writing in EFL is far better predicted by the independent variables than was Writing in Dutch; percentage of variance explained of Writing in English is 75%, compared to Writing in Dutch 51%.

4 DISCUSSION AND CONCLUSIONS Our hypotheses are not fully confirmed. Although statistically significant, the difference in Dutch Writing Proficiency between the NS and NNS students is rather small (Table 1). Differences in language abilities and speed of access are much larger. However, our main research questions related to the structural relations between writing as a complex construct and its constituent subskills. Writing in English had the same structural relations with its component skills in both groups, as was expected, but we had not expected to find that the same would be the case for Dutch 120 SCHOONEN, ET AL.

Writing Proficiency. Three explanations come to mind. The first is a statistical one. When no statistical differences can be found, it is important to consider a possible lack of statistical power to show that substantial differences are statistically significant. The NS group (N = 281) seems to be large enough, but the NNS group (N = 57) is relatively small for these kinds of SEM analyses. It is difficult to estimate the statistical power in SEM, because it is difficult to define which differences are large and which are small. For example, could we be dealing with differences in just one parameter (e.g. a regression weight) or with differences in a whole set of parameters. When we confine ourselves to the structural model, i.e. the regression of Writing Proficiency on its constituent variables, we have estimated that we have a statistical power of more than .95 to detect a situation in which all six regression weights in the NNS group are .20 higher than in the NS group. Thus for such situations the statistical power is more than satisfactory. However, more subtle differences, such as when just two predictors (e.g. the speed measures) have a regression weight of .20 higher in the NNS group than in the NS group are much harder to detect. Statistical power is .11 in the analysis of Dutch Writing Proficiency and .33 in the analysis of English Writing Proficiency, which is rather low. So one can only conclude that if any differences exist between the two groups, they are not large enough to show up in the present analyses. A second explanation concerns the status of the NNS group. Although these students come from immigrant families and were raised in a different language from Dutch, most (65%) of these 13-/14-year olds were educated in the Dutch school system and learned to read and write in Dutch; few students (21%) entered Dutch educational system later than grade 3. The exposure to written Dutch and, more importantly, their experience in writing Dutch – activities in which students engage mainly in the context of assignments in the school curriculum, and seldom in non- educational contexts – may have been equally large for most students, on average, as was pointed out in the Introduction. Thus, Dutch ‘literacy’ (in this restricted sense) may have been alike for the two groups. Differences in componential breakdown of Writing Proficiency between NS and NNS students, if existent at all, may have dis- appeared in earlier years. A replication with younger students could corroborate (or falsify) this second explanation. When more of our longitudinal data become available, we will test whether this apparent similarity between the two groups remains. Of course, a third explanation is that what we have found is what is actually the case: the relative weights of the components of writing proficiency are basically the same for all these 13-/14-year old writers, whether they write in Dutch as their first language or as their second language, i.e., their language of instruction and schooling. And, as we argued in the Introduction, no differences between the two groups are expected in EFL writing. Process studies with think-aloud protocol analyses point in the same direction, that is, that the writing processes are basically the same in L1 and L2 writers (see Cumming, 2001 for a review). However, it is entirely possible, that the differential weights of the subskills would have been different if students of lower proficiency levels, not only in English but also in Dutch, had been tested. Our analyses show that the predictive value of the speed or fluency measures is negligible within the context of this set of predictors. Note, however, that the regres- KNOWLEDGE AND SPEED IN L1, L2 AND EFL WRITING 121 sion weights are dependent of the set of predictors. The fluency measures correlate substantially with writing proficiency as is indicated by the bivariate correlations (Table 4 and 6). However, they hardly have an additional explanatory value in predicting writing proficiency in a set of predictors including the knowledge variables (Table 4 and 6). The relatively large number of knowledge variables and substantial correlations between these (linguistic and metacognitive knowledge) variables on the one hand and the speed measures on the other6 make the latter variables seemingly superfluous in explaining writing proficiency, as it was measured in our study. Most experimental studies have shown an effect of writing span measures and related cognitive fluency measures on writing fluency, effects on writing quality are less unequivocal (cf. Ransdell & Levy, 1999). One of the writing assignments in each language was written on a (laptop) computer, and during writing all keystrokes were logged. Future analyses of these log files may show that our speed measures have more (unique) explanatory power in predicting writing fluency rather than in predicting writing quality. It may be somewhat surprising that the lexical variables (Vocabulary Knowledge and Speed of Word Retrieval) seem less important and less related to writing proficiency than the grammatical or morpho-syntactical variables (Grammatical Knowl- edge and Speed of Sentence Building). This is contrary to what is usually found in reading research. A tentative explanation could be that in the initial processes of writing (i.e., generating the content of the message) writers only come up with concepts and propositions that they have words for. This would be a kind of self ‘cen- sorship’ that appears early in the language production process avoiding difficulties at the lexical level. In a study of valid indicators of second language proficiency, Cumming and Mellow (1996) found that their lexical richness measure of compositions was not indicative of second language proficiency, and they suggested that there might be a kind of ceiling in the predictive value of such lexical measures. Similarly, it might be the case that differences in vocabulary knowledge may not manifest themselves in relatively restricted written compositions in which the vocabulary needed may have been too limited. However, formulating sentences inevitably involves grammar and hence calls upon one’s grammatical knowledge, even if it is inadequate or insufficient. Thus, grammar may be a more ‘discriminative’ predictor of the quality of the written text. More detailed (lexical and grammatical) analyses of the texts written by the students could shed more light on this tentative explanation (cf. Cumming & Mellow, 1996), as could experimental or think aloud studies. Both kinds of studies are currently conducted within the NELSON project. Snellings (in prep.) has trained lexical retrieval of a set target words to enhance EFL writing proficiency of students, and Stevenson (in prep.) has collected think aloud protocols along with keystroke logs of both NS and NNS students writing in Dutch and English. In conclusion, we found no essential differences in the pattern of regression weights of Dutch and EFL writing proficiency on the three types of predictor variables (lan-

6 For Dutch Word Retrieval and Sentence Building these correlations with latent knowledge variables range from .17-.36 and .41-.58 respectively. For English these correlations range from .25-.49 and .36-.61. 122 SCHOONEN, ET AL. guage knowledge, retrieval speed, and metacognitive knowledge) between the NS and the NNS writers. The question remains whether there is a threshold level of language knowledge – and perhaps a subsequent threshold for the speed with which language knowledge can be retrieved – below which component skills play a different role. If so, it probably will be at a very low level of L2/EFL language proficiency. It remains to be further investigated at which levels of language knowledge, speed of processing, and metacognition such a threshold might exist and whether we are dealing with one or more thresholds.

ACKNOWLEDGEMENTS The data presented in this chapter come from the first round of measurement in a longitudinal study, called NELSON, into the development of reading and writing proficiency in Dutch as an L1 or L2 and in EFL. General information about the NELSON project is available at www.sco-kohnstamminstituut.uva.nl/nelson. The authors wish to thank Alister Cumming, and the anonymous reviewers for their comments on earlier versions of this chapter. They also want to thank Conor Dolan for his advice on statistical matters. This research was financially supported by the Netherlands Organization for Scientific Research (NWO).

EARLY EXPOSURE TO AN L2 PREDICTS GOOD L1 AS WELL AS GOOD L2 WRITING

M. ROSARIO ARECCO* & SARAH RANSDELL*

*Florida Atlantic University, USA

Abstract. Recent research shows that bilinguals with relatively advanced skills tend to have cognitive advantages compared to their monolingual peers (Bialystok, 1992; Hakuta, 1986; Ransdell, Arecco & Levy, 2001). The research described in this chapter focuses on the relation between the age of first exposure to an L2 and the development of higher cognitive skills, such as skills used in L1 and L2 writing. Spanish-English, Estonian-English and Polish-English bilinguals were divided into two groups according to the age of their first exposure to their L2. Participants in the early L2 group were first exposed to an L2 before the age of twelve. The late L2 group was comprised of participants who were exposed to an L2 at the age of twelve or older. Both groups wrote two 10-minute essays in a counterbalanced design. One essay was in L1 and the other in L2 (English). Compared with the late L2 group, the early L2 group produced better quality writing with more fluency in both L1 and L2. One explanation for the observed differences is that participants in the early L2 exposure group may simply be more advanced in their English language skills. A second possibility is that early exposure to an L2 enhances general cognitive abilities. The present findings that early exposure to an L2 is correlated with better L1, as well as L2 writing fluency and quality, indicates support for the improvement of general cognitive, as well as linguistic skills.

Keywords: bilingualism, critical period, early L2 exposure, cognitive abilities, writing fluency, writing quality.

1 INTRODUCTION L2 studies have shown that cognitive advantage can be linked to bilingual ability (see Hakuta, 1986, for review). Bilinguals have been shown to have advantages over monolinguals in various cognitive tasks, e.g., concept formation, rule-discovery, the allocation of new strategies for specific tasks, verbal originality, divergent thinking, and creative thinking (Hamers, 1996; Nayak, Hansen, Krueger, & McLaughlin, 1990; Ransdell, Arecco & Levy, 2001; Sasaki, 1993). Aronsson (1981) argued that bilingual children develop metalinguistic awareness, the ability to separate language

M. Arecco & S. Ransdell (2002). Early exposure to an L2 predicts good L1 as well as L2 writing. In G. Rijlaarsdam (Series ed.) & S. Ransdell & M. Barbier (Vol. Eds.), Studies in Writ- ing, Volume 11: New Directions for Research in L2 Writing, 123 – 131. © 2002 Kluwer Academic Publishers. Printed in the Netherlands. 124 ARECCO & RANSDELL form from meaning, earlier than their monolingual peers. Bilinguals have the experience of using two different grammars, two different ways of constructing linguistic form and this aids their development. Aronsson found that pre-school bilingual children were more successful than monolingual children when asked to make grammatical corrections and when asked to eliminate syntactic errors. She suggested, as have others, that children become more sensitive to grammatical errors when learning to inhibit the use of one language while using another (Galambos & Goldin-Meadows, 1990). There is also evidence that early exposure to an L2 facilitates many aspects of bilingual development, especially phonological knowledge. The present study investigates the relation between the timing of L2 exposure and subsequent written language production, not only in L2, but also in L1.

1.1 Early L2 exposure and L1 skills As little as two or more years of formal training in a second language has been linked to superior L1 skills (Bialystok, 1992; Thomas, 1988; Mohanty & Babu, 1983). Bialystok (1992) observed that better control of attention to linguistic input is achieved by bilingual children and suggested that selective attention is enhanced by early experience with two languages. She administered a task to comparable mono and bilingual children between three and five years of age that tested conceptualization of number under a perceptually-distracting condition. Children were presented with towers made out of either Lego or Duplo blocks. Lego and Duplo blocks differ in only one aspect. Duplo blocks are eight times the volume; therefore, a tower made out of Duplo blocks will be twice as large. Children were told that the towers represent apartment buildings and that one family lives in each block (floor) regardless if the blocks are big or small. They were also told that the answer was dependent on the number of blocks, not on the height of the towers. Two apartment towers, one made of Lego blocks and the other made of Duplo blocks, were shown to each child. Children were asked to indicate which tower could hold more families. Bilingual children consistently performed better in this task than do monolingual children. The study suggests that the superior performance demonstrated by bilingual children is attributable to early experience with two languages, the learning of language structure for two different linguistic systems, and increased control of attention to linguistic processing associated with secondary language acquisition (Bialystok, 1992). Bialystok (1992) concluded that bilinguals show cognitive advantages over monolinguals, and that these advantages are linked to the various skills developed in the course of learning a second language. Because bilinguals are used to seeing and hearing the names of things in two different languages, they are better prepared to process arbitrariness found in referential forms of language. Bilinguals are accustomed to communicating with different people in different languages, which may facilitate the access to different language forms that have the same meaning. Bilin- guals also experience different customs and social relationships. Using language that is deeply rooted in culture leads to a greater awareness of semantic as well as pragmatic dimensions of word meanings. All of these different experiences potentially give bilinguals knowledge about the structure and function of language earlier than EARLY L2 EXPOSURE AND COGNITIVE ABILITIES 125

L1s (Bialystok, 1992). Bialystok (1988) also found bilingual advantage on tasks requiring a high control of linguistic processing. She attributed the results of her study to the early exposure to dual language systems and the bilingual's frequent attention to formal aspects of language. Vygotsky (1962) stated that because bilinguals are able to express the same thought in different languages they can develop a greater awareness of the arbitrary properties of language than monolingual children.

1.2 Bilingualism and metalinguistic awareness According to Durgunoglu (1997), metalinguistic awareness refers to the attention given to the structural characteristics of a language instead of the utilization of the language for communicative purposes. Mohanty and Babu (1983) attempted to find an effect of bilingualism on metalinguistic abilities in the Kond tribal culture of In- dia. They tested 180 bilingual and monolingual subjects from the same tribal culture and from three different age groups (10-12, 12-14, 14-16). There were 30 bilinguals and 30 monolinguals in each age group. All bilingual subjects were balanced bilinguals, meaning they were highly proficient in both languages, as tested by a translation task. Mohanty and Babu administered two tests to all participants, a nonverbal intelligence test and a metalinguistic ability test to measure three aspects of metalinguistic ability: the understanding of the arbitrariness of language, meaning and ref- erent relationship, and the non-physical nature of words. He found that bilinguals performed better at the three tasks of the metalinguistic abilities test across all age groups. They attributed their findings to the effect that second language knowledge has on thought processes. They stated that bilinguals’ ability to encode thoughts in two languages rather than one enables them to develop a different view towards language and its properties. Mohanty and Babu further cite a study by Ben-Zeev (1977) in which she states that bilingual children develop a different kind of analysis towards language which transfers to other forms of cognitive structures. Nayak, Hansen, Krueger and McLaughlin (1990) stated that multilinguals might have certain skills that may help them in the processing of linguistic stimuli more effectively than monolinguals. They tested 24 multilingual and 24 monolingual adults in tasks involving language- learning strategies using an artificial linguistic system, which consisted of natural language-like grammar structures of varying degrees of complexity. Nayak et al. (1990) found that multilinguals performed better than monolinguals in learning the rules for syntax in an artificial language system. They attribute this finding to the experience multilinguals have with different language systems and their willingness and ability to search for rules. They also found a superior cognitive flexibility effect in multilinguals when trying to learn the rules of the artificial language. Multilin- guals used a greater variety of strategies to learn the new language system than monolinguals suggesting that their superiority in language learning skills could be attributed to their greater flexibility in switching strategies (Nayak et al., 1990). Nayak quotes a study by Ramsay (1980) in which 10 multilingual adults were compared against 10 monolingual adults in learning of a foreign language unknown to both groups (in this study, French). Ramsay (1980) found that multilinguals were 126 ARECCO & RANSDELL

‘successful learners’. They found useful techniques sooner than monolinguals did in learning a new language. For example, multilinguals practiced aloud and verbalized with no hesitation on the mental processes they were using. Thomas (1988) found that English-Spanish bilingual college students showed superiority over English monolingual students when learning French in a formal setting. The bilingual students performed significantly better than monolingual counterparts on tests of vocabulary and grammar. Thomas postulated that this outcome was due to the possibility that bilinguals used their metalinguistic awareness skills to enhance their performance on the assignments focused on language form. Furthermore, bilinguals wrote more understandable essays in French than did the monolinguals. Thomas suggests that metalinguistic awareness in bilinguals facilitated the careful obser- vance of linguistic output focusing their attention on the message. Further analyses by Thomas (1988) revealed that the bilingual students who had studied Spanish in a formal setting for at least two years had advantages over the bilingual students who had learned Spanish informally at home. The former group performed significantly better on the grammar test but not on the vocabulary tests. Thomas (1988) suggests that the experience of language in a formal setting has more impact on the sensitivity to grammar than on the recognition of cognates (vocabulary). With this past research in mind, the goal of the present study is to investigate the relation between early exposure to a second language and writing abilities in L1, as well as in L2. Experience in more than one language may endow the second language learner with a greater awareness of the structure and function of language than monolinguals (Bialystok, 1992). It is predicted that bilinguals who are exposed to a second language at an early age will write better in quality and fluency than bilinguals who were exposed to a second language later in life.

2 METHOD 2.1 Participants The study included 40 Spanish-English bilinguals, 40 Polish-English bilinguals and 20 Estonian-English bilinguals. The Spanish sample was composed of student volunteers from psychology classes at Florida Atlantic University. The Polish sample was composed of psychology students from Florida Atlantic University and adults from the Polish community living in South Florida. The Estonian sample was composed of student volunteers from psychology classes at Tallinn Pedagogical Univer- sity in Estonia. All students, except for some adults from the Polish sample, participated for extra credit in a college class. The average age of participants was 26.7 years. Of the Spanish sample, the average age was 26, and the Polish sample had an average age of 39. Estonian participants had a mean age of 21. The Spanish sample rated themselves as 80% in second language composite skill, which includes rating on how they write, understand, speak and understand written text in a second language (SD = 13.4), the Polish sample rated themselves as 77% in second language composite skill (SD = 15.2) and the Estonian sample rated themselves as 67% in second language composite skill (SD = 18.6). EARLY L2 EXPOSURE AND COGNITIVE ABILITIES 127

All participants were divided in two groups based on a median split depending on the age of first exposure to L2. The participants in the early L2 exposure group (n = 58) were first exposed to a second language before the age of 11, and the participants in the late L2 exposure group (n = 54) were exposed to L2 after the age of 12. In the early L2 exposure group, there were 31 Spanish-English bilinguals, one Pol- ish-English bilingual and 26 Estonian-English bilinguals. In the late L2 exposure group, there were 9 Spanish-English bilinguals, 35 Polish-English bilinguals, and 10 Estonian-English bilinguals. The early L2 exposure group rated themselves as 76% (SD = 15.2) in second language composite skill, comparable to the late L2 exposure group, 72% (SD = 19.1). Exposure age was determined on the basis of a single question asking participants to indicate how old they were when they ‘learned’ their L2.

2.2 Materials Participants were given a self-report of language fluency survey on which they rated their abilities in L1 and in L2 on a 4-point scale with four different dimensions: expressing thoughts in spoken language, understanding written language, understanding spoken language and expressing thoughts in written language. A nonverbal intelligence test was also given, the Cattell Culture-Fair test, which consists of 50 four- choice questions each of which include a series of simple line drawings (Cattell & Cattell, 1963). Estonian, Spanish and Polish translations of the Nelson-Denny Read- ing Comprehension Subtest, form G, (Brown, Fishco, & Hanna, 1993) were created by professional translators. A program called Fauxword (Levy & Ransdell, 1995) measured writing fluency by capturing each keystroke as the participant types and replaying them on request in real or fast time. The program measures words typed per minute by recording complete words created during writing, even those later deleted and calculating a final word count. Typing to dictation is used as a measure of simple words per minute (wpm) and is used to qualify the obtained writing fluency measure.

2.3 Design and procedure Testing occurred in one session. Two 10-minute essays were written, in a counterbalanced design, as single tasks, one in L1 and one in L2. During the session, participants signed consent forms in L1 and then completed these tasks in the following order: the Cattell Culture-Fair test for 20 minutes, and the Nelson-Denny reading comprehension subtest in L1 for 20 minutes. After a 5-minute break, they were asked to write two baseline essays, one in L1 and the other in L2. Dependent variables included writing fluency (words per minute) and the quality scores of the essays. Writing fluency was measured by using Fauxword (see Levy & Ransdell, 1995). Writing quality was based on the essay-sorting task described by Madigan (1991), in which each essay is evaluated holistically and rated based on organization, clarity, and development of ideas. Writing quality scores were derived from the average of two peer raters evaluating holistically each essay in English or an English translation. Each rater evaluated 20 essays at a time and re-evaluated them again one 128 ARECCO & RANSDELL week later. Rating procedure reflected the raters’ own concept of writing quality. Reliability coefficient for the two independent peer raters was r >.80. Intra-raters reliability coefficient was r >.90 for both raters. Raters were blind as to experimental condition. Quality scores of the English translated essays were highly correlated with the quality scores of the same essays prior to being translated, r >.90. The participants were told to write the 10-minute essays as if they were writing for a grade to a professor. A topic card was placed in front of the participant on top of the PC and it was also read aloud. The topics were in the following counterbalanced order: Imagine the best possible college class in an imaginary university with unlimited funds. The other topics took a similar form but were about a college professor, a holiday vacation, a career after graduating from university, and a boyfriend/girlfriend. They were also asked to compare their views to other possible opinions and were told when 5 minutes were remaining, and then when 1 minute was remaining to finish their essays.

3 RESULTS As shown in Table 1, a significant relationship was found between age of first exposure to L2 and writing fluency in L1 and L2. The earlier the age of first exposure to L2, the more fluent the essay writing in L1 and L2.

Table 1: Correlations between age of exposure to L2 and other variables.

1. 2. 3. 4. 5. 6.

1. Age of exposure L2 2. Non verbal intelligence test -.320 3. Reading comp. test in L1 -.305 .155 4. Fluency essay L1 -.570 .117 .231 5. Fluency essay L2 -.377 .123 .201 .773 6. Quality essay L1 -.300 .179 .495 .331 7. Quality essay L2 .250 .215 .125 .301 .564 .480

Underlined correlations are statistically significant at p< .05 level.

There was also a strong relationship among writing measures. Writing measures were positively correlated within and across languages. Writing fluency in L1 was positively correlated with writing fluency in L2. Fluency in writing was consistent in both languages for participants. Similarly, writing quality in L1 was positively correlated with writing quality in L2, in other words, the higher the quality in L1 writing, the higher the quality in L2 writing and vice versa. Moreover, writing quality in L1 was positively correlated with writing fluency in L1 and writing quality in L2 was positively correlated with writing fluency in L2. An ANOVA with age of exposure as a factor revealed significant differences between groups. As shown in Table 2, the early L2 exposure group performed better in the non-verbal intelligence test (M = 50.8, SD = 8.5) than the late L2 exposure group (M = 43.4, SD = 11.5), F (1, EARLY L2 EXPOSURE AND COGNITIVE ABILITIES 129

98) = 13.02, p< .01. Furthermore, the early L2 exposure group performed better in the reading comprehension test in L1 (M = 67.8, SD = 20.1) than the late L2 exposure group (M = 57.1, SD = 16.2), F (1, 85) = 7.67, p < .01.

Table 2: Means (and Standard Deviations) for groups from different conditions.

Early L2 Late L2 Exposure Exposure F p n =58 n = 54

Non Verbal Intelligence Test 50.8 8.5 43.4 11.5 13.02 0.00 Reading Comprehension Test in L1 67.8 20.1 57.1 16.2 7.67 0.01 Self Report Composite Score in L1 98.3 99.9 0.8 4.87 0.03 Self Report Composite Score in L2 75.6 15.2 72.1 19.1 1.20 0.28 Self Report Writing Score In L1 96.5 10.9 99.5 3.4 3.60 0.06 Self Report Writing Score In L2 63.6 21.7 63.7 22.8 0.00 0.98 Fluency essay L1 13.1 4.7 8.3 5.4 25.00 0.00 Fluency essay L2 10.9 4.0 8.2 4.8 10.15 0.00 Quality essay L1 67.7 14.3 57.7 21.4 7.88 0.01 Quality essay L2 55.5 18.1 47.3 23.3 4.22 0.04

Underlined F-values are statistically significant at p < .05 level.

There were also significant differences between groups in writing fluency and quality. The early L2 exposure group wrote more fluently in the L1 essay (M = 13.1 wpm, SD = 4.7) than the late L2 group (M = 8.3 wpm, SD = 5.4), F (1, 107) = 25.0, p<.01. Similar results were found for essays in L2; the early L2 group wrote significantly more fluently in L2 (M = 10.9, SD = 4.0) than the late L2 group (M = 8.2, SD = 4.8), F (1, 105) = 10.15, p< .01. In writing quality, the early L2 exposure group wrote better in L1 (M = 67.7, SD = 14.3), than the late L2 group (M = 57.7, SD = 21.4), F (1, 101) = 7.88, p< .01. Related results were found for L2 writing quality. The early L2 exposure group wrote significantly better in L2 (M = 55.5, SD = 18.1) than the late L2 group (M = 47.3, SD = 23.3), F (1, 105) = 4.22, p< .05. Further- more, as seen in Figure 1, a correlation analysis revealed a negative relationship between writing quality in L1 and age of exposure to L2 (r > -0.30, p < .05) 130 ARECCO & RANSDELL

4 DISCUSSION

100

90 Q U 80 A L 70 I T 60 Y 50 L 1 40

10 0 10 20 30 40 50 AGE EXP to L2

Figure 1: Scatterplot – Correlation between writing quality in L1 and age of exposure to L2.

The purpose of the present study was to investigate the extent to which age of first exposure to a second language is correlated with the fluency and quality of writing in L1 as well as in L2. Bilinguals with three very different language bases were divided into two groups depending on their age of first exposure to a second language. They were given two argumentative essays to write in L1 and L2. The group with an early age of exposure to a second language wrote significantly better in L1 and L2 in quality and fluency than the group who were exposed to L2 at a later age. Correla- tion does not imply causation but the link is compelling, especially in light of the critical period hypothesis. The critical period hypothesis predicts that early exposure to an L2 will lead to stronger abilities in L2, especially phonological skills because of a predisposition and biological readiness to learn sound systems before puberty (Flege, 1999). Cultural and social factors can also make learning a second language easier for children than adults. Children simply have less inhibitions regarding communication. There has also been evidence that, all other things equal, adults actually learn an L2 more readily than children. The present study sheds light on the mechanisms governing early exposure to an L2 and prediction of subsequent L1 skills. If early exposure to L2 affects L1, then general cognitive mechanisms are implied as well as language-specific ones. It is generally assumed that those exposed to an L2 at an early age simply have stronger language-skill specific skills, namely, different degrees of bilingualism. Advanced and highly-skilled bilinguals are usually exposed to a second language at an early age while more weakly-skilled bilinguals are typically exposed to a second EARLY L2 EXPOSURE AND COGNITIVE ABILITIES 131 language at a later age. Ransdell, Arecco and Levy (2001) found that cognitive advantages are much more pronounced in advanced bilinguals. It is also possible that early exposure to a second language may potentiate greater general cognitive abilities beyond purely verbal ones. If general metalinguistic awareness is a necessary condition for greater cognitive abilities in bilinguals, then it should be domain- general in its effects. Metalinguistic awareness encompasses a large number of general domains, including phonological awareness, syntactic awareness, and semantic awareness. Bilinguals learn to suppress an unwanted language in order to speak another depending on the appropriate context, for example, many Spanish-English children living in United States learn to speak English in school and Spanish at home. Successful writers must also possess a high level of divergent thinking skills. Ricciardelli (1992) summarizes 20 different studies in which bilinguals were found to perform better than monolinguals in tasks requiring high levels of divergent thinking. She states that these findings are consistent with other studies where bilingual superiority, e.g., metalinguistic awareness, and concept formation, was shown. Ric- ciardelli (1992) attributes these findings to the bilingual threshold theory developed by Cummins (1976) that states that bilinguals need to acquire a high level of proficiency in a second language to develop better general cognitive abilities than their monolingual peers. The results from the present study contribute to knowledge of links between age of exposure to a second language and superior L2 (English) writing skills among Spanish-English, Polish-English, and Estonian-English college students. Of course, writing expertise and education are unique to each bilingual subject. Writing is a process that requires a wide range of cognitive processes. Writing in an L2 is a process that requires a wide range of additional social processes. This correlational study suggests that advanced bilingualism can improve the quality and fluency of writing production in L2 as well as in L1. It is likely that early exposure to a second language can foster general cognitive abilities as well but causal mechanisms await sorely needed experimental and longitudinal correlational evidence.

ACKNOWLEDGMENTS

The authors would like to Gert Rijlaarsdam, Richard Coff, and an anonymous reviewer for their comments on an earlier draft of this chapter. The second author also thanks Toomas Niit at Tallinn Pedagogical University in Tallinn, Estonia and the United States Fulbright Scholar Program for hosting her stay in Estonia.

THE EFFECTS OF TRAINING A GOOD WORKING MEMORY STRATEGY ON L1 AND L2 WRITING

SARAH RANSDELL*, BEVERLY LAVELLE*, & C. MICHAEL LEVY**

*Florida Atlantic University, USA **University of Florida, USA

Abstract. The main purpose of this chapter is to describe the effects of training working memory strategies on writing performance. The research compares two writing strategies, an ‘all-at-once’ strategy derived from previously-collected analyses of skilled writers and a ‘step-by-step’ derived from less skilled writers (Levy & Ransdell, 1995). The effectiveness of training these writing strategies was tested in four different subgroups of students: Advanced college students writing in L1, beginning students in L1, advanced students in L2, and beginning students in L2. The main research question was whether an all-at- once writing strategy (non-linear processing) could lead to higher writing quality and fluency than a step- by-step strategy (linear processing). Non-linear processing is characterized by continuous planning, text generation, and revision and is associated with good writing. Poor writing involves a linear process of planning first, generating text, and revising last. One hundred and twenty-five participants wrote three 10- minute essays in their own untrained strategy plus the two trained strategies. The all-at-once strategy yielded higher writing fluency and quality than the step-by-step condition across all four subgroups of writers. The effect of training a good working memory strategy was most potent on the least skilled writers in terms of both college experience and English language writing experience.

Keywords: college composition, linear processing, non-linear processing, writing subprocesses, writing training, working memory, writing strategy, training interventions, writing signatures, online data analyses, think-aloud protocols.

S. Ransdell, B. Lavelle, & M. Levy (2002). The Effects of Training a Good Working Memory Strategy on L1 and L2 Writing. In G. Rijlaarsdam (Series editor) & S. Ransdell & M. Barbier (Vol. editors), Studies in Writing, Volume 11: New Directions for Research in L2 Writing, 133 – 144. © 2002 Kluwer Academic Publishers. Printed in the Netherlands. 134 RANSDELL, LAVELLE & LEVY

1 INTRODUCTION 1.1 The writing strategies This chapter presents the results of a training program for introducing specific Eng- lish composition strategies to four different subgroups of L1 and L2 college writers. The study attempts to improve students’ writing across three levels of English language experience, native speakers, and beginning and advanced L2 learners from a variety of L1 backgrounds. In addition, the participants represent two levels of college experience: Third and fourth-year undergraduates represent the advanced level, and first-year students, the beginning level. The four subgroups of writers were included in the training study to determine the generalizability of these procedures for improving a variety of L1 and L2 writers’ fluency and quality. Clearly, no one writing strategy will always produce the highest quality output. However, within the short-term expository writing of the type studied here, there are online data that link a non-linear processing style with the most successful writing (Levy & Ransdell, 1995). Levy and Ransdell collected keystroke records of word-processed writing to determine the temporal patterns of good and poor writers. They discovered a continuum anchored by two types, non-linear and linear. Non-linear processing is characterized by continuous planning, text generation, and revision and is associated with good writing. Poor writing involves a linear process. The linearity can be quite extreme with a limited amount of planning preceding a relatively large amount of text generation. Revision, if done at all, is an afterthought. Why do skillful writers move continuously between writing subprocesses while less skilled writers do not? Better working memory control promotes nonlinear processing and so such control may help account for more successful writing. There- fore, those who have better working memory control should more easily apply a continuous, all-at-once writing method than those with low working memory control. When less skilled writers are asked to plan, they produce written material that closely resembles the eventual product (Burtis, Bereiter, Scardamalia, & Tetroe, 1983). Relatively more skilled writers have far less correspondence between their early drafting and their final product. There is also some evidence that fluency in an L2 can facilitate the co-ordination of nonlinear processing due to increased working memory control (Ransdell, Arecco, & Levy, 2001). Ransdell and her colleagues found that bilinguals were able to maintain L1 writing quality and fluency in the presence of unattended irrelevant speech while maintaining a concurrent 6-digit memory load. In previous research, over 100 monolinguals of comparable verbal and nonverbal L1 skill to these bilinguals reduced quality and fluency under these same conditions (Ransdell, Levy & Kellogg, 1996). It was concluded that processing fluency in another language may confer long-term working memory benefits during dual-task language conditions. These benefits cannot be addressed without direct comparisons of L1 and L2 writers under experimental training procedures. It is often the case that more extensive college experience is associated with better writing. While it is difficult to ascertain whether more college experience is the cause of better writing or whether better writers tend to stay in college, the correlation is worth exploring. This study aims to generalize a training effect in both mono- EFFECTS OF TRAINING A GOOD STRATEGY 135 lingual L1 students and beginning and advanced L2 learners of English who are either starting or finishing an undergraduate degree. Three word-processed essays were collected from these students under three experimental conditions, an untrained baseline, a step-by-step strategy and an all-at-once strategy. The step-by-step strategy is characteristic of writers associated with poorer writing fluency and quality, especially first-year college students in beginning English composition courses. The linearity of their writing is the most apparent temporal feature of their drafting. Ba- sic writers tend to plan sparsely, even when asked to do so before writing per se. Then they fall into a pattern of ‘knowledge telling’ whereby they simply recount the contents of their knowledge regarding the topic at hand (Bereiter & Scardamalia, 1987). More skilled writers, on the other hand, tend to plan, generate text, and revise in rapid, frequently changing subprocess epochs (Levy & Ransdell, 1995). This form of nonlinear processing facilitates knowledge transforming because it allows the writer to construct meaning while engaged in text generation (Tynjala, Mason & Lonka, 2001).

1.2 Online evidence of writing signatures Online data reveal that each person has a specific ‘writing signature’ associated with persistent differences in writing quality and fluency (Levy & Ransdell, 1995). In- cluded in this ‘writing signature’ are individual differences in working memory ability that promote or inhibit nonlinear processing. There have been no studies of writing signature data in L2 writing. This study therefore includes monolingual, as well as bilingual writers generating text in their L2. Recent research has discovered that bilinguals may have an advantage over monolinguals in working memory abilities such as information suppression and capacity (Ransdell, Arecco & Levy, 2001). Ericsson and Delaney (1999) have shown that expertise in many domains can lead to better working memory strategies and even development of a kind of long-term working memory. According to Ericsson and Delaney, long-term working memory is a highly skilled form of access to, and manipulation of, knowledge structures within long-term memory. Kellogg (2001) has recently applied the concept of long- term working memory to the domain of writing. A high degree of domain-specific knowledge was found to lessen interference on a secondary task during writing. It is likely that a high degree of domain-specific knowledge in an L1 and an L2 can also be an advantage under training situations that require the manipulation of strategies within working memory. On-line analysis of word-processed records and think-aloud protocols of skillful college writers are measured to determine writing signatures. A writing signature replayed in real-time that shows nonlinear processing is associated with good fluency and quality. It has been said that the writing process is more important to a good written product than knowledge of essay form (Murray, 1980). In past research, Kellogg (1990) compared writing with and without a prewriting phase. The prewriting phase included explicit planning before text generation. It was hypothesized that prewriting, and other forms of planning before actual text generation, should improve fluency and quality. Kellogg found evidence that outlining and 136 RANSDELL, LAVELLE & LEVY planning improved style and content but this outcome was limited to a specific type of writing that did not require substantial idea generation and organization in the first place. Elbow (1981) viewed prewriting as well as other planning activities as harmful and concluded that they should not be done. He proposed that the creation of a first draft, in which planning, translating, and reviewing interact extensively, improves the quality and fluency of writing (1981). This type of writing has been labeled opportunistic by Hayes-Roth & Hayes-Roth (1979). Opportunistic writing has been said to be effective because the composing itself is a way to think through a written composition (Horton, 1982).

1.3 Writing as a learning tool An entire volume in the Kluwer Studies on Writing series has recently been devoted to the topic of writing as a learning tool (Tynjala, Mason & Lonka, 2001). Most researchers have concluded that tasks that stimulate or require goal formulation and planning will be more effective than those that do not promote knowledge building as the text unfolds (Cumming, Busch & Zhou, this volume; Knudson, 1991). In the domain of reading comprehension, there are clear individual differences associated with more or less skill that parallel the types of writing in this study. Magliano and his colleagues have shown that poor readers process sentences in a highly localized and linear way whereas good readers build knowledge by constantly referring back to previous and future text in order to understand the sentence they are currently reading (Magliano, Little, & Graesser, 1993). It is an empirical question whether planning will be more effective if done simultaneously with text production and revision or whether is should be done as a preliminary step to writing per se. Cognitive studies on attention suggest that tasks that require the writer to plan and form goals before writing will lead to better written results due to reduced demands on attentional capacity (Kellogg, 1990). This idea leads to predictions of a reduction in demands on attention and memory that may, in turn, facilitate writing (Kellogg, 1990). Other research suggests that nonlinear movement through various writing subprocesses is much more characteristic of skilled college writers than a linear, ‘plan first, revise last’ strategy (Levy & Ransdell, 1995). Writers in Kellogg’s studies were free to continue to expand upon the prewriting plan they created and were very likely to do so. This was not then a pure test of linear processing. In the present study, writers are encouraged to completely separate planning, text generation and revision in the step-by-step condition. The main research hypothesis in the present study is that an all-at-once writing strategy will lead to higher writing quality and fluency than a step-by-step strategy. A second question is whether both native L1 writers and L2 learners can show improved writing English quality and fluency in the all-at-once condition. The logic of the research was to try to impose the type of writing signature shown to be representative of more or less skillful writers. Writing fluency and quality should be reduced in a ‘step-by-step’ strategy condition because it mimics the strategy most characteristic of poor writers. Correspondingly, writing should be facilitated in an ‘all-at- once’ strategy condition because it mimics the strategies of good writers. Those with EFFECTS OF TRAINING A GOOD STRATEGY 137 greater college experience and stronger L2 English skills should be better able to generate high quality, fluent essays in each condition.

2 METHOD 2.1 Participants A total of 125 college students wrote three English language essays in the study. Twenty advanced monolingual writers, 43 beginning monolingual writers, 32 advanced bilinguals writing in L2, and 30 beginning L2 learners participated in all experimental conditions. Advanced monolinguals consisted of students from an upper-level psychology class, and beginning monolinguals were from a beginning college English composition class, both in the United States. The upper-level psychology class was for psychology majors nearing the completion of their undergraduate degree. The beginning English composition class was designed to teach basic writing skills to all community college students who were in their first semester of college. Advanced bilinguals were 13 Estonian native speakers and 19 Finnish native speakers at the University of Jyväskylä, Finland, who were fluent in their L1, Eng- lish, and at least one other language. Beginning bilinguals participated from several ESOL (English as a Second Language) classes in the same American community college as the beginning monolinguals taking composition in their L1. The ESOL classes were generally taken as their first formal exposure to English writing composition. All students volunteered to participate in exchange for extra credit in their classes and some tips about improving their English composition skills.

2.2 Design and strategies The study utilized a within-subjects design where participants wrote three 10-minute essays on the computer on three expository topics. A reminder card left on the monitor stated one of the following topics for each condition in randomized order. Topic A card read ‘Imagine the best possible boyfriend/girlfriend. What would this person be like and why would he or she be so great? Be sure to compare your view with other possible opinions’. Topic B card read ‘Imagine the best possible college professor. What would this person be like and why is he or she so great? Be sure and compare your view with other possible opinions’. Topic C card read ‘Imagine the best possible holiday vacation. Where would you go, who would you take with you, and what would you do? Be sure to compare your view with other possible opinions’. Participants were first asked to write as they normally do on the computer to establish their baseline writing strategies, then they were asked to write two 10- minute essays each under the two strategy conditions. The two strategies were ‘step- by-step’ and an ‘all-at-once’. The strategy used for the second and third essays, was determined by counterbalanced order. All essays were written in English. The step-by-step strategy consisted of instructions to plan in the first 10% of the essay writing time, type in the middle 70%, and revise at the end in the last 20% of the time. The all-at-once strategy consisted of instructions to plan, type, and revise continuously throughout the essay. Instructions on the different strategies were fol- 138 RANSDELL, LAVELLE & LEVY lowed by observation of an on-line video sample of a step-by-step and an all-at-once session of writing. An instruction card was left on the student’s workstation with the following reminders. In the all-at-once strategy, the card read ‘Plan, write, and revise as often as you possibly can throughout the entire essay. Use all three processes continuously’. In the step-by-step strategy, during the first minute, the following card was visible, ‘The first step is planning. Planning is creating ideas for what you are going to write about before you actually start writing complete sentences. During the next seven minutes, the visible card read, ‘The second step is writing. Writing is when you put your planned ideas into complete sentences and paragraphs. This is simply the time you spend typing the essay’. In the final 2 minutes of the step-by- step strategy condition, the card read ‘The third step is revising. Revising is when you correct your grammar, spelling, or typing errors. You may also change the meaning of sentences or their organization’. In the pre-writing video, a student peer was depicted using each of the strategies while providing a think-aloud protocol exemplifying the key components of each strategy. The key components of the all- at-once strategy were statements like the following: ‘This sentence I just wrote makes me think of a new idea’; ‘I am revising as I go’; and ‘Changing my text makes me change my thinking’. Step-by-step statements included: ‘In this first section, I must only plan what to write’; ‘Now I can only generate sentences, but not change any of them’; and ‘Revising is saved for last’.

2.3 Individual difference measures Several measures were used to detect individual differences in the participants before writing training. See Table 1 for the English reading comprehension and baseline writing fluency and quality by participant subgroup. The advanced monolingual group included 3rd- and 4th-year college students who were native English speakers. The beginning monolingual group included first-year English composition students. The advanced bilinguals were international students who spoke at least two languages and were at least 3rd- or 4th- year college students. The beginning L2 learners were international students who spoke at least two languages and were first-year college students with limited English writing experience. As is shown in Table 1, the advanced monolinguals were better in English reading comprehension, writing fluency, and quality than the other groups. The beginning monolinguals were not reliably different from the advanced bilinguals on any of these pretraining measures and were intermediate in performance to the other two groups. The beginning L2 learners were significantly poorer in reading and writing in English than the other two groups. It should be noted that the four subgroups of writers were within one standard deviation of performance on Nelson-Denny English reading comprehension (Brown, Fishco, & Hanna, 1993) indicating a characteristically normal range of verbal ability for college students. Advanced monolinguals were found to have better pre-training reading comprehension, writing fluency, and writing quality than the three other groups. Advanced bilinguals were comparable in English, their second or third language, to beginning monolinguals, writing in their native language. Not surprisingly, beginning second EFFECTS OF TRAINING A GOOD STRATEGY 139 language learners were the weakest group. From these data, it would seem, that one would first need an advanced proficiency in English and second, an advanced bilingual ability in order to produce a fluent and high quality written product. This study also provides confirmation of the strong interrelations typically found among measures of reading comprehension, writing fluency and writing quality. The interrelations can be generalized across L1 and L2 boundaries for those with low and high college experience and English experience. Reading comprehension, in fact, accounted for 38% of the variance in writing fluency and 25% of the variance in writing quality.

Table 1. English reading comprehension. baseline writing Fluency and Quality by subgroup.

Monolinguals Bilinguals Advanced Beginning Advanced Beginning (n = 20) (n = 43) (n = 32) (n = 30)

Reading Comprehension 32 24 25 12 Fluency (wpm) 21 15 15 10 Quality (%) 63 51 57 35

Means underlined are not reliably different from one another at p < .05.

2.4 Determination of essay quality and fluency Essays were scored using two dependent measures because past research has shown that fluency is a necessary but not sufficient characteristic of high quality writing (Levy & Ransdell, 1995). Writing fluency was defined as the number of words generated per minute controlling for typing speed and including those words typed before the final draft but deleted. This was determined by a FauxWord program that collected all the participant keystrokes and allowed the researchers to view an exact replay of the essay as it was created (Levy & Ransdell, 1995). Fluency determined from this online method is an improvement because words generated, but then deleted are an indication of potential knowledge transformation and effort that go unnoticed when simple number of words generated or words per minute product-type calculations are used. The second measurement was writing quality, which was determined through a reliable essay sort procedure with trained peer raters adapted from Johnson, Linton, and Madigan (1994). Essay sort is a holistic method, utilizing two separate raters who divide the essays into groups based on the rater’s concept of an essay’s organization, clarity, and development of ideas. Essay sort yielded inter-rater reliability (kappa) coefficients ranging from .71 to .92. Essays were scored on a five-point scale based upon comparison to other essays in the sample. Each rater then reviewed a subset of 30 essays each rating session until the whole group was completed. This 140 RANSDELL, LAVELLE & LEVY task was performed twice on each set, each time separated by one week. Essay quality and fluency are generally correlated with one another in the range of about .50 - .60 and also account for about 25% of the variance in Nelson-Denny reading comprehension scores (Levy & Ransdell, 1995).

3 RESULTS A three-way within-subjects analysis of variance was conducted on the fluency and quality data in two separate analyses. There was no effect of whether the participant received the order with step-by-step first or second and no interaction of order with experimental condition for either fluency or quality analyses. The results for fluency are presented first.

3.1 Writing fluency Figure 1 shows fluency in words per minute (WPM) as a function of the three experimental conditions and by subgroup.

15 WPM 10

0 Adv Begin Adv Bilinguals Begin L2 learners Monolinguals Monolinguals

Baseline Step by Step All at Once

Figure 1. Words created per minute (WPM) as a function of training condition for each group.

Writing fluency was a main effect for all subgroups, F (2, 142) = 7.51, p < .001. In pairwise comparisons, the step-by-step condition (14.83 wpm) was reliably slower than the all-at-once condition (15.1 wpm), F (1,73) = 12.11, p < .001. The step-by- step condition (13.1 wpm) was also reliably slower than the baseline (14.7 wpm), F (1,75) = 9.56, p < .003. Fluency in the baseline condition (14.8 words per minute, wpm) was not reliably slower than in the all-at-once condition (15.1 wpm), F < 1. EFFECTS OF TRAINING A GOOD STRATEGY 141

There was also a significant interaction between the main effect of experimental condition and subgroup, F (2,142) = 2.14, p < .05, but only in the comparison of the baseline versus the step-by-step condition. The two most extreme groups, the advanced monolinguals and the beginning L2 learners showed better fluency in the baseline than the step-by-step condition. The other two subgroups, beginning monolinguals and advanced bilinguals showed no such difference. In the baseline condition, the advanced monolinguals had the best writing fluency overall (20.9 wpm). The beginning monolinguals (14.9 wpm) and advanced bilinguals (15.1 wpm) were equivalent in fluency and the beginning L2 learners (9.7 wpm) were the slowest. A stepwise multiple regression analysis was also performed on the fluency data. A model including subgroup, reading comprehension, and fluency in the baseline and step-by-step conditions accounted for 70% of the variance in fluency in the all-at- once strategy condition, F (4,110) = 66.05, p < .001. The unique predictors of all-at- once fluency were fluency in the other two conditions even when both subgroup and reading comprehension are accounted for.

3.2 Writing quality As with writing fluency scores, the quality of an essay was influenced by the writing strategy employed. A marginal main effect of strategy condition for quality was found, F (2, 105) = 2.57, p < .08. Across all subgroups writers generated 52% quality in the baseline condition, 49% quality in the step-by-step, and 53% quality in the all-at-once condition. There was also a main effect of subgroup: The advanced monolinguals wrote higher quality essays than the other groups, F (3, 106) = 19.49, p < .001. The advanced bilinguals and the beginning monolinguals were comparable in quality and the beginning L2 learners were the poorest. Pairwise comparisons using Bonferroni adjustment showed that only the beginning L2 learners were reliably, but marginally, better in the all-at-once condition than the other conditions, F (2,22) = 2.80, p < .08. For advanced monolinguals, the all-at-once strategy produced the highest quality essays, followed by their original untrained strategy, and least quality was produced when the step-by-step strategy was used. For the beginning monolinguals and advanced bilinguals, there was no reliable difference between the strategies implemented except for the all-at-once strategy producing slightly higher quality essays. For the beginning bilinguals, the all-at-once strategy was most effective, followed by the baseline strategy, and finally by the step-by-step. The effect of training a good working memory strategy was most potent on the least skilled writers in terms of both college experience and English language writing experience. A stepwise multiple regression analysis was also performed on the quality data. A model including subgroup, reading comprehension, and quality in the baseline and step-by-step conditions accounted for 52% of the variance in quality in the all-at- once strategy condition, F (4,109) = 28.86, p < .001. The unique predictors of all-at- once quality were quality in the other two conditions when entered after both subgroup and reading comprehension. Figure 2 depicts quality scores by subgroup and experimental condition. 142 RANSDELL, LAVELLE & LEVY

10 Quality (%)

0 Adv Monolinguals Begin Monolinguals Adv Bilinguals Begin L2 learners

Baseline Step by Step All at Once

Figure 2. Essay sort writing quality (%) as a function of training condition for each subgroup.

4 DISCUSSION These data indicate that an all-at-once strategy facilitates higher fluency in both L1 and L2 writers. Writing quality is also higher in the all-at-once strategy condition compared to the step-by-step strategy for both L1 and L2 writers. The all-at-once strategy had a facilitating effect causing writers to write more fluently and well. The step-by-step strategy had an inhibiting effect causing writers to slow down and write more poorly. Advanced bilinguals were found to be as fluent in English, their L2, as beginning monolinguals are in English, their L1. This finding suggests that writing and general cognitive abilities in L1 transfer to L2, at least at some point in L2 acquisition and development. Writing quality was found to be higher for those with more college experience, but not necessarily more English experience because there was higher performance in the advanced monolinguals and the advanced bilinguals compared to the advanced and beginning monolinguals. Not surprisingly, beginning bilinguals have the poorest L2 writing. Beginning bilinguals write less fluently than the other groups of writers in part due to their lack of English language knowledge. This could be the result of many students learning to write better English as they progressed through college. It could also result from a scenario where no student learns to write better, but that those who were poor writers at the initial testing withdrew from college shortly afterwards. What is surprising is that the beginning monolingual writers are no more fluent nor write of any higher quality in L1 than the advanced bilingual writers writing in their L2. Perhaps this is another example of the general cognitive benefits of bilingual skill (i.e., Arecco & Ransdell, this volume). Or perhaps general writing ability and knowledge of rhetori- EFFECTS OF TRAINING A GOOD STRATEGY 143 cal conventions plays at least as great a role as second language knowledge in creating text. A good writing strategy can be trained, and it can improve writing performance. This was evidenced by a significant main effect of essay condition found for all four groups of students. The all-at-once strategy yielded higher writing fluency and quality than the step-by-step condition and the main difference between the conditions was the nonlinearity of the all-at-once strategy. There is now corroborating evidence that nonlinear processing has analogous effects in the writing and reading domains. Good readers show constant movement within the text they are trying to compre- hend (Magliano, Little, & Graesser, 1993) and good writers show constant movement within the text they are trying to create. It was confirmed that writing quality is not necessarily higher for those with more English experience, but is in fact higher for those with more college experience. All of the different types of writers could be trained, but advanced second language experience was also predictive of good L2 writing. These results have implications for future instruction in writing. Students can benefit from writing strategies known to produce better quality writing. Training studies in composition have a long tradition, but there has been a failure to apply research findings. Furthermore, there has been no research directly comparing training in L1 and L2 among college students. Hillocks (1984) conducted a meta-analysis of nearly 12,000 elementary to college-aged monolingual students in training studies conducted from 1963 to 1982. He analyzed the duration, mode of instruction, and focus of instruction in these studies designed to train good composition practices and found that the dimensions of effective instruction were different from educational practice. Educational research has been hampered by a lack of transfer from the research laboratory to the school classroom. More studies that provide concrete training effects that are applicable to improving composition classrooms are needed. It may be that certain strategies work better with particular types of people dependent upon their working memory control, language experience, time allowances, and preferences. There is no global ‘best’ strategy. In this initial study, students with more college experience writing in L1, as well as for those who were advanced bilinguals writing in L2, wrote the best. And the all-at-once strategy proved the superior method for producing fluent, high quality essays. The training effect can be generalized to monolingual writers in their L1 and to bilingual writers in their L2. It can also be generalized to bilingual writers of English as a second language who represent native language groups very different from English. In sum, there are clearly more similarities than differences when comparing the factors that can lead to good L1 and L2 writing. One of the themes of this volume is that L2 writing can be as strong as L1. There are clearly general cognitive advantages to strong composition experiences in both L1 and L2. This study provides empirical evidence that L2 writing benefits from the same continuous revision processes characterized by high quality L1 writing. A general consensus is evolving that conventions play at least as great a role in creating text, as does second language. 144 RANSDELL, LAVELLE & LEVY

ACKNOWLEDGEMENTS We thank the participating students and their instructor Richard Appelbaum for their involvement. We also thank Gert Rijlaarsdam, Richard Coff, and one anonymous reviewer for helpful comments on an earlier draft of this chapter. Finally we thank the United States Fulbright Scholar Program for funding the first author’s stay in Estonia, and Florida Atlantic University for funding the first author’s stay in Finland.

A COMPARISON BETWEEN NOTETAKING IN L1 AND L2 BY UNDERGRADUATE STUDENTS

MARTINE FARACO*, MARIE-LAURE BARBIER** & ANNIE PIOLAT**

*Laboratory of Speech and Language, ** Centre of Research of Cognition, Language and Emotion, Aix-en-Provence, France

Abstract. In this chapter notetaking is considered as a way to understand L2 writer’s cognitive resources and strategies within a task constraining working memory. Three main linguistic activities are involved in notetaking: understanding, writing and reading. Indeed, notetakers need to store what they hear in order to transcribe it, while at the same time, they have to understand the speaker to change and write down the information. In such a situation, the processing of information and the allocation of resources require the notetaker to perform simultaneous cognitive activities. In the present experiment, the main goal is to identify and analyze the strategies of adaptation in a situation that seems particularly dependent on proficiency level in L2. To identify performances and strategies in such a situation, a comparison is undertaken between native notetakers in French and non-native notetakers more or less fluent in French. The strategies developed by these students during their notetaking are analysed from three main sources of information: a real-time analysis of the relationships between the teacher's discourse and notetaking by the student (notetaking activity is recorded with a graphic tablet), an analysis of the notes produced, and an analysis of the memorization of the information. Results illustrate that notetaking is an important resource-demanding task that depends on the target language fluency, as the most competent non-natives participants do not have the same performances and strategies in notetaking as do less competent ones.

Keywords: L2 acquisition, notetaking activity, working memory, comprehension, written production, study of pauses

M. Faraco, M. Barbier, & A. Piolat (2002). A comparison between notetaking in L1 and L2 by undergraduate students. In: G. Rijlaarsdam (Series ed.) & S. Ransdell & M. Barbier (Vol- ume eds.), Studies in Writing, Volume 11: New Directions for Research in L2 Writing, 145 – 167. © 2002. Kluwer Academic Publishers. Printed in the Netherlands.

146 FARACO, BARBIER & PIOLAT

1 INTRODUCTION More and more students are motivated to obtain qualifications at universities that are internationally recognized. These students pursue university programs within foreign countries in order to be totally immersed in a second language (L2). This situation requires twice as much learning for students. They must learn an L2 so as to achieve an oral and written bilingual competence, and they must become skilled within each domain field necessary to obtain the corresponding qualifications. The pedagogical support of students in such a complex learning situation requires that we better understand the strategies they use to adapt themselves and to study in L2 as efficiently as in their first language (L1). Hence, the general purpose of this research is to analyze further these strategies used by students to follow their lessons in L2. The goal is to examine how these students can understand and select the information they listen to by taking notes at the same time in order to refer to them afterwards for the exams. The main hypothesis is that notetaking is already an important resource demanding activity in L1. It should be even more constraining in L2. For college students, undergraduate and graduate alike, notetaking consists of writing down verbal information that is heard or read. The notes taken will be reviewed afterwards in order to prepare oneself in knowledge testing situations (multiple-choice questions, essay writing, etc.). As it is shown in Boch’s research (1998) as well as in Van Meter, Yokoi & Pressley (1994), students develop a very complex and sophisticated idea of this activity, of its roles and its influence on the success of their studies. The research interests of psychologists and educational psychologists on notetaking are mainly pedagogically orientated. The results of various studies are convergent. For example, it is shown that taking notes has proven more successful in knowledge testing than hearing the same information several times (Kiewra, Benton, Kim, Risch & Christensen, 1995). The effects of notetaking on learning have been studied according to different contexts of notetaking and processes developed by notetakers to transcribe the information. Some of these are the following: Situation of access to the information (reading a text, speech listening; see Morgan, Lilley & Boreham, 1998; Rickards, Fajen, Sullivan & Gillepsie, 1997); techniques for notetaking (verbatim, outline matrix, see Kiewra & al., 1995); incitement to take notes (intentional indication or not from the teacher, see Boch, 1998, 1999); cognitive features of notetakers (Hadwin, Kirby & Wooddhouse, 1999; Ruhl & Suritsky, 1995); type of reviewing activity with notes (read it, rewrite it, see Slotte & Lonka, 1999); and the type of knowledge tests the notetakers are submitted to in order to reformulate the information they have been working on (multiple choice question, essay-writing, see Oakhill & Davies, 1991; Slotte & Lonka, 1999, 2001; Lahtinen, Lonka & Lindbloom-Ylaenne, 1997). It is necessary to underline that studies on the nature of the processes and knowledge involved in notetaking are not considered enough within a psycholinguistic perspective. However, the study of notetaking, particularly when listening to speech, constitutes a relevant situation for the analysis of treatments involved in complex linguistic mechanisms. NOTETAKING IN L1 AND L2 147

1.1 Working memory and notetaking The speed of language production is much faster by speaking than by writing. An adult's speech is about 2-3 words per second. When dictating speech, the speed slows down by only a few hundredths of a second. However, to produce the same content, the rate of writing is about 0.3 - 0.4 words/second (Piolat, 2001). Therefore, notetakers transcribing a lecture’s content must reduce the significant difference between the speed of what they hear and the speed of their writing. The notetakers’ adaptation strategy consists of reducing the information by trun- cating the writing production. Indeed, notetakers can simultaneously abbreviate the surface characteristics of their production (letters, words, parts of sentences and whole sentences) and the ideas considered as important ones. The notes must however, when reviewed, remain comprehensible and unambiguous. The shortening processes must therefore be regular and stable in order to be automatically used for written transcriptions and for the restitution of missing information when notes are reviewed. In general, notetakers have a certain amount of conventional and personal skills in their native language which allow them to pick up as much information as possible despite the time constraints (i.e., abbreviations, icons, semio-graphic signs, up making of pages, etc.; see Boch, 1998, 1999; Piolat, 2001). The notetakers then develop varied strategies that are also related to their level of domain knowledge and their cognitive abilities (Hadwin, Kirby & Woodhouse, 1999). All the choices and decisions associated with notetaking remain difficult. Notetakers are at the same time interpreters and writers, having to choose and to quickly transcribe the information dictated into written notes they can understand. These choices affect the translating process as much as the ideas that are considered important. The particularity of notetaking is that the activities of comprehension and writing are done in real time and simultaneously. It is therefore interesting for psychologists to pay more attention to the fundamental role of working memory in such a complex activity. Language understanding and production are largely deliberate activities and are therefore resource demanding (Baddeley, 2000; Kintsch, 1988; Kellogg, 1999, 2001; Lea & Levy, 1999; Torrance & Jeffery, 1999). Notetaking would be particularly complex (and therefore more resource demanding) since it requires notetakers to deal with these parallel understanding and production processes, each one demanding a substantial cognitive effort. The quantity of cognitive resources would then correspond to the mental energy available for an individual notetaker at a given moment and for a specific cognitive activity. This cognitive capacity could vary over time depending on the task conditions and the performance achieved by the notetaker. Gathercole and Baddeley (1993) have measured this capacity with different types of working memory span tests in order to display its constraining role on the language processing in comprehension and verbal production tasks. Notetakers have to cope with several simultaneous demands that frequently go beyond their working memory capacity. This overloading would be even more constraining for some notetakers who have less cognitive skills, such as less competence in the language in which the lecture is given (e.g., a second language). As competitive demands are made on working memory, the effective activation of the two processes (understanding and writing) involved in notetaking can be very re- 148 FARACO, BARBIER & PIOLAT stricted. To pursue their activity, notetakers have to develop suitable strategies in order to adjust their level of comprehension, as well as the quantity and the content of what they write. For example, some notetakers can prefer a better understanding of the information, but write down fewer ideas. Others can decide to transcribe as much information as possible, but defer the understanding until afterwards, when rereading their notes. This last strategy, which involves a quick speed of production, is a priori easier for a native speaker who has automated the processes of transcription (handwriting and execution). It is less obvious for a writer transcribing in L2. A strategic use by the notetakers of their cognitive resources is therefore crucial. It is then necessary to analyze these notetakers’ strategies in order to describe the way they assign time and effort to the different processes involved in notetaking. In an overload situation, they will have to distribute efficiently their executive resources between the two different processes (comprehension and writing). They can then shift between two goals. Either they develop a strategy focused on the information comprehension (e.g., selecting and organizing what is said to structure the speech; Allison & Tauroza, 1995; Buck, 1992; Hansen, 1994). Or they adopt a strategy focused on word for word transcription (e.g., verbatim notetaking in order to retain and transcribe as many words and knowledge as possible). It remains to examine whether and how notetakers succeed in transferring these possible strategies in the field of L2 notetaking. Do they use specific palliative strategies, or do they develop a combination of transferred and new strategies?

1.2 Notetaking in L2 writing Students who practice their L2 for a long time have acquired some linguistic skills (Bell, 1995; Ellis, 1994) that could allow them to benefit from their notetaking strategies developed in their L1 (Shaw & Ting Kun Liu, 1998). A facilitative transfer (Berman, 1994; Gass, 1996) between the two languages (native and L2) should occur within the L2 processes. However, within the specific context of lecture comprehension in L2, some studies have shown that students have difficulties in locating rhetoric signs that guide the detection of the main ideas within the lecturer’s speech (Chaudron, Loschky & Cook, 1994; Flowerdew & Miller, 1992; King, 1994). These signs should be detected as quickly as possible for notetakers in order to know what they must write down. Moreover, both the lecture’s comprehension and the activation of writing processes are not optimum for a majority of students in L2 (Kern, 1994; Larios, Murphy & Marín, this volume; Sasaki, this volume; Schoonen, van Gelderen, de Glopper, Hulstijn, Snelling, Simis & Stevenson, this volume; Weiss- berg, 2000; Whalen & Ménard, 1995; Zimmermann, 2000). Notetaking even requires specific regular writing procedures such as abbreviations and a spatial organization of the ideas. As illustrated in Faraco (1997a & b, 2000) and Clerehan (1995), students have great difficulties in writing down their notes in L2 for two reasons. Firstly, they ignore some of the abbreviation rules in their L2s and they do not possess sufficiently automated procedures. Secondly, they have problems in selecting and situating the different lecture’s information. These students are overloaded and adopt a strategy NOTETAKING IN L1 AND L2 149 focused on the microstructure of the lecture’s content. When they try to speed up their notetaking, they often write down some words in L1 or they produce neolo- gisms (Faraco, 1997a & b; Clerehan, 1995; Dunkel & Davis, 1994), as if their linguistic abilities in comprehension and writing cannot be adapted to the notetaking task. One of the most common interpretations for explaining student's difficulties in transferring their L1 notetaking strategies, and in preserving their linguistic abilities in L2, is based on the idea of an additional cognitive cost. From several studies about production and reading in a foreign language, Perdue and Gaonac'h (2000) suggested that problems related to L2 activities would not be directly related to a dysfunction of the activated processes. They would be relied instead upon an insufficient automation of surface linguistic treatments, which would make the L2 activity particularly demanding in cognitive resources (Barbier, 1998a; Ransdell, Arecco & Levy, 2001; Segalowitz & Segalowitz, 1993). The question is still to know how notetakers manage their cognitive resources as they become progressively more involved in comprehension and transcription. There are even fewer studies which have tried to evaluate cognitive resources involved in writing production in real time, by comparing native speakers with L2 learners undertaking the same writing task. Some results are however available, particularly on temporal features of text production (Barbier, 1998a & b, 1999; Foulin, 1995; Griffiths & Beretta, 1991; Warren, 1996; Larios, Murphy & Maren, this volume). Namely, the study of pauses has the advantage of examining a natural event that can be easily observed in a writing task, for instance by recording it on a graphic tablet (Chesnet, Guillabert & Espéret, 1994; see Method afterward). The frequency of pauses can be considered as a significant variable to better understand the cognitive load evolving during L2 text production compared to L1. Indeed, the executive process is likely to be constrained by conceptual and/or linguistic treatments during text production (Kellogg, 1999). By considering this executive process as similar for adults in L1 and L2, variations of pauses frequency would inform about the cognitive load involved in planning and controlling these treatments: the more pauses are frequent, the more writers can’t cope with the parallel activation of executive and other writing processes. As a complementary variable, the length of a pause denotes the time needed by writers to elaborate in real time solutions for realizing their production. Their variation would also inform about the cognitive load of writing processes (Foulin, 1995; Matsuashi, 1981; Schilperoord, 1996). Especially, the analysis of the length of pauses according to their localization in the text produced allows one to situate areas where processes are particularly resource demanding (Schilperoord, 1996). In that way, some data is regular within the literature: processes involved during a pause located between the highest structural units of the text (propositions and sentences) would be the most resource demanding. Other data is available from the analysis of the length and frequency of the pauses in L2, compared to in L1. For instance, Barbier (1998a & b) has analysed the written productions of French-speaking participants who were more or less advanced in their learning of English at the university (Second or Master year). These participants had to compose an argumentative text in French (L1) or in English (L2). From the analyses of the length and frequency of the pauses, it appears that at all levels of processes, writing activities needed more cognitive resources for writers in English, including 150 FARACO, BARBIER & PIOLAT the most advanced ones in this language. When L2 writers are exposed to a cognitive overload they assign more resources to the surface aspects, which are immediately useful to the writer, at the expense of more conceptual aspects of their production. This focus on the surface aspects should have a major impact on the strategies developed in notetaking situations. Concerning notetaking, non-natives at the beginning of their studies at university often ignore the abbreviations commonly used by natives (see the abbreviation in French of words ending with ‘ion’ and ‘ment’, or the use of the consonantal skeleton of a word, etc.). Other tools for quick translation as icons are less available to them. To compensate for this lack of knowledge, non-natives could use some processes transferred from their L1, as truncation for example, or notation of the first letter of a word, organization into pages, code switching (to replace a French lexicon or to structure the teacher's speech; Faraco, 1997 a & b, 2000). But these palliative strategies remains limited, because of their instability. The lack of automaticity in capturing ideas certainly leads to a reduced volume of noted words, especially since in L2. The comprehension process is also more costly and requires more time. Faraco (2000) points out a poorer comprehension of the logical structure of the lecture which rarely appears in the non-natives’ notetaking. Finally, L2 notetakers are in a particularly complex situation to manage and they need more cognitive resources at all levels of processing (production and comprehension). They also have to adapt their strategies to co-ordinate the requirements in such a situation.

1.3 Main goals of the research The general purpose of this research is to point out how working memory affects notetaking as a complex linguistic activity that requires simultaneously understanding and writing. More precisely, the main goals of the study are to identify and analyze the following questions. To what extent do notetakers’ difficulties in L2 produce an additional cognitive load rather than simply reflect a deficiency in notetaking processes? Which kind of palliative strategies are developed by L2 notetakers, and to what extent do notetakers rely on their L1 by transferring notetaking knowledge and procedures in L2? Furthermore, the strategies of adaptation and the use of L1 knowledge about notetaking seem particularly dependent on the proficiency level in L2. By studying the notetakers' activities at different stages of their learning of French, the interest is to better understand how non-natives compensate and adapt their strategies according to their experience of notetaking in L2. Therefore, to understand the non-natives’ capacity to cope with comprehension, selection and translation all throughout notetaking, three sources of data have been analysed: chronometric features of the notetaking, the volume and content of notes produced, and the lecture’s comprehension measured after the notetaking. If difficulties in L2 notetaking are effectively dependent upon a greater cognitive load, then the chronometric analysis should reveals more frequent and longer pauses for non- native notetakers compared to natives. Indeed, if non-native’s translation processes are more resource demanding, then these participants should need more time to anticipate and select the semantic content of their notes. For their part, the volume and NOTETAKING IN L1 AND L2 151 content analysis of notes produced as data about lecture’s comprehension should reveal some information about palliatives strategies used by non-natives. If they have more difficulties in understanding the lecture, non-natives should select less information, and then write down a lesser quantity of information (words and ideas) than natives. They should also understand and memorize afterwards less information than natives. But the most-advanced non-natives, whose translation process should be less resource demanding, should be able to refer to their L1 knowledge (by using code switching for example) and transfer some notetaking procedures (truncation or notation of the first letter of a word, organization into the page).

2 METHOD 2.1. Participants Nine French native and eighteen non-native students from the University of Provence participated in this experiment. The French native students were regular first-year students in psychology. They were asked to answer a questionnaire addressing the following three question areas: • age and academic standard, mother tongue and foreign languages known, number of years spent in a foreign country; • experience in notetaking in French and in foreign languages, within the academic context, and in listening or reading situations; • knowledge of literature and in particular of Umberto Eco’s work. All the native participants had learned two foreign languages during their school attendance. None of them was advanced in these foreign languages learning and none of them had lived more than a year in a foreign country. They had little (or even no) knowledge of Umberto Eco's work. They all said that they have always taken notes in French. The non-native students were first-year students within the European program in LEA (‘Langues Etrangères Appliquées’ for ‘Applied Foreign Languages’), which is integrated within the Erasmus-Socrates Program of exchanges associating eight European universities. This program is based on a trilingual approach, with the study of two foreign languages and various subjects as law, economics, management, international trade (depending on the university). After four years of these studies, students receive an international education and obtain three European master's degrees. In the present experiment, the non-native students came from Germany and Spain. Information about their competence, as for non-native participants, has been collected with questionnaire before the beginning of the experiment addressing the following six areas: • age and academic standard, mother tongue and foreign languages known; • experience in notetaking in French and in foreign languages, within the academic context, and in listening or reading situations; • number of years devoted to learning French; 152 FARACO, BARBIER & PIOLAT

• self-evaluation on a four-point scale about the proficiency level in oral (speaking, speech understanding) and written French (reading and writing); • scores from a French language test that students had taken at the beginning of their academic year, in the context of the French course as a foreign language. This test was essentially based on grammatical competencies in French; • knowledge of literature and in particular of Umberto Eco’s work. The non-native participants were divided according to their L2 fluency (in less competent (NN-) and more competent in French (NN+), based on their scores from the French language test. The results for the non-native students group were divided on both sides of the median score (group size = 18; median score = 42). This division was made in accordance with the questionnaire concerning their experience in French language and their auto-evaluation (see Table 1):

Table 1. Characteristics of non-native participants (mean and standard deviation). Scores in the language test distribution, auto-evaluation scores, and number of years of learning French by subgroup (Level of L2 fluency).

NN+ NN – n = 9* n = 9*

M sd M sd

Language test score 26.56 10.49 57.67 8.03 Auto-evaluation score 2.72 5.05 3.19 0.60 Number of years learning French 10.33 5.05 6.33 2.12

*3 German; 6 Spanish

As with native, non-native participants had little (or even no) knowledge of literature nor of Umberto Eco’s work. They were all accustomed to notetaking in their mother tongue but not in French.

2.2. Material and Apparatus 2.2.1. Content of the lecture The lecture used for this experiment was an extract of a course which was initially made for French first-year students at the University of Provence, and recorded on an audiotape for postal tuition. This lecture lasted 12 minutes. The transcription of this lecture came to the amount 1792 words (fluency = 149.33 words/minutes). The content of the lecture was about a comment by Umberto Eco of his novel The Name of the Rose (see Appendix). The teacher gave information about the author; explanations about the subject, announcement of the contents and the plan of the course, argumentation (at first… next… for example…), recall of the most important ideas that students must retain. Then, the lecture presented the main features of an oral NOTETAKING IN L1 AND L2 153 teaching situation, despite non-verbal features (gestures, eye-contact, McNeill, 1992) because of materials constraints (to control the lecturer’s presentation, participants had to work with a tape-recorder). But the two other modalities of human communication (Guaïtella et Santi, 1994), verbal and oral lecturer’s performance, which certainly guide the students’ notetaking, were preserved within this lecture. For verbal features, the teacher performed by accentuating the logical structure of the pedagogical discourse (with connectives, examples, quotations…), and by using repetitions, reformulations, side sequences of explanations, etc. The speech also presented oral features, as over articulations, intonative contours, insistences (emphasis accent, lengthening of the vowels, silent or filled pauses…), etc. The analysis of the participant’s notetaking activity first required examination of the content of the lecturer's oral speech. Therefore, this one was written down and its content was characterized by using the judges' method. Five judges had to fragment the content of the lecture into three categories of segments: basic semantic units (units covering only one basic information, as propositional representations which constitute a sentence-by-sentence information and then the microstructure of the text; van Dijk & Kintsch, 1983), conceptual units (when several units came together about the same conceptual content which is ordered hierarchically among others for representing the macrostructure of the text, van Dijk & Kintsch, 1983), and the major units (when several conceptual units form one of the major ideas of the speech that lead to a representation of the emergent structure of the text). The judges’ classification of the lecture’s content was very similar to one to another. Few discordances (only about 6 basic units) have been observed. In each case, the choice of the majority of judges has been retained. Finally, the lecturer's oral speech contained 7 major units that included 16 conceptual units with 63 basic units (see Appendix).

2.2.2. Comprehension test A comprehension test has been elaborated to give an account of the amount and the nature of the information recorded by the writers during notetaking. This test contained a free question about general comprehension (‘What are the main ideas you have retained from this lecture?’). Thirteen multiple-choice questions were presented (7 questions with 1 correct answer, 5 questions with 2 correct answers, and 1 question with 3 correct answers). The majority of these questions evaluated the participant’s comprehension of the lecture (‘What kind of writing style does the lecturer talk about’? ‘What are the two key-ideas developed by Umberto Eco about writing and reading functions?’). Some of the questions also asked for precise information given in the lecture (‘What is the date of publication of the novel The Name of the Rose?’).

2.2.3. Notetaking on the graphic tablet Participants had to take their notes on a graphic tablet (WACOM) connected to a PC computer. The G-Studio Software (Chesnet, Guillabert & Espéret, 1994) allowed for the creation of a record of the participants' graphic activity. Based on the tablet record, a chronometric analysis of their text setting afterwards was possible. Then, participants’ writing activity were recorded in real time and all the interruptions of 154 FARACO, BARBIER & PIOLAT their writing activity could be located. With this software, a pause was identified when writers stopped their graphic activity, even if the point of the pen rested on the graphic tablet. Pauses corresponding to a ‘calligraphy co-ordination’ (as to link different graphemes, or to realize the ‘t’ stroke for example) have not been considered. In general, the duration of these pauses were shorter than 250 milliseconds (Barbier, 1998a).

2.3. Procedure Each participant was individually tested for about 45 minutes. From the very beginning of the session, participants were told they would have four tasks to do: a) complete a questionnaire about their experience of the French language; b) listen to a recorded academic lecture and take notes on the main information; c) review their notes and then d) take a test on the information they had memorized about the lecture. a) Each participant (native or non native) completed a written questionnaire about his or her experience for about 15 minutes (see 2.1.). b) The instructions of the experimental task itself were given to participants, as they would have to take notes from the recorded academic lecture, and afterwards fill out a written comprehension test about the contents of the lecture, without their notes. Participants were presented with a short description of the lecture contents (the lecturer spoke about Umberto Eco’s novel The name of the rose, and about the comments the author himself made on this novel). The word ‘apostille’ which was cited in the lecture and which is generally unknown even for natives who prefer to use the term ‘postface’, was explained and spelt out to the participants. Otherwise, they were asked not to pay attention to spelling mistakes. Some technical aspects were then detailed (about the manipulation of the pen for the graphic tablet and the way of changing the sheet of paper). Before starting their notetaking, participants were asked to train themselves to use the pen for 5 minutes. They could put the graphic tablet into all the positions they like. They just had to make sure that the sheet of paper stayed fixed on it. Dur- ing their notetaking activity, and in order to change the page, participants had to press upon the tablet with the pen in a space provided to ‘change the page’. Then, the experimenter quickly replaced it by fixing a new sheet of paper on the graphic tablet. The prior one was put close to the participants so that they could continue to refer to it. c) Once the notetaking was completed, the participants took the time they needed to review their notes. d) Afterwards the participants took the written comprehension test. The experimenter read each question out loud and progressively in order to prevent the participants from looking back to the previous questions and to insure an ex- haustive reading of the questions. NOTETAKING IN L1 AND L2 155

3 RESULTS Results are presented according to four categories of dependent variables: the temporal characteristics of the notetaking activity, the volume and content of notes produced, the notetaking procedures, and the comprehension performances.

3.1. Chronometric analysis 3.1.1. Scoring Data concerning the frequency and the duration of pauses have been grouped according to their location (see Appendix). The purpose was to analyze the frequency of pausing at the different levels of the lecturer’s speech content, and then, to evaluate the time spent by the participant for processing at each of these levels. Five type of pauses have been identified according to their location: • Intra-word pauses: mid-position in the word; • Inter-words pauses: between the final position of a word and the beginning of the writing of the next word; • Inter-basic units pauses: between the end of a basic semantic unit and the beginning of the first word of the next basic unit; • Inter-conceptual units pauses: between the end of a conceptual unit and the beginning of the first word of the next conceptual unit; • Inter-major units pauses: between the end of a major unit and the beginning of the first word of the next major unit. Then pauses were linked with one or the other of these five locations which can be regrouped within two categories:

Intra-word and inter-word pauses were considered as local pauses, indicating linguistic treatments by notetakers (as orthographic ones). A word was defined as any group of graphemes (noun, verb, adjective, article) separated by blanks once the lecture was written down. These pauses were included within the basic units and were never located at these or other units’ boundaries. Thus, they were not taken into consideration for the analysis of pauses linked to semantic treatment (see next paragraph). Some criteria have been followed to measure the duration of these local pauses. Intra-word pauses corresponded to one or several pauses within the same word. Furthermore, when the inter-words pause was indicated by a punctuation mark (full stop, comma) the length of the pause was equal to the sum of the lengths of pauses on both sides of the punctuation mark (Foulin, 1995; Barbier, 1998a). Concerning the percentage of intra and inter-words pauses, these have been calculated from the total number of words heard by the participants. Then, the percentage of intra-word pauses has been calculated by dividing the number of these pauses by the total number of words contained in the lecturer’s speech (1792 words). The percentage of inter-word pauses has been calculated by dividing the number of these pauses by the number of potential inter-word locations (1793 inter-word spaces).

156 FARACO, BARBIER & PIOLAT

Pauses at inter-basic, inter-conceptual and inter-major units were considered as global pauses, all of them indicating the notetaker’s semantic treatments about the contents of the lecturer’s speech. Then, the analysis of these pauses needed (1) the segmentation by judges of the lecturer’s speech in three categories of semantic units (basic, conceptual and major ones; see for their description § 2.2.1.), and (2) the setting of the location criteria of the pauses. Each pause was identified only once, according to an ‘inclusive hierarchy’ (Foulin, 1995; Barbier, 1998a). To avoid counting the same pause several times, the units boundaries have been coded exclusively: when a pause could take a position at two or more levels, it was coded as a pause only for the highest level of the hierarchy. The initial pause of a major unit for example, which is also the initial pause of both a conceptual unit and a basic one, has been exclusively coded as a major unit’s initial pause. For each participant, the duration of these three types of semantic pauses has been calculated as for the duration of the local pauses. Their percentages have been calculated by dividing the number of pauses for each type of location by the number of potential locations corresponding within the lecturer’s speech (see Appendix: 64 inter basic units; 17 inter- conceptual and 8 inter-major units).

3.1.2. Frequency and duration of local treatments Results show that inter-word pauses are significantly more frequent than intra-word pauses, F(2,24) = 86.59 p<.0001 (see Table 2).

Table 2. Percentage of pauses (and standard deviations) according to local treatments (intra- words and inter-words) and to the target language fluency (natives, non-natives+ and non- natives—).

N NN + NN – Total

Type of local pauses M sd M sd M sd M sd

Intra-words 11.0 06.2 25.1 15.0 26.1 12.6 20.7 13.4 Inter-words 39.6 15.8 51.3 16.1 58.7 24.4 49.9 20.1 Total 25.3 18.8 38.2 20.2 42.4 25.2

The target language fluency has a significant effect on frequency of a local treatment, F(2,24) = 3.807, p<.04. More precisely, simple effects analyzed reveals that intra-word pauses were less frequent for natives (M = 11%) than for the most competent non-natives (M = 25.1%; F(1,16) = 6.852; p<.02) and for less competent ones (M = 26.1%; F(1,16) = 10.415; p<.006). Considering inter-word pauses frequency, the analysis of simple effects shows that natives tend to pause between words less frequently (M =39.6%), compared to the less competent non-natives (M = 58.7%; F(1,16) = 3.905; p<.066; see Table 2). NOTETAKING IN L1 AND L2 157

The two factors (type of local pauses and language fluency) do not interact significantly concerning the frequency of pauses. The length of intra and inter-word pauses does not differ significantly (see Table 3). It still does not vary according to language fluency. The two factors (type of local pauses and language fluency) do not interact significantly concerning the length of local pauses.

Table 3. Mean length of pauses in seconds (and standard deviations) according to local treatments (intra-words and inter-words) and to the target language fluency (natives, non- natives+ and non-native—).

N NN + NN – Total

Type of local pauses M sd M sd M sd M sd

Intra-words .74 .30 .62 .12 .73 .18 .69 .21 Inter-words .70 .16 .73 .15 .83 .21 .75 .18 Total .72 .24 .67 .14 .78 .19

3.1.1 Frequency and duration of global treatments Results show that the frequency of pauses vary significantly for each type of global pauses, F(2,48) = 61.695 p<.0001 (see Table 4). The pauses situated at the inter- major units are the most frequent (M = 93.1%), followed by the pauses situated at the inter-basic units (M= 68.5%), and the pauses situated at the inter-conceptual units which are a lot less frequent ones (M= 54.7%).

Table 4. Percentage of pauses (and standard deviations) according to global treatments (basic units, conceptual and major units) and to the target language fluency (natives, non- natives+ and non-natives—).

N NN + NN – Total

Type of global pauses M sd M Sd M sd M sd

Inter-basic units 57.5 8.7 59.6 5.5 47.1 18.2 54.7 12.9 Inter-conceptual units 70.0 9.3 69.6 6.1 65.9 27.6 68.5 16.6 Inter-major units 92.9 17.5 96.8 6.5 89.7 15.1 93.1 13.7 Total 73.5 19.2 75.3 17.0 67.5 26.9

The frequency of global pauses does not vary significantly according to the language fluency. The two factors (type of global pauses and language fluency) do not interact significantly concerning the frequency of pauses. 158 FARACO, BARBIER & PIOLAT

The length of pauses varies significantly for each type of global pauses, F(2,48) = 4.694 p<.014 (see Table 5). The pauses situated at the inter-major units are the long- est ones (M = 9.02 s), followed by the pauses situated at the inter-conceptual units (M = 7.32 s) and the pauses situated at the inter-basic units which are the shortest ones (M = 5.49 s). Without considering their location, the length of pauses also varies according to the language fluency, F(2,78) = 3.960 p<.023 (see Table 5). The analysis of simple effects reveals that less competent non-natives pause longer (M = 10.53s.) than natives (M = 5.26s.; F(1,52) = 4.958; p<.03); they tend to pause longer than the most competent non-natives (M = 6.03s.; F(1,52) = 3.466; p<.0683). The two factors (type of global pauses and language fluency) do not interact significantly concerning the duration of pauses.

Table 5. Mean length in seconds (and standard deviations) according to global treatments (basic units, conceptual and major units) and to the target language fluency (natives, non- natives + and non-natives—).

N NN + NN – Total

Type of global pauses M sd M sd M sd M Sd

Inter-basic units 4.77 (2.72) 4.58 (2.35) 7.11 (5.04) 5.49 (3.63) Inter-conceptual units 5.68 (2.97) 6.31 (3.67) 9.66 (10.16) 7.32 (6.51) Inter-major units 5.33 (3.01) 7.20 (4.94) 14.53 (17.50) 9.02 10.99) Total 5.26 (2.82) 6.03 (3.81) 10.53 (11.98)

Finally, an analysis concerning the frequency and duration of the 5 types of locations of pauses has been realized (without considering the two categories of treatments – local or global). Regardless of the target language fluency, results show that the pauses located at the word units boundaries (intra-word, inter-words) are more occasional and are shorter than those located at the boundaries of larger semantic units (inter-basic units, inter-conceptual and major units), F(4,96) = 100,788 (see Tables 2 & 4); p <.0001; F(4,96) = 17,574; p <.001 (see Tables 3 & 5). Global pauses (inter-basic units, inter-conceptual and major units) are on average 10 times longer than local pauses (intra-word, inter-words). They are also on average twice as frequent. The proficiency level in French (L2) also has a global effect on the mean length of pauses, without considering the categories of pauses. Less competent non-natives pause longer (M = 6.63 sec.) than natives (M = 3.44 sec) and than most competent non-natives (M = 3.88 sec.), F(2,132) = 3,015; p<.0524. NOTETAKING IN L1 AND L2 159

3.2 Volume and content of notes produced 3.2.1 Scoring Considering the lecturer’s speech, several variables have been used to analyse the volume and the contents of the notes produced by participants. The volume of the notes corresponds to the total percentage of words selected (e.g., number of words in the notes on the 1792 words contained within the lecturer’s speech). Moreover, the percentage of words noted within each category of units has been evaluated (e.g., number of words within basic units, within conceptual units and within major units noted by participants). The contents of the notes have been analysed thoroughly by considering the mean percentage of basic, conceptual and major units noted from the total number of each of these units within the lecturer’s speech. The mean percentages of basic and conceptual units per major units have also been considered. In addition, for each participant, the number of basic units that were misinterpreted has been calculated (number of misinterpretations).

3.2.2 Words selection About the percentage of words produced, contrast analysis reveals that natives (M = 14.9%) note down more words than all non-natives together (M = 12%), F(1.25) = 4.081, p = .0542 (see table 6). Considering the number of words within each type of unit, contrast analysis shows that only the most competent non-natives produced less words within basic units (M = 3.8) than natives (M = 5.3), F(1,16)=10.169, p<.006 (see table 6). But all the non-natives produced less words within conceptual units (M = 10.3) than natives (M = 12.8), F(1,25)= 4.205, p =.0509 (see table 6). All the non-natives produced also less words within major units (M = 29.25) than natives (M = 36.3), F(1,25)= 4.518, p< .05 (see table 6).

Table 6. Mean percentage of words produced according to the target language fluency, in consideration of the number of words contained within the lecture and within each basic, conceptual and major units produced.

N NN + NN – Total

Percentage of words produced M sd M sd M sd M

Total 14.9 3.2 12.4 2.9 11.6 4.2 31.6 Within basic units 5.3 1.2 3.8 0.7 4.9 2.0 13.0 Within conceptual units 12.8 2.4 10.3 2.5 10.7 3.3 4.7 Within major units 36.3 8.8 31.3 6.5 29.3 8.8 11.3

160 FARACO, BARBIER & PIOLAT

For all of these variables, scores between the most competent non-natives and the less competent non-natives do not differ significantly (as was found in Ransdell, Lavelle & Levy, this volume).

3.2.3 Content selection Contrast analysis reveals that the most competent non-natives (M = 63.8%) tend to note more basic units than less competent non-natives and natives altogether (M = 53.2%), F(1.25) = 4.24, p =.05 (see Table 7). The target language fluency has no significant effect on the percentage of notation of conceptual and major units.

Table 7. Mean percentage of basic, conceptual and major units noted, according to the target language fluency.

N NN + NN – Total

Ratio of Units M sd M sd M sd

Basic semantic units 56.2 9.9 63.8 11.5 50.3 15.8 56.8 Conceptual units 79.9 7.9 83.8 5.0 75.2 14.5 79.6 Major units 82.7 12.5 79 8.7 84 5.9 81.9

The target language fluency has a significant effect on the notation of basic units within major ones, F(2,24) = 5.38, p<.013 (see table 8). The most competent non- natives produce more basic units within major ones (M = 8.1) than less competent non-natives and natives together (M = 6.4), F(1,25) = 8.689, p<.007. The target language fluency has also a significant effect on the notation of conceptual units as considered within major units, F(2,24) = 4.073, p=.03 (see table 8). The most competent non-natives produced more conceptual units within major units (M=3.1) than less competent non-natives and natives together (M = 2.7), F(1,25) = 6.18, p<.02.

Table 8. Mean percentage of basic and conceptual and units per major units according to the target language fluency.

N NN + NN – Total

Structuration of content by major units M sd M sd M sd

Basic units 6.8 1.1 8.1 1.4 5.9 1.6 6.9 Conceptual units 2.8 3.0 3.1 4.0 2.5 4.2 2.8

NOTETAKING IN L1 AND L2 161

Moreover, the number of misinterpretations varies significantly according to the target language fluency, F(2,24) = 3,713; p<.04. The less competent non-natives produced significantly more misinterpretations (M = 2.4) than the most competent non-natives (M = 0.7), F(1,16) = 6.294, p<.03. The difference between natives (M = 1.2) and most competent non-natives is not significant, F(1,16) = 0.831, p = .3755.

3.3 Notetaking procedures 3.3.1 Scoring For each participant, variables related to the procedures for notetaking have been considered: the number of abbreviations (words which were not written in a complete form), the number of icons (arrows, hyphen, etc), and the amount of basic units that were reformulated (number of reformulations with other words than those pronounced by the lecturer).

3.3.2 Results The target language fluency has no significant effect on the amount of abbreviations and icons. However, contrast analysis shows that natives tend to produce more abbreviations (M = 25.2) than less competent non-natives (M = 11.2; F(1.16) = 3.198, p<.10). The difference between the natives and the most competent non-natives is insignificant. Contrast analysis shows also that natives tend to produce more icons (M = 1.7) than less competent non-natives (M = 0.2; F(1.16) = 4.225, p<.06; see table 9). For both of these groups, the differences compared to the most competent non-natives are insignificant. Furthermore, the number of reformulations varies significantly according to the target language fluency, F(2,24) = 3.802, p<.04 (see table 9). The most competent non-natives produce less reformulations (M = 4.2) than natives (M = 10.3; F(1,16) = 6.231, p<.03) and than less competent non-natives (M = 16.3; F(1,16) = 6.085, p<.03).

Table 9. Mean number of abbreviation, icon produced, and reformulation according to the target language fluency

N NN + NN – Total

Notetaking procedures M sd M sd M Sd

Abbreviation 25.2 20.2 25.3 31.9 11.2 11.8 20.6 Icon 1.7 2.1 6.7 11.2 0.2 0.4 2.9 Reformulation 10.3 6.6 4.2 3.2 16.3 14.4 10.2

162 FARACO, BARBIER & PIOLAT

3.4 Comprehension performance The participant’s comprehension of the lecturer’s speech has been evaluated by analysing the score that each participant obtained on the comprehension test (see § 2.2.2.). This score has been calculated by adding one point for each correct answer to the corresponding item. Remember that 13 multiple-choice questions were presented (7 questions with 1 correct answer, 5 questions with 2 correct answers, and 1 question with 3 correct answers). Thus, the participant could attain a maximum of 20 points. Results show that the target language fluency had a significant effect on the scores obtained on the comprehension test (F(2,24) = 8.284, p<.002). Contrast analysis shows that the mean score for less competent non-natives (M = 9.3 (1.8) is significantly lower than that for more competent non-natives (M = 13 (2.7); F(1,16) = 11.256, p<.004), in comparison to that of the natives (M = 12.1 (1.1); F(1,16) = 14.859, p<.002). The difference between natives and competent non-natives is not significant, F(1,16) = 0.844, p = .3718). As shown in Table 10, a significant relationship was found between comprehension scores and processing variables as abbreviation, icons and reformulation. The more abbreviations and icons produced, the better the comprehension scores. In- versely, the more reformulation of the lecturer speech, the worse the comprehension scores.

Table 10. Correlations between comprehension scores and processing variables (abbreviation, icon, and reformulation).

1. 2. 3. 1. Comprehension scores 2. Abbreviation 0.63 3. Icon 0.62 0.79 4. Reformulation - 0.46 - 0.36 - 0.25

Underlined correlations are statistically significant at p<.05 level.

4 DISCUSSION Several results need to be considered in this experiment. First of all, the number of words' delivered by the notetakers is contrasted between natives and non-natives. The translation process in L2 is particularly resource demanding. This cognitive load does not lead to longer intra- and inter-word pauses, but to much more frequent pausing. Non-native notetakers do hesitate at intra- and inter-words boundaries nearly twice as long as natives. The notes' content processing is also more resource demanding for non-natives than for natives. However, to anticipate the content of their notes in L2, notetakers do not pause as frequently. They pause longer, especially to plan the conceptual and the main units of organization. Actually, non- NOTETAKING IN L1 AND L2 163 natives need twice as long as natives to anticipate the organization of units contents (conceptual and major) during their notetaking. These first results show that notetakers in L2 (especially less competent ones) have difficulties in following the lecturer’s speech and in taking notes. This seems to be related to (1) difficulties with local processing (word translation) as suggested by intra and inter-word’s pauses' frequency; and (2) problems in managing the whole of the task and understanding the lecture’s content as it is suggested by the increased length of pauses concerning semantic units. One can observe that these difficulties are additional to those generally observed during notetaking by natives. Indeed, for L1 notetakers also, distributions of pauses’ length and frequency have the same pattern according to their localization. They are far longer and more frequent at semantic processing which is particularly resource demanding in notetaking. Results concerning the volume and content of notes produced are consistent with the data about the understanding and translating cognitive load all along notetaking in L2. They also clarify the strategies used by notetakers to adapt themselves to such a complex situation. Indeed, notetakers difficulties in L2 are in understanding the lecture and in translating it at the same time. This leads them to avoid risks of translating and to take fewer notes than in L1. This data concern both advanced and less advanced non-natives, but it is not dependent on the same profile of notetaking strategy. Indeed, the less competent participants in L2 show greater difficulties in notetaking as they select fewer semantic units (basic and conceptual) from the lecturer’s speech. They use fewer abbreviations and icons but they do more reformulations. Results concerning the lecture’s comprehension reveal that these participants also introduce more misinterpretations within their notes and have difficulties in memorizing the lecture’s contents. These data confirm, as Faraco (2000) pointed out for L2 non-advanced notetakers, that there is a poorer comprehension of the macrostructure of the lecture compared to the natives’ notetaking. By contrast, the results taken from the most competent non-natives were better compared to that of the least competent ones. They even write a little more semantic (basic and conceptual) units than natives. These performances are obvious in considering the ratio of units written down (the number of selected ideas with regard to the whole lecture’s content; see § 3.2.3.). Notetaking procedures too are contrasted according to language fluency (between the least and the most competent L2 participants). As with native notetakers, the most competent non-natives have developed abbreviation procedures and they prefer to use icons rather than to reformulate the lecturer’s ideas. These last notetaking procedures lead to a better comprehension and memorization of the lecture content’s. Then, these participants did not refer much more to their L1 during L2 notetaking by using procedures as code switching (with some words reformulated in L1). They do not use directly their L1 but they apply more efficiently their notetaking skills previously acquired in another language. The whole set of these results is fairly consistent. From three variables about notetakers performances, it is shown that L2 notetaker’s difficulties are not directly dependent upon a dysfunction of the activated processes. It is much more an additional cognitive load which would raise difficulties for non-native students to develop successful notetaking strategies (Barbier, 1998a; Ransdell, Arecco & Levy, 2001; Segalowitz & Segalowitz, 1993). If the notetakers are not as advanced in L2, 164 FARACO, BARBIER & PIOLAT they are in a situation of cognitive load for two reasons. Firstly, the linguistic processing L2 is highly demanding and they have not developed as much automated procedures (in comprehension and writing) as have more advanced L2 students. Sec- ondly, they do not use the specific procedures of notetaking, either because they do not succeed in transferring their L1 notetaking skills or either because they are not sufficiently trained in applying these skills with efficiency in another language, as L2 advanced seem capable of doing. By contrast, these last ones have a different notetaker’s profile and perform better, namely because they understand better the lecture’s content and succeed in managing the different processes. At last, they take advantage of their notetaking by memorizing much more information afterwards. These various results illustrate how on-line analyses of notetaking with a focus on information comprehension, selection and translation modes can inform psycho- linguists about the L2 expertise. In this way, this experiment supports studies concerned with the role of working memory in writing (Kellogg, 1999; Lea & Levy, 1999; Torrance & Jeffery, 1999; Ransdell, Lavelle & Levy, this volume) thus notetaking can be considered in particular as a highly demanding writing task.

ACKNOWLEDGMENTS The authors would like to thank Cyril Deniaud for his efficient and decisive help in the experiment, by assuring the connexion of the G-Studio Software with the graphic tablet. We also would like to thank Sarah Ransdell, Gert Rijlaarsdam, Jacques Vauclair, and two anonymous reviewers for their relevant comments on earlier drafts of this chapter. For further correspondence: [email protected] aix.fr, [email protected], and [email protected] NOTETAKING IN L1 AND L2 165

APPENDIX The lecture was originally in French. Only a part of it is presented here after [missing information is indicated as (…)]. The transcription of the features in the lecturer’s speech is described here. a) The number of each units’ boundaries within the lecturer’s speech: the total number of words (1792); the total number of spaces between words (1793); the total number of spaces between basic units (64); the total number of spaces between conceptual units (17); the total number of spaces between major units (8); b) The potential localization of pauses within the lecturer’s speech transcription (see afterwards; Note: Word and intra-words pauses are between [3] basic units boundaries within conceptual or major unit; [4] conceptual units boundaries within major unit; and [5] major unit boundaries)

BEGINNING: The Apostille of The Name of the Rose, the novel of Umberto Eco

[5] Umberto Eco is a contemporary Italian semiotician, specialist in the study of the signs and a medie- valist. [3] Since 1982 (date of publication of The Name of the Rose) he becomes a novelist too.

[5] In an interview, he explains the transition to novelist for his because he enjoys the challenges of something new [3] and he distinguishes the two writings: writing of his novels, writing of the essays without organizing them into a hierarchy. ------[4] On one hand the theoretical writing: is looking for a solution to a problem, in defending conclusions by arguing his point. [3] Thus from this writing, there is only one possible argumentation, one only good interpretation ------[4] While in the writing of a novel, the author sets the scenes for contradictions and ambiguities, the author doesn’t commit himself and leaved space for the reader’s interpretation

[5] Our first set of remarks about the apostille of The Name of the Rose will be to think about its functions ------[4] From that we have just said, one can understand the difference what Eco makes and which exists, between the novel The Name of the Rose and the apostille [3] (…) [3] And, Eco adds to it a preface, no not a preface, a postface, an apostille. That is to say an addition made in the margin of the narration. [3] So not being a preface, it is not intended to be read before, [3] but it is a postface written afterwards, to answer to the readers’ letters (that is what justifies in part its presence). [3] It is a collective answer [3] but which doesn’t answer however to all questions. [3] In effect, Eco explains that he refuses to unlock the interpretation because it would be taken as the of the novel that he criticizes (see the page 6). [3] and he refuses also to answer questions that he judges as unnecessary (page 84).

166 FARACO, BARBIER & PIOLAT

[5] So what does Eco do in his postface? [3] and firstly I will give the main arguments. ------[4] First of all he justifies the writing of it (that confirms in some way that this writing is not obvious). ------[4] After he tells why and how he has written his novel. [3] First why a historic novel situated in the Middle-Age, [3] And afterwards how he wrote such a novel.

[5] And his answers to these questions were given in precise indications about the writing of The Name of the Rose [3] by delegating somebody to speak, somebody who is a narrator/character, an old monk who spoke about an episode from his youth [3] (…) + [3] (…) ------[4] Another technique is, he thinks, so the author writes thinking about a reader, [3] a reader who falls into the trap which is set for him, a reader who becomes an accomplice with the author (or that he tries to become the accomplice with the author) [3] Thus the author thinks about a reader, not about an identified reader who responds to his demands, [3] At the contrary, Eco defends himself in having made a commercial book [3] (…) + [3] (…)

[5] Hence two key-ideas on which our second point will be: [3] on one hand in writing is in constructing a reader, [3] on the other hand in reading is in giving the text a sense and to interpret it, so as to construct a signi- fication.

[5] Second point about the apostille: novel and reading. ------[4] Firstly, thus, to write is to build the mind of the reader. [3] and the construction of the reader is the title of a chapter (page 54) which concludes theoretically the previous comments [3] (…) + [3] (…) ------[4] In building the reader it can be developed into three propositions. ------[4] The first one, one writes thinking about a reader, not a precise reader, but about the possibility to be read one day (it is that he explains on pages 56-57). [3] (…) + [3] (…)+[3] (…) + [3] (…)+[3] (…) + [3] (…)+[3] (…) [3] Thus the writer thinks about he effect he will produce, that he wants to produce in any case, [3] and he talks with that he calls a model reader, that is to say a virtual reader, abstract and who will bring up to date the different interpretative possibilities of the text. ------[4] Second proposition: one can write, one can write, that one of the possibilities, is to satisfy the known wish of a majority of readers or to make the reader discover that he must to want, says Eco. [3] And at that moment the writer speaks with a new reader. [3] And the example taken by Eco is that of Manzoni, the great Italian author, who radically revitalizes the historic novel of the XIX° century. [3] (…) + [3] (…)+[3] (…) [3] Thus there are the two possibilities, there is two possibilities, two choices, two possibilities: to answer the request of the reader or on the contrary to construct a new reader ------[4] (…) NOTETAKING IN L1 AND L2 167

[3] It is the role, for example, who will play in The Name of the Rose, the choice of a naïve narrator who will ask questions, the good questions, the questions which would ask the reader himself or the choice of the paraleipsis figure that we will go back over.

[5] Second remark, to read is to construct significations. [3] As the reader (as the author, sorry) constructs the world of his novel, by keeping his characters in keeping/consonant with the epoch chosen, [3] the reader, him, having to reconstruct this world from the data of the text [3] (…) + [3] (…) [3] Construct significations, it is also because literature is complex and ambiguous, it is to generate several possible interpretations. It is to make sure that the text generates several possible interpretations. ------[4] The examples given by Eco are this of the title which, he says, provide always a kind of guidance to the reading [3] (…) + [3] (…)+ [3] (…) ------[4] Another example, the reader can operate not planned parallels. [3] (…) [3] And the parallel generates interpretations that the author had not thought about, had not planned. ------[4] Another example of these pluralities of interpretations. [3] The reader can finally perceive or not the explicit or implicit parallels [3] that keeps up the novel with the other texts of the literature [3] Eco evokes the parallels of his novel with the detective novel (it is page 62, page 63), [3] with the story of the apocalypse (page 33), [3] with the stories of Borgès (page 34) [3] and according to his own culture the reader will be able to make the text have conversations with such or such other. ------[4] To conclude this second point. One sees that the reading is a full activity [3] and an interpretative activity, a hermeneutic activity [3] and an activity at the same time free and constraint. ------[4] The reader is forced by the text to adopt a certain rhythm (see page 49) [3] and constraint for example by Eco’s novel to read Latin (there are long theological discussions about poverty). [3] The reader is changed by his reading of a novel which don’t answer his wishes ------[4] but in another side he is free to interpret as he likes: the title, [3] to judge the behaviours of the characters, [3] to evaluate the degree of irony that this historical novel comports which is at the same time a parody of a detective novel, of a novel of investigation. END

COLLABORATIVE WRITING IN L2: THE EFFECT OF GROUP INTERACTION ON TEXT QUALITY

FOLKERT KUIKEN & INEKE VEDDER

University of Amsterdam, the Netherlands

Abstract. This chapter investigates the role of group interaction in L2 writing. It is assumed that text quality in L2 is positively affected by collaborative dialogue. When learners are given the opportunity to reconstruct a text together, which has been read to them by their teacher, their joint product will be better than an individual reconstruction. In order to test this hypothesis, three groups of L2 learners of English, Dutch, and Italian were submitted to a dictogloss task. The focus of the study was on the grammatical and lexical complexity of the text produced by the learners and on the linguistic strategies they used during the text reconstruction phase. However, a positive effect of interaction could not be established. This may partly be due to the cross-sectional design of the study. Other factors which may have had an impact on the obtained results are also discussed, such as level of L2 proficiency, difficulty of the text, group dynamics and the general problem of measuring the role of interaction in L2 writing. In the final section of the chapter it is suggested that a qualitative analysis of the discussions between the learners might shed some light on the benefits of a collaborative writing task like dictogloss.

Key words: collaborative dialogue, dictogloss, grammatical complexity, interaction, language strategies, lexical richness, metacognition, noticing, output hypothesis, text reconstruction

1 INTRODUCTION In this chapter the results and implications of a study on collaborative writing in a second language (L2) are presented. The study was conducted to investigate the effect of group interaction on text quality in L2. We will report on an experiment carried out among three groups of adult and adolescent learners of Dutch, English and Italian as a second language. These learners were submitted to a collaborative writing task, based on the dictogloss procedure. Our main research question is whether the syntactic and lexical quality of the text produced by the learners is positively affected by the way learners interact with

F. Kuiken & I. Vedder (2002). Collaborative writing in L2: The effect of group interaction on text quality. In: G. Rijlaarsdam (Series ed.) & S. Ransdell & M. Barbier (Volume eds.), Stud- ies in Writing, Volume 11: New Directions for Research in L2 Writing, 169 – 188. © 2002. Kluwer Academic Publishers. Printed in the Netherlands.

170 KUIKEN & VEDDER each other. Moreover, we are interested in the kind of grammatical and lexical strategies which L2 learners use while being engaged in a collaborative writing activity. With respect to our research question, the study addresses three issues involved in the debate on the effects of instruction and the conditions which may be facilitative for L2 acquisition: noticing, interaction and metacognition. With regard to noticing, much recent research in Second Language Acquisition (SLA) has investigated the question of whether conscious attention to a particular linguistic form may promote the acquisition of that form (Doughty and Williams 1998; VanPatten 1996; VanPatten & Cadierno 1993). Noticing a linguistic form in the input is thought to operate as a necessary, though not sufficient condition for processing and acquisition to take place. As stated by Schmidt (1990, 1995), Skehan (1998) and Robinson (1995, 1996), there seems to be a connection between learners’ awareness of linguistic forms in the input and successful learning. A second line of research, closely related to the concept of noticing, concentrates on the role of interaction and output. As claimed by Swain’s Output Hypothesis, output may influence noticing. An important role in this respect is attributed to collaborative dialogue, i.e. dialogue in which learners are engaged in negotiating meaning and language building. Collaborative dialogue mediates L2 learning. Through collaborative dialogue, learners engage in co-constructing their L2 and in building knowledge about it. Collaborative language production tasks may therefore prompt learners to deepen their awareness of linguistic rules. This may trigger cognitive processes that may both generate new linguistic knowledge and consolidate existing knowledge (Swain 1985, 1998; Swain & Lapkin 2000a, 2000b). Also metacognition is supposed to have a facilitative effect on L2 acquisition. As assumed by Ellis (2000), there are properties in a learning task that will predispose or induce learners to engage in certain types of language use and mental processing that are beneficial for acquisition. Verbalization of problems in contexts in which learners are engaged in meaningful interaction may help learners to understand the relation between form and meaning. As a consequence, particular learning tasks, like problem-solving and collaborative writing activities, may positively affect the acquisition of L2 knowledge, since these activities lead to a greater metacognitive awareness (Long & Robinson 1998). Noticing, interaction and metacognition, as important conditions for L2 learning, are the theoretical assumptions underlying dictogloss. Dictogloss is a pedagogical task in which learners are encouraged to reflect on their own output. In section 2, the dictogloss procedure as an example of a collaborative writing activity will be illustrated. In section 3, Skehan’s Information Processing Model, on which our study was based, will be discussed. Section 4 contains a description of the research question and design of the study. In Section 5, the results of the product and process analysis of dictogloss are summarized. The product analysis concerns the grammatical and lexical complexity of the written text produced by the L2 learners in close co-operation. In this process, we refer to the linguistic and interactional strategies that have lead to the production of the text. The chapter concludes with a discussion. COLLABORATIVE WRITING IN L2 171

2 COLLABORATIVE WRITING IN L2: DICTOGLOSS Collaborative writing tasks in L2 or L1 are often assigned as a way to foster reflective thinking. The very act of explaining and defending ideas to one’s peers is thought to force students to take critical positions on their own ideas and writing. In a couple of studies concerning academic writing by university students in L1, the effects of collaborative writing on the development of reasoning skills and metacognitive awareness were examined. In a study conducted by Higgins, Flower & Petraglia (1992) on the role of metacognitive reflection in student collaboration, a significant correlation was found between the amount of reflective conversation and the quality of students’ writing plans. In a similar study by Keys (1994, 1995) on the development of scientific reasoning skills in a naturalistic collaborative writing context, the changes in students’ reasoning discourse were analyzed. Over time, participants improved their ability to compose scientific explanations. Moreover, collaborative writing encouraged students to construct their own understandings of science concepts. Although there are some research findings regarding the cognitive and psychological aspects of collaborative writing, these studies are mainly concerned with collaborative writing in L1. Few studies have explored the specific nature of collaborative writing in L2. Furthermore, little is reported on the linguistic aspects and the factors that may influence the success of various writing techniques. This applies to the dictogloss technique as well. Although in the case of dictogloss, the writing task is more constrained by the input because L2 learners have to produce a written version of a text that first has been read to them by the teacher. Also a dictogloss task can be considered as an example of a collaborative writing activity. Apart from a few studies conducted by Izumi, Bigelow, Fujiwara & Fear- now (1999), Kuiken & Vedder (2000, in press), Swain (1998) and Swain & Lapkin (2000a, 2000b), there is but little information on the degree to which dictogloss may promote L2 acquisition or on the specific linguistic areas which may be affected. As stated in section 1, the focus of the present study is both on the grammatical and lexical complexity of the text produced by the learners during a dictogloss task and on the linguistic strategies that are used. Wajnryb (1990) provides a detailed description of the dictogloss procedure from a pedagogical point of view. First, a short text is read at normal speed to a group of L2 learners. This text, which is either a constructed or an authentic one, is intended to provide practice in the use of particular linguistic forms or constructions. While the text is being read, learners take notes; they then work together in small groups to reconstruct the initial text from their shared resources. Together they find out what they know, should know and do not know about the target language. After the reconstruction phase, the final version is compared with the original text, and then analyzed and commented upon by the teacher. As an illustration of dictogloss and how it works, two examples are presented below for L2 Italian. Example (1) contains an original text read by the teacher, A tutta velocità! (‘At high speed!’). Example (2) contains the reconstructed version made by one of the groups. The translation into English is ours (see Appendix 1). A tutta ve- locità! is a text on car racing and cycle racing in Italy. Comparison of the two ver- 172 KUIKEN & VEDDER sions shows that the reconstructed version is much shorter and grammatically and lexically simplified. Moreover it contains a number of grammatical and lexical errors, such as populare instead of popolari, and corso instead of corsa.

(1) A TUTTA VELOCITÀ! (original version)

L’automobilismo e il ciclismo sono due sport molto popolari in Italia. Molti appassionati seguono i bolidi da corsa impegnati nelle gare di Formula 1 in Italia o all’estero. Le due gare più importanti sono il Grand Prix di San Marino, che si corre sul circuito di Imola, in provincia di Bologna, e il Gran Premio di Monza, vicino a Milano. La competizione ciclistica italiana più importante è il Giro d’Italia, che richiama un gran numero di persone ai bordi delle strade che percorre. Si corre a giugno, dura una ventina di giorni e va dalle Alpi alla Sicilia. Da alcuni anni esiste anche un Giro d’Italia internazionale femminile. Il colore della maglia della prima in classifica è rosa, come per gli uomini. Il ciclismo non è seguito soltanto stando seduti in poltrona, ma è anche praticato attivamente soprattutto nell’Italia settentrionale e centrale. La sera, al ritorno dal lavoro, molti inforcano la bicicletta e pedalano per qualche chilometro. C’è chi preferisce invece fare jogging, prima o dopo il lavoro. Anche se gli uomini sono ancora la maggioranza le donne si dedicano sempre di più a queste attività sportive. Oggi si ha molta cura del proprio corpo e per mantenersi in forma si va in palestra a praticare il body building; è una ginnastica che si fa con l’uso di attrezzi, regolati sulla forza individuale.

From: G. Martina & A. Pagliarulo (1996). In Italia. Amsterdam: Intertaal, p. 53.

(2) A TUTTA VELOCITÀ! (reconstructed version)

Gli sport, soprattutto il cyclismo e l’automobilismo sono molto populare in Italia. Quattro esempi del automobilismo sono la Formula Uno, il Gran Prix di San Marino, il Grand Premio di Monza, e il Grand Prix d’Imola. Il corso di bycicletta più importante è il Giro d’Italia. Questo corso è in giunio e va degli Alpi alla Sicilia. C’è anche un corso internazionale femminile. Per tutta la gente, anche gli uomini, il colore dei vestiti è rosa. In generale il cyclismo è molto populare in Italia, soprattutto al centro d’Italia, per esempio dopo il lavoro e la sera. Qualche sport per mantenersi in forma sono il bodybuilding e la ginnastica. Qui si tratta alla forza individuale.

The excerpt under (3) is the initial part of a discussion between Nico, Bart and Rona, three Dutch university students of Italian, while they were reconstructing A tutta velocità. Example (3) illustrates how the three participants, during the reconstruction of the content of the initial text, were discussing both content (C), lexicon (L), spelling (S), pragma-rhetorics (P) and grammar (G). The discussion between the students was mainly in Dutch, with a few words in Italian (between quotation marks); the translation into English is ours.

COLLABORATIVE WRITING IN L2 173

(3) Nico: ‘Ci sono molti...’ [C] What are fans or supporters? [L] Bart: Eh... ‘Tifosi’. [L] Nico: The word fans doesn’t exist, I think. [L] Rona: ‘Tifosi’. [L] Nico: ‘Tifosi’? [L] Bart: ‘Tifosi’. [L] Rona: I don’t think that was in the text. [C] Nico: With ‘f’or with... [S] Bart: No, I think I had something else than ‘tifosi’. [C] Nico: Well, it doesn’t matter. With ‘f’ probably? [S] ‘Molti tifosi di Formula uno?’ Yes? Or... [C] Bart: No..., yeah, perhaps that many people are [C] watching that sport and that... something something... abroad? Nico: Yeah, but what I’ve written down is that there [C] are many fans, but we didn’t tell of what kind, I think it was of that... of that ‘Formula uno’. Bart: Or, or both. [C] Rona: I think both. [C] Nico: Both? ‘Ci sono molti tifosi di...’ [C] Rona: But that is the same again, like the first [P] sentence, more or less. Nico: Yeah, but that doesn’t matter. [P] Rona: Well, that’s the way she (= the teacher) said it. [P] Nico: Yes, she said it like that. So, what shall [P] we write down: of these two sports? Bart: ‘Di...’ [P] Rona: Yes. [P] Nico: ‘Di questi...’ [P] Bart: Or something like that. [P] Nico: ‘Questi due sport? Sporte?’ [G] Bart: ‘Sport’, because it’s a foreign word. [G] Rona: Yes. ‘Sport’, it remains the same, I think. [G] Nico: You’re right: ‘questi due sport’. [G]

As can be seen from the example, the students were not sure about the Italian equivalent of ‘fans’ (tifosi). They also vacillated between the plural form sporte and sport. Bart proposed to use sport, being a non-Italian word with no separate plural form. Rona and Nico accepted his solution. As a result of collaborative dialogue, the linguistic forms tifosi and sport were probably ‘noticed’ by the three learners. This noticing may subsequently stimulate the process of language acquisition. It may be hypothesized that while learners interact with each other, their language ability improves, as far as their morpho-syntactic, lexical and pragma-rhetorical skills are concerned. In the following section, Skehan’s Information Processing Model, in which noticing occupies a central position, will be discussed. 174 KUIKEN & VEDDER

3 SKEHAN’S INFORMATION PROCESSING MODEL According to Skehan’s Information Processing Model, various influences affect noticing, such as the frequency and salience of the input, classroom instruction, task demands on processing resources, individual differences between learners in processing ability, and readiness to pay attention to certain linguistic forms (Skehan 1998:52). The Information Processing Model emphasizes input processing and the interaction of input features, via noticing, with the interlanguage system of the L2 learner. Skehan’s model, which was adapted somewhat for the purposes of our study, can be described as follows (see Figure 1): 1. Input qualities: the more frequent and prominent a form, the more likely it is to be noticed in the input. 2. Fo- cused input: noticing may be influenced by both instruction and the selective effects of tasks, which make particular forms salient. 3. Modified input: noticing may be triggered by interaction, which may induce learners to search for alternative forms and to modify their output, in order to be understood by other learners (Long 1983, 1996). As a consequence, output becomes modified input and influences noticing, as can be seen from the arrow in the model from ‘output/interaction’ to ‘modified input’. 4. Task demands on processing resources: the cognitive complexity of a language task may influence noticing as well, in terms of making it more or less likely to occur. 5. Internal factors: individual differences (IDs) between learners in processing ability, and readiness to pay attention to certain linguistic forms may also have an impact on noticing. 6. Working memory and long-term memory: working memory is activated by the various influences operating upon noticing. The result of noticing then becomes available for modification and incorporation into long-term memory (Robinson 1995).

Input

Qualities Working Frequency Memory Salience Noticing

INTERACTION Focused Input Long-term Instruction Memory OUTPUT Selective Tasks Effects INPUT

Modified Task demands Internal Input on Factors Processing Readiness Resources Processing Capacity

Figure 1: Types of noticing based on Skehan (1998). COLLABORATIVE WRITING IN L2 175

Skehan’s Information Processing Model applies to dictogloss in the following ways. The text read to the learners (‘input’ in Skehan’s model) contains particular linguistics forms and structures, which occur frequently and stand out prominently in the input (‘input qualities’). In the classroom, learners’ attention has already been drawn to it. Through its characteristics, the dictogloss task itself makes these structures more prominent (‘focused input’). While interacting with each other, learners modify their output in order to be understood (‘modified input’). In terms of cognitive complexity, the task demands of dictogloss on processing resources are probably neither too high nor too low, so it may be assumed that they will not overload the learners’ internal system in such a way that noticing is less likely to occur (Robin- son, in press). Between the learners of each group there are presumably internal differences in processing capacity and readiness. Collaborative attentional focusing and co-construction of knowledge may lead to a change in the underlying interlanguage system. Working memory is activated and as a result of noticing, the structures may be stored into long-term memory. Acquisition of the grammatical structures under investigation may therefore be facilitated.

4 RESEARCH QUESTION AND DESIGN 4.1 Research question In our study the relationship between text quality and the way learners interact with each other is explored. Before we say something about the nature of this relationship, two preliminary questions are taken into consideration. The first one has to do with the process that is going on while learners are reconstructing the text: which strategies do learners use in order to reconstruct the original text? The term strategy can be defined as follows: by ‘strategy’ we mean every utterance during the reconstruction phase aimed to overcome a communicative or a linguistic problem. We distinguish between strategies referring to language content, language form (lexicon, spelling, pragma-rhetorics and grammar) and to the procedure of dictogloss (see example 3 in section 2). In this chapter we concentrate on language form, in particular on the nature of the lexical and grammatical strategies used by the learners. The second preliminary question concerns the product of the dictogloss procedure: to what extent do learners succeed in reconstructing the original text? It is highly probable that the reconstructed text will be simpler than the original one. Learners will not be able to reproduce the full text because of memory problems; moreover they will encounter lexical and morphosyntactic problems when reconstructing the text. In order to find out whether these assumptions are true the original and the reconstructed texts will be compared with each other in terms of reproduction of content, grammatical complexity and lexical richness. Our main research question concerns the relationship between the strategies used during the interaction process and the outcome of the collaborative writing task: to what extent is there a connection between the various linguistic strategies used by the learners on the one hand and the grammatical complexity and lexical richness of the reconstructed texts on the other? We hypothesize that interaction and metacogni- 176 KUIKEN & VEDDER tive reflection on language form, text content and the writing process itself result in more noticing and, as a consequence, in a better knowledge of certain grammatical and lexical forms. It may therefore be expected that there will be a positive effect from the use of grammatical and lexical strategies on the grammatical complexity and lexical richness of the reconstructed text. In other words, the more of these strategies, the more complex and rich will be the text in grammatical and lexical respects.

4.2 Design and subjects The subjects involved in the study consisted of three groups of L2 learners, aged between 16 and 35 years: 1) 11 students of Dutch as a second language coming from various linguistic backgrounds who had been studying Dutch for almost one year (DSL); 2) 15 Dutch high school students who were in their fifth year of learning English (ESL); 3) 14 Dutch university students who were in the middle of their first year of Italian (ISL). All learners were screened as having obtained an intermediate level in their L2 proficiency. They were asked to perform a dictogloss task. As explained in section 2, the teacher read a short text. The learners were told to listen and to take notes; then they had to reconstruct together a written version of the original text in small groups consisting of three or four students. The instruction the learners got was that the reconstructed version had to be grammatically and lexically correct and as complete as possible with respect to the content of the original text. No specific instruction was given regarding text length, number of sentences, textual organization and structure. This chapter reports on two groups of DSL students (D1 and D2), and on three groups of ESL (E1, E2 and E3) and ISL students (I1, I2 and I3). For DSL a text called De Bibliotheek (‘The Library’) was used. For ESL and ISL two texts were used: for ESL Jungle Boy (text 1) and Phobia Poll (text 2); for ISL A tutta velocità (‘At high speed’, text 1) and Videogiochi (‘Videogames’, text 2). These texts were selected in close co-operation with the teachers who checked them for text difficulty, content familiarity and interest level.

4.3 Strategies used during text reconstruction In order to be able to investigate the interactional process during the reconstruction phase of dictogloss, we analyzed the strategies used by the learners. Each strategy was coded along two dimensions: type of strategy and type of domain (see Figure 2). COLLABORATIVE WRITING IN L2 177

Form Content Procedure Rest

Grammar Lexicon Spelling Pragma- rhetorics

Strategies

Searching Transfer Deduction Avoidance Rule knowledge Other

Figure 2: Strategies used during text reconstruction

We identified three levels of analysis. On the first level we distinguished between language form, content, procedure and a rest category in which utterances were scored which could not be attributed to one of these three categories. On the second level of analysis, language form was subdivided into lexicon, spelling, pragma- rhetorics and grammar, as illustrated in example (3) in section 2. With respect to lexicon and grammar, we made a distinction between the following six strategies: searching, transfer, deduction, avoidance, rule knowledge and other.

Strategy Examples

Searching (Search): ‘What are fans or hooligans?’ to look for the correct word or structure. ‘I can write down something like i ciclisti professionali possono parti… participere or participare, something like that.’ Transfer (Trans): ‘Yeah, but I think in Latin it’s something to make use of knowledge of L1 or any other with parti… something with ere in any language. case.’ ‘And se… In French it is also something like en se, I believe.’ Deduction (Deduct): ‘It should be vicino a Milano and not vicino to deduce a word or structure based on anal- da. Didn’t we have before something like ogy with similar cases. vicino alla stazione?’ ‘Doesn’t rosa go the same as blu? I think these words do not change.’ Avoidance (Avoid): ‘Why don’t you write down ha fatto, that’s 178 KUIKEN & VEDDER

to avoid a word or structure on purpose and less risky.’ replace it eventually by something else. ’Or shall we skip that part?’ Rule knowledge (Rule knowl): ‘Participere, I should say, if it was partici- to make use of knowledge about lexical rules pare it would be emphasized.’ like collocations of a word or about gram- ‘And if you write down nei Alpi…No, then it matical rules. should be gli Alpi. But that’s wrong, because mountains are mostly feminine.’ Other: ‘Does anybody have an idea how you could in this category were instances scored which say that?’ could not be attributed to one of the other ‘Yeah, that doesn’t sound bad.’ strategies mentioned, but where the discussion on a lexical or grammatical problem was going on.

5 RESULTS 5.1 Process analysis With regard to what happened during the reconstruction of the text, we started by examining the use of strategies referring to form, content and procedure. The results of this analysis are shown in Table 1. In this table (as well as in Tables 2-4) first the raw total scores of the strategies used are given. The other scores are given in percentages in order to make comparisons between the different groups easier. These scores show that the groups differ in their use of strategies, especially from one language to another.

Table 1: The use of strategies (in %) during text reconstruction in Dutch, English as Second Language and Italian as a Second Language.

DSL ESL ISL Text 1 Text 2 Text 1 Text 2 D1 D2 E1 E2 E3 E1 E2 E3 I1 I2 I3 I1 I2 I3

Raw total 463 444 137 111 136 236 287 204 343 191 576 238 161 350 Form 35.1 39.8 20.4 36.9 13.9 18.6 28.2 35.3 25.4 26.7 47.8 28.1 32.7 34.9 Content 58.4 53.3 67.9 49.5 80.7 77.0 64.5 59.3 76.5 72.2 49.6 68.0 64.9 60.9 Procedure 5.7 6.5 11.7 13.5 7.4 3.4 4.5 2.9 0.9 - 1.2 3.4 - - Rest 0.9 0.5 - - - 0.8 2.8 2.5 - 1.1 1.7 0.4 - 2.0

From Table 1 it can be inferred that the groups differed somewhat in their attention given to form, varying from 13.9% for E3 (text 1) to 47.8% for I3 (text 1). It turned out that in the majority of cases, learners discussed the content of the text: they did so at least half of the time and some groups even spent three quarters or more on COLLABORATIVE WRITING IN L2 179

discussing content (E3 on text 1, E1 on text 2 and I1 on text 1). Little attention was paid to procedural issues. The groups did not only differ in the amount of time spent on discussing language form, but also in the domains of language form they discussed. The results of the analysis of the use of formal strategies (lexicon, spelling, pragma-rhetorics, grammar) are presented in Table 2.

Table 2: The use of formal strategies (in %) during text reconstruction.

DSL ESL ISL Text 1 Text 2 Text 1 Text 2 D1 D2 E1 E2 E3 E1 E2 E3 I1 I2 I3 I1 I2 I3

Raw total 165 167 28 41 20 44 81 72 81 51 274 67 54 122 Lexicon 29.1 27.2 17.9 22.0 95.0 18.2 40.7 20.8 50.6 25.5 50.4 46.3 24.1 65.6 Spelling 44.2 35.2 53.6 2.4 5.0 27.3 38.3 44.4 14.8 51.0 19.3 7.5 40.7 - Pragm.rhetor 10.9 14.8 14.3 24.4 - 22.7 12.3 27.8 7.4 - 7.3 6.0 3.7 21.3 Grammar 15.8 22.7 14.3 51.2 - 31.8 8.6 6.9 27.2 23.5 23.0 40.3 31.5 13.1

Some groups mainly paid attention to lexical problems. This was particularly the case for E3 (95% on text 1), whereas I1 and I3 spent about half of their discussions on language form on lexical issues. Other groups paid relatively more attention to spelling. This was the case for D1 (44.2%) and D2 (35.2%) as well as for E1 (53.6% on text 1) and I2 (51%; 40.7%). Pragma-rhetorics was not so much discussed although E2 spent a substantial amount of time on it (24.4% on text 1) as well as E3 (27.8% on text 2), but for some other groups this was not a point of discussion at all (E3 and I2 on text 1). The groups also largely differed in talking about grammar. In group E3, this subject was not touched upon in text 1, whereas half of the formal strategies used by E2 in discussing the same text were devoted to grammatical issues. Regarding the use of lexical and grammatical strategies, the results make clear that the learners did not so much use transfer, deduction and avoidance, but most of the time, a general searching strategy. Also rule knowledge occurred, more regarding grammar than with respect to lexicon. Often a substantial part of the discussions on grammar and lexicon could not be attributed to one of the fore mentioned strategies and were therefore scored in the category ‘other’ (see Tables 3 and 4). The results regarding the use of the different strategies employed during the reconstruction phase of dictogloss can be summarized as follows: 1) Half of the time or more was spent on the reconstruction of content (Table 1). 2) Learners differed in their attention devoted to language form, one group paying little attention to it (E3 on text1), another group (I3 on text 1) nearly half of the time (Table 1). 180 KUIKEN & VEDDER

3) Learners differed also very much in their attention given to lexicon, spelling, pragma-rhetorics and grammar (Table 2). Some groups mainly paid attention to lexical problems (E3, I1, I3). Other groups paid more attention to spelling (D1, D2, E1, I2). Pragma-rhetorics were not much discussed. The attention for grammar varied from none (E3 on text 1) to half of all the discussions on formal issues (E2 on text 1). 4) In their use of lexical strategies, learners mostly seemed to prefer a general searching strategy above more specific strategies like transfer, deduction, avoidance and rule knowledge (Table 3). A substantial part of the discussions had to be scored in the category ‘other’. The same is true for the use of grammatical strategies (Table 4), although some groups relied heavily on rule knowledge (E2, E3 on text 2, I2 on text 2, I3).

Table 3: The use of lexical strategies (in %) during text reconstruction.

DSL ESL ISL Text 1 Text 2 Text 1 Text 2 D1 D2 E1 E2 E3 E1 E2 E3 I1 I2 I3 I1 I2 I3

Raw total 48 48 5 9 19 8 33 15 39 13 99 31 13 80 Searching 70.8 60.4 100 66.7 68.4 62.5 66.7 66.7 71.8 100 71.7 48.4 100 68.8 Transfer 4.2 ------20.5 - 2.2 - - 1.25 Deduction ------Avoidance - 8.3 ------7.7 - 5.1 - - - Rule kno. - 10.4 ------6.5 - - 17.5 Other 25.0 20.8 - 33.3 31.6 37.5 30.3 33.3 - - 14.5 51.6 - 12.5

Table 4: The use of grammatical strategies (in %) during text reconstruction.

DSL ESL ISL Text 1 Text 2 Text 1 Text 2 D1 D2 E1 E2 E3 E1 E2 E3 I1 I2 I3 I1 I2 I3

Raw total 26 40 4 21 - 14 7 5 22 12 63 27 17 16 Searching 84.6 50.0 25.0 9.5 - 35.7 14.3 - 36.7 91.7 38.1 63.0 52.9 6.3 Transfer - - - 9.5 ------3.2 - - - Deduction ------Avoidance ------1.6 - - - Rule kno. 7.7 5.0 - 52.4 - - 38.6 100 36.7 8.3 34.9 - 47.1 75.0 Other 7.7 45.0 75.0 28.6 - 64.3 57.1 - 27.3 - 25.4 37.0 - 8.3

COLLABORATIVE WRITING IN L2 181

5.2 Product analysis As far as the analysis of the product of collaborative writing i.e. the linguistic quality of the reconstructed texts is concerned, we have examined three aspects: reconstruction of content, grammatical complexity and lexical richness. All texts were scored for these three aspects by the two researchers and a research assistant. In the rare cases where the raters differed from each other, they easily came to an agreement in discussing the problem.

5.2.1 Reconstruction of content We started by comparing the number and percentage of the content elements of the reconstructed texts with those of the original text. The learners had to reconstruct the text as well as they could. So the more content elements represented in the reconstructed text, the better the result. What content elements are can be illustrated by means of an example. The first sentence of the original text of A tutta velocità! is: L’automobilismo e il ciclismo sono due sport molto popolari in Italia (see example (1) in section 2). In this sentence we distinguished four content elements: l'automobilismo, il ciclismo, (due sport) molto popolari and in Italia. In the reconstructed text under (2) in section 2 all these four elements were reproduced: Gli sport, soprattutto il cyclismo e l’automobilismo sono molto populare in Italia. However, towards the end of the reconstructed text, whole sentences contained in the original text are missing, like Il ciclismo non è seguito soltanto stando seduti in poltrona, ma è anche praticato attivamente sopratutto nell’Italia settentrionale e centrale. The results concerning the reconstruction of content are presented in Table 5, where Or stands for the original text.

Table 5: Number of content elements in the original (OR) and reconstructed texts.

DSL ESL ISL Text 1 Text 2 Text 1 Text 2 OR D1 D2 OR E1 E2 E3 OR E1 E2 E3 OR I1 I2 I3 OR I1 I2 I3

Total 12 10 6 18 10 8 11 12 10 11 10 38 17 25 20 17 13 14 9 Percent. 83 50 55 44 61 83 92 83 45 66 53 76 82 53

The ESL and ISL learners were better in reproducing the content of text 2 than of text 1. It might be that one of the texts was considered more difficult than the other. Besides that there were differences between the groups concerning the number of content elements. D1 for instance reproduced 83% of the content elements, whereas D2 only reproduced half of it. 182 KUIKEN & VEDDER

5.2.2 Grammatical complexity We assumed that, if only for reasons of limited memory space, the reconstructed texts would be simpler than the original ones. First we investigated if the reconstructed texts were grammatically less complex than the original version. Grammati- cal complexity was scored by means of a general measure (MLU = mean length of utterance); more specifically we looked at the structure of the sentence, the noun phrase, the adverbial phrase and the verb phrase (see Table 6).

Table 6: Grammatical complexity.

DSL ESL ISL Text 1 Text 2 Text 1 Text 2 Or D1 D2 Or E1 E2 E3 Or E1 E2 E3 Or I1 I2 I3 Or I1 I2 I3

General N sentenc. 8 6 6 7 7 7 10 6 6 6 7 13 9 14 10 9 8 9 7 N words 91 78 84 116 71 68 90 111 77 70 86 233 120 154 135 148 98 101 87 MLU 11.4 13.0 14.0 16.6 10.1 9.7 9.0 18.5 12.8 11.7 12.3 17.9 13.3 11.0 13.5 16.4 12.3 11.2 12.4

Sentence complexity Simple 5 3 2 1 4 4 4 1 2 1 2 2 7 8 6 3 3 3 3 Compound 3 3 4 6 3 3 6 5 4 5 5 11 2 6 4 6 5 6 4 Co-ordin. - 2 2 6 1 3 1 1 3 2 1 5 1 2 3 4 3 2 2 Subordin. 3 2 4 6 2 3 7 12 4 6 8 11 1 6 1 12 5 8 3

Noun Phrase Simple 20 14 5 19 8 8 14 20 19 15 21 28 22 20 18 12 10 18 7 Modified 8 7 6 9 7 6 6 14 6 7 8 36 14 22 21 25 15 11 10

Adverbial Phrase 13 14 10 16 9 7 14 5 6 8 8 30 12 17 19 17 8 8 12

Verb Phrase Simple 3 3 3 13 7 8 12 8 7 7 12 27 11 17 11 20 9 18 8 Compound 6 5 9 7 3 4 5 6 5 4 2 3 - 1 2 3 4 - 3

What we found is that in general, the reconstructed texts were indeed simpler than the original ones. They contained less words and the mean length of utterance of the reconstructed texts was lower, except for D1 and D2. With respect to sentence complexity, the number of simple sentences in the reconstructed texts was the same or higher than in the originals – again with the exception for D1 and D2 – while the number of co-ordinated and subordinated sentences was lower in most cases. Regarding the use of noun phrases, the number of simple noun phrases was generally lower in the reconstructed versions than in the originals. Simple noun phrases consist of nouns, eventually preceded by a determiner (‘monkeys’, ‘a boy’, ‘his sur- COLLABORATIVE WRITING IN L2 183

roundings’), as opposed to modified noun phrases which contain a noun that is specified by another element (‘a little boy’, ‘a tribe of monkeys’, ‘his hands clenched’). The number of modified noun phrases used in the reconstructions turned out to be lower as well. This was particularly the case for ESL and ISL. For the complexity of the adverbial phrases (‘in Africa’, ‘at the age of one’, ‘around him’) we also found that the number of adverbial phrases was nearly always lower in the reconstructed versions. With respect to verb phrases the number of simple predicates exceeded the number of compound predicates for ESL and ISL, but not for DSL. By simple predicates we mean verb phrases consisting of a single verb (‘finds’, ‘believes’, ‘mothered’), while compound predicates contain two or more verbs (‘has been found’, ‘is believed to have lost’, ‘has been mothered’). In nearly all cases compound predicates were used to a lesser extent in the reconstructed versions than in the original ones. To summarize we can conclude that in accordance with our expectations, the sentences of the reconstructed texts were shorter, they contained less co-ordinated and subordinated sentences, less simple and modified noun phrases, less adverbial phrases, and less simple and compound verb phrases than the original texts.

5.2.3 Lexical richness Regarding lexical richness, we analyzed all content words that did not belong to the 2000 most frequent words of Dutch, English and Italian. A cut-off point of 2000 words was chosen because this amount corresponds to what can generally be considered a basic productive vocabulary for L2 learners. As can be seen in Table 7, lexical richness was examined in terms of the number of retained items, additions and substitutions.

Table 7: Lexical richness.

DSL ESL ISL Text 1 Text 2 Text 1 Text 2 OR D1 D2 OR E1 E2 E3 OR E1 E2 E3 OR I1 I2 I3 OR I1 I2 I3

N items 16 15 5 19 7 8 15 20 16 11 15 44 19 30 22 35 23 19 12 Retained 14 2 3 2 9 11 9 9 12 16 9 21 16 10 Addition 0 0 1 1 2 1 0 0 0 2 2 0 0 0 Substitution 1 3 3 5 4 4 2 6 7 12 11 2 3 2

Frequency of substituted items Higher 1 2 2 4 2 2 2 3 3 7 5 2 1 2 Lower 0 0 0 0 0 0 0 0 0 0 1 0 0 0 Same 0 1 1 1 1 2 0 3 4 5 5 0 0 0

184 KUIKEN & VEDDER

Retained items are those words that the learners reproduced from the original text. By additions, we mean words that were introduced in the reconstructed text without a clear reference to the original text. Substitutions are words that were replaced by other ones (e.g. ‘environment’ instead of ‘surroundings’). These substitutions might be more frequent, less frequent or belonging to the same frequency category as the lexical items used in the original texts. We assumed that the reconstructed texts would differ from each other in the number of retained items and in the number of more or less frequent substitutions used by the learners. What we found is that lexically the reconstructions were always less varied than the originals. There were however differences between the reconstructed texts. Some versions were rather poor, containing about a third of the lexical items of the original (D2 and E1 on text 1, I3 on text 2), while others were lexically more varied, with three-quarters or more of the words from the original text (D1, E1 on text 2, E3 on text 1 and 2). In many cases lexical simplification strategies were applied, since the frequency category of the substituting items in the reconstructions was, except for one case (I3 on text 1), always higher or the same as the frequency of the lexical items in the originals.

5.3 The relationship between process and product In order to answer our main research question, we compared the grammatical and lexical strategies used by the learners during the reconstruction phase with the grammatical complexity and lexical richness of the reconstructed texts. The result of this comparison is shown in Table 8. In this table, the numbers of grammatical and lexical strategies are presented as well as the percentages of the strategies used during the reconstruction phase. These percentages are compared with the mean length of utterance (MLU) as a global measure of grammatical complexity and with the number of lexical items found in the reconstructed texts as a measure of lexical richness. Our assumption was that the more learners discuss grammatical and lexical issues, the more complex will be the text and the more lexical items it will contain. On the basis of our data, such a straightforward conclusion cannot be drawn. As far as grammatical complexity is concerned, group D2 uses more grammatical strategies and has a higher MLU as compared to D1. Group E3 does not make use of grammatical strategies in text 1 and has also the lowest MLU in comparison with E1 and E2. Group E1 uses the most grammatical strategies in text 2 and has a higher MLU than that of E2 and E3. But in all other cases, no clear relation can be established between the grammatical strategies used by the groups and their MLU’s. A similar picture emerges with respect to lexical richness. For the learners of DSL and ESL on text 1, the higher the percentage of lexical strategies used, the more lexical items are contained in the text. But for the learners of ISL and ESL on text 2 this relation was not observed. This means that we have to conclude that no unambiguous relationship between the strategies used and the reconstructed texts can be established. COLLABORATIVE WRITING IN L2 185

Table 8: Relationship product-process.

DSL ESL ISL Text 1 Text 2 Text 1 Text 2 D1 D2 E1 E2 E3 E1 E2 E3 I1 I2 I3 I1 I2 I3

Grammatical complexity Grammatical strategies N 26 40 4 21 - 14 7 5 22 12 63 27 17 16 % 15.8 22.7 14.3 51.2 - 31.8 8.6 6.9 27.2 23.5 23.0 40.3 31.5 13.1 MLU 13.0 14.0 10.1 9.7 9.0 12.8 11.7 12.3 13.3 11.0 13.5 12.3 11.2 12.4

Lexical richness Lexical strategies N 48 48 5 9 19 8 33 15 39 13 138 31 13 80 % 29.1 27.2 17.9 22.0 95.0 18.2 40.7 20.8 50.6 25.5 50.4 46.3 24.1 65.6 Nitems 15 5 7 8 15 16 11 15 19 30 22 23 19 12

6 DISCUSSION Our findings show that, with regard to the use of strategies, half of the time was spent on the reconstruction of content. Moreover, learners differed in their attention dedicated to language form, one group paying hardly any attention to it, another group nearly half of the time. They also differed very much in their attention given to lexicon, spelling, pragma-rhetorics and grammar. Concerning lexical richness, there also appeared to be differences. Lexically, the reconstructed versions were always less varied than the originals. Some reconstructed versions were rather poor, while others were lexically more varied. Lexical simplification strategies also played a part, since the frequency category of the substituting items in the reconstructed texts was always higher or the same as that of the lexical items in the originals. Regarding the connection between these strategies and the grammatical complexity and lexical richness of the reconstructed texts, no positive effect has been found. This may be due to the cross-sectional design of the experiment. It seems likely that some time passes between the first noticing of a structure and the actual use of it. So whether group interaction results in the end in a more complex and richer use of language may have to be verified in a longitudinal study. But there are other factors that complicate the study of the exact nature of the effect of interaction on text quality. Firstly, it seems likely that both the amount of time spent on discussions on grammatical and lexical problems and the kind of strategies used by the learners are affected by the learners’ level of L2 proficiency. In this study we assumed that the 186 KUIKEN & VEDDER more learners discuss grammatical and lexical issues, the more complex and lexically richer the text will be. This idea could not be confirmed. In principle it is even possible that some learners may not need to talk so much about grammatical and lexical problems, simply because they know that what they are writing is correct. In this sense, text quality and use of strategies may be determined by L2 proficiency. One may wonder however whether this was the case for our learners, as the same learners presented a different behavior on text 1 and text 2. This brings us to a second factor that may influence the research findings, namely text difficulty. It is possible that the number and kind of strategies used by the learners depend on the complexity of the text. A similar text effect could be noticed in our analysis of the reconstruction of content (section 5.2.1), where we demonstrated that the ESL and ISL learners were better in reproducing content in one text than in the other. A third factor has to do with what may be called group dynamics. As stated by Long (1996), interaction seems to take place especially in case of an information gap, when learners with different levels of L2 knowledge question each other’s linguistic proposals. For less proficient learners, this information gap is certainly an advantage: they are able to take profit from the correct solutions proposed by more advanced learners. The opposite, however, may not always be the case: an incorrect structure proposed by a less proficient learner may be accepted by other learners simply because he or she has a more extroverted personality and more social pres- tige. Although our study was not focused on this kind of group dynamics, we noticed that the role of some of the learners was more passive during the reconstruction phase, while others played a much more active part in the discussions. One may wonder whether learners necessarily have to engage actively in interaction in order to learn. In a study on task-based interaction and incidental vocabulary learning Newton (1995) found that explicit negotiation of word meaning appeared to have less impact on vocabulary acquisition than use of words in the communicative process of completing the task. However, this doesn’t mean that negotiation was of no use; it may be, for example, that interaction takes place more often in case of more problematic words that are more difficult to acquire anyway. It is also possible that a discussion of a certain linguistic item (e.g. the correct use of a preposition) may lead to an increased awareness in other areas (for instance word order). As a consequence, the hypothesized increase in metacognitive awareness may not be the same for all learners and the effects may be both direct and indi- rect. Furthermore, particular forms that are used during the interaction may not necessarily occur in writing. Another complicating factor has to do with the general problem of measuring the role of interaction in L2 writing. In the current study, we have distinguished various strategies and a number of measures on grammatical and lexical complexity. The results show that in this way it is difficult to demonstrate the effects of interaction. For that reason, the design of the research will be modified in a follow-up study. We will then distinguish between an experimental group that is required to interact while processing input and a control group that is not. We will limit our focus to the acquisition of one or two particular linguistic structures. In order to say something about the benefits of interaction, the pre-existing linguistic knowledge of the L2 learners COLLABORATIVE WRITING IN L2 187 will be measured on a pre-test, whereas gains in proficiency will be determined by means of a post-test and a retention test after the treatment has taken place. Besides a quantitative analysis of the data, we will also make a detailed qualitative study of the discussions during the reconstruction phase. Example (3) in section 2 may serve to illustrate how a qualitative analysis of lexical and grammatical discussions may lead to an increased linguistic awareness. In this example Nico, Bart and Rona were trying to find the Italian equivalent of the word ‘fans’ or ‘supporters’. As we have seen, Nico didn’t know the word and asked his fellow-students. Bart provided the right word: tifosi. Although Nico didn’t seem to know the word tifosi, he didn’t want to use the word ‘fans’. In the end Nico was convinced by the others to use tifosi: after Rona’s approval and Bart’s reaffirmation of it, Nico decided to write it down. In the same excerpt the three learners were discussing a grammatical problem: the plural of sport. Nico was not sure how to formulate in Italian the sentence: ‘there are many fans of these two sports’. Bart suggested di and Rona agreed with him. Nico then asked: di questi? and di questi due sport, sporte?, showing his doubt about the plural of sport. Bart, who knew the grammar rule, explained to the others: ‘sport, because it is a foreign word.’ In both cases Nico signalled a problem and Bart happened to know the answer, giving additional evidence by formulating the rule regarding the plural of sport. Nico and Rona, who became aware of this rule, might remember it, even if they heard it for the first time. Maybe they would use it the next time in a similar context. What is fundamental is that in the interactional process, in a joint effort, learners search for the correct solution. In producing the target language they may notice the gap between what they want to say and what they can say. Group interaction may therefore have triggered the process of noticing. This noticing may then promote language acquisition. A detailed qualitative analysis may reveal if this is indeed the case. Franken and Haslett (this volume) also demonstrated that measures related to the quality of arguments showed a significantly positive effect from talk. In anticipation of further empirical results, a collaborative writing task, such as dictogloss, seems to have a number of pedagogical advantages. Language skills are not practiced in isolation, but are integrated. Learners become aware of their linguistic shortcomings and are encouraged to help each other and to search together for a solution. 188 KUIKEN & VEDDER

APPENDIX 1

AT HIGH SPEED! (ORIGINAL VERSION) Car racing and cycle racing are two very popular sports Italy. Many fans follow the racing cars engaged in the Formula 1 races in Italy or abroad. The most important car races are the Grand Prix of San Marino, on the track of Imola, in the province of Bologna, and the Gran Premio of Monza, near Milan. Italy’s most important cycling contest is the Giro d’Italia, drawing large numbers of people alongside the roads where the cyclists pass. The Giro is hold in June, it lasts about twenty days and it goes from the Alps down to Sicily. Since a couple of years their exists also an international Giro d’Italia for women. The color of the pullover of the one who arrives first is pink, just as for the men. Cycle racing is not only followed while seated in an arm-chair, but is actively practised, especially in northern and central Italy. In the evening, back from work, many persons get on their bicycle to cycle for a couple of kilometers. Others prefer to go jogging, before or after work. Even if men still have a majority, women have taken up these sport activities more and more. Today people take much care of their body and to keep themselves in form, they go to various sport schools to practise body building, a kind of physical training done with apparatuses, regulated on individual force.

AT HIGH SPEED! (RECONSTRUCTED VERSION) Sports, especially cycle racing and car racing are very popular in Italy. Four examples of car racing are the Fomula Uno, the Gran Prix of San Marino, the Grand Pre- mio of Monza and the Grand Prix of Imola. The most important cycling race is the Giro d’Italia. This race is hold in June and goes from the Alps down to Sicily. There is also an international cycling race for women. For everybody, also for men, the color of the clothes is pink. In general cycle racing is very popular, especially in central Italy, for instance after work and in the evening. Two sports to keep yourself in form are bodybuilding and gymnastics. What matters here is individual force.

INVESTIGATING LEARNERS’ GOALS IN THE CONTEXT OF ADULT SECOND- LANGUAGE WRITING

ALISTER CUMMING, MICHAEL BUSCH, & ALLY ZHOU

Ontario Institute for Studies in Education/University of Toronto, Canada

Abstract. Focusing on goals in the context of adults learning to write in English in university contexts, we first review the literature published on this topic and relevant psychological theories. We then present analyses of interview and stimulated recall data from 6 adults with various language backgrounds preparing for studies at a university in Canada. In this context, learners seem to conceptualize their learning of writing primarily in terms of learning the English language (e.g., as vocabulary and grammar), of heuristic strategies that aid composing (e.g., planning, editing, or writing quickly), and of gaining mastery over conventional types of written genres (e.g., argumentative, narrative, or other kinds of expository texts). Although presumably related to the curriculum the participants were studying, these goals for learning to write in a second language are inherently complex and diverse as well as contextually and personally situated. We propose criteria for identifying the structure of goals in this domain: the goal itself (e.g., as an aspect of language, rhetoric, composing processes, or genres); the origin of the goal, dilemmas related to it, and motives guiding it; situational conditions the goal appears in; responsibility for fulfilling the goal; materials, resources, or people that may assist in realizing the goal; and strategies for acting on the goal. We suggest various prospects for principled future research on goals for second-language writing.

Keywords: goals, English as a second language, writing, adults, university contexts, interviews, stimulated recalls.

1 INTRODUCTION Understanding the goals that learners have to improve their writing in a second language is important not only for the design of appropriate instructional approaches and curriculum policies but also for advancing knowledge about the psychology of learning in this domain. With these purposes in mind we conducted the case study documented in the present chapter to describe the goals that a sample of adult learners of English as a Second Language (ESL) expressed about their writing. We aim to use this information to prepare for future research that could analyze such goals lon-

A. Cumming, M. Busch, & A. Zhou (2002). Investigating learners’ goals in the context of adult second-language writing. In: G. Rijlaarsdam (Series ed.) & S. Ransdell & M. Barbier (Volume eds.), Studies in Writing, Volume 11: New Directions for Research in L2 Writing, 189 – 208. © 2002. Kluwer Academic Publishers. Printed in the Netherlands. 190 CUMMING, BUSCH & ZHOU gitudinally as adult ESL students progress through their academic studies at university.

1.1 Studies of goals in second-language writing Several studies have demonstrated that goals serve as a viable, effective focus for learning and instruction in second-language writing. For example, Cumming (1986) prompted, through demonstrations of writing processes and feedback on their compositions, a class of engineering students to individually select and monitor several goals each for improving their ESL writing – finding most could readily do so in this context. After several weeks of practice applying their goals while they composed, most of these students could sense when they had accomplished the goals, then they could select personally-relevant new ones to act on. In many instances, improve- ments in the qualities of the students’ written texts, as indicated by analytic ratings comparing their pre-post composition tests, proved to relate directly to the type of goals the students had chosen and acted on (i.e., in respect to rhetorical, logical, or grammatical features of their texts). Hoffman (1998) followed similar procedures with ESL university students in a writing course in New Zealand, modeling exemplary writing behaviors then prompting students to set relevant goals and to monitor and document them. She found the goals that students selected focused either on qualities of their written texts or on their strategies for composing. The students used a variety of different tactics to strive to achieve these goals, ranging from uses of resource books and assistance from others to self-monitoring and organizational activities. But the students varied in their successes with their stated goals, depending on their abilities in English and in writing, the effort they invested in the goals, as well as the extent to which they were able to reflect realistically on their own literate behaviors. Haneda (2000) appealed to the concept of goals, within the framework of activity theory (Leont’ev, 1979), to explain differences among Canadian learners of Japa- nese in their uses of feedback from teacher-student conferences to revise drafts of their written texts, depending on the differing motives that individuals had for learning Japanese and the proficiency they had attained in that language. Donato and McCormick (1994) – likewise conceptualizing goals within the framework of Vy- gotskian sociocultural theory – found students declaring and refining personal goals for improving their French in the U.S., as well as reflecting on their strategies for realizing them, in the context of their writing regular self-assessment entries in a portfolio in a university-level French course. Rather than conducting any kind of sensitization or training with the learners (as in Cumming, 1986 or Hoffman, 1998), Donato and McCormick observed these behaviors to emerge in the context of students writing up individual portfolios. Students tended to become more focused in their goal and strategy development over time as well as better able to connect their self-assessments to evidence about their language performance, seemingly because of the mediating influences of regularly reflecting on and writing about them. Considering educational situations more broadly, all adult learners probably use some kinds of goals, with varying degrees of self-awareness and explicitness, to INVESTIGATING LEARNERS’ GOALS 191 guide their learning, and these probably differ greatly from person to person and from situation to situation. People’s goals may draw on a range of source influences, but perhaps most conspicuously students might try to match the aims modeled, suggested, or advocated by their instructors, for example, as has been shown in detail for second-language writing in the context of dialogue journals (e.g., Nassaji & Cumming, 2000; Peyton, 1993). Nonetheless, surprisingly little is known about this phenomenon in respect to second-language writing. This lack of information is un- settling if one considers that other research on second-language writing has shown (incidentally) that learners may vary their goals for task performance greatly, depending on their perceptions of the value or purpose of a writing task (Casanave, 1992; Cumming, 1995; Leki, 1995; Prior, 1998; Silva, 1992; Spack, 1997; Uzawa & Cumming, 1989). Moreover, instructors may hold or implement radically different views of learning to write, even within the same educational program, if they differ in their understandings of the goals and situations of their ESL students (Cumming, 2001, In Press a; Leki & Carson, 1997; Shi & Cumming, 1995; Zamel, 1995). In- deed, Leki (2001) reviewed several published case studies of ESL learners to highlight how the personal goals these individuals had for writing were thwarted by the more powerful goals asserted in the educational contexts in which they found themselves. These concerns are especially perplexing in view of the position that statements about intended outcomes for learning (stated in terms that resemble goals) have come to take around the world in curriculum policies that prescribe specific standards, benchmarks, or target attainment levels that learners are expected to achieve uniformly in their programs of study. As Brindley (1998), Cumming (2001), and Valdes, Haro and Echevarriarza (1992) have argued in respect to second- language writing, scarcely any empirical research or theoretical foundations inform these curriculum frameworks, despite their making relatively explicit pronounce- ments about what students should be able to do in writing at fixed stages in their studies. This is a major pedagogical reason why research is needed to determine how learners themselves conceptualize their goals for learning (or other, related goals), what efforts they and their instructors might make to achieve them, whether they are in fact achieved in the forms prescribed, and what the extent of variation might be among these elements.

1.2 Psychological theories of goals The concept of goals has a strong foundation and tradition in the psychology of learning. For instance, Austin and Vancouver (1996) and Locke and Latham (1990, 1994) provide thorough reviews of previous research on goal constructs, demon- strating the extent of their history and conceptualization; their linkages to major theories of social psychology, such as self-efficacy theory (Bandura, 1986), control theory (Carver & Scheier, 1981), or situated cognition (Brown, Collins & Duguid, 1989); and their value in guiding the design of empirical inquiry into learning. Re- lated concepts are also integral to John Dewey’s conceptualizations of motivation for learning under conditions of reflection, effort, and interest, for example, as described by Prawat (1998: 215-217). Goal theory provides a theoretically rigorous 192 CUMMING, BUSCH & ZHOU means of identifying people’s learning intentions in phenomenologically grounded terms arising from learners’ self-evaluations of their task performance and purposes. Moreover, goals can form a principled basis to assess learning for individuals and within or across groups as well as a way of tracing changes in learning over time and in differing contexts. In certain domains of education, such as mathematics instruction, much concrete evidence on the value of goal setting has accumulated. For instance, Ames (1992), Fuchs, Karns, Hamlett, Katzroff, and Dutka, (1997), and Schunk (1996) have demonstrated that learners trained to use goals for self- evaluation or task mastery can significantly improve their performance on mathematics problems. Likewise, Vye, Goldman, Voss , Hmelo, Williams, and the Cogni- tion and Technology Group at Vanderbilt (1997) have found close relations between the goals students set for solving complex math problems, the logical arguments they formulate to interpret these problems, and their successes in solving the problems accurately. Two conceptualizations of students’ goals for learning are particularly relevant to written composition and to the present study. One is Bereiter and Scardamalia’s (1989) elaborations, in regards to learning in a variety of domains, on their extensive studies of written composition (Bereiter & Scardamalia, 1987). They argue for the value of a particular orientation to learning they call intentional learning, grounded on concepts of goals rather than of strategies. For example, Ng and Bereiter (1991) distinguished three types of goals in analyses of children’s approaches to algebra problems: (1) task completion goals (which involved just accomplishing given tasks), (2) instructional goals (aimed at determining the main points of teaching, presented either implicitly or explicitly to learners), and (3) knowledge-building goals, based on individuals’ efforts to construct their personal sense of phenomena or the world. They argued that the latter type of goals serves people’s learning more significantly than do the other two types of goals, which are typically promoted through educational tasks, but which quickly become routine and involve minimal mental effort (cf. Doyle, 1983). For studies of written composition, Bereiter and Scardmalia (1987) argued that the latter type of goals is foundational to a psychological model of writing as knowledge transforming, arguing that it bears the potential for learning that could truly lead to expertise, unlike the more commonly adopted, basic model of knowledge telling to compose (i.e., just telling one’s knowledge about a topic) that is widely promoted in school activities for writing. Bereiter (1997) argued that learners need to utilize a knowledge-transforming mode of thinking to be able to construct knowledge and abilities that transfer across the bounds of simply performing particular tasks. Cumming (1995) argued for applications of these ideas to teaching second-language writing. A second, important conceptualization of goals appears in Leont’ev’s (1972, 1979; Engeström, 1991; Engeström & Cole, 1997; Engeström, Miettinen & Puna- maki, 1999) activity theory, aspects of which numerous theorists have recently adopted to explain the teaching and learning of writing (e.g., Russell, 1995), literacy (e.g., Winsor, 1999) and of second languages (e.g., Lantolf, 2000). For instance, Gillette (1994: 210) claimed that the ‘differences in L2 achievement’ she documented among adult learners of French arose ‘primarily as a function of student goals in the course of instruction’ depending on each ‘learner’s social history and INVESTIGATING LEARNERS’ GOALS 193 the use value ascribed to foreign languages in his or her environment.’ Leont’ev (1972, trans. 1979: 46) defined ‘activity’ in Soviet psychology as the central ‘unit of life’, which functions to ‘orient the subject in a world of objects,’ forming ‘a system with its own structure, its own internal transformation, and its own development.’ Activity theory places people’s goals for actions in a central position in reference to their broad motives for structuring an activity, the particular operations they undertake to do so, the objects they orient themselves toward to perform an activity, and the prior events they have experienced that enable them to perform in the manners they do. Activity is a fundamentally holistic conceptualization of human behavior situated in a sociocultural context. But people’s goals are but one aspect of this complex system, so they must be understood in reference to the overall activity structure and the circumstances under which people act (Leont’ev, 1972, trans. 1979: 60-66; Zinchenko & Gordon, 1976, trans. 1979: 99-101, 124-127). To promote applications of activity theory, Leont’ev (1972, trans. 1979) described several key characteristics of goals: Goals direct and connect people’s actions (p. 60); goals exist, not in a finite state, but as series of multiple, related goals (p. 61) and as a long-term process of ‘testing goals through action’ (p. 62), some goals may only be evident retrospectively (i.e., after an action is completed, p. 62), people can consider their goals abstractly and consciously (unlike the details of their actions, pp. 62-63), and goal formation is integral to human activity (though laboratory-type tasks often remove this condition, thus unwittingly stripping away integral dimensions of the psychological phenomena under study, p. 62).

1.3 Implications for research on second-language writing These psychological conceptualizations of goals have important implications for studies of second-language writing. First, people’s goals have differing characteristics, which may affect their abilities to perform, learn, or change. For this reason, understanding the goals that students have, for example, for second-language writing, is fundamental to understanding not only how they approach learning but also to evaluating how well they can do so, or be helped to do so, under the conditions they experience. Second, goals integrally relate to (and may perhaps even determine) the strategic operations that people undertake in performing specific tasks. For this reason, it may make little sense to try to document strategies for performing tasks like second-language writing or characteristics of the texts that people produce, in general terms or in the abstract, unless they are analyzed in reference to the goals people have to motivate and guide their task performance as well as other essential aspects of these activity structures and the contexts in which they are embedded. In Leont’ev’s terms, most research on second-language writing has addressed either the level of strategic operations for the activity of writing (i.e., composing processes) or the resulting products of the activity (i.e., written texts), but not the full, integral nature of the activity itself. That is, nearly all of the many accounts of composing processes or of texts written in second languages that have appeared in recent decades (e.g., reviewed in Cumming, In Press b; Silva, 1993) have been conducted in contexts created for the purposes of research (i.e., controlled experimental tasks) 194 CUMMING, BUSCH & ZHOU which do not preserve the integrity of the natural activity system (e.g., defined by personal motives, goals, etc.) in which people actually write in a second language. As a consequence, this research depicts a limited view of what second-language writing is, perhaps more akin to the standardized contexts of tests or examinations rather than to what people experience in academic studies, work, or other settings (Parks & Maguire, 1999). Such research is not therefore able to adequately explain why people write the way they do, nor how they may learn to do so (except perhaps to perform a task under the terms commanded by researchers, or as an instance of the transfer of literate abilities generally across bilingual contexts, as argued in Cummins, 1991). A related limitation is the artificial separation of learning from teaching in studies of second-language writing (e.g., observed in Cumming, 1998; Cumming & Riazi, 2000, Grabe, 2001; Grabe & Kaplan, 1996). Without a theoretical framework (such as activity theory) to link teaching with learning, researchers have studied each as independent aspects of second language writing, not knowing quite how to explain one in terms of the other, despite their obvious interrelatedness as activities central to the enterprise of education. (But intimations of their interrelatedness have appeared in, for example, studies of discrete instructional functions, such as tutoring in Aljaafreh & Lantolf, 1994, dialogue journal writing in Nassaji & Cumming, 2000, or feedback on writing in Conrad & Goldstein, 1999.) A third problem is that of accounting for sources of variability among individuals and across sociolinguistic contexts for learning to write in second languages (Connor, 1996; Hornberger, 1989; Li, 1996). The situations for writing in second-languages are so diverse internationally that the characteristics of any one group of learners in any one setting need to be accounted for systematically in considering how findings from research with them might apply to other situations. To counter these limitations, research on second-language writing needs to move toward analyses of the full, naturally-occurring activity systems in which people experience writing in a second language. Particularly, such research needs to document learners’ goals and motives in the contexts of their specific social conditions – to demonstrate how these relate to the actions people perform while composing, the uses they make of instruction and available resources, the characteristics of the texts they produce, the achievements they make over time, and the major sources of variation with sociocultural contexts. These purposes are in line with arguments expressed by Rueda and Moll (1994) and Trembley and Gardner (1995) for studies of motivation in the psychology of second-language learning generally, by Leki and Carson (1997) in respect to ESL writing in particular, and by Waters (1996) in reviewing research that has surveyed the needs for learning among ESL students in North American universities. We undertook the present study intending it to be an initial step in these directions.

2 METHOD Seeking case study data to prepare a preliminary description of the goals that adult ESL learners have for improving their writing in English, we solicited volunteer participants from a local ESL program, gathered from them personal profile, inter- INVESTIGATING LEARNERS’ GOALS 195 view, and stimulated recall data (based on a sample of their recent writing), then transcribed, coded, and interpreted these data.

2.1 Participants We sought adult learners of English with a variety of backgrounds, experiences, and motives for studying English. Six young adults, described with pseudonyms in Table 1, volunteered to be interviewed for about 1 and a half hours individually in return for our producing a written profile of their goals for ESL writing and our providing advice to improve their writing after the research was completed. They ranged from 17 to 33 years of age, represented 5 different mother tongues, had studied previously in different educational systems in diverse parts of the world, had resided in Canada or the U.S. for periods from a few weeks to 1 year (though Ali had lived in the U.S for 4 years as a child), and intended to pursue different future academic studies and careers. All were studying in the same pre-university, intensive ESL program at a local university, and (judging from their class placements and the TOEFL scores they reported, in Table 1) their proficiency in English was just at the point where they would be able to begin academic programs in English at the university. The participants all appeared to have sufficient proficiency in English to conduct the interviews competently, though they displayed some minor constraints in their abilities to express themselves in English.

Table 1. Profile of participants.

Person Age L1 TOEFL Degree Occupation Career plan

Gertrude 33 German 257* Diploma in Transportation Unsure engineering engineer Ali 23 Farsi 590 University Student of engi- MBA program neering then manager Claudia 17 Italian None High school Student Laywer Yun-hui 24 Korean 620 B.A. Student International relations Kanaporn 23 Thai 613 B.A. Translator Accountant in trading company Eun-Joo 28 Korean Not B. Sc. Engineer Teach English available

* Gertrude’s score of 257 was on the recent Computer-based TOEFL (Test of English as a Foreign Lan- guage), whereas the TOEFL scores for Ali, Yun-hui, and Kanaporn were on the older, pencil and paper version of this test, which has a different scale. The scores here are all within a similar range across the two scales for this standard test of English proficiency. Claudia had not taken this test and Eun-Joo did not report her score to us, though both students had been placed in the same class through placement tests in their ESL program. 196 CUMMING, BUSCH & ZHOU

2.2 Instruments We developed and refined, through two rounds of piloting on ourselves then with several graduate students, a standard interview schedule. This consisted of three parts: (1) a profile of each person’s prior education and experiences learning and writing English; (2) a series of open-ended questions, followed by prompts to elicit examples, about aspects of English writing each person might be trying to improve (in general and in respect to specific types of written texts, processes for composing, revising and editing, using sources of information, and the grammar and vocabulary of the English language), the people and situations they interact with when they write in English, and their future plans for academic studies and work; and (3) a stimulated recall (cf. Gass & Mackey, 2000; Smagorinsky, 1994) in which each participant described, paragraph by paragraph to the interviewer, a recent piece of writing in English they had brought to the interview, prompted by questions about what they were trying to achieve in the writing, how they did this, and how they might have done so better. One of us (Busch) conducted all of the interviews over a period of several weeks, audio-taping them.

2.3 Analyses After transcribing the audio-taped interviews in full, we analyzed them using a constant-comparative methodology (following procedures in Miles & Huberman, 1994) to develop a scheme, grounded in the interview data, to describe the nature and structures of goals reported by all 6 ESL learners. Each of us (the 3 researchers) reviewed the transcripts, proposed categories and subcategories to describe the goals reported in the interviews, then we met to revise and refine these categories several times while applying them to samples of the transcribed interviews. We treated the turn (i.e., each student’s full reply to an interview question) as our unit of analysis, using polytonic coding for each turn (i.e., multiple codings of each turn, where appropriate). One of us (Zhou) then attempted to code samples of the data using the computer program NVivo (Qualitative Data Solutions and Research, 2001) to depict the hierarchical structure that emerged among the coded data. Once she was relatively satisfied with the codes we had developed, and she felt able to apply them reliably, a second member of our team (Busch) independently coded in NVivo sections from the beginning, middle, and end of each person’s interview transcript (forming about 20% of the total data). Finding the two researchers agreed in their categorizations on 93% of the 290 turns they had independently coded, Zhou then coded the remaining data in NVivo.

3 FINDINGS Our principal findings appear as the taxonomy of goals in Appendix A. The ordering of this taxonomy distinguishes (1) goals, (2) situational conditions for realizing goals, (3) orientations to goals, and (4) dilemmas associated with goals. This is a considerably more complex and situationally embedded image of goal structures related to second-language writing than we had expected to document, but we be- INVESTIGATING LEARNERS’ GOALS 197 lieve this four-fold framework is necessary to do justice to the conceptualizations of their goals that our informants conveyed during the interviews. Each participant reported between 186 to 293 statements that we coded as relating to their goals for learning to write in English, producing a total of 1,536 coded statements in the full data set.

3.1 Goals The goals expressed by the interviewees were mostly immediate goals for improving their writing in English, but some goals also referred to future, anticipated requirements for academic studies, particular tests, or intended careers. (See Appendix A.) We distinguished 27 subcategories of immediate goals for ESL writing: The most frequently mentioned concerned improving vocabulary (e.g., ‘I’m trying to expand, develop my vocabulary’) or grammar (e.g., ‘to improve my grammar’). Participants also referred frequently to improving various aspects of their rhetoric (e.g., ‘how to organize, ya, my idea before getting to write’), composing processes (‘to improve my speed’), and abilities to handle specific genres of texts (e.g., ‘how to compare two things’). Their future goals referred either to their intended academic studies (e.g., ‘I will go to an accounting program’), passing particular tests (e.g., ‘I would like to take the TOEFL after this course’), or career plans (e.g., ‘I want to work for international organization like the U.N.’).

3.2 Situational conditions for goals The interviewees typically described their goals for improving their writing in Eng- lish in reference to particular situations. The largest numbers of these statements referred to the students’ present situations in ESL classes (e.g., ‘The ESL program. We do a lot of writing.’), though numerous references were also made to past situations (e.g., ‘I’ve had to do presentations at my university.’) as well as contexts such as particular places for composing (e.g., ‘in my room’) or writing for tests, at work, or in academic courses. In considering this aspect of the students’ goals, we tried to distinguish whether participants described their goals as originating with themselves, their peers, their teachers, or work activities. The origin of most goals appeared to be associated with the students themselves in the context of their being students (e.g., ‘I’d like to improve my speed.’ ‘I want to make my writing more interesting.’), though some also seemed to originate from teachers (e.g., ‘Even the teachers have been telling us how to write notes.’), and a few goals were described as originating from peers (e.g., ‘That’s where my friend came to help me. And she said that I should start with the thesis and write the first sentence, which says what I’m going to talk about.’). We did not identify any instances of their current goals that had originated from work activities, though some of the people with work experience alluded to these situations during their interviews (e.g., ‘I know how to write a good, clear letter, and write a business contract.’ ‘If I’m a manager, I have to give the special writings to my higher manager. Or if it’s to my lower employees, I have to hand in, to tell them what to do and what not to do.’). 198 CUMMING, BUSCH & ZHOU

3.3 Orientations to goals We further distinguished how the interviewees said they oriented themselves to their goals in respect to (1) responsibilities for their goals and (2) strategies used to realize goals. In terms of responsibilities, we coded all but one of the 162 statements referring to responsibilities for goals as indicating that the interviewees perceived themselves as the agents to fulfill the goals (e.g., ‘I try, always to make short sentences, try to make things simple and logic, always’). Only one statement appeared to refer to a teacher as having some responsibility for a goal, but even that was superseded by the student saying she had later assumed her own responsibility for the goal (‘Ya, she [i.e., the teacher] did, but I develop my own way of doing it. So I think, I mean for me it’s better.’). We did not find any instances of any of the students mentioning that their peers or any other people might have responsibilities for their goals. The strategies that the students described to realize their goals were extremely varied. We categorized these at a global level in terms of: (1) seeking assistance from particular types of people, (2) self-regulation, (3) stimulation, (4) use of tools, and (5) language practice. The people that the students reported seeking assistance from included, in relatively equivalent numbers, their peers (e.g., ‘Usually, sometimes when we write something, we just share them with our classmates.’), teachers (e.g.., ‘No one checked it. Only the teacher.’), and various other people (i.e., tutors, relatives, friends, e.g., ‘Sometimes the stories I wrote I sent to my sister. She said, you had a lot of verbal problems.’). A range of strategies involving self-regulation was mentioned. The most numerous of these, cited several times by each inter- viewee, were: to make an outline prior to writing (e.g., ‘Ya, I make an outline, of course, and I try to keep it.’), editing and revising drafts of compositions (e.g., ‘Ya, after I finish writing and maybe two times, at least two times I read it again and check some mistakes.’), and planning (e.g., ‘Ya, well, I actually think about what I want to say. That’s my, that’s my. Sometimes it helps when I think deeply about what I want to say.’). Less frequently mentioned were strategies for regulating themselves through: trying not to confuse ideas, writing simply, using visual supports, synthesizing ideas, using examples, using different words from source texts, and trying to be neutral on controversial topics. Strategies mentioned for stimulating themselves to realize their goals varied from person to person, but these included playing music, drinking coffee, walking around, and talking with people. The interviewees also referred to various mediating tools they made use of for their writing (cf. Wells, 1999). All the students mentioned books and magazines, computers, dictionaries, and newspapers, and some also cited crossword puzzles, stationary, closed captioning on TV, and the Internet. Most of the interviewees also cited practicing to use English as a strategy, notably in reference to writing (e.g., ‘I think the best thing is write as much as you can.’), reading (e.g., ‘I try to improve my writing by...read more books.’), and improving vocabulary (e.g., ‘I want to improve my vocabulary so now I read a lot of books.’). INVESTIGATING LEARNERS’ GOALS 199

3.4 Dilemmas related to goals Certain of the statements the students made about their goals were phrased in the form of dilemmas, problems, or challenges that they perceived themselves facing. By recognizing problems in a particular aspect of their abilities to write in English, the students may be preparing a foundation to form goals later that they might try to fulfill (e.g., ‘I have problems with vocabulary. That’s my huge problem.’ ‘I never think what I want to write. I know the subject. I just write it down.’). Or, in some instances, they may have been indicating that they had already achieved particular goals (e.g., ‘So I think I don’t really have problems for grammar or sentence structure.’). In other instances, such statements may indicate that the students felt they need help to guide themselves (e.g., ‘Actually, I don’t know how to improve my weak points.’; ‘I don’t know... I don’t even know what are my problems.’). As evident in the preceding examples, many of these dilemmas related to English grammar or vocabulary, each of which are enormously vast and complex domains to contem- plate learning. The diverse foci of other dilemmas (see examples in Appendix A) involved: aspects of rhetoric, such as developing logic and ideas, organizing texts, and adhering to stylistic conventions; composing processes, such as getting started on a composition, generating ideas to write about, editing and revising, taking too much time to write, and managing time; working with particular text types (notably those assigned in their classes, it seems), such as free writing or comparison, argumentative, or narrative essays; or dealing with diverse issues such as directness, poor handwriting, or adhering to formulas for composing. (These categories mostly du- plicate those described above as goals or strategies, but we have highlighted them here because they were expressed explicitly as dilemmas.) Interestingly, all of the students made several statements, typically phrased as dilemmas, about their affective states related to their writing (e.g.,’I don’t like to show my writing. I cannot stand criticism.’ ‘Maybe I’m too panicked. You know, it’s just hard for me to do something again and again.’).

3.5 Complexity of and variability in goals The four categories above represent central features of these goals in a taxonomic fashion. But viewing them in this atomistic way also obscures the relations among these aspects of goals, the relations between these variables and each interview case as a whole, the variability evident from person to person, and the inherent complexity and uniqueness of each person’s goals and the dilemmas they associated with them. For example, when asked the initial question in their interviews, ‘In general, what are you trying to learn or improve in your writing in English?’, each person gave a qualitatively different kind of response. Claudia described how her goals focus on improving her sentence structure and word choice while writing in English, trying to reach standards she has in her L1 writing: Okay, I think I’m trying to improve my style. Correcting my free writing is okay. I mean my grammar is not bad now. So I’m trying to give a lot a kind of style. It’s something I like. When I write in my own language, I pay a lot of attention to the details or 200 CUMMING, BUSCH & ZHOU

the style. I can’t really do it now in English cause I, you know, I’m just starting to learn it...Like, ah, the structure of the sentence. In, I mean, I really like to write complex period, complex sentences and ... using like particular words, more expressive words. (Claudia) Yun-Hui similarly expressed her goals in regards to English vocabulary and her expressive abilities, but she emphasized her future intention to study at graduate school: Generally, ah, kind of hard to say in general, because I have specific reason to, um, you know, improve writing in English. Cause I’ll be able to go to graduate school. I know my speaking in English is not perfect. It’s not good enough. But so my speaking is better than writing, I think. I want to...write... I mean, ya. Sometimes when I try to write something in English, I feel that I don’t know many vocabularies. So even though, ah, you know, this is so hard to express my opinion or something. (Yun-hui) Kanaporn referred to how she was trying to read widely to improve her English vocabulary and knowledge and thereby expand her world view: I try to improve my writing by...read more books. Ya. And then trying to broaden my view because I need only prefer the fiction books, novels, now. I try to read more like science or non-fiction, and history. Many, many things... I read to gain know, to know more about everything around, the story and the vocabulary, ya. (Kanaporn) Eun-Joo similarly emphasized goals of vocabulary acquisition through reading as well as uses of mnemonics and improving her argumentative abilities, but she observed how her instructors’ feedback had shaped these perceptions: Ah, I want to improve the academic vocabulary especially. From last course my writing teacher have a comment about my writing: a little bit, not so big vocabulary. And, also, ah, my thinking is now so. I mean the writing itself. I have a lack of develop my opinion. And so now I’m trying to a...to read a lot of books, more books. And...and then, there are lots of new vocabulary. And try to memorize all of them. Like a writing, writing several times, and say the new word several times. Ya...I don’t like keep a list. I don’t keep a list when I read new words. I try to memorize and say every time. (Eun- Joo) Gertrude focused on her composing processes, addressing how she was trying to improve her fluency and control over, and self-satisfaction with, her planning and editing while she wrote in English: I’d like to, first of all, I’d like to improve speed. It takes me a long time to start. For example, if I write an essay too long, ah, like two hours I’m sitting... front of my paper. And so I’d like to...And also if I have done it, I’m not satisfied with the quality. Ah actually, it’s very rarely that I like my essays. So I’d like to improve that. If I see an essay of another person, I can tell if it is great or not, but if I see my essay, I always think it’s not good. (Gertrude) In contrast, Ali described how he had been trying to move his goals from reaching a certain score on the written component of the test of English used for university ad- missions to more academically oriented skills such as note-taking and composing essays: For the TOEFL exam we do have a test of English, TWE. The first thing that I’m doing is preparing myself for that. Other than that, you, I think, even if I, that’s not so hard to pass if you just the right, the... It has a certain rule, you know, learning that was not so hard. So you can just get an average number of 4 out of 6. But when we gain through INVESTIGATING LEARNERS’ GOALS 201

school how to take notes, how to write essays, I think these are much, much more important than TWE. You know, how to write a note, how to write an essay. (Ali) As these brief extracts from the interviews indicate, multiple considerations enter into even a simple statement of a person’s goals for writing improvement. The sub- stance and qualities of goals differ from person to person, and goals appear in different combinations. People’s orientations to their goals intermingle with their senses of their situational conditions for writing and self-development. For example, Eun- joo, Ali, and Yun-hui emphasize how the educational circumstances they are in, or they anticipate being in, have shaped their goals for writing improvement, whereas Claudia, Gertrude, and Kanaporn express their goals in terms that seem to arise immediately from their personal intentionality or self-concepts. The tools people call upon to assist them range from material resources, such as books (for Kanaporn, Eun-Joo), to self-control strategies or mental heuristics (for Claudia, Gertrude, Eun- Joo). Complexity and variability are more the norm in these data than are the commonalities or explicit hierarchies of relations suggested in the taxonomy in Appen- dix A.

4 DISCUSSION In the foreground of these students’ accounts of their goals for writing was their present situation in ESL classes. In this context, writing seems to be conceptualized primarily in terms of learning the English language (i.e., as vocabulary and grammar), of heuristic strategies that aid composing (e.g., planning, editing, or writing quickly), and of gaining mastery over conventional types of written genres (e.g., argumentative, narrative, or other kinds of expository texts). Correspondingly (and perhaps inevitably because they formed their current activity systems), the students stated their goals for learning in these terms. We marvel at the complexity, diversity, and enormous scope of these skills and abilities, wondering how the students are able to focus their resources and efforts into concrete goals that make learning them feasible. In this sense, we are surprised at the highly constructive view of learning second-language writing that these students expressed, conceptualizing their goals almost exclusively as their own responsibility to manage, often through self- regulatory strategies, though acknowledging that their goals may originate with themselves as well as their teachers, peers, and others as well as be realized with the aid of such people and diverse tools and resources. Although these learners seemed relatively reflective and cognizant of their current skills and abilities related to second-language writing, none seemed to have high levels of expertise in this domain, nor did we identify any instances in our data of goals that involved knowledge- transforming (in the manner described by Bereiter & Scardamalia, 1987, 1989), though there were hints of some students having strong motives for learning to write in English that may in the future lead them toward such goals, for example, to succeed in their academic studies or to realize their career plans. But the students did not express, at the time of our interviews, any concrete understandings of the requirements for writing in these contexts, even when they were prompted to speculate 202 CUMMING, BUSCH & ZHOU about them during the interviews, presumably because these were not part of their current activity systems. These observations, we believe, represent the major implications arising from these data and analyses: Goals for learning to write in a second language are necessarily complex and diverse as well as contextually and personally situated. These characteristics also feature in the four previous studies of goals in second-language writing reviewed at the beginning of this article. The importance of contextual situations, diversity, complexity, and personal responsibilities for learning feature in Cumming’s (1986) and Hoffman’s (1998) accounts of their sensitizing ESL learners to become aware of, select, and monitor personal goals to improve their writing as well as in Haneda’s (2000) and Donato and McCormick’s (1994) accounts of learners progressively developing, reflecting on, and acting on their goals for writing in Japanese and in French, respectively. These characteristics, however, contrast with the kinds of uniform outcomes recently proclaimed as standards, benchmarks, or attainment levels in many curricula for language education and other subjects around the world. Goals are embedded within activity systems. What then are goals in these data? In the process of coding the interview data, we observed that in order to identify something verifiably as a goal, we felt we had to be able to determine the substantive focus of the goal, the origin of the goal, the conditions the goal appeared in, who assumed responsibility for it, what strategies and resources might be used to deal with the goal, and what dilemmas the goal corresponded to. This suggests a certain minimal structure to the nature of goals in this domain, which may be helpful in identifying and tracing the development and reali- zation of goals in future studies of second-language writing. This structure appears to involve:

1) the goals themselves (e.g., in terms of language, rhetoric, composing processes, or genres), 2) origins of the goals, dilemmas related to them, and motives guiding them, 3) situational conditions the goals appear in, 4) responsibility for fulfilling the goals, 5) resources (i.e., materials, information, and people) that may assist in realizing the goals, and 6) strategies for acting on the goals.

In viewing these multiple facets of learners’ goals, we realize that a truly insightful study of goals in this domain would need such criteria to identify students’ goals then to trace how students act on them over time in their writing tasks and in courses of language study and subsequently of academic study. That is an aim for our future research. To this end, the present case study is able to contribute a basis on which we (and hopefully others, as well) might be able to identify and distinguish learners’ goals. However, the interview and recall data that we collected here provide only a mere glimpse of learners’ goals and the activity structures in which they may exist. We need to determine with greater precision the trajectory and timeline of each specific goal in the history of individuals’ development of their writing abilities in the second language. INVESTIGATING LEARNERS’ GOALS 203

Ideally, we might expect (following the tenets of activity theory) to see learners first formulate dilemmas related to their writing abilities, then to formulate these more concretely as goals, next to transform them to operations that they start to use while writing, then ultimately use them as tools that are routinely integral to their composing. But that remains a process to investigate and verify in the future. Similarly, describing the larger activity structures in which particular goals are embedded is a challenge, akin to putting into operational form the idea of a discourse community (cf. Beaufort, 1997; Berkenkotter & Huckin, 1995). Such research would require, in addition to interview and stimulated recall data (as in the present study), extended observations of classroom contexts, analyses of students actually performing writing tasks and of the written texts they produce, and information from other people, such as instructors and other students (and perhaps even family or employers), who also constitute the activity systems in such settings. In short, research on goals for writing in a second language needs to be conceived from the relatively panoramic yet detailed perspective on the contexts of writing in language and academic studies obtained, for example, in studies such as Prior (1998), Riazi (1997) or Spack (1997). We also realize, in retrospect, that there are various things that we should have done or asked about in the interviews in order to obtain a fuller image of people’s goals and their corresponding activity systems. One was the students’ attitudes toward responses to their writing, particularly the kinds and aspects of feedback they sought or found valuable, who they expected such feedback from (teachers, peers, or others), and how they made use of it. We should have also sought more detailed information on the students’ uses of tools – including communication technologies (e.g., computers or the Internet), information resources (e.g., dictionaries, books, etc.), and cognitive heuristics for composing – and how they made use of them to act on their goals. We realize as well that we should standardize our interviewing procedures to make them consistent for each student and to produce reliable, interrelated categories of goals; pursue more in-depth discourse analyses of the interview data; and refine our procedures for stimulated recalls to get more information from these and to know how people made use, in their studies and other situations, of the ideas they mentioned in these recalls. Another key issue requiring evaluation is people’s commitments to their goals (Austin & Vancouver, 1996). How firmly are learners tied to particular goals and how urgently do they feel they need to fulfill them? A further issue is determining how learners see particular goals in relation to others. The taxonomy of goals in Appendix A represents a hierarchical structure that we, as researchers, perceived in the data. But to be able to make sense of individual students’ goals we would need to know how each person fits their goals together, from their perspective, into a whole that puts a priority on certain goals over others to guide their activities. Two issues that intrigued us, and which surely warrant further investigation, are (1) the relations of learners’ goals to their formulations of dilemmas about their writing and (2) the links between the relatively rational status of goals and the more emotionally-based affective states which the present learners expressed frequently about their writing, and which as Clachar (1999) rightly charges, have been largely ignored in studies of second-language writing. Although the present analysis of goals for learning to write in a second language is considerably fuller than previous studies on this topic, it is inherently exploratory as 204 CUMMING, BUSCH & ZHOU well as bound to the interview contexts and the temporal situations of just these 6 ESL students. So it would be premature to attempt any applications of the goals listed in Appendix A to teaching. Indeed, we realize our next inquiry in this direction has to be descriptive, exploratory studies of ESL writing instructors to determine how they might conceptualize goals for their students’ learning, to determine the correspondences and differences between their conceptualizations and those of their students, and to map out how these fit into activity systems in their classrooms. Likewise, we need to consider how instructors in particular academic disciplines view goals for ESL writers in their courses. We are especially curious to see if these relations are as dissonant as, for example, Cliff (2000) portrays them to be for students undertaking university studies in general. A major challenge in thinking about applications of goals to educational practices will be to avoid the kinds of decontex- tualized, atheoretical proclamations about learning that have arisen from studies of communication strategies in second languages (e.g., as criticized by Firth & Wag- ner, 1997, and summarized in Kasper & Kellerman, 1997) or inventories of strategies for learning languages (e.g., as criticized by Ellis, 1994, and summarized in Oxford, 1990; Wenden & Rubin, 1987).

ACKNOWLEDGEMENT We thank the participating students and their instructor for their involvement and Merrill Swain for helpful comments on an earlier draft of this chapter. We are pres- ently continuing this research through Social Sciences and Humanities Research Council of Canada standard grant 410-2001-0791, Goals and Activities for ESL Writing: Learning and Instruction in the Pre-University/University Interface, for which we gratefully acknowledge funding to Alister Cumming.

APPENDIX

TAXONOMY OF GOALS AND OF CONDITIONS, ORIENTATIONS, AND DILEMMAS RELATED TO THEM, WITH EXTRACTS OF EXEMPLARY STATEMENTS FROM THE PRESENT DATA

1. GOALS 1.1. Immediate Goals for Writing 1.1.1. Language Kanaporn: Oh not the format. The language itself 1.1.1.1.Language/Vocabulary Claudia: I'm trying to expand, develop my vocabulary. 1.1.1.2. Language/Grammar Ali: I'm really not trying to improve my grammar. I want to, but you know, it's not like something that comes in first. 1.1.2. Rhetoric 1.1.2.1. Rhetoric/Logic and Ideas Gertrude: The second thing I try is not to confuse ideas. Just prove one idea and then the next. Doesn't matter if there is some similarities or not. 1.1.2.2. Rhetoric/Organization Kanaporn: Ya. And the how to organize, ya, my idea before getting to write because I have only 30 minutes or 40 minutes for the test to, ya, mainly xxx for the test. 1.1.2.3. Rhetoric/Style and Format Claudia: OK. I think I'm trying to improve now my style. Cor- INVESTIGATING LEARNERS’ GOALS 205

recting my free writing is OK. I mean my grammar is not bad now. So I'm trying to give a lot a kind of style. It's something I like. When I write in my my own language, I pay a lot of attention to the details or the style. I can't really do it now in English cause I, you know, I'm just starting to learn it. 1.1.3. Composing Processes 1.1.3.1. Composing Proc- Kanaporn: I have to improve my speed. esses/Writing Speed 1.1.3.2. Composing Proc- Ali: She did. Yes. [The teacher talks about planning and making esses/Planning outlines in class.] 1.1.3.3. Composing Processes/Editing Ali: My teacher said, you have mistakes in these. I was just and Revising checking those, but naturally I never go after my essay. You know. I just like to hand it in to someone and he reviews it for me. That's more, much more like, um I don't know… easier for me. 1.1.4. General Improvement Claudia: Um I think I need to improve everything. I mean you always need to improve your. But I think now I need to improve my vocabulary more than grammar. Not just vocabulary cause I need to improve my grammar too… 1.1.5. Content or Topic Eun-Joo: Anyway the teacher always gives us a topic. 1.1.6. Use of Tools Ali: Two fingers, but right now I'm trying to use, I don't know, other fingers. I trying on it. I'm trying to slow down because I can't really type real fast with two fingers. But I'm trying to slow down and use my other fingers too. Because I've seen people around here who just type so fast. 1.1.7. Genre 1.1.7.1. Genre/Academic Writing Gertrude: At the moment I'm going to do a Masters course at the University of Toronto, so I'm trying to improve my academic writing. 1.1.7.2. Genre/Essay Yun-hui: For me I want to, I wanna write a good essay. You know. For graduate school. And so that's way I chose the academic skills class. 1.1.7.3. Genre/Business Documents Kanaporn: I should start from essay and then a longer term paper. Reports and the paper and get to business writing. But I think business writing is not emotional, not creative. Just to be direct. 1.1.7.4. Genre/Reports Ali: We had to do a summary of a report. You know I did that today. It was a summary of XXX.com. I really enjoyed writing that. I tried to remember what XXX said in class. How to summarize a report. Write the important things. Don't use like exact sentences, exact words. I try to remember all those. So that was the best. 1.1.7.5. Genre/Lecture Notes Ali: But when we gain through school how to take notes, how to write essays. I think those are more, much more important than TWE. You know. How to write a note. How to write an essay. 1.1.7.6. Genre/Text Types Eun-Joo: And so I think every, every kind of writing I want to. I have to improve. 1.1.7.6.1 Genre/Text Eun-Joo: Ya. And we learn about argumentative. How can we Types/Argumentative write or develop the argumentative essay. 1.1.7.6.2. Genre/Text Types/Compare Gertrude: She explained. How to compare two things. And then and Contrast the possibility to go one point. The first element. That's her and that's me. And then second point. That's her and that's me. In- stead of saying that's me and that her. And then say everything about me and everything about her to compare. We made it like that point. And then her… We just had to do this one part. So it's not the whole essay. And to start and prepare this, we have to xxx everything we could think of that could be different 206 CUMMING, BUSCH & ZHOU

between xxx. 1.1.7.6.3. Genre Claudia: We just starting on, I don't know, introduction or con- /Text Types/Free Writing clusion or just free writing or narratives. 1.1.7.6.4. Genre Gertrude: And we should have written narrative, but something /Text Types/Narrative special in our life. But I didn't do that because… I… I couldn't find a topic, actually. 1.1.8. Understand the Culture Kanaporn: I need to understand the culture more. 1.2. Future Goals 1.2.1. Academic Goals 1.2.1.1. Attend University Kanaporn: I will go to an accounting program this September. 1.2.1.2. Pass Tests Claudia: Ya. I wanna. I would like to take the TOEFL after this course 1.2.2. Career Goals Yun-hui: I want to work for international organization like the UN or something like that.

2. CONDITIONS 2.1. Place Claudia: I can write on the street. I mean lying down on grasses. At home 2.2. Context 2.2.1. Tests Yun-hui: Actually, I have my TOEFL course already. Score already. And I took two times TWE. But it was awful. I mean the score, so I was disappointed. 2.2.2. Work Kanaporn: Most of them are documents for ah more xxx the trading company. And its. Most of them are nearly the same. That I can keep it the format and maybe change only some details and some dates. 2.2.3. Academic Classes Kanaporn: Yes, because like everyone, my classmates, they are all working people. They they come and the class I don't like that kind of class. It's not practical. 2.2.4. ESL Classes Gertrude: The ESL program. We do a lot of writing. 2.2.5. Past Situations Gertrude: Ah yes. I've had to do presentations at my university. 2.2.6. Present Situations Kanaporn: Sometimes she have, she told us to exchange. 2.3. Origin of Goals 2.3.1. Self Gertrude: I'd like to. First of all I'd like to improve speed. 2.3.2. Peers Gertrude: And that's where my friend came to help me. And she said that I should start with the thesis and write the first sentence, which says what I'm going to talk about. And then make the connections. And so like that it works. 2.3.3. Teachers Ali: Even the teachers have been telling us how to write the notes so you can review them after. 2.3.4. Work or Professional Activities [no examples coded]

3. ORIENTATIONS 3.1. Responsibility for Fulfilling Goals 3.1.1. Self Gertrude: I try, always try to make short sentences, try to make things simple and logic, always. 3.1.2. Peers [no examples coded] 3.1.3. Teachers Claudia: Ya, she did, but I develop my own way of doing it. So I think, I mean for me it's better. 3.1.4. Others [no examples coded] 3.2. Strategies 3.2.1. Seek Assistance from People Gertrude: Well, it depends on the people. Not everyone wants to read my outline. 3.2.1.1. Seek Assistance from Peers Claudia: Ya, usually sometimes when we write something, we just share them with our classmates. 3.2.1.2. Seek Assistance from Teach- Kanaporn: No one checked it. Only the teacher. INVESTIGATING LEARNERS’ GOALS 207 ers 3.2.1.3. Seek Assistance from a Pri- Eun-Joo: Oh, can I say I have a private lesson? A private tutor. vate Tutor 3.2.1.4. Seek Assistance from Others Ali: Not necessarily. No. Sometimes the stories I wrote I sent to my sister. She said you had a lot of verbal problems. You don't know what kind of words you should use. Maybe in what tense. Sometimes like that. But no never like in email writing. 3.2.2. Self-Regulation 3.2.2.1. Use Outline Gertrude: Yes. Ahh, OK. Ya. I make an outline, of course. And I try to keep it. Ya, that's what I'm doing. 3.2.2.2. Try not to confuse ideas Gertrude: The second thing I try is not to confuse ideas. Just prove one idea and then the next. 3.2.2.2. Planning Gertrude: Ya, well, I actually think about what I want to say. That's my, that's my. Sometimes it helps when I think deeply what I want to say. 3.2.2.3. Write Simply Gertrude: What is grammar. Of course I try to improve it. But when I'm writing, I'm trying to write simple. 3.2.2.4. Use Visual Supports Gertrude: What I like as well is to show it on the graphic. So I try to find visual methods to explain something. And that's what I like. 3.2.2.5. Editing and Revising Yun-hui: Ya, after I finish writing and maybe two times. At least two times I read it again and check some mistakes. 3.2.2.6. Synthesize Claudia: I try to synthesize… everything. I mean as much as possible. 3.2.2.7. Use Examples Kanaporn: And you think of some small example to add to this points. 3.2.2.8. Use Different Words Eun-Joo: Oh ya, right. I… I do not try to write the same words. Use same words. I try to use similar sentence. Similar words. I try to use another words. 3.2.2.9. Try to be Neutral Eun-Joo: I want to avoid criticism from… from the other people. I mean about this special the gay people issue. So I try to be neutral. 3.2.3. Stimulation Yun-hui: Nothing. [In reply to question about whether she uses any stimulating activities such as drinking coffee, playing music, etc.] 3.2.3.1. Coffee Gertrude: I go make coffee 3.2.3.2. Movement Gertrude: Walk around. 3.2.3.3. Talk Gertrude: Chat with people. 3.2.3.4. Music Gertrude: Oh, I can't listen to music when I'm writing. No.. Perhaps it would help. 3.2.4. Improving Vocabulary Eun-Joo: I mean, ah,… I want to improve my vocabulary so now I read a lot of books. 3.2.5. Use of Primary Tools 3.2.5.1. Books and Magazines Yun-hui: If I have a difficult topic to write on, maybe books are really helpful for me. 3.2.5.1. Computers Eun-Joo: Ya. With pencil and paper. And then I use computers. 3.2.5.2. Crossword Puzzles Kanaporn: Oh I have the crossword book. Many, many 3.2.5.3. Dictionaries Yun-hui: I mean. Whenever I found new vocabulary I look up dictionary. And I write down and … I try to memorize it. 3.2.5.4. Workplace Documents Claudia: From Amnesty books, Amnesty reports. That's where I get information. 3.2.5.5. Stationery Kanaporn: I don't like to write with pencil if it's for the final paper I will hand in. That's the bad habit. 3.2.5.6. Newspapers Kanaporn: To read the newspaper. Some critics corner. To more idea. 3.2.5.7. Captions on TV Eun-Joo: And a oh when I watch TV, I use the function cap- tion. 208 CUMMING, BUSCH & ZHOU

3.2.5.8. Internet Kanaporn: I used the dictionary, the Internet. Ya 3.2.6. Practice Ali: I think the best thing is write as much as you can. 3.2.7. Reading Kanaporn: I try to improve my writing by… read more books. Ya. And then trying to broaden my view because I need only prefer the fiction Books, novels now. I try to read more like science or non-fiction. And history. Many many things.

4. DILEMMAS 4.1. Dilemmas/Language 4.1.1. Grammar Claudia: So I think I don't really have problems for grammar or sentence structure. 4.1.2. Vocabulary Claudia: I have problems with vocabulary. That's my huge problem 4.2. Dilemmas/Rhetoric 4.2.1. Logic or Ideas Gertrude: This essay I have to write this week and I can't formulate my ideas very well.

4.2.2. Organization Eun-Joo: Always in my brain there are many many thoughts and ideas, but I cannot arrange everything you know. This is introduction. This is body paragraph and this is conclusion. 4.2.3. Style or Format Claudia: I don't really follow any particular format. Like I write, I don't know… It depends on the situation. It depends on me. Usu- ally I don't follow a particular order. 4.3. Dilemmas/Composing Proc- esses 4.3.1. Getting Started Yun-hui: For me, it's really difficult to start it, but once I start, I just keep writing and finish. 4.3.2. Editing and Revising Eun-Joo: I'm a little bit. I don't like reading my essay again and again because it's a very painful (laughs) to…to read my essay. 4.3.3. Writing Speed Gertrude: It takes me a long time to start. For example if I write an essay too long. Ah like two hours I'm sitting… front of my paper. 4.3.4. Time Management Gertrude: I didn't find the time and ah I didn't feel like writing. I don't have an idea. 4.4. Dilemmas/Text Types 4.4.1. Free Writing Gertrude: Because xxx don't like to tell me to do free writing. Perhaps I could do free writing, but not in this, not this system. 4.4.2. Compare and Contrast Gertrude: Well, I have… Each essay I have trouble. 4.4.3. Argumentative Gertrude: And sometimes we do write at the moment give an argumentative writing to… ah… that's what I'm stuck in now. 4.4.4. Narrative Gertrude: And we should have written narrative, but something special in our life. But I didn't do that because… I… I couldn't find a topic, actually. 4.5. Dilemmas/Can't Solve Writing Eun-Joo: Actually, I don't know how to improve my weak points. Difficulties 4.6. Dilemmas/Uncertainty Ali: I don't know… I don't even know what are my problems. 4.7. Dilemmas/Content or Topic Gertrude: That's why I don't like free writing. That's what we do in school. She said just write if you don't know what to write, write. I don't know what to write. 4.8. Dilemmas/Affective States Ali: Maybe I'm too panicked. You know, it's just hard for me to do something again and again. 4.9. Dilemmas/Handwriting Claudia: I know I have a really bad handwriting. I'm. It's really bad. Everyone complains about that. 4.10 Dilemmas/Follow Rules or Eun-Joo: Sometimes I don't like that kind of structure. To follow Formula the rules. I don't. Sometimes it bothers me. I don't care about it and just to write down something. 4.11. Dilemmas/Directness Eun-Joo: Ya, directness. So that's the way I am. I have a difficulty in writing essays.

WHEN AND WHY TALKING CAN MAKE WRITING HARDER

MARGARET FRANKEN & STEPHEN HASLETT

Massey University, New Zealand

Abstract. This chapter reports on a quasi-experimental study investigating the effects of talk on the writing of argument texts by second language students in a New Zealand high school. Results are compared on variables related to three conventional text quality measures (communicative quality, ideas and organisation, and grammatical accuracy and complexity), and also on variables specifically related to the rhetorical organisation of argument texts (frequency of claims, grounds and elaboration of grounds, as identified by Toulmin, Rieke, and Janki, 1984). The results of the study indicate that the opportunity to work with a peer before and during writing had limited and specific effects on the texts the students wrote. Working in a solitary way resulted in significantly higher mean scores for linguistic accuracy and complexity (a general text quality measure). Significant and positive effects for opportunity to talk with a peer were seen in the quantity of grounds-related propositions (an argument text measure), but only when students wrote texts that appeared to require more domain-specific knowledge to support the claims made in their argument texts. This chapter seeks to explain these effects with reference to the notion of attentional resources and aspects of noticing. A number of task and learner factors are explored to explain the way in which attentional resources are divested and the degree to which learners attend to or notice aspects of the language and content experienced in the interaction.

Keywords: second language writing, information processing models, noticing, attentional resources

1 INTRODUCTION Two significant notions from cognitive psychology are frequently drawn on to explain general aspects of second language acquisition. These are attention and noticing. Research has found that attention and noticing are central for second language acquisition development to occur (Robinson, 1995; Schmidt, 1990, 1994), and that the attentional resources of second language learners are limited (VanPatten, 1990). De Bot (1996: 550) states that ‘Attention can be viewed as a limited set of mental resources that have to be shared by various processing activities....’. The corollary of this is that second language learners need to make decisions about where resources are to be allocated and that decisions to attend to one area are usually made at the expense of attending to another.

M. Franken & S. Haslett (2002). Talking can make writing harder. In: G. Rijlaarsdam (Series ed.) & S. Ransdell & M. Barbier (Volume eds.), Studies in Writing, Volume 11: New Direc- tions for Research in L2 Writing, 209 – 229. © 2002. Kluwer Academic Publishers. Printed in the Netherlands. 210 FRANKEN & HASLETT

Much recent research on attentional resources has been focused on speaking. (See for instance de Bot’s (1996) article reviewing research on the output hypothesis.) However, a consideration of attentional resources is now also being represented in more specific skill areas such as in listening research (Brown, 1995; Goh, 2000). Goh’s study for instance identified that a large number of students in her study experienced listening comprehension problems as a result of lack of attention in the perception phase (as identified by Anderson, 1995). While early accounts of the writing process incorporated a view of attentional resources in that they claimed that less expert writers were likely to focus on form rather than meaning (Perl, 1979, on first language composing; Zamel, 1983, on second language composing), few more recent studies directly address the issue of attentional resources. This chapter, as a post hoc exploration of the findings of a quasi-experimental writing study, makes reference to the literature on attention and noticing. The study reported in the chapter initially set out to statistically evaluate whether or not opportunities to interact with a partner lead to a better text than when those opportunities are not present. The second language writers were students in a New Zealand high school. The study was prompted by studies such as that by Sweigart (1991). Swei- gart makes reference to literature explicating claims for the benefits of interaction. He concludes however, ‘While much of this work is highly suggestive of the utility of exploratory talk, no empirical studies have sought to directly measure its effects’ (1991: 470). Sweigart’s empirical study investigated three different conditions: class discussion and teacher-led discussion (the talk conditions) and listening to a lecture with no discussion. The variable of student writing ability (a three point scale administered by the class teacher) was an additional factor in the research design. The effect of the above conditions on summary writing and argument writing was tested. As this study was conducted in mainstream classroom context, Sweigart evaluated the argument texts in terms of a holistic but essentially rhetorically based rubric. In addition to the variable of text quality, he also examined the effect of talk on topic knowledge and attitudes to writing. The important findings from Sweigart for the present study are that there was a statistically positive effect for both the talk conditions on the measure of topic knowledge (measured before writing took place). The two talk conditions also resulted in better text scores (as realised rhetorically). What is also of interest is his comment about a small group of students whose protocols were analysed. He found that ‘attention to monitoring is less for students when they compose after the talk conditions compared to the lecture condition. The decrease in monitoring is paral- leled by an increase in the attention to reviewing, generating, and planning in the talk condition’ (1991: 484). The study presented in this chapter also tests the effects of interaction on the rhetorical features of argument texts as in Sweigart’s study – ‘thesis and elaboration with supporting evidence’ (1991: 492). In addition it considers the pragmatic level of the text (on a measure of communicative quality) and features of language form (grammatical accuracy and complexity). The investigation of effects of speaking tasks on writing for second language learners is complex but productive. It allows researchers to think beyond SLA re- TALKING CAN MAKE WRITING HARDER 211 search and to link findings with those in other areas such as in Sweigart’s study. It also allows researchers to explore a range of variables that add to SLA explanations of the role of interaction (see for instance Ellis, 1999a). Research investigating the distribution of attentional resources and acquisition in the SLA tradition has presented the opposition of meaning and form. The latter is interpreted as a variable in much of the research so that many research studies test whether or not certain aspects of form, most frequently aspects of syntax, have been acquired by the learners in the research studies (see for instance research related to the output hypothesis such as Pica, Holiday, Lewis, & Mogenthaler, 1989; Swain, 1995; Swain & Lapkin, 1995). With reference to more conventional areas of concern ‘whether it is a feature of pronunciation, a lexical item, a grammatical rule or some other aspect of language’ (Ellis, 1999c: 234), it is important to consider that interaction may have differential effects. As Ellis goes on to state, ‘It is not self-evident that acquisition of these different features involves the same processes. Indeed there is ample evidence that they do not’ (Ellis, 1999c: 234). Sato (1986), in making the claim that interaction may contribute little to the acquisition of morphosyntax, draws attention to the need to consider a range of variables. Speaking to writing tasks require such consideration. The complexity of the speaking to writing context lies in the fact that we are concerned with not only intake i.e. what learners notice and take into short term memory (Ellis, 1999c: 237) and acquisition but also the retrieval of acquired knowledge in the new context of writing. In addition, while uptake may well occur in the process of interaction, this may not endure. As Ellis (1994: 283) comments, ‘Evi- dence that learners improve the grammaticality of their utterances when pushed does not of course constitute evidence that acquisition takes place’. It is also the case that writing itself poses particular affective challenges for students. Attention is not just a cognitive issue but also an affective one. Elley’s research (cited in Ellis 1999b: 60) reminds us of this fact. Elley, in his vocabulary study found that children’s degree of involvement in the content of the story appeared to have a large role in determining which particular items were learned. These complexities need to be taken into account when interpreting statistical results and results related to group data. The following section outlines the way in which the effects of interaction on the written argument texts of second language high school students were evaluated.

2 METHOD 2.1 Participants The twenty students who acted as subjects for the study came from a mixed sex year 12 class at an urban New Zealand high school. As such they were an intact group. Five were girls and fifteen were boys. The majority of students were from Asia: Taiwan, Hong Kong, Malaysia and Korea. The remaining students were from India, Mexico, Macedonia and Fiji. The students had been in New Zealand for a period of two months to two years at the time the study was conducted. 212 FRANKEN & HASLETT

2.2 Procedure The teaching in the programme was carried out by the class teacher and another teacher (the first author). The class teacher was experienced and had had both TE- SOL (teaching English to speakers of other languages) and LAC (language across the curriculum) training. She had volunteered to participate in the programme and hence was assumed to be sympathetic and supportive of the programme. The teacher and the first author were randomly assigned to one of the two conditions: one when students were given opportunities to work in pairs and the other when students had little or no opportunity to work with a partner, and remained with that condition throughout the programme. The study required there to be two parallel groups, group 1 and group 2. In order to do this, two pretests were administered. The texts written in the pretest sessions were rated according to Toulmin, Rieke, and Janki’s (1984) analytical categories for argument texts (as modified by McCann, 1989) and a holistic measure of quality of argument text based on the model proposed by Knudsen (1992). The scores on both types of measures were summed, converted to z scores, and ranked. From the ranked list, pairs with closely aligned scores were formed and then randomly assigned to groups. Means were calculated for the summed scores of each group. Group 1 scores showed a mean of 35.8 whereas group 2 had a mean of 36.5. Language proficiency data that was both relatively recent and gathered at the same time for all students was used to confirm the essential parity of the two groups. The present study fitted into a wider academic skills programme which aimed to prepare students for both the demands of the curriculum and the IELTS test. The study lasted six weeks, covering one topic per week. The instructional sequence, or sequence of tasks used to complete one topic, involved a total of three periods per week amounting to two and a half hours. The first session was a double period during which students generally completed all prewriting tasks. The third period was used for students to write their texts. Prewriting tasks included the critical study of a sample argument text (see Franken, 1997a, for an example of the text study procedure used), brainstorming and mapping. During writing, students had access to a cue card that provided writing prompts (see Bryson & Scardamalia, 1991, for an example of this task type). Instructions for the completion of each task in the two conditions differed only in that the students on alternate occasions were to carry out the tasks with a peer (see Appendix 1 for a sample list of instructions). The sequence of tasks, in booklet form, contained clear instructions. Neverthe- less the procedures for completing the task sequence were modelled before the programme began. Teacher involvement during the sessions was minimal. One teacher introduced each new topic to the whole class before each session and made brief and general comments on previous work before students either moved off to work independently or with a partner. Students chose a partner from within their groups, as it was believed that the students would feel more motivated and more inclined to talk with self-selected peers. The only restriction was that on occasions after the first, when students worked col- laboratively, they were asked to choose a peer who differed from the one chosen on the first occasion. TALKING CAN MAKE WRITING HARDER 213

2.3 Research design In an experimental study, the variable of interaction is problematic. In a naturalistic setting with a class of students there exists great variability in the amount and quality of interaction each student engages in when instructed to work with a partner. Rather than attempt to control such a variable, it was felt that the variable needed to be accurately named. It is in effect an opportunity variable. Interaction between each of the student pairs on each occasion was taped as was any interaction that occurred between the teacher and students when they were expected to work independently. Examination of this data at the end of each week suggested to the teacher and the researchers that essential differences between the interaction and no interaction conditions (as defined above) were being maintained. Figure 1 below summarises the programme and the research design. As can be seen in Figure 1, the two groups experienced both the interaction and no interaction conditions but at different times.

Week

1 2 3 4 5 6

Group 1 I NI I NI I NI Group 2 NI I NI I NI I Level of Resourcing Non-resourced Semi-resourced Resourced Topic General Semi-domain-specific Domain-specific

I = the interaction condition; NI = the no interaction condition.

Figure 1. Programme plan.

A second set of independent variables: non-resourced writing, semi-resourced writing, and resourced writing, arose from the way in which the programme was designed. As can be seen from Figure 1, the first two weeks of the programme involved working on a general writing task without input or textual support (non- resourced writing). In the second two weeks, students were given a fact sheet, list of facts relevant to the topic, as support and were required to write on more specific topics related to General Science (semi-resourced writing). The last two weeks’ writing involved specific Economics/History type writing tasks which were accompanied by written texts (resourced writing). The variable referred to as resourcing represented not only the degree of textual support, but also domain specificity and experience with the argument genre as it occurred last in the programme. It was therefore a combination of factors including occasion, resourcing, and topic. The researchers could not avoid the confounding of these three variables, while at the same time addressing the ethical demands of providing a pedagogically logical and valid progression of work. To assess whether better texts were produced, two sets of dependent measures were designed. The first set comprised three general analytical scores: communicative quality, ideas and organisation, and grammatical accuracy and complexity. Each of these three measures was based on a scale of fifteen (as developed by Hartfiel, 214 FRANKEN & HASLETT

Hughey, Wormuth, & Jacobs, 1985). The second set of measures relating specifically to argument texts were frequency measures of the constituents: claims, grounds and elaboration of grounds (as identified by Toulmin, Rieke, & Janki, 1984). While the latter is not strictly a qualitative measure of the effectiveness of argument, as in the case of the holistic rubrics used by researchers such as Knudsen (1992), McCann (1989) and Sweigart (1991), it can be argued that quantity of constituents and their relationship is critical for a good argument (see for instance an explanation of Chris- tensen’s view (cited in Coe, 1988). ‘Christensen’s basic point is that expert writers generate more information than novice writers do and they add this extra information to their basic statements as modifiers (typically, he argues, creating cumulative sentences). This extra information, he asserts is most commonly extra detail, on a lower level of generality than the basic statements being modified, and its inclusion creates what he calls texture, what makes us describe a piece of writing as being ‘well developed’ or as having ‘depth’.’ (Coe, 1988: 68) The reliability of the ratings given for text quality measures was estimated by calculating the percent of agreements achieved by two raters (one of which was the writer), for a sample of fifteen percent of the texts (texts produced on occasions one, three and five for a random sample of nine students). The level of agreement between the two raters ranged between 78% (for frequency of grounds) and 94% (for communicative quality, and grammatical accuracy and complexity). The results gained from the ratings were analysed as a repeated measures design. Group was the between subjects main effect, and level of resourcing, and level of interaction were the repeated measures. The design then was a doubly repeated measures design. The repeated measure effects level of resourcing and level of interaction were tested in the multivariate repeated measures mode (see for example, Littell, Freund, & Spector, 1991), rather than as a split plot design i.e. the design used for each independent variable remains a univariate ANOVA. There is another possible interpretation of the design: group one always experienced interaction before non-interaction; group two always experienced non- interaction before interaction, so that the groups differed in the sequence of these two (I-NI or NI-I), and level of resourcing is the only ‘within’ factor in the design. These two interpretations are possible because there are only two groups. This issue is discussed further in the results section to follow.

3 RESULTS The mean scores on all three of the general quality measures: grammatical accuracy and complexity (AC), ideas and organisation (IO), and communicative quality (CQ) are given in Table 1 below. Note that the scores obtained for each of the interaction conditions within each level of resourcing were obtained on different occasions. So that, for instance, interaction was experienced on week one by group one and on week two by group two (see Figure 1). Table 1 presents summed scores across occasions. TALKING CAN MAKE WRITING HARDER 215

Table 1. Scores for general measures of text quality by level of resourcing and interaction (Means, standard deviation and standard error).

Non-resourced Semi-resourced Resourced

AC IO CQ AC IO CQ AC IO CQ

With interaction M 6.30 7.15 7.60 6.35 7.00 7.15 7.45 8.50 8.55 SD 2.05 2.06 2.50 2.43 2.94 2.32 2.24 2.44 2.56 SE 0.46 0.46 0.56 0.54 0.66 0.52 0.50 0.55 0.57 Without interaction M 6.65 7.35 7.75 6.85 7.65 7.80 8.35 8.35 9.20 SD 2.32 2.30 2.17 2.58 2.70 2.59 2.94 2.72 2.76 SE 0.52 0.51 0.49 0.58 0.60 0.58 0.66 0.61 0.62

AC = accuracy and complexity; IO = ideas and organisation; CQ = communicative quality

The fact that there were only two groups, one for each sequence of resourcing, leads to some confounding. This decision was a logistical one. However, groups did have very similar scores at the beginning and at each change in level of resourcing, so that they seemed to be very similar. Differences found have consequently been interpreted in terms of level of interaction and level of resourcing rather than simply as differences between groups. Note too that the formal testing of differences between groups (adjusted for level of resourcing and level of interaction) is not statistically significant. An examination of the mean scores in Table 1 indicates, not surprisingly, that the level of resourcing for the writing task appeared to be an important influence. The students appeared to be writing better texts by the time they were working on the fifth and sixth writing occasions as represented by the resourced condition. This observation is borne out by the results of the ANOVA, indicating significantly higher scores for all three areas. This indicates a positive effect from textual resources. It should also be remembered that the resourced writing condition represented not only level of textual support for writing but also increased experience with the genre and the sequence of tasks. With respect to interaction, a visual inspection of the descriptive statistics, presented in Table 1, shows that in all three conditions, the non-resourced, the semi- resourced and the resourced, the mean scores that resulted when students worked alone were, with one exception, higher than when they had the opportunity to work with a peer. The results from the ANOVA for the grammatical accuracy and complexity scores are given below in Table 2. As mentioned above, high levels of significance for resourcing can also be seen in addition to those for interaction. The results of the ANOVA for ideas and organisation, and communicative quality are not tabulated however, since these indicated that while these two effects were significant for 216 FRANKEN & HASLETT grammatical accuracy and complexity, they were not significant for either ideas and organisation or communicative quality. Note that the ANOVAs for grammatical accuracy and complexity, ideas and organisation, and communicative quality are not independent because these three variables are correlated. The maximum correlations (i.e. unadjusted for level of resourcing or interaction) are given in Appendix 2.

Table 2. Effects on grammatical accuracy and complexity scores.

Effect Num df Den df Type II SS MS F P

Group 1 16 3.42 3.42. 0.50 0.49 Subject (group) 16 - 110.29 6.90 Interaction 1 16 9.32 0.01 Resourcing 2 15 22.12 0.00 Interaction*Group 1 16 1.08 0.31 Interaction*Resourcing 2 15 0.34 0.72 Resourcing*Group 2 15 1.66 0.22 Resourcing*Interaction*Group 2 15 1.64 0.23

Underlined F-scores are statistically significant p < .01

Even given the high correlation between any pairing of grammatical accuracy and complexity, ideas and organisation, and communicative quality, grammatical accuracy and complexity produces significant effects while ideas and organisation, and communicative quality do not. The significance of interaction and level of resourcing for grammatical accuracy and complexity are both so strong that the significance remains even after adjusting for the effect of this correlation with ideas and organisation, and communicative quality. Table 3 reports the means, standard deviations, and standard errors of measurement. They statistically support the visual analysis that it was the no interaction condition that resulted in higher mean scores for grammatical accuracy and complexity. Counter to expectation, the opportunity to interact with a peer before and during writing did not appear to contribute to the linguistic resources accessible to the students as students’ texts were more accurate and more complex when they worked alone. This effect remains under either of the two possible interpretations of the design discussed above. Note that interaction is not a significant effect statistically if resourcing is ignored (as in the standard errors in Table 3), so that the effect of interaction may be masked if resourcing is not also considered. The second set of measures (relating specifically to argument text structure) were three frequency measures including claims, grounds, and elaboration of grounds. The mean scores attained for each of these measures are presented in Table 4 below. Visual inspection of Table 4 likewise indicates greater scores on each of the three variables for at least the non-resourced and semi-resourced writing conditions. However a different pattern emerges for the resourced writing that occurred at the TALKING CAN MAKE WRITING HARDER 217 end of the programme, indicating higher scores for interaction on the grounds related measures.

Table 3. Scores for grammatical accuracy and complexity by interaction.

Interaction Yes No

M 6.70 7.28 SD 2.27 2.69 SE 0.29 0.35

Table 4. Scores for argument measures of text quality by resourcing and interaction.

Non-resourced Semi-resourced Resourced

Variable CL GR EGR CL GR EGR CL GR EGR

With interaction M 5.70 5.35 7.95 10.60 5.55 3.75 8.35 5.30 5.90 SD 2.30 2.80 6.76 3.00 3.24 3.96 3.70 2.36 4.44 SE 0.51 0.63 1.51 0.66 0.72 0.89 0.83 0.53 0.99 Without interaction M 7.40 6.40 7.60 11.55 6.00 3.10 10.35 3.90 3.70 SD 2.72 3.87 5.38 4.11 3.42 2.31 2.76 2.40 3.92 SE 0.61 0.87 1.20 0.92 0.76 0.52 0.62 0.54 0.88

CL= claims; GR = grounds; EGR = elaboration of grounds

The following discussion proceeds by explaining the findings associated with each of the three measures: claims, grounds and elaboration of grounds. Note that from Appendix 2, these three measures have low correlation with each other and with grammatical accuracy and complexity, ideas and organisation and communicative quality. The ANOVAs below are essentially independent of each other and of those presented previously. The ANOVA performed on the frequency data relating to number of claims indicated significance for interaction and for resourcing. Results of the ANOVA procedures including resourcing and interaction are described in Table 5. It would seem that the frequencies are sufficiently high that ANOVA and generalised linear models would have yielded similar results. These results present a similar pattern to those for grammatical accuracy and complexity. As the level of interaction is of most interest, the means, standard deviations and standard errors associated with the level of interaction are presented below in Table 218 FRANKEN & HASLETT

6. These indicate that again, like the results for the measures of grammatical accuracy and complexity, the without interaction condition resulted in a greater number of claims being realised in students’ texts irrespective of level of resourcing (see Table 6). Again as for grammatical accuracy and complexity, these conclusions hold regardless of the interpretation of the design, and the effect of interaction may be masked if resourcing is not also considered.

Table 5. Effects on claims scores.

Effect Num df Den df Type II SS MS F p

Group 1 16 22.56 22.56 4.00 0.06 Subject (group) 16 - 90.32 5.65 Interaction 1 16 7.22 0.02 Resourcing 2 15 33.33 0.00 Interaction*Group 1 16 1.02 0.33 Interaction*Resourcing 2 15 0.27 0.77 Resourcing*Group 2 15 0.16 0.85 Resourcing*Interaction *Group 2 15 2.46 0.12

Underlined F-scores are statistically significant p < .05

Table 6. Scores for frequency of claims by interaction.

Interaction Yes No

M 8.22 9.77 SD 3.61 3.65 SE 0.47 0.47

As suggested above in Table 6, the area of grounds and elaboration of grounds reflected an interesting and distinct pattern when ANOVA was carried out since statistical interactions were found. For grounds, a level of interaction by group effect sig- nified that the level of interaction was producing a significantly different effect on the production of grounds for one of the groups. A three way statistical interaction was also observed between resourcing, interaction and group. This indicates that in addition to a specific effect of interaction on one of the groups, a specific effect could also be located for one of the groups with one of the levels of resourcing. The test for significance showed p = 0.0357 and p = 0.0005 respectively as is seen in Table 7. TALKING CAN MAKE WRITING HARDER 219

Table 7. Effects on grounds scores.

Effect Num df Den df Type II SS MS F p

Group 1 16 1.54 1.54 0.22 0.64 Subject (group) 16 - 111.61 6.98 Interaction 1 16 0.01 0.94 Resourcing 2 15 3.14 0.07 Interaction*Group 1 16 5.26 0.04 Interaction*Resourcing 2 15 3.04 0.08 Resourcing*Group 2 15 1.89 0.18 Resourcing*Interaction*Group 2 15 13.13 0.00

Underlined F-scores are statistically significant p < .05

The means, standard deviations and standard errors associated with the levels of resourcing, the levels of interaction, and the two groups are presented for comparison in Table 8 below. Table 8 indicates that in the non-resourced condition, group two had a higher mean score for frequency of grounds in the interaction condition, while for group one, it was the without interaction condition that resulted in higher mean scores. A similar pattern is seen in the semi-resourced condition. However, for the resourced condition for group one, the interaction condition resulted in significantly higher scores than the without interaction condition. Group two on the other hand had higher mean scores for the without interaction condition.

Table 8. Scores for frequency of grounds by resourcing, interaction and group.

Non-resourced Semi-resourced Resourced

Interaction

Group Yes No Yes No Yes No

1 M 4.20 8.10 4.10 7.00 5.50 1.90 SD 1.55 4.41 1.97 3.43 2.68 1.37 SE 0.49 1.39 0.62 1.09 0.85 0.43 2 M 6.50 4.70 7.00 5.00 5.10 5.90 SD 3.34 2.41 3.68 3.27 2.13 1.20 SE 1.06 0.76 1.16 1.03 0.67 0.38 Total M 5.35 6.40 5.55 6.00 5.30 3.90

Group 1 had sequence I-NI, group 2NI-I.

220 FRANKEN & HASLETT

In the study, sequence (I-NI for group one; or NI-I for group two) was confounded with a topics factor as the groups worked on different topics on each occasion. There were six occasions in all, each with a different topic. As explained above (see Figure 1), students worked alternatively with interaction and without interaction. The first topic for instance, was carried out by group one with interaction, while group two worked without interaction. On the second occasion, the second topic was again done under two conditions, but group one experienced the without interaction condition and group two the interaction condition. Hence the effects of the group factor may be accounted for in terms of the topic differences. The following graph, Figure 2, shows the frequency of grounds for each of the two groups in relation, not to the treatment or to level of resourcing, but in terms of the topic. Figure 2 shows that group one and group two perform similarly on most of the topics. However, there is a most noticeable difference at topic six. The graph shows group two scoring much higher on topic six than group one. For group two, this co- incided with the interaction condition.

9 8 7 6 5

Score 4 3 2 1 0 123456 Topic

Group 1 Group 2

Figure 2. Mean frequency of grounds for topics across levels of resourcing.

A possible interpretation of this is that group two benefited from interaction when required to write on a topic on which few of the students would have had much content and domain-specific knowledge. In this context, grounds may have been easier to derive from a partner and from the textual input by way of a partner. Another way of expressing this pattern is to see that group one were disadvantaged without the help of peers to work with. Working in a solitary way may have meant less opportunity to generate ideas and little access to the ideas in the source reading. What is interesting however is that group two were not similarly disadvantaged for topic 5 which was at a similar level of complexity to topic 6. The difference then is either simply a difference between groups, or can be explained by the benefit of the sequence NI-I over I-NI, when faced with two domain specific topics. The confound- TALKING CAN MAKE WRITING HARDER 221 ing in the design does not allow a distinction to be made between these two possibilities. There is some further discussion of topic six below. For the frequency of elaboration of grounds, a similar pattern of results of the ANOVA procedures are reported below in Table 9. The level of resourcing factor again proved to be significant, but most importantly similar statistical interactions were observed (level of resourcing, level of interaction, and group).

Table 9. Effects on elaboration of grounds scores.

Effect Num df Den df Type II SS MS F p

Group 1 16 150.38 150.38 4.49 0.05 Subject (group) 16 - 535.40 33.46 Interaction 1 16 2.55 0.13 Resourcing 2 15 6.29 0.01 Interaction*Group 1 16 2.46 0.14 Interaction*Resourcing 2 15 1.55 0.24 Resourcing*Group 2 15 0.48 0.63 Resourcing*Interaction*Group 215 3.800.046

Underlined F-scores are statistically significant p < .05

The descriptive statistics given in Table 10 show a similar pattern to the scores for grounds for group one. In the two contexts (non-resourced writing and semi- resourced writing), their scores for the without interaction condition exceeded those for the interaction condition. However for group two there appears to be a more pervasive influence of interaction. Group two has consistently higher scores on frequency of elaboration of grounds in the interaction condition for all three levels of resourcing. A topic factor alone could explain the results obtained for the first two levels of resourcing which cover topics one to four as the two groups irrespective of whether they are experiencing the interaction or without interaction condition have a similar pattern of achievement as can be seen in Figure 3 below. On topic five however, the difference in scores for frequency of elaboration of grounds becomes less. At topic six, which was also carried out with the support of the greatest level of resourcing, a different pattern emerges. Students working together with this level of resourcing and on this topic produced higher scores for elaboration of grounds. Topic six, rather more than even topic five, was one in which few students would have had content and domain-specific knowledge. Group two working on this topic with interaction enjoyed a significant advantage over group one who worked in a solitary way. The effect parallels that for frequency of grounds. 222 FRANKEN & HASLETT

Table 10. Scores for frequency of elaboration of grounds by resourcing, interaction and group.

Non-resourced Semi-resourced Resourced

Interaction

Group Yes No Yes No Yes No

1 M 6.10 7.50 1.80 3.40 4.70 2.20 SD 5.07 6.74 1.69 2.76 2.91 4.13 SE 1.60 2.13 0.53 0.87 0.92 1.31 2 M 9.80 7.70 5.70 2.80 7.10 5.20 SD 7.96 3.97 4.67 1.87 5.47 3.22 SE 2.52 1.26 1.48 0.59 1.73 1.02 Total M 7.95 7.60 3.75 3.10 5.90 3.70

In summary, for linguistic accuracy and complexity, and frequency of claims, the opportunity to work in a solitary way appeared to result in gains that were greater than those when working with a partner. However, it also appears that when faced with a topic on which students have little content and domain-specific knowledge, interaction is significantly helpful for realising in text, grounds-related material, arguably the most important aspect of a good argument text. In other words the effect of interaction seems to be mediated by the type of topic students are writing on. In this study those effects were observed by argument text constituent analysis – a more direct observation than Sweigart’s (1991) independent measurement of topic knowledge.

Frequency 4

0 123456 Topic

Group 1 Group 2

Figure 3. Mean frequency of elaboration of grounds for topics across levels of resourcing. TALKING CAN MAKE WRITING HARDER 223

4 DISCUSSION The positive effects on grammatical accuracy and complexity found for solitary activity can also be approached from a point of view that seeks to explain why no significant effects were found for interaction. This finding runs counter to claims related to the output hypothesis: one, that interaction can lead to improved grammatical performance (Pica, 1988; Pica et al., 1989), and two, that interaction enhances fluency (Swain & Lapkin, 1995). Also important to address is the fact that there were no effects from interaction on communicative quality. One may also have expected better performance on communicative quality on the basis of the Mangels- dorf’s (1989) claim that speaking activities for writing make writers more aware of the need to communicate effectively to a reader. The positive effects on grounds related material brought about by interaction is perhaps not surprising given Sweigart’s (1991) finding that topic knowledge and the effectiveness of the argument improved. Discussion of the findings in particular for grammatical accuracy and complexity draws on the notion of attentional resources. Noticing on the other hand may account for content related benefits. The following section explores what may be the factors affecting the way that attentional resources are divested and the degree to which learners attend to or notice aspects of the language and content experienced in interaction. The factors, reminiscent of Mosenthal's (1983) model, include the task, the material and the writer. For the purposes of this discussion, task and material factors are combined.

4.1 Task and material factors Task factors refer not only to the manner in which information is presented (Sweller, 1994: 304), and the way in which tasks are sequenced.

4.1.1 The task sequence While the task sequence in this programme contained tasks that built on each other in terms of increasing cognitive complexity, content continuity, and genre continuity, the mere fact of switching from speaking to writing may have resulted in constraints on performance. ‘Task shifting’, the moving from one type of task to another results in competition for attentional resources (Ellis, 1994: 390). In this study students were required to operate in two modes, one with speaking and one without. The speaking they were required to do was highly structured, and in terms of Doyle’s (1983) analysis of tasks, students knew how and why they were to perform the speaking. Nonetheless the tasks in the task sequence would not have been frequently encountered by students given the findings of surveys into school-based writing (Applebee, 1984; Philips, 1985). The fact that the interaction experienced in the task sequence was novel meant that it required a ‘task shift’ and that it had an associated ‘cost’ (Meiran, 1996). The cost was a decrement in performance. Specifi- cally, the decrement appeared to be manifest in the grammatical accuracy and complexity of the texts the students wrote. 224 FRANKEN & HASLETT

4.1.2 The nature of the genre and content While there was genre continuity in the task sequence in that all the prewriting tasks also focused on aspects of argument structure, argument poses particular problems for writers. If a student is familiar with the genre one may predict that they will be able to automatise certain aspects of the writing task. However an unfamiliar genre and one which is difficult such as argument, may impose additional cognitive load (Sweller, 1994; Foster & Skehan, 1996), and as Sweigart found in his (1991) study, caused an attentional shift to focus on generating and planning aspects of composing behaviour. Many studies investigating students' ability to write argument texts, in particular, have illustrated difficulties across a range of age levels (McCann, 1989; Knudsen, 1989; Coirier, 1996). McCann (1989) reports NAEP (National Assessment of Educational Progress) results from the USA in the decade 1970-1980, which reveal that ‘students have much more difficulty with persuasive writing tasks, which involve argument, than with narrative, descriptive, or expository tasks’ (McCann 1989: 62). Another aspect of genres, particularly important in speaking to writing tasks, is the role of audience. Rumelhart’s (1980) distinction between closed and open genres is referred to here. If the genre is closed (writer determined, as in narrative) one may expect there to be positive transfer from speaking to writing (Rumelhart, 1980). Conversely benefits when working with a genre open to critical response from others (audience determined, as in argument) is less likely to transfer effectively (Bere- iter & Scardamalia, 1982; Knudsen, 1989). Closely related to the nature of the genre are the complexity of the content (Can- dlin, 1987) and the interactivity of the content (Sweller, 1994). Complex content or information in a task is likely to make that task more difficult. Students may experience additional cognitive load. Candlin (1987: 21) describes a particular aspect of content complexity in the following explanation related to a receptive task, ‘Tasks which require learners to follow a clear chronological sequence referring to individual actions of individual characters, will clearly be cognitively less demanding than a task in which there is no such clear development and where the picture is complicated by multiple actions and multiple actors’. Added to the notion of content complexity is the parameter of content interactivity identified and discussed by Sweller (1994). Interactivity refers to the degree to which information is interrelated or discrete. Material that is discrete and has low interactivity has little impact on processing capacity (Sweller 1994: 304, 305), while content needed for argument has a high degree of interactivity particularly when generated not only from textual resources, but also with a partner. Another aspect that could be considered to be part of task and material factors is the nature of the interaction that students are expected to engage in. If interaction that students engage in during interaction adds little to the task at hand, its redundant nature may bring about negative effects, or ‘extraneous cognitive load’ (Sweller, 1994: 303). If students engage in interaction that is perceived to be necessary as opposed to redundant, it may be more likely that there will be positive effects. Clearly the perception of content as being important to complete the task will affect notic- TALKING CAN MAKE WRITING HARDER 225 ing, and may have played an important role in the way in which students interacted on topic six.

4.2 Writer factors The following discussion focuses on factors associated with the writers themselves. While the present study did not empirically account for individual writer factors in that it dealt with group data, it is nonetheless important to look at individual factors both for their potential to further understand the complexity of the issue, and to sig- nal future avenues for research.

4.2.1 Language proficiency Possibly the most important and often explored individual variable is that of language proficiency. De Bot states, ‘In language production by NSs (native speakers), most attention goes to higher processes, such as the co-ordination of intentions; lower, automatic processes on the morphosyntactic and phonological/articulatory level receive hardly any attention’ (1996: 550). Some writers maintain that for second language learners with low proficiency the inverse occurs, that is, in contexts where demands are made on attentional resources, it is not the lower level or surface considerations of form that are negatively affected, but rather the higher level or meaning-related aspects of language. In writing research, for instance there is evidence to show that inexpert writers divest attentional resources in this way (see for instance Scardamalia, Bereiter, & Goelman, 1982). Learners with greater proficiency would presumably be closer to the case of native speakers whom de Bot (1996: 550) says experience the following: ‘Most attention goes to higher processes such as the co-ordination of intentions; lower automatic processes on the morphosyntactic and phonological/articulatory level receive hardly any attention’. In speaking to writing tasks, the same effects can be predicted from the present study. For the relatively proficient second language students in this study, the demand placed on them by communicating with a partner appeared to have resulted in ss, rather than more attention being paid to the surface features of their texts. Thus, the level of proficiency of the students in this study may have been a factor in determining how attentional resources were distributed. The level of proficiency of students participating in classroom interaction will affect the nature of the benefits that interaction will bring to the texts subsequently written by those students.

4.2.2 Prior content knowledge Prior content knowledge is most often seen as a facilitative factor. The content knowledge a writer brings to the writing task is clearly a factor in the successful production of text (see for instance Eigler et al, 1990; Sweigart, 1991). There is evidence that domain-specific topics vary considerably in terms of their requirement for knowledge, input, and information processing, prior to and during writing (see for instance the volume of articles edited by Carretero & Voss, 1994). De Bot (1996) 226 FRANKEN & HASLETT makes the claim that interaction makes learners notice gaps in their linguistic knowledge and that noticing gaps may trigger cognitive processes which generate linguistic knowledge. It would appear that under certain circumstances, talk can trigger memory searches of prior knowledge, it can facilitate the reorganisation of prior knowledge, and it can raise consciousness of knowledge gaps. However it would appear from the present study that a necessary prerequisite for this to occur is that gaps in prior knowledge exist. When students have much prior knowledge of the topic, this may affect their evaluation of the usefulness of interactive task as discussed above. This is a possible explanation for the fact that in the present study, it was only when students worked with difficult topics that positive effects from working with a partner were found in the content of their texts. Student engagement has been found to correlate with the learning of particular aspects of language such as vocabulary (Elley cited in Ellis, 1999b: 60). Bruffee (1984) proposes an interesting factor that may affect student engagement and hence noticing, this being the nature of the underlying beliefs that students bring to a speaking task. This affects that nature of the interaction or discourse (Bruffee, 1984). If students engage in ‘abnormal discourse’ as opposed to ‘normal’ discourse the level of engagement may increase. Abnormal discourse is defined by Bruffee as occurring when beliefs are different. Franken (1997b: 208) observes that dissonance may not only apply to beliefs but also when content or prior knowledge is different.

4.2.3 Interactional expertise In the same way that there is variability in language proficiency, so too do students vary in terms of their ability and willingness to be productive in pair or group interaction. In other words, students have varying levels of ‘interactional expertise’. The notion of interactional expertise is related to the concept of scaffolding. The term, scaffolding, describes the role of an expert in assisting a learner to solve problems in the zone of proximal development as described by Vygotsky (Gredler, 1997). ‘Scaf- folding is the process of controlling the task elements that are initially beyond the learner’s capacity’ (Gredler 1997: 365). In relation to pair or group interaction, it is the expert’s talk that provides the means by which task elements can be brought within other learners’ capacity. When less expert students are paired or grouped with an expert interactant, one may expect that interactant to bring benefits that may affect the quality of subsequent writing. While this factor was not explored in the present study, independent assessment of students’ interactional expertise would help understand both the limitations and potential of interaction.

5 CONCLUSION The chapter has presented and attempted to interpret the findings of a study suggesting that interaction has limited and specific effects on subsequent written texts. It proposes a number of constraints on language performance when second language students are asked to move from speaking to writing, and has indicated where constraints on language performance are likely to lie and why they may occur. A num- TALKING CAN MAKE WRITING HARDER 227 ber of factors likely to have affected noticing particularly of aspects of content have also been considered. In interpreting the results from the present study, a number of issues come to light. One of these relates to balancing the demands of experimental research design with those of real classrooms, real teachers and real students. Group data is persuasive in validating results. However care should be taken in the way in which students are divided into groups and the way in which groups are formed. Where logis- tically possible, designs that avoid potential confounding of group variables with other situational variables should be strived for. While group data should continue to be valid data, there is the need to move beyond group data and examine in more detail the variety of individual learner factors that may affect the way learners respond to opportunities to engage in interaction. Another issue is the importance of considering a range of dependent variables that may be sensitive to opportunities to interact with peers. It may be that in studies such as these, holistic rubrics or rating scales of variables such as grammatical accuracy and complexity – a response to Sato’s (1986) claim that interaction may contribute little to the acquisition of morphosyntax – are still not tuned to the appropriate level of language to indicate effects. While both the rhetorically based rubric used by Sweigart and the frequency analysis of argument text constituents used in the present study appeared sensitive to effects, other variables related to argument text features need to be explored. One last consideration and caution for future empirical research is that effects of interaction may take time to take hold. 228 FRANKEN & HASLETT

APPENDIX 1: SAMPLE INSTRUCTIONAL MATERIALS AND RESOURCE MATERIALS Note: Brackets indicate additional instructions for talk condition.

This is a set of tasks to help you write an argument text. The topic is:

FOR OVERSEAS STUDENTS STUDYING IN NEW ZEALAND THERE ARE BENEFITS AND DISADVANTAGES. THE BENEFITS OUTWEIGH THE DISADVANTAGES.

Task one What do you know/think about the topic? Write your ideas in the table (as you tell your partner). Add new ideas and change your old ones (as you hear your partner talking). Put these in the bottom of the table.

Task two Look at the argument text below. This is a good example of an argument text. What makes the text a good text? Circle the things that are good. Say why they are good. Write your reasons on the side of the page (as you tell your partner). Add new ideas and change your old ones (as you hear your partner talking).

Task three Map important information that will be useful for writing your text. (When you have finished, explain your map to your partner). Add new ideas and change old ones (as you hear your partner talking).

Task four Begin writing your text. When you have written three sentences, read them carefully. Use the cue card to help you improve your text. Make changes to your text if you need to. Tick any cues you have used from the cue card. (Work with your partner as you are writing your text. After you have written three sentences, read them to your partner. Your partner will cue you to improve your text. Make changes to your text with the help of your partner.)

Task five Use the criteria you have listed on task sheet two. Prove (to your partner) that you have met the criteria. Put a tick beside the criteria on task sheet two each time your writing shows that good thing. See how many ticks you can get.

Task six Spend some time now making any changes to your text that you feel are necessary.

Task seven Without using your text, explain your argument to the teacher (or to your partner). TALKING CAN MAKE WRITING HARDER 229

APPENDIX 2: CORRELATIONS BETWEEN VARIABLES

1. 2. 3. 4. 5.

1. Accuracy and complexity 2. Ideas and organisation .76 3. Communicative quality .87 .78 4. Claims .12 .08 .07 5. Grounds .18 .24 .21 -.02 6. Elaboration of grounds .07 .23 .13 -.18 .38

A PROBLEM-POSING APPROACH TO USING NATIVE LANGUAGE WRITING IN ENGLISH LITERACY INSTRUCTION

ELIZABETH QUINTERO

New York University, USA

Abstract. The intent of this research is to document ways in which students learning English as a new language can become more effective writers, readers, and participants in English literacy by using their native languages and literacies. This chapter reports on an on-going qualitative study of contexts that use problem-posing and critical literacy for English Language Learners. This problem-posing, often described as critical pedagogy, combines reflective thinking, information gathering, collaborative decision making, and personal learning choices. The format allows for use of students’ native languages and literacies while studying English literacy. Furthermore, the problem-posing approach allows for multiple forms of authentic evaluation and assessment. This approach is not a prescriptive lesson planning format, but is a way to facilitate student choice and generative work which is related to students’ lives whatever the age or context of the students. Three case study examples are discussed and a preliminary analysis is made using the framework of critical theory. Critical literacy, as defined by the author, is a process of constructing meaning and critically using language (oral and written) as a means of expression, interpretation and/or transformation of our lives and the lives of those around us. This framework led to the categoriza- tion of findings into three categories of: expression, interpretation, and transformation. These three categories can be guides for discussion both learning in a general, deep content sense and in terms of academic literacy skills.

Keywords: critical literacy, English literacy, native language literacy, oral and written language, problem-posing, and qualitative study.

1 INTRODUCTION

The research evidence is very clear that first-language development provides crucial support for second-language development. The more that students are given positive opportunities for L2 development, the better they will succeed academically in L2. (Ovando & Collier, 1998:115)

E. Quintero (2002). A problem-posing approach to using native language writing in English literacy instruction. In: G. Rijlaarsdam (Series ed.) & S. Ransdell & M. Barbier (Volume eds.), Studies in Writing, Volume 11: New Directions for Research in L2 Writing, 231 – 244. © 2002. Kluwer Academic Publishers. Printed in the Netherlands. 232 QUINTERO

The intent of this research is to document ways in which students learning English as a new language can become more effective writers, readers, and participants in English literacy as well as in their native language literacies. Issues of language and critical literacy in the context of home culture and culture of English Language Learners' new learning environments in both schools and communities are considered. This chapter reports on an on-going qualitative study of contexts that use problem-posing and critical literacy for English Language Learners. This problem- posing, often described as critical pedagogy, combines reflective thinking, information gathering, collaborative decision making, and personal learning choices. The method helps students express their voice in oral and written ways (Torres-Guzmán, 1993). Kuiken and Vedder (this volume) show that when a writing task involves collaboration there are a number of pedagogical advantages. This collaboration helps to assure the language skills are integrated rather than practiced in isolation (Kuiken & Vedder, this volume). Furthermore, the problem-posing approach allows for multiple forms of authentic evaluation and assessment. This listening approach is not a prescriptive lesson-planning format, but is a way to facilitate student choice and generative work that is related to students’ lives whatever the age or context of the students. The theoretical construct used by Cummins (1999) to describe instruction for language learning and academic achievement has three components. The three are 1) Focus on Message, 2) Focus on Language, and 3) Focus on Use. Cummins (1999) maintains that The Focus on Language component should put controversial issues such as the appropriate ways and time to teach L2 grammar in a context of Lan- guage Awareness. He advocates that under this category of Language Awareness, students must develop critical language awareness that involves exploring the relationships between language and power. He says the following. In short, a focus on formal features of the target language should be integrated with critical inquiry into issues of language and power. Also to be effective, a focus on language must be linked to extensive input in the target language (e.g. through reading) and extensive opportunities for written and oral use of the language. (p. 32)

All over the world critical literacy is being studied in terms of theory, practice, pol- icy and research (Comber & Kamler 1997; Muspratt, Luke, & Freebody, 1997; Freire & Macedo, 1987). The widespread attention does not mean scholars agree upon the definition or the extent to which the theory and the practice can or should be separated. Comber and Kamler (1997) state, If critical literacy is to mean anything to educators in the nineties, a greater self consciousness about language will need to be developed; a metalanguage for talking about how language constructs both a representation of experience and a positioning of readers and writers in relations of power. (p.1) The issues of power, what students learn to read and write, and their access to trans- formative use of their literacy skills are exaggerated in the classroom contexts of multilingual learners. The researcher defines critical literacy as a process of constructing meaning and critically using language (oral and written) as a means of expression, interpretation and/or transformation of our lives and the lives of those TALKING CAN MAKE WRITING HARDER 233 around us. This chapter is a description of design, implementation, and evaluation of an instructional method used with native and non-native language English composition and literacy classes. The contextual reality of many English Literacy classrooms is that there are a large variety of native languages present. This is a plus when a classroom teacher speaks the native language of the students. However, it is almost always impossibility for each teacher to be knowledgeable of every language represented in the classroom. Reyes (1992) documents that rarely is it crucial for the classroom teacher to be proficient in all languages represented by students in the classroom. Yet, through the method explained here, any classroom teacher can or- chestrate a meaningful English Literacy lesson that includes writing in the students’ native language – even when the teacher does not know the native language. This approach, problem-posing and using native language writing as a scaffolding technique to promote English literacy, while exposing students to important content information, gives value to each student’s personal experience. The approach allows for adaptations to be made regarding language and background knowledge such as cultural experience and experience in formal literacy instruction. It also allows for flexibility in terms of educational aspirations of the students. A congressionally mandated longitudinal study (Ramírez, Yuen, Ramey, & Pasta, 1991) a decade ago found that in many, if not most, of bilingual, and English as a Second Language Classrooms in the U.S.A., students were forced to be passive learners with teacher dominance of virtually all classroom activities. The researchers reported the following. In over half of the interactions that teachers have with students, students do not produce any language...When students do respond, typically they are providing simple information recall. Rather than being provided with the opportunity to generate original statements, students are asked to provide simple discrete close ended or patterned responses. Not only does this pattern of teacher/student interaction limit a student’s opportunity to create and manipulate language freely, but it also limits the student’s ability to engage in more complex learning...(pp. 421-422). The intention with this chapter is to document ways in which native language writing can be used to enhance English literacy instruction. The most convincing evidence of alternative theoretical propositions regarding bilingual writing comes from research of dual language or two-way bilingual immersion programs (Cummins, 1999). Evaluations of these programs have consistently revealed over the course of elementary school for both language minority and language majority students. For example, outcomes from the Oyster Bilingual School in Washington, DC illustrate this. The bilingual program involves instruction in both Spanish and English for about half of the lesson time. Students read and write in both languages each day so that development of literacy in both languages in simultaneous. The academic results of this program shows, for example, in grade 3 level Reading, Mathematics, Language and Sciences scores were 1.6 – 1.8 median grade equivalents above norms (percentiles 74-81). The grade 6 grade equivalents were 4.4 – 6.2 above norms (percentiles 85 – 96) (Freeman, 1998). Freeman (1998) provides detailed discourse analyses that illustrate how the interactions between educators and students in Oys- ter bilingual school ‘refuse’ the discourse of subordination that characterizes the treatment of minorities in most conventional schools. 234 QUINTERO

2 RATIONALE The problem-posing (Freire, 1985) teaching method is a way to facilitate student choice and generative work that is integrally related to students’ lives. The importance of personally meaningful contexts for writing activities for all learners has been argued by many (Fine, 1991; Giroux, 1987; Greene, 1995). In many settings where students have been previously unsuccessful in school, the need to create meaningful links between school and community has been proven effective (Reyes, 1992). Evidence shows that the social context of the setting in which problem- posing is implemented enhances the opportunity to focus on the often dramatic tran- sitions that the students experience as they move from one sociocultural context to another (Quintero & Rummel, 1998; Rummel & Quintero, 1997; Shor; 1987). The method lends itself to focus events in students' personal lives while they are learning English and other content area subjects. Comber and Kamler (1997) maintain that we must continue to examine the different versions of critical literacy that emerge and show the complex pictures of what pedagogies for critical literacy look like in different settings and educational contexts. These researchers believe that detailed narratives are needed which show the negotiations of critical literacies in sites around the world. In addition, they (Comber & Kamler; 1997) say that it is urgent that we continue to discuss the way in which power is exercised through textual practices during these times of demands for accountability in the production of students with particular sets of competencies and in the context of managerialist discourses which threaten to inhibit the freedom and power of individual and group student text in the literacy classroom. It has been shown that acculturation and language acquisition is impacted by the process of aligning new societal expectations and requirements with previous cultural norms, individual perceptions, and experiences is preeminent in immigrants' lives. Yet these urgent issues are often ignored (Ullman, 1997; Zou, 1998). Franken and Haslett (this volume) report research documenting that when students are faced with a topic on which they have little content and domain-specific knowledge, interaction is significantly helpful for understanding text. By virtue of the fact that many immigrant students come from such a variety of backgrounds with such different ‘funds of knowledge’ as Moll (1987) reports, it is almost inevitable that many occasions will arise in the literacy classroom when one or more students have a lack of background knowledge on a topic. Edelsky (1986) maintains that writing is a recursive process in which students actively hypothesize about the aspects of writing that can be applied across languages. DeSilva (1998) and Perez (1998) maintain that children can develop their knowledge of second language writing and speaking conventions by using what they understand about writing in their native language. Many researchers and classroom teachers have shown that language interaction in the form of student-centered discussions and demonstrations, language experience story writing and reading, and holistic literacy development allow for native language literacy to enhance English language literacy development (Edelsky, 1986; Goodman & Goodman, 1987; Ha- kuta, 1986; Hakuta & Garcia, 1989; Hudelson, 1994; Peregoy & Boyle, 2000). TALKING CAN MAKE WRITING HARDER 235

In academic contexts, the notion that L1 interferes with the learning of L2, is known to not be supported by research evidence (Larsen-Freeman & Long, 1991). It is clear that L1 serves a function in early L2 acquisition, but it is a supportive role rather than a negative one (Ovando & Collier, 1998). In the beginning stages of L2 acquisition, acquires lean on their L2 knowledge to analyze patterns in L2, and the6 subconsciously apply some structures from L2 to L2 in the4 early stages of interlanguage development. Most linguists look upon this process as a positive use of L1 knowledge. (Ovando & Collier, 1998: 95) Ovando and Collier (1998) go on to state the importance of educating school staff and parents who are unfamiliar with this information about the importance of students using native language in learning. Research of Cashion and Eagan (1990) in Canadian French immersion programs and of Verhoeven (1991) among minority language students in the Netherlands shows that transfer across languages is two-way (from L1 to L2 and then back from L2 to L1) if the sociolinguistic and educational conditions are right. In this short chapter, examples of three classrooms using the problem posing/critical literacy approach are presented. Students in these classrooms participate in activities that involved writing in both native languages and English. The three classroom examples were chosen to illustrate how this method of instruction en- courages students ‘to learn about and interpret the world and reflect upon themselves in relation to people and events around them...and to explain, analyze, argue about, and act upon the world’ (Hudelson, 1994: 130) while they negotiate their English literacy development using their native language as a scaffold. One classroom is a middle school classroom of 22 English Language Learners with a mixture of native languages including Spanish, Hmong, Burmese, and Somali. Another classroom is a high school/adult basic education class of 15 English Language Learners who speak Hmong, Vietnamese, Russian, Chinese, Spanish, and Polish as native languages. The third classroom is a first grade classroom with 23 students from several native language groups.

3 METHOD The design of the study reported here is qualitative. According to Denzin and Lin- coln (2000), qualitative research is an interdisciplinary, transdisciplinary, and sometimes counterdisciplinary field. They say that the word qualitative implies an emphasis on processes and meanings that are not rigorously measured or measured at all in terms of quantity, amount, intensity, or frequency. The qualitative design supports the study of the intersection of languages (native and target), culture, instructional contexts, and the critical use of language by the learners. The design is the most appropriate for looking at teaching and learning that uses a critical theory framework. This small on-going study uses an observational case study format described by Bogdan and Biklen (1998). Central to this research design are data-collection techniques such as participant-observation and formal and informal interviews. It is believed that by using these methodological techniques, the intersection of teaching 236 QUINTERO and learning, the intersections of different languages and literacies, and the intersections of the relationships among students and teachers can be documented in the type of rich contextual detail that will be useful to educators in the field. Further, because the study is designed to learn from a varied, but specific population of learners in a context with a critical literacy focus, the design lent itself to the data collection and analysis purposes that can show a variety of outcomes. The data is analyzed using a framework of critical theory. The data are examined in terms of whether or not the students are using literacy to construct meaning and critically use language (oral and written) as a means of expression, interpretation and/or transformation in their lives. Of course, the expression and transformations include evidence of the students’ learning academic English literacy. In addition, the framework also demands that the politics of knowledge and the origins of sustained inequities of modern society be examined by learners in the context of their studies.

4 OBSERVATIONAL CASE STUDIES The classes observed and studied used the problem-posing activities of this methodology that provide a nonlinear framework in which complex information can be connected with students’ lives and a context which supports critical literacy. The three cases, which represent three different age groups of learners, in three different schools, in three different communities, all demonstrate native language writing using problem posing methodology in lessons with an autobiographical narrative theme. They use reflection, discussion, drawing, writing in their native languages, writing in English, and oral language as they explain their work to others. In the process of participating in various activities to create the autobiographical narratives, the students use many strategies necessary for successful academic English (McCloskey & Stack, 2000). Freire’s (1973) problem-posing format of listening, dialogue, action is utilized in flexible ways. Students participate in the activities by relating their own experience and/or choose other aspects related to the topic of study to further their reading, writing, and background knowledge. They reflect upon the community they have entered, pursue their particular needs of English language literacy, and learn important content knowledge required at their particular level of schooling. Through the process of participating in the activities and readings, and reflection, the literacy student is participating in a complex form of autobiographical narrative. The students’ participation in this method is autobiographical in the sense that the student writer is constantly encouraged to reflect on her/his own experience. Personal narrative illustrates the voices of the participants, in their cultural contexts as they tell about their experiences and explain ongoing efforts at agency and transformation. In the Listening part of the class there may be an introduction by the teacher that includes social and historical information related to the topic or theme of study. Then the teacher suggests related individual reflective journaling activities by students to reacquaint themselves with their own past experiences. This also includes prereading activities for focusing purposes. Then, the students may then listen to specific information about the writing tasks ahead. TALKING CAN MAKE WRITING HARDER 237

In the Dialogue part, the students’ dialogue with each other about their reflective writings or discuss other tasks given by the teacher. The students dialogue with peers through small and large group discussions. In the Action part of the lesson, the students create their compositions in their native language. Then using oral language in both their native language and the target language of English they explain (not translate directly) what they wrote. Then after this oral language interaction, they tackle the task of rewriting the assignment in English. The problem-posing activities are recursive activities, not a lesson plan in a preconceived, rigid sequence.

4.1 Case one The first ‘Classroom Case’ was compiled with the help of 7th grade English Lan- guage Learners and their teacher. The school is in a large midwestern city and in a neighborhood that is home to many recent immigrants. In this particular class, the majority of students are native speakers of Spanish, some from the United States, Mexico and Central America. There are a few Hmong speakers, one speaker of Burmese, and one speaker of Somali in the class.

Listening The teacher showed the class a photo of herself when she was seven years old and then pulled out a map of Minnesota. She showed them the small Northern Minnesota town where her own immigrant family lived when she was a child. Then she told a story of how when she was seven, her father lost his job on the railroad and there was no other work in their town. So the family decided to move to California. Her dad went first and saved money, and then sent for the family some months later. Her mother, with her four young children, flew on an economy flight leaving at midnight and arriving in California at dawn. ‘I still remember how impressed I was by the sight of the toy city below...and how happy my dad was to see us,’ the teacher explained. The teacher then showed the drawing she had made in four parts that visually told her story which was drawn on one large sheet of butcher paper.

Dialogue Here the teacher asked the students ‘What are sentences to explain each of the pictures?’ (They worked on keeping the story in past tense and had to learn a couple of new words like airport and toy).

Action Teacher then gave students the a lesson guide that included the following instructions: Your Story Assignment 1) Write a story about your childhood in your 1st language. Write the story in the past tense. 2) Write the same story in English. Divide it into 4 parts to go with 4 drawings. 238 QUINTERO

3) Draw 4 pictures that show the 4 parts of your story. 4) Tell your story in English to 3 other students and 1 teacher. 5) Rewrite your story in the present tense. 6) Hand everything in to your teacher to get a grade. This guide, along with the task required of reading their stories to others, satisfies an important component of providing opportunities for the development of academic English. Cummins (1994) outlines a framework for English Language Learners that high- lights distinction between conversational and academic communication. He notes that persuading another person of one’s point of view and writing an essay demands cognitively demanding skills necessary for academic language use. The teacher commented that ‘It was lots of fun, very interactive, and encouraged the students to think and speak in meaningful complex sentences. I’ll definitely do something like this again!’ A student who had emigrated from Mexico to Minnesota told her story first with pictures, then she wrote the story in her native language of Spanish, and then she wrote the story in English. A student from El Salvador told his story of going from El Salvador to Mexico to the United States. His first work sample was in drawing form, the second was a story written in his native language of Spanish, and the third, a story as he wrote it in English.

4.2 Case two The second ‘Classroom Case’ was compiled with the help of high school/adult basic education English Language Learners and their student teacher. The school is in a mid-sized midwestern city. In this particular class, the students are native speakers of Hmong, Polish, Chinese, Vietnamese, Spanish, and Russian.

Listening The teacher told a story about a childhood experience in which she had learned a personal lesson: My brother and I were fighting about something – he started throwing rocks at me and I ran and hid in my dad’s truck thinking he would stop, but he threw one anyway and it cracked the windshield. When my dad came out he was very angry at us for breaking the window and told us were no longer going fishing and we would have to pay for the broken window with our allowance. We both cried because we really wanted to go fishing and because it would take us forever to pay for the window. I learned that my brother and I should fight and that we should help each other to stay out of trouble, even if we’re mad at each other. Dialogue Using open-ended questions, the teacher led a discussion about the teacher’s story and asked students to discuss in small groups a similar type of story from their own lives.

Action She then asked the students to write their story in their native language and draw a summary of their stories. Then the students took turns telling their stories in English. TALKING CAN MAKE WRITING HARDER 239

Finally, all students wrote a paragraph in English (not a translation of their native writing, but an English writing version appropriate for their level of development in English literacy) about their ‘learning’ story. An adult student recently arrived from Poland and just beginning her English language study, wrote an extensive story in her native Polish. At the bottom of the page she drew small drawings to depict the story and then began to paraphrase her story in her newly acquired English to a classmate. A Rubric/Scoring Guide used to evaluate the students writing with this activity follows. Any scoring guide can be adapted according to grade level, instructional goals, and assessment requirements at either national or state level. In this case, the teacher used the rubric guide as a further teaching tool and as a strategy for students to use to participate in self-evaluation and peer evaluation. The students were initially confused by some of the language used in the guide. Yet, once the teacher discussed and explained aspects covered by the guide, the students spent much time and energy self-evaluating and revising their writing. As Kuiken and Vedder (this volume) in this collection report in their particular study, What is fundamental is that in the interactional process, in a joint effort, learners search for the correct solution. In producing the target language they may notice the gap between what they want to say and what they can say.

(Kuiken & Vedder, this volume: 179)

4.2.1 Exceptional writing: Demonstrates outstanding competency

• Tells a coherent; uses a variety of narrative with sophistication • Orients the reader by describing the scene and people • Has an authentic voice, revealing an attitude toward the incident • Has exceptional sentence structure, uses language with imagination, is generally free from errors in grammar, usage, and conventions of written English

4.2.2 Effective writing: Demonstrates clear competency

• Describes a well-told incident but may lack coherence • Implies or states the significance, but may be superficial or tacked on • Reveals an earnest storyteller • Adequate control of sentence structure • May have a few errors in grammar usage and conventions of Written English

4.2.3 Basic writing: Demonstrates basic competency

• Tells of a specific incident, but shows a limited use of strategies, generally brief • May neglect narrative • May imply or state significance in a limited way • Fails to relate incident with appropriate detail 240 QUINTERO

• May have errors in grammar usage and conventions of Written English

4.2.4 Developing writing: Demonstrates beginning competency

• May present the incident in a general or fragmentary way • May limit or omit reference to context • Reflection is superficial or missing • Shows minimal evidence of personal involvement • Little or no details; sentences may be short or confusing • May have a variety of repeated errors in grammar usage and conventions of Written English; these errors cause confusion

4.2.5 Beginning writing: Demonstrates very little competency

• Reader may have to infer incident • Focus may be on others rather than self • Context very limited or missing • Little or no significance • Little or no evidence or personal involvement • Frequent lapses in sentence control • May have pervasive errors in grammar usage and conventions of Written Eng- lish

At the time of this writing, the student teacher was using this rubric as a structured guide for self-evaluation for the students. They were benefiting from practicing their use of academic English terminology and grammar as it applied to their personal stories.

4.3 Case three The following observational case documents a lesson in a first grade classroom in an urban school in a large midwestern city. The teacher is a twenty-year veteran teacher within the district, a woman of Irish American decent. The student teacher with her is a Hmong male who is from the community where the school is situated. The students in the class consist of sixteen Hmong children, three African American children, and one child from South America. For the following lesson, the teacher used the storybook, Whispering Cloth, by Pegi Deitz Shea, illustrated by Anita Reggio and stitched by Youa Yang

Listening The teacher began by gathering the children around her in the classroom center area where she unfolded several quilts. She reminded them of previous discussions and stories they had shared about quilts. Then she showed a weaving from Ireland and explained that it was from the country her family came from. Then, she held up a TALKING CAN MAKE WRITING HARDER 241 large, colorful ‘story cloth’ which had been made by one of the school staff’s relatives who is Hmong.

Dialogue The teacher asked: ‘Do you think a quilt could tell a story? Do you think you can hear a story from a cloth?’ The children discussed briefly what they thought about the question.

Action Then the teacher passed the folded cloth around the circle, so that each first grader could ‘listen’ to the cloth. Then, she showed the class the book and told them just a little about the book. ‘It is a story about a Hmong girl and her grandmother who live in a refugee camp in Thailand. Grandmother is teaching Mai how to make story cloths and Mai creates one that tells her story.’ Then the teacher showed the bilingual glossary in the book with Hmong words and English translations, and explained that she would read the story in English in a few minutes, but that first, the student teacher would read it in Hmong.

Listening The story was read in Hmong. The students who did not understand Hmong appeared to be fascinated by the reading in spite of not comprehending.

Dialogue The teacher then asked: ‘Can you guess what the story was about based upon Mr. Z’s intonations and the pictures?’

Action Then the teacher read the story in English. The student teacher explains to the students that he has written a letter to their families explaining what they are learning about. The letter is written in English, Hmong, and Spanish. In the letter, the teachers ask the parents if their child could either share a story cloth, quilt, or other artifact that tells a family story. When the items are brought to school, extension activities will be implemented. The class will make a class story cloth with a contribution from each student’s drawing and writing (native language or English or both) during the next few days. In an interview, the teacher talks about the problem-posing aspects of her work with her first graders: I think one area that has really helped strengthen me in taking risks in curriculum has been embracing an integrated approach, and interdisciplinary strategies. For example, ...writing really was a turning point for me. You can incorporate a lot of learning into the writing process and writing strategies just by following the brainstorming, prewriting activities, what your follow up is. You don’t have to be ‘married to a manual’, you don’t have to take the curriculum and follow every scope and sequence activity. There are things that I can do to put together the objectives that are more appropriate to the needs of the students in my classroom and not let the manual dictate that. It’s well worth the risk to go beyond that and to extend curriculum by embracing an integrated approach. .... I’m probably not a traditional teacher.... Whether they are first graders or 242 QUINTERO

third graders they really respond when children write genuinely from their hearts in their home language and in English. To give children a voice is a risk because a lot of teachers want control. I think particularly with students at risk, you have to help let them know that it is okay to talk about things or it is okay for them to question because who wants robot kids? Along with that you have a climate that offers children the opportunity to question and you have to set up an atmosphere where it’s okay to make mistakes. I really applaud kids that are willing to take a risk in not knowing what the outcome will be.

These teachers combine their documentation and recording of children’s literacy development in several ways. For example, they regularly keep anecdotal notes on five-by-seven cards to document what the children know and can do. They are an important aid in helping guide progress in language and literacy. One card documented a student who had recently arrived from South America:

11/4 First English words!!!: ‘Shut up.’ 11/15 Named colors on the patterned calendar: orange, blue, black. 12/1 Shows a grasp of basic English words: Numbers to 20, many colors, school words. 1/8 To Tong, ‘That's my chair. It has my sweater.’

The teachers also keep a literacy portfolio on each student. This assessment of writing considers such criteria as: content, organization, voice and style, use of language, and use of conventions, with scores ranging from 5 (high) to 1 (low). The dimensions used for evaluation are created from a combination of Tierney's (1991: 134) continuum of portfolio descriptors and portfolio evaluation guidelines from Performance and Portfolio Assessment for Language Minority Students (1992) published by the National Clearinghouse for Bilingual Education. Portfolios are evaluated on the following criteria:

• fluency or increase in length • variety of topic especially topics with bicultural content • presence of voice especially as expressed through cultural detail • flexibility in types of writing, approaches, problem solving and depth of reading responses • communicative sense or seeming awareness of a writing audience • syntactic development including increase in types of sentences written, an increase in complexity of sentences and a growing awareness of punctuation use.

Another very important evaluative indicator for the teachers in this program is the reaction of the parents to the learning situation for their children. The teacher and student teacher, in this case, regularly interview parents. In particular, there has been enthusiastic response to native language writing and speaking in the classroom while their children study the regular state curriculum in English. A few parent comments (translated from Hmong) show this. TALKING CAN MAKE WRITING HARDER 243

‘My son speaks both Hmong and English to me, but I am happy to see that the Hmong language is being taught.’

‘We only talk in Hmong so I know they can speak it well, but the are just learning to read and write it. It is great that they are learning to and write their own language. It will be useful.’ Whereas in this first grade classroom, during the time of observation, much assessment to date had been anecdotal, state standardized tests were administered later in the school year. This is a school that has an ELL population of 68%. The number of families who are living at or beneath poverty level is 86%. In the state tests at Grade 3 and 5 in Spring 200l the students scored higher than many suburban schools in reading. Grade 3 reading scores went from 8% in 2000 to 47% in 2001. Their literacy program is balanced at all levels. They teach skills within the context of shared reading and writing, guided reading and writing and independent reading and writing.

5 DATA ANALYSIS AND PRELIMINARY FINDINGS Looking at these three briefly described examples and the data collected, a preliminary analysis can be made using the framework of critical theory. Critical literacy, as defined by the author, is a process of constructing meaning and critically using language (oral and written) as a means of expression, interpretation and/or transformation of our lives and the lives of those around us. This framework led to the catego- rization of findings into three categories of: expression, interpretation, and transformation. These three categories can be guides for discussion both learning in a general, deep content sense and in terms of academic literacy skills. In terms of expression, by using a theme of personal history narrative, the teachers gave all students an opportunity to be experts in the telling of their story. By en- couraging a three-tiered approach of drawing, native language writing, and English writing, the activity gave opportunity for all students, regardless of literacy proficiency, to tell their stories. In most cases in each of the classes the students were more expressive in terms of the content and details of their stories when writing in their native languages or drawing. In terms of interpretation, this method of using problem-posing gives ample opportunity for the students to reflect upon and discuss with others their interpretations of events of their immigration. In addition, in the third stage of the activity, when they were using English and telling what they had written to a classmate, interpretation of the story based upon the language choices made became a very dynamic academic process. Finally, in terms of transformation, in this type of activity the opportunities are great. Every immigrant and refugee has experienced a transformation by moving from one country and culture to another. Learning to negotiate the new culture and new language provides for daily transformations. And very specifically, by attending to their own development of English literacy and academic English the students are generating more opportunities for further transformation. In activities in which cul- 244 QUINTERO ture, power, language, prior knowledge and current dilemmas are addressed are the opportunities for transformation the greatest. For example in the 7th grade class, Students 1 and 2 expressed their stories, interpreted them, and documented transformation in terms of their families’ immigration experiences. In terms of their literacy skills, the teacher documented through field notes that there was a variety of evidence of growth, and the teacher being grateful that the activity pushed them ‘to think and write in more complex and meaningful sentences’ than they had previously done. The second case showed students discussing history, politics, family childrearing practices in their examples. In terms of their writing in English, after having used their native languages in writing their initial version, they practiced the skill of interpretation without using direct translation. The teacher reported marked fluency in English writing in the weeks after the activity was done. The third case showed that the first graders were entranced with the learning task of thinking about whether or not a cloth can ‘tell a story.’ They participated actively in the discussion about who can tell stories, can an audience understand stories even when the language used to tell the story is not understood. Then, in terms of skills, they exhibited a working knowledge of interpretation, listening, decoding skills, and vocabulary development. All of the skills enhanced will be influential in the students’ telling, writing, and reading of their ‘own stories’ as they create the classroom story cloth.

6 CONCLUSION In conclusion, it is hoped that these brief glimpses into the on-going study will add to the field of literacy instruction for English Language Learners. First, it is hoped that the examples illustrate the endless potential for content learning and specific English skill learning using native language writing. Secondly, by describing the three very different examples of classroom uses of native language writing by teachers who do not speak the language of most of their students, it illustrates that such a native language writing component can be implemented to complement literacy lessons in many contexts. Finally, the author hopes that in the future, students will not be shocked when their teachers included native language writing as a part of their assignments, ‘What? Maestra (teacher), you want me to write in Spanish?’ Change in the deep structure will come only when educators walk into their classrooms burdened not by the anger of the past and the disdain of the present, but with their own identities focused on transforming the social futures towards which their students are traveling. (Cummins, 1999: 2)

REFERENCES

Abbott, R. D., & Berninger, V. W. (1993). Structural equation modeling of relationships among developmental skills and writing skills in primary- and intermediate-grade writers. Journal of Educational Psychology, 85 (3), 478-508. Acock, A.C. (1997). Working with missing values. Family Science Review, 10 (1), 76-102. Adams, M. J. (1990). Beginning to read: Thinking and learning about print. Cambridge, MA: MIT Press. Adams, M. J., & Huggins, A. W. (1985). The growth of children's sight vocabulary: A quick test with education and theoretical implications. Reading Research Quarterly, 20, 262-281. Akyel, A. (1994). First language use in EFL writing: Planning in Turkish vs. planning in English. Interna- tional Journal of Applied Linguistics, 4 (2), 169-176. Akyel, A. (1994). First language use in EFL writing: Planning in Turkish vs. planning in English. Inter- national Journal of Applied Linguistics, 4, 169-176. Alderson, J. C. (1984). Reading in a foreign language: A reading or a language problem? In J. C. Alder- son & A. H. Urquhart (Eds.), Reading in a foreign language (pp. 1-24). London: Longman. Aljaafreh, A., & Lantolf, J. (1994). Negative feedback as regulation and second language learning in the zone of proximal development. Modern Language Journal, 78, 465-483. Allison, D., & Tauroza, S. (1995). The effect of discourse organisation on lecture comprehension. English for Specific Purposes, 14 (2), 157-173. Ames, C. (1992). Classrooms: Goals, structures, and student motivation. Journal of Educational Psychol- ogy, 84, 261-271. Anderson, J. R. (1995). Cognitive psychology and its implications (4th ed.). New York: Freeman. Anzai, Y., & Uchida, N. (1981). Kodomo wa Ikani Sakubun o Kakuka [How children produce writing]. Japanese Journal of Educational Psychology, 29, 323 - 332. Appel, R., & Vermeer, A. (1998). Speeding Up Second Language Vocabulary Acquisition of Minority Children. Language and Education, 12(3), 159-173. Applebee, A. N. (1984). Contexts for learning to write: Studies of secondary school writing instruction. Norwood, NJ: Ablex. Arndt, V. (1987). Six writers in search of texts: a protocol-based study of L1 and L2 writing. ELT Jour- nal, 41(4), 257-266. August, D. & Hakuta, K. (1997). Improving schooling for language-minority children. Washington, DC: National Academy Press. Austin, J. T., & Vancouver, J. B. (1996). Goal constructs in psychology: Structure, process, and content. Psychological Bulletin, 120, 338-375. Bachman, L. (1990). Fundamental considerations in language testing. Oxford: Oxford University Press. Baddeley, A. (2000). The episodic buffer: a new component of working memory? Trends in Cognitive Sciences, 4 (11), 417-423. Bandura, A. (1986). Social foundations of thought and action: A social-cognitive theory. Englewood Cliffs, NJ: Prentice-Hall. Barbier, M.-L. (1996). Functional features of the text production in a second language, compared to the first language. Proceedings of the European Writing Conferences, Barcelona, October 23-25. Barbier, M.-L. (1998a). Rédaction en langue première et en langue seconde. Indicateurs temporels et coût cognitif [Writing in first and second language. Temporal variables and cognitive load]. Paris: Presses Universitaires du Septentrion. Barbier, M.-L. (1998b). Rédaction en langue première et en langue seconde: comparaison de la gestion des processus et des ressources cognitives [Comparison between writing processes and cognitive resources in first and second language]. Psychologie Française, 43(4),361-370. Barbier, M.-L. (1999). Written production in first and second language: Activation of writing processes and cognitive ressources. In E. Espéret & L. Tolchinsky (Eds.), Proceedings of the 1998 European Writing Conference (pp. 369-374). Poitiers: MSHS. Beaufort, A. (1997). Operationalizing the concept of discourse community: A case study of one institutional site of composing. Research in the Teaching of English, 31, 486-529. Belinchón, M., Rivière, A. & Igoa, J. M. (1992). Psicología del lenguaje: Investigación y teoría. Madrid: Trotta Bell, J. S. (1995). The relationship between L1 & L2 literacy: Some complicating factors. TESOL Quar- terly, 29 (4), 687-704.

246

Benton, S. L., Kraft, R. G., Glover, J. A., & Plake, B. S. (1984). Cognitive capacity differences among writers. Journal of Educational Psychology, 76(5), 820-834. Bereiter, C. (1980). Development in writing. In L.W. Gregg and E.R. Steinberg (Eds.), Cognitive processes in writing (pp. 73-93). Hillsdale, NJ: Erlbaum. Bereiter, C. (1997). Situated cognition and how to overcome it. In Kirshner, D. & Whitson, J. (Eds.), Situated cognition: Social, semiotic, and psychological perspectives (pp. 281-300). Mahwah, NJ: Erlbaum. Bereiter, C., & Scardamalia, M. (1982). From conversation to composition: The role of instruction in a developmental process. In R. Glaser (Ed.), Advances in instructional psychology (Vol. 2, pp. 1-64). Hillsdale, NJ: Lawrence Erlbaum Associates. Bereiter, C., & Scardamalia, M. (1987). The psychology of written composition. Mahwah, NJ: Erlbaum. and Associates. Bereiter, C., & Scardamalia, M. (1989). Intentional learning as the goal of instruction. In L.B. Resnick (Ed.), Knowing, learning and instruction: Essays in honor of Robert Glaser (pp. 361-392). Hillsdale, NJ: Erlbaum. Berkenkotter, C., & Huckin, T. (1995). Genre knowledge in disciplinary communication: Cogni- tion/culture/power. Hillsdale, NJ: Erlbaum. Berman, R. (1994). Learners' transfer of writing skills between languages. TESL Canada Journal, 12, 29- 46. Block, D. (2000). Problematizing interview data: voices in the mind’s machine. TESOL Quarterly, 34, 757-763. Boch, F. (1998). Les étudiants de premier cycle: discours sur leur prise de notes. In C. Fintz (Ed.), Contributions à la didactique du français dans l'enseignement supérieur: Bricolage ou rénovation ? [Post-graduate students’ speech about their notetaking. In C. Fintz (Ed.), Contributions for French instruction at university] (pp. 109-124) Paris: L'Harmattan. Boch, F. (1999). Pratiques d'écriture et de réécriture à l'Université. La prise de notes, entre le texte source et le texte cible [Writing at University. Notetaking: between primary text and target one]. Lille: Presses Universitaires du Septentrion. Bogdan, R. C., & Biklen, S.K. (1998). Qualitative research for education: An introduction to theory and methods (3rd edition). Boston: Allyn & Bacon. Bollen, K. A. (1989). Structural equations with latent variables. John Wiley & Sons: New York.Brindley, G. (1998). Outcomes-based assessment and reporting in language learning programmes: A review of the issues. Language Testing, 15, 45-85. Bosher, S. (1998). The composing processes of three South East Asian writers at the post-secondary level: an explanatory study. Journal of Second Language Writing, 7, 205-241. Breetvelt, I., Van den Bergh, H., Rijlaarsdam, G. (1996). Rereading and generating and their relation to text quality. An application of multilevel analysis on writing process data. In G. Rijlaarsdam, H. van den Bergh, & M. Couzijn (Eds.), Theories, models and methodology in writing research (pp. 10 - 21). Amsterdam: Amsterdam University Press. Brindley, G. (1998). Outcomes-based assessment and reporting in language learning programmes: A review of the issues. Language Testing, 15, 45-85. Britton, J., Burgess, T., Martin, N., McLeod, A., & Rosen, H. (1975). The development of writing abilities (11-18). London: Mcmillan. Broekkamp, H., and van den Bergh, H. (1996). Attention strategies in revising a foreign language text. In G. Rijlaarsdam, H. Van den Bergh, & M. Couzijn (Eds.), Theories, models and methodology in writing research (pp. 170-181), Amsterdam: Amsterdam University Press. Brooks, E. (1985). Case studies of the composing processes of five “unskilled” English-as-a-second- language writers. Unpublished doctoral dissertation. New York University. Brown, G. (1995). Dimensions of diffculty in listening comprehension. In D. Mendelsohn & J. Rubin (Eds.), A guide for the teaching of second language listening (pp. 59-73). San Diego, CA: Dominic Press. Brown, J. I., Fishco, V., & Hanna, G. (1993). Nelson-Denny reading test: part II comprehension test. Chicago, IL: The Riverside Brown, J. S., Collins, A., & Duguid, P. (1989). Situated cognition and the culture of learning. Educa- tional Rsearcher, 18, 32-42. Brown, J., Collins, A., & Duguid, P. (1989). Situated cognition and the culture of learning. Educational Researcher, 18, 1, 32-41.

REFERENCES 247

Bruce, B. (1983). Three perspectives on writing. Technical Report No. 41. Center for the Study of Read- ing, Champaign, IL: USA. Bruffee, K. (1984). Collaborative learning and the ‘conversation of mankind’. College English, 46, 635- 652. Bryson, M., & Scardamalia, M. (1991). Teaching writing to students at risk for academic failure. In B. Means, C. Chelemer, & M. S. Knapp (Eds.), Teaching advanced skills to at risk students: Views from research and practice (pp. 141-167). San Francisco: Jossey-Bass Publishers. Buck, G. (1992). Listening comprehension: construct validity and trait characteristics. Language Learn- ing, 42 (3), 313-357. Burtis, P. J., Bereiter, C., Scardamalia, M., & Tetroe, J. (1983). The development of planning in writing. In G. Wells & B. M. Kross (Eds.), Explorations in the development of writing (pp. 153-174). Hillsda- le, NJ: Erlbaum. Candlin, M. (1987). Towards task-based learning. In C. Candlin and D. Murphy (Eds.), Language learning tasks (pp. 5-22). Englewood Cliffs, NJ: Prentice Hall. Carretero, M., & Voss, J. F. (Eds.). (1994). Cognitive and instructional processes in History and the social sciences. Hillsdale, NJ: Lawrence Erlbaum. Carroll, B.J. (1980). Testing communicative performance. Oxford: Pergamon. Carson, J. & Kuehn, P. (1992). Evidence of transfer and loss in developing second language writers. Language Learning, 42, 157-182. Carson, J., Carrell, P., Silberstein, S., Kroll, B., & Kuehn, P. (1990). Reading-writing relationships in first and second language. TESOL Quarterly, 24, 245-66. Carter, M. (1990). The idea of expertise: An exploration of cognitive and social dimensions of writing. College Composition and Communication, 41, 265 – 286. Carver, C. & Scheier, M. (1981). Attention and self-regulation: A control theory approach to human behavior. New York: Springer-Verlag. Casanave, C. (1992). Cultural diversity and socialization: A case study of a Hispanic woman in a doctoral program in sociology. In D. Murray (Ed.), Diversity as a resource: Fefining cultural literacy (pp. 148-180). Alexandria, VI: TESOL. Cashion, M. & Eagan, R. (1990). Spontaneous reading and writing in English by students in total French immersion : summary of final report. English Quaterly, 22(1), 30-44. Castley, A. (1998). Practical Spelling. New York: Learning Express. Chall, J.S., Jacobs, V.A., & Baldwin, L.E. (1990). The Reading crisis: Why poor children fall behind. Cambridge, MA: Harvard University Press. Chaudron, C., Loschky, L & Cook, J. (1994). Second language listening comprehension and lecture notetaking. In J. Flowerdew (Ed.), Academic listening: Research perspectives (pp. 75-92). Cambridge: Cambridge University Press. Chenoweth, N. A. & Hayes, J. R. (2001). Fluency in writing. Written Communication,18, 80-98. Chesnet, D., Guillabert, F., & Espéret, E. (1994). G-Studio: un logiciel pour l'étude en temps réel des paramètres temporels de la production écrite [G-Studio : a software program for on-line analysis of temporal variables during written production]. L'Année Psychologique, 94, 283-294. Chin, E. (1994). Redefining “context” in research on writing. Written Communication, 11, 445-482. Clachar, A. (1999). It’s not just cognition: The effect of emotion on multiple-level discourse processing in second-language writing. Language Sciences, 21, 31-60. Clay, M. M. (1988). What did I write? Exeter, N.H. Heinemann. Clerehan, R. (1995). Taking it down: Notetaking for L1 and L2 students. English for Specific Purposes, 14(2), 137-155. Cliff, A. F. (2000). Dissonance in first-year student's reflections on their learning. European Journal of Psychology of Education, 15, 49-60. Coe, R. (1988). Toward a grammar of passages. Carbondale and Edwardsville: Southern Illinois Univer- sity Press. Cohen, A. D., Weaver, S. J., & Li, T-Y. (1998). The impact of strategies-based instruction on speaking a foreign language. In A. D. Cohen (ed.), Strategies in learning and using a second language (pp. 107- 156). London: Longman. Cohen, A.D. & Brooks-Carson, A. (2001). Research on direct vs. translated writing processes: Implica- tions for assessment. Modern Language Journal, 85, 169-188.

248

Coirier, P. (1996). Composing argumentative texts: Cognitive and or text complexity. In G. Rijlaarsdam, H. van den Bergh and M. Couzijn (Eds.), Theories, models and methodology in writing research (pp. 317-338). Amsterdam: Amsterdam University Press. Comber, B., & Kamler, B. (1997). Critical literacies: Politicising the language classroom. Paper submitted for interpretations. Language and Literacy Research Centre, University of South Australia, Un- derdale. Connor, U. (1996). Contrastive rhetoric: Cross-cultural aspects of second-language writing. Cambridge: Cambridge University Press. Conrad, S., & Goldstein, L. (1999). ESL student revision after teacher-written comments: Text, contexts, and individuals. Journal of Second Language Writing, 8, 147-179. Cook, V. J. (1991). Second language learning and language teaching. London: Edward Arnold. Cooper, P.L. (1984). The assessment of writing ability: A review of research. Princeton, N.J.: Educational Testing Service. (GRE Board Research Report GREB No. 82-15R / ETS Research Report 84-12.) Cumming, A. (1986). Intentional learning as a principle for ESL writing instruction: A case study. TESL Canada Journal, Special Issue 1, 69-83. Cumming, A. (1989). Writing expertise and second language proficiency. Language Learning, 39, 81- 141. Cumming, A. (1990). Metalinguistic and ideational thinking in second language composing. Written Communication, 7(4), 482-511. Cumming, A. (1995). Fostering writing expertise in ESL composition instruction: Modeling and evaluation. In D. Belcher & G. Braine (Eds.), Academic writing in a second language: Essays on research and pedagogy (pp. 375-397). Norwood, NJ: Ablex. Cumming, A. (1998). Theoretical perspective on writing. Annual Review of Applied Linguistics, 18, 61, 78. Cumming, A. (2001). The difficulty of standards, for example in second language writing. In T. Silva & P. Matsuda (Eds.), On second language writing (pp. 209-229). Mahwah, NJ: Erlbaum. Cumming, A. (In press a). Experienced ESL/EFL writing instructors' conceptualizations of their teaching: Curriculum options and implications. To appear in B. Kroll (Ed.), Exploring the dynamics of second- language writing. Cambridge: Cambridge University Press. Cumming, A. (In press b). Learning to write in a second language: Two decades of research. To appear in R. Manchon (Ed.), Writing in the L2 classroom: Issues in research and pedagogy. Special Issue of Cuadernos de Filologia Inglesa, University of Murcia, Spain. Cumming, A., & Mellow, D. (1996). An investigation into the validity of written indicators of second language proficiency. In A. Cumming & R. Berwick (Eds.), Validation in language testing (pp. 72- 93). Clevedon, UK: Multilingual Matters. Cumming, A., & Riazi, A. (2000). Building models of adult second-language writing instruction. Learn- ing and Instruction, 10, 55-71. Cumming, A., Rebuffot, J. & Ledwell, M. (1989). Reading and summarizing challenging texts in first and second languages. Reading and Writing: An Interdisciplinary Journal, 2, 201-219. Cumming, A., Rebuffot, J., & Ledwell, M. (1989). Reading and summarizing challenging texts in first and second languages. Reading and Writing, 2, 201- 219. Cummins, J. (1980). The cross-lingual dimensions of language proficiency: implications for bilingual education and the optimal age question. TESOL Quaterly, 14, 175-187. Cummins, J. (1991). Interdependence of first and second language proficiency in bilingual children. In E. Bialystok (Ed.), Language processing in bilingual children (pp.70-89). Cambridge: Cambridge Uni- versity Press. de Beaugrande, R. A. & Dressler, W.V. (1981). Introduction to text linguistics. New York: Longman de Bot, K. (1996). Review article: The psycholinguistics of the output hypothesis. Language Learning, 46, 529-555. de Groot, A. M. B., & Kroll, J. F. (1997). Tutorials in bilingualism: Psycholinguistic perspectives. Mah- wah, NJ: Lawrence Erlbaum Associates. DeSilva, A. D. (1998). Emergent Spanish writing of a second grader in a whole language classroom. In B. Perez (Ed.), Sociocultural contexts of language and literacy. Devine, J., Railey, K., & Boshoff, P. (1993). The implications of cognitive models in L1 and L2 writing. Journal of Second Language Writing, 2, 203-225. Donato, R., & McCormick, (1994). A sociocultural perspective on language learning strategies: The role of mediation. Modern Language Journal, 78, 453-464.

REFERENCES 249

Doughty C., & J. Williams (eds) (1998). Focus on form in classroom second language acquisition. Cam- bridge: Cambridge University Press. Doyle, W. (1983). Academic work. Review of Educational Research, 53, 159-199. Duncan, S., & De Avila, E.A. (1987). LAS (Language Assessment Scales). Monterey, CA: McGraw-Hill. Dunkel, P.A. & Davis, J.N. (1994). The effect of rhetorical signalling cues on the recall of English lecture information by speakers of English as a native or second language. In J. Flowerdew (Ed.), Academic listening: Research perspectives (pp. 55-74). Cambridge: Cambridge University press. Durgunoglu, A.Y. & Oney, B. (1999). A cross-linguistic comparison of phonological awareness and word recognition. Reading and Writing, 11, 281-299. Durgunoglu, A.Y., Nagy, W.E., & Hancin-Bhatt, B.J. (1993). Cross-language transfer of phonological awareness. Journal of Educational Psychology, 85, 453-465. Edelsky, C. (1982). Writing in a bilingual program: The relation of L1 and L2 texts. TESOL Quarterly, 16, 211-228. Edelsky, C. (1982). Writing in a bilingual program: The relation of L1 and L2 texts. TESOL Quarterly, 21, 677-696. Edelsky, C. (1986). Writing in a bilingual program: Habia una vez. Norwood, NJ: Ablex. Ehri, L.C. (1997). Learning to read and learning to spell are one and the same, almost. In C. Perfetti, L. Rieben & M. Fayol (Eds.), Learning to spell. (pp. 237-269). Mahwah, NJ: Erlbaum Associates. Eigler, G., Jechle, T., Merziger, G., & Winter, A. (1990). Knowledge and text production. In H. Mandl, E. de Corte, N. Bennett and H. F. Friedrich (Eds.), Learning and instruction: European research in an international context (Vol. 2.2, pp. 341-356). Oxford: Pergamon Press. Elbow, P. (1981). Writing with power. New York: Oxford University Press. Ellis, R. (1994). The study of second language acquisition. Oxford: Oxford University Press. Ellis, R. (1999a). Learning a second language through interaction. Amsterdam: John Benjamins. Ellis, R. (1999b). Factors in the incidental acquisition of second language vocabulary from oral input. In R. Ellis (Ed.), Learning a second language through interaction (pp. 35-61). Amsterdam: John Ben- jamins. Ellis, R. (1999c). Retrospect and prospect. In R. Ellis (Ed.), Learning a second language through interaction (pp. 233-257). Amsterdam: John Benjamins. Ellis, R. (2000). Task-based research and language pedagogy. Language Teaching Research, 4(3), 193- 220. Ellis, R. (Ed.) (1994a). Learning a second language through interaction. Amsterdam: John Benjamins. Engber, C. A. (1995). The relationship of lexical proficiency to the quality of ESL compositions. Journal of Second Language Writing, 4(2), 139-155. Engeström, Y. & Cole, M. (1997). Situated cognition in search of an agenda. In Kirshner, D. & Whitson, J. (Eds.), Situated cognition: Social, semiotic, and psychological perspectives (pp. 301- 309). Mahwah, NJ: Erlbaum. Engeström, Y. (1991). Overcoming the encapsulation of school learning. Learning and Instruction, 1, 243-259. Engeström, Y., Miettinen, R. & Punamaki, R. (Eds.) (1999). Perspectives on activity theory. Cambridge: Cambridge University Press. Ericsson, K. A., & Delaney, P. F. (1999). Long-term working memory as an alternative to capacity models of working memory in everyday skilled performance. In A. Miyake. & P. Shah (Eds.), Models of working memory: mechanisms of active maintenance and executive control, (pp. 257-297). Cam- bridge: Cambridge University Press. Ericsson, K. A., & Simon, H. A. (1993). Protocol analysis: Verbal reports as data. [Revised edition]. Cambridge, MA: The MIT Press. Fagan, W.T. & Hayden, H.M. (1988). Writing processes of French and English of fifth grade French immersion students. The Canadian Modern Languages Review, 44, 653-668. Faigley, L, Cherry, R. D., Jolliffe, D. A., & Skinner, A. M. (1985). Assessing writers’ knowledge and processes of composing. Norwood, NJ: Ablex. Faigley, L. (1986). Competing theories of process: A critique and a proposal. College English, 48, 527- 542. Faraco, M. (1997a). Technique de prise de notes en français spécialisé. [Notetaking techniques in French for specific purpose]. Le Français dans le Monde, 287, 38-40.

250

Faraco, M. (1997b). Etude longitudinale de la prise de notes d'un cours universitaire français: le cas d'étu- diants étrangers d'un cursus européen [Longitudinal study about notetaking during a French university lecture : some european students cases]. Association for Specific Purposes, 15-18, 41-54. Faraco, M. (2000). Prises de notes: Quelles compétences pour les européens ? [Notetaking : which com- petences for Europeans ?], In L. Collès, J.-L. Dufays, G. Fabry & C. Maeder (Eds.), Didactique des langues romanes, Le développement de compétences chez l'apprenant [Roman language didactics. Learner’s competencies development] (pp. 107-112). Dordrecht: De Boeck. Fayol, M. (1991). From sentence production to text production: Investigating fundamental processes. Fayol, M. (1999). From on-line management problems to strategies in written composition. In G. Ri- jlaarsdam & E. Espéret (Series eds.) & M. Torrance & G. Jeffery (Vol. eds.), Studies in writing: volume 3.The Cognitive Demands of Writing. Processing Capacity and Working Memory in Text Pro- ductionl. [pp. 13-23]. Amsterdam: Amsterdam University Press. Fine, M. (1991). Framing dropouts: Notes on the politics of an urban public high school. Albany, NY: SUNY Press. Firth, A. & Wagner, J. (1997). On discourse, communication, and (some) fundamental concepts in SLA research. Modern Language Journal, 81, 285-300. Fitzgerald, J., & Shanahan, T. (2000). Reading and writing relations and their development. Educational Psychologist, 35, 39-50. Flavell, J. H. (1985). Cognitive development. Englewood Cliffs, NJ: Prentice Hall. Flege, J. E. (1999). Age of learning and second language speech. In D. Birdsong (Ed.), New perspectives on the critical period for second language acquisition. (pp. 101-132). Mahwah, NJ: Erlbaum. Fleishman, J., & Benson, J. (1987). Using LISREL to evaluate measurement models and scale reliability. Educational and Psychological Measurement, 47(4), 925-939. Flower, L. (1994). The construction of negotiated meaning. A social cognitive theory of writing. Carbon- dale: Southern Illinois University press. Flower, L. S. & Hayes, J.R. (1980). The dynamics of composing: Making plans and juggling with constraints. In L.W. Gregg & E.R. Steinberg (Eds.) Cognitive processes in writing (pp. 31-50). Hillsdale, N.J.: Lawrence Erlbaum. Flower, L., & Hayes, J. R. (1981). A cognitive process theory of writing. College Composition and Com- munication, 32, 365-387. Flower, L., Schriver, K. A., Carey, L., Haas, C., & Hayes, J. R. (1992). Planning in writing: The cognition of a constructive process. In S. P. Witte, N. Nakadate & R. D. Cherry (Eds.), A rhetoric of doing: Essays on written discourse in honor of James L. Kinneavy (pp. 181 - 243). Carbondale: Southern Il- linois University Press. Flower, L.S. & Hayes, J.R. (1983). A cognitive model of the writing process in adults. (ERIC Document Reproduction Service No. ED 240608) Flowerdew, J. & Miller, L. (1992). Student perceptions, problems and strategies in L2 lectures. RELC Journal, 23(2), 60-80. Foster, P., & Skehan, P. (1996). The influence of planning and task type on second language performance. Studies in Second Language Acquisition, 16, 299-323. Foulin, J.N. (1995). Pauses et débits: les indicateurs temporels de la production écrite [Pauses and fluency : temporal variables for written production]. L'Année psychologique, 95, 483-504. Francis, N. (2000). The shared conceptual system and language processing in bilingual children: Findings from literacy assessment in Spanish and Náhuatl. Applied Linguistics, 21, 170-204. Franken, M. (1997a). What makes this a good text? Be your own judge. In M. Lewis (Ed.), New ways in teaching adults (pp. 38-42). Alexandria, VA: Teachers of English to Speakers of Other Languages. Franken, M. (1997b, December). Second language learners talk for writing. Paper presented at New Zealand Association for Research in Education Conference, Auckland, New Zealand. Freire, P. (1973). Education for Critical Consciousness. New York: Seabury. Freire, P. (1985). The politics of education. Granby, MA: Bergin & Garvey. Freire, P., & Macedo, D. (1987). Literacy: Reading the word and the world. Granby, MA: Bergin & Garvey. Friedlander, A. (1990). Composing in English: Effects of a first language on writing in English as a second language. In B. Kroll (Ed.). Second language writing. Research insights for the classroom (pp. 109-125). Cambridge: Cambridge University Press.

REFERENCES 251

Fuchs, L., Fuchs, D., Karns, K., Hamlett, C., Katzroff, M., & Dutka, S. (1997). Effects of task-focused goals on low-achieving students with and without learning disabilities. American Educational Re- search Journal, 34, 513-543. Galambos, S. J. & Goldin-Meadow, S. (1990). The effects of two languages on levels of metalinguistic awareness. Cognition, 34, 1-56. Galbraith, D. & Rijlaarsdam, G. (1999). Effective strategies for the teaching and learning of writing. Learning and Instruction, 9, 93-108. Galbraith, D. (1996). Self-monitoring, discovery through writing and individual differences in drafting strategy. In G. Rijlaarsdam, H. van den Bergh, & M. Couzijn (Eds.), Theories, models and methodology in writing research (pp. 121-141). Amsterdam: Amsterdam University Press. Galbraith, D. (1999). Writing as a knowledge-constituting process. In G. Rijlaarsdam & E. Espéret (Se- ries eds.) & M. Torrance & D. Galbraith (vol. eds.), Studies in writing: volume 4. Knowing what to write (pp. 139 –159). Amsterdam: Amsterdam University Press. Gardner, H. (1983). Frames of mind. New York: BasicBooks. Garton, A., & Pratt, C. (1989). Learning to be literate. Oxford, UK: Blackwell. Gaskill, W. (1986). Revising in Spanish and English as a second language: A process oriented study of composition. Unpublished doctoral dissertation, University of California, Los Angeles. Gass, S. & Mackey, A. (2000). Stimulated recall methodology in second language research. Mahwah, NJ: Erlbaum. Gass, S. (1996). Second language acquisition and linguistic theory: the role of language transfer. In W. C. Ritchie & T.K. Bhatia (Eds.), Handbook of second language acquisition (pp. 317-348), San Diego: Academic Press. Gathercole, S.E., & Baddeley, A.D. (1993). Working memory and language. Hillsdale, NJ: Lawrence Erlbaum Associates. Genesee, F. (1994). Educating second language children : The whole child, the whole curriculum, the whole community. Cambridge: Cambridge University Press. Goh, C. C. M. (2000). A cognitive persective on language learners’ listening comprehension problems. System, 28, 55-75. Goodman, K., & Goodman, Y. (1987). What is whole in whole language? Portsmouth, NHö Heinemann. Gosden, H. (1996). Verbal reports of Japanese novices’ research writing practices in English. Journal of Second Language Writing, 5, 109-128. Goswami, U. (1991). Learning about spelling sequences: The role of onsets and rimes in analogies in reading Child Development, 62, 1110-1123. Gough, P. B., & Tunmer, W. E. (1986). Decoding, reading, and reading disability. Remedial and Special Education, 7(1), 6-10. Grabe, W. (2001). Notes towards a theory of second language writing. In T. Silva & P. K. Matsuda (Eds.), On second language Grabe, W., & Kaplan, R. B. (1996). Theory and practice of writing: An applied linguistic perspective. Harlow, UK: Longman. Graham, S., & Harris, K. R. (1997). Self-regulation and writing: Where do we go from here? Contempo- rary Educational Psychology, 22, 102-114. Gredler, M. E. (1997). Learning and instruction: Theory into practice (3rd ed.). Upper Saddle River, NJ: Prentice Hall. Greene, M. (1995). Notes on the search for coherence. In Beane, J. (Ed.), ASCD Yearbook (pp. 139-145). Alexandria, VA: ASCD. Griffiths, R., & Beretta, A. (1991). A controlled study of temporal variables in NS-NNS lectures. RELC Journal, 22(1), 1-19. Guaïtella I., & Santi S. (1994). Une approche ‘phonétique’ de la communication multimodale et de l’interaction [A phonetic approach to the multimodal communication and interaction]. Dialangue, 5, 64-71. Guilford, J.P. (1973). Fundamental Statistics in Psychology and Education. Tokyo: McGraw-Hill Koga- kusha, Ltd. Hadwin, A. F., Kirby, J. R., & Woodhouse, R. A. (1999). Individual differences in notetaking, summarization and learning from lectures. Alberta Journal of Educational Research, 45 (1), 1-17 Hakuta, K. (1986). Mirror of language: The debate on bilingualism. NY: Basic Books. Hall, C. (1987). Revision strategies in L1 and L2 writing tasks: A case study. Unpublished doctoral dissertation, University of Utah.

252

Hall, C. (1990). Managing the complexity of revision among languages. TESOL Quarterly, 24, 245-266. Hall, E. (1991) Variations in composing behaviours of academic ESL writers in test and non-test situations. TESL Canada Journal, 8, 9-33. Hamp-Lyons, L. (1990). Second language writing: Assessment issues. In B. Kroll (Ed.), Second language writing research: Insights for the classroom (pp. 69-87). Cambridge: Cambridge University Press. Haneda, M. (2000). Negotiating meaning in writing conferences: An investigation of a university Japa- nese-as-a-foreign language class. Unpublished Ph.D. thesis, Department of Curriculum, Teaching and Learning, University of Toronto. Hansen, C. (1994). Topic identification in lecture. In J. Flowerdew (Ed.), Academic listening: Research perspectives (pp. 131-145). Cambridge: Cambridge University press. Hartfiel, V.F., Hughey, J.B., Wormuth, D.R. & Jacobs, H.L. (1985). Learning ESL composition. Rowley, MA: Newbury House. Hashiuchi, T. (1995). Paragurafu Raitingu Nyuumon [Introduction to paragraph writing]. Tokyo: Ken- kyuusha. Hayes, J. R. (1996). A new framework for understanding cognition and affect in writing. In C. M. Levy & S. Ransdell (Eds.), The science of writing: Theories, methods, individual differences, and applications (pp. 1 - 27). Mahwah, NJ: Lawrence Erlbaum. Hayes, J. R., & Flower, L. S. (1980). Identifying the organization of writing processes. In L. Gregg & E. Steinberg (Eds.), Cognitive process in writing (pp. 3 - 30). Hillsdale, NJ: Lawrence Erbaum Associ- ates. Hayes, J. R., & Flower, L. S. (1983). Uncovering cognitive processes in writing: An introduction to protocol analysis. In P. Mosenthal, L. Tamar & S. A. Walmsey (Eds.). Research on writing (pp. 206 - 220). New York: Longman. Hayes, J. R., & Nash, J. G. (1996). On the nature of planning in writing. In C. M. Levy & S. Ransdell (Eds.), The science of writing: Theories, methods, individual differences, and applications (pp. 29 - 55). Mahwah, NJ: Lawrence Erbaum. Hayes-Roth, B., & Hayes-Roth, F. (1979). A cognitive Model of Planning. Cognitive Science, 3, 275-310. Hazenberg, S., & Hulstijn, J. H. (1996). Defining a minimal receptive second-language vocabulary for non-native university students: An empirical investigation. Applied Linguistics, 17 (2), 145-163. Henry, K. (1996). Early L2 writing development: A study of autobiographical esays by university-level students of Russian. The Modern Language Journal, 80, 309-326. Hermans, H. J. M. (1996). Voicing the self: From information processing to dialogical interchange. Psy- chological Bulletin, 119, 31-50. Higgings, L., Flower, L. S., & Petraglia, J. (1992). Planning text together. The role of critical reflection in student collaboration. Written Communication, 9 (1), 48-84. Hillocks, G. (1984). What works in teaching composition: A meta-analysis of experimental treatment studies. American Journal of Education, 93(1), 133-170. Hillocks, G. (1986). Research on written composition: New directions for teaching. Urbana, IL: ERIC Clearinghouse on Reading and Communication skills. Hirose, K., & Sasaki, M. (2000). Effects of teaching metaknowledge and journal writing on Japanese university students’ EFL writing. JALT Journal, 22, 94 -113. Hoffmann, A. (1998). An exploratory study of goal setting and the nature of articulated goals in second language writing development. New Zealand Studies in Applied Linguistics, 4, 33-48. Holquist, M. (1990). Dialogism: Bakhtin and his world. London: Rouletge. Hoover, W. A., & Tunmer, W. E. (1993). The components of reading. In G. B. Thompson, W. E. Tunmer, & T. Nicholson (Eds.), Reading acquisition processes (pp. 1-19). Philadelphia: Multilingual Matters. Hornberger, N. (1989). Continua of biliteracy. Review of Educational Research, 59, 271-296. Horton, S. R. (1982). Thinking through writing. Baltimore: Johns Hopkins Press. Isaac, S., & Michael, W. B. (1981). Handbook in research and evaluation. San Diego: EdITS. Ivanic, R. & Camps, D. (2001). I am how I sound: Voice as self-representation in L2 writing. Journal of second language writing, 10, 3-33. Iwahara, S. (1997). Kyouiku to Shinri notameno Suikeigaku [Inferential statistics for education and psychology]. Tokyo: Nihonbunkakagakusha. Izumi, S., Bigelow, M., Fujiwara, M., & Fearnow, S. (1999). Testing the output hypothesis. Effects of output on noticing and second language acquisition. Studies in Second Language Acquisition, 21, 421-452.

REFERENCES 253

Jacobs, H. L., Zinkgraf, S. A., Wormuth, D. R., Hartfiel, V. F., & Hughey, J. B. (1981). Testing EFL composition: A practical approach. Rowley, MA: Newbury House. Janssen, D., van Waes, L., & van den Bergh, H. (1996). Effects of thinking aloud on writing processes. In C. M. Levy & S. Randsdell (Eds.), The science of writing: Theories, methods, individual differences, and applications (pp. 233 -250). Mahwah, NJ: Lawrence Erbaum Associates. Jeffery, G. (1996). Writing in a second language: the implications of cognitive load on the writing process. Paper presented at the Second language and acquisition conference. Southampton, July. Johns, A. M. (1990). L1 composition theories: implications for developing theories of L2 composition. In B. Kroll (Ed.), Second language writing research: Insights for the classroom (pp. 24-36). Cam- bridge: Cambridge University Press. Johnson, J. S., & Newport, E. L. (1989). Critical period effects in second language learning: The influence of maturational state on the acquisition of English as a second language. Cognitive Psychology, 21, 60-99. Johnson, S., Linton, P., & Madigan, R. (1994). The role of internal standards in assessment of written discourse. Discourse Processes, 18, 231-245. Jones, C.S., & Tetroe, J. (1987). Composing in a second language. In A. Matsuhashi (Ed.), Writing in real time: Modelling production processes (pp. 34-57). Norwood, NJ: Ablex Jöreskog, K., & Sörbom, D. (1996a). LISREL 8: User’s Reference Guide. Chicago, IL: Scientific Soft- ware International, Inc. Jöreskog, K., & Sörbom, D. (1996b). PRELIS 2: User’s reference guide. A program for multivariate data screening and data summarization; a preprocessor for LISREL. Chicago, IL: Scientific Software In- ternational, Inc. Juel, C., Griffith, P. L., & Gough, P. B. (1986). Acquisition of literacy: A longitudinal study of children in first and second grade. Journal of Educational Psychology, 78(4), 243-255. Kasper, G. & Kellerman, E. (Eds.) (1997). Communication strategies: Psycholinguistic and sociolinguistic perspectives. London: Longman. Kasper, L. F. (1997). Assessing the metacognitive growth of ESL student writers. TESL-EJ Journal, 3, 1- 20. Kellogg, R. T. (1990). Effectiveness of prewriting strategies as a function of task demands. American Journal of Psychology, 103 (3), 327-342. Kellogg, R. T. (1996). A model of working memory in writing. In C. M. Levy & S. Ransdell (Eds.), The science of writing: Theories, methods, individual differences and applications (pp. 57-71). Mahwah, New Jersey: Lawrence Erlbaum Associates. Kellogg, R. T. (1999). Components of working memory in text production. In G. Rijlaarsdam & E. Espéret (Series Eds.) & M. Torrance & G. Jeffery (Volume eds.), Studies in writing: volume 3. The Cognitive Demands of Writing. Processing Capacity and Working Memory in Text Production (pp. 43-61). Amsterdam: Amsterdam University Press. Kellogg, R.T. (2001). Commentary on Part II: Processing and Development of Expertise in Writing. In G. Rijlaarsdam (Series Ed.) & D. Alamargot & L. Chanquoy (Volume Eds.), Studies in Writing, Vol. 9, Through the Models of Writing (pp. 219-228). Dordrecht: Kluwer Academic Publishers. Kern, R.G. (1994). The role of mental translation in second language reading. Studies in Second Lan- guage Acquisition, 16, 441-461. Keys, C.W. (1994) The development of scientific reasoning skills in conjunction with collaborative writing assignments. An interpretive study of 6-9th-grade students. Journal of Research in Science Teaching, 31 (9), 1003-1022. Keys, C.W. (1995) An interpretive study of students use of scientific reasoning during a collaborative report writing intervention in 9th grade general science. Science Education, 79 (4), 415-435. Khaldieh, S. (2000). Learning strategies and writing processes of proficient vs. less-proficient learners of Arabic. Foreign Language Annals, 33, 522-534. Kiewra, K.A., Benton, S.L., Kim, S., & Risch, N. & Christensen, M. (1995). Effects of note-taking format and study technique on recall and relational performance. Contemporary Educational Psychology, 20, 172-187. King, P. (1994). Visual and verbal messages in engineering lecture: notetaking by postgraduate L2 students. In J. Flowerdew (Ed.), Academic listening: Research perspectives (pp. 219-240). Cambridge: Cambridge University press. Kintsch, W. (1988). The role of knowledge in discourse comprehension: A construction-integration model. Psychological Review, 2, 163-182.

254

Knudsen, R. E. (1989). Effects of instructional strategies on children's informational writing. Journal of Educational Research, 83, 91-96 Knudsen, R. E. (1991). Effects of writing experience, grade, and reading level on children’s narrative writing. Journal of research and Development in education, 24 (3), 45-52. Knudsen, R. E. (1992). The development of written argumentation: An analysis and comparison of argumentative writing at four grade levels. Child Study Journal, 22, 167-184. Kobayashi, H., & Rinnert, C. (1992). Effects of first language on second language writing: Translation versus direct composition. Language Learning, 42, 183-215. Kobayashi, H., & Rinnert, C. (1996). Factors affecting composition evaluation in an EFL context: Cul- tural rhetorical pattern and readers’ background. Language Learning, 46, 397-437. Kobayashi, H., & Rinnert, C. (1998). Development of L2 writers’ discourse revision skills. Paper presented at AAAL Annual Conference. Seattle, Washington. March 14-17. Koda, K. (1993). Task-induced variability in FL composing: Language-specific perspectives. Foreign Language Annals, 26, 332-346. Kramsch, C. (2000). Social discursive constructions of self in L2 learning. In J. P. Lantolf (Ed.), So- ciocultural theory and second language learning (pp. 133-153). Oxford: Oxford University Press. Krapels, A. R. (1990). An overview of second language writing process research. In B. Kroll (Ed.), Sec- ond language writing: Research insights for the classroom (pp. 37 - 56). Cambridge: Cambridge University Press. Krings, H.P. (1994). What do we know about writing processes in L2? The state of the art. In K. Heinz Pogner (Ed.), More about writing. Odense Working Papers in Language and Communication, 6, 83- 114. Kuiken, F., & Vedder, I. (2000). The interface between grammar and lexicon. AILA '99. Tokyo, August 1- 6, Waseda University. Cd-rom. Kuiken, F., & Vedder, I. (in press). Focus on form and the role of interaction in promoting language learning. Proceedings of the Colloquium Instructed Second Language Learning, Brussels, August 24th-26th, 2000. Lahtinen, V., Lonka, K., & Lindbloom-Ylaenne, S. (1997). Spontaneous study strategies and the quality of knowledge construction. British Journal of Educational Psychology, 67 (1), 13-24 Lantolf, J. (2000). Second language learning as a mediated process. Language Teaching, 33, 79-96. Laufer, B., & Nation, P. (1995). Vocabulary size and use: Lexical richness in L2 written production. Applied Linguistics, 16(3), 307-322. Lay, N. (1982). Composing processes of adult ESL learners. TESOL Quarterly, 16, 406. Lay, N. (1988). The comforts of the first language in learning to write. Kaleidoscope, 4, 15-18. Lea, J., & Levy, C. M. (1999). Working memory as a resource in the writing process. In G. rijlaarsdam & E. espéret (Series eds.) & M. Torrance & G. Jeffery (volume eds.), Studies in writing: volume 3. The Cognitive Demands of Writing. Processing Capacity and Working Memory in Text Production (pp. 63-82). Amsterdam: Amsterdam University Press. Leki, I. (1992). Understanding ESL writers: A guide for teachers. London and Portsmouth: Heinemann. Leki, I. (1995). Coping strategies of ESL students in writing tasks across the curriculum. TESOL Quar- terly, 29, 235-260. Leki, I. (2000). L2 writing: a commentary. Learning and Instruction, 10, 101-105. Leki, I. (2001). Hearing voices: L2 students’ experiences in L2 writing courses. In T. Silva & P. Matsuda (Eds.), On second language writing (pp. 17-28). Mahwah, NJ: Erlbaum. Leki, I., & Carson, J.(1997). “Completely different worlds”: EAP and the writing experiences of ESL students in university courses. TESOL Quarterly, 31, 39-69. Leont’ev, A. (1972, trans. 1979). The problem of activity in psychology. In J. Wertsch (Trans. & Ed.), The concept of activity in Soviet psychology (pp. 37-71). Armonk, NY: Sharpe. Leont’ev, A. (1978). Activity, consciousness, and personality. Englewood Cliffs, NJ: Prentice Hall. Levelt, W.J.M. (1989). Speaking: From intention to articulation. Cambridge, Mass: MIT Press. Levy, C. M., & Marek, P. (1999). Testing Components of Kellogg’s Multicomponent Model of Working Memory in Writing: The Role of the Phonological Loop. In G. Rijlaarsdam & E. Espéret (Series eds.) & M. Torrance & G. Jeffery (volume eds.), Studies in writing: volume 3. The Cognitive Demands of Writing. Processing Capacity and Working Memory in Text Production (pp. 25-41). Amsterdam: Amsterdam University Press. Levy, C. M., & Ransdell, S. E. (1995). Is writing as difficult as it seems? Memory and Cognition, 23, 767-779.

REFERENCES 255

Levy, C. M., & Ransdell, S. E. (1996). Writing signatures. In C. M. Levy & S. E. Ransdell (Eds.), The science of writing: Theories, methods, individual differences and applications (pp. 149-161). Mah- wah, NJ: Erlbaum. Li, X. (1996). “Good writing” in cross-cultural context. Albany, NY: State University of New York Press. Littell, R. C., Freund, R. J., & Spector, P. C. (1991). SAS system for linear models (3rd ed.). Cary, NC: SAS Institute Inc. Locke, E. A. & Latham, G. P. (1990). A theory of goal-setting and task performance. Englewood Cliffs, NJ: Prentice-Hall. Locke, E. A., & Latham, G. P. (1994). Goal setting theory. In H. F. O'Neil & M. Drillings (Eds.), Motiva- tion: Theory and research (pp. 13-29). Hillsdale, NJ: Erlbaum. Long, M. H. & Robinson, P. (1998). Focus on form: theory, research and practice. In C. Dougerty & J. Williams (Eds.), Focus on form in classroom second language acquisition (pp. 15-63). Cambridge: Cambridge University Press. Long, M.H. (1983). Native speaker/non-native speaker conversation and the negotiation of comprehensible input. Applied Linguistics, 4, 126-141. Long, M.H. (1996). The role of the linguistic environment in second language acquisition. In: W. Ritchie & T. Batia (Eds.), Handbook of second language acquisition. New York: Academic Press. Magliano, J. P., Little, L. D., & Graesser, A. (1993). The impact of comprehension instruction on the calibration of comprehension. Reading Research and Instruction, 32(3), 49-63. Manchón, R., Roca de Larios, J., & Murphy, L. (2000). An approximation to the study of backtracking in L2 writing. Learning and Instruction, 10, 13-35. Manchón, R.M. (1997). Learners’ strategies in L2 composing. Communication and Cognition, 30, 91- 114. Manchón, R.M. (in press). Second language Learners’ composing strategies. A review of research. Inter- national Journal of English Studies Mangelsdorf, K. (1989). Parallels between speaking and writing in second language acquisition. In D. Johnson & D. Roen (Eds.), Richness in writing: Empowering ESL students (pp. 134-145). London: Longman. Matsuhashi, A. (1981). Pausing and planning: the tempo of written discourse production. Research in the Teaching of English, 15, 113-114. Mc Donough, S. H. (1999). Learner strategies. Language teaching, 32, 1-18. McCann, T. M. (1989). Student argumentative writing: Knowledge and ability at three grade levels. Re- search in the Teaching of English, 23, 62-76. McCutchen, D. (1996). A capacity theory of writing: Working memory in composition. Educational Psychology Review, 8(3), 299-325. McCutchen, D. (2000). Knowledge, processing, and working memory: Implications for a theory of writing. Educational Psychologist, 35(1), 13-23. McCutchen, D., Covill, A., Hoyne, S. H., & Mildes, K. (1994). Individual differences in writing: Implica- tions of translating fluency. Journal of Educational Psychology, 86(2), 256-266. McNeill D. (1992). Hand and mind, what gestures reveal about thought. Chicago: The University of Chicago Press. Meiran, N. (1996). Reconfiguring of processing mode prior to task performance. Journal of Experimental Psychology: Learning, Memory and Cognition, 22, 1423-1442. Miles, M., & Huberman, A. (1994). Qualitative data analysis: An expanded sourcebook (2nd ed.). Thou- sand Oaks, CA: Sage. Miller, K. S. (2000). Academic writers on-line: investigating pausing in the production of text. Language Teaching, 4, 123-148. Mohanty, A. K., & Babu, N. (1983). Bilingualism and metalinguistic ability among Kond tribals in Orissa, India. Journal of Social Psychology, 121, 15-22. Mohanty, A. K. (1992). Bilingualism and cognitive development of Kond tribal children: Studies on metalinguistic hypothesis. Pharmacopsychoecologia, 5 (1-2), 57-66. Moragne e Silva, M. (1988). A study of composing in a first and a second language. Texas Papers in Foreign Language Education, 1, 132 – 151. Morgan, C.H., Lilley, J.D. & Boreham, N.C. (1998). Learning from lectures: the effect of varying the detail in lecture handouts on note-taking and recall. Applied Cognitive Psychology, 2, 115-122.

256

Mosenthal, P. (1983). Defining classroom writing competence: A paradigmatic perspective. Review of Educational Research, 53, 217-251. Moss, P. A. (1994). Can there be validity without reliability? Educational Researcher, 23, 5-12. Murray, D. (1980). Writing as process: How writing finds its own meaning. In T. R. Donovan & B. W. McClelland (Eds.), Eight Approaches to teaching Writing (pp. 3-20). Urbana, IL: National Council of Teachers of English. Muspratt, S., Luke, A., Freebody, P. (1997). Constructing critical literacies: Teaching and learning textual practice. Creskill, NJ: Hampton Press. Myers, M. (1980). A procedure for writing assessment and holistic scoring. Technical Report. Urbana, IL: National Council of Teachers of English. Nagata, Y., & Yoshida, M. (1997). Toukeiteki Tajuuhikakuhou no Kiso. [Introduction to statistical multiple comparison]. Tokyo. Saientistosha. Nassaji, H., & Cumming, A. (2000). What’s in a ZPD? A case study of a young ESL student and teacher interacting through dialogue journals. Language Teaching Research, 4, 95-121. National Clearing House for Bilingual Education (1992). Performance and Portfolio Assessment for Language Minority Students. Washington D.C. Nayak, N., Hansen, N., Krueger, N., & McLaughlin, B. (1990). Language-learning strategies in monolingual and multilingual adults. Language Learning, 40(2), 221-244. Newton, J. (1995). Task-based interaction and incidental vocabulary learning: A case study. Second Lan- guage Research, 11, 159-177. Ng, E., & Bereiter, C. (1991). Three levels of goal orientation in learning. Journal of the Learning Sci- ences, 1 (3-4), 243-271. Norris, J. M., & Ortega, L. (2000). Effectiveness of L2 instruction: A research synthesis and quantitative meta-analysis. Language Learning, 50, 417-528. Nystrand, M. (1989). A social-interactive model of writing. Written Communication, 6, 66-85. Oakhill, J., & Davies, A.M. (1991). The effects of test expectancy on quality of note-taking and recall of text at different times of day. British Journal of Psychology, 82, 179-189. Oney, B., & Durgunoğlu, A.Y.(1997). Beginning to read in Turkish: A phonologically transparent orthography. Applied Psycholinguistics, 18, 1-15. Ovando, C. J., & Collier, V. P. (1998). Bilingual and ESL classrooms: Teaching in multicultural contexts. New York: McGraw-Hill. Oxford, R. (1990). Language learning strategies: What every teacher should know. Rowley, MA: New- bury House. Parks, S., & Maguire, M. (1999). Coping with on-the-job writing in ESL: A constructivist-semiotic perspective. Language Learning, 49, 143-175. Penningroth, S. L., & Rosenberg, S. (1995). Effects of a high information processing load on the writing process and the story written. Applied Psycholinguistics, 16, 189-210. Pennington, M. C., & So, S. (1993). Comparing writing process and product across two languages: a study of 6 Singaporean university students writers. Journal of Second Language Writing, 2, 41-63. Perdue, C., & Gaonac'h, D. (2000). Acquisition des langues secondes [Second languages acquisition]. In M. Kail & M. Fayol (Eds.), L'acquisition du langage, le langage en développement au delà de trois ans [Language acquisition, language development after three years old] (vol. 2, pp. 215-246). Paris: PUF. Perez, B. (1998). Writing across writing systems. In B. Perez (Ed.), Sociocultural contexts of language and literacy (pp. 58-65). Mahweh, NJ: Erlbaum. Perl. S. (1979). The composing processes of unskilled college writers. Research in the teaching of Eng- lish, 13, 317-336. Peyton, J. (1993). The development of beginning writers: Six student profiles. In J. Peyton & J. Staton (Eds.), Dialogue journals in the multilingual classroom: Building language fluency and writing skills through written interaction (pp. 47-99). Norwood, NJ: Ablex. Philips, D. (1985). A month's writing in four classrooms. Writing Research Report No. 1, Wellington: NZCER. Pica, T. (1988). Interlanguage adjustments as an outcome of NS-NNS negotiated interaction. Language Learning, 38, 45-73. Pica, T., Holliday, L., Lewis, N., & Morgenthaler, L. (1989). Comprehensible output as an outcome of linguistic demands on the learner. Studies in Second Language Acquisition, 13, 343-376.

REFERENCES 257

Pierce, L., & O'Malley, J.M. (1992). Performance and Portfolio Assessment for Language Minority Stu- dents. Washington D.C: National Clearinghouse for Bilingual Education. Piolat, A. (2001). La prise de notes [Notetaking]. Paris: Presses Universitaires de France. Piolat, A., & Pélissier, A. (1998). Etude de la rédaction de textes: contraintes théoriques et méthodes de recherche [Cognitive approach to written production : theoretical constraints and research methods]. In A. Piolat & A. Pélissier (Eds.)., La rédaction de textes, Approche cognitive [ Written production, Cognitive approach] (pp.225-270). Neuchâtel: Delachaux & Niestlé. Pittard, V. (1999). Knowledge, ideas and the social situation of writing. In G. Rijlaarsdam & E. Espéret (Series Eds.) & M. Torrance & D. Galbraith (Volume Eds.), Studeis in Writing. Volume 4: Knowing what to write: conceptual processes in text production (pp. 161-172). Amsterdam: Amsterdam Uni- versity Press Porte, G. (1996). When writing fails: How academic context and past learning experiences shape revision. System, 24, 107-116. Porte, G. (1997). The etiology of poor second language writing: The influence of perceived teacher preferences on second language revision strategies. Journal of Second Language Writing, 6, 61-78. Pozo, J. I. (1989). Teorías cognitivas del aprendizaje. Madrid: Morata. Prawat, R. (1998). Current self-regulation views of learning and motivation viewed through a Deweyan lens: The problems with dualism. American Educational Research Journal, 35, 199-224. Prior, P. (1998). Writing/disciplinarity: A sociohistoric account of literate activity in the academy. Mah- wah, NJ: Erlbaum. Prior, P. (2001). Voices in text, mind, and society: Sociohistoric accounts of discourse acquisition and use. Journal of Second Language Writing, 10, 55-81. Qi, D. S., (1998). An inquiry into language-switching in second language composing processes. The Canadian Modern Language Review, 54, 143-435. Qualitative Data Solutions and Research. (2000). NVivo. Thousand Oaks, CA: Sage. Quintero, E., & Rummel, M. (1998). American voices: Webs of diversity. Columbus, OH: Merrill Educa- tion. Raimes, A. (1985). What unskilled writers do as they write: A classroom study of composing. TESOL Quarterly, 19, 229-258. Raimes, A. (1987). Language proficiency, writing ability, and composing strategies: A study of ESL college student writers. Language Learning, 37, 439 – 469. Raimes, A. (1987). Language proficiency, writing ability, and composing strategies: A ge student writers. Language Learning, 37, 439-468 Raimes, A. (1998). Teaching writing. Annual Review of Applied Linguistics, 18, 142-167. Ramanathan, V., & Atkinson, D. (1999). Ethnographic approaches and methods in L2 writing research: A critical guide and review. Applied Linguistics, 20, 44 - 70. Ramirez, J. D., Yuen, S. D., Ramey, D. R., & Pasta, D.J. (1991). Final report: Longitudinal study of structured English immersion strategy, early-exit and late-exit transitional bilingual education programs for language-minority children (Vols. I and II). San Mateo, CA: Aguirre International. Ramsay, R.M.G. (1980). Language-learning approach styles of adult Multilinguals and successful language learners. Annals of the New York Academy of Sciences, 75, 73-96. Ransdell, S., & Levy, C. M. (1999). Writing, reading, and speaking memory spans and the importance of resource flexibility. In G. Rijlaarsdam & E. Espéret (Series eds.) & M. Torrance & G. Jeffery (Eds.), Studies in writing: volume 3. The Cognitive Demands of Writing. Processing Capacity and Working Memory in Text Production (pp. 99-113). Amsterdam: Amsterdam University Press. Ransdell, S. E. & Levy, M. (1998). Writing in a second language and the limits of working memory. Paper presented at the 1998 European Writing conference of the EARLI Special Interest Group on writing. Poitiers. France. Ransdell, S. E., Arecco, M. R. & Levy, C. M. (2001). Bilingual long-term working memory: The effects of working memory loads on writing quality and fluency. Applied Psycholinguistics, 22, 117-132. Raykov, T., & Marcoulides, G. A. (2000). A first course in structural equation modeling. Mahwah, NJ & London: Lawrence Erlbaum Associates. Reese, W. H. (1997). Counterbalancing and other uses of repeated-measures in latin square designs: Analyses and interpretations. Journal of Experimental Child Psychology, 64, 137-158. Reichelt, M. (1999). Toward a more comprehensive view of L2 writing: Foreign language writing in the U.S. Journal of Second Language Writing, 9, 181-204.

258

Reid, J. (1990). Responding to different topic types: A quantitative analysis from a contrastive rhetoric perspective. In B. Kroll (Ed.), Second language writing: Research insights for the classroom (pp. 191-210). New York: Cambridge University Press. Reyes, M. (1992). Questioning venerable assumptions: Literacy instruction for linguistically different students. Harvard Education Review, 62(4), 427-444. Riazi, A. (1997). Acquiring disciplinary literacy: A social-cognitive analysis of text production and learning among Iranian graduate students of education. Journal of Second Language Writing, 6, 105-137. Ricciardelli, L. A. (1992). Creativity and bilingualism. Journal of Creative Behavior, 26(4), 242-254. Rickards, J.P., Fajen, B.R., Sullivan, J.F., & Gillespie (1997). Signaling, Notetaking, and field independence-dependence in text comprehension and recall. Journal of Educational Psychology, 89, 508-517. Rijlaarsdam, G., & van den Bergh, H. (1996) The dynamics of composing- An agenda for research into an interactive model of writing: many questions, some answers. In C.M. Levy and S. Ransdell (Eds). The science of writing. Theories, methods, individual differences, and applications (pp. 107-125). Mahwah, NJ: Erlbaum. Rinnert, C. & Kobayashi, H. (2001). Differing perceptions of EFL writing among readers i n Japan. The Modern Language Journal, 85, 189-209. Robinson, P. (1995). Review article. Attention, memory and the “noticing” hypothesis. Language Learn- ing, 45, 283-331. Robinson, P. (Ed.) (1996). Task complexity and second language syllabus design: Data-based studies and speculations. Special Issues of University of Queensland Working Papers in Language and Linguis- tics. Robinson, P. (in press). Task complexity, cognition and second language syllabus design: a triadic framework for examining task influences on SLA. In P. Robinson (Ed.), Cognition in second language instruction. Cambridge: Cambridge University Press. Roca de Larios, J. , Marín, J. & Murphy, L. (2001). A temporal analysis of formulation processes in L1 and L2 writing. Language Learning, 51, 497-538. Roca de Larios, J., Murphy, L., & Manchon, R. (1999). The use of restructuring strategies in EFL writing: A study of Spanish learners of English as a foreign language. Journal of Second Language Writing, 8, 13 – 44. Rueda, R., & Moll, L. C. (1994). A sociocultural perspective on motivation. In H. F. O'Neil & M. Drill- ings (Eds.), Motivation: Theory and research (pp. 117-137). Hillsdale, NJ: Erlbaum. Ruhl, K. L., & Suritsky, S. (1995). The pause procedure and/or an outline: Effect on immediate free recall and lecture notes taken by college students with learning disabilities. Learning DisabilityQuarterly, 18 (1), 2-11. Rumelhart, D. E. (1980). Schemata: The building blocks of cognition. In R. Spiro, B. Bruce & W. Brewer (Eds.), Theoretical issues in reading comprehension (pp. 33-58). Hillsdale, NJ: Lawrence Erlbaum Associates. Rummel, M., & Quintero, E. (1997). TeachersÆ reading/teachersÆ lives. Albany, NY: State University of New York Press. Russell, D. (1995). Activity theory and its implications for writing instruction. In J. Petraglia (Ed.), Reconceiving writing, rethinking writing instruction (pp. 51-78). Mahwah, NJ: Erlbaum. Sanders, M., & Meijers, G. (1995). English as L3 in the Elementary School. ITL: Review of Applied Lin- guistics, 107-108, 59-78. Sasaki, M. (1993). Relationships among second language proficiency, foreign language aptitude, and intelligence: A protocol analysis. Language Learning, 43(4), 469-505. Sasaki, M. (2000). Toward an empirical model of EFL writing processes: An exploratory study. Journal of Second Language Writing, 9, 259 -291. Sasaki, M., & Hirose, K. (1996). Explanatory variables for EFL students’expository writing. Language Learning, 46(1), 137-174. Sato, C. (1986). Conversation and interlanguage development: Rethinking the connection. In R. Day (Ed.), Talking to learn (pp. 23-45). Rowley, MA: Newbury House. Scardamalia, M., & Bereiter, C. (1991). Literate expertise. In K. A. Ericsson & J. Smith (Eds.), Toward a general theory of expertise: Prospects and limits. (pp. 172-194). Cambridge: Cambridge University Press. Scardamalia, M., & Paris, P. (1985). The function of explicit discourse knowledge in the development of text representations and composing strategies. Cognition and Instruction, 2, 1-39.

REFERENCES 259

Scardamalia, M., Bereiter, C., & Goelman, H. (1982). The role of production factors in writing ability. In M. Nystrand (Ed.), What writers know: The language, process, and structure of written discourse (pp. 173-210). NY: Academic Press. Schereiner, C. (1997). A portrait of the student as a young writer: Reevaluating Emig and the process movement. College Composition and Communication, 48, 86-104. Schilperoord, J. (1996). The distribution of pause time in written text production. In G. Rijlaarsdam, H. Van den Bergh & M. Couzijn (Eds.), Theories, models and methodology in writing research (pp. 542-556), Amsterdam: Amsterdam University Press. Schmidt, R. (1990). The role of consciousness in second language learning. Applied Linguistics, 11, 129- 158 Schmidt, R. (1994). Deconstructing consciousness: In search of useful definitions for applied linguistics. AILA Review, 11, 11-26. Schmidt, R. (1995). Attention and awareness in foreign language learning and teaching. Honolulu: Uni- versity of Hawai'i Press. Schoonen, R., & De Glopper, K. (1999). The assessment of writing proficiency: Problems and possibilities. In W. Decoo, M. Simons & A. Stepanyan (eds.), L’enseignement curriculaire des langues étrangères (pp.191-212). Echanges INTAS vol. 3. Lier: Van In. Schoonen, R., & De Glopper, K. 1996). Writing performance and knowledge about writing. In G. Rij- laarsdam, H. van den Bergh & M. Couzijn (Eds.), Theories, models and methodology in writing research (pp. 87-107). Amsterdam: Amsterdam University Press. Schunk, D. (1996). Goal and self-evaluative influences during children’s cognitive skill learning. Ameri- can Educational Research Journal, 33, 359-382. Segalowitz, N.S., & Segalowitz, S.J. (1993). Skilled performance, practice, and the differenciation of speed-up from automatization effects: evidence from second language word recognition. Applied Psycholinguistics, 14, 369-385. Sengupta, S. (2000). An investigation into the effects of revision strategy instruction on L2 secondary school learners. System, 28, 97-113. Shaw, P., & Ting Kun Liu, E. (1998). What develops in the development of second language writing? Applied Linguistics, 19, 225-254. Shi, L., & Cumming, A. (1995). Teachers' conceptions of second language writing instruction: Five case studies. Journal of Second Language Writing, 4, 87-111. Shor, I. (1987). Freire for the Classroom: A Sourcebook for Liberatory Teaching. Portsmouth, NH: Boynton/Cook. Sijtstra, J. (Ed.) (1997). Balans van het taalonderwijs aan het einde van de basisschool 2. Uitkomsten van de tweede taalpeiling einde basisonderwijs [Account of the language education at the end of primary school 2. Results of the second language assessment at the end of primary school]. Arnhem: Cito. Silva, T. (1989). A critical review of ESL composing process research. Paper presented at the 23rd An- nual Meeting of the Teachers of English to Speakers of Other Languages. San Antonio, Texas. March 7-11. Silva, T. (1990). Second language composition instruction: Developments, issues, and directions in ESL. In B. Kroll (Ed.), Second language writing: Research insights for the classroom (pp. 11 - 23). Cam- bridge: Cambridge University Press. Silva, T. (1992). L1 vs L2 writing: ESL graduate students’ perceptions. TESL Canada Journal, 10(1), 27- 46. Silva, T. (1993). Toward an understanding of the distinct nature of L2 writing: The ESL research and its implications. TESOL Quarterly, 27, 657-677. Silva, T., & Matsuda, P. (2001). On second language writing. Mahwah, NJ: Erlbaum. Skehan, P. (1998). A cognitive approach to language learning. Oxford: Oxford University Press. Skibniewski, L. (1988). The writing processes of advanced foreign language learners in their native and foreign languages: Evidence from thinking-aloud and behavior protocols. Studia Anglica Posnanien- sia, 21, 177 – 186. Slotte, V., & Lonka, K. (1999). Review and process effects of spontaneous note-taking on text comprehension. Contemporary Educational Psychology, 24 (1), 1-20 Slotte, V., & Lonka, K. (2001). Note taking and essay writing. In G. Rijlaarsdam (Series ed.), P. Tynjälä, L. Mason & K. Lonka (Volume eds.), Studies in Writing, Vol. 7, Writing as a Learning tool: Inte- grating Theory and Practice (pp. 131-143). Dordrecht: Kluwer Academic Publishers.

260

Smagorinsky, P. (1994). Think-aloud protocol analysis beyond the black box. In P. Smagorinsky (Ed.), Speaking about writing: Reflections on research methodology (pp. 3 - 19). Thousand Oaks, CA: Sage. Smith, V. (1994). Thinking in a foreign language: An investigation into essay writing and translation by L2 learners. Tübingen: Verlag. Snow, C. E., & Hoefnagel-Hohle, M. (1978). Critical period for language acquisition: Evidence from second language learning. Child Development, 49, 1263-1279. Sommers, N. I. (1980). Revision strategies of student writers and experienced adult writers. College Composition and Communication, 31, 378-388. Spack, R. (1997). The acquisition of academic literacy in a second language: A longitudinal case study. Written Communication, 14, 3-62. SPSS Incorporated. (1993). SPSS Base System, Release 6.x. Chicago, IL: SPSS. SPSS Incorporated. (1994). SPSS 6.1 Base System User’s Guide. Macintosh Version. Chicago, IL: SPSS. Stanovich, K.E. (1991). Word recognition: Changing perspectives. In R. Barr, M.L. Kamil, P.B. Mosen- thal & P.D. Pearson (Eds.), Handbook of Reading Research, Vol II (pp. 418-452). New York/London: Longman. Stotsky, S. (1995). The uses and limitations of personal or personalised writing in writing theory, research and instruction. Reading Research Quarterly, 30, 758-776. Stratman, J., & Hamp-Lyons, L. (1994). Reactivity in concurrent think-aloud protocols: Issues for research. In P. Smagorinsky (Ed.), Speaking about writing: Reflections on research methodology (pp. 89 - 112). Thousand Oaks, CA: Sage. Swain, M. (1985). Communicative competence: some roles of comprehensible input and comprehensible output in its development. In S. Gass & C. Madden (Eds.), Input in second language acquisition (pp. 235-253). Rowley MA: Newbury House. Swain, M. (1995). Three functions of output in second language learning. In G. Cook & B. Seidlhofer (Eds.), Principle and practice in applied linguistics: Studies in honour of H.G. Widdowson (pp. 125- 144). Oxford: Oxford University Press. Swain, M. (1998). Focus on form through conscious reflection. In C. Doughty & J. Williams (Eds.), Fo- cus on form in classroom second language acquisition (pp. 15-63). Cambridge: Cambridge Univer- sity Press. Swain, M., & Lapkin, S. (1995). The development of fluency in advanced learners of French. Applied Linguistics, 17, 84-119. Swain, M., & Lapkin, S. (2000a). Focus on form through collaborative dialogue: exploring task effects. In M. Bygate, P. Skehan & M. Swain (Eds.), Researching pedagogic tasks: second language learning, teaching and testing. (pp. 99-110). Harlow, Essex: Longman. Swain, M., & Lapkin, S. (2000b). Task-based second language learning: the uses of the first language. Language Teaching Research 4(3), 251-274. Swarts, H., Flower, L. S., & Hayes, J. R. (1984). Designing protocol studies of the writing process: An introduction. In R. Beach & L. S. Bridwell (Eds.), New directions in composition research (pp. 53 – 71). New York: The Guilford Press. Sweigart, W. (1991). Classroom talk, knowledge development, and writing. Research in the Teaching of English, 25, 469-496. Sweller, J. (1994). Cognitive load theory, learning difficulty, and instructional design. Learning and Instruction, 4, 295-312. Tabachnick, B. G., & Fidell, L. S. (1996). Using multivariate statistics (3rd ed.). New York: Harper- Collins. Takeuchi, K. (Ed.). (1989). Toukei Gaku Jiten [Dictionary of Statistics]. Tokyo: Toyokeizaishinpousha. Tannenbaum, R. J., Rosenfeld, M., & Breyer, F.J. (1996). Linking TOEIC scores to self-assessments of English-language abilities: A study of score interpretation. Unpublished manuscript submitted to Test of English for International Communication. Educational Testing Service. Tarone, E., Downing, B., Cohen, A., Gillette, S., Murie, R., & Dailey, B. (1993). The writing of SE Asian American students in secondary school and university. Journal of Second Language Writing, 2, 149- 172. Thomas, J. (1988). The role played by metalinguistics awareness in second and third language learning. Journal of Multilingual and Multicultural Development, 9, 235-246.. Thomas, M. (1994). Assessment of L2 proficiency in second language acquisition research. Language Learning, 44, 307-336.

REFERENCES 261

Thorson, H. (2000). Using the computer to compare foreign and native language writing processes: A statistical and case study approach. Modern Language Journal, 84, 155-170. Tierney, R., Carter, M., & Desai, L. (1991). Portfolio Assessment in the Reading-Writing Classroom. Norwood, MA: Christopher Gordon. Timbur, G. (1994). Taking the social turn: Teaching writing post-process. College Composition and Communication, 45, 108-118. Ting, R. Y. (1996). Looping forward: Drafting in my own language. English Language Teaching Journal, 50, 135-142. Torrance, M., & Jeffery, G. (1999). Writing processes and cognitive demands. In G. Rijlaarsdam & E. espéret (Series Eds.) & M. Torrance & G. Jeffery (Volume eds.), Studies in writing: volume 3. The Cognitive Demands of Writing. Processing Capacity and Working Memory in Text Production (pp. 1-11). Amsterdam: Amsterdam University Press. Torrance, M., Thomas, G., & Robinson, E. J. (1999). Individual differences in the writing behaviour of undergraduate students. British Journal of Educational Psychology, 69, 189-199). Torrance, M., Thomas, G., & Robinson, E. J. (2000). Individual differences in undergraduate essay- writing strategies: A logitudinal study. Higher Education, 39, 181-200. Torres-Guzman, M. E. (1993). Critical pedagogy and Bilingual/Bicultural Education Special Interest Group update. NABE News, 17(3), 14-15. Toulmin, S. E., Rieke, R., & Janki, A. (1984). An introduction to reasoning (2nd ed.). NY: Macmillan. Treiman, R. (1993). Beginning to spell. New York: Oxford University Press. Trembley, P. F., & Gardner, R. C. (1995). Expanding the motivation construct in language learning. Modern Language Journal, 79, 505-520. Tynjälä, P., Mason, L. & Lonka, K. (2001). Writing as a Learning Tool: Integrating theory and practice. Dordrecht: Kluwer Academic Publishers. Uchida, N. (1986). Sakubun no Shinrigaku: Sakubun no Kyoujuriron eno Shisa [Psychology of writing: Implications for pedagogical theories of writing]. Kyoikushinrigaku Nenpoo, 25, 162 - 177. Uchida, N. (1989). Kodomo no Suikou Hooryaku no Hattatsu: Sakubun ni Okeru Jikonai Taiwa no Katei [Development of strategies used by children in rewriting compositions: Dialectical processes in writing]. Ochanomizu University Studies in Arts and Culture, 42, 75 - 104. Uchida, N. (1990). Gengo no Sanshutsu [Production of language]. In N. Uchida (Ed.), Gengo Kinou no Hattatsu [Development of language ability] (pp. 183 - 221). Tokyo: Kaneko Shobou. Ullman, C. (1997). Social Identity and the Adult ESL Classroom. Eric Digest. National Clearinghouse on Literacy Education. October 1997, EDO-LE-98-01. Uzawa, K. (1996). Second language learners’ processes of L1 writing, L2 writing, and translation from L1 into L2. Journal of Second Language Writing, 5(3), 271-294. Uzawa, K., & Cumming, A. (1989). Writing strategies in Japanese as a foreign language: Lowering or keeping up the standards. The Canadian Modern Language Review, 46, 178-194. Valdés, G., Haro, P. & Echevarriarza, M. (1992). The development of writing abilities in a foreign language: Contributions toward a general theory of L2 writing. Modern Language Journal, 76, 333-352 Vallen, T., & Stijnen, S. (1991). Ethnic minorities and ethnic minority languages in Dutch primary schools: Problems and challenges. Language and Education, 5(2), 113-123. van den Bergh, H., De Glopper, K. & Schoonen, R. (1988). Directe metingen van schrijfvaardigheid: validiteit en taakeffecten. In F.H. van Eemeren & R. Grootendorst, Taalbeheersing (pp. 370-378). Dordrecht: Foris Publications. [Direct measures of writing ability: validity and task effects.] van Dijk, T.A., & Kintsch, W. (1983). Strategies of discourse comprehension. New York: Academic Press. Van Meter, P., Yokoi, L., & Pressley, M. (1994). College students' theory of note-taking derived from their perceptions of note-taking. Journal of Educational Psychology, 86, 323-338. VanPatten, B. (1990). Attending to content and form in the input: An experiment in consciousness. Stud- ies in Second Language Acquisition, 12, 287-301. VanPatten, B. (1996). Input processing and grammar instruction. New York: Ablex. VanPatten, B., & Cadierno, T. (1993). Explicit instruction and input processing. Studies in Second Lan- guage Acquisition, 15(2), 225-243. Verhallen, M., & Schoonen, R. (1993). Lexical knowledge of monolingual and bilingual children. Ap- plied Linguistics, 14(4), 344-363. Verhoeven, L. (1991). Acquisition of biliteracy. AILA Review, 8,61-74

262

Victori, M. (1995). EFL writing knowledge and strategies: An integrative study. Unpublished doctoral dissertation. Departamento de Filología Inglesa. Universitat Autónoma de Barcelona. Victori, M. (1999). An analysis of writing knowledge in composing: a case study of two effective and two less effective writers. System, 27, 537-555. Victori, M., & Lockhart, W. (1995). Enhancing metacognition in self-directed language learning. System, 23(2), 223-234. Vye, N., Goldman, S., Voss, J., Hemlo, J., Williams, S. & Cognition and Technology Group at Vander- bilt. (1997). Complex mathematical problem solving by individuals and dyads. Cognition and In- struction, 15, 435-484. Vygotsky, L.S. (1962). Thought and Language. Cambridge: M.I.T. Press Wajnryb, R. (1990). Grammar dictation. Oxford: Oxford University Press. Warren, E. (1996). The significance of pauses in written discourse: A comparison of native speaker and learner writing. In A. Archibald & G. Jeffery (Eds.), Second language acquisition and writing: A multidisciplinary approach (pp. 152-168). Southampton: University of Southampton. Waters, A. (1996). A review of research into the needs in English for academic purposes of relevance to the North American higher education context. TOEFL Monograph Series 6. Princeton, NJ: Educa- tional Testing Service. Way, D. P., Joiner, E. G., & Seaman, M. A. (2000). Writing in the secondary foreign language classroom: The effects of prompts and tasks on novice learners of french. Modern Language Journal, 84, 171- 184. Weissberg, B. (2000). Developmental relationships in the acquisition of English syntax: writing vs. speech. Learning and Instruction, 10, 37-53. Wells, G. (1999). Dialogic inquiry: Towards a sociocultural practice and theory of education. Cam- bridge: Cambridge University Press. Wenden, A., & Rubin, J. (1987). Learner strategies in language learning. Englewood Cliffs, NJ: Prentice Hall. Wertsch, J. (Ed. & Trans.) (1979). The concept of activity in Soviet psychology. New York: Sharpe. Whalen, K. (1988). Pilot study on the nature of difficulties in written expression in a second language: process or product? Bulletin of the CAAL, 10, 51-59. Whalen, K., & Ménard, N. (1995). L1 and L2 writer's strategic and linguistic knowledge: A model of multiple-level discourse processing. Language Learning, 45 (3), 381-418. Widdowson, H.G. (1983). New starts and different kinds of failure. In A.Freedman, I. Pringle, and J. Yalden (Eds.), Learning to write: First language/Second language (pp. 34-47). London: Longman. Winsor, D. (1999). Genre and activity systems. Written Communication, 16, 200-224. Witte, S. P. (1992). Context, text, intertext: toward a constructivist semiotic of language. Written communication, 9, 237-308. Witte, S.P., Nakadate, N., & Cherry, R.D. (1992). The rhetoric of doing. Carbondale: Southern Illinois University Press. Yau, M. S. S. (1991). The role of language factors in second language writing. In L. Malavé and G. Du- quette (Eds.), Language, culture and cognition: a collection of studies in first and second language acquisition (pp. 266-283). Clevedon, Avon: Multilingual matters. Zamel, V. (1982). Writing: The process of discovering meaning. TESOL Quarterly, 16, 195 – 209. Zamel, V. (1983). The composing process of advanced ESL students: Six case studies. TESOL Quarterly, 17, 165-187. Zamel, V. (1995). Strangers in academia: The experiences of faculty and ESL students across the curriculum. College Composition and Communication, 46, 506-521. Zimmerman, R. (2000). L2 Writing: subprocesses, a model of formulating and empirical findings. Learn- ing and Instruction, 10, 73-99. Zinchenko, V., & Gordon, V. (1976, trans. 1979). Methodological problems in the psychological analysis of activity. In J. Wertsch (Trans. & Ed.), The concept of activity in Soviet psychology (pp. 72-133). Armonk, NY: Sharpe.

AUTHOR INDEX

Abbott, R. D., 82, 103, 247 Bollen, K. A., 54, 248 Acock, A. C., 111, 247 Boreham, N. C., 146, 257 Adams, M. J., 82, 85, 247 Bosher, S., 13, 14, 15, 17, 18, 19, 34, 39, 46, Akyel, A., 14, 15, 16, 18, 19, 20, 30, 46, 47, 47, 50, 51, 52, 53, 55, 248 247 Breetvelt, I., 66, 248 Alderson, J. C., 105, 247 Breyer, F. J., 77, 262 Aljaafreh, A., 194, 247 Brindley, G., 191, 248 Allison, D., 148, 247 Britton, J., 37, 248 Ames, C., 192, 247 Broekkamp, H., 4, 248 Anderson, J. R., 212, 247 Brooks, E., 15, 16, 17, 18, 19, 20, 23, 24, 25, Anzai, Y., 53, 57, 59, 60, 66, 68, 73, 247 26, 27, 34, 36, 39, 46, 47, 248, 249 Appel, R., 102, 247 Brooks-Carson, A., 15, 16, 17, 18, 20, 34, 46, Applebee, A. N., 225, 247 47, 249 Arecco, M. R., i, 7, 9, 13, 47, 123, 131, 134, Brown, G., 212, 248 135, 142, 149, 163, 259, 277 Brown, J., 46, 191, 248 Arndt, V., 4, 13, 17, 18, 19, 23, 31, 34, 41, 45, Brown, J. I., 127, 138, 248 47, 51, 52, 247 Bruce, B., 87, 249, 260 Aronsson, K., 123, 124 Bruffee, K., 228, 249 Atkinson, D., 78, 259 Bryson, M., 214, 249 August, D., 83, 247, 256 Buck, G., 148, 249 Austin, J. T., 191, 203, 247 Burgess, T., 248 Burtis, P. J., 134, 249 Babu, N., 124, 125, 257 Bachman, L., 72, 74, 247 Cadierno, T., 170, 263 Baddeley, A. D., 147, 247, 253 Camps, D., 47, 254 Baldwin, L. E., 86, 249 Candlin, M., 226, 249 Bandura, A., 191, 247 Carey, L., 55, 58, 252 Barbier, M. L., i, 3, 4, 6, 9, 10, 14, 47, 149, Carrell, P., 15, 47, 249 154, 155, 156, 163, 247, 277 Carretero, M., 228, 249 Beaufort, A., 203, 247 Carroll, B.J., 14, 249 Belinchón, M., 38, 247 Carson, J., 14, 15, 16, 18, 19, 20, 34, 46, 47, Bell, J. S., 16, 17, 19, 35, 46, 47, 148, 247 191, 194, 249, 256 Benson, J., 111, 114, 252 Carter, M., 46, 58, 249, 263 Benton, S. L., 104, 146, 248, 255 Carver, C., 191, 249 Ben-Zeev, S., 125 Casanave, C., 191, 249 Bereiter, C., 3, 21, 22, 24, 30, 34, 37, 44, 56, Cashion, M., 237, 249 61, 64, 68, 82, 103, 106, 134, 135, 192, 202, Castley, A., 110, 249 226, 227, 248, 249, 258, 260, 261 Cattell, A. K., 127 Beretta, A., 149, 253 Cattell, R. B., 127 Berkenkotter, C., 203, 248 Chall, J. S., 86, 87, 249 Berman, R., 14, 15, 16, 18, 20, 27, 43, 47, 69, Chaudron, C., 148, 249 148, 248 Chenoweth, N. A., 14, 15, 16, 17, 19, 20, 27, Berninger, V. W., 82, 103, 247 33, 47, 249 Berwick, R., 250 Cherry, R. D., 42, 73, 251, 252, 264 Bialystok, E., 123, 124, 125, 126, 250 Chesnet, D., 149, 153, 249 Bigelow, M., 171, 254 Chin, E., 47, 249 Biklen, S. K., 237, 248 Christensen, M., 146, 216, 255 Birdsong, D., 252 Clachar, A., 204, 249 Block, D., 13, 248 Clay, M. M., 81, 249 Boch, F., 146, 147, 248 Clerehan, R., 148, 149, 249 Bogdan, R. C., 237, 248 Cliff, A. F., 204, 249 Coe, R., 216, 249

264

Cognition and Technology Group at Eigler, G., 227, 251 Vanderbilt., 192, 264 Elbow, P., 136, 251 Cohen, A., 13, 47, 262 Ellis, R., 15, 148, 170, 204, 213, 225, 228, 251 Cohen, A.D., 15, 16, 17, 18, 20, 34, 46, 47, 68, Engber, C. A., 103, 251 249 Engeström, Y., 192, 251 Coirier, P., 226, 250 Ericsson, K. A., 3, 52, 59, 135, 251, 260 Cole, M., 192, 251 Espéret, E., 149, 153, 247, 249, 252, 253, 255, Collier, V. P., 233, 237, 258 256, 259 Collins, A., 46, 191, 248 Comber, B., 234, 236, 250 Fagan, W. T., 13, 16, 17, 18, 19, 20, 31, 33, 34, Connor, U., 194, 250 47, 251 Conrad, S., 194, 250 Faigley, L,, 12, 73, 251 Cook, G., 262 Fajen, B. R., 146, 260 Cook, J., 249 Faraco, M., i, 6, 9, 14, 15, 17, 18, 36, 40, 47, Cook, V. J., 44, 148, 250 148, 149, 150, 163, 251, 252, 277 Cooper, P. L., 114, 250 Fayol, M., 21, 104, 251, 252, 258 Couzijn, M., 248, 250, 253, 261 Fearnow, S., 171, 254 Covill, A., 104, 257 Fidell, L. S., 62, 262 Cumming, A., ii, 3, 4, 7, 8, 10, 11, 12, 13, 15, Fine, M., 236, 252 16, 17, 18, 19, 20, 23, 24, 25, 29, 30, 31, 32, Fintz, C., 248 34, 36, 37, 39, 41, 43, 44, 45, 46, 47, 49, 50, Fishco, V., 127, 138, 248 51, 52, 55, 58, 66, 68, 71, 72, 77, 104, 106, Fitzgerald, J., 81, 82, 95, 252 120, 121, 122, 136, 190, 191, 192, 194, 202, Flavell, J. H., 26, 252 204, 250, 258, 261, 263, 277 Flege, J. E., 130, 252 Cumming, A.,, ii, 52, 58, 277 Fleishman, J., 111, 114, 252 Cummins, J., 4, 22, 32, 40, 45, 131, 194, 234, Flower, L., 3, 21, 37, 44, 46, 50, 52, 55, 58, 70, 235, 240, 246, 250 103, 171, 252, 254, 262 Flowerdew, J., 148, 249, 251, 252, 254, 255 Dailey, B., 13, 47, 262 Foster, P., 226, 252 Davies, A. M., 146, 258 Foulin, J. N., 149, 155, 156, 252 Davis, J. N., 149, 251 Francis, N., 45, 252 Day, R., 260 Franken, M., 187 De Avila, E. A., 86, 251 Franken, M., ii, 8, 10, 214, 228, 236, 252, 277 de Beaugrande, R. A., 24, 250 Freebody, P., 234, 258 de Bot, K., 212, 227, 250 Freire, P., 234, 236, 238, 252, 261 De Glopper, K., 148, 277 Freund, R. J., 216, 257 de Groot, A. M. B., 2, 250 Friedlander, A., 13, 15, 16, 17, 18, 19, 20, 29, Delaney, P. F., 135, 251 30, 37, 47, 252 Desai, L., 263 Fuchs, D., 192, 253 DeSilva, A. D., 236, 250 Fuchs, L., 253 Donato, R., 190, 202, 250 Fujiwara, M., 171, 254 Doughty C., 170, 251, 262 Downing, B., 13, 47, 262 Galambos, S. J., 124, 253 Doyle, W., 192, 225, 251 Galbraith, D., 24, 77, 253, 259 Dressler, W.V., 24, 250 Gardner, H., 68, 76, 253 Duguid, P., 46, 191, 248 Gardner, R. C., 194, 263 Duncan, S., 86, 251 Garton, A., 81, 253 Dunkel, P. A., 149, 251 Gaskill, W., 17, 20, 31, 32, 33, 43, 44, 47, 253 Durgunoğlu, A. Y., i, 82, 88, 258 Gass, S., 148, 196, 253, 262 Dutka, S., 192, 253 Gathercole, S. E., 147, 253 Genesee, F., 1, 253 Eagan, R., 237, 249 Gillespie, 260 Echevarriarza, M., 47, 191, 263 Gillette, S., 13, 47, 192, 262 Edelsky, C., 16, 17, 18, 31, 33, 45, 47, 236, Glover, J. A., 104, 248 251 Goelman, H., 227, 261 Ehri, L. C., 82, 251 Goh, C. C. M., 212, 253 Goldin-Meadow, S., 124, 253

AUTHOR INDEX 265

Goldman, S., 192, 264 Hoffmann, A., 254 Goldstein, L., 194, 250 Holliday, L., 258 Goodman, K., 236, 253 Holquist, M., 46, 254 Goodman, Y., 236, 253 Hoover, W. A., 82, 254 Gordon, V., 193, 263, 264 Hornberger, N., 194, 254 Gosden, H., 42, 253 Horton, S. R., 136, 254 Goswami, U., 95, 253 Hoyne, S. H., 104, 257 Gough, P. B., 82, 253, 255 Huberman, A., 196, 257 Grabe, W., 12, 21, 23, 38, 52, 76, 77, 78, 103, Huckin, T., 203, 248 104, 194, 253 Hudelson, S., 236, 237 Graesser, A., 136, 143, 257 Huggins, A. W., 85, 247 Gredler, M. E., 228, 253 Hughey, J. B., 15, 216, 254, 255 Greene, M., 236, 253 Hulstijn, J. H., i, 6, 109, 148, 254, 277 Gregg, L., 248, 252, 254 Griffith, P. L., 82, 255 Igoa, J. M., 38, 247 Griffiths, R., 149, 253 Isaac, S., 53, 254 Guilford, J. P., 117, 253, 262 Ivanic, R., 47, 254 Guillabert, F., 149, 153, 249 Iwahara, S., 64, 66, 254 Izumi, S., 171, 254 Haas, C., 55, 58, 252 Hadwin, A. F., 146, 147, 253 Jacobs, H. L., 15, 58, 61, 216, 254, 255 Hakuta, K., 83, 123, 236, 247, 253 Jacobs, V. A., 86, 249 Hall, C., 14, 15, 16, 17, 18, 19, 23, 32, 33, 34, Janki, A., 211, 214, 216, 263 36, 41, 43, 44, 47, 247, 249, 252, 253, 254, Janssen, D., 53, 255 256, 257, 264 Jechle, T., 251 Hall, E., 14, 15, 16, 17, 18, 19, 23, 32, 33, 34, Jeffery, G., 34, 103, 147, 164, 252, 255, 256, 36, 41, 43, 44, 47, 247, 249, 252, 253, 254, 259, 263, 264 256, 257, 264 Johns, A. M., 12, 254, 255 Hamers, J. F., 123 Johnson, J. S., 255 Hamlett, C., 192, 253 Johnson, S., 86, 139, 255, 257 Hamp-Lyons, L., 17, 53, 254, 262 Joiner, E. G., 15, 47, 264 Haneda, M., 190, 202, 254 Jolliffe, D. A., 73, 251 Hanna, G., 127, 138, 248 Jones, C. S., 3, 4, 14, 15, 16, 17, 18, 19, 31, 32, Hansen, C., 148, 254 33, 36, 43, 47, 51, 52, 105, 106, 255 Hansen, N., 123, 125, 258 Jöreskog, K., 111, 112, 255 Haro, P., 47, 191, 263 Juel, C., 82, 255 Hartfiel, V. F., 15, 216, 254, 255 Hashiuchi, T., 56, 254 Hayden, H. M., 13, 16, 17, 18, 19, 20, 31, 33, Kamler, B., 234, 236, 250 34, 47, 251 Kaplan, R., 12, 21, 23, 38, 52, 76, 77, 78, 103, Hayes, J. R., 3, 9, 14, 15, 16, 17, 19, 20, 21, 104, 194, 253 27, 33, 37, 44, 47, 50, 52, 55, 58, 61, 64, 69, Karns, K., 192, 253 77, 103, 249, 252, 254, 262 Kasper, G., 204, 255 Hayes-Roth, B., 136, 254 Kasper, L. F., 15, 20, 26, 28, 47, 255 Hayes-Roth, F., 136 Kasper, L.F., 14 Hayes-Roth, F., 254 Katzroff, M., 192, 253 Hazenberg, S., 109, 254 Kellerman, E., 204, 255 Hemlo, J., 264 Kellogg, R. T., 104, 134, 135, 136, 147, 149, Henry, K., 14, 15, 16, 17, 20, 28, 33, 40, 41, 164, 255, 256 47, 254 Kern, R. G., 44, 148, 255 Hermans, H. J. M., 46, 254 Keys, C. W., 171, 255 Higgins, L., 171 Khaldieh, S., 16, 17, 18, 20, 26, 27, 47, 255 Hillocks, G., 23, 143, 254 Kiewra, K. A., 146, 255 Hirose, K., 3, 4, 13, 14, 15, 16, 18, 19, 25, 31, Kim, S., 146, 255 32, 33, 40, 43, 46, 47, 53, 55, 56, 58, 62, 63, King, P., 148, 255 68, 77, 106, 254, 260 Kintsch, W., 104, 147, 153, 255, 263 Hoefnagel-Hohle, M., 262 Kirby, J. R., 146, 147, 253

266

Kirshner, D., 248, 251 Manchón, R.M., 12, 13, 14, 16, 17, 18, 19, 29, Knudsen, R. E., 214, 216, 226, 256 37, 46, 47, 257 Knudson, R. E., 136 Marcoulides, G. A., 112, 117, 259 Kobayashi, H., 4, 14, 15, 16, 17, 18, 20, 28, 29, Marek, P., 104, 256 33, 41, 43, 47, 68, 69, 256, 260 Marín, J., 13, 47, 148, 260 Koda, K., 15, 29, 256 Martin, N., i, 248 Kraft, R. G., 104, 248 Mason, L., 135, 136, 261, 263 Kramsch, C., 46, 256 Matsuda, P., 1, 250, 253, 256, 261 Krapels, A. R., 12, 22, 49, 65, 256 Mc Donough, S. H., 12, 257 Krings, H. P., 33, 256 McCann, T. M., 214, 216, 226, 257 Kroll, B., 15, 47, 249, 250, 252, 254, 255, 256, McCloskey M., 238 260, 261 McCormick,, 190, 202, 250 Kroll, J. F., 2, 250 McCutchen, D., 104, 257 Kross, B. M., 249 McLaughlin, B., 123, 125, 258 Krueger, N., 123, 125, 258 McLeod, A., 248 Kuehn, P., 14, 15, 16, 18, 19, 20, 47, 249 Meijers, G., 260 Kuiken, F., ii, 8, 10, 171, 234, 241, 256, 277 Meiran, N., 225, 257 Mellow, D., 121, 250 Lahtinen, V., 146, 256 Ménard, N., 4, 14, 16, 17, 18, 19, 20, 24, 31, Lantolf, J., 192, 194, 247, 256 32, 33, 34, 43, 47, 51, 52, 106, 148, 264 Lapkin , S., 13, 16, 17, 18, 19, 20, 28, 47, 170, Merziger, G., 251 171, 213, 225, 262 Michael, W. B., i, ii, 10, 53, 254, 277 Larsen-Freeman, D., 237 Miettinen, R., 192, 251 Latham, G. P., 191, 257 Mildes, K., 104, 257 Laufer, B., 103, 256 Miles, M., 196, 257 Lay, N., 13, 16, 17, 18, 19, 29, 36, 37, 47, 256 Miller, K. S., 13, 16, 17, 18, 19, 29, 37, 46, 47, Lea, J., 45, 104, 147, 164, 256 257 Ledwell, M., 15, 47, 51, 52, 250 Miller, L., 148, 252 Leki, I., 2, 9, 37, 38, 191, 194, 256 Miyake, A., 251 Levelt, W.J.M., 38, 256 Mohanty, A. K., 124, 125, 257 Levy, C. M., i, 3, 4, 6, 7, 9, 10, 13, 14, 15, 17, Moll, L. C., 194, 236, 260 18, 20, 33, 36, 45, 47, 66, 77, 104, 121, 123, Moragne e Silva, M., 13, 16, 17, 18, 19, 24, 31, 127, 131, 133, 134, 135, 136, 139, 140, 147, 33, 34, 41, 43, 45, 46, 47, 51, 52, 257 149, 160, 163, 164, 254, 255, 256, 257, 259, Morgan, C. H., 146, 257 260 Morgenthaler, L., 258 Lewis, N., 213, 252, 258 Mosenthal, P., 225, 254, 258, 262 Li, T-Y., 68, 194, 249, 257 Moss, P. A., 20, 258 Li, X., 68, 194, 249, 257 Murie, R., 13, 47, 262 Lilley, J. D., 146, 257 Murphy, L., i, 5, 9, 13, 16, 46, 47, 50, 51, 55, Lindbloom-Ylaenne, S., 146, 256 83, 148, 149, 249, 257, 260, 277 Linton, P., 139, 255 Murray, D., 135, 249, 258 Littell, R. C., 216, 257 Muspratt, S., 234, 258 Little, L. D., 42, 102, 136, 143, 179, 242, 257 Myers, M., 87, 258 Locke, E. A., 191, 257 Lockhart, W., 3, 264 Nagata, Y., 68, 258 Long, M. H., 170, 174, 186, 237, 251, 257 Nagy, W. E., 88, 251 Lonka, K., 135, 136, 146, 256, 261, 263 Nagy, W.E., 88, 251 Loschky, L., 148, 249 Nakadate, N., 42, 252, 264 Luke, A., 234, 258 Nash, J. G., 61, 64, 69, 254 Nassaji, H., 191, 194, 258 Macedo, D., 234, 252 Nation, P., 103, 256 Madigan, R., 255 Nayak, N., 123, 125, 258 Magliano, J. P., 136, 143, 257 Newport, E. L., 86, 255 Maguire, M. (1999)., 194, 258 Newton, J., 186, 258 Manchón, R., 12, 13, 14, 16, 17, 18, 19, 29, 37, Ng, E., 192, 258 46, 47, 257 Norris, J. M., 21, 258 Nystrand, M., 46, 47, 258, 261

AUTHOR INDEX 267

Oakhill, J., 146, 258 Reid, J., 63, 260 Oney, B., 251, 258 Reyes, M., 235, 236, 260 Ortega, L., 21, 258 Ricciardelli, L. A., 131, 260 Ovando, C. J., 233, 237, 258 Rickards, J. P., 146, 260 Oxford, R., 204, 247, 249, 251, 253, 256, 258, Rieben, L., 251 261, 262, 263, 264 Rieke, R., 211, 214, 216, 263 Rijlaarsdam, G., 10, 24, 44, 66, 131, 144, 164, 248, 250, 252, 253, 255, 256, 259, 260, 261, Paris, P., 24, 247, 248, 258, 259, 260 263, 277 Parks, S., 194, 258 Rinnert, C., 4, 14, 15, 16, 17, 18, 20, 28, 29, Pasta, D. J., 235, 259 33, 41, 43, 47, 68, 69, 78, 256, 260 Pélissier, A., 9, 259 Risch, N., 146, 255 Penningroth, S. L., 104, 258 Rivière, A., 247 Pennington, M. C., 4, 13, 15, 16, 17, 18, 19, Robinson, E. J., 36, 170, 174, 175, 211, 257, 20, 23, 31, 33, 41, 44, 47, 258 260, 263 Perdue, C., 149, 258 Robinson, P., 36, 170, 174, 175, 211, 257, 260, Perez, B., 236, 250, 258 263 Perfetti, C., 251 Roca de Larios, J., 5, 9, 13, 14, 16, 17, 18, 19, Perl, S., 212, 258 20, 31, 33, 36, 42, 45, 46, 47, 50, 51, 55, 67, Petraglia, J., 171, 254, 260 83, 91, 257, 260 Peyton, J., 191, 258 Rosen, H., 248 Pica, T., 213, 225, 258 Rosenberg, S., 104, 258 Pierce, L., 259 Rosenfeld, M., 77, 262 Piolat, A., i, 6, 9, 10, 14, 47, 147, 164, 259, Rubin, J., 204, 248, 264 277 Rueda, R., 194, 260 Pittard, V., 46, 259 Ruhl, K. L., 146, 260 Plake, B. S., 104, 248 Rumelhart, D. E., 226, 260 Porte, G., 13, 16, 17, 18, 19, 20, 23, 26, 27, 37, Rummel, M., 236, 259, 260 38, 46, 47, 259 Russell, D., 192, 260 Pozo, J. I., 17, 259 Pratt, C., 81, 253 Prawat, R., 192, 259 Sanders, M., 102, 260 Pressley, M., 146, 263 Sasaki, M., i, 3, 4, 5, 7, 9, 13, 14, 15, 16, 17, Prior, P., 47, 191, 203, 227, 259 18, 19, 20, 24, 25, 27, 28, 31, 32, 33, 36, 40, Punamaki, R., 251 43, 46, 47, 49, 53, 54, 55, 56, 57, 58, 60, 62, 63, 68, 72, 77, 106, 123, 148, 254, 260, 278 Sato, C., 213, 229, 260 Qi, D. S., 14, 15, 16, 17, 18, 19, 29, 37, 46, 47, Scardamalia, M., 3, 21, 22, 24, 34, 37, 44, 56, 259 61, 65, 68, 82, 103, 106, 134, 135, 192, 202, Quintero, E., ii, 8, 9, 10, 236, 259, 260, 277 214, 226, 227, 248, 249, 260, 261 Scheier, M., 191, 249 Raimes, A., 14, 15, 16, 17, 18, 19, 20, 23, 24, Schereiner, C., 12, 261 25, 26, 29, 36, 37, 40, 41, 42, 43, 44, 46, 47, Schmidt, R., 170, 211, 261 50, 51, 52, 55, 259 Schoonen, R., i, 6, 9, 102, 103, 108, 114, 148, Ramanathan, V., 78, 259 261, 263, 278 Ramey, D. R., 235, 259 Schriver, K. A., 55, 58, 252 Ramirez, J. D., 259 Schunk, D., 192, 261 Ramsay, R. M. G., 125, 259 Seaman, M. A., 15, 47, 264 Ransdell, S., i, 3, 4, 6, 7, 9, 10, 13, 14, 15, 16, Segalowitz, N. S., 149, 163, 261 17, 18, 20, 33, 34, 36, 37, 45, 47, 66, 77, Segalowitz, S. J., 149, 163, 261 104, 121, 123, 127, 131, 133, 134, 135, 136, Sengupta, S., 15, 16, 17, 18, 19, 20, 28, 261 139, 140, 142, 149, 160, 163, 164, 254, 255, Shanahan, T., 81, 82, 95, 252 256, 257, 259, 260, 277 Shaw, P., 14, 17, 18, 19, 27, 47, 63, 148, 261 Raykov, T., 112, 117, 259 Shi, L., 191, 261 Rebuffot, J., 15, 47, 51, 52, 250 Shor, I., 236, 261 Reese, W. H., 259 Sijtstra, J., 102, 261 Reichelt, M., 16, 259 Silberstein, S., 15, 47, 249

268

Silva, T., 1, 3, 12, 13, 14, 15, 16, 17, 18, 19, Ting, R-Y., 13, 14, 16, 17, 18, 19, 27, 33, 40, 20, 23, 30, 32, 33, 47, 49, 52, 56, 63, 106, 47, 63, 148, 261, 263 191, 194, 250, 253, 256, 261 Torrance, M., 36, 103, 147, 164, 252, 253, 255, Simon, H. A., 3, 52, 59, 251 256, 259, 263 Skehan, P., 170, 173, 174, 175, 226, 252, 261, Torres-Guzman, M. E., 263 262 Toulmin, S. E., 211, 214, 216, 263 Skibniewski, L., 15, 17, 18, 23, 24, 31, 32, 47, Treiman, R., 95, 263 51, 52, 261 Tunmer, W. E., 82, 253, 254 Skinner, A. M., 73, 251 Tynjälä, P., 263 Slotte, V., 146, 261 Smagorinsky, P., 13, 52, 196, 262 Uchida, N., 53, 57, 59, 60, 66, 68, 73, 247, 263 Smith, J., 14, 15, 16, 17, 18, 19, 25, 37, 39, 41, Ullman, C., 236, 263 42, 43, 44, 45, 46, 47, 260, 262 Uzawa, K., 13, 15, 16, 17, 18, 19, 20, 23, 24, Smith, V., 14, 15, 16, 17, 18, 19, 25, 37, 39, 25, 26, 29, 31, 36, 45, 46, 47, 51, 52, 106, 41, 42, 43, 44, 45, 46, 47, 260, 262 191, 263 Snow, C. E., 262 So, S., 4, 13, 15, 16, 17, 18, 19, 20, 23, 31, 33, Valdes, B., 191 41, 44, 47, 70, 71, 73, 79, 104, 110, 120, Valdés, G., 14, 15, 16, 17, 18, 19, 30, 31, 40, 165, 166, 173, 181, 185, 198, 199, 200, 201, 47, 263 204, 205, 206, 207, 208, 209, 216, 239, 258 Vallen, T., 102, 263 Sommers, N. I., 38, 262 van den Bergh, H., 4, 53, 248, 250, 253, 255, Sörbom, D., 111, 112, 255 260, 261, 263 Spack, R., 191, 203, 262 van Waes, L., 53, 255 Spector, P. C., 216, 257 Vancouver, J. B., 191, 203, 247 Stack, L., 238 VanPatten, B., 170, 211, 263 Stanovich, K. E., 105, 262 Vedder, I., ii, 8, 10, 171, 234, 241, 256, 278 Stijnen, S., 102, 263 Verhallen, M., 102, 263 Stotsky, S., 41, 262 Verhoeven, L., 237, 263 Stratman, J., 53, 262 Vermeer, A., 102, 247 Sullivan, J. F., 146, 260 Victori, M., 3, 14, 15, 16, 17, 19, 20, 23, 24, Suritsky, S., 146, 260 25, 26, 27, 36, 39, 43, 44, 45, 46, 47, 104, Swain, M., 13, 16, 17, 18, 19, 20, 28, 47, 170, 264 171, 204, 213, 225, 262 Voss, J., 192, 264 Swarts, H., 50, 262 Voss, J. F., 228, 249 Sweigart, W., 212, 213, 216, 224, 225, 226, Vye, N., 192, 264 227, 229, 262 Vygotsky, L. S., 125, 228, 264 Sweller, J., 225, 226, 262

Wajnryb, R., 171, 264 Tabachnick, B. G., 62, 262 Walmsey, S.A., 254 Takeuchi, K., 60, 262 Warren, E., 149, 264 Tamar, L., 254 Waters, A., 194, 264 Tannenbaum, R. J., 77, 262 Way, D. P., 15, 16, 18, 19, 20, 29, 47, 91, 264 Tarone, E., 13, 15, 16, 18, 20, 28, 47, 262 Weaver, S. J., 68, 249 Tauroza, S., 148, 247 Weissberg, B., 148, 264 Tetroe, J., 3, 4, 14, 15, 16, 17, 18, 19, 31, 32, Wells, G., 198, 249, 264 33, 36, 43, 47, 51, 52, 105, 106, 134, 249, Wenden, A., 204, 264 255 Wertsch, J., 256, 264 Thomas, G., 14, 36, 124, 126, 262, 263 Whalen, K., 4, 14, 16, 17, 18, 19, 20, 24, 31, Thomas, J., 14, 36, 124, 126, 262, 263 32, 33, 34, 43, 47, 51, 52, 106, 148, 264 Thomas, M., 14, 36, 124, 126, 262, 263 Whitson, J., 248, 251 Thorson, H., 16, 17, 18, 19, 31, 33, 34, 35, 46, Widdowson, H. G., 15, 262, 264 47, 263 Williams, J., 170, 192, 251, 257, 262, 264 Tierney, R., 244, 263 Williams, S., 170, 192, 251, 257, 262, 264 Timbur, G., 12, 263 Winsor, D., 192, 264 Ting Kun Liu, E., 14, 17, 18, 19, 27, 47, 148, Winter, A., 251 261 Witte, S. P., 42, 46, 252, 264

AUTHOR INDEX 269

Yuen, S. D., 235, 259 Woodhouse, R. A., 147, 253 Wormuth, D. R., 15, 216, 254, 255 Zamel, V., 4, 12, 13, 15, 16, 17, 19, 23, 26, 36, 39, 41, 44, 46, 47, 49, 50, 51, 52, 55, 66, Yau, M. S. S., 3, 15, 16, 17, 18, 19, 33, 39, 47, 191, 212, 264 264 Zhou, Y., ii, 7, 136, 196, 278 Yokoi, L., 146, 263 Zimmerman, R., 9, 74, 264 Yoshida, M., 68, 258 Zinchenko, V., 193, 264 Zinkgraf, S. A., 255

SUBJECT INDEX accessibility, 104, 105 151, 153, 154, 155, 162, 163, 164, 212, 247, adults, 7, 16, 31, 87, 125, 126, 130, 149, 189, 248, 249, 253, 255, 257, 260, 261, 263 195, 252, 258 constraints, 6, 12, 13, 21, 24, 32, 33, 39, 44, age, 2, 7, 16, 28, 32, 45, 55, 56, 102, 123, 124, 55, 103, 106, 117, 118, 147, 153, 195, 225, 125, 126, 127, 128, 129, 130, 131, 151, 183, 228, 229, 252, 259 195, 226, 233, 234, 238, 250 content knowledge, 227, 238 analysis context, 2, 4, 7, 8, 11, 15, 16, 29, 31, 34, 40, process analysis, 170 41, 42, 46, 47, 72, 102, 120, 131, 148, 151, product analysis, 170 152, 171, 187, 189, 190, 191, 193, 197, 201, argumentative text, iii, iv, 8, 51, 149, 250 212, 213, 222, 233, 234, 236, 238, 242, 245, assessment, 11, 14, 15, 19, 20, 41, 61, 69, 71, 249, 251, 256, 257, 259, 264 74, 102, 114, 190, 228, 233, 234, 241, 244, critical literacy, 233, 234, 236, 237, 238 245, 248, 249, 250, 252, 255, 258, 261 critical period, 123, 130, 252 attention, 2, 3, 4, 12, 23, 24, 25, 27, 28, 31, 32, cross-language transfer, 6, 81, 83, 100 33, 40, 42, 44, 45, 50, 51, 53, 66, 67, 69, 71, cultural factor, iv, 35 76, 77, 78, 87, 106, 108, 124, 125, 126, 136, data collection, 11, 16, 17, 32, 57, 108, 238 147, 154, 170, 174, 175, 178, 179, 180, 185, deduction, 177, 179, 180 200, 205, 211, 212, 213, 227, 234 design attentional resource, 211, 212, 213, 225, 227 repeated measures design, 216 audience, 17, 19, 23, 29, 31, 36, 39, 42, 46, dictogloss, 169, 170, 171, 175, 176, 179, 187 109, 226, 244, 246 editing, 23, 25, 26, 27, 37, 189, 196, 198, 199, bilingualism, 125, 257 200, 201 Bonferroni correction, 62, 63, 64, 66, 68 EFL ciscourse community, 42, 203, 247 Japanese EFL learner, 49, 54, 55 claim, 32, 54, 213, 225, 228, 229 Efl writing, 6, 27, 49, 54, 55, 58, 76, 101, 106, coding of ideas, 23, 33 118, 120, 121, 247, 250, 254, 260, 264 coding system, 50, 60 English as a second language, 1, 8, 50, 143, cognates, 40, 109, 126 189, 252, 253, 255 cognitive ESL writing, 16, 190, 194, 195, 197, 204, 250 abilities, 56 executive process, 149 load, 104 expert writer, 21, 44, 49, 52, 55, 70, 72, 212, cognitive abilities, 1, 4, 5, 56, 123, 131, 142, 216 147 flexibility, 9, 23, 37, 125, 235, 244, 259 Cognitive abilities, 56 fluency, 6, 7, 15, 20, 25, 26, 29, 33, 37, 42, 43, cognitive approach, 261 49, 63, 68, 101, 105, 107, 111, 113, 118, cognitive capacity, 104, 106, 147 120, 121, 123, 126, 127, 128, 130, 131, 133, cognitive complexity, 174, 175, 225 134, 135, 136, 139, 140, 141, 142, 145, 152, cognitive constraint, 104 156, 157, 158, 159, 160, 161, 162, 163, 200, cognitive load, 8, 104, 149, 150, 162, 163, 164, 225, 244, 246, 252, 257, 258, 259, 262 226, 247, 255 focused input, 175 Cognitive load, 104 foreign language writing, 1 cognitive processing, 1, 3, 4, 104 formulation, 9, 11, 21, 23, 25, 29, 31, 32, 33, cognitive resource, 4, 6, 9, 30, 105, 106, 145, 37, 40, 42, 43, 44, 136, 260 147, 148, 149, 150, 247 genre, 18, 31, 34, 35, 46, 215, 217, 225, 226 collaboration goal setting, 4, 7, 61, 192, 254 collaborative dialogue, 169, 170, 173, 262 grammar, 14, 28, 36, 82, 87, 91, 101, 103, 104, collaborative writing task, 169, 175, 187 105, 107, 121, 125, 126, 138, 172, 175, 177, communicative quality, 211, 212, 215, 216, 179, 180, 185, 187, 189, 196, 197, 199, 200, 217, 218, 219, 225 201, 205, 207, 208, 234, 241, 242, 249, 256, composition 263 college composition, 133 grammatical accuracy, 211, 212, 215, 216, comprehension, 6, 9, 82, 86, 127, 129, 136, 217, 218, 219, 220, 225, 229 138, 139, 140, 141, 145, 147, 148, 149, 150, grammatical complexity, 169, 175, 176, 181, 184, 185

272

grammatical knowledge, 103, 107, 121 198, 212, 216, 217, 218, 220, 222, 223, 227, ground 228, 229, 235, 238, 241, 245, 248, 249, 254, elaboration of ground, 211, 216, 218, 219, 256, 264 220, 223, 224 lexical retrieval, 37, 104, 105, 107, 121 grounds, 19, 26, 211, 216, 218, 219, 220, 221, lexical richness, 103, 121, 169, 175, 176, 181, 222, 223, 224, 225, 231 183, 184, 185 group dynamics, 169, 186 lexical strategies, 170, 176, 180, 184 ideas and organisation, 211, 215, 216, 217, linguistic knowledge, 3, 4, 6, 8, 39, 83, 84, 218, 219 101, 103, 104, 105, 106, 107, 111, 170, 186, information gap, 186 228, 264 information processing model, 211 linguistic threshold, 105 input linguistics, 1, 175, 250, 261, 262 input processing, 174, 263 LISREL, 111, 114, 252, 255 modified input, 174, 175 literacy, 8, 9, 21, 22, 27, 34, 35, 45, 46, 81, 82, interaction, ii, 1, 8, 15, 23, 26, 36, 37, 38, 39, 83, 88, 89, 90, 91, 120, 192, 233, 234, 235, 40, 43, 44, 105, 106, 140, 141, 169, 170, 236, 237, 238, 241, 244, 245, 246, 247, 249, 174, 175, 185, 186, 187, 211, 212, 213, 215, 250, 252, 255, 258, 260, 262 216, 217, 218, 219, 220, 221, 222, 223, 224, development, 81, 82, 83, 236, 237, 244 225, 226, 227, 228, 229, 235, 236, 239, 251, local planning, 49, 61, 68, 69, 71, 72 253, 256, 258 long-term memory, 135, 174, 175 interactional expertise, 228 material factor, 225, 226 interviews, 5, 13, 17, 29, 189, 195, 196, 197, mean length of utterance, 182, 184 198, 199, 201, 202, 203, 237 measurement model, 112, 115, 116, 118, 252 knowledge metacognition, 122, 169, 170, 264 domain-specific knowledge, 8, 135, 211, metacognitive awareness, 170, 171, 186 222, 223, 224, 236 metacognitive knowledge, 6, 23, 27, 28, 101, personal knowledge, 26 103, 104, 106, 107, 121, 122 rule knowledge, 177, 179, 180 metaknowledge, 26, 53, 56, 82, 254 strategic knowledge, 26, 106 MLU, 182, 184, 185 task knowledge, 26, 104 modularity, 38 vocabulary knowledge, 6, 81, 100, 102, monitoring, 28, 39, 44, 49, 72, 73, 82, 106, 103, 121 190, 212, 253 knowledge transfer, 3 multiple data sources, 5, 49, 53, 76 L1 and L2 writing process, 5, 11, 35, 51, 55 native language literacy, 233, 236 L2 notetaking activity, 6, 145, 153, 154, 155 acquisition, 9, 142, 145, 170, 171, 237 noticing, 8, 169, 170, 173, 174, 175, 176, 185, early L2 exposure, 7, 123, 127, 128, 129 187, 211, 212, 227, 228, 229, 254, 260 proficiency, 5, 10, 11, 13, 14, 22, 27, 28, novice writer, 22, 49, 52, 55, 56, 57, 63, 71, 29, 30, 38, 39, 40, 41, 43, 44, 45, 46, 51, 216 52, 53, 63, 67, 68, 76, 169, 176, 185, oral and written language, 233 186, 262 orthography, 95, 103, 105, 107, 258 proficiency., 5, 10, 11, 13, 28, 67, 76, 176, output hypothesis, 169, 212, 213, 225, 250, 185, 186 254 language pausing, 23, 25, 29, 31, 49, 54, 155, 162, 257 knowledge, 101, 102, 103, 104, 105, 107, pausing behavior, 29, 49, 54 113, 122, 125, 142, 143 performance loss, 34 proficiency, 22, 39, 41, 85, 102, 112, 121, planning, 3, 7, 9, 20, 21, 22, 23, 24, 25, 27, 28, 122, 227, 228, 250, 260 29, 30, 31, 32, 33, 36, 37, 39, 42, 43, 44, 45, strategies, 169 46, 47, 51, 55, 56, 59, 60, 61, 64, 65, 67, 69, latent variable, 111, 112, 113, 114, 115, 116, 70, 71, 72, 73, 74, 79, 80, 105, 106, 133, 117, 118, 248 134, 135, 136, 138, 149, 189, 198, 200, 201, level, iii, 4, 6, 13, 14, 18, 22, 24, 26, 28, 29, 32, 205, 212, 226, 233, 234, 247, 249, 252, 254, 33, 34, 35, 36, 39, 41, 43, 44, 49, 50, 51, 58, 257 61, 62, 63, 66, 68, 74, 81, 84, 85, 88, 100, global planning, 49, 55, 60, 61, 68, 69, 70, 102, 104, 105, 106, 107, 108, 121, 122, 128, 71, 72, 74 129, 131, 134, 137, 145, 147, 148, 150, 152, thematic planning, 61, 69 156, 158, 162, 169, 176, 177, 185, 190, 193, planning time, 27, 64, 65

SUBJECT INDEX 273 pragma-rhetorical skill, 173 reading and writing strategy, 101, 103, 104, primary trait, 109 107, 110 problem solving, 244, 264 simplification strategy, 184, 185 problem-posing, 8, 233, 234, 235, 236, 238, structural equation modelling, i, 101, 111 239, 243, 245 structural model, 115, 116, 117, 120 process writing instruction, 27, 49, 53, 54, 55, study of pauses, 145, 149 56, 63, 73 task demands, 29, 31, 46, 174, 175, 255 processing text linear processing, 133, 136 text characteristics, 101, 103, 104, 107, 108, nonlinear processing, 133, 134, 135, 143 110 processing load, 258 text difficulty, 176, 186 processing resources, 174, 175 text quality, iii, 8, 20, 34, 37, 66, 169, 175, process-product, 7 185, 186, 211, 212, 216, 217, 219, 248 product-oriented studies, 53, 55, 69 text reconstruction, 169, 176, 177, 178, 179, psycholinguistics, 1, 2, 4, 250 180 qualitative study, 187, 233, 234 think-aloud, 1, 3, 5, 13, 17, 19, 20, 28, 43, 50, rating, 15, 30, 81, 87, 103, 109, 126, 139, 229 51, 52, 53, 57, 106, 120, 133, 135, 138, 262 reactivity, 53 think-aloud protocol, 1, 3, 5, 20, 50, 51, 52, Reactivity, 53 120, 133, 135, 138, 262 readiness, 130, 174, 175 thought recursiveness, 23, 44 fragmentation of thought, 33 register, 40 topic, 8, 18, 21, 23, 24, 29, 30, 31, 33, 37, 39, rehearsing, 23, 25, 31, 44 42, 56, 58, 62, 64, 68, 69, 76, 87, 90, 114, reliability, 2, 11, 19, 20, 108, 109, 113, 114, 128, 135, 136, 189, 192, 204, 205, 206, 207, 116, 128, 139, 216, 252, 258 208, 212, 214, 215, 222, 223, 224, 225, 227, rereading, 20, 26, 29, 47, 66, 67, 68, 148 228, 230, 236, 238, 244, 260 research methods in psychology, 1 training intervention, 133 resourcing, 25, 215, 216, 217, 218, 219, 220, transcription, 3, 25, 33, 43, 148, 149, 152, 165 221, 222, 223, 224 transfer, 1, 3, 4, 5, 9, 11, 19, 20, 21, 22, 31, 32, restructuring, 29, 40, 50, 67, 260 34, 35, 40, 45, 46, 84, 90, 91, 95, 97, 98, retrieval speed, 101, 122 100, 142, 143, 148, 151, 177, 179, 180, 192, retrospective accounts, 59 194, 226, 237, 248, 249, 251, 253 revising, 3, 7, 23, 24, 26, 39, 43, 46, 56, 57, translating from L1 to L2, 68 106, 133, 138, 196, 198, 199, 241, 248 university contexts, 189 rhetorical demands, 29, 42 word retrieval, 101, 103, 104, 105, 107, 114 rhetorical refining, 25 working memory, 6, 7, 30, 33, 37, 44, 45, 104, salience, 174 106, 133, 134, 135, 141, 143, 145, 147, 150, sample size, 16, 50, 76, 91 164, 174, 247, 251, 255, 257, 259 second language writing, 17, 194, 211, 236, writing 250, 253, 254, 256, 259, 261, 264 ability, 3, 5, 11, 14, 15, 16, 20, 22, 25, 32, sentence building, 101, 103, 104, 105, 107 35, 38, 39, 40, 41, 42, 43, 44, 45, 46, 52, sentence complexity, 182 53, 62, 63, 64, 73, 101, 102, 103, 106, skilled performance, 42, 251 142, 212, 250, 259, 261, 263 social factors, 9, 45, 130 assignment, 32, 106, 108, 109, 111, 113, socio-cognitive perspective, 35 115, 121, 255 spelling, 3, 4, 5, 6, 54, 81, 82, 83, 84, 85, 86, behavior, 5, 25, 32, 49, 50, 53, 54, 59, 77, 88, 89, 90, 91, 95, 96, 99, 100, 103, 110, 190 138, 154, 172, 175, 177, 179, 180, 185, 253 component, 246 stimulated recall, 5, 13, 17, 19, 49, 53, 54, 55, componential nature of writing, 103 189, 195, 196, 203 expertise, 3, 4, 5, 9, 22, 27, 49, 50, 51, 55, stimulated recall protocols, 5, 19, 49, 54 56, 64, 67, 72, 74, 76, 131, 250 strategic competence, 72, 74, 76 fluency, 53, 55, 63, 73, 104, 121, 123, 127, strategy, 4, 6, 25, 29, 37, 60, 65, 67, 95, 104, 128, 129, 133, 135, 138, 139, 141, 143 106, 133, 135, 136, 137, 138, 141, 142, 143, model, 3, 6, 21, 22, 42, 77, 101, 103 147, 148, 163, 175, 176, 179, 180, 190, 199, performance, iii, 5, 22, 40, 46, 59, 103, 107, 241, 253, 259, 261 133, 143

274

process, iii, iv, 3, 4, 5, 8, 9, 12, 13, 21, 22, strategy, 7, 50, 58, 60, 72, 76, 133, 134, 25, 30, 33, 35, 36, 39, 40, 41, 45, 46, 49, 136, 141, 143 50, 51, 52, 53, 54, 55, 56, 57, 59, 72, 73, subprocess, 7, 133, 134, 136 74, 76, 77, 78, 104, 106, 120, 135, 148, task, 3, 6, 7, 8, 9, 13, 18, 23, 26, 30, 42, 46, 149, 176, 190, 212, 243, 247, 248, 249, 50, 77, 87, 100, 105, 108, 149, 164, 171, 252, 254, 255, 256, 258, 261, 262, 263 191, 202, 203, 213, 215, 217, 226, 227, proficiency, 5, 6, 38, 50, 101, 104, 105, 234, 238, 253, 256 107, 112, 120, 121, 122, 261 training, 76, 133, 138 prompts, 108, 109, 214 writing ability, 52 quality, 7, 8, 15, 56, 93, 104, 121, 123, 128, writing expertise, 56 129, 130, 133, 134, 135, 136, 138, 139, writing process, 52, 53 142, 143, 259 empirical model of writing process, 49 signature, 7, 77, 133, 135, 136 writing strategy, 58 skill, 1, 5, 9, 11, 16, 22, 23, 25, 26, 27, 29, writing task, 77 30, 31, 34, 39, 40, 41, 42, 51, 66, 98, written production, 2, 9, 15, 145, 149, 249, 103, 131, 137, 247, 248, 258 252, 256, 259

LIST OF CONTRIBUTORS

Sofia Arino Marti, Associate Professor of English Language at the Univerity Rovira i Virgili, Tarragona, Spain, [email protected]. M. Rosario Arecco, Graduate of Psychology at the Florida Atlantic University, Ft. Lauder- dale, USA, [email protected]. Marie-Laure Barbier, Associate Professor of Psychology, Institut Universitaire de Formation des Maîtres de Lyon (University Institute for Teacher's Formation), Lyon, France, marie- [email protected]. Michael Busch, Ph.D. candidate, Second Language Education, Ontario Institute for Studies in Education, University of Toronto, Toronto, Canada, [email protected]. Alister Cumming, Professor and Head, Modern Language Centre, Ontario Institute for Studies in Education, University of Toronto, Toronto, Canada, [email protected]. Aydin Y. Durgunoglu, Professor of Psychology, University of Minnesota Duluth, Duluth, USA, [email protected]. Martine Faraco, Teacher of French as a Foreign Language and Researcher at the Laboratory Speech and Language, University of Provence, Aix-en-Provence, France. Margaret Franken, Senior Lecturer in Language Studies, Massey University, Palmerston North, New Zealand, [email protected] Amos van Gelderen, Senior Researcher in Language Education at the University of Amster- dam, Amsterdam, the Netherlands, [email protected]. Kees de Glopper, Professor of language and communication at the University of Groningen, Groningen, The Netherlands, [email protected]. Steven Haslett, Associate Professor and Director of the Statistics Research and Consulting Centre at Massey University, Palmerston North, New Zealand, [email protected] Jan Hulstijn, Professor of Second Language Acquisition at the University of Amsterdam, Amsterdam, the Netherlands, [email protected]. Folkert Kuiken, Assistant Professor of Second Language Acquisition at the University of Amsterdam, Amsterdam, the Netherlands, [email protected]. Beverly Lavelle, Graduate of psychology at the Florida Atlantic University, Ft. Lauderdale, USA, [email protected]. Javier Marin, Associate Professor of Psychology at the University of Murcia, Spain, [email protected]. Montserrat Mir, Assistant Professor of Spanish and Foreign Language Methodology at Illinois State University, Normal, USA, [email protected]. Liz Murphy, Associate Professor of English at the University of Murcia, Spain, liz- [email protected]. Annie Piolat, Professor of Psychology, University of the Provence, Aix-en-Provence, France, [email protected]. Elisabeth Quintero, Associate Professor of Early Childhood and Elementary Education with a focus on Literacy and Family Literacy in Multilingual, Multicultural Communities and Schools. Also developing cross-disciplinary Early Childhood Masters Program with So- cial Justice focus, at New York University, USA, [email protected]. Sarah Ransdell, Associate Professor of Psychology at Florida Atlantic University, Ft. Lauder- dale, USA, [email protected]. Gert Rijlaarsdam, Professor of Language Education at the University of Amsterdam and the Utrecht University, the Netherlands, [email protected]. Julio Roca De Larios, Associate Professor of Education at the University of Murcia, Spain, [email protected].

276

Miyuki Sasaki, Professor of Applied Linguistics at Nagoya Gakuin University, Seto, Japan, [email protected]. Rob Schoonen, Senior Researcher of Language Education and Assistant Professor of Second Language Acquisition at the University of Amsterdam, Amsterdam, the Netherlands, [email protected]. Annegien Simis, Ph.D. candidate at the University of Amsterdam, Amsterdam, the Nether- lands, [email protected]. Patrick Snellings, Ph.D. candidate at the University of Amsterdam, Amsterdam, the Nether- lands, [email protected]. Marie Stevenson, Ph.D. candidate at the University of Amsterdam, Amsterdam, the Nether- lands, [email protected]. Ineke Vedder, Assistant Professor of Second Language Acquisition at the University of Am- sterdam, Amsterdam, the Netherlands, [email protected]. Ally Zhou, Ph.D. candidate, Second Language Education, OISE/University of Toronto, To- ronto, Canada, [email protected].