Argumentative L2 text in context: An exploratory study in Australia and

Grace H. Y.

RESEARCH REPORTS

General Editor: Gregory James

VOLUME FIVE

Argumentative L2 text in context: An exploratory study in Australia and Hong Kong

Grace H. Y. Wong

LANGUAGE CENTRE Hong Kong University of Science and Technology

This report is a shortened, edited version of the author’s thesis, ‘Argumentative L2 text in context: An exploratory study in Australia and Hong Kong’ for which she was awarded the degree of PhD at Macquarie University, Sydney, 2002.

Language Centre Hong Kong University of Science and Technology Copyright © November 2003. All rights reserved. ISBN 962-7607-22-3

Postal Address: Language Centre, Hong Kong University of Science and Technology, Clear Water Bay, , Hong Kong SAR, CHINA

Telephone: (852) 2358 7880

Facsimile: (852) 2335 0249

Contents

Editorial Foreword ix

Acknowledgements xi

Synopsis xiii

Chapter One: Introduction 1

Chapter Two: Language learning in context: Australia and Hong Kong 3

Second language competence 3 Competence versus performance 3 Defining language proficiency 3 Contexts of developing second language competence 6 The place of SLC and the socio-cultural factors it presents 6 Personal/psychological factors 9 ESL versus EFL language learning 9 Learning English in Hong Kong 10 Learning English as an International student in Australia 11 Background of international students in Australian universities 11 Problems faced by international students 11 English proficiency 11 Relationships 13 Implications for the present study 14

Chapter Three: Teaching and assessing academic writing 15

Students’ academic writing needs 15 Research on students’ academic writing 16 Contrasts between good writing and poor writing 16 Rhetorical functions identified in students’ academic writing 16 Recurrent rhetorical functions in good essays 16 Problematic rhetorical functions 17 Academic conventions 17 Referencing 17 Critical attitudes toward published work 18 Cultural differences in discourse patterns 18 Assessing academic writing 19 Reliability of holistic rating 19 Validity of holistic rating 19 Extrinsic factors 20 Content/subject matter 20 Grammatical accuracy 20 Organisation 21 Questions and issues 22 Chapter Four: The structure of argumentative text and the study of 23 rhetorical relations

The structure of argumentative text 23 Global structure 23 Local structures 25 Segmentation of text 28 Using the sentence as the unit of segmentation 28 Using the T-unit as the unit of segmentation 28 Using the proposition as the unit of segmentation 29 Using the F-unit as the unit of segmentation 30 The hierarchical integration of argument and the study of rhetorical relations 31 Rhetorical Structure Theory Analysis 33 Functional Role Analysis 34 Communicative Function Analysis 36 Synthesis of previous approaches of studying rhetorical relations 37 Categorisation of rhetorical functions 37 Previously used categorisation systems of rhetorical functions 37 Categorisation system for this investigation: Rhetorical Function Analysis 40 Categorisation at the lower levels 43 Categorisation at the upper levels 44

Chapter Five: Research design and methods for analysis 45

Data collection 45 Research questions 45 Rhetorical Function Analysis (RFA): Method for investigating the hierarchical integration of argument (Research Questions 1 and 3) 47 Segmentation of text 47 Categorisation of rhetorical functions 47 Graphic representation of Rhetorical Function Analysis 47 Summary of the findings of Rhetorical Function Analysis 51 Validation of Rhetorical Function Analysis 56 Grammatical Accuracy Analysis (GAA): Method for investigating syntactic accuracy (Research Questions 2 and 4) 56 Previously used methods for studying grammatical accuracy in student writing 56 Categorisation system for this investigation: Grammatical Accuracy Analysis 57 Validation of Grammatical Accuracy Analysis 59 Hypotheses of the present investigation 60

Chapter Six: Results and discussion: Rhetorical Function Analysis 61

Introduction 61 Hypothesis 3 61 The hypothesis 61 Results 62

vi Integration of rhetorical relations at Functional Unit/Simple Sentence/Complex Sentence Level into rhetorical relations at Propositional Segment Level 62 Integration of rhetorical relations at Propositional Segment Level into rhetorical relations at Paragraph Level 62 Integration of rhetorical relations at Paragraph Level into rhetorical relations at Macrostructure Level 63 Discussion 64 The most frequently occurring relations in the high-rated essays 64 The most frequently occurring relations in the low-rated essays 65 Sequences of three rhetorical functions 67 Match with the Argumentative Response Structure 67 Adherence to academic conventions 69 Hypothesis 1 71 The hypothesis 71 Results 72 Integration of rhetorical relations at Functional Unit/Simple Sentence/Complex Sentence Level into rhetorical relations at Propositional Segment Level 72 Integration of rhetorical relations at Propositional Segment Level into rhetorical relations at Paragraph Level 72 Integration of rhetorical relations at Paragraph Level into rhetorical relations at Macrostructure Level 73 Discussion 74 The most frequently occurring relations and macrostructures identified in the Australian and Hong Kong essay groups 74 Language environment for the Australian subjects 75

Chapter Seven: Results and discussion: Grammatical Accuracy 77 Analysis

Introduction 77 Hypothesis 4 77 The hypothesis 77 Results 77 Discussion: Similarities and differences in error types between the high-rated and low-rated essays 78 Hypothesis 2 81 The hypothesis 81 Results 81 Discussion 82 Similarities in error types between the Australian and Hong Kong essays 82 Differences in error types between the Australian and Hong Kong essays 83 Plagiarism in the Hong Kong low-rated essays 85 Further implications of Grammatical Accuracy Analysis 86 Conclusion 86

vii Chapter Eight: Conclusion 89

Synthesis of findings 89 Comparison between Rhetorical Function Analysis and Grammatical Accuracy Analysis 90 Implications for teaching 91 Reactive measures 91 Proactive measures: Towards a model for the teaching of local and global coherence 91

References 95

Appendix 1: Taxonomy of Rhetorical Function Analysis 105

Rhetorical Functions at the Lower Levels 105 Rhetorical Functions at Paragraph Level 109 Rhetorical Functions at Macrostructure Level 111

Appendix 2: HKH4 – Text and Rhetorical Function Analysis 115

Appendix 3: Taxonomy of Grammatical Accuracy Analysis 121

Appendix 4: Frequency of rhetorical functions identified at the lower and upper levels of Australian/Hong Kong high-/low-rated essays 123

Appendix 5: Frequency of grammatical errors identified in Australian/ Hong Kong high-/low-rated essays 131

Appendix 6: Comparison of frequencies and relative rankings of error types: High-rated essays versus low-rated essays/Australian essays versus Hong Kong essays 135

viii Editorial Foreword

Readable writing must not only be clear in what it communicates, but must also satisfy the demands of the discipline. While academics have an intuitive knowledge of what consti- tutes good academic prose, this knowledge is seldom articulated explicitly and much less often brought to the attention of students. Students are usually only instructed in general terms, if at all, as to the requirements of their written work. These instructions usually ex- tend to advice on the required content of macrostructural components such as the follow- ing which might be required in an experimental report: Introduction, Methodology, Results and Discussion. They may require that the arguments be logical and consistent but rarely do they elaborate on how these instructions are to be carried out or advise on what kinds of sentences can carry what sort of information or on the ordering of items of information. (Kalder et al. 1996:1)

A key concept in the understanding of intersentential relations is that of rhetorical func- tion. In this Report, Grace Wong uses Rhetorical Function Analysis as a tool to construct a ‘rhetorical map’ of a text to provide a diagrammatic network of the hierarchies of rela- tionships, and thus show its patterns of coherence and cohesion. With data from the writing of Hong Kong students in Hong Kong and in Australia, she shows that the latter group, after three years away from their home environment, “were able to produce longer essays, with more accurate tense forms and idiomatic ex- pressions …. [T]heir increased exposure to English has helped them to advance along the path of approximation toward the target language”. Wong points out, however, that many fossilised errors remain, and that the Hong Kong students in Australia still have difficul- ties in structuring their writing. As an exploratory step towards a longitudinal study of the interlingual acquisition of rhetorical functions, Wong’s Report provides valuable groundwork for further research of importance in language pedagogy.

Reference

Kaldor, S., Herriman, M. & Rochecouste, J. 1996. The academic teacher and the student writer: Raising textual awareness across disciplines. Different Approaches: Theory and Practice in Higher Education. Proceedings HERDSA Conference 1996. Perth, 8–12 July. [On-line] Available at: www.herdsa.org.au/confs/1996/kaldor.html

ix x Acknowledgements

I would like to extend my thanks to my supervisor, Associate Professor Pam Peters, for her guidance in helping me situate my research not only in the educational context, but also in wider contexts – social, cultural and psychological – and hence make the enquiry more meaningful and relevant than it would have been. May I also take this opportunity to thank all my colleagues and students who have rendered help to me. I am grateful to Professor Gregory James, Director of Language Centre of the Hong Kong University of Science and Technology, for his pastoral care, in being immensely interested in my research and giving me support in many ways. I would also like to thank Dr Man Yu Wong of the Mathematics Department for helping me with the compilation and interpretation of statistics. Special thanks also go to Pionie Foo, Joyce Lee, Li Po-lung, Ellen Tsang and Flavia Lai for validating the two methods used in this research. Thanks are due, too, to my student helpers, Ada Hui, Ace Hung and Penny Yu, for their hard work. The help from all these colleagues and students has made me realise that completing a thesis is not just a matter of “personal growth”, but also a collaborative effort, in which many friends play an active part. Last but not least, I extend my deepest gratitude to my family: my parents, husband and sons. Without their understanding, patience and forbearance, the completion of this work would not have been possible. I am especially indebted to my husband for supporting my study and self-actualisation wholeheartedly throughout all these years. The unfailing support from my family is indeed another collaborative effort I have witnessed in the process of my own “personal growth”.

Grace H. Y. Wong Hong Kong November 2003

xi

xii Synopsis

Many ESL/EFL learners find it difficult to write argumentatively, because unlike L1 stu- dents, who usually only have to attend to the quality and organisation of the arguments, L2 students also have to ensure that their language is grammatically correct and accurate. Since it is difficult for L2 learners to improve their accuracy quickly, the question ex- plored here is whether the use of an effective argumentative structure and/or a favourable language environment can be conducive to successful L2 writing. The objectives were threefold: firstly, to explore how effectively L2 tertiary students build up their universe of argumentative discourse, from the lowest discourse level to the highest; secondly, to examine whether compared with an EFL setting, an ESL environment can bring im- provement in organisation and grammatical accuracy to students’ writing; and finally, to test the correlation between the holistic grades awarded to essays and their quality as de- termined by analytical measures of argumentative coherence and grammatical accuracy. All the subjects of this study were Hong Kong Chinese, first-year EAP students at Macquarie University, Australia, and the Hong Kong University of Science and Tech- nology, thus studying in an ESL and EFL environment respectively. The data set con- sisted of twenty-four essays written on an argumentative topic, six high-rated and six low-rated from each group. Two tools were used to investigate the differences between the high-rated and low- rated essays, and also those between Australian and Hong Kong essays: Rhetorical Func- tion Analysis (RFA) was devised to investigate the hierarchical integration of argument, and Grammatical Accuracy Analysis (GAA) was used to study syntactic accuracy. RFA showed that the high-rated essays displayed effective argumentative structures at different textual levels, with appropriate in-depth development of ideas, which all con- tribute to the development of arguments. The low-rated essays, however, were character- ised by plagiarism and contextually unidentifiable relations. Their macrostructures did not contribute to the scaffolding of arguments, as these are unclear, contradictory, or even lacking. GAA revealed a surprising similarity in the types of error made by the Australian and Hong Kong subjects, suggesting that not a great deal of development had taken place in the Australian subjects’ English after they had left Hong Kong. Yet there were signifi- cant differences in the frequency of some categories of error, showing that the Australian subjects were able to produce more accurate tense forms and idiomatic expressions. There was also less plagiarism in their texts. Though they may not have taken full advan- tage of the ESL environment, their increased exposure to English seemed to have helped them achieve better approximations toward the target language norms. The findings also showed that independent measures of argumentative coherence and grammatical accuracy correlated significantly with each other, and also with teachers’ holistic assessments of essay quality. This supports the view that pragmatic competence and grammatical competence are interrelated. Effective argumentative structure and grammatical accuracy complement each other; either one of them is a necessary but not sufficient condition for writing success. The fact that increased exposure to everyday English could help the Australian subjects advance a little in grammatical accuracy, but not in organisation, supports the hypothesis that it takes longer to develop L2 learners’ writing competence than oral competence. It also suggests that L2 students need to be taught argumentative structures explicitly. Once they master these, they can concentrate on grammatical accuracy, and become better able to produce essays which are organisa- tionally and linguistically sound, to convey their arguments.

xiii xiv Chapter One: Introduction

In recent years, there has been a great deal of interest in second language writing. It is not easy for learners to become proficient in a second language, and among the four lan- guage skills, the writing skill can be said to be the most difficult to acquire. As a ‘pro- duction’ skill, it is intrinsically more demanding than the ‘reception’ skills of reading and listening (Biggs & Watkins 1996:278), and as a more permanent form of communi- cation than speech, it requires lengthy formal training and sustained attention to detail. My interest in the teaching of second language writing stems from the facts that I am an L2 speaker of English on the one hand, and an L2 English teacher on the other. When I started teaching a class of Form Six science students (the equivalent of Year 12 in Aus- tralia) in a secondary school in Hong Kong in the 1980s, one of the worrying facts was that these students could not write in an acceptable manner at all. Their grammar was faulty; their tone was inappropriate; their register was wrong; but above all, they did not know how to construe or organise their ideas. This was because throughout their study at all levels, they had been given notes to memorise and to ‘reproduce’, so as to get high marks in the internal and external examinations. Yet the Advanced Level (A/L) Use of English (UE) Examination is very demanding. There are five sections, two of which deal with writing: Section A is a conventional paper, asking students to express their personal opinions in an argumentative manner toward current events or topics of local relevance, and Section E, entitled ‘Practical Skills for Work and Study’, is an integrated paper which requires the reading of several documents, summarising them and responding to them. The demands of the UE Examination thus go beyond the thinking and organisation abilities of most candidates. However, UE is the all-important subject because even if students should get Grade ‘A’ in all the other subjects, if they fail in the UE Examination, they cannot enter univer- sity, except in very rare circumstances. The onus of helping students pass the UE Examination is thus put on the shoulders of English teachers. If they fail to help their students get a pass, they have to bear the blame from students, parents, other subject teachers and their principals. In 1993, I started teaching at a new university in Hong Kong, the Hong Kong Uni- versity of Science and Technology (HKUST), where undergraduates major either in Sci- ence, Engineering or Business Studies, all taught in English. In the first and second years of its foundation, the university was a genuine English-medium university, with all lec- tures and tutorials conducted in English. But most first-year students had difficulty in following them. From its third year, the size of the student population increased. Tutori- als grew in size, and in many situations became ‘mini-lectures’, which were designed to supplement, or even to repeat part of, the main lectures. These were often given in Can- tonese. Therefore using in tutorials was a ‘win-win’ situation, and this has continued until now. Among the four language skills in English, reading can be said to be the skill most commonly practised among first-year students. However, when the students read, they do not always aim at thorough understanding or application. Sometimes they aim only at passing a test or examination. Continuous assessment has been part of HKUST’s assess- ment policy. This has three implications for teachers and students: first, the lecturers have a heavy load in setting and marking examinations. To find a solution, some lectur- ers make use of multiple-choice questions as part of their assessment tool. Students soon find that they are back in their ‘good old days’ of secondary school, skimming over books to look for information to memorise in order to get high marks in examinations.

1 When the students have such a heavy study load already, they find that they cannot spare time to work hard in improving their language skills. Some of them feel that they do not really need to study English for ‘academic purposes’. Others realise that they need Eng- lish after graduation. However, since life is usually dictated by urgency rather than prior- ity, many students do little to improve their English. Secondary students may lack proficiency in English to tackle the learning, but they still have the motivation to develop it, at least in order to pass the Use of English Exami- nation. However, at university, students may still lack language proficiency, and their motivation to improve may in fact wane. On going to Australia in 1995, I found that some of the ethnic Chinese students, who had come from Hong Kong and studied there for several years, seemed to be truly bilin- gual. The education system in Australia also seems to encourage more independent en- quiry and critical thinking. Students are given the freedom to write in a creative way. All these phenomena seem very encouraging, and it is the expectation of many migrant fami- lies, including Chinese from Hong Kong, that such a language environment will be con- ducive to helping L2 learners to develop their language competence much more than if they finish their schooling in a foreign language environment like that of Hong Kong. However, when I had a chance to teach the ‘Writing Skills’ class organised by the Linguistics Department of Macquarie University and audit their EAP lessons, I soon dis- covered that the English learning situation at tertiary level in Australia was not a homo- geneous as that at the primary and secondary levels. There were, of course, students who had a high proficiency in English; on the other hand, there were some whose English writing ability was surprisingly low, as reflected in the grades they received in their EAP assignments. Their spoken English was acceptable, but they spoke little English outside the classroom. All these students shared a similar background, having come from Hong Kong to study in Australia about three years previously. Confronted with seemingly conflicting pictures at different educational levels, I was motivated to find out whether there was any difference in the level of English proficiency between the Australian students whose EAP lessons I audited, and their Hong Kong counterparts at HKUST. I was particularly interested in investigating their performance in writing, to determine whether there was any difference in the organisation and gram- matical accuracy in the argumentative writing of these two groups.

2 Chapter Two: Language learning in context: Australia and Hong Kong

SECOND LANGUAGE COMPETENCE

There is no blueprint for acquiring competence in a second language. As Gass & Selinker (2001:329) maintain, “[o]ne of the most widely recognized facts about second language learning is that some individuals are more successful in learning a second lan- guage than other individuals.” According to Ellis (1994), such differences are in part so- cially determined and in part explained by psychological factors.

Competence versus performance

A distinction is made by Chomsky (1965) between linguistic competence and per- formance. Competence is the mental representation of linguistic rules which constitute the learner’s internal grammar. Performance, on the other hand, consists of the use of this grammar in the comprehension and production of the language. For Ellis (1994:13), the main goal of second language acquisition research is “to characterize learners’ underlying knowledge of the L2, i.e. to describe and explain their competence. However, learners’ mental knowledge is not open to direct inspection; it can only be inferred by examining samples of their performance.” This is the approach taken here, in the investigation of the actual written output of tertiary students. However, al- though researchers can make inferences about the underlying linguistic knowledge a learner possesses, these are largely untestable. The focus of this investigation is therefore on the writing proficiency level of learners as displayed in their product, in terms of their text organisational strategies and grammatical accuracy. In this work, the term ‘compe- tence’ is used interchangeably with the term ‘proficiency’.

DEFINING LANGUAGE PROFICIENCY

Just what constitutes language proficiency is the centre of debate. Oller (1976) suggests that proficiency is unitary and calls the essential character of proficiency Expectancy Grammar, which “sequentially orders linguistic elements in time and in relation to extral- inguistic elements in meaningful ways” (Oller 1979:34). Cummins (1979:198), however, distinguishes two types of proficiency: Cognitive/Academic Language Proficiency (CALP) and Basic Interpersonal Communication Skills (BICS). He observes that it takes language minority students a much longer time to attain grade- or age-appropriate profi- ciency levels in academic skills than it does for them to acquire face-to-face communica- tion skills in English, as well as sociolinguistic appropriateness (Cummins 1983/1986: 152). Canale & Swain (1980) extended Cummins’ model, by proposing that apart from sociolinguistic competence, communicative competence also consists of grammatical and strategic competence. Grammatical competence includes knowledge of pronuncia- tion, spelling, vocabulary, as well as rules of word or sentence formation, while strategic competence involves the use of a repertoire of ‘coping’ strategies to avoid communica- tion breakdown. In a revision, Canale (1983) added a fourth element, discourse compe- tence, to the 1980 model. This includes the mastery of cohesion (use of appropriate con- junctions) and coherence (logical sequencing and effective organisation).

3 The components of language proficiency have therefore been expressed in increas- ingly specific ways, for example, from being viewed as a single concept (Oller) to a two- fold concept (Cummins), a threefold concept (Canale & Swain) and a fourfold concept (Canale). In 1994, Ellis seemed to swing back to the dichotomy, by remarking that com- municative competence entails both linguistic (or grammatical) competence and prag- matic competence (Ellis 1994:13). Yet Ellis’ definition of ‘pragmatic competence’ actu- ally encompasses the knowledge of using language to construct discourse and to perform speech acts in socially appropriate ways. It seems that the term has been used as an um- brella term to incorporate sociolinguistic competence, strategic competence and even discourse competence. There is thus no apparent consensus as to what communicative competence exactly entails: “a definitive analysis of communicative competence is just as elusive as was lan- guage proficiency” (Larsen-Freeman & Long 1991:38), although it is generally agreed that communicative competence consists of the knowledge users of a language have in- ternalised to enable them to understand and produce messages in that language. The discussion of language proficiency which is most relevant to the present work derives from a reflection of Cummins (1983/1986:152), that a division of language profi- ciency into two categories, BICS and CALP, is an oversimplification. He proposes a new framework which interprets language proficiency along two continua (ibid.:153; Figure 1 refers).

Cognitively undemanding

A C

Context- Context- embedded reduced

B D

Cognitively demanding

Figure 1: Range of contextual support and degree of cognitive involvement in communicative activities.

The horizontal continuum expresses how much contextual support is available in the communication of meaning. Context-embedded communication derives from personal interaction in a shared reality, such as the everyday world outside the classroom, in which the participants negotiate meaning actively. On the other hand, in context-reduced communication, such as the interaction which takes place in the classroom, the shared reality cannot be assumed. In order to minimise the risk of misinterpretation, the mes- sages must be put across precisely and explicitly. Examples of communicative events along this continuum, from left (context-embedded) to right (context-reduced), include engaging in a discussion, writing a personal letter, and writing an academic essay. The vertical continuum addresses the developmental aspects of communicative profi- ciency in terms of the level of active cognitive involvement. The upper end indicates

4 communicative activities in which the linguistic tools are proficiently mastered. Little cognitive involvement is therefore required. At the lower end are communicative tasks in which the tools are not yet automatised, and thus require active cognitive involvement. Examples of tasks which necessitate active cognitive involvement include writing an es- say, or persuading another person to accept a viewpoint. Since this study is about argumentative writing at tertiary level, it falls within quad- rant D in Cummins’ model, that is, a communicative event which takes place in a con- text-reduced situation, and is cognitively demanding. Cummins (1983/1986:156) re- marks that “[t]he more context-reduced a particular task … the longer it will take L2 learners to achieve age-appropriate performance”. Coupled with the fact that academic writing is cognitively demanding, it would take L2 learners a longer time to acquire this skill than it does for them to master face-to-face oral skills, according to Cummins’ model. In this study, I examine whether this hypothesis is valid or not in the development of the writing skills of the Chinese students in Australia and Hong Kong. Another point which is worth further examination here is the relationship between grammatical competence and communicative competence. For Ellis (1994:13), commu- nicative competence entails both grammatical competence and pragmatic competence, and the level of a learner’s grammatical competence is closely related to that of his/her pragmatic competence. This viewpoint is different from the findings of Swain (1985/1986). Her subjects were children whose first language was English, and who were studying Grade 6 in a French immersion programme. Both oral and literacy based tasks were assigned for the subjects to complete (cf. Table 1):

Table 1: Types of tasks for testing different linguistic traits, from Swain (1985/1986:118). Traits Grammar Discourse Sociolinguistic Structured interview Film retelling and - requests Oral production argumentation - suggestions - complaints Multiple choice 45 items 29 items 28 items 2 narratives 2 notes Written production directive 2 letters

The results showed that although the immersion students had learnt French for seven years, their grammatical competence was not equivalent to that of the native speakers. Only their levels of discourse and sociolinguistic competence were similar to those of the latter. Swain (ibid.:135) concludes that discourse and sociolinguistic competence do not rely heavily on grammar for their realisation. The present investigation is fundamentally an enquiry into a special type of prag- matic competence, that is, the writing competence (CALP) of L2 learners. An analysis of the subjects’ grammatical competence is incorporated as a means of triangulation for the main focus of the study, how tertiary students organise their arguments in extended es- says. The design allows us to investigate the relative validity of both Ellis’ viewpoint and Swain’s findings in the context of learning academic writing at tertiary level, that is, whether academic writing competence and grammatical competence are interrelated.

5 CONTEXTS OF DEVELOPING SECOND LANGUAGE COMPETENCE

The place of SLC and the socio-cultural factors it presents

Social and cultural factors influence the choice of languages to be learned, as well as the differences in the level of language proficiency learners achieve (Berns 1990:3). A number of social factors, such as gender and age, may affect the acquisition of a second language. Various studies (e.g. Burstall 1975; Slavoff & Johnson 1995) have found that females are better language learners than males. In Boyle’s (1987) study of 490 ethnic Chinese university students in Hong Kong (257 males and 233 females), the female subjects achieved significantly higher means in ten tests of general L2 proficiency. The main reason might be that females have more positive attitudes to learning a second language than males (cf. Gardner & Lambert 1972; Spolsky 1989). As for age differences, it is commonly believed that children are more likely to attain native-like proficiency in an L2 than are teenagers and adults. According to the ‘Critical Period Hypothesis’ (Gass & Selinker 2001:335), there is an age-related point, which is usually set at puberty, beyond which it generally becomes difficult or even impossible to gain full mastery of a second language. Apart from the age of onset of learning, the age of arrival in the country of the target language is also an important factor. In a study reported in 1989, Johnson & Newport investigated learners’ acquisition of syntax based on different ages of arrival in the L2 country, ranging from 3 to 39. Their finding was that learners’ proficiency was linearly related to the age of arrival only up to puberty. Another important factor in learning an L2 is the length of stay in the country of the target language. Slavoff & Johnson (1995) examined the acquisition of English by differ- ent groups of Chinese, Japanese, Korean and Vietnamese children, who arrived in the United States between the ages of 7 and 12. They found that length of stay, as opposed to age of arrival, was an important variable in predicting knowledge of English syntax. Similar findings were obtained by Bialystok (1997) in two studies, one of which investi- gated the learning of English syntax by Chinese speakers. However, for a group of for- eign workers in Germany, the length of stay in the new country functioned as a major factor in their L2 acquisition only during the first two years. After that, the acquisition was overridden by other social factors such as

• age at time of immigration, • professional training in the country of origin, • number of years of formal education, • contact with Germans during leisure time, and • contact with Germans at work. (Heidelberger Forschungsprojekt 1978)

While the first three factors listed here are social, the last two are cultural, which repre- sent a different approach to understanding language learning. Cultural factors include learner attitudes toward

• the target language, • the target-language speakers, • the target-language culture, • the social values of learning the second/foreign language, • the particular uses of the target language, • themselves as members of their own culture. (Ellis 1994:198)

6 Such attitudes may influence how much effort a learner makes in learning a sec- ond/foreign language, and this in turn has an impact on the level of language competence he/she develops. Several of the factors listed above are included in Schumann’s (1978a, 1978b) Acculturation Model, which predicts that if learners acculturate, they will learn, and vice versa. The first social variable in the Acculturation Model is the extent to which a group is or is not dominant over another. There may be two completely different situations: the first in which the L2 group is dominant (as in a colonial context), and the second in which the L1 group is dominant (as one in which emigrants find themselves). It seems that in the former case, learning is less likely to take place. The second social variable in the Acculturation Model is the extent to which a group integrates. If one fails to acculturate, one cannot succeed in learning a second language. For Ellis (1994:207), the extent to which a group integrates, or the degree of acculturation, is related to “the intersection of [their] views about their own ethnic identity and those about the target-language culture”. The combinations of positive and/or negative attitudes predict the degree of success in mastering the L2 (cf. Table 2):

Table 2: Attitudes and L2 learning, from Ellis (1994:208). Attitudes toward Native culture Target culture Additive bilingualism + + Subtractive bilingualism - + Semilingualism - - Monolingualism + -

Key: + = positive attitudes; - = negative attitudes

The ideal outcome – additive bilingualism – is likely to occur when learners have a posi- tive view of their own ethnic identity as well as the target-language culture. They thus maintain their L1, adding the L2 to their linguistic repertoire, and become “balanced bi- linguals”. The opposite is true in subtractive bilingualism, in which case the learners have a low respect for their ethnic identity and are eager to assimilate in the target- language culture. In the end they replace their L1 with L2. Semilingualism occurs when learners fail to develop full proficiency in either their L1 or their L2, as they have nega- tive attitudes towards both their own culture and the target-language culture. Finally, monolingualism, or failure to acquire L2, is generally associated with a strong ethnic identity and negative attitudes toward the target-language culture.

Table 3: HKCEE grade combination statistics for Chinese Language and English Language (Syllabus B) in all day-school candidates (1993), n = 64,924. From So (1998:169). English Language

Grades A/B/C Grades D/E Grade F

Chinese Language Grade A/B/C 4,007 6.2% 8,093 12.5% 347 0.5% Grade D/E 1,310 2.0% 20,269 31.2% 7,058 10.9% Grade F 33 0.1% 4,462 6.9% 19,345 29.8%

The phenomenon of semilingualism can be seen in Hong Kong, except that it may be more apt to describe the attitude of local citizens toward the Chinese and English cul- tures as one of indifference, rather than that of negativity. Before 1997, when Hong

7 Kong was still ruled by the British, many local people were unable to identify themselves with either the British or the Chinese authority. The end result was that many of them, including students, were proficient neither in English nor Chinese. This is illustrated by the norm-referenced results in English and Chinese Language achieved by Hong Kong secondary students in 1993 and 1995 respectively in the two major local public examina- tions, the Hong Kong Certificate of Education Examination (HKCEE) and the Hong Kong Advanced Level Examination (HKAL) (cf. Tables 3 and 4). From Table 4 it can be seen that in 1993, of the 64,924 students who took HKCEE, 19,345, or nearly 30%, of the total number of candidates, attained a ‘F’ grade in both English and Chinese Language. Only a very small percentage (4,007, or 6.2%) obtained Grade A/B/C in the same examination. The situation improved a little when the same batch of students took the HKAL two years later, in 1995 (cf. Table 4).

Table 4: HKAL Examination grade combination statistics for Chinese Language & Culture and Use of English in all day-school candidates (1995), n = 22,209. From So (1998:169). English Language

Grades A/B/C Grades D/E Grade F

Chinese Language Grade A/B/C 1,888 7.88% 2,425 10.12% 122 0.51% Grade D/E 1,660 6.92% 11,721 48.88% 2,206 9.2% Grade F 36 0.16% 1,371 5.72% 780 3.25%

Of the 22,209 candidates, 780, or 3.25%, were awarded Grade F, while 1,888, or 7.88%, attained Grade C or above. However, in both 1993 and 1995, most candidates did not perform well, since 31.22% and 48.88% of the total candidature obtained only a passing grade (D or E) in both English and Chinese in HKCEE and HKAL respectively. The fig- ures in the tables thus show that “balanced bilingualism is not one of [the students’] characteristics” (So 1998:168). After the sovereignty of Hong Kong was returned to China in 1997, the identity crisis has become less severe for the local people. Yet many of them are still not keen on as- similating the Chinese culture, not to speak of the English culture, now that the British influence has faded. Their writing and oral skills in both English and Chinese remain poor, as lamented by the erstwhile head of Hong Kong’s Education Commission, Rosanna Wong Yick-ming, in 2001 (as reported by Gary Cheung in the South China Morning Post, 3rd May, 2001). But it is possible that immigration to, or studying in, another country may change learners’ attitudes towards their L1 and/or L2 learning, and hence their competence. Some of them may be eager to assimilate into the target-language culture, and hence de- velop subtractive bilingualism. Others may acquire a new perspective on their own eth- nic identity, and in turn a more positive attitude toward their culture. If they become in- terested in the L2 culture at the same time, they may develop additive bilingualism. However, if they become alienated from the target-language culture, monolingualism may persist. Of course, for some learners, studying abroad makes no impact on their atti- tude. They may thus retain their semilingualism. In this study, we will investigate which type of bilingualism our Australia-based subjects attained after they had left Hong Kong.

8 Personal/psychological factors

Besides sociological factors, language learning is also affected by a multitude of per- sonal/psychological factors. Among those listed by Gass & Selinker (2001:329–71) are Aptitude, Anxiety, Locus of control, Personality, Learning strategies, and Motivation. Of these, motivation is said to be “the second strongest predictor of success, trailing only aptitude” (ibid.:349). According to Gardner (1985:50), motivation involves four aspects: “a goal, effortful behaviour, a desire to attain the goal and favourable attitudes toward the activity in question.” Ellis (1994:516) also supports this view: “It is the need to get meanings across and the pleasure experienced when this is achieved that provides the motivation to learn an L2.” Motivation has already been mentioned as playing an important role in sec- ond/foreign language learning for Hong Kong students. The interplay between this factor and acculturation, as well as other psychological and socio-cultural factors affecting a learner’s attitudes toward learning a second/foreign language will be further examined, to determine their relative impact on students in Australia and/or Hong Kong.

ESL VERSUS EFL LANGUAGE LEARNING

Many scholars make a distinction between (English as a Second Language) and EFL (English as a Foreign Language). Two such, Crystal (1995) and McArthur (1996), con- nect the distinction with the official status of a language. Hence, in his section, “A Short Glossary of EL Terms”, Crystal (op. cit.:108) defines EFL as

English seen in the context of countries where it is not the mother tongue and has no spe- cial status, such as Japan, France, Egypt and Brazil and ESL as

English in countries where it holds special status as a medium of communication ... The term has also been applied to the English of immigrants and other foreigners who live within a country where English is the first language.

Others, such as Hartmann & James (1998:50), connect the distinction with whether the language is dominant or not. They define ESL as

those varieties of English which are used by speakers for whom it is not the NATIVE LAN- GUAGE, usually in a country where it is the endoglossic, or dominant language, or in coun- tries where it has an acknowledged function and EFL as

those varieties of English which are used by learners for whom it is not the NATIVE LAN- GUAGE, usually outside a country where it is the dominant language. [caps sic]

Thus the distinction made between ESL and EFL provides contrasting contexts for learn- ing, in which various socio-cultural and psychological factors, discussed earlier, are in- volved. Hartmann & James’ definitions for ESL and EFL will be used as the basis of our understanding of these two terms in this study.

9 LEARNING ENGLISH IN HONG KONG

The definition of ESL provided by Crystal does not mention Hong Kong, but he cate- goryised Hong Kong under “ESL Countries” elsewhere in his book. This is not difficult to understand, as Hong Kong seems to fit the first part of his definition for ESL as a place where English holds “special status as a medium of communication”. This view is shared by a number of researchers. However, Falvey (1998) argues against this view for Hong Kong, and in so doing, challenges the definitions of both Crystal (1995) and Ellis (1997). His distinction between ESL and EFL is as follows:

EFL [is] used to refer to language which is learned primarily in the classroom with little assistance from the language environment. ESL, in contrast, [is] used to describe language which is used by the majority of people in a country, or territory as their lingua franca, as the language of the church, government, the press and judiciary. (Falvey op. cit.:75)

He goes on to argue that after the return of sovereignty of Hong Kong to China in 1997, English was no longer used as a lingua franca in domains such as the judiciary in Hong Kong. Falvey’s arguments are echoed by So (1998:168), who argues that “[b]y demo- graphic background Hong Kong is a largely monolingual society; it is only through a twist of history and by design that a degree of individual bilingualism is in evidence in our society”. Not only is the degree of individual bilingualism limited, but also there is a relatively high degree of social distance between the Chinese community and the mostly English-speaking expatriate community in Hong Kong (Luke & Richards 1982:53). This phenomenon has not changed since the return of the sovereignty of Hong Kong to China in 1997, the year in which the Hong Kong data for this study was collected (cf. Pen- nington 1998; Li et al. 2000). So (1998:161) also made this point very clear: “Indeed, even today, interaction between these two communities has remained minimal: Hong Kong has always been a city with two identifiable and separate communities.” Falvey (1998:76) emphasised that very few people had in fact achieved genuine bi- lingualism in Hong Kong. Although until 1997 more than 90% of secondary schools in Hong Kong claimed themselves to be “English-medium”, this only meant that textbooks and written examinations were in English. Most of the other components in teaching/ learning were conducted in Cantonese, or in a mixed code (by adding some English words to Cantonese), which is thought to be the worst of all modes of instruction (Gib- bons 1987) and is detrimental to students’ English attainment. The massive expansion of tertiary education since the 1980s has further aggravated the situation. In the past, only between 2% and 5% of the population could enter univer- sity. From 1989, this proportion rose to as many as 18%. These young people are now drawn from a much larger catchment area, including the poorer districts, where Canton- ese is almost used exclusively (Falvey 1998:80). The expansion has tremendous implica- tions for the use of English at the university. One year after the expansion took place, Bruce (1990:19) remarked that academic staff had to simplify their lectures in face of students’ language problems. In 1998, Falvey observed that Chinese was used as the me- dium of instruction in many undergraduate classes (op. cit.:79–80). A telephone survey conducted in 2001, in which 547 young people aged 15 to 29 in Hong Kong were ques- tioned, reported that 30% of the interviewees found that English was inherently difficult to learn. Up to 20% of them even expressed fear of learning this language. Most respon- dents indicated that they limited their use of English to the classroom; one-third claimed to have no opportunity to speak English (Schwartz & Li 2001). This discussion illustrates two points. First, the degree of multilingualism is small in Hong Kong. This logically leads to the second point: “As the population of Hong Kong

10 are almost all Cantonese speakers, English is used so little that few students have ac- quired the level of fluency in the language which qualifies it as a second language” (Kember & Gow 1990:361). English is taught as a foreign language rather than a second language. This is the basis for regarding Hong Kong as an EFL rather than ESL envi- ronment for learning English, pace Crystal and others mentioned earlier.

LEARNING ENGLISH AS AN INTERNATIONAL STUDENT IN AUSTRALIA

The status of learning English in Australia does seem to fit into the latter part of Crys- tal’s (1995:108) definition for ESL in that it is “the English of immigrants and other for- eigners who live within a country where English is the first language”. At first sight, this looks less controversial than is the case of Hong Kong. However, the situation in Austra- lian universities is complicated.

Background of international students in Australian universities

Australian universities provide education to many international students, who are now “an established component of the student body” in these institutes (Brooks & Adams 2001:1). Macquarie University, where some of the data for this research were collected, is no exception (cf. Table 5):

Table 5: International students as a percentage of enrolments in Macquarie University, from Macquarie University (1999); Macquarie University (2000). International students as a percentage International students as a percentage of total enrolments of undergraduate enrolments 1998 1999 1998 1999 9.4% 11.4% 4.4% 6.3%

Most of these international students come from Southeast Asian countries, who represent as many as 70% of the total number of international students in all programmes, and even up to 80% of the total number of international undergraduate students (Macquarie University 1999, Macquarie University 2000). These students face many difficulties, such as undergoing acculturation, as well as facing challenges brought by changes in study methods (Volet & Renshaw 1996). But their poor English proficiency (Nixon 1996) is the most fundamental of all.

Problems faced by international students

English proficiency

In recent years, many university lecturers in Australia have expressed concern about the low English standard of international students. Most Australian universities require a level of 6.0 in all the four language skills of the IELTS (International English Language Testing System). But Brooks & Adams (2001:4) believe that this requirement may be “problematic”, as in the IELTS scale, 6.5 is the “lower score for less linguistically de- manding courses”. Coley (1999:13) also remarks that “Australian universities are pitch- ing their IELTS entry requirements at language ability levels which are at best only ‘probably acceptable’.” He even draws the conclusion that “it is indeed possible for

11 NESB [Non-English speaking background] students to be at an Australian university without being able to speak, write, read or understand English at the required level” (ibid.:15). English language proficiency has a great bearing on the international students’ aca- demic work. Reid et al. (1998) argue that problems experienced by international students may have more to do with levels of English language proficiency than with styles of teaching and learning. University statistics indicate that when English language profi- ciency is not an issue (for example, in courses such as Accounting), international stu- dents tend to obtain better results than do local students (Volet & Renshaw 1996:217). Volet and Renshaw’s study is complemented by that of Brooks & Adams (2001), who investigated whether any relationship could be established between students’ fre- quency of speaking English and their academic performance in a first-year Business course. The first set of data, “frequency of speaking English”, was collected from a con- venience sample of 144 students, which represented 51% of the class cohort. The sub- jects were asked to respond to three questions:

1. Are you an overseas student or not overseas student? 2. Are you male or female? 3. How often do you speak English?

For the last question, students had four response options: “All the time”, “Most of the time”, “Some of the time” and “As little as possible”. To Brooks & Adams (op. cit.:6), this question served as a measure of the respondents’ ATE (Attitude toward English). The responses were converted to a four-point Likert-type scale, with “All the time” = 4, and “As little as possible” = 1. The second set of data, “academic performance”, was taken from the final results of the 285 students enrolled on the course, which were meas- ured as a mark out of 100, and linked to whether they were international or local students. Finally, statistical tests were used to see if there were differences between international and local students in terms of their ATE, as well as their academic performance. The results from the first set of data, “frequency of speaking English”, show that the international students speak English only occasionally (cf. Table 6):

Table 6: ATE response frequency of international students, from Brooks & Adams (2001:7). ATE (Attitude Toward English) Gender of N As little as Some of Most of the All of the international students possible the time time time Male 14 0% 43% 36% 21% Female 18 22% 50% 28% 0%

From the table it can be seen that 43% of the male international students indicated that they spoke English “some of the time”, and another 36% of them said that they spoke English “most of the time”. Only 21% of them reported that they spoke English “all of the time”. The situation of the female respondents was even worse. As many as 22% of them indicated that they spoke English “as little as possible”, and not a single one of them said that they spoke the language “all of the time”. This does not support earlier findings that compared to their male counterparts, females have more positive attitudes learning an L2 and are thus better language learners. According to Brooks and Adams, the correlation between “frequency of speaking English” and “academic performance” was confirmed in the contrasting results for local

12 and international students (see Table 7 for comparative statistics on their ATE and course results graded out of 100 marks):

Table 7: International and local students compared in terms of ATE and marks, from Brooks & Adams (2001:9). ATE Mark Category of students Mean N Std dev Mean N Std dev International 2.8 14 0.8018 53.0 27 11.2817 Male Local 3.8 53 0.4945 61.4 98 6.3059 International 2.1 18 0.7254 53.6 29 6.1930 Female Local 3.7 59 0.5587 61.6 131 5.9732

Key: Range of ATE scores: 1–4 1 = As little as possible 2 = Some of the time 3 = Most of the time 4 = All the time

Statistical correlations were however not shown in Brooks & Adams’ table, so no causal relationship can be established. The relatively small number of international students ef- fectively prevents tests of significance being carried out. There were also problems in Brooks and Adams’ third survey question: while the first three options given to students (“all the time”, “most of the time” and “some of the time”) refer to frequency, the last option, “as little as possible”, is a comment on attitude. The grounds on which Brooks & Adams interpreted subjects’ “frequency of speaking English” as “Attitude toward Eng- lish” (ATE) are also unclear. Despite these issues, the research raises the question as to what makes these international students so reluctant to speak in English. This in turn leads to another legitimate question: What sort of relationship do these students have with the L1 English-speaking population in Australia?

Relationships

International students have great difficulty in establishing friendships with L1 speakers. This was pointed out as early as 1984 by Bradley & Bradley in the report entitled Prob- lems of Asian students in Australia:

On the issue of friendship with Australians, nearly every overseas student who had been in Australia for any length of time felt unable to form close friendships with Australians, es- pecially Anglo-Australians. (p. 261)

Few changes in this domain seemed to have taken place in the 1980s and 1990s. A large- scale survey of students’ study-related and personal experiences involving 1,250 under- graduates (436 international and 814 local students) enrolled in three universities in South Australia (Adelaide, Flinders, and the University of South Australia) found “mix- ing with Australians” ranked the ninth most often cited problem (Mullins et al. 1995:218). While two-thirds of the international students surveyed indicated that it was a minor or serious problem, another 15% of them rated it as a serious problem. This find- ing is relevant to the present investigation, as a sizeable number of the international stu- dent respondents of Mullins et al.’s study came from Hong Kong. Some international students put the blame for their difficulty of establishing friend- ship on Australians alone, while others felt that both groups had shared responsibility:

13 Some Australians are ‘snobby’ or ‘just don’t like Asians’. … Some attribute these difficul- ties to a communication problem and shyness on both sides, as well as strangeness caused by cultural differences. … It seems clear that the expectations and obligations of friendship are different in Southeast Asia. (Bradley & Bradley 1984:261–2)

Another reason why some of the international students, notably those coming from Hong Kong, are not keen to establish friendships with the local people is because they intend to return to their own countries once they have obtained their qualifications (Ballard & Clanchy 1991:21. Barker et al. (1991:83) amplify the kinds of context in which Asian students have the greatest difficulty: “taking the initiative in conversations, dealing with angry people, and interacting with the opposite sex”. This is echoed in Hellmundt et al.’s (1998:333) ob- servation that many international students in Australia are “isolated, non-participating and under-performing”. Although the last part of their comment is not true for all (cf. Volet & Renshaw 1996), these students are often isolated from fellow Australian stu- dents, a problem reported by many researchers.

Implications for the present study

According to the literature we have examined, learners in a ‘majority language context’ (such as Chinese students at Australian universities) vary enormously in the extent to which they approximate to the target language proficiency, and ‘classroom learners’ (such as the Hong Kong subjects) often fail to develop much functional language ability. While there is still opportunity for the former to attain native-speaker-like proficiency, the chance for the latter to do so is very slim. However, the Australia-based students do not necessarily take advantage of the language environment. In theory, they should be able to speak, as well as to write, better than their Hong Kong counterparts if immersion takes place. Yet our survey of the Australian research has presented us with a not-so- positive picture of the international students there. The students’ English proficiency is often less than satisfactory, and their tendency to mix with students from their own places of origin may well jeopardise their chances of improving spoken English as well as written English. The benefits of the ESL environment cannot be taken for granted. Faced with these conflicting views of the ESL environment, the former based on ‘theory’ and the latter seemingly on ‘reality’, this study aims to investigate the writing performance of Chinese students in Australia, as compared with their Hong Kong coun- terparts. The questions to be addressed are:

1. whether the argumentative essays written by these Australian students are better than those written by their Hong Kong counterparts in terms of organi- sation; and if this is not so, what the factors accounting for this are;

2. whether the argumentative essays written by these Australian students are better than those written by their Hong Kong counterparts in terms of the de- gree of grammatical accuracy; and if this is not so, what the factors ac- counting for this are.

From these questions stem two of the hypotheses of the present investigation. Together with two others, these will be presented and developed in Chapter 5.

14 Chapter Three: Teaching and assessing academic writing

STUDENTS’ ACADEMIC WRITING NEEDS

At university, students are required to submit a number of written assignments, including essays, reports, projects, case studies and research papers. Such tasks make many de- mands on ESL students, because they require a high level of argumentative skills and writing proficiency, as well as an understanding of the academic writing conventions. This is vividly expressed by a subject in Briguglio’s 1998 study:

[N]ot just [assistance with] writing, but expressing myself in intellectually mature lan- guage, in academic language. Because sometimes, that’s what I think is a bit difficult for non English-speaking background people, to make a distinction between, for example, academic language, non-academic language, and slang. (p. 6)

This student’s remark echoes the findings of a large-scale study of international and local students conducted in Australia (Mullins et al. 1995), which revealed that ‘Writing as- signments/reports’ was a serious problem for only six per cent of the local Australian respondents, and thus ranked as low as fifteenth in their list of ‘serious problems’. In contrast, 17% of the international students considered academic writing a serious prob- lem. It ranked eighth in their list. This problem was also a persistent one; some students who had attended university for over a year still reported that they found it difficult to write assignments or reports (ibid.:213). Tertiary students in Hong Kong also report that they find academic writing difficult. In an investigation I conducted in the Spring semester of 2001, all the first-year Engi- neering students who attended the ‘English for Academic Purposes’ course at HKUST and Technology were required to write a portfolio to evaluate their own writing skills. Among the 108 students I taught, 99% indicated that they found they lacked the writing skills to cope with the requirements of their university study (Wong 2001a). They identi- fied three areas which needed most improvement: organisation, grammar and vocabulary. The first of these is itemised by Jordan (1997:7). He notes that successful academic writ- ing involves a set of study skills including:

• planning • writing drafts, revising • summarising, paraphrasing and synthesising • continuous writing in an academic style • organising writing appropriately • using quotations, footnotes, bibliography, and • finding and analysing evidence; using data appropriately.

Among these skills, the ones most important for successful academic writing are organi- sation, referencing (e.g. the use of quotations from secondary sources), as well as appro- priate use and analysis of data.

15 RESEARCH ON STUDENTS’ ACADEMIC WRITING

Contrasts between good writing and poor writing

Research into good and poor academic writing shows that they differ in two ways: firstly, they differ in their development of ideas, and secondly, in the degree to which they con- form to the conventional global and local structures of argumentative text. Witte (1983) found that good essays have more parallel progression than poor essays, that is, proficient writers elaborate on a few crucial ideas only. Poor writers, on the other hand, use more “sequential progression”. They keep on introducing new topics, includ- ing irrelevant ones. Witte’s finding was echoed by Lindeberg’s (1985a:339), who dis- covered that in-depth idea movement is associated with good writing, and vice versa:

Poor writers … keep to a very limited range of levels of generality, or … make a long- winded dive into depths of detail. … [G]ood writers tend to have fairly orderly patterns of neat descent, whereas poor writers reveal ragged, inconsistent patterns.

The suggestion that in-depth idea development contributes to writing success was echoed by the results of an investigation carried out in Venezuela (Villegas 1998, 1999), in which a group of first-year college students were taught the macrostructure of argumen- tative text. In the post-test, their use of organisational strategies (such as the elaboration strategy) had helped them write more coherent essays. Other kinds of research have focused on the degree to which good and poor academic essays conform to the conventional global and local structures of argumentative text. One example of this type of study is Connor (1987). It was found that the high-rated es- says contained more of these conventional argumentative structures, such as the ‘Prob- lem-Solution’ macrostructure and the local structure ‘Claim + Justification + Induction’.

Rhetorical functions identified in students’ academic writing

Recent research into discourse structure has focused on rhetorical functions in texts, that is, the semantic relations between sentences and propositions, or larger discourse units. Through them the arguments are developed continuously in better texts. By the same to- ken, the sequence of rhetorical functions in poor writing is problematic, disrupting the flow of the discourse. These findings are most relevant to the current investigation, as the methodology of studying organisation, one of our focuses, involves the examination of rhetorical functions (cf. Chapter 4).

Recurrent rhetorical functions in good essays

The research into the most prevalent rhetorical functions identified in academic writing has divergent findings. While Lindeberg’s (1988, 1994) study found that ‘Evaluate’ was the most common rhetorical function in academic writing, Stuart-Smith & Busbridge (1996) found that ‘Elaboration’ and ‘Evidence’ were the most frequent relations in their data set. On the other hand, Lindeberg’s second finding that the ‘Assert-Specify- Evaluate-Cause’ pattern recurs in good academic writing was very similar to that of Connor (1987), who discovered that ‘Claim + Justification + Evaluation’ was a common pattern in student essays.

16 At the global level, Stuart-Smith & Busbridge discovered that good essays display the same pattern, “that is, CENTRAL NUCLEUS in the INTRODUCTION with the BODY of the essay relating to the CENTRAL NUCLEUS or the whole INTRODUCTION … by the relations of ELABORATION or EVIDENCE” (op. cit.:26; caps sic). Poor essays, however, display a smaller frequency of these patterns at both local and global levels.

Problematic rhetorical functions

Lindeberg (1988, 1994) discovered that poor essays contain a higher frequency of “non- functional units”, that is, units with no identifiable rhetorical role in the context. She noted various underlying causes, such as problems of logic, cohesion, vocabulary, and/or syntax. Stuart-Smith & Busbridge (1996), on the other hand, looked at unsuccessful rela- tions from the perspective of communicative effects, where “the student writer intended a particular relation yet the corresponding effect was not achieved” (ibid.:28). In their data set, only one ‘Unsuccessful Relation’ was discovered in the good essays, but there were ten such in the poor essays. Kaldor et al. (1998:51) also identified problems in rhe- torical relations in student writing:

Often signalling will be inconsistent with the information contained in the sentence, or the reader may not be able to establish relationships with earlier sentences because they may not be there or may be too distant.

The problematic categories of rhetorical functions identified by all these researchers thus often signal a break in coherence. The presence of such problematic, or unsound, rhetori- cal relations is not conducive to further integration of arguments.

Academic conventions

Apart from the discoursal challenge in organising an argumentative text, L2 writers often have problems in the academic conventions of university writing. Jordan (1997) points out that the convention of being critical toward published work, as well as referencing conventions, are areas in which students experience particular difficulties when writing academic essays.

Referencing

In many countries in Asia, including China, there has historically been a great respect for the printed word, as well as for authoritative figures. Biggs & Watkins (1996:278), for example, point out that acknowledging sources is not always required or expected by Chinese teachers. However, they quickly cite yet another reason for plagiarism, or failure to acknowledge the source :

[W]hat appears to be plagiarism often occurs as a result of writing in a second language. Students who want to make a point particularly clearly see paraphrasing the source as a strange thing to do when the source itself makes the point better than they ever could re- word it in an imperfectly mastered language. (ibid.:278–9)

17 In other words, plagiarism may be the result of students’ diffidence about their own lan- guage proficiency. When they go to study in the English-speaking world in places such as Australia, North America or Britain, they face considerable difficulty, as plagiarism is considered a serious ‘academic crime’. Rather, the ability to provide citations and to ref- erence one’s discussion is crucial in academic writing, since it establishes one’s aca- demic credentials (Jordan 1997:115).

Critical attitudes toward published work

Chinese students’ reverence for published work has also brought to some of them another difficulty in writing academic discourse. The educational traditions in China and those of the Western world are very different:

College teachers in Western institutions assume that their duty is to develop their students as independent learners … . The traditions of [Chinese] scholarship attest to knowledge as wisdom … there is no scope for critical questioning and analysis … (Ballard & Clanchy 1991:24)

In writing academic essays, many Chinese students merely include citations from books or articles without critiquing or commenting on them (Cortazzi 1990), and thus often fail to fulfil the analytical task required of them. Richards & Skelton (1991:40) draw the fol- lowing conclusion: “Overseas students evaluate less, and evaluate less critically.” This remark has been echoed by those who have studied the attitudes of Asian learners in Australia (cf. Ballard & Clanchy 1991). Smith et al. (1998:273) believe that such an ap- proach reflects the students’ “anxious reaction to the different Western teaching envi- ronment”. One further reason which has not been suggested above is a learner’s poor language proficiency. Their non-critical approach may reflect a lack of language rather than thought.

Cultural differences in discourse patterns

Some researchers link the phenomenon of the non-critical attitude of the Asian students with the global structure of the argumentative essays these students produce. For exam- ple, Kaplan (1966:4–8, 10) suggested that the Anglo-Saxon pattern, often termed ‘linear’, places great value on the writer taking a stance, analysing evidence and finally arriving at an evaluation. The Oriental pattern, on the other hand, is described as ‘circular’, since the topic is discussed from all perspectives, with no clear conclusion or recommendation made at the end of the essay. Kaplan’s finding about the direct approach to argumentation of the Anglo-Saxon pat- tern found support in the studies of Clyne (1981) and Connor (1987). Malcolm & Honjio (1988) also discovered that in writing problem-solution essays, English L1 speakers typically followed a pattern with clearly marked sections:

18

Brief introduction to the topic ↓ Identification and discussion of important issues and problems ↓ Possible solutions ↓ Evaluation of solutions

Chinese students, by contrast, produce this pattern:

Lengthy introduction to the topic ↓ Presentation of the facts related to the topic ↓ Presentation of the facts which oppose the view just given ↓ Indefinite conclusion

The ‘Indefinite conclusion’ in the Oriental pattern shows that Chinese students are not predisposed to settle the issues raised. Kaplan (1966:2) believes that the pattern reflects the students’ own cultural background. Perhaps the students investigated had been influ- enced by the scholarly tradition of their own country, where there is “no requirement to argue … [nor] to reach clear-cut conclusions” (Ballard & Clanchy 1991:32–3).

ASSESSING ACADEMIC WRITING

An aspect closely related to the teaching of academic writing is the issue of assessment. Holistic rating, defined as “any procedure which stops short of enumerating linguistic, rhetorical, or informational features of a piece of writing” (Cooper 1977:4), is a popular tool in assessing writing. Yet many teachers are concerned with two issues, scorer reli- ability and construct validity.

Reliability of holistic rating

One of the assumptions underlying holistic rating is that given a scale which describes the characteristics of a text at all levels, trained raters will assess the texts in the same or similar ways every time, i.e. holistic rating should have reliability. Yet scorer reliability is much affected by differences in the perceptions and attitudes of the raters and the kind of training they receive (Milanovic et al. 1995:93).

Validity of holistic rating

Milanovic et al. (ibid.) maintain that “marking is not simply a matter of reliability, but also concerned with the issue of construct validity”. The validity of holistic rating is in- terconnected with a whole range of textual properties such as content, sentence structure

19 and organisation, as well as external aspects like its length, layout and appearance. Other extrinsic factors in holistic rating are associated with the raters themselves.

Extrinsic factors

Some raters base their judgements on the appearance of the papers, rather than essay quality. Stewart & Grobe (1979:213), for example, found that the results of holistic judgement were positively correlated with essay length. Layout also seemed to lead to prejudices in some raters even before they had looked at the content (Milanovic et al. 1995:106). On the other hand, papers in poor handwriting can generally receive lower holistic ratings (Charney 1984:75; Vaughan 1991:114). Charney (loc. cit.) comments that these aspects of texts “are easy to pick out but are irrelevant to ‘true writing ability’”. Some extrinsic factors come from the context of marking itself, and might also turn out to be hindrances to judgement of “true writing ability”. One such factor is the ran- dom sequencing of essays (Vaughan 1991:121; Milanovic et al. 1995:106). Another is the background of the raters. Newcomb (1977) also found that holistic judgements could be affected by the rater’s nationality and sex. The first finding, concerning nationality, was echoed by Connor-Linton (1995), who discovered that while Japanese EFL teachers valued word choice, Americans focused more on intersentential discourse. Newcomb’s second finding, sex as a variable in holistic rating, was confirmed by Vann et al.’s (1991) study, which found that compared with their male counterparts, female raters were less likely to attribute errors to carelessness, or be irritated by erroneous language. Lexical density is another characteristic which affects holistic rating. Nold & Freed- man (1977), Grobe (1981), Neilson & Piche (1981), Laufer (1994) and Laufer & Nation (1995) have all found that teachers give better grades to texts with high levels of lexical density, that is, essays which contain a larger number of low-frequency words. Yet the factors which attract most attention from raters are content/subject matter, grammatical accuracy, and organisation.

Content/subject matter

In several investigations, content has been found to be the most important rating criterion. For example, in Vaughan’s (1991:114) study, a number of English teachers, all experi- enced in holistic rating in the same university system, were asked to grade six essays ho- listically, as well as make verbal comments on them. It was found that raters made most comments in regard to content. This supports Freedman’s (1981) finding, in which prob- lems with content emerged as the most frequent issue raised by assessors. Content is not only valued by some English teachers; it is also an important rating criterion to many subject specialists.

Grammatical accuracy

Grammatical accuracy is a further aspect which was found to be emphasised by English teachers. Stewart & Grobe (1979), for instance, discovered that good writing was associ- ated with the mechanics, that is, freedom from spelling errors and other surface errors. This result was supported by Sparks (1988), who investigated whether grammatical ac- curacy had a bearing on raters’ perception of ESL writing quality. His findings, based on

20 the holistic ratings of 30 ESL essays by two experienced teachers, revealed that high- rated writing was correlated with grammatical accuracy. In Lukmani’s (1995) study, it was discovered that English teachers valued grammatical accuracy more than did subject specialists, who paid more attention to coherence. These results differed from those of Bridgeman & Carlson (1983) and Hamp-Lyons (1988), who found that in rating essays, English teachers placed most attention to text organisation. Comparatively, grammatical accuracy is relatively easier to judge. Ballard & Clanchy (1991:33) include spelling, grammar, idiom, register, format and styles of referencing in a set of “objective criteria” by which an assignment can be assessed. Although the no- tions of what constitute an excellent piece of writing and a very poorly written essay dif- fer from one rater to another, very few will mistake a grammatically accurate essay for an inaccurate one, or vice versa.

Organisation

Organisation undoubtedly plays an important role in judgements of writing quality. In Connor & Lauer’s (1986) investigation, it was found that organisation had the highest correlation with the holistic scores. This confirmed the earlier findings of Lautamatti (1978, 1980), Connor (1984) and Connor & Farmer (1985). The differing focuses of English teachers and subject specialists in assessing aca- demic writing have been examined in other studies (e.g. Bridgeman & Carlson 1983; Keller-Cohen & Wolfe 1987; Hamp-Lyons & Reed 1990). In Bridgeman & Carlson’s study, the evaluation criteria for written work used by English teachers at university were compared with those used by subject specialists. Organisation was the most important rating criterion for the English teachers. However, the subject specialists placed ‘paper organisation’ only sixth. They viewed ‘quality of content’ as the most important, though it was ranked only ninth in importance by the English teachers. Similar findings were obtained by Hamp-Lyons (1988). To explore whether English teachers and subject specialists in Hong Kong also put divergent emphases on what counts as writing quality, I interviewed the programme co- ordinator and three assessors of the ‘Writing and Speaking through the Curriculum’ Pro- gramme at HKUST (Wong 2001b). The staff noted that while Business lecturers placed great importance on content, they themselves found content and organisation equally im- portant. At first glance, their view seemed to differ from that of the English teachers we examined earlier, who ranked organisation much more highly than content. However, when asked to elaborate, the interviewees said that for them, ‘content’ embodies all as- pects of essay quality, except grammar. Two of the assessment criteria they set for ‘con- tent’ are that an essay should be “well-developed and well-organised, and include an ef- fective introduction and conclusion”. These two criteria are, in fact, related to organisa- tion, rather than content. This investigation highlights a serious problem in the validity of different language constructs in holistic rating. It is possible that teachers’ judgements of one aspect of a text, such as content, are confounded with others, such as organisation. In any case the organisation of a text is inherently difficult to judge. Assessors may respond differently to it, as noted by Hoey (1991:166):

The same text may be found coherent by one reader and incoherent by another, though an overwhelming consensus can be achieved for most naturally-occurring texts.

21 The problem results from the fact that “the unity of text is not identifiable with a combi- nation of linguistic features and will never be absolute” (Falvey 1993:42). Therefore the validity of the rating of organisation poses a major question for teachers, and test admin- istrators within the field of language assessment. Yet another problem in the appraisal of organisation and textual coherence emerged in the study of rating reported by Wong (1995), in which some raters admitted that they confused the concept of ‘coherence’ with that of ‘cohesion’. This tallies with the findings of some researchers (e.g. Field & Yip, 1992; Field 1993, 1994; Milton & Tsang 1993; Milton 2001) that in Hong Kong, a great deal, if not too much, attention is paid to the teaching of cohesion, and on the use of cohesive devices. The emphasis on micro- organisation – and on grammatical accuracy – tends to mask the importance of the macro-organisation of discourse. Some raters indicate that they tend to assume that a script which is full of errors is badly organised as well (Wong 1995:157). Thus the valid- ity of holistic rating could be questioned, because raters may be using different rating criteria. This confirms Vaughan’s (1991:120) observation that “[d]espite their similar training, different raters focus on different essay elements and perhaps have individual approaches to reading essays”.

Questions and issues

The preceding discussion has reviewed some of the problems in holistic rating and how its validity can be questioned. This suggests the need to investigate whether organisation, as well as the coherence and integration of arguments, go together with grammatical ac- curacy in contributing to writing quality, and whether these qualities can be found in es- says which receive a high grade in holistic rating. Apart from the two research questions raised at the end of Chapter 2, two more can be formulated:

Question 1

Do the high-rated essays display a higher level of coherence and integration of arguments than the low-rated essays; and if this is/is not so, what are the factors accounting for this?

Question 2

Do the high-rated essays display a greater degree of grammatical accuracy than the low-rated essays; and if this is/is not so, what are the factors accounting for this?

The discussion also highlights the importance of analysing grammatical accuracy sepa- rately from textual organisation, though both are involved in successful writing. Com- pared to grammatical accuracy, discourse organisation in general, and the structure of argumentative texts in particular, are much less understood. Systematic ways of analys- ing argumentative discourse will be discussed and developed in Chapter 4, with the im- mediate aim of developing a method for investigating the integration of arguments. The ultimate aim is to devise a model for judging and/or teaching coherent writing.

22 Chapter Four: The structure of argumentative text and the study of rhetorical relations

THE STRUCTURE OF ARGUMENTATIVE TEXT

A text is a structural equivalent of language in real use which conveys meaning in all four senses of Hymes’ (1972) communicative competence (whether a text is: possible, feasible, appropriate, and performed), and which suggests a topic of discourse… Grabe & Kaplan (1996:40)

A text is thus a communicative event, whose message must be conveyed successfully from the writer to the reader. Grabe & Kaplan (op. cit.:42) suggest that the text structure has a major role to play in ensuring successful communication, and it should be exam- ined at different levels:

texts convey information to the reader on many different levels. Any theory of writing will, therefore, have to provide explanations about such multilevel communication …

This principle is central to the design of the present study.

Global structure

Grabe & Kaplan (1996:42) suggest that “understanding and producing texts will not only require knowledge of the surface structuring of texts but also of the underlying textual structuring”. The underlying structure of a text was earlier defined by Grabe (1985:110) in terms of coherence as “the underlying relations that hold between assertions (or propositions) … [which] contribute to the overall discourse theme”. He also notes the one thing that various well-known text analysis models have in common is the idea that there are three interacting features essential to a coherent text:

• a discourse theme (or thesis); • a set of relevant assertions relating logically among themselves by means of subordination, coordination and superordination …; and • an information structure imposed on the text to guide the reader in under- standing the theme or intent of the writer …

Cooper & Matsuhashi (1983:12) state, “Any discourse of more than a sentence or two requires a global plan … about the purpose for the writing and about the readers … for the writing.” This view is supported by several others, who have chosen to use alterna- tive terms for the ‘global plan’, such as ‘schema’ (Callow & Callow 1992), ‘frame’ (Frederiksen 1986) and ‘discourse structure’ (Meyer 1992). The main argument of these researchers regarding the global structure is that it is useful for guiding the writer to write and the reader to understand the message. The global structure of discourse is also important in distinguishing one genre from another:

Logic, conception, and execution all differ as a writer shifts from one of these [discourse] types to the other. … The major difference … is not at the sentence level, though it is there as well, but at the discourse level; at the level of the global structure or plan of the piece. (Cooper & Matsuhashi 1983:14–15)

23 Linguists generally categorise texts as narrative, expository or argumentative. It is the argumentative genre we are investigating in this study. Many linguists link the text type and structure of a piece of writing with its audience and purpose. As Cooper & Matsuhashi (1983:12) note, “a particular purpose, audience and occasion for writing lead to a certain discourse type with its characteristic structure”. Their argument is that the discourse structure “can only be learned through extensive ex- perience with each discourse type as a reader and writer” (ibid.:15). Meyer (1992:80) further points out that if the writer selects an appropriate structure to convey the message and also cues the readers into this structure, it will help the reader in constructing a cog- nitive representation of the text which is similar to what is intended by the writer. Meyer’s view is supported by Grabe (1996:1), who states that “the text provides strong guidelines for the coherent interpretation that is intended by the writer, and that is typically well perceived by the reader”. He elaborates (ibid.:10):

Genres begin as functional conventions that serve meaningful purposes, because they pro- vide defaults for recognizing basic background information about text while letting a reader focus on the newer more-specific information.

This is in line with Longacre’s (1992:110) observation that “[a] schema is discourse-type specific but is not discourse specific in terms of the content of a particular text”. Some researchers emphasise that the global structure also contributes to the coher- ence of the text. McCutchen & Perfetti (1982) maintain that the global structure, or what they call “text form constraint (genre)”, is important in that it actually constitutes one of the four major sources of discourse coherence. (The other three are topic knowledge, topic coherence and local connectedness.) Lautamatti (1990), on the other hand, believes that the global structure creates a special type of coherence, called Propositional Coher- ence, achieved by developing the discourse topic with the aid of cognitive frames, by storing “information relating to different objects, events etc. in a hierarchical, and thus maximally economical, way” (ibid.:35). Meyer (1992:80) also considers discourse struc- ture of paramount importance, as it is “the organization that binds [the text] together, and delivers its overall organization”, by specifying the logical connections among ideas, as well as the subordination of some ideas to others. The global structure thus seems to contribute to the integration of the text in several ways. Tirkkonen-Condit (1989:417) states that the dominant global structure in argumen- tative text should be an Argumentative Response Structure. Such a structure manifests itself in the argumentative text as three types of sequence: Question-Answer, Remark- Response and Problem-Solution. She observes that the usual sequences at the global level are Remark-Response and Problem-Solution, and in particularly the latter. She be- lieves that it is ideal for the description of the argumentative text to be embedded in the description of the argumentation process, which is described as “an instance of the cog- nitive process of problem-solving” (Kummer 1972:29): the writer/speaker (S) assumes that the reader or hearer (H) has an undesirable initial position in a state of affairs. S’s goal is to change the initial position in H’s mind so that it finally approaches S’s own position. This goal is achieved via a number of sub-goals – the single arguments, or the claims, of the argumentation. The end product, the text, is “a constellation of miniature texts which all contribute to the ultimate goal, the solutions” (Tirkkonen-Condit 1984:224). The miniature texts themselves are in turn organised into three global units, the Topic Unit, the Elaboration Unit and the Conclusion Unit. But apart from the Problem-Solution sequence, the Remark-Response sequence is indeed a recurrent structure in the argumentative text, since argumentation is often trig-

24 gered as a response to a remark, an assertion, or a claim. Guba & Lincoln (1989) suggest that argumentative inquiry often takes place in value-laden contexts, and that in selecting an issue, values are involved. If indeed values are involved in argumentation, it is evident that a certain position is to be taken:

[T]he function of argumentation is to put forward possible reasons and then to test those reasons. The possible belief is embodied in the position …. The possible reasons for the belief are embodied in the arguments given. (Meiland 1989:187–8)

The argument which conveys the position taken by a speaker/writer in response to a re- mark/claim typically embodies its own argumentative structure. In fact, the element ‘claim’ not only forms the structural key to an argument. It may appear at other levels of the text, as the argument is further developed. That is why Tirk- onnen-Condit (1989:447) believes that the “typical sequence of functions” in argumenta- tive text is ‘Claim + Justification + Induction’, which she terms as the interactive pattern in the ‘Problem’ unit in the Problem-Solution Structure (Tirkonnen-Condit 1984:230). She is thus in effect saying that there are local structures in argumentative text. Linde- berg (1994:139) makes this point very clear, when she refers to these structures as “local rhetorical functions”: “A well formed sequence of local rhetorical functions form a global rhetorical structure”. In other words, theoretically speaking, the arguments at the global level, and those at the levels beneath it, should be integrated. The interplay of higher and lower levels of argument can also be seen in writers’ use of ‘evaluation’. Evaluation is one of components in the sequence Situation-Evaluation, a variant of the Fundamental Information Structure (FIS) (Winter 1976). This variant re- sembles Tirkkonen-Condit’s Remark-Response sequence in presenting first the ‘back- ground’ to the discussion, followed by an evaluative response. But evaluation is often used in the lower levels of discourse as well, when writers make a judgement on what they themselves, or other people, have said earlier in the text.

Local structures

So far we have been reviewing what we might call top-down approaches to text organi- sation, starting with the global structure, and finding structured components within it. Other text linguists have developed bottom-up approaches to discourse structure, to en- hance our understanding of how arguments are built up from the local structures. The last thirty years have witnessed parallel approaches in discourse analysis which are largely “intuited analyses of the relations … according to a limited set of predetermined proposi- tions” (Winter 1992:139). Such approaches include:

(a) Clause Relation Theory: Winter (1971, 1976, 1977, 1992), Hoey (1979, 1983, 1991) and Jordan (1984, 1992), (b) Rhetorical Structure Theory: Mann & Thompson (1986, 1987, 1988) and Thompson & Mann (1987), (c) Functional Role Theory: Lieber (1979), Lindeberg (1988) and Albrechtsen et al. (1991); and (d) Communicative Function Analysis: Wong (1993).

In respect of Clause Relation Theory, which provides a foundation for understanding lo- cal structures, Winter believes that the ‘clause’ is the basis of discourse structure. Yet his use of the term is not simply that of traditional grammar. His ‘clause relations’ are a way

25 of referring to local semantic relations: “A clause relation is how we interpret the infor- mation of one clause in the light of other clauses.” (Winter 1974:59) Jordan (1992:179) makes a similar point: “A clause relation is the meaning between two coherent stretches of text.” Earlier, in 1983, Hoey, another proponent of the Clause Relation Theory, related the concept of ‘clause relation’ to the hierarchical structure of discourse: “a discourse may be made up of clause relations which are themselves members of larger clause rela- tions which are in turn members of an overall relation” (ibid.:32). In other words, the overall clause relation of a text (be it ‘Situation-Evaluation’, or the other variants of the Fundamental Information Structure) is built up from smaller clause relations. This is echoed by Albrechtsen et al. (1991:91), when they expound the concepts of local coher- ence and global coherence:

local coherence applies to formal, semantic and pragmatic relations between neighbouring clauses or functional units, and global coherence relates to higher order units dominating such interclausal relationships.

Albrechtsen et al.’s definition for ‘local coherence’ as “formal, semantic and pragmatic relations between neighbouring clauses or functional units” elaborates what Hoey re- ferred to as “clause relations”. What is more enlightening is that in saying that there are “higher order units dominating [the] interclausal relationships”, Albrechtsen et al. are providing an explanation here of how clause relations can be “members of larger clause relations which are in turn members of an overall relation”. The “higher order units” re- fer to the global structure of discourse, which we discussed in the last section. Other linguists have also commented on the rhetorical relationship between particular textual segments at the lower discourse levels. Aston (1977:487), for example, stated, “one of the expectations as to the development of argumentative discourse is for state- ments to be followed by explanations, assertions by justifications”. In more general terms, Lindeberg (1994:139) maintained that the choice of local rhetorical structures is closely related to the discourse structure at the global level. She further (ibid.:141) stressed the significance of examining the local structures: “The sequences of functions essentially reflect the strategies that the students used to build up their universe of dis- course, to develop the topics, ideas, concepts or arguments they introduced.” It was from Lindeberg’s statement that one of the objectives of this study was derived: to investigate how effectively L2 tertiary students build up their universe of argumentative discourse, from the lowest level to the highest; that is, to study the argumentative structure not only at the global level and the clausal/sentential level, but also at the levels in between the two. While the actual shape and size of the local rhetorical structures are left open, their function is clear. Lindeberg (1985b:85) suggests that local structures, like the global structure, are constrained, or guided, by the specific writing task, as well as the discourse type: “different tasks call for different development strategies, and hence different coher- ence patterns”. Enkvist (1990:23) also states that “[m]any kinds of texts have conven- tional patterns and hence conventional strategies, in them, optimal coherence results from conforming to the optimal strategy”. These postulates regarding the local structures being guided by the genre and the writing task will be investigated in the present study. As for the local structures of argumentative text, researchers believe that these are also guided by the task and/or the genre. Toulmin et al. (1984:14) maintain that argu- mentation is “the whole activity of making claims, challenging them, backing them up by producing reasons, criticising those reasons, rebutting those criticisms, and so on”. Three elements, ‘Claim’, ‘Data’ and ‘Warrant’, are included in Toulmin’s (1958) ‘model

26 of informal logic’. ‘Claims’ are “assertions put forward publicly for general acceptance” (Toulmin et al. 1984:29); ‘Data’ are “the facts we appeal to as a foundation for the claim”; and ‘Warrants’ are “general, hypothetical statements, which can act as bridges” (ibid.:97–8). Aston (1977), on the other hand, connects local rhetorical structures with speech acts, believing that the analysis of speech acts is essential in uncovering larger communicative units, or “speech events”. He quotes Widdowson (1973:74):

certain combinations of acts such as definitions, classifications, generalisation, qualifica- tion, and so on, which in many cases constitute larger communicative units like explanations, descriptions and reports, and which may be said to reflect the actual methodology of scientific enquiry.

He also cites Sinclair & Coulthard (1975) with respect to ‘acts’ combining to form ‘moves’, ‘moves’ to form ‘exchanges’, ‘exchanges’ to form ‘transactions’, and ‘transac- tions’ to form ‘speech events’. To apply this to argumentative discourse, Aston (1977: 474) postulates that acts combine to realise ‘famigerated’ activities, such as the evalua- tion of hypotheses and the accounting for observed phenomena. He states that evaluation is related directly to a frequent activity in argumentative text, the discussion of hypothe- ses (ibid.:487). This is because when a writer introduces a reported assertion, the reader will expect the writer to go on to evaluate the assertion. The two acts (Reported Asser- tion-Evaluation) combine to form a Claim. Aston points out that sometimes the Evalua- tion may be inferred if it is omitted. He gives an example to illustrate this: “According to Chomsky, generative and interpretative semantics are notional variants. He ignores the fact that …” (ibid.:496). His conclusion is that “the reporting of an assertion in argu- mentative discourse entails an evaluation of it, whether this is done explicitly or not” (ibid.), reminiscent of Aristotle’s use of ‘enthymeme’, “[whose] commonest contempo- rary sense is that of an argument in syllogistic form, in which one premise is unstated, and implied by the conclusion” (James 2001:70). In elaborating on the ‘accounting for phenomena’, Aston suggests that there is a ‘re- cursive possibility’ of explanatory relationships in extended discourse. He states that an Observation can be realised by a Claim-Justification. Since Justification is in turn real- ised by Observation, recursion occurs (Aston 1977:490). Figure 2 refers:

Observation Claim

Justification Observation Claim

Justification etc.

Figure 2: Explanatory relationships in extended discourse: recursion.

It was Tirkkonen-Condit (1984) who took a step further to develop what Toulmin and Aston had explored, particularly the notions of ‘Claim’ and ‘Justification’ into ‘Claim + Justification + Induction’. She calls this the typical sequence of functions in argumenta- tive text. Her postulation was confirmed by Connor’s (1987). Tirkkonen-Condit too

27 linked this typical sequence with the global structure of argumentative text. Thus the in- terrelation between local and global structures of argumentative text has ample theoreti- cal and empirical foundation. In the models for the local structures of argumentative text described above, there are two common elements: ‘Claim’ and ‘Justification’. The third element, although termed differently by Toulmin (‘Warrant’) and Tirkonnen-Condit (‘Induction’), refers to the element of ‘Evaluation’ made on the basis of the justification. Therefore although the models look different, they are essentially similar.

SEGMENTATION OF TEXT

There are divergent methods for segmenting text. Three of them are most commonly used, which involve the use of

• a grammatical unit (the sentence or the T-unit), • a semantic unit (the proposition), or • a grammatical/semantic unit (the F-unit). as the basic unit of segmentation. In this section, each of these methods will be intro- duced and evaluated, so as to justify the final decision made regarding the method of segmentation used in the present study.

Using the sentence as the unit of segmentation

At first sight, the sentence might appear to be an appropriate segmentation unit because it is easily identified as a syntactic structure. Yet its discoursal value is not to be assumed, as pointed out by McCarthy (1991:152):

a number of things in clause and sentence grammar have implications for the discourse as a whole, in particular, word order, cohesion, and tense and aspect … the sentence will have no special status other than a grammatical and orthographic unit.

Earlier, in 1983, Hunt had already raised objections to using the sentence as a segmenta- tion unit for analysing student writing, on the basis that some students are weak in punc- tuation. Furthermore, the indiscriminate use of and by some writers in co-ordinated sen- tences makes this unit of segmentation even more problematic. Hunt’s view is particu- larly relevant to our own data, i.e. student writing, which has often been criticised for misuse of punctuation and overuse of cohesive devices. For all these reasons, the ortho- graphic sentence is unsatisfactory as the basis of segmentation.

Using the T-unit as the unit of segmentation

Hunt’s own solution was to use the T-unit as the unit of text segmentation. He (1970:4) defined the T-unit as “one main clause plus any subordinate clause or nonclausal struc- ture that is attached to or embedded in it”, and argued that segmenting a text into T-units means segmenting it into its shortest grammatically feasible (i.e. grammatically inde- pendent) units. Consider this example (from Hunt 1983:101):

28 I like the movie we saw about Moby Dick the white whale the captain said if you can kill the white whale Moby Dick I will give this gold to the one that can do it and it is worth sixteen dollars they tried and tried but while they were trying they killed a whale and used the oil for the lamps they almost caught the white whale.

According to Hunt, this long sentence consists of six T-units:

1. I like the movie we saw about Moby Dick the white whale. 2. The captain said if you can kill the white whale, Moby Dick, I will give this gold to the one that can do it. 3. And it is worth sixteen dollars. 4. They tried and tried. 5. But while they were trying they killed a whale and used the oil for the lamps. 6. They almost caught the white whale.

Other analysts agree that the T-unit has several advantages as a segmentation unit. T- units can be identified objectively, in terms of simple and complex sentences, or co- ordinated sentences with subject deletion (O’Donnell 1976:32; Watson 1983:129–130). Secondly, they are not affected by poor punctuation. Lastly, since T-unit length incorpo- rates the mean words per clause and the number of clauses per T-unit, it is a useful index of syntactic complexity (O’Donnell 1976:32; Hunt 1977; Witte & Davis 1980:6). However, the T-unit does not seem to be an appropriate unit of segmentation for the current study. For one thing, it is generally used as an index for measuring syntactic ma- turity, which is not our focus here. For another, since it is “one main clause plus any subordinate clause or nonclausal structure that is attached or embedded in it”, the T-unit, compared with the F-unit, to be introduced in the next section, is less delicate in reveal- ing the development of the argumentative discourse. It may also bring confusion to the calculation of the ratio of the number of units to the number of functions, as some T- units may consist of only the main clause while others may have a subordinate clause or a nonclausal structure attached or embedded in them. Indeed, Hunt (1983:107) himself suggests that “it may well be that, among high- school and college students, clause length is a better index of maturity than is T-unit length”, for the reason that “in the early grades both clause length and number of clauses increase but for full maturity it is clause length that will grow, not the other index”. This suggestion is relevant to the current study, as the data sets were produced by “college students”. For this and other reasons, the T-unit was not chosen as our segmentation unit.

Using the proposition as the unit of segmentation

A proposition is a statement expressing a judgement or opinion. Crombie (1985:13) ex- plains that the semantic relations involved in text analysis are, minimally, two proposi- tions. Each member of a semantic relation is often encoded as a separate clause, since this is frequently the linguistic unit used to encode a single proposition. However, a se- mantic relation may also be encoded as a group of clauses; cf. her illustration (ibid.):

John’s playing squash and Mary’s weeding the garden while Tim’s chopping wood and Sam’s preparing the dinner but Jane isn’t doing anything.

Alternatively, a semantic relational member may be encoded as a proposition embedded in a single clause, for example:

29

Her exaggerations make him furious. (i.e. she exaggerates (reason)

Reason – Result he becomes furious (result)

This suggestion of using propositions as the segmentation unit is however criticised by Lindeberg (1988:64): “even a very short text contains a considerable number of proposi- tions. Thus a whole essay broken down into sets of propositions would soon become quite cumbersome to handle …” Using propositions as the unit of segmentation would indeed be cumbersome for the analysis of a large number of essays. What is more impor- tant, if the principle of segmentation relies solely on semantic content, it is not very reli- able, as too much subjective judgment may be involved in the process. Therefore the proposition will not be used as the unit of segmentation in the present study.

Using the F-unit as the unit of segmentation

For Lindeberg (1988), the most appropriate unit of segmentation is the functional unit of discourse, or F-unit, a suggestion first put forward by Lieber in 1979. Interestingly, the F-unit involves using both grammatical and semantic notions in text segmentation. It is defined as “the set of clauses or clause equivalent serving an identifiable rhetorical func- tion in written discourse” (Lieber 1979:Abstract:i). A variety of grammatical structures, such as co-ordinate structures and subordinate structures, can constitute F-units. Lieber’s segmentation principles (1979:93–5) are presented below:

1. Coordinate structures 1.1. Full clauses joined by coordinating conjunctions or marks of punctuation constitute separate F-units. e.g. a. John broiled the steak, b. and Bob made the salad.

1.2 Clauses exhibiting gapping in a non-initial member constitute separate F- units. e.g. a. Nick prepared the main course, b. and Tony the desert [sic].

1.3 Clauses containing conjoined verbal structures will be segmented into more than one F-unit. e.g. a. He finished his work b. and left immediately.

1.4 Conjoined nonverbal elements within a clause will be segmented into separate F-units when an overt marker indicating a change of rhetorical function is present (i.e. but, except, or an adverbial marker or preposi- tional phrase). e.g. a. She is highly qualified b. and therefore suitable for the position.

30 2. Subordinate structures 2.1 Adverbial subordinate clauses and clause equivalents, with the exception of temporal and locative structures e.g. a. They had to hire new teachers b. (in order) to handle the expected increase in enrollment.

2.2 Nonrestrictive relative clauses and sentence relatives e.g. a. The set of F-units includes reduced non-restrictive relatives, b. which are generally called appositives.

2.3 Nonrestrictive appositives (i.e. reduced nonrestrictive relative clauses) e.g. a. The latest procedure can be found in Professor Wirth’s new book, b. Airport Management in Developing Countries.

2.4 Nonrestrictive appositives of exemplification, identification and renaming e.g. a. He’s made all his arrangements with the new managing company, b. namely, Walter and Samuels.

2.5 Absolute constructions related to adverbial clauses or non-restrictive relatives e.g. a. The day being sunny, b. he decided to play golf.

The F-unit, compared to the T-unit, is a more appropriate unit for discourse segmentation since it is defined not by grammar alone, but by rhetorical functions as well. Hunt (1970:4) notes that “[a]ny complex or simple sentence would be one T-unit, but any compound or compound-complex sentence would consist of two or more T-units”. Yet in the case of the F-unit, a complex sentence would consist of two or more F-units, just like a compound sentence. Therefore in using the F-unit, the potential problem of confusion of calculations connected with the T-unit can be avoided. Another strong reason for our adopting the F-unit as our segmentation unit lies in the aim of this study, which is close to Lindeberg’s objective of tracing propositional coher- ence and thus the integration of arguments in student text (Lindeberg 1988:63).

THE HIERARCHICAL INTEGRATION OF ARGUMENT AND THE STUDY OF RHETORICAL RELATIONS

We have seen how the global structure and local structures of text in general, and those of argumentative text in particular, are closely related. This is an important notion, be- cause if these structures are not integrated, the final product will not be a coherent text. Thus, a coherent text must satisfy the requirements of both local and global coherence:

The requirements of global coherence say, “Given the overall goals I am trying to accom- plish, what can I say next that will serve them?” Local coherence says, “Given what I just said, what can I say that is related to it?”. (Agar & Hobbs 1982:7)

Thompson & Mann (1987:80) suggest that texts are not just strings of clauses, but rather

31 consist of hierarchically organised clauses which relate to one another in different ways. They are, in effect, saying that the clauses in a text are hierarchically integrated. Earlier, Hoey (1983:32) also asserted, “discourses are organised … at least in part in a hierarchi- cal manner”. Jacobs (1990:154) further relates this to academic discourse and states that academic prose has a hierarchical structure, tall rather than wide. The concept of ‘hierarchical blocking’ was introduced by Lindeberg (1988:158), who maintained that there are large semantic ‘blocks’ in the text. To support her view, she cited Hoey’s (1983:32) observation that a discourse is made up of clause relations which are themselves members of larger clause relations. She also noted that hierarchical block- ing had been discovered in Tirkkonen-Condit’s (1985) study of Problem-Solution struc- ture in L1 professional argumentative articles in English, and in Thompson & Mann’s (1987) as well as Mathiessen & Thompson’s (1988) studies of professional expository writing in English. However, she added,

hierarchical ‘blocking’ in the modeling of a text is not so simple, and it is often impossible in less well-formed or less structured essays. … It is rare … for student essays to be so clearly organised into hierarchical blocks. (Lindeberg 1988:158–9)

The reason for this is explained by Albrechtsen et al. (1991:89):

[I]n transferring ideas to paper, basically hierarchical cognitive structures have to be trans- lated into basically sequential linguistic form. It is very difficult to carry out this process in such a way that both sequential (local) and hierarchical (global) coherence are taken care of simultaneously by one linguistic sequence.

They present these ideas graphically in a ‘Dynamic Model for Written Communication’:

SENDER TEXT RECEIVER

COGNITIVE GLOBAL ‘TOP-DOWN’ HIERARCHY COHERENCE PROCESSING

= = =

SEQUENCE LOCAL ‘BOTTOM-UP’ COHERENCE PROCESSING

Figure 3: Dynamic model for written communication, from Albrechtsen et al. (1991:90).

In the figure, the problems experienced by the receiver in top-down and/or bottom-up processing are presented by the broken lines. What the solid lines mean the authors do not make clear, however. Albrechtsen et al. (op. cit.:89–90) suggest that L2 student writ- ers develop different coherence strategies (associative, focal and hierarchical) to cope with the requirements of local and global coherence at different stages of their learning. Yet they offer no explanation as to how the L2 writers at fairly advanced levels, such as our subjects, can be helped to develop hierarchical coherence in their texts. An adapta- tion of their model will be made in Chapter 8, to make suggestions as to how tertiary students can be helped to improve the local and global coherence in their writing.

32 Since it is relatively more difficult to analyse hierarchical ‘blocking’, or the hierar- chical integration of argument, in student text, finding the appropriate model, or method, is important. We will now review three discourse analysis methods used for analysing rhetorical relations within texts, namely Rhetorical Structure Theory Analysis, Func- tional Role Analysis, and Communicative Function Analysis. These three methods repre- sent two different approaches to discourse analysis, bottom-up and top-down. From another perspective, they also represent two divergent ways of studying text, studying it at the local levels only, and studying it at local and global levels.

Rhetorical Structure Theory Analysis

Rhetorical Structure Theory (RST) Analysis adopts a bottom-up approach to analysing text, from the local levels to the global level. The analysis at the local levels reveals the relationship between individual units, whereas that at the global level shows how large segments of a text are built up from smaller textual segments below. The first and second steps of RST Analysis are similar to many methods of discourse analysis. The first involves segmenting the text into units for analysis, or text spans, which can be clauses, or a cluster of units at higher levels (Stuart-Smith & Busbridge 1996). The second step entails the categorisation of the units according to the rhetorical functions they perform in the discourse. In other words, the semantic relations between clauses, sentences or larger discourse units are studied. In fact the term ‘relations’ is used by Mann & Thompson (1987:3) to refer to rhetorical functions. Mann and Thompson have a unique view of identifying rhetorical relations: their starting point is ‘nuclearity’. They believe that the text can be broken down into pairs of text spans, in each of which there is a member which is “more essential to the writer’s purpose than the other” (op. cit.:31), labelled the ‘nucleus’, while the other member which is “incomprehensible if independent of the other, … [and] is more suitable for substitution than the other” (ibid.) is labelled the ‘satellite’. To perform RST Analysis, they propose using a set of relations, which are listed in full in the next section. These rhetorical relations are presented through diagrams, which consist of two basic elements, vertical lines and arcs. The vertical lines are used to indi- cate the nuclei, while the arcs are utilised to connect units held together by a certain rela- tion, and are labelled with rhetorical functions. The following is Thompson & Mann’s (1987:95–97) illustration of RST Analysis, operating on a gardening text:

Bouquets in a basket – with living flowers 1. There is a gardening revolution going on. 2. People are planting flower baskets with living plants, 3. mixing many types in one container for a full summer of floral beauty. 4. To create your own “Victorian” bouquet of flowers, 5. choose varying shapes, sizes, and forms, besides a variety of complementary colors. 6. Plants that grow tall should be surrounded by smaller ones and filled out with others that tumble over the side of a hanging basket. 7. Leaf textures and colors will also be important. 8. There is the silver-white foliage of dusty miller, the feathery threads of the lotus vine floating down from above, the deep greens, or chartreuse, and even the widely varied foliage colors of the coleus.

33 The RST diagram of this text is as follows: 1 – 8 background

1 – 3 4 – 8 elaboration purpose

4 1 2 – 3 5 – 8 elaboration elaboration elaboration

2 3 5 6 7 – 8 elaboration

7 8 Figure 4: RST analysis of the “Bouquets in a Basket” text, from Mann & Thompson (1987:97).

At the most general level, this text provides background information about the “garden- ing revolution” (Units 1–3). The rest of the text describes the method of arranging flower baskets with living plants. The description is achieved by the infinitive clause of purpose in Unit 4 (“[t]o create your own ‘Victorian’ bouquet of flowers”), together with the ‘elaboration’ relations found in many parts of the text. This illustrates how RST Analysis “identifies hierarchic structure … [and] describes the relations between text parts in func- tional terms, identifying both the transition point of a relation and the extent the items related” (Mann & Thompson 1987:2). Such an analysis method thus recognises the need to understand the intermediate structures in an argument, and how they connect into the hierarchy. However, it does have limitations. The first is that although RST Analysis can be used for diagnosis at local, intermedi- ate, as well as global levels, the distinctions between these levels are fluid. This does no harm to its graphic representation of individual texts, but it means that the method does not provide a means for text analysts to compare the findings at a specific level. The sec- ond limitation lies in the graphic representation. With its many lines and arcs, and the labelling of relationships in an extremely small font size to save space, the graphic repre- sentation is not user/reader-friendly.

Functional Role Analysis

Functional Role Analysis is another bottom-up approach to text analysis, which focuses on text relations at the local levels. Lieber (1979:Abstract:i) defines “functional units” as “the set of clauses or clause equivalents serving an identifiable rhetorical function in written discourse”. The first two steps of Functional Role Analysis are similar to those of RST Analysis: the text is first segmented into functional units, which are then catego- rised according to their rhetorical functions. Yet while RST Analysis can be used to trace rhetorical relations from the local levels to the global level, Functional Role Analysis only focuses on local structures (Lindeberg 1988:160). Her sample analysis provides an illustration of the ‘local’ nature of her method:

34 Text BPSD9 Sequences of Functional Roles 1. Is the British parliamentary system democratic? 1. Question 2. Every democratic country in the western world have some kind 2. Assert of parliament. 3. People can vote as they like 3. Assert b. and if they are not satisfy decisions made by the parliament. b. Qualify (3c) c. is it very easy to vote in the next elections for a new c. Result (3–3c) candidate.

Figure 5: Sample analysis using Functional Role Analysis, from Lindeberg (1988:147–8).

Although this method concentrates on the local level, it has the potential to be extended to the global level. Such an idea was suggested by Lindeberg (1988) herself as well as by Albrechtsen et al. (1991), her co-researchers in the NORDWRITE Project. Since the analyses carried out on the pilot data of the Project did not produce consistent results, Albrechtsen et al. suggested that “a more global approach seemed to be the key to an im- provement of the discriminatory power of the analysis”. They gave an illustration (op. cit.:86):

Extract from an essay rated ‘high’ 11: A large advantage in the modern world is that we have doctors and hospital, b: so that we can get medical help c: if we need that. 12: Before many children died a short time before they were born, b: because they were ill and couldn’t get any help. 13: Many people couldn’t afford to pay a doctor. 14: If you got cancer e.g., b: there wasn’t anything the doctor could do for you. 15: Now they have got more knowledge, b: and you can live a long, normal life, c: even if you get cancer.

11: ASSERT b: RESULT (11) c: QUALIFY (11b) 12: ASSERT b: CAUSE (12) 13: ASSERT CONTRAST 14: QUALIFY (11-11c) b: ASSERT SPECIFY (11-11c)

15: CONTRAST (14b) b: RESULT (15) CONTRAST c: QUALIFY (15b) (14-14b)

Column 1 Column 2 Column 3

By adding Columns 2 and 3, Albrechtsen et al. claim that they have extended the Func- tional Role Analysis from a local to a global analysis. Yet from their demonstration, we can see that what they have actually done is extended their method only up to the ‘inter- mediate’ level, where the bigger ‘meaning clusters’ lie, but not really up to the global level, where paragraphs connect with one another to form the macrostructure. Therefore,

35 although this idea of conducting discourse analysis beyond the local levels was adopted in the present investigation, the method suggested by Albrechtsen et al. was adapted to a great extent to serve the specific purposes of our study, and to make it a method which really analyses rhetorical relations from the local levels to the global level.

Communicative Function Analysis

A third method used to analyse rhetorical relations is Communicative Function Analysis (CFA) (Wong 1993). Interestingly, this method combines both RST and Functional Role analyses, and studies texts at both the local and global levels. A totally different ap- proach, the top-down approach, was taken: “The understanding of the text … begin[s] with a top down approach instead of assigning a communicative function to a functional segment by relating it to segments in close vicinity.” (Wong 1993:76) The procedure of CFA is first to divide the text at the global level (labelled as ‘Mac- rostructure’ in CFA), to match with “problem-situation-solution-evaluation” macrostruc- ture. This is followed by segmentation of the text at the local levels (labelled as ‘Sub- sentence Level’ and ‘Sentence Level’). The next step involves the compilation of an analysis table by assigning communicative functions to the functional segments (Wong op. cit.:75). Then statistical information is compiled. Finally, RST diagrams are drawn according to the analysis table. The first three steps are illustrated in Table 8:

Table 8: Communicative Function Analysis, from Wong (1993:App. p. 8). Sub- Sentence Sent. Sentence Macro- Para Text 1–3 Level No. Level structure Functions Functions 1 1a Hong Kong, a fast-moving city, Qualify Assert Problem 1b The pressure of life are forcing Hong Kong people 1c To forget their traditional roles Qualify 1d By neglecting their old people, Specification 1e In the sense that the senior are Comparison not enjoying the same authority as before. Gone are the days when having 2a an old in family Temporal Affirmation Was regarded as a blessing and Overlap (1) 2b source of pride.

Wong justifies his method by saying that if the bottom-up approach (such as RST Analy- sis) is taken, “the assignment of functional roles will be extremely difficult, chaotic, as well as localized, instead of having a global view” (ibid.:76). Yet RST Analysis can be used to analyse the local as well as global levels. When, on the contrary, an analyst be- gins text analysis with a top-down approach, s/he might project too much individual in- terpretation of the macrostructure of the text. This relies on subjective analysis of large blocks of text. Another drawback of Wong’s system is that unlike approaches such as RST, it cannot reveal clearly how local functions are ultimately built up into global ones, because in CFA there is no level between the Sentence Level and Macrostructure Level.

36 Synthesis of previous approaches to studying rhetorical relations

In the preceding paragraphs, we have examined three methods for studying rhetorical relations. Each of them seems to have limitations, which can bring considerable difficul- ties to text analysis. It was therefore decided that a new discourse analysis method, Rhe- torical Function Analysis (RFA), would be devised for the purposes of the present inves- tigation. This method addresses the problems discussed above. First, it analyses text from the local levels to the global level, and can therefore be used to investigate the hierarchi- cal integration of argument. In adopting this method, the analyst first identifies the rhe- torical relations of neighbouring units at the lowest level of discourse and traces the de- velopment of these relations through intermediate propositions to the paragraph. Finally the macrostructure is reviewed, to determine whether the relations at the highest dis- course level match the conventional global argumentative structures, such as the Argu- mentative Response Structure (Tirkkonen-Condit 1989). It thus keeps the merits of the bottom-up approach, of being more objective, without losing sight of the overall perspec- tive of the text. Also, RFA is superior to RST Analysis in that the exact level of analysis can be specified. A further advantage is that the graphic representation of RFA is much clearer than that of RST Analysis. RFA shares the same preliminary steps with many discourse analysis methods, namely:

Step 1: Segmentation of text Step 2: Categorisation of segmented units according to the rhetorical functions they perform in developing ideas in the discourse Step 3: Text analysis.

CATEGORISATION OF RHETORICAL FUNCTIONS

Previously used categorisation systems of rhetorical functions

The six categorisation systems to be reviewed here were devised to examine argumenta- tive text, but they differ in four respects. The first difference is related to the nature of the study each taxonomy was designed for. Most of these taxonomies were devised to inves- tigate student writing: Crombie (1985), Lindeberg (1988), Wong (1993) (closely mod- elled on Crombie) and Kaldor et al. (1998). Mann & Thompson’s (1986) framework, on the other hand, was designed for the investigation of professional expository prose, while Cooper & Matsuhashi’s (1983) taxonomy was mapped out with a combination of targets (that is, student text and professional text) in mind. A second kind of difference lies in the format of the taxonomies. All these systems, except Crombie’s, contain only freestanding functions. In Crombie’s system, both free- standing and paired-up functions exist. Examples of the former include ‘Chrono-logical sequencing’ and ‘Simple comparison’, and of the latter, ‘Reason-result’ and ‘Conces- sion-contra-expectation’. Crombie’s system is the only one to contain broad categories of semantic functions, such as ‘Temporal Relations’, ‘Cause & Effect’ and ‘Truth & Valid- ity’. A third difference lies in the elaborateness of categorising the rhetorical functions. There is a large range, from the least elaborate system, Cooper & Matsuhashi (containing only fourteen functions), to the most elaborate, Kaldor et al. (containing 37 functions). This disparity in ‘elaboration’ is related to what each of these researchers interprets as

37 the most important rhetorical functions in argumentative text. In the following, the six categorisation systems are tabulated for easy comparison:

Table 9: Six existing categorisation systems of rhetorical functions. Cooper & Mann & Crombie Lindeberg Wong Kaldor et al. Matsuhashi Thompson (1985) (1988) (1993) (1998) (1983) (1986) Motivation Question Initiation Query Evaluate Evaluation Evaluate Evaluation Cite Solutionhood Solution Background Background Circumstance Circumstance Summarise Summary Summarise Summary Justification Justify/ Support Predict Propose TEMPORAL RELATIONS Narrate Temporal Temporal Overlap Overlap Chronological Sequencing Sequencing Sequencing MATCHING Compare Simple Comparison Compare Comparison Contrast Simple Contrast Contrast Contrast CAUSE & EFFECT Cause| Reason- Volition/ Cause- Reason| Cause of Result Result Non-volition Result Result Conse- Cause|Result quence Means-result Specify- Means| Result Result Conclude Grounds- *|Conclude Specify| *|Conclusion Conclude/ Conclusion Conclude Generalise/ Consolidate Means-purpose *|Purpose Specify|* Means| Purpose Condition- Condition| Qualify| Condition| Condition of Consequence Result Result Result Conse- Otherwise Qualify Hypothesis quence TRUTH & VALIDITY State|* Statement/ Assert * Assert| Introduce Affirmation Affirmation State|* Statement- Assert|* Assert|Denial Denial Denial-correction Denial| Correction Concession Concessive❘ * Concession❘ Contra- expectation

38 Cooper & Mann & Crombie Lindeberg Wong Kaldor et al. Matsuhashi Thompson (1985) (1988) (1993) (1998) (1983) (1986) ALTERNATION Contrastive Contrastive Alternation alternation Supplementary Supple- Alternation mentary Alternation BONDING Coupling Joint Co-specify Coupling Contrastive Coupling Coupling State| Statement- Evidence Assert| Assert Exemplify Exemplify Exemplification Specify Exempli- fication Statement- Assert| Qualify Qualify/ Exception Qualify Refine List Classify Neutralise/ Distinction PARAPHRASE Paraphrase Restatement Restate Paraphrase Paraphrase/ Recapi- tulate/ Restate AMPLIFICA- TION Define Term Enablement Specify Specification Define Specification Add Predicate Elaboration/ Specify Specification Augment/ specification Enablement Counter/ Expand on/ Restrict Label SETTING Qualify/ Event/ Qualify Qualify Localise Describe State-location Qualify/ Event-direction Qualify Qualify Describe Qualify/ Event-manner Qualify Qualify Describe Bridge Zero Component Zero Rela- tion/ Function

Key: BOLD Categories of semantic functions (Crombie 1985) * indicates the absence of a corresponding term (Cooper & Matsuhashi 1983; Mann & Thompson 1986; Lindeberg 1988; Wong 1993) - indicates interacting rhetorical functions (Crombie 1985) | indicates two separate rhetorical functions (Cooper & Matsuhashi 1983; Lindeberg 1988) / indicates (an) alternative rhetorical function(s) (Mann & Thompson 1986; Kaldor et al. 1998)

39 The table shows that the systems devised by Cooper & Matsuhashi and Lindeberg are not as complex as others, with rather too few categories for analysing argumentative dis- course. On the other hand, the categorisations of Crombie, Wong and Kaldor et al. are too elaborate. It may be hard to locate incidence of the less common ones (such as in the case of ‘Denial-correction’ and ‘Event-direction’ in Crombie), and this will render part of the tool useless. Some of the categories in Kaldor et al. overlap with each other in ap- plication, for example, ‘Restrict’ and ‘Specify’ (Kaldor et al. 1998:47):

RESTRICT: Links earlier general statement to a specific case SPECIFY: Focuses on one member (hyponym) of a previously mentioned larger class or specify a particular aspect relating to the task

Though the definitions make clear the distinction, in their applications the two are not necessarily distinguishable. Among the six systems examined, Mann & Thompson’s (1986) scheme seems to be the most appropriate in terms of relative elaborateness (neither ‘over-’ nor ‘under- elaborate’). Yet the definitions of some of its categories are rather obscure, e.g. ‘Enable- ment’ (“R[eader] comprehending S[atellite] increases R’s potential ability to perform the action presented in N[ucleus]”) (Mann & Thompson 1987:54–6), which cannot be easily applied. The appropriateness of the rhetorical function inventory for the type of text is an im- portant issue. Since professional text was the only focus in Mann & Thompson’s study and part of the focus in Cooper & Matsuhashi’s, their taxonomies are less relevant to the research of students’ writing. That of Kaldor et al. was not selected because some cate- gories which are important for analysing tertiary student writing, such as Evaluation, were not included in it. The categorisation systems of Lindeberg and Crombie seem to be most relevant, as the genres they studied were closest to that of the present study. Neither Lindeberg’s nor Crombie’s taxonomy could however be directly adopted in this study, as Lindeberg was found to be ‘under-elaborate’ and Crombie ‘over-elaborate’. It was therefore decided that a new taxonomy would be adapted from Lindeberg and Crombie, to keep the merits of both systems, and at the same time strike a balance in terms of elaborateness. Another reason for the adaptation was that the analyses of Linde- berg and Crombie were only local in nature, which made it impossible for us to rely on them alone to examine hierarchical integration of argument.

Categorisation system for this investigation: Rhetorical Function Analysis

Rhetorical Function Analysis (RFA) allows us to analyse the rhetorical function of each discourse unit from the lowest up to the highest textual level. In this system, four lower levels are distinguished. They include Functional Unit/Simple Sentence (FU/SS) Level, Complex Sentence (CS) Level, as well as Propositional Segment (PS) Level. The upper levels comprise Paragraph (P) Level and Macrostructure (M) Level. The categorisation of these levels is in accordance with the theories presented earlier. By identifying the ar- gumentative structures at different textual levels and connecting them with the macro- structure, RFA extends the local approach taken by existing systems. The Functional Unit (FU) is the basic unit of analysis employed in this study, and hence Functional Unit/Simple Sentence (FU/SS) Level is also the lowest level of our analyses. A Functional Unit refers to either a clause or a simple sentence. In the case of the former, the clause either subsumes, or combines with, (an)other clause or clauses to

40 form a complex sentence. As an example of ‘subsumption’ from a Hong Kong student text, consider:

HKH2 – Paragraph 1 S1a Nowadays, television are common in every family b and they become a necessity in people’s daily life.

In this extract, the writer makes an Assertion in S1a, and presents the Result in S1b. Thus S1b is subsumed by S1a to form a complex sentence, whose rhetorical function is also an Assertion. Let us now examine an example of ‘combination’ from a low-rated text written by an Australia-based student:

AL1 – Paragraph 3 S16 He says that these studies vary in terms of the role of prior anger is important in determining aggressive behaviors. S17a If prior anger is a precondition of aggressive b then the effects to T.V. violence are in doubt.

S17a is the Condition of S17b, while S17b is the Result of S17a. Yet when these two clauses combine, they serve the rhetorical function of ‘elaborating’ S16, so in the graphic representation, the sign will be used in the graphic representation of RFA to em- brace these two clauses to show that they have formed a new rhetorical function (i.e. ‘Elaboration (16)’) at Complex Sentence Level. These examples show that no matter whether subsumption or combination takes place, a complex sentence results. Thus the second level in our categorisation is Complex Sentence (CS) Level. The concepts of ‘subsumption’ and ‘combination’ of rhetorical functions are important in our method of analysis, as these take place not only at the lowest discourse level (FU/SS), but also at every other level up to the paragraph. The Proposition Segment (PS) Level is the next level in our hierarchy. A sentence (either simple or complex) may subsume, or combine with, a neighbouring sentence or sentences to form a ‘meaning cluster’ (or a ‘propositional segment’) inside a paragraph, or an orthographic paragraph itself. In the case of the former, further subsumption(s) or combination(s) between propositional segments may in turn take place, to form a para- graph. The PS Level may therefore consist of several sub-levels, or only one level. At the upper textual levels, a distinction is made between the rhetorical function at Paragraph Level and that of Macrostructure Level, because they stand for different rela- tions. The ‘Rhetorical Function at Paragraph Level’ refers to the rhetorical relationship between two or more (usually neighbouring) paragraphs. These may feed into a distin- guishable macrostructure, that is, the global structure. In this case, we take the further step of judging the roles of all the paragraphs in the text (labelled as ‘Rhetorical Func- tions at Macrostructure Level’) from the perspective of how well they match up with one of the conventional global argumentative structures, the Argumentative Response Struc- ture identified by Tirkkonen-Condit (1989). The taxonomy of RFA is presented on the following page. The definitions and exam- ples of the rhetorical functions included in this taxonomy are presented together with the abbreviated forms used in the graphic representation, in Appendix 1.

41

Rhetorical Function Analysis (RFA)

Functions at the Lower Levels: Functions at the Upper Levels

FU/SS/CS/PS Level Paragraph Level

TRUTH & VALIDITY TRUTH & VALIDITY

*Assertion (Ass) *Evaluation (Eval)

Justification DISCOURSE DEVELOPMENT

Concession (Conc) Background (Back) *Evaluation (Eval) *Metatextual Statement (MeS) Reported Evaluation (RepE) *Elaboration (Elab)

CAUSE & EFFECT OTHERS

Cause (Cau) *Unidentified Rhetorical Function (UNID)

Result (Resu) *Unfulfilled Rhetorical Function (UNFU) Condition (Cond) Purpose (Purp)

DISCOURSE DEVELOPMENT

*Elaboration (Elab) Exemplification (Exem) Macrostructure Level Comparison (Comp) TRUTH & VALIDITY/ Coupling (Coup) DISCOURSE DEVELOPMENT Alternation (Alt) *Assertion (Ass) Summary (Sum) *Evaluation (Eval) Restatement (ReS) Question (Ques) Logical Conclusion (LCon) Answer (Ans) Chronological Frame (ChF) Problem (Pro) *Metatextual Statement (MeS) Solution (Sol)

OTHERS OTHERS

*Unidentified Rhetorical Function (UNID) *Unidentified Rhetorical Function (UNID) *Unfulfilled Rhetorical Function (UNFU) *Unfulfilled Rhetorical Function (UNFU)

Key: * Rhetorical Functions used for analysis purposes in both lower and upper levels.

42 Twenty-one rhetorical functions are used for the purpose of analysis at the lower textual levels. At the upper levels, six relations are employed at the Paragraph Level, and eight at the Macrostructure Level. A few of the functions for the lower levels are used recur- sively up to the paragraph, and in some cases above it, depending on the particular mac- rostructure. These include Assertion, Evaluation, Elaboration, Metatextual Statement, Unidentified Rhetorical Function and Unfulfilled Rhetorical Function, all marked with an asterisk in the foregoing list. At all levels, the rhetorical functions are grouped under Broad Categories, since the ‘Broad categories of semantic functions’, first introduced by Crombie (1985), are the most useful for organisational purposes. All these rhetorical functions are freestanding relations, which, compared with paired-up ones, provide more flexibility to the categorisation.

Categorisation at the lower levels

Because the genre under study is argumentative EAP essays, the two Broad Categories which are most relevant, namely, ‘Truth & Validity’ and ‘Cause & Effect’, are included at the lower levels. Besides these two Broad Categories, the category ‘Discourse Devel- opment’ is also included, as it is important for any genre. Rhetorical functions which cannot be included in any of the three Broad Categories mentioned above are grouped under ‘Others’. These include Unidentified Rhetorical Function (UNID) and Unfulfilled Rhetorical Function (UNFU). The term Unidentified Rhetorical Function is used to refer to a discourse unit whose rhetorical function cannot be worked out from the context. The following example is taken from AL1, an Australian low-rated essay. The discourse unit which has an unidentifiable rhetorical function has been bolded:

S6a The responsible parents are worried their children’s future, b factors which raises related to this subject.

The expression in S6b makes it impossible for the reader to guess what this clause means, and in turn to work out from the context the rhetorical relation of S6b with, for example, S6a. Therefore S6b was categorised as Unidentified Rhetorical Function (UNID). This problematic function is parallel to the ‘Non-function Units’ and ‘Zero Re- lation/Function’ in the categorisation system of Lindeberg (1988, 1994) and Kaldor et al. (1998) respectively. The key factor which helps us to understand the concept behind Unfulfilled Rhetori- cal Function (UNFU) is also the context. An Unfulfilled Rhetorical Function is a dis- course unit which signals to the reader through contextual clues that it is intended to serve a certain rhetorical function. The rhetorical function signalled has not been fulfilled, and thus cannot meet the expectation of the reader. This might be due to an erroneous choice of vocabulary, or a grammatical error, which causes a communication breakdown. A further example comes from essay AL1:

S4 Because children confronted too much time watching T.V. S5 The alarm was causes is T.V. violent effects on children are the issue.

Sentence 4 starts with the causal conjunction, Because. This gives the reader the expecta- tion that this sentence is the Cause of Sentence 5. Yet the wrong lexical choice of con- fronted in S4 makes the Cause incomprehensible. Therefore S4 was categorised as ‘UNFU [Cause (5)]’, that is, an ‘Unfulfilled Rhetorical Function’, which is presumably the Cause of S5. This problem is referred to as ‘Unsuccessful Relations’ in Stuart-Smith

43 & Busbridge’s (1996) categorisation system, and ‘Semantic/rhetorical mismatch’ in Kal- dor et al.’s (1998) scheme.

Categorisation at the upper levels

The Paragraph Level consists of six rhetorical functions. Among them, Evaluation is the only rhetorical relation grouped under ‘Truth & Validity’. Three functions are grouped under ‘Discourse Development’: Background, Metatextual Statement and Elaboration. Unidentified Rhetorical Function and Unfulfilled Rhetorical Function are also included at Paragraph Level, grouped under ‘Others’. This is because in some essays, the rhetori- cal function of a paragraph in relation to the other paragraphs cannot be worked out from the context, or the paragraph does not perform the rhetorical function the reader may ex- pect from the context. The categorisation at Macrostructure Level is based on the three sequences of ‘Ar- gumentative Response Structure’ (ARS) (Remark-Response, Question-Answer and Prob- lem-Solution) (Tirkonnen-Condit 1989). As mentioned earlier, argumentation can be triggered as a response to a remark, an assertion, or a claim. ‘Assertion’ has thus been added as a descriptor at this level. The use of ‘Evaluation’ at Macrostructure Level re- flects the Situation-Evaluation structure identified by Winter (1976). Hence the follow- ing rhetorical functions have been devised for the analysis at Macrostructure Level in this investigation: Assertion, Evaluation, Question, Answer, Problem and Solution. Simi- lar to Paragraph Level, Unidentified Rhetorical Function and Unfulfilled Rhetorical Function are included at Macrostructure Level, as some paragraphs fail to contribute to the formation of a distinguishable and/or appropriate macrostructure.

44 Chapter Five: Research design and methods for analysis

DATA COLLECTION

The data for the present study were collected between 1996 and 1997 from two groups of subjects: first-year students at Macquarie University, Sydney, who were enrolled in an ‘English for Academic Purposes’ (EAP) course in the Spring Semester, 1996, and their counterparts in Hong Kong, who took an EAP course at the Hong Kong University of Science and Technology (HKUST) in the Fall Semester, 1997. Thus the study took place in the first semester of each subject’s university life. There were both male and female subjects, aged between 18 and 21, and majoring in various fields of study. All the sub- jects, both in Sydney and Hong Kong, were ethnic Chinese, born and mostly schooled in Hong Kong. While the HKUST group was studying in Hong Kong (an EFL environ- ment), the Macquarie group had gone to Australia (an ESL environment) and begun to study at high school there three years prior to the present study. A three-year period of study in Australia was made one of the criteria for selecting the subjects because this cor- responds with the duration of time a student needs to spend to complete senior secondary education in Australia. This ensured that the Sydney-based students had all had equal duration of exposure to the ESL environment. The data collected comprised argumentative essays written by the subjects outside class, that is, not written under any time constraint. The essay question was: “Does watching violence on television have any effect on children’s behaviour?” The reference materials used by the subjects to answer the question were three articles, Freedman (1988), Milavsky (1988) and Singer & Singer (1988). The essay question was originally assigned for the Australian subjects as part of their course requirements. The help of the teacher who taught the EAP course at HKUST was enlisted to integrate the same as- signment into the course there. The groups of essays were not equivalent in length, the Australian ones averaging around 1,200 words, and those from Hong Kong some 600 to 800 words. This disparity was unavoidable, because of the conditions under which the data were collected and the different requirements of the courses. Since the essays in each cohort were different in length, in carrying out the statistical analyses, the percent- ages relative to the length of each script, instead of the raw number of occurrences, are compared. The essays collected were rated holistically, in each case by an independent L1 rater, who was the course tutor of the Australian and Hong Kong groups respectively. From each of these two groups, six high-rated (Grade ‘A’) essays as well as six low-rated (Grade ‘C’) ones were identified. The whole data set thus consisted of twenty-four es- says. The coding identifies each script by its place of origin, as well as the grade it has received from the rater, high versus low. Thus AL1, for example, stands for Australian low-rated script 1, and HKH6 Hong Kong high-rated script 6.

RESEARCH QUESTIONS

There were three objectives in the current study. The first was to examine how L2 terti- ary students build up their universe of argumentative discourse, from the lowest to the highest. To carry out this investigation, Rhetorical Function Analysis (RFA), as de- scribed in Chapter 4, was devised. A comprehensive demonstration of the method is pro- vided below. The method was validated by four independent raters.

45 A different kind of control on the method was to employ ‘triangulation’, the findings validated by reference to those of an independent method, Grammatical Accuracy Analy- sis (GAA), described and demonstrated below. An analysis of the students’ grammatical accuracy is also highly relevant to each one’s overall writing competence. It was there- fore of interest to investigate whether the results obtained from GAA complemented, or conflicted with, those derived from RFA. The second research objective involved comparisons to determine whether being educated in an ESL environment would bring about differences in (i) argumentative structuring and (ii) grammatical accuracy in the essays these students produced, when compared with data from an EFL setting. The two research questions raised earlier in Chapter 2 were rephrased as follows, based on the literature we surveyed in Chapter 4:

Question 1

Do the argumentative essays (both high-rated and low-rated) written by first-year Australian L2 university students display a higher level of hierarchical integra- tion of argument than those written by their Hong Kong counterparts?

Question 2

Do the argumentative essays (both high-rated and low-rated) written by first-year Australian L2 university students display a greater degree of grammatical accu- racy than those written by their Hong Kong counterparts?

The third objective of the study was to examine the correlation between the grades awarded to students’ writing and the level of organisation and grammatical competence evidenced in these essays. Similarly, the two research questions raised in Chapter 3 were rephrased on the basis of the literature review presented in Chapter 4:

Question 3

Do the high-rated argumentative essays written by first-year Australian L2 and Hong Kong university students display a higher level of hierarchical integration of argument than the low-rated essays?

Question 4

Do the high-rated argumentative essays written by first-year Australian L2 and Hong Kong university students display a greater degree of grammatical accu- racy than the low-rated essays?

46 RHETORICAL FUNCTION ANALYSIS (RFA): METHOD FOR INVESTIGATING THE HIERARCHICAL INTEGRATION OF ARGUMENT (RESEARCH QUESTIONS 1 AND 3)

Segmentation of text

The ‘Functional Unit’ was chosen as the segmentation unit for this study. For text analy- sis to be carried out, each essay in the data sets was first segmented into sentences, and numbered according to the student writer’s punctuation. The sentences were in turn seg- mented into F-units, according to Lieber’s segmentation principles, presented in Chapter 4. Each of the F-units was assigned a letter (a, b, c, etc.) for identification purposes. To illustrate the segmentation principles, part of Paragraph 1 of HKH4, has been segmented into F-units as below. (The full text of HKH4 is included in Appendix 2.)

HKH4 – Paragraph 1 1 The question whether TV viewing has a negative effect on children’s behaviour is controversial. 2 There are two points of view. 3 One of these is that television viewing of violent programming leads to aggressive behaviour of children. 4 This is extracted from ‘Some Hazards Of Growing Up In A Television Environment’ by Jerome L. Singer & Dorothy G. Singer. 5 And the other is that there is no direct relationship between TV viewing and aggres- sive behaviour. 6 This is extracted from ‘Television Violence And Aggression’ by Jonathan L. Freed- man.

Categorisation of rhetorical functions

The second step of Rhetorical Function Analysis was to categorise the F-units according to their rhetorical functions, using the taxonomy developed for this purpose presented in Chapter 4. A sample analysis of rhetorical functions at all levels of HKH4, together with the full text, is included in Appendix 2. In the next section, extracts from the full analysis will be used to illustrate the graphic representation of RFA, whereas the complete ‘Summary of Rhetorical Function Analysis’ of HKH4 will also be shown.

Graphic representation of Rhetorical Function Analysis

To present the results of RFA graphically, after the rhetorical functions (RFs) at FU/SS or CS Level had been identified checks were made to see whether these RFs were sub- sumed by, or combined with, one another to form a RF at a higher level, PS Level, that is, whether these RFs were members of larger rhetorical relations. The RFA of Paragraph 6, HKH4, provides a good illustration:

HKH4 – Paragraph 6 34 In conclusion, Singer & Singer suggested that heavy television viewing puts children at risk of increased aggression and restlessness at least in the short run. 35a On the contrary, Freedman suggested that all the experiments have many variables. b so we cannot conclude that there is a direct relationship between TV violence and aggressive behaviour.

47 36 Above all, I will agree with Freedman. 37 It is because in different experiments, they have different outcomes. 38 Sometimes, the outcomes contradict our assumption. 39 Moreover, increased aggressive behaviour may be due to, let say, insufficient sleep. 40a Therefore, we can’t even convince ourselves b to draw a fair conclusion c and prove that there is a causal link between TV violence and aggressive behaviour.

HKH4 – Rhetorical Function Analysis

Level PARAGRAPH 6

M Ans (1)

P Eval (1–35 : Para. 1–5)

PS Sum Eval (1–33) (1–35)

CS Comp LCon (34) (36–39)

FU/SS Sum Just LCon Eval Just Elab Coup LCon Purp Coup (1–35) (35b) (35a) (1–35) (36) (37) (38) (36–39) (40a) (40b)

Sent 34 35a 35b 36 37 38 39 40a 40b 40c

Key: Ans Answer Elab Elaboration Purp Purpose Comp Comparison Eval Evaluation Sum Summary Coup Coupling Just Justification Sent Sentence LCon Logical Condition

Sentence 34 (S34) gives a summary of the view of Singer & Singer, made in Sentences 1–33. Since it is a simple sentence, it was categorised as ‘Summary (1–33)’ at Simple Sentence (SS) Level. S35, on the other hand, is a complex sentence, consisting of two functional units, S35a and S35b. At Functional Unit (FU) Level, S35a provides a justifi- cation for the logical conclusion made in S35b. Yet when these two clauses combine at Complex Sentence (CS) Level, they serve as a contrast to S34, summarising the view of Freedman, which opposes that of Singer & Singer. So the branching is used to em- brace S35a and S35b to show that they have formed a new rhetorical function at CS Level, i.e. ‘Comparison (34)’. Since S35 is used to contrast to S34, it is subsumed by S34 to form the meaning cluster of ‘Summary (1–33)’ at Propositional Segment (PS) Level. Graphically, the subsumption is indicated by the brace. S36 to S40 form a much bigger meaning cluster. In S36, the writer is evaluating his/her earlier arguments. This sentence was thus categorised as ‘Evaluation (1–35)’. S37 provides a justification for S36, while S38 and S39 elaborate S37. Finally, a logical conclusion is made in S40. Therefore all these three sentences (S37 to S40) are

48 clusion is made in S40. Therefore all these three sentences (S37 to S40) are subsumed by S36 to form ‘Evaluation (1–35)’ at PS Level. The next step of RFA is to check whether further subsumption takes place to form the RF at Paragraph (P) Level, or yet another RF at PS Level. Since S36–S40 provide an overall evaluation to every argument presented so far, including the summary made in S34–35, they naturally subsume S34–35. The whole paragraph thus serves the function of ‘Evaluation (1–35)’. In this paragraph, the rhetorical relations are found to be part of larger rhetorical relations, which are in turn members of even larger relations. The RFA of Paragraph 6 also illustrates the fact that the RF at P Level refers to the rhetorical function of a paragraph in relation with one or more other paragraphs – usually neighbouring. For example, Paragraph 6 was categorised as ‘Evaluation’ of S1–35, or Paragraphs 1–5. Yet this paragraph is in turn “a member of an overall relation”, in its attempt to answer the question raised at the beginning of the essay, “whether TV viewing has a negative effect on children’s behaviour is controversial” (S1). Therefore its RF at M Level was categorised as ‘Answer (1)’, as we were judging the rhetorical function of this paragraph from the perspective of the macrostructure of the text. Since no subsump- tion or combination of rhetorical functions takes place at M Level, only the categorisa- tion was put down, but not the signs for subsumption/combination. The Paragraph Level is the level at which the categorisations for the lower and upper levels connect. We will use the RFA of Paragraph 3, HKH4, to demonstrate this point:

HKH4 – Paragraph 3 14 However, Freedman (1986) has a different opinion. 15 He said different laboratory experiments have different outcomes. 16 Some of them are based on some factors, like anger. 17a The aggression may occur b just because the person is angry. 18 Anger is apparently necessary in some studies and not in others. 19a Therefore, he thought that we can’t make sure the aggressive behaviour of chil- dren is due to anger or actually the heavy TV viewing. b so there is no direct causal link between TV violence and aggressive behaviour.

S14 is a ‘Comparison’ of S11–S13 (i.e. Paragraph 2), since it states that Freedman does not agree with Singer & Singer. S15, which relates Freedman’s opinion, is a ‘Reported Evaluation’ of S14. S16–S17 elaborate S14 (by introducing the role played by anger in some experiments), while S18–S19 present Freedman’s evaluation and conclusion re- spectively. At PS Level, S15–S19 combine together and form ‘Elaboration (14)’ and are thus subsumed by S14. Therefore, Paragraph 3 primarily provides a contrast with its neighbouring paragraph (Paragraph 2). Yet if we trace further, we will find that Para- graph 3 elaborates S1–S10, especially S8–S10, which state:

8 However, they all use similar experiments to do the research. 9 They are laboratory experiments, field experiments and natural experiments. 10a In this paper, I will compare these three types of experiments one by one b and make a judgement at the end.

49

HKH4 – Rhetorical Function Analysis

Lev el PARAGRAPH 3

M Ans (1)

P Comp = Elab (12–13) (1–10 : para. 1)

PS Comp Elab (12–13) (14)

CS Elab LCon (16) (15–17)

Comp RepE Elab Elab Cau RepE LCon LCon FU/SS (12–13) (14) (15) (16) (17a) (17) (15–17) (19a)

Sent 14 15 16 17a 17b 18 19a 19b

Key: Ans Answer Elab Elaboration RepE Reported Evaluation Cau Cause Eval Evaluation Sent Sentence Comp Comparison LCon Logical Condition

Therefore in the graphic representation, the primary function of this paragraph, ‘Com- parison (11–13: Para. 2)’, is linked with its ‘secondary function’, ‘Elaboration (1–10: Para. 1)’, by the = sign. Since this sign can indicate the connection between the categori- sations for the lower/intermediate and upper levels, it is important in the graphic repre- sentation of RFA. In compiling the frequency of RFs at P Level, only the ‘secondary function’ of a paragraph was counted. The analysis at PS Level of Paragraph 3 provides illustration of a sign which was used for graphic clarity: the signs and were used to indicate subsumption and combination of rhetorical functions respectively. In Paragraph 3, a new sign, was put above S14, to show that unlike S15–S19, this rhetorical function has not combined with, or subsumed, another function.

50 Summary of the findings of Rhetorical Function Analysis

The graphic representation of RFA can show clearly whether clause relations are mem- bers of larger rhetorical relations. However, in order to compare one script with another, we needed to find a method to summarise the findings of the RFA of each script. There- fore, a ‘Summary of Rhetorical Function Analysis’ was prepared for each. This summary provides evidence or support for the analysis and presents the Macrostructure Analysis:

Part I: Basic information on Functional Units and Rhetorical Functions Part II: Problems of Rhetorical Functions at Functional Unit/Simple Sen- tence/Complex Sentence Level Part III: Rhetorical Functions at Propositional Segment Level and Paragraph Level Part IV: Macrostructure Analysis.

Part I: Basic information on Functional Units and Rhetorical Functions

This section presents summary information from each script, including its total number of words and sentences. Four tables are used to summarise the information of the rhetori- cal units and rhetorical functions. The first shows the number of sentences respectively consisting of 1 F-unit, 2 F-units, 3 F-units, and so on. The other three tables indicate what sorts of rhetorical functions were identified at different levels. The following is the set of summaries tabulated for HKH4, for illustration purposes.

Part I: Basic information on Functional Units and Rhetorical Functions: HKH4

Total number of words: 686 Total number of sentences: 40 Total number of Functional Units: 54

Table10: Make-up of sentences in terms of functional units, Script HKH4. Types of sentence Number Percentage Sentences consisting of 1 Functional unit 28 70% Sentences consisting of 2 Functional units 10 25% Sentences consisting of 3 Functional units 2 5% Total 40 100%

Most of the sentences in this essay are simple sentences. Even the complex sentences consist of only two clauses (cf. Table 11). Tables 12 to 14 show that all the rhetorical functions at various levels in HKH4 are sound. Elaboration (28.4%) and Logical Conclu- sion (11.11%) are the most frequently occurring functions at the lower levels, suggesting that the writer is able to elaborate ideas and draw interim conclusions. The Elaboration strategy is extended to Paragraph Level, as evidenced by the high percentage of Elabora- tion there (57.14%). At Macrostructure Level, only two types of rhetorical functions were identified, Question (16.67%) and Answer (83.33%). The macrostructure of the es- say is thus ‘Question-Answer’, which matches one of the conventional global structures of argumentative text.

51 Table 11: Frequency of rhetorical functions identified at the lower levels, Script HKH4. Total FU/SS CS PS Rhetorical Function frequency in Level Level Level each category TRUTH & VALIDITY 1 1 2 Assertion 0 (1.85%) (6.67%) (2.47%) 2 2 Justification 0 0 (3.70%) (2.47%) 2 2 Concession 0 0 (3.70%) (2.47%) 4 2 6 Evaluation 0 (7.41%) (13.33%) (7.41%) 6 3 1 10 Reported Evaluation (11.11%) (25%) (6.67%) (12.35%) CAUSE & EFFECT 1 1 Cause 0 0 (1.85%) (1.23%) 2 2 Result 0 0 (3.70%) (2.47%) 1 1 Condition 0 0 (1.85%) (1.23%) 1 1 Purpose 0 0 (1.85%) (1.23%) DISCOURSE DEVELOPMENT 14 3 6 23 Elaboration (25.93%) (25%) (40%) (28.40%) 1 1 2 Exemplification 0 (1.85%) (6.67%) (2.47%) 4 3 3 10 Comparison (7.41%) (25%) (20%) (12.35%) 4 4 Coupling 0 0 (7.41%) (4.94%) Alternation 0 0 0 0 1 1 2 Summary 0 (1.85%) (6.67%) (2.47%) Restatement 0 0 0 0 7 2 9 Logical Conclusion 0 (12.96%) (16.67%) (11.11%) 2 2 Chronological Frame 0 0 (3.70%) (2.47%) 1 1 2 Metatextual Statement 0 (1.85%) (8.33%) (2.47%) Total FU/SS CS PS Rhetorical Function frequency in Level Level Level each category OTHERS Unidentified 0 0 0 0 Rhetorical Function Unfulfilled 0 0 0 0 Rhetorical Function Total frequency 54 12 15 81 at each level (100%) (100%) (100%) (100%)

52 Table 12: Frequency of Rhetorical Functions identified at Paragraph Level, Script HKH4. Rhetorical Function Paragraph Level TRUTH & VALIDITY Evaluation 1 (14.29%) DISCOURSE DEVELOPMENT Background 1 (14.29%) Elaboration 4 (57.14%) Metatextual Statement 1 (14.29%) OTHERS Unidentified Rhetorical Function 0 Unfulfilled Rhetorical Function 0 Total frequency at each level 7 (100%)

Table 13: Frequency of Rhetorical Functions identified at Macrostructure Level, Script HKH4. Rhetorical Function Macrostructure Level TRUTH & VALIDITY Assertion 0 Evaluation 0 DISCOURSE DEVELOPMENT Question 1 (16.67%) Answer 5 (83.33%) Problem 0 Solution 0 OTHERS Unidentified Rhetorical Function 0 Unfulfilled Rhetorical Function 0 Total frequency at each level 6 (100%)

Part II: Problems of Rhetorical Functions at Functional Unit/Simple/Complex Sentence Levels: AL1

7 The question is are T.V. violent contribute aggression. UNFU [Assertion]

Grammatical problem (This sentence may be rewritten as: “The question is whether T.V. violence contributes to children’s aggression.”)

8 The concept was argued among academics prominent sociological and psychological re- searchers. UNFU [Elaboration (7)]

Referential problem due to wrong choice of word, which confuses the reader as to what the word “concept” refers to. (It may be rewritten as: “The question was argued among academ- ics, prominent sociological and psychological researchers.”)

Key: [ ] the Rhetorical Function presumed from the context, yet unfulfilled due to grammatical mistakes or referential problems

The problematic rhetorical relations are often brought about by grammatical errors, and are categorised separately in the Grammatical Accuracy Analysis (GAA), presented at the end of this chapter.

53 Part III: Rhetorical Functions at Propositional Segment Level and Paragraph Level

One of the purposes of this part of the summary is to record the rhetorical functions at the highest Propositional Segment Level, to show whether they are members of larger rhetorical relations at Paragraph Level.

Part III: Rhetorical Functions at Propositional Segment Level and Paragraph Level: HKH4 – Paragraphs 1 to 6

Problems RFs at high- at RFs at Problems at Para. Focus est PS Level highest P Level P level PS Level

Background Assertion Background P1 + + + Nil Nil (1–11) Metatextual Metatextual Metatextual Statement Statement Statement

Elaboration Laboratory (1–10)* Elaboration P2 experiments: ‘Pro’ + Nil (1–10: (11–13) TV effects – Singer Reported Nil Para.1)* & Singer Evaluation (11)

Comparison Comparison Laboratory (12–13)* (12–13)* = P3 experiments: ‘Con’ + Nil Elaboration Nil (14–19) TV effects – Elaboration (1–10: Freedman (15) Para.1)

Elaboration Field experiments: (1–10)* + Elaboration P4 ‘Pro’ & ‘Con’ TV Reported Nil (1–10: Nil (20–26) effects – Singer & Evaluation Para.1)* Singer vs Freedman (20) + Com- parison (22)

Elaboration Natural (1–10)* + experiments: ‘Pro’ & Reported Elaboration P5 ‘Con’ TV effects – Evaluation Nil (1–10: Nil (27–33) Singer & Singer vs (27) + Com- Para.1)* Freedman parison (28– 30)

Summary P6 Conclusion: ‘Con’ (1–33) + Evaluation Nil (34–40) TV effects Evaluation (1–35) Nil (1–35)

Key: * Rhetorical Functions built up and sustained from Local Levels to Intermediate Level, and finally to Paragraph Level

54 The first column of the table lists the paragraphs and sentences analysed. The second is a summary of the main focus of a certain paragraph. The third column presents the sub- sumption or combination of rhetorical functions at the highest Propositional Segment (PS) Level. In HKH4, there is no problem in the combinations/subsumptions of rhetori- cal functions at PS Level. Therefore, sound relations in turn emerge at Paragraph (P) Level. If there are problematic rhetorical relations at PS Level, the ‘RF at P Level’ will turn out to be an unsound relation as well. These problems are recorded in Columns 4 and 6, so as to explain clearly to the reader how well, or how poorly, the arguments of an essay are integrated. To illustrate the usage of these two columns, the summary of Para- graph 1, AL1, is included below:

Part III: Rhetorical Functions at Propositional Segment Level and Paragraph Level: AL1 – Paragraph 1

Problems at RFs at High- RFs at P Problems at P Para. Focus Highest PS est PS Level Level Level Level Expression prob- S4 should lems (e.g. S7, have been S10, S11) and Assertion Background written as a referencing prob- + Background + clause and lems (e.g. S8) P1 (1–11) Result (1–2) + Metatextual combined blur the rhetorical + UNFU Statement with S5 to function of the UNFU form S4a & second half of 4b this paragraph (S7–S11)

The first half of Paragraph 1 presents the background of the whole essay, yet the expres- sion and referencing problems in the second half of this paragraph, S7–S11, make it dif- ficult for the analyst to identify its rhetorical function. Therefore the ‘RFs at P Level’ are ‘Background + UNFU’. All the underlying reasons for the labelling of UNFU are re- corded in the last column, ‘Problems of RFs at P Level’, which provides easy reference for the reader.

Part IV: Macrostructure Analysis

Part IV: Macrostructure Analysis – HKH4 Paragraph RF at Paragraph Level RF at Macro Level 1 Background + Metatextual Statement Question 2 Elaboration (1–10: Para.1) Answer (1) 3 Comparison (12–13) = Elaboration (1–10: Para.1) Answer (1) 4 Elaboration (1–10: Para.1) Answer (1) 5 Elaboration (1–10: Para.1) Answer (1) 6 Evaluation (1–35) Answer (1)

The Macrostructure Analysis has two functions: firstly, to show whether the macrostruc- ture of a text is made up of sound rhetorical relations at Paragraph Level which are con- ducive to the integration of arguments; and secondly, to examine whether the macro- structure identified matches any of the conventional global structures of argumentative essays, ‘Question-Answer’, ‘Remark-Response’, or ‘Problem-Solution’ (Tirkkonen- Condit 1989). The Macrostructure Analysis of HKH4 is presented as above.

55 The question of “whether TV viewing has a negative effect on children behaviour” is raised in Paragraph 1. Therefore the ‘RF at M Level’ of this paragraph was categorised as ‘Question’. The writer attempts to provide an answer in Paragraphs 2–6, by relating and evaluating the evidence presented by two researchers. Therefore all these paragraphs were categorised as ‘Answer’. The macrostructure of this essay thus matches one of the conventional global argumentative structures, ‘Question-Answer’. This partly accounts for the high grade it received.

Validation of Rhetorical Function Analysis

Discourse analysis methods, including RST and RFA devised for this research, involve judgements on the part of analysts. They are, as Mann & Thompson (1987:4) point out, “judgments of plausibility rather than certainty”. Although the decisions are based on predetermined categorisations, they cannot be considered to be “judgments of certainty”. So in order to ensure that the diagnoses made by RFA are at least “judgments of plausi- bility”, this method was validated in May 1998. Four raters, all experienced ESL teach- ers teaching at HKUST Language Centre, were invited to use RFA to analyse four texts, one from each data set, namely, AH4 (a high-rated Australian text), AL2 (a low-rated Australian text), HKH1 (a high-rated Hong Kong text), and HKL3 (a low-rated Hong Kong text). This selection was made so that the data for the validation covered the range of high-rated and low-rated essays written in both Australia and Hong Kong. A training session was conducted, during which the rationale behind RFA, as well as its operation methods, were thoroughly explained. Then the raters applied RFA on Para- graphs 1 and 2 of HKH1. As noted by several researchers (e.g. Mann & Thompson 1988), ‘good’ texts are always analysable. Starting with a well written text would therefore help the raters to understand the operation methods. The inter-rater reliability of RFA was 0.79. Hatch & Lazaraton (1991) suggest that for inter-rater reliability, a “strong” correlation should be taken as 0.8. Thus we consid- ered the inter-rater reliability of RFA to be sufficiently strong, and satisfactory for a new method. We have seen from this section that RFA, can meet Albrechtsen et al.’s (1991:81) expectation of being “sensitive to the interplay between local and global coherence”, and thus reveal the hierarchical integration of argument clearly.

GRAMMATICAL ACCURACY ANALYSIS (GAA): METHOD FOR INVESTIGATING SYNTACTIC ACCURACY (RESEARCH QUESTIONS 2 AND 4)

Previously used methods for studying grammatical accuracy in student writing

This section reviews several methods of categorising grammatical problems: Bunton (1989), Milton (2001) and Peters & Gassmann (1995). Bunton’s data were collected from essays written by his Form 5 to 7 students in Hong Kong. The errors presented are not categorised or ranked, but only grouped according to type in the second part of the book, where exercises are given. Bunton (op. cit.:Preface) admits that

The dictionary-style format does not allow for the inclusion of errors that cannot be linked to a key word … The main categories that cannot be included in this format are mistakes with tenses and subject-verb agreement …

56 This categorisation was not considered systematic and exhaustive enough for the pur- poses of the current study. Milton’s taxonomy was devised on the basis of two large Hong Kong corpora. Since his study set out to “identify the main variant features of the written interlanguage … of Hong Kong students of English” (Milton op. cit.:Summary), the categorisation of his taxonomy had to be very fine-grained. However, the main focus of the comparisons made in the present enquiry was more concerned with essay quality in terms of overall syntactic competence, which makes a fine breakdown of error types unnecessary. We therefore decided not to base our scheme on Milton’s categorisation. The taxonomy on which we chose to model our scheme was the tool used in Peters & Gassmann (1995) to carry out a large-scale study which aimed at investigating the acqui- sition of writing skills of students advancing through the high school system in Australia.

Categorisation system for this investigation: Grammatical Accuracy Analysis

In developing the taxonomy for the present study, Grammatical Accuracy Analysis (GAA), adaptations were made to Peters & Gassmann’s scheme. Instead of consisting of as many as 49 “problem subclasses” grouped under four broad categories (‘Spelling’, ‘Morphology’, ‘Punctuation’ and ‘Grammar’), the errors were revised into many fewer categories in GAA. On the other hand, we created new categories, because of our need to identify common errors made by Hong Kong students:

Wrong Word Redundant Word Conjunctive Devices Sequencing, and Total Communication Breakdown.

The taxonomy of GAA consists of nineteen categories. The definitions and examples of error types included in this taxonomy are detailed in Appendix 3.

Taxonomy of Grammatical Accuracy Analysis

① Derivational Morphology ② Spelling ③ Tenses ④ Number Agreement ⑤ Nonfinite Forms ⑥ Voice ⑦ Wrong Word ⑧ Missing Word ⑨ Redundant Word ⑩ Collocation ⑪ Articles ⑫ Conjunctive Devices ⑬ Anaphoric/Cataphoric Pronouns ⑭ Sequencing ⑮ Word Punctuation

57 ⑯ Internal Punctuation within & between phrases or clauses ⑰ Sentence-final Punctuation ⑱ Tangled Construction ⑲ Total Communication Breakdown.

To answer our Research Questions 2 and 4, the errors in each script were analysed ac- cording to GAA. For the purposes of convenience and neatness, the numbers assigned to each error category, as above, were used in the analysis instead of the spelling in full. The errors were then counted and tabulated. When the same error recurred in the same script, as in the case of the word problem being misspelt as *problen for the second or third time, the error was only counted once. The application of GAA is more straight- forward than RFA because its categorisation is based strictly on grammatical entities. The following is a demonstration of GAA, conducted on the first two paragraphs of AL1, the low-rated Australian essay we examined earlier:

AL1 – Grammatical Accuracy Analysis

Paragraph 1

1. Television is becoming a part of our social life, a source of recreation and entertainment. ① 2. It provides a variety of programs such as violence and non-violence, real and fantasy, comedy and tragic stories. ⑧ 3a T.V. is so popular∧ b every home at least has a set in the room. ⑨ ⑦ 4 Because children confronted too much time watching T.V. ⑦ ⑨ ⑧ ⑧ ① ⑧ 5 The alarm was∧ causes∧ is T.V. violent ∧ effects on children. ⑧ 6a The responsible parents are worried∧ their children’s future. ⑧ ⑨ ⑤ ⑧ b ∧ factors which raises∧ related to this subject. ⑫ ④ ⑧ 7 The question is are T.V. violent contribute∧ aggression. ⑦ ③ ⑫ ① 8 The concept was argued among academics∧ prominent sociological and ① psychological researchers. ⑤ ⑨ ⑨ 9 The evidence and information present by these experts finding are of different ⑨ opinions. ④ ④ ⑨ 10a The experiment and sources obtained was in a separate categories and patterns. ⑨ ⑯ b These involved in laboratory research∧ field experiments and correlational studies in the field.

58 ③ 11a We examine some articles ⑤ ⑮ ④ ⑧ b gather evidence from these expert researchers finding∧ results.

Paragraph 2 ③ ① 12a First we look at Freedman, an author and psychologist researcher ⑧ b ∧writes in his article “Television violence and aggression: what the evidence shows” (1988).

13 Jonathan Freedman argues that there is little or no positive evidence to support the ⑤ assertion that T.V. violence led to increases aggression. ④ 14a Freedman look at three types of studies ⑯ ④ b to support his argument, laboratory research, field experiment and natural ① experiment studies in the field.

Key: repeated error, which has not been counted again ① Derivational Morphology ③ Tenses ④ Number Agreement ⑤ Nonfinite Forms ⑦ Wrong Word ⑧ Missing Word ⑨ Redundant Word ⑫ Conjunctive Devices ⑮ Word Punctuation ⑯ Internal Punctuation

The most recurrent error identified in the extract above is Missing Word, whose fre- quency is as high as 10 in these two paragraphs, which contain only about 250 words. The other recurrent error categories include Wrong Word, Redundant Word, Number Agreement, Tenses and Derivational Morphology, each of which has occurred three or four times in this short extract.

Validation of Grammatical Accuracy Analysis

GAA was validated in the same way as for RFA before the main study was carried out. The raters used for RFA were involved again, and the same four scripts employed for standardisation (AH4, AL1, HKH1 and HKL3). Compared with the RFA training session, the raters had fewer queries regarding the application of GAA. This was unsurprising, since as seasoned English teachers, the raters will have had a great deal of experience in judging the grammatical accuracy of texts, whereas carrying out discourse analysis through RFA was new to them. The inter-rater reliability of GAA, 0.85, was even higher than that of RFA (0.79). Following Hatch & Lazaraton’s (1991) suggestion, the inter-rater reliability of GAA is strong. The fact that GAA has generated a higher agreement among raters was not sur- prising, for the reason mentioned above. Compared to the other criteria for rating essays, grammatical accuracy is also relatively easier to judge.

59 HYPOTHESES OF THE PRESENT INVESTIGATION

The above is an account of the research design and methodology of the present investiga- tion. With these developed and validated, we can formulate four hypotheses derived from the research questions raised at the beginning of this chapter:

Hypothesis 1

The argumentative essays (both high-rated and low-rated) written by first-year Australian L2 university students do not display a higher level of hierarchical integration of argument than those written by their Hong Kong counterparts.

Hypothesis 2

The argumentative essays (both high-rated and low-rated) written by first-year Australian L2 university students do not display a greater degree of grammatical accuracy than those written by their Hong Kong counterparts.

Hypothesis 3

The high-rated argumentative essays written by first-year Australian L2 and Hong Kong university students do not display a higher level of hierarchical in- tegration of argument than the low-rated essays.

Hypothesis 4

The high-rated argumentative essays written by first-year Australian L2 and Hong Kong university students do not display a greater degree of grammatical accuracy than the low-rated essays.

60 Chapter Six: Results and discussion: Rhetorical Function Analysis

INTRODUCTION

This chapter presents the findings of Rhetorical Function Analysis (RFA), used to com- pare the level of hierarchical integration of argument, firstly, between the Australian and Hong Kong essays (Hypothesis 1), and secondly, between the high-rated and low-rated essays (Hypothesis 3). The findings were analysed by Poisson Regression, to determine whether there was any statistical difference among the four groups of essays. The level of significance was set at 0.05 (cf. Seliger & Shohamy 1989:220).

HYPOTHESIS 3

The hypothesis

The high-rated argumentative essays written by first-year Australian L2 and Hong Kong university students do not display a higher level of hierarchical integration of argument than the low-rated essays.

To test this hypothesis, I had to investigate whether there were any differences between the high-rated and low-rated essays in terms of their percentages of

• rhetorical relations at Functional Unit (FU)/Simple Sentence (SS)/Complex Sentence (CS) Level which are themselves members of larger rhetorical relations at Propositional Segment (PS) Level; • rhetorical relations at Propositional Segment Level which are themselves mem- bers of larger rhetorical relations at Paragraph (P) Level; and • rhetorical relations at Paragraph Level which are themselves members of larger rhetorical relations at Macrostructure (M) Level.

All the rhetorical relations mentioned refer to identifiable and fulfilled relations, since only sound rhetorical relations at a lower discourse level bring about sound relations at a higher level, and thus contribute to the hierarchical integration of argument. To compare the high-rated and low-rated essay groups at each level, we adjusted the total number of subsumptions and/or combinations of sound rhetorical relations by the total number of subsumptions/combinations of all types of rhetorical relations (including unsound relations, UNID and UNFU), identified at the same level; that is,

Total number of subsumptions/combinations of sound RFs * Total number of subsumptions/combinations of all RFs **

Note * This excludes the subsumptions/combinations which involve UNID and UNFU ** This includes the subsumptions/combinations which involve UNID and UNFU

61 Results

Integration of rhetorical relations at Functional Unit/Simple Sentence/Complex Sentence Level into rhetorical relations at Propositional Segment Level

Table 14: Testing of Hypothesis 3 at Propositional Segment Level: high-rated = low-rated. Average of percentages of Range of percentages subsumptions/combinations of Essay Standard of subsumptions/ sound rhetorical functions groups deviation combinations of sound (per total number of rhetorical functions all rhetorical functions) H (n = 12) 100% 0 0 L (n = 12) 69.97% 0.1872 0.5714

Key: H High-rated essay group (both Hong Kong & Australian L2) L Low-rated essay group (both Hong Kong & Australian L2)

In the high-rated essay groups, the average of the percentages of the subsump- tions/combinations of sound rhetorical relations (per total number of the subsumtions/ combinations of all rhetorical relations) is 100%. The standard deviation, as well as the range of these percentages, is 0. This means that the high-rated essays are homogeneous, in that all their rhetorical functions at Functional Unit/Simple Sentence/Complex Sen- tence Level which are subsumed by, or combine with, one another are sound ones. In the low-rated essay groups, the average percentage is only 69.97%. The standard deviation is 0.1872, and the range 0.5714, suggesting that compared with the high-rated essays, there is a greater variation in terms of integration of arguments within the low-rated groups. Since the difference between the high-rated and low-rated essay groups is highly signifi- cant (P < 0.0001), the null hypothesis was rejected. Thus, we conclude:

Compared with the low-rated essays, the high-rated argumentative essays written by first- year Australian L2 students and their Hong Kong counterparts have a larger percentage of rhetorical relations at Functional Unit/Simple Sentence/Complex Sentence Level which are themselves members of larger rhetorical relations at Propositional Segment Level.

Integration of rhetorical relations at Propositional Segment Level into rhetorical relations at Paragraph Level

Table 15: Testing of Hypothesis 3 at Paragraph Level: high-rated = low-rated. Average of percentages of Range of percentages subsumptions/combinations of Essay Standard of subsumptions/ sound rhetorical functions groups deviation combinations of sound (per total number of rhetorical functions all rhetorical functions) H (n = 12) 100% 0 0 L (n = 12) 43.65% 0.1843 0.5458

Key: H High-rated essay groups (both Hong Kong & Australian L2) L Low-rated essay groups (both Hong Kong & Australian L2)

62 In the high-rated essay groups, the average of the percentages of the subsump- tions/combinations of sound rhetorical relations (per total number of the subsump- tions/combinations of all rhetorical relations) is 100%. The standard deviation, as well as the range of these percentages, is 0. This means that the high-rated essays are homogene- ous, in that all their rhetorical functions at Propositional Segment Level which are sub- sumed by, or combine with, one another are sound ones. In the low-rated essay groups, on average, only 43.65% of their rhetorical functions at PS Level which are subsumed by, or combine with, one another are sound ones. The standard deviation is 0.1843, and the range 0.5458, suggesting that compared with the high-rated essays, the low-rated ones are not so homogeneous in terms of integration of arguments. Since the difference be- tween the high-rated and low-rated essay groups is highly significant (P < 0.0001), the null hypothesis was rejected. Thus, we conclude:

Compared with the low-rated essays, the high-rated argumentative essays written by first- year Australian L2 students and their Hong Kong counterparts have a larger percentage of rhetorical relations at Propositional Segment Level which are themselves members of larger rhetorical relations at Paragraph Level.

Integration of rhetorical relations at Paragraph Level into rhetorical relations at Macrostructure Level

Table 16: Testing of Hypothesis 3 at Macrostructure Level: high-rated = low-rated. Average of percentages of Range of percentages of subsumptions/combinations of Essay Standard subsumptions/ sound rhetorical functions groups deviation combinations of sound (per total number of rhetorical functions all rhetorical functions) H (n = 12) 100% 0 0 L (n = 12) 40.42% 0.1999 0.5904

Key: H High-rated essay groups (both Hong Kong & Australian L2) L Low-rated essay groups (both Hong Kong & Australian L2)

In the high-rated essay groups, the average of the percentages of the subsump- tions/combinations of sound rhetorical relations (per total number of all rhetorical rela- tions) is 100%. The standard deviation, as well as the range of these percentages, is 0. This means that the high-rated essays are homogeneous, in that all their rhetorical func- tions at Paragraph Level which are subsumed by, or combine with, one another are sound ones. In the low-rated essay groups, on average, only 40.42% of their rhetorical functions which are subsumed by, or combine with, one another are sound ones. The standard de- viation is 0.1999, and the range 0.5904, suggesting that compared with the high-rated essays, the low-rated ones are not so homogeneous in terms of integration of arguments. Since the difference between the high-rated and low-rated essay groups is highly signifi- cant (P < 0.0001), the null hypothesis was rejected. Thus, we conclude:

Compared with the low-rated essays, the high-rated argumentative essays written by first- year Australian L2 students and their Hong Kong counterparts have a larger percentage of rhetorical relations at Paragraph Level which are themselves members of larger rhetorical relations at Macrostructure Level.

63 Discussion

The combined results of the three analyses above show that Hypothesis 3 could be re- jected:

Compared with the low-rated essays, the high-rated argumentative essays written by first-year Australian L2 and Hong Kong university students display a higher level of hier- archical integration of argument.

The most frequently occurring relations in the high-rated essays

The high-rated texts in the data sets have a wide repertoire (23 categories) of rhetorical relations (see Appendix 4). The rhetorical functions which have the highest frequency at different textual levels are presented in Table 17:

Table 17: Rhetorical Functions of the highest frequencies identified at different textual levels in the highly essay groups. High-rated essay groups Australian L2 Hong Kong Level Rhetorical Functions Rhetorical Functions of the highest Frequency of the highest Frequency frequency frequency Functional Elaboration 130 (19.85%) Elaboration 71 (23.51%) Unit/Simple Result 87 (13.28%) Result 31 (10.26%) Sentence Coupling 73 (11.15%) Comparison 31 (10.26%)

Elaboration 55 (31.07%) Elaboration 23 (31.94%) Complex Result 26 (14.69%) Comparison 10 (13.89%) Sentence Evaluation 21 (11.86%) Reported Evaluation 9 (12.50%)

Elaboration 33 (45.21%) Elaboration 45 (29.22%) Propositional Comparison 8 (10.96%) Evaluation 19 (12.34%) Segment Evaluation 7 (9.59%) Result 16 (10.39%) Reported Evaluation 7 (9.59%) Lower Levels: Elaboration 230 (23.33%) Elaboration 127 (28.41%) Total Result 129 (13.08%) Comparison 49 (10.96%) Elaboration 46 (62.16%) Elaboration 29 (61.70%) Paragraph Evaluation 19 (25.68%) Evaluation 7 (14.89%)

Evaluation 68 (91.89%) Answer 33 (80.49%) Macrostructure Assertion 6 (8.11%) Question 6 (14.63%)

At the lower levels, the rhetorical function with the highest frequency identified in both high-rated groups is Elaboration (Australia: 23.33%; Hong Kong: 28.41%). This result matches Stuart-Smith & Busbridge’s (1996) finding, but not Lindeberg’s (1988) observa- tion that that “Evaluate” is the most frequently occurring relation in good essays. At Paragraph Level, Elaboration remains the most frequently occurring relation (62%) in the two high-rated groups. At Macrostructure Level, Evaluation stands out as the most frequently occurring relation (91.89%) for the Australian group, followed by Assertion (8.11%). For the Hong Kong group, Answer is the rhetorical function which has the highest frequency (80.49%), followed by Question (14.63%). This suggests that there is a difference between the dominant macrostructures identified in these two groups, to be

64 discussed later. These results cannot be compared with the findings of any earlier re- search, which concentrated on the lower levels only. The reason why Elaboration is the most frequently occurring rhetorical function at Paragraph Level is related to the organisational strategy of the writers of the high-rated essays: eight out of twelve preferred to front their argumentative stance in a metatextual statement in the introduction, and used the following paragraphs to elaborate on that stance. To use HKH1 as an example, the writer fronts his stance in S4a–b in the first paragraph. Paragraphs 2 to 10 are all elaborations of S4a–b:

4a Although some experts say that these television programs are not the major factor causing this phenomenon, b there are many evidences and experiments show that television programs have this negative effect.

Paragraph Rhetorical Function at Paragraph Level 1 Background (S5–42) + Metatextual Statement

2 Elaboration (S4)

3 Elaboration (S6–9: Para. 2) = Elaboration (S4)

4 Comparison (S10–12) = Elaboration (S6–9: Para. 2) = Elaboration (S4)

5 Elaboration (S4)

6 Elaboration (S17–18: Para. 5) = Elaboration (S4) Comparison (S19–20: Para. 6) = Elaboration (S17–18: Para. 5) = Elabo- 7 ration (S4) 8 Elaboration (S4)

9 Elaboration (S25–26: Para. 8) = Elaboration (S4) Comparison (S27–28: Para. 9) = Elaboration (S25–26: Para. 8) = Elabo- 10 ration (S4) 11 Evaluation (S6–30)

12 Solution (S31–38: Para. 11) = Solution (S1–4: Para. 1)

Many proficient writers in this study fronted their stance in the introduction and then went on to elaborate it before coming to a conclusion. Their pattern therefore conforms to the “linear” Anglo-Saxon pattern (Kaplan 1966, 1987). Such a finding contrasts with Kaplan’s observations, and those of Malcolm & Honjio (1988) and Ballard & Clanchy (1991), that Chinese students typically produce a “circular” pattern, with a lengthy intro- duction and an indefinite conclusion.

The most frequently occurring relations in the low-rated essays

There are 25 categories of rhetorical relations in the low-rated essays, exceeding that of the high-rated groups by two (Unidentified Rhetorical Function, Unfulfilled Rhetorical Function). See Appendix 4; the rhetorical functions which have the highest frequency at different levels are summarised in Table 18.

65 Table 18: Rhetorical Functions of the highest frequencies identified at different textual levels in the low-rated essay groups. Low-rated essay groups Australian L2 Hong Kong Level Rhetorical Functions Rhetorical Functions of the highest Frequency of the highest Frequency frequency frequency Elaboration 131 (22.51%) Elaboration 48 (19.67%) Functional Unfulfilled Rhetorical 90 (15.46%) Unfulfilled Rhetorical 48 (19.67%) Unit/Simple Function Function Sentence Coupling 46 (7.90%) Unidentified Rhetorical 18 (7.38%) Function Elaboration 55 (34.81%) Unfulfilled Rhetorical 19 (27.94%) Unfulfilled Rhetorical 27 (17.09%) Function Complex Function Elaboration 17 (25%) Sentence Result 13 (8.23%) Unidentified Rhetorical 9 (13.24%) Function Unfulfilled Rhetorical 54 (33.75%) Unfulfilled Rhetorical 19 (37.25%) Function Function Propositional Unidentified Rhetorical 33 (20.63%) Elaboration 7 (13.73%) Segment Function Unidentified Rhetorical 5 (9.80%) Elaboration 28 (17.5%) Function Assertion 5 (9.80%) Elaboration 214 (23.78%) Unfulfilled Rhetorical 86 (23.69%) Lower Unfulfilled Rhetorical 171 (19%) Function Levels: Function Elaboration 72 (19.83%) Total Unidentified Rhetorical 89 (9.89%) Unidentified Rhetorical 32 (8.82%) Function Function Elaboration 36 (36.73%) Unfulfilled Rhetorical 22 (42.31%) Unfulfilled Rhetorical 33 (33.67%) Function Paragraph Function Elaboration 12 (23.08%) Unidentified Rhetorical 15 (15.31%) Metatextual Statement 6 (11.54%) Function Unfulfilled Rhetorical 33 (33%) Unfulfilled Rhetorical 23 (57.5%) Function Function Macro- Answer 27 (27%) Answer 6 (15%) structure Unidentified Rhetorical 15 (15%) Question 5 (12.5%) Function

At the lower levels, the most frequently occurring relations in the Australian L2 low- rated essay group are, in descending order, Elaboration (23.78%), Unfulfilled Rhetorical Function (UNFU) (19%) and Unidentified Rhetorical Function (UNID) (9.89%). Their relative rankings even remain exactly the same at Paragraph Level, constituting 36.73%, 33.67% and 15.31% respectively of the total number of functions identified there. At Macrostructure Level, UNFU emerges as the most frequently occurring relation (33%). The patterns in the Hong Kong low-rated essay group are similar, with UNFU and UNID ranking among the top three at all levels, except at Macrostructure Level. UNFU consti- tutes as much as 23.69%, 42.31% and 57.5% respectively at Paragraph Level and Macro- structure Level. This is the major difference between the high-rated and low-rated essay groups. The high-rated groups do not contain any UNID or UNFU, whereas the low- rated ones display a large number of these problematic rhetorical relations at all levels. This partly accounts for their lower level of argumentative integration, and also reflects the writers’ weakness in grammar, as will be discussed in the next chapter. Such findings coincide with the conclusions of Stuart-Smith & Busbridge (1996:28), Lindeberg (1988, 1994) and Kaldor et al. (1998:51).

66 Sequences of three rhetorical functions

As we saw in Chapter 4, three elements in common in different models of local struc- tures of argumentative texts are: Claim, Justification and Evaluation. In the categorisa- tion of the present study, the terms Justification and Evaluation remain the same, while the term Assertion has been adopted to stand for Claim. The pattern ‘Assertion + Justifi- cation + Evaluation’ is evident at Propositional Segment Level in the data sets of our study, as evidenced in Table 19:

Table 19: Frequency of the ‘Assertion + Justification + Evaluation’ pattern identified in the high- rated and low-rated essay groups. High-rated essay groups Low-rated essay groups Essay Frequency Essay Frequency AH1 7 AL1 1 AH2 4 AL2 2 AH3 9 AL3 2 AH4 8 AL4 2 AH5 13 AL5 1 AH6 6 AL6 3 Sub-total 47 Sub-total 11

HKH1 3 HKL1 1 HKH2 3 HKL2 0 HKH3 4 HKL3 1 HKH4 4 HKL4 0 HKH5 4 HKL5 1 HKH6 2 HKL6 2 Sub-total 20 Sub-total 5

TOTAL 67 TOTAL 16

The frequency of the ‘Assertion + Justification + Evaluation’ sequence identified in the high-rated essays greatly exceeds that of the low-rated ones. This suggests that the proficient students understood that argumentation is “the whole activity of making claims, … backing them up by producing reasons, criticising those reasons, [and] rebutting those criticisms” (Toulmin et al. 1984:14). Their essays have better integrated arguments. The writers of the low-rated essays, on the other hand, might not have fully understood the concept of argumentation, or were not skilful enough to make use of the pattern, since incomplete patterns, in which one of the three elements – Assertion, Justification, or Evaluation – is missing, were identified in some low-rated texts.

Match with the Argumentative Response Structure

At first sight, an argumentative response structure (Tirkkonen-Condit 1989:417), or in- tent, could be identified in all the twenty-four texts in our data sets, either manifested in the sequence Question-Answer or Remark-Response. All the writers present the essay plan in the first paragraph, whether they would like to find out an answer to the question of the effect of television, or to respond to researchers’ remarks, or findings, regarding this issue. However, Macrostructure Analysis shows that in the low-rated texts, problems emerge in the ‘response’ or ‘answer’. The evaluations are either tangled or empty, with no real substance:

67 Table 20: Problems of evaluation identified in low-rated essays. Essay Problem AL3 Body contains only Reported Evaluation without interim evaluations, which AL6 makes the evaluation in the conclusion unconvincing. HKL3 Lopsided arguments HKL5 AL5 Contradiction between the Body (containing ‘Con’ arguments) and the Con- HKL2 clusion (with a ‘Pro’ stance) AL3 AL5 Unclear evaluation due to expression problems AL6 HKL4

On the other hand, the writers of other low-rated essays seemed to be taking an avoid- ance strategy and were not inclined to make an evaluation at the end of the essay. AL1 is a good example:

AL1 – Paragraph 16 79 In conclusion, it take a very difficult decision to judge these experts comment. 80 The findings are two ways results. 81 There is speculation that violent T.V. programs are bad for children, 82a perhaps we are not experts to express our own opinion, b but I think the situation are complex c and the findings from these experts are in balance. 83a It might needs long term investigation to find out the truth b as like no one can predict the weather around us.

In this sense the macrostructures identified in the low-rated texts are ‘unfulfilled’, as they fail to fulfil the functions they signal to perform, for instance, being the ‘response’ to a ‘remark’. This is another reason why compared with the high-rated essays, the low-rated ones have a lower degree of integration of arguments. A breakdown of the macrostruc- tures identified in our data, both ‘fulfilled’ and ‘unfulfilled’, is presented in Table 21. The macrostructures of all the twelve low-rated essays are ‘unfulfilled’. The Remark- Response sequence is evident in five of them, and Question-Answer in seven. In the high-rated texts, the Remark-Response sequence was identified in eight, and the Ques- tion-Answer in three. The essay topic might have played an important role in influencing the students’ selection of macrostructure. It was framed as a question: “Does watching violence on television have any effect on children’s behaviour?” Some subjects would find it appropriate to use the Question-Answer sequence, while others preferred to incor- porate the question as part of the background, and to relate the ‘remarks’ made by differ- ent researchers. Interestingly, there is a hybrid sequence of Remark-Response-Solution in HKH1. This seems to suggest that some Hong Kong tertiary students may think that suggesting solutions is an essential part of expressing argumentative opinions toward current issues, which indeed has been my experience. To summarise, the essays in our data sets present the global argumentative structure. Three sequences are identified, whose frequencies manifested in both ‘fulfilled’ and ‘un- fulfilled’ macrostructures are, in descending order, Remark-Response (13), Question- Answer (10) and Remark-Response-Solution (1). This does not support Tirkkonen- Condit’s (1989:417) finding that Problem-Solution is the most frequently occurring se- quence employed in argumentative genre. Our most useful discovery is the identification of ‘unfulfilled’ macrostructures in most of the low-rated texts.

68 Table 21: Frequency of the macrostructures identified in the data sets. Hong Sequences Hong Australian Kong Australian Macro- of Argumentative Kong low- high-rated high- low-rated Total structure Response rated essays rated essays Structure essays essays Problem-Solution 0 0 0 0 0 ‘Fulfilled’ Remark-Response 6 2 0 0 8 Macro- Question-Answer 0 3 0 0 3 structure Remark-Response- 0 1 0 0 1 Solution Sub-total 6 6 0 0 12 Problem-Solution 0 0 0 0 0 ‘Unfulfilled’ Remark-Response 0 0 3 2 5 Macro- Question-Answer 0 0 3 4 7 structure Sub-total 0 0 6 6 12 Total 6 6 6 6 24

Adherence to academic conventions

With regard to two conventions attached to academic writing, the focus in this investiga- tion was whether our subjects “evaluated less, and evaluated less critically”, and whether our subjects acknowledged the sources of their citations. As shown in Table 17, Evaluation emerges as one of the relations of the highest fre- quency at nearly every level in the Australian high-rated essay group, and at Proposi- tional Segment and Paragraph Levels in the Hong Kong group. Yet in the low-rated groups, it only constitutes 2 to 3% of the total number of rhetorical functions identified at all levels, except at the Macrostructure Level of the Australian group, where its fre- quency amounts to 12% (cf. Table 22). However, as I have discussed, the evaluations made are either inconclusive or not sufficiently clear, or contain lopsided or contradic- tory arguments, and thus do not contribute to the integration of arguments.

Table 22: Frequency of evaluation identified at different textual levels in low-related essays. Essay group Australian Hong Kong

low-rated essays low-rated essays Level Function Unit / 12 6 Simple Sentence (2.06%) (2.46%) 6 1 Complex Sentence (3.80%) (1.47%) 3 1 Propositional Segment (1.88%) (1.96%) 21 8 Lower Levels: Total (2.33%) (2.2%) 2 2 Paragraph (2.04%) (3.85%) 12 1 Macrostructure (12%) (2.5%)

The question of whether our subjects on the whole “evaluated less” is more complex. Consider the following example, extracted from AL3, an Australian low-rated essay:

69 AL3- Paragraph 6 30 An experiment conducted by Freshbach and Singer (1971) offered evidence re- garding two groups of boys living in two separate cottages. 31a One group was shown violent programs b and the other was shown a non-violent program. 32a After two weeks of experiments, the boys who were initially more aggressive showed an increased in aggression b after viewing violent program. 33 Whereas, the boys who watched the non-violent program had showed lower lev- els of aggression. 34a It can be seen from this experiment that when children who were initially more aggressive are exposed to violent programs they tend to be more aggressive, b and thus television seems to trigger the aggression that the children had from other influences.

S34 was categorised as an Unidentified Rhetorical Function, because from the context we cannot tell whether the evaluation is made by the writer, or whether s/he is only citing Freshbach & Singer (without indicating the source). Since such problems appear in most of the low-rated texts, but are absent from the high-rated essays (cf. Table 20), it sug- gests that the non-critical approach of these Chinese students may be more of a problem of language than attitude, their undue respect for authority, as suggested in Chapter 3.

Table 23: Frequency of Unidentified Rhetorical Functions in low-rated essays brought about by unclear/missing sources. Australian Hong Kong Frequency Frequency low-rated essays low-rated essays AL1 2 HKL1 1 AL2 1 HKL2 3 AL3 3 HKL3 2 AL4 6 HKL4 0 AL5 0 HKL5 1 AL6 0 HKL6 2 Sub-total 12 Sub-total 9

As for referencing conventions, many writers of the low-rated essays failed to provide part of the source (the year of publication and/or the page number), as exemplified in S11 of AL4:

11 Singer & Singer point out theoretically that television viewing of violent programs definitely increases children’s aggression.

In extreme cases, the entire source has been left out, which is evident in Sentences 19 and 22 in the following extract:

AL5 – Paragraph 5 19 The experimental studies in field setting carried out with small viewers have shown that television viewing can have both prosocial and aggressive effect. 22 The Naturalistic Experiment indicates that direct observations of aggressive be- haviour of the children increased after introduction of television.

The functional units with such a referencing problem were categorised as ‘Unfulfilled Rhetorical Function [Reported Evaluation]’, because they fail to perform the functions

70 they signal (e.g., through the use of quotation marks). While this type of unsuccessful relation is not found in the high-rated scripts, it appears in most of the low-rated ones:

Table 24: Frequency of Unfulfilled Rhetorical Functions in low-rated essays brought about by unclear/missing sources. Australian Hong Kong Frequency Frequency low-rated essays low-rated essays AL1 1 HKL1 0 AL2 1 HKL2 2 AL3 0 HKL3 4 AL4 7 HKL4 0 AL5 1 HKL5 2 AL6 1 HKL6 3 Sub-total 11 Sub-total 11

The stage of schooling of our subjects and their place of origin, Hong Kong, made it unlikely that they had been greatly influenced by Chinese scholarly tradition. A more plausible reason for this referencing problem is again the poor language proficiency of some of the subjects. With the emergence of these two referencing problems, the quality of their argumentative EAP essay is affected, since the essays lack the writer’s viewpoint, and/or the arguments presented are not supported with clear and identifiable sources of information. All the factors we have discussed here, contribute to, or disrupt, the hierarchical inte- gration of argument in the argumentative essays in our data sets.

HYPOTHESIS 1

The hypothesis

The argumentative essays (both high-rated and low-rated) written by first-year Australian L2 university students do not display a higher level of hierarchical integration of argument than those written by their Hong Kong counterparts.

To test this hypothesis, I needed to investigate whether there were any differences be- tween the Australian L2 essays and Hong Kong essays in terms of their percentages of

• rhetorical relations at Functional Unit (FU)/Simple Sentence (SS)/Complex Sentence (CS) Level which are themselves members of larger rhetorical relations at Propositional Segment (PS) Level; • rhetorical relations at Propositional Segment Level which are themselves mem- bers of larger rhetorical relations at Paragraph (P) Level; and • rhetorical relations at Paragraph Level which are themselves members of larger rhetorical relations at Macrostructure (M) Level.

To compare the Australian L2 and Hong Kong essay groups at each level, we adjusted the total number of subsumptions and/or combinations of sound rhetorical relations by the total number of subsumptions/combinations of all types of rhetorical relations (in- cluding unsound relations, UNID and UNFU), identified at the same level.

71 Results

Integration of rhetorical relations at Functional Unit/Simple Sentence/ Complex Sentence Level into rhetorical relations at Propositional Segment Level

Table 25: Testing of Hypothesis 1 at Propositional Segment Level: Australian L2 = Hong Kong. Average of percentages of sub- Range of percentages sumptions/combinations of Essay Standard of subsumptions/ sound rhetorical functions groups deviation combinations of sound (per total number of rhetorical functions all rhetorical functions) A (n = 12) 87.69% 0.1648 0.4167 HK (n =12) 82.28% 0.2355 0.5714

Key: A Australian L2 essays (both high-rated & low-rated) HK Hong Kong essays (both high-rated & low-rated)

In the Australian L2 essay groups, 87.69% of their rhetorical relations at Functional Unit /Simple Sentence/Complex Sentence Level which are subsumed by, or combine with, one another are sound ones, and in the case of the Hong Kong essay groups, the average percentage is 82.28%. For the Australian groups, the standard deviation is 0.1648, and the range of these percentages 0.4167. For the Hong Kong groups, the standard deviation is 0.2355, and the range 0.5714. This shows that neither of the groupings is homogene- ous in terms of integration of arguments at the lower levels. This is because each of them consists of both high-rated and low-rated essays. The difference between the Australian groups and the Hong Kong groups is statistically significant (P = 0.0441), just below the threshold of 0.05 set. It allows us to establish:

Compared with those written by their Hong Kong counterparts, the essays written by first- year Australian L2 university students have a larger percentage of rhetorical relations at Functional Unit/Simple Sentence/Complex Sentence Level which are themselves mem- bers of larger rhetorical relations at Propositional Segment Level.

Yet the significance is not at the highest level and the basis for establishment is therefore not strong.

Integration of rhetorical relations at Propositional Segment Level into rhetorical relations at Paragraph Level

Table 26: Testing of Hypothesis 1 at Paragraph Level: Australian L2 = Hong Kong. Average of percentages of Range of percentages subsumptions/combinations of Essay Standard of subsumptions/ sound rhetorical functions groups deviation combinations of sound (per total number of rhetorical functions all rhetorical functions) A (n = 12) 72.25% 0.3264 0.8125 HK (n = 12) 71.41% 0.3171 0.75

Key: A Australian L2 essays (both high-rated & low-rated) HK Hong Kong essays (both high-rated & low-rated)

72 In the Australian L2 essay groups, 72.25% of rhetorical relations at Propositional Seg- ment Level which are subsumed by, or combine with, one another are sound ones, and in the case of the Hong Kong groups, the average percentage is 71.41%. These two per- centages are very close. For the Australian groups, the standard deviation is 0.3264, and the range of these percentages 0.8125. For the Hong Kong groups, the standard deviation is 0.3171, and the range 0.75. This suggests that neither of these two groupings is homo- geneous in terms of integration of arguments at Propositional Segment Level, as each consists of both high-rated and low-rated essays. Since the difference between the Aus- tralian groups and the Hong Kong groups is not statistically significant (P = 0.9203), we cannot reject the null hypothesis. Thus, we conclude:

There is no significant difference between the Australian and Hong Kong essays in their percentage of rhetorical relations at Propositional Segment Level which are themselves members of larger rhetorical relations at Paragraph Level.

Integration of rhetorical relations at Paragraph Level into rhetorical relations at Macrostructure Level

Table 27: Testing of Hypothesis 1 at Macrostructure Level: Australian L2 = Hong Kong. Average of percentages of Range of percentages subsumptions/combinations of Essay Standard of subsumptions/ sound rhetorical functions groups deviation combinations of sound (per total number of rhetorical functions all rhetorical functions) A (n = 12) 72.25% 0.3264 0.8125 HK (n = 12) 68.17% 0.3551 0.8571

Key: A Australian L2 essays (both high-rated & low-rated) HK Hong Kong essays (both high-rated & low-rated)

In the Australian L2 essay groups, 72.25% of rhetorical relations at Propositional Seg- ment Level which are subsumed by, or combine with, one another are sound ones, and in the case of the Hong Kong groups, the average percentage is 68.17%. For the Australian groups, the standard deviation is 0.3264, and the range of these percentages 0.8125. For the Hong Kong groups, the standard deviation is 0.3551, and the range 0.8571. This sug- gests that neither of these two groupings is homogeneous in terms of integration of ar- guments at Paragraph Level, as each consists of both high-rated and low-rated essays. Since the difference between the Australian and the Hong Kong groups is not statistically significant (P = 0.4803), we cannot reject the null hypothesis. Thus, we conclude:

There is no significant difference between the Australian and Hong Kong essays in their percentage of rhetorical relations at Paragraph Level which are themselves members of larger rhetorical relations at Paragraph Level.

Based on these analyses, we concluded that Hypothesis 1 could not be rejected, since the result for the first element was only marginally significant, and those for the other two were not statistically significant. Thus:

There is no substantial difference in the level of hierarchical integration of argument dis- played in the argumentative essays (both high-rated and low-rated) written by first-year

73 Australian L2 university students as compared with those written by their Hong Kong counterparts.

If we carry out a cross-hypothesis analysis, we will find that for Hypothesis 1, the stan- dard deviations (SDs) and the ranges of the percentages of subsumptions/combinations of sound rhetorical functions (Australian L2 vis-à-vis Hong Kong essays) are much greater than those for Hypothesis 3 (high-rated vis-à-vis low-rated essays). This is under- standable, as in the case of the former each of the two groupings consists of both high- rated and low-rated essays, which explains the greater differences within these groupings in terms of integration of arguments. The greatest disparity of percentages is detected at Paragraph Level, especially in the comparison between the Australian and the Hong Kong essay groups (Hypothesis 1). For example, the SDs for the Australian groups and Hong Kong groups for Hypothesis 1 at Paragraph Level are 0.3264 and 0.3551 respec- tively. These are the same as, or slightly greater than, their SDs at Propositional Segment Level, 0.3264 and 0.3171 respectively, but are much greater than those at FU/SS/CS Level, 0.1648 and 0.2355 respectively. This is not surprising, as it is easier for a writer to ensure that the ideas of his/her essay are coherent at the clausal/sentential levels com- pared with the upper levels, especially Paragraph Level.

Discussion

The most frequently occurring relations and macrostructures identified in the Australian and Hong Kong essay groups

As mentioned earlier, in both high-rated essay groups, AH and HKH, Elaboration ranks the top at the lower levels, and Elaboration and Evaluation at Paragraph Level. Yet at Macrostructure Level, Evaluation and Assertion rank the highest in the Australian group, and Answer and Question in the Hong Kong group (cf. Table 17). The two low-rated groups, AL and HKL, are also similar in terms of the types of relation at different levels. UNID and UNFU rank the top at the lower levels, and UNFU at Paragraph and Macro- structure Levels (cf. Table 18). Since the most frequently occurring rhetorical functions identified at all levels between the two high-rated groups, and those between the low- rated groups, are very similar, there are no obvious differences between the Australian and Hong Kong groups as two conglomerations.

Table 28: Frequency of types of macrostructures identified in Australian and Hong Kong essays. Macrostructure Australian essay groups Hong Kong essay groups ‘Fulfilled’ High-rated 6 High-rated 6 Macrostructure Low-rated 0 Low-rated 0 ‘Unfulfilled’ High-rated 0 High-rated 0 Macrostructure Low-rated 6 Low-rated 6 Totals 12 12

When we further compare the frequency of the types of macrostructure identified in the Australian and Hong Kong essay groups (A vs HK), we find that their results are identi- cal: six ‘fulfilled’ and six ‘unfulfilled’ macrostructures in each group (cf. Table 28). This provides further evidence that there are no differences in the level of hierarchi- cal integration of argument displayed in the Australian and Hong Kong essay groups. Such a result may also be explained by the extent to which the Australian subjects used English inside and outside the classroom.

74 Language environment for the Australian subjects

To discover more about the subjects of the current study, interviews were conducted with one of the EAP tutors at Macquarie University, Peter Roger, and some of the Australian L2 subjects. The comments made by Roger were especially relevant, since some of our subjects had attended his tutorial sessions and he had assessed their essays. He stated that from his frequent encounters with his overseas students, he observed that it is possible to live in Australia and use English only in the classroom. Bearing in mind Falvey’s (1998:75) definition of EFL, “language which is learned primarily in the classroom with little assistance from the language environment”, many of these Australian students might be properly regarded as EFL students in these terms. Roger also remarked that foreign students in Australia do not gain additional L2 competence from everyday use of English in social encounters. They find it difficult to establish friendship with L1 speakers. This echoes the observations made by Bradley & Bradley (1984), Barker et al. (1991), Mullins et al. (1995) and Biggs & Watkins (1996). When there is an absence of equal discourse roles between learners and L1 speakers, in- ternational students who are already collectivistically inclined will become more attached to their own ethnic groups and less willing to adapt to the L1 culture (Ellis 1994:187). It seems, then, that living in Australia makes little difference at all to these students. They behave in much the same way as when they were in Hong Kong. There might be only one major difference: for Hong Kong students, “the opportunity and need to use English [in Hong Kong] to communicate were low” (So 1998:161; my emphasis). Now that some of these students had gone to Australia, the opportunities for them to use English were high, if they looked for them. Yet their motivation for using the language remained low. Many have neither integrative nor instrumental motivation to learn English, and do not capitalise on the ESL environment in Australia. They thus make little progress in their English. This may partly explain why they do not develop better argumentative skills, and why no difference was found in the hierarchical integration of argument dis- played in the Australian L2 essays as compared to the Hong Kong texts.

75 76 Chapter Seven: Results and Discussion: Grammatical Accuracy Analysis

INTRODUCTION

This chapter presents the findings of Grammatical Accuracy Analysis (GAA), devised to compare the level of syntactic accuracy, firstly, between the Australian and Hong Kong essays (Hypothesis 2) and secondly, between the high-rated and low-rated essays (Hy- pothesis 4). As with the findings obtained from RFA, those derived from GAA were ana- lysed by Poisson Regression, to determine whether there is any statistical difference among the four groups of essays. The level of significance was set at 0.05.

HYPOTHESIS 4

The hypothesis

The high-rated argumentative essays written by first-year Australian L2 and Hong Kong university students do not display a greater degree of grammatical accuracy than the low- rated essays.

To test this hypothesis, I first analysed the errors according to the taxonomy of GAA, counting the total number of grammatical errors of each essay group. This total was then adjusted by the total number of words, between the high-rated essay groups and the low-rated essay groups, that is,

Total number of grammatical errors Total number of words

Results

Table 29: Testing of Hypothesis 4: high-rated = low-rated. Average of percentages of Essay Standard Range of percentages of grammatical errors groups deviation grammatical errors (per total number of words) H (n = 12) 2.74% 0.00782 0.02907 L (n = 12) 8.49% 0.04591 0.1376

Key: H High-rated essay group (both Hong Kong & Australian L2) L Low-rated essay group (both Hong Kong & Australian L2)

The table shows that the average of the percentages of grammatical errors (per total number of words) for the high-rated essay groups is only 2.74%. The SD is 0.00782, and the range of these percentages 0.02907, which means that the high-rated essays are ho- mogeneous. In contrast, the average percentage for the low-rated essay groups is 8.49%. The SD is 0.04591, and the range 0.1376. Thus, compared with the high-rated essay groups, the low-rated groups are less homogeneous in terms of their percentages of grammatical errors. Since the difference between the high-rated and low-rated essay groups is highly significant (P < 0.0001), the null hypothesis was rejected. Thus:

77 Compared with the low-rated essays, the high-rated argumentative essays written by first-year Australian L2 and Hong Kong university students have a smaller percentage of grammatical errors. They thus display a greater degree of grammatical accuracy than the low-rated essays.

Discussion: Similarities and differences in error types between the high- rated and low-rated essays

The results for Hypothesis 4 confirm that there is a correlation between grammatical ac- curacy and success in argumentative writing. This supports the findings of Stewart & Grobe (1976), Sparks (1988) and Lukmani (1995) that grammatical accuracy has a bear- ing on raters’ perception of ESL writing quality. To investigate further the reasons accounting for these results, two types of compari- son were made. First, the frequency, percentage and relative ranking of each error type in the high-rated groups (H) were compared with those in the low-rated groups (L). The second type of comparison involved the use of tests of statistical significance. It was found that there are similarities between the groupings in the relative rankings of the error types (see Appendix 5 for the frequency of errors in individual essays in each group; Appendix 6 for the full comparison of frequencies, percentages, and relative rank- ings of the error types). Two categories emerge as the most frequently occurring errors in both H and L: Agreement, and Internal Punctuation within and between Phrases and Clauses. If we further break the percentages down into those of the Australian High-rated Group (AH), Australian Low-rated Group (AL), Hong Kong High-rated Group (HKH) and Hong Kong Low-rated Group (HKL), we find that in all groups except AH, Number Agreement ranks first. The rankings of Internal Punctuation and Articles are between the first and sixth in every group. These three categories, when combined, constitute half or nearly half in the case of AH and HKH (50.25% and 44.96% respectively), and about one-third of the total number of errors identified in both AL and HKL (36.22% and 32.59% respectively). Cf. Table 30:

Table 30: Frequencies, percentages and relative rankings of the most recurrent errors in the four essay groups. Essay groups Error Australian high- Australian Hong Kong Hong Kong category rated essays low-rated high-rated low-rated (AH) essays (AL) essays (HKH) essays (HKL) Number 36 120 35 60 Agreement (18.09%) (18.9%) (27.13%) (18.99%) (Nag) [2nd] [1st] [1st] [1st] Internal 41 71 12 23 Punctuation (20.6%) (11.18%) (9.3%) (7.28%) (IP) [1st] [2nd] [3rd] [4th] 23 39 11 20 Articles (Art) (11.56%) (6.14%) (8.53%) (6.33%) [3rd] [6th] [4th] [5th] Total 100 230 58 103 (Nag+IP+Art) (50.25%) (36.22%) (44.96%) (32.59%)

Key: The numbers in parentheses are the percentage of each error type in the respective grouping. The numbers in square brackets are the relative ranking of each error type in the respective grouping.

78 Although the common error types of the high-rated and low-rated essay groups fall into similar patterns, their grades are very different. To investigate possible reasons for this, we have to examine the absolute values (frequencies) of these error categories. Such comparisons are very revealing. For example, Agreement constitutes 18.09% and 18.9% of the total frequency of errors in AH and AL respectively, yet the raw number of this error type is only 36 in the former, as compared to 120 in the latter. In HKL it is a good deal lower, in percentage (18.99%), than in HKH (27.13%), yet there are almost twice as many Agreement errors in them (HKL: 60 vs HKH: 35). These kinds of disparity be- tween the percentages and the raw numbers apply to the errors in Internal Punctuation and Articles. When the raters were marking the scripts, they will have been aware of the large numbers of errors in certain scripts, and awarded low (and high) grades accordingly. On the other hand, when the data were further submitted to statistical tests of signifi- cance, significant differences emerged between the H and L groupings in two error types, Spelling (P = 0.0198) and Collocation (P = 0.0237), in terms of the percentages in their respective groupings; see Table 31:

Table 31: Frequencies, percentages and relative rankings of the error categories with statistical differences in the high-rated and low-rated essay groups. Error Essay groups category High-rated essays (H) Low-rated essays (L) 22 31 Spelling (6.71%) (3.26%) [5th] [13th] 32 52 Collocation (9.76%) (5.47%) [4th] [7th]

Key: The numbers in parentheses are the percentage of each error type in the respective grouping. The numbers in square brackets are the relative ranking of each error type in the respective grouping.

The frequencies of Spelling and Collocation errors are higher in the low-rated essays than the high-rated ones. Such a result is interesting, as these two error categories are dif- ferent in nature. Problems in spelling are generally considered to be a more superficial “mechanical” type of error (Stewart & Grobe 1979) – obvious and ubiquitous. On the other hand, Collocation errors, defined in this study syntagmatically as “Wrong choice of, redundant, or missing, word which is against conventional idiom or usage” (see Appen- dix 3), cover a greater range, including erroneous use of vocabulary, prepositions, or other idiomatic aspects of the language. Compared with Spelling, this error type is much more subtle, individual and deep-seated. The results of our statistical analysis show that the proficient writers made fewer mistakes in both types of error, superficial as well as subtle, which partly accounts for their high grades. Errors in Spelling and Collocation may therefore have had a particular influence on raters’ assessments. This supports Bal- lard & Clanchy’s (1991:33) view that “Spelling”, as well as “Idiom” and “Register”, among others, form a set of “objective criteria” by which an essay is rated. Another factor which may also account for the differences in grades received by the high-rated and low-rated essays is that three error types can only be found in the low- rated essays. These include Anaphoric/Cataphoric Pronouns, Tangled Construction and Total Communication Breakdown. Their frequencies, percentages and relative rankings are presented below in Table 32:

79 Table 32: Frequencies, percentages and relative rankings of the error categories identified in the low-rated essay groups only. Essay Groups Australian Hong Kong Error Australian Hong Kong high- high- category low-rated low-rated rated rated (AL) (HKL) (AH) (HKH) 42 17 Anaphoric/ 0 (6.61%) 0 (5.38%) Cataphoric Pronouns [5th] [8th] 12 5 Tangled Construction 0 (1.89%) 0 (1.58%) [15th] [16th] 20 9 Total Communication 0 (3.15%) 0 (2.85%) Breakdown [12th] [12th]

Total: 74 31 0 0 Each essay group (11.65%) (9.81%)

Key: The numbers in parentheses are the percentage of each error type in the respective grouping. The numbers in square brackets are the relative ranking of each error type in the respective grouping.

In both AH and HKH, the frequencies for all these error types are zero. They also consti- tute only 11.65% and 9.81% of all the errors identified in the AL and HKL respectively. Yet when I performed statistical tests of significance, I found that the average percent- ages of these types of error in the low-rated groups are significantly different from zero (P < 0.0001). This underscores the importance of this difference between the high-rated and low-rated groups. What is more, these error types, especially the errors in Tangled Construction and Total Communication Breakdown would have exerted an influence on the teachers’ ratings because they all affect the reading and comprehension of the text, and detract from the quality of the arguments. Consider this example of Tangled Con- struction, identified in a low-rated Australian L2 essay (AL1):

5 The alarm was causes is T.V. violent effects on children are the issue.

What the writer may have intended to say here is that anxiety was raised over whether television violence has effects on children. Although the reader can guess at the intended meaning, it takes considerable effort, and things still seem confused. The situation is even worse when the meaning cannot be communicated at all, as in the case of Total Communication Breakdown (from AL1 again):

77a a person in authority, has chosen to show a violent programs, b in the position is more responsible than of industry representative c and parents who quick to reduce any potential of negative effects of viewing in children activities.

In this sentence, the writer has not made it clear why “a person in authority” is more re- sponsible than the “industry representative”, and more basically, who these two terms refer to. One more problem is that S77c does not seem to be related to S77a and S77b. All these seriously affect the argument of the text.

80 Errors in Anaphoric/Cataphoric Pronouns also cause “major obstructions to readabil- ity”, as noted by Kaldor et al. (1998:57–8). This is evident from an extract taken from a low-rated Hong Kong essay, HKL2:

HKL2 – Paragraph 2 2a From the past researches, long time in TV viewing can cause shorts sighted, b but people concerns much about it as well as the negative effect of TV violence on child. 3 Is it as worst as people think? 4 Yet, it is just a pessimistic perspective.

In S2b, the pronoun it refers to ‘short-sightedness’. Therefore the reader finds the it in S3 and S4 very confusing – Does this refer to ‘short-sightedness’ again or to the negative effect of TV violence on child? The low-rated essays contain many more grammatical errors than the high-rated ones. Furthermore, only the low-rated essays present the types of error that obstruct the read- ing and comprehension of the text. Therefore two factors seem to affect teachers’ holistic rating: the quantity of errors in the essays, as well as their ‘quality’, or gravity, especially whether the errors affect the reader’s comprehension of the text.

HYPOTHESIS 2

The hypothesis

The argumentative essays (both high-rated and low-rated) written by first-year Australian L2 university students do not display a greater degree of grammatical accuracy than those written by their Hong Kong counterparts.

To test this hypothesis, I compared the total number of grammatical errors, adjusted by the total number of words, between the Australian essay groups (A) and the Hong Kong essay groups (HK).

Results

Table 33: Testing of Hypothesis 2: Australian L2 = Hong Kong. Average of percentages of Range of Standard Essay groups grammatical errors percentages of deviation (per total number of words) grammatical errors A (n = 12) 5.34% 0.04339 0.1386 HK (n = 12) 5.88% 0.04555 0.1428

Key: A Australian L2 essays (both high-rated & low-rated) HK Hong Kong essays (both high-rated & low-rated)

The averages of the percentages of grammatical errors (per total number of words) in both groupings are very similar: 5.34% for the Australian, and 5.88% for the Hong Kong groups. The standard deviations are also very similar for both (0.04339 (Australian) and 0.04555 (Hong Kong)), as are the ranges (0.1386 (Australian) and 0.1428 (Hong Kong)), thus showing that both groupings are highly homogeneous in terms of their percentages

81 of errors. Since the difference between the groupings is not statistically significant (P = 0.5212), we cannot reject the null hypothesis. Thus:

There is no significant difference in the degree of grammatical accuracy displayed in the argumentative essays (both high-rated and low-rated) written by first-year Australian L2 university students as compared with those written by their Hong Kong counterparts.

Discussion

Similarities in error types between the Australian and Hong Kong essays

To discover the reasons accounting for the results, the two approaches mentioned earlier were used again to compare the percentages and relative rankings of the error categories in the Australian groups (A) with those of the Hong Kong ones (HK). First, the error types in each grouping were ranked according to their percentages, then the data were submitted to statistical tests. When the relative rankings of the error categories were compared, it was found that there were similarities between A and HK. As already observed, Number Agreement, Internal Punctuation and Articles constitute a large percentage of the total number of er- rors in all the four groupings in this study. These three categories, when combined, con- stitute 39.57% of the total number of errors identified in the Australian groupings, and 36.18% in the Hong Kong groupings. Indeed, Number Agreement, which ranks first in both A and HK, alone already accounts for one-fifth of the total error frequencies (18.71% in A, and 21.35% in HK). We have noted that studying in an ESL environment did not seem to have helped these subjects very much to make improvements in structur- ing their arguments. The same seems to be the case in relation to their grammatical accu- racy. The similar error patterns discovered may indicate that our Australian subjects, who were of Hong Kong origin, had formed stable error patterns, or “relatively fixed de- fects”, in their L2, before they had left Hong Kong:

a learner’s degree of proficiency can legitimately be conceived as a ‘system’ created by the learner for himself. This system is not invariant although it may have certain relatively fixed defects which, after, Selinker (1972), are often referred to as ‘fossilizations’. (Stern 1983:355)

These results echo Milton’s (2001:109) finding that the English interlanguage of Hong Kong students is homogeneous: “Hong Kong students develop an L2 that is accommo- dated remarkably well to the demands and constraints of their educational environment.” One of the “constraints” of the Hong Kong educational environment is certainly the fact that English is not taught as an L2 there. Despite this, the local educational environ- ment imposes a constant “demand” on students: in order to be able to enter university, they have to pass the Use of English examination, which is a difficult task for many stu- dents. Therefore Hong Kong students are keen to “fudge grammatical competence and to impose communicative dynamism on texts”, which is said to have shaped the character- istics of the “Hong Kong interlanguage” (Milton ibid.:109). The reasons for the difficulty experienced by the Australian subjects in making a breakthrough in improving the accuracy of their L2 are complex, involving an interplay of socio-cultural and psychological factors. The socio-cultural factors highlighted in Chapter 2 include the learners’ view about the target-language culture. The psychological factor is motivation, whether the learner feels that there is “the need to get meanings

82 across and the pleasure experienced” (Ellis 1994:516). We have also seen that there is a relatively high degree of social distance between the Chinese community and the Eng- lish-speaking community (Luke & Richards 1982:53; So 1998:161; Pennington 1998; Li et al. 2000). It seems that even after some of these students (such as our Australian sub- jects) have gone abroad to study, they remain ‘indifferent’, and even become hostile, to the target-language culture, despite increased exposure. The reasons for this are again complex, the most plausible one being the lack of motivation. They may neither feel “the need to get meanings across” (being ‘self-sufficient’, mixing only with people of their own language community, and anticipating returning to Hong Kong after graduation) nor experience the “pleasure” associated with communicating with the L1 speakers in Aus- tralia (partly owing to their low English proficiency, and partly owing to the unfriendly, or lukewarm, attitudes of some Australians). This fits the Acculturation Model proposed by Schumann (1978a, 1978b) to a certain extent, in that the first social variable for suc- cess in L2 learning is the extent to which a group integrates. Yet the second social vari- able postulated by Schumann, that learning is less likely to take place in a situation in which the L2 group is dominant (as in a colonial context) than in one in which the L1 group is dominant (as in a country of immigrants) has not been confirmed in this study. It is true that there is little success in L2 learning in Hong Kong, yet studying in a country, such as Australia, in which the L1 group is dominant, does not by itself automatically guarantee success in L2 learning either. It still depends on how willing the learner is to mix with the native speakers there, and to expose him/herself to the target language. Cf. Milton (2001:110):

Psychological factors, such as variations in learners’ attitudes and intrinsic aspects of the L2 may affect learnability, but it also seems clear that sociolinguistic factors are major de- termining factors in the features we find in students’ written production.

The interplay of socio-cultural and psychological factors in L2 learning is intricate.

Differences in error types between the Australian and Hong Kong essays

Table 34: Frequencies, percentages and relative rankings of the error categories with statistical differences in the Australian and Hong Kong essay groups. Error Essay groups category Australian essays (A) Hong Kong essays (HK) 20 22 Tenses (2.4%) (4.94%) [13th] [8th] 44 40 Collocation (5.28%) (8.99%) [8th] [3rd]

Key: The numbers in parentheses are the percentage of each error type in the respective grouping. The numbers in square brackets are the relative ranking of each error type in the respective grouping.

Although the Australian L2 and Hong Kong essays display similarities in error type, there are two categories which show significant statistical differences between these two groupings: Tenses (P = 0.0278) and Collocation (P = 0.0451). Cf. Table 34.

83 Tenses rank 13th in the Australian group and 8th in the Hong Kong group, and Col- location 8th in the Australian group and 3rd in the Hong Kong group. This suggests that the Australian L2 subjects were better able to handle tense forms and to create more ac- curate collocations and idiomatic expressions. Therefore although they might not have taken full advantage of studying in the L1 environment, their increased exposure to spo- ken and written English seemed to have enhanced their resources in the target language. The collocation errors that they produce suggest that at least they have more options to draw upon. Consider this sentence from a low-rated Australian essay, AL2:

7a However, some researchers argue with strong evidence and experiment that TV violence would only be a short-term effect b or even has nothing to do with the real world violence.

In producing the bolded clause, the writer seems to have combined two structures:

(1) TV violence would only have a short-term effect … (2) … TV violence would only be of a short-term effect …

This type of error is certainly more sophisticated than mechanical ones such as Spelling. They are also evident in the high-rated Australian essays. Take for example AH4, its first sentence reads:

1 With the seemingly increased portrayal of graphic violence on television, the ef- fects of it on children’s behaviour and attitudes has become of great concern to many people, particularly parents.

The writer seems to have combined two structures to produce the syntactic blend (bolded):

(1) … has been of great concern … (2) … has become a great concern

This can be explained by Wolfe-Quintero et al.’s (1998:4) observation that “[s]econd language learners write more grammatically and lexically complex sentences as they be- come more proficient”. Peters & Gassmann (1995:139) noted errors which display the effects of students tackling new expressions:

[Students’] syntactic problems are symptoms of trying to develop more sophisticated ways of talking about what they observe ...

Although students trying to capture new expressions in their writing inevitably make mistakes, it is part of the process of acquiring new resources in the target language. Seen from this perspective, the Australian L2 subjects who combined two structures in their writing and made mistakes in the process would seem to be more advanced along their paths of approximation toward the target language standard than those who produce only surface errors.

84 Plagiarism in the Hong Kong low-rated essays

It is true that the number of unclear/missing sources in the Hong Kong scripts is the same as that in the Australian scripts. Yet while the instances of plagiarism are restricted to one or two sentences in the Australian scripts, there are two Hong Kong low-rated essays, HKL1 and HKL3, which contain a whole extract ‘lifted’ from the references. These scripts contain noticeably fewer grammatical errors, compared to the other Hong Kong low-rated essays, such as HKL2 and HKL4:

Essay Total Number of grammatical errors HKL1 36 HKL2 101 HKL3 16 HKL4 90

An example of plagiarism in HKL1 is shown below. All the bolded sentences and clauses have been lifted from Singer & Singer (1988:173), and strung together with minimal contribution from the student:

HKL1 – Paragraph 3 On the other side of the coin, television can make benefits on the growth and develop- ment of children mind. The television industry, confronted with concerns by psycholo- gists and other behavioral scientists, as well as by some religious leaders and con- sumer advocates, about the quality of programming available generally, and specifi- cally programming for children, calls for clear social science evidence that viewing of such materials produce any effect at all on children. According to the tests held by Federal Communications Commission, television can have important positive effects on children by giving an example of a child who, having recently seen only one demon- stration on TV of the Heimlich maneuver – a method for dislodging bones or other ob- jects in the throat that threaten suffocation, was able to use it shortly afterward to save the life of another child. It could produce a valuable social learning experience.

Some student writers contribute relatively more themselves, while still blending stretches from the original. Consider this example extracted from HKL3:

HKL3: 1a Most parents are more likely to march on school boards b to ban books that contain unsuitable materials to children, c but they rarely look closely at what their children are watching.

The bolded clauses here have been lifted from two sentences in Paragraphs 2 and 3 re- spectively in Singer & Singer (again shown by bolding in the following), and interwoven as one sentence in the students’ text:

Paragraph 3, Singer & Singer (1988:172) … why are parents more likely to march on school boards to ban books such as ‘Catcher in the Rye’ or ‘Down These Mean Streets’, but to respond with relative indif- ference to the excessively violent content that characterizes daily television program- ming?

85 Paragraph 2, Singer & Singer (1988:171) On the other hand, we have indications that parents prefer not to look closely at what their children are watching or to consider television as a potential influence or haz- ard …

Compared with the writer of HKL1, the writer of HKL3 has been more skilful in his/her plagiarism, yet still not confident enough to digest the ideas presented in the resource materials and write the essay alone.

Further implications of Grammatical Accuracy Analysis

The findings of Grammatical Accuracy Analysis have important implications for L2 learning. Stern (1983:355) remarks:

From the point of view of pedagogy, the key issue is that the interlanguage in many in- stances is too fossilized, too idiosyncratic, and does not move reliably through better and better approximations towards target language norms.

The fact that the Australian essays display more ‘advanced errors’ and less serious pla- giarism seems to suggest that although the Australian subjects might not be able to re- duce the total number of errors to a significant level, their English is not “too fossilised”. They seemed to have more resources, although they were not always able to make the correct choice among competing grammatical forms and/or idiomatic expressions. They also seemed to be making “better and better approximations towards target language norms” than someone who produces frequent surface errors such as faulty subject-verb agreement, and plagiarises extensively. The key issue is that their “move towards target language norms” is not yet “reliable” enough. The reason for this may lie in their motiva- tion in learning English. Milton (2001:110) suggests,

students … are nevertheless willing and able, if given timely and reliable support, to de- velop greater grammatical, discoursal and communicative competence in the L2.

While there is no doubt about these learners’ ability, their willingness cannot be taken for granted. The deciding factor for L2 writing success is not merely students’ ability, but also their willingness (motivation). If they are willing, or better, determined, to develop, coupled with the “timely and reliable support” they receive, they should be able continu- ously to improve their L2 competence and in the end achieve success.

CONCLUSION

The results from Grammatical Accuracy Analysis have allowed us to reject Hypothesis 4: compared with the low-rated essays, the high-rated ones displayed a higher degree of grammatical accuracy. On the other hand, Hypothesis 2 could not be rejected, since GAA showed there was no difference in the degree of grammatical accuracy in the Australian L2 essays and Hong Kong essays. Grammatical Accuracy Analysis was used as a means of triangulation, to determine whether its results complemented, or conflicted with, those obtained from Rhetorical Function Analysis. We now have similar findings derived from these two independent methods, GAA and RFA. To establish further whether there is any statistical correlation between them, the Pearson Correlation Coefficient was calculated.

86 In the calculation, the percentages of sound rhetorical functions and grammatical errors were used instead of the raw totals of each, which are influenced by the total number of rhetorical functions and grammatical errors respectively, and hence need to be adjusted. The results are presented in Table 35:

Table 35: Correlation between the percentage of sound rhetorical functions and the percentage of grammatical errors identified at different textual levels. Discourse level Correlation coefficient Propositional Segment Level -0.56 (P = 0.0037) Paragraph Level -0.74 (P < 0.0001) Macrostructure Level -0.74 (P < 0.0001)

The correlation coefficient is -0.56 at Propositional Segment Level (P = 0.0037), -0.74 at both Paragraph Level and Macrostructure Level (P < 0.0001 in both cases). We can thus conclude that the percentage of sound rhetorical functions is inversely correlated with that of grammatical errors, and the correlation is statistically significant at all the three levels. In other words, the more grammatical errors an essay contains, the fewer sound rhetorical functions emerge at the higher levels of the discourse structure. This is because the unsound rhetorical functions, Unidentified Rhetorical Function (UNID) and Unful- filled Rhetorical Function (UNFU), are brought about by several underlying factors, the most likely being the writer’s poor grammatical proficiency. These statistics confirm the correlation between the results derived from RFA and GAA respectively, thus supporting Ellis’ (1994) view that L2 learners’ pragmatic com- petence and grammatical competence are interrelated, rather than Swain’s (1985/ 1986:135) claim that discourse competence does not rely heavily on grammar for its re- alisation. Effective argumentative structuring and grammatical accuracy in fact comple- ment each other in making L2 writing successful. Since the high-rated essays display a higher level of hierarchical integration of argu- ment and a greater degree of grammatical accuracy than the low-rated ones, we can fur- ther conclude that there are correlations between teachers’ holistic ratings and the results derived from RFA and GAA respectively.

87 88 Chapter Eight: Conclusion

SYNTHESIS OF FINDINGS

The objectives of the present investigation were threefold. The first was to investigate how L2 tertiary students build up their universe of argumentative discourse, from the lowest level to the highest. The second objective was to examine whether compared with an EFL setting, an ESL environment would bring a difference in the writing proficiency in both organisation and grammar in L2 student writing. Finally, this study aimed at ex- amining the correlation between the grade awarded to student writing and the level of structural organisation and grammatical accuracy evidenced in these texts. All these ob- jectives have been fulfilled. The RFA tool has shed light on the structure of L2 argumentative writing in several ways. We have discovered that the arguments in high-rated essays are hierarchically in- tegrated, as shown by the high frequency of rhetorical functions such as Elaboration, each of which has a role to play in both its immediate and larger contexts of the dis- course, and hence simultaneously contribute to local and global coherence. A large num- ber of conventional coherence patterns of argumentative text (‘Assertion + Justification + Evaluation’) have also been identified in the high-rated texts. On the other hand, the low- rated essays display many instances of ‘Unidentified’ and ‘Unfulfilled’ rhetorical func- tions, which make it impossible for a significant number of coherence patterns to be formed at Propositional Segment Level. There is also strong evidence of plagiarism, and a lack of clear, logical and/or genuine evaluations. Therefore although macrostructures can still be identified in these low-rated essays, they do not contribute to the scaffolding of arguments. It is as if the macrostructures were just attached there, playing no function at all. The second finding is that an ESL environment cannot guarantee development of writing ability. Socio-cultural factors such as the learner’s views of the target-language culture, as well as psychological factors such as motivation, may have an important role to play in the proficiency level a learner reaches. It seemed that even after our subjects had gone to Australia to study, they remained ‘indifferent’, and even became hostile, to the target-language culture, despite increased exposure to it. The reasons were again complex, the most plausible one being their lack of motivation, as these students might neither feel “the need to get meanings across” (because they anticipated returning to Hong Kong to work after graduation), nor experience the “pleasure” associated with communicating with the L1 speakers in Australia. As a result of the interplay of socio- cultural and psychological factors, our Australian L2 subjects seem not to have capital- ised on the ESL environment as much as they might to improve their English, and their recurrent errors were very similar to those made by their Hong Kong counterparts. Yet these Australia-based students were able to produce longer essays, with more accurate tense forms and idiomatic expressions. Their essays also display less serious plagiarism, compared to those written by the Hong Kong subjects. All this suggests that their in- creased exposure to English has helped them to advance along the path of approximation toward the target language in some respects. However, increased exposure to everyday English has not helped the Australian students acquire effective strategies for structuring their writing. The EAP courses they attend may not help much either, if they are taught little more than the basic skills of essay writing, that is, the writing of introductions (background, thesis statements) and conclusions. While elementary L2 students may find these skills useful, they are far too basic to help tertiary students such as our subjects to

89 cope with the demands of academic writing. Rather, these students need to be taught what coherent argumentative writing really entails. This study underscores Cummins’ (1983) hypothesis, that it takes L2 learners longer to acquire academic writing skills than oral skills, since argumentative writing is indeed a communicative event which takes place in a context-reduced situation, and is cognitively demanding. Our third finding is that the results derived from Rhetorical Function Analysis and Grammatical Accuracy Analysis correlated significantly with each other. Effective ar- gumentative structure and grammatical accuracy complement each other; either one of them is a necessary but not sufficient condition for writing success. When we further compare our measures of argumentative coherence and grammatical accuracy, through RFA and GAA, with teachers’ holistic assessments of essay quality, we also find signifi- cant correlations. The design of the study did not encompass the relative emphasis placed by the assessors on organisation or grammatical accuracy when making their holistic judgements. Yet the emphasis of some teachers on grammatical accuracy may account for the instances of plagiarism in the low-rated scripts.

COMPARISON BETWEEN RHETORICAL FUNCTION ANALYSIS AND GRAMMATICAL ACCURACY ANALYSIS

Two analytical tools, Rhetorical Function Analysis (RFA) and Grammatical Accuracy Analysis (GAA), were developed for this research for different purposes. Both tools per- formed their primary function satisfactorily, yet it was found that RFA is more valid a tool for diagnosing the argumentative academic text for three reasons. The first consideration which makes RFA a superior tool to GAA is that the latter cannot reflect the problem of plagiarism, since it is designed to categorise grammatical errors only. RFA, however, has the means to identify this problem directly. When a writer fails to provide the source of a citation, the functional unit with such a problem is categorised as ‘Unfulfilled Rhetorical Function [Reported Evaluation]’, indicating that the unit has failed to perform the function it signals to perform (through the use of quota- tion marks, for example). This is an important parameter, since plagiarism is a serious problem in some student texts, especially the low-rated ones, as evidenced in our investi- gation. Another consideration is the scope of the two tools. GAA was modelled on the anno- tation system employed in Peters & Gassmann (1995), which categorises syntactic prob- lems at the surface, and does not “go beyond language problems embedded in single sen- tences” (ibid.:137). GAA is likewise a tool which can only be applied in the ‘local’ con- text. RFA, in contrast, can be used to diagnose the impacts of local discourse structures on the global structure. This makes it superior to its predecessors such as Functional Role Analysis (Lieber 1979; Lindeberg 1988; Albrechtsen et al. 1991). A third value of RFA is its ability to relate local and global structures in the same process, and to reveal different aspects of essay quality simultaneously. When the results of text analysis are presented graphically through RFA, the analyst can quickly discover which rhetorical relations contribute to both local and global coherence, or disrupt them. Through the final summary of RFA, the underlying causes of the problematic relations, such as low grammatical accuracy and poor referencing skills, can be detected. Thus RFA can perform a triple function, revealing local and global discourse structures, re- flecting the quality of arguments, and exposing underlying problems at the same time. Since RFA can perform more with less, it presents itself as an efficient and effective dis- course analysis tool for research and perhaps also assessment purposes (for example, in

90 validating results obtained from holistic rating). It can be easily adapted from argumenta- tion to other genres, such as exposition, by adding and/or deleting some of its categories. Rhetorical Function Analysis is one type of close textual analysis – sometimes criti- cised as narrow, since the understanding of texts requires multiple perspectives (Candlin & Hyland 1999:2). Although close textual analysis is the primary methodology utilised in this research, the analysis has been situated in the wider contexts of the text’s produc- tion – not only the educational context, but also the linguistic, socio-cultural, and the psychological contexts for L2 students based in Australia and Hong Kong.

IMPLICATIONS FOR TEACHING

An ESL environment does not, then, guarantee success in academic writing. Other fac- tors such as motivation may influence the development of this skill. Teachers should be aware of the ‘push’ and ‘pull’ forces which enhance/constrain L2 learning, and adopt re- active as well as proactive measures by making use of GAA and RFA to attain their goal.

Reactive measures

‘Reactive measures’ are understood here as appropriate remedial action taken by teachers, as a response to the findings of error analysis. Teachers can adopt GAA as a tool to ana- lyse the errors made by students, to provide substance for the writing curriculum, as well as the ‘what’, ‘when’ and ‘how’ of remedial grammar teaching. RFA suggests a different frame of operation. It provides comprehensive information regarding different aspects of essay quality, which teachers can use to analyse the argumentative structure of student essays and help individual students according to their specific needs, through written feedback and/or face-to-face conferencing sessions.

Proactive measures: Towards a model for the teaching of local and global coherence

The most significant contribution RFA can make to the teaching of writing lies in proac- tive measures, that is, action taken by teachers as a heuristic to students’ learning, to en- hance students’ understanding of text structure, and their ability to create sound argu- mentative texts themselves. As we have found, the statistical correlation between the re- sults derived from RFA and GAA suggests that effective argumentative structure and grammatical accuracy are both necessary conditions for writing success. But our findings show that of these two elements, it is the organisational structure which requires more explicit teaching, since increased exposure to L1 cannot help every ESL student to ac- quire the strategies for structuring. Yet the teaching of basic organisational skills alone does not guarantee writing success either. To help students better integrate their argu- ments, teachers should introduce to them the most important part of effective argumenta- tive structuring: the principle of hierarchical integration of argument – the fact that each linguistic unit, be it a clause, a sentence, a propositional segment, or a paragraph, has a definite role to play in both its immediate context, and also in the larger context of the whole text. With adequate exposure and guidance, students will then be able to plan their clauses/sentences/propositional segments/paragraphs and macrostructures strategically, by asking themselves questions such as “Given the overall goals I am trying to accom-

91 plish, what can I say next that will serve them? Given what I just said, what can I say next that is related to it?”, as suggested by Agar & Hobbs (1982:7). Albretchsen et al.’s ‘Dynamic model for written communication’ (cf. Chapter 3, Figure 3) is thus adapted into a ‘Model for the teaching of local and global coherence’:

TEACHING OF

HIERARCHICAL INTEGRATION OF ARGUMENTS

LOCAL/GLOBAL ARGUMENTATIVE STRUCTURES

SENDER TEXT RECEIVER

COGNITIVE GLOBAL ‘TOP-DOWN’ HIERARCHY COHERENCE PROCESSING

SEQUENCE LOCAL ‘BOTTOM-UP’ COHERENCE PROCESSING

Figure 6: Model for the teaching of local and global coherence.

The model shows how the teaching of argumentative structures and their integration can contribute to success in global coherence in L2 writing, and in turn the top-down proc- essing of the raters. The teaching can also enhance students’ skills in achieving local co- herence, which will in turn facilitate the bottom-up processing of the raters. Therefore in this model, solid lines are used to link up ‘Global Coherence’ and ‘Top-down Process- ing’ on one hand, and ‘Local Coherence’ and ‘Bottom-up Processing’ on the other, to replace the broken lines in the Albrechtsen et al. model. In this way our model satisfies the “guiding principle in model building”, “to make explicit the factors that contribute directly to students’ second language writing” (Cumming 1998:68). An additional benefit of adopting this model in teaching is that once students master the argumentative structures, they will not have to struggle with their organisational strategies and language at the same time. This is especially useful for students writing argumentative text, which was found to have the highest correlation with syntactic com- plexity, as compared to descriptive and narrative essays (cf. Veliz 1999). They can then concentrate on grammatical accuracy, to produce argumentative essays which are organi- sationally and linguistically sound. Therefore RFA not only has research but also peda- gogical value. Cumming (op. cit.:66) notes,

learning to write may largely be a process of “personal growth in social context.” For this reason, cultural, contextual, and individual differences cry out to be better appreciated and understood if educational theories and practices are to be relevant to these aims.

92 What I hope to have achieved through this study is: firstly, to have contributed to the un- derstanding of the interplay of “cultural, contextual, and individual differences” in L2 students’ experience of learning to write (how ethnic Chinese students originating from Hong Kong learn to write in English in two different socio-cultural and linguistic con- texts, Australia and Hong Kong); and secondly, to have demonstrated the value of two methodologies: one for integrating the analysis of local and global coherence, and the other for identifying students’ grammatical errors. With these two methodologies, L2 students can be helped to achieve local and global coherence simultaneously in their writing, and to deal with their errors systematically.

93 94 References

Agar, M. & Hobbs, J. R. 1982. Interpreting discourse: Coherence and the analysis of ethno- graphic interviews. Discourse Processes, 5, 1–32. Albrechtsen, D., Evensen, L. S., Lindeberg, A. & Linnarud, M. 1991. Analysing developing dis- course structure: The NORDWRITE Project. In R. Phillipson, E. Kellerman, L. Selinker, M. S. Smith & M. Swain (Eds), Foreign/Second language pedagogy research: A commemora- tive volume for Claus Færch (pp. 79–91). Clevedon & Philadelphia PA: Multilingual Matters. Aston, G. 1977. Comprehending value: Aspects of argumentative discourse. Studi italiani di lin- guistica teorica ed applicata, 6 (3), 465–509. Ballard, B. & Clanchy, J. 1991. Assessment by misconception: Cultural influences and intellec- tual traditions. In L. Hamp-Lyons (Ed.), Assessing second language writing in academic con- texts (pp. 19–35). Norwood NJ: Ablex. Barker, M., Child, C., Gallois, C., Jones, E. & Callan, V. J. 1991. Difficulties of overseas stu- dents in social and academic situations. Australian Journal of Psychology, 43 (2), 79–84. Beaugrande, R. A. de. 1980. Text, discourse and process. Toward a multidisciplinary science of texts. Norwood NJ: Ablex. Belcher, D. & Braine, G. 1995. Academic writing in a second language: Essays on research and pedagogy. Norwood NJ: Ablex. Berns, M. 1990. ‘Second’ and ‘foreign’ in second language acquisition/foreign language learning: A sociolinguistic perspective. In B. VanPatten & J. F. Lee (Eds), Second language acquisi- tion/foreign language learning (pp. 3–11). Clevedon & Philadelphia PA: Multilingual Mat- ters. Bialystok, E. 1997. The structure of age: In search of barriers to second language acquisition. Second Language Research, 13, 116–37. Biggs, J. B. & Watkins, D. A. 1996. The Chinese learner in retrospect. In D. A. Watkins & J. B. Biggs (Eds), The Chinese learner: Cultural, psychological and contextual influences (pp. 269–85). Hong Kong: Comparative Education Research Centre, Faculty of Education, Uni- versity of Hong Kong. Boyle, J. 1987. Sex differences in listening vocabulary. Language Learning, 37, 273–84. Bradley, D. & Bradley, M. 1984. Problems of Asian students in Australia: Language, culture and education. Canberra: Australian Government Publishing Service. Bridgeman, B. & Carlson, S. 1983. Survey of academic writing tasks required of graduate and undergraduate foreign students. Princeton NJ: Educational Testing Service. Briguglio, C. 1998. Non-English speaking background students’ perceptions of their linguistic and educational needs. Perth: Curtin University of Technology. 2000. Generic skills: Attending to the communication skills needs of international students. Paper presented at the Inaugural Ceremony, Lifelong Learning Conference, Yepoon. July. Brooks, G. & Adams, M. 2001. Spoken English usage: Looking at its relationship to the aca- demic performance of first-year students in a business unit. Unpublished report. Sydney: Business Studies Department, Macquarie University. Bruce, N. 1990. EL2-medium education in a largely monolingual society: The case of Hong Kong. Hongkong Papers in Linguistics and Language Teaching, 13, 9–23. Bunton, D. 1989. Common English errors in Hong Kong. Hong Kong: Longman. Burstall, C. 1975. Factors affecting foreign-language learning: A consideration of some relevant research findings. Language Teaching and Linguistics Abstracts, 8, 105–25. Callow, K. & Callow, J. 1992. Text as purposive communication: A meaning-based analysis. In W. C. Mann & S. A. Thompson (Eds), Discourse description: Diverse linguistic analyses of a fund-raising text (pp. 5–37). Amsterdam: Benjamins. Canale, M. 1983. From communicative competence to communicative language pedagogy. In J. C. Richards, & R. W. Schmidt (Eds), Language and communication (pp. 2–25). London & New York: Longman.

95 Canale, M. & Swain, M. 1980. Theoretical bases of communicative approaches to second lan- guage teaching and testing. Applied Linguistics, 1, 1–47. Candlin, C. N. & Hyland, K. (Eds). 1999. Writing: Texts, processes and practices. London & New York: Longman. Charney, D. 1984. The validity of using holistic scoring to evaluate writing: A critical overview. Research in the Teaching of Writing, 18, 65–81. Cheung, G. 2001. Rosanna Wong says English is the key issue. South China Morning Post, 3rd May, p. 4. Chomsky, N. 1965. Aspects of the theory of syntax. Cambridge MA: MIT Press. Christie, F. 1987. Genres as choice. In I. Reid (Ed.), The place of genre in learning: Current de- bates (pp. 22–34). Melbourne: Centre for Studies in Literacy Education, Deakin University. Clyne, M. 1981. Culture, discourse and communicative competence. Proceedings of the Second Study Skills Conference (pp. 101–15). Melbourne: La Trobe University. Coley, M. 1999. The English language entry requirements of Australian universities for students of non-English speaking background. Higher Education Research & Development, 18 (1), 7– 17. Connor, U. 1984. A study of cohesion and coherence in English as a second language students’ writing. Papers in Linguistics: International Journal of Human Communication, 17 (3), 301– 16. 1987. Argumentative patterns in student essays: Cross-cultural differences. In U. Connor & R. B. Kaplan (Eds), Writing across languages: Analysis of L2 texts (pp. 57–71). Reading MA: Addison-Wesley. Connor, U. & Farmer, M. 1985. The teaching of topical structure analysis as a revision strategy: An exploratory study. Paper presented at the American Educational Research Association Conference, Chicago, April. Connor, U. & Lauer, J. 1986. Understanding persuasive essay writing, linguistic/rhetorical ap- proach. In L. S. Evensen (Ed), Nordic research in text linguistics and discourse analysis (pp. 245–66). Trondheim: Tapir. Connor-Linton, J. 1995. Crosscultural comparison of writing standards: American ESL and Japanese ESL. World Englishes, 14 (1), 99–115. Cooper, C. R. 1977. Holistic evaluation of writing. In C. R. Cooper & L. Odell (Eds), Evaluating writing: Describing, measuring, judging (pp. 3–32). Urbana IL: National Council of Teach- ers of English. Cooper, C. R. & Matsuhashi, A. 1983. A theory of the writing process. In M. Martlew (Ed.), The psychology of written language (pp. 3–40). Chichester: Wiley. Cortazzi, M. 1990. Cultural and educational expectations in the language classroom. In B. Harri- son (Ed.), Culture and the language classroom. ELT Documents 132 (pp. 54–65). London: Modern English Publications/British Council. Crombie, W. 1985. Process and relation in discourse and language learning. Oxford: Oxford University Press. Crystal, D. 1995. The Cambridge encyclopaedia of the English language. Cambridge: Cambridge University Press. Cumming, A. 1998. Theoretical perspectives on writing. Annual Review of Applied Linguistics, 18, 61–78. Cummins, J. 1979. Cognitive/academic language proficiency, linguistic interdependence, the op- timal age question and some other matters. Working Papers on Bilingualism 19, 197–205. 1983. Language proficiency and academic achievement. In J. W. Oller, Jr. (Ed.), Issues in language testing research (pp. 108–26). Rowley MA: Newbury House. Reprinted in J. Cummins & M. Swain (Eds), Bilingualism in education (pp. 138–61). New York: Longman, 1986. Education Commission (Hong Kong) 1995. Education Commission Report No. 6 (Parts 1 & 2). Enhancing language proficiency: A comprehensive strategy. Hong Kong: Government Printer. Ellis, R. 1994. The study of second language acquisition. Oxford: Oxford University Press. 1997. Second language acquisition. Oxford: Oxford University Press.

96 Enkvist, N. E. 1984. Contrastive linguistics and text linguistics. In J. Fisiak (Ed.), Contrastive linguistics: Prospects and problems (pp. 45–67). The Hague: Mouton. 1990. Seven problems in the study of coherence and interpretability. In U. Connor & A. M. Johns (Eds), Coherence in writing: Research and pedagogical perspectives (pp. 9–28). Alex- andria VA: Teachers of English to Speakers of Other Languages. Entwistle, N. J. & Ramsden, P. 1982. Student learning in its context. In N. J. Entwistle & P. Ramsden (Eds), Understanding student learning (pp. 1–5). New York: Nichols. Falvey, P. 1993. Towards a description of corporate text revision. PhD, University of Birming- ham. 1998. ESL, EFL & language acquisition in the context of Hong Kong. In B. Asker (Ed.), Teaching language and culture: Building Hong Kong on education (pp. 73–85). Hong Kong: Addison Wesley Longman. Field, Y. 1993. Piling on the additives: The Hong Kong connection. In R. Pemberton & E. S. C. Tsang (Eds), Studies in lexis (pp. 247–68). Hong Kong: Language Centre, Hong Kong Uni- versity of Science and Technology. 1994. Cohesive conjunctions in the English writing of Cantonese speaking students from Hong Kong. Australian Review of Applied Linguistics, 17 (1), 125–39. Field, Y. & Yip, L. 1992. A comparison of internal cohesive conjunction in the English essay writing of Cantonese speakers and native speakers of English. RELC Journal, 23, 15–28. Frederiksen, C. H. 1986. Cognitive models and discourse analysis. In C. R. Cooper & S. Greenbaum (Eds), Studying writing: Linguistic approaches (pp. 227–67). Beverly Hills CA: Sage. Freedman, J. L. 1988. Television violence and aggression: What the evidence shows. In Oskamp, S. (Ed.), Television as a social issue (pp. 144–62). London: Sage. Freedman, S. 1981. Influences on evaluators of expository essays: Beyond the text. Research in the Teaching of Writing, 15, 245–55. Gardner, R. C. 1985. Social psychology and second language learning: The role of attitudes and motivation. London: Arnold. Gardner R. C. & Lambert, W. 1972. Attitudes and motivation in second language learning. Row- ley MA: Newbury House. Gass, S. & Selinker, L. 2001. Second language acquisition: An introductory course (2nd Edition). Mahwah NJ: Erlbaum. Gibbons, J. 1987. Code-mixing and code choice: A Hong Kong case study. Clevedon: Multilin- gual Matters. Grabe, W. 1985. Written discourse analysis. In R. B. Kaplan (Ed.), Annual Review of Applied Linguistics, 5, 101–23. 1996. Genre analysis, discourse coherence and literacy instruction. Presentation at the 11th World Congress of Applied Linguistics, Jyväskylä, August. Grabe, W. & Kaplan, R. B. 1996. Theory and practice of writing. London: Longman. Grobe, C. 1981. Syntactic maturity, mechanics and vocabulary as predictors of quality ratings. Research in the Teaching of Writing, 15, 75–86. Guba, E. G. & Lincoln, Y. S. 1989. Fourth generation evaluation. Newbury Park CA: Sage. Halliday, M. A. K. & Hasan, R. 1976. Cohesion in English. London: Longman. Hamp-Lyons, L. 1988. The product before: Task-related influences on the writer. In P. Robinson (Ed.), Academic writing: Process and product (pp. 35–46). ELT Documents No. 129. London: MEP/Macmillan. (Ed.). 1991. Assessing second language writing in academic contexts. Norwood NJ: Ablex. Hamp-Lyons, L. & Reed, R. 1990. Development of the new Michigan Writing Assessment (Re- port to the College of LS & A, University of Michigan). Ann Arbor MI: English Composi- tion Board. Hartmann, R. R. K. & James, G. 1998. Dictionary of lexicography. London & New York: Routledge. Hatch, E. & Lazaraton, A. 1991. The research manual: Design and statistics for applied linguis- tics. New York NY: Newbury House.

97 Heidelberger Forschungsprojekt. 1978. The acquisition of German syntax by foreign migrant workers: Heidelberger Forschungsprojekt ‘Pidgin Deutsch’. In D. Sankoff (Ed.), Linguistic variation: Models and methods (pp. 1–22). New York: Academic Press. Hellmundt, S., Rifkin, W. & Fox, C. 1998. Enhancing intercultural communication among busi- ness communication students. Higher Education Research & Development, 17 (3), 333–44. Hoey, M. 1979. Signalling in discourse. Birmingham: English Language Research, University of Birmingham (Discourse analysis monograph No. 6). 1983. On the surface of discourse. London: Allen & Unwin. 1991. Patterns of lexis in text. Oxford: Oxford University Press. Hunt, K. W. 1970. Syntactic maturity in school children and adults. Monographs of the Society for Research in Child Development, 35 (1). 1977. Early blooming and late blooming syntactic structures. In C. R. Cooper & L. Odell (Eds), Evaluating writing: Describing, measuring, judging (pp. 91–104). Urbana, IL: Na- tional Council of Teachers of English. 1983. Sentence combining and the teaching of writing. In M. Martlew (Ed.), The psychol- ogy of written language (pp. 99–125). Chichester: Wiley. Hymes, D. 1972. On communicative competence. In J. Pride & A. Holmes (Eds), Sociolinguis- tics (pp. 269–93). Harmondsworth & New York: Penguin. Jacobs, S. 1990. Building hierarchy: Learning the language of the science domain, ages 10–13. In U. Connor & A. M. Johns (Eds), Coherence in writing: Research and pedagogical perspec- tives (pp. 151–68). Alexandria VA: Teachers of English to Speakers of Other Languages. James, G. 2001. Many a slip …: A dictionary of shaky grounds and missed connections. Vocabu- lary for discussion of argument. Hong Kong: Language Centre, Hong Kong University of Science and Technology. Johns, A. 2000. What are genres? And why should teachers care? In A. Furness, H. Y. G. Wong & L. Wu (Eds), Penetrating discourse: Integrating theory and practice in second language teaching (pp. 9–20). Hong Kong: Language Centre, Hong Kong University of Science and Technology. Johnson, J. & Newport, E. 1989. Critical period effects in second language learning: The influ- ence of maturational state on the acquisition of ESL. Cognitive Psychology, 21, 60–99. Jordan, M. P. 1984. Rhetoric of everyday English texts. London: Allen & Unwin. 1992. An integrated three-pronged analysis of a fund-raising letter. In W. C. Mann & S. A. Thompson (Eds), Discourse description: Diverse linguistic analyses of a fund-raising text (pp. 171–226). Amsterdam: Benjamins. Jordan R. R. 1997. English for academic purposes: A guide and resource book for teachers. Cambridge: Cambridge University Press. Kaldor, S., Herriman, M. & Rochecouste, J. 1998. Tertiary student writing. Sydney: National Centre for English Teaching and Research, Macquarie University. Kaplan, R. 1966. Cultural thought patterns in intercultural education. Language Learning, 16, 1– 20. 1987. Cultural thought patterns revisited. In U. Connor & R. B. Kaplan (Eds), Writing across languages: Analysis of L2 texts (pp. 9–22). Reading MA: Addison-Wesley. Keller-Cohen, D. & Wolfe, A. 1987. Extended writing in the College of Literature, Science and the Arts: Report of a faculty survey. Ann Arbor MI: English Composition Board. Kember, D. & Gow, L. 1990. Cultural specificity of approaches to study. British Journal of Edu- cational Psychology, 60, 356–63. Kummer, W. 1972. Aspects of a theory of argumentation. In E. Gulich & W. Raible (Eds), Textsorten: Differenzierungskriterien aus linguistiächer Sicht (pp. 25–49). Frankfurt am Main: Athenaeum. Larsen-Freeman, D. & Long, M. 1991. An introduction to second language acquisition research. London: Longman. Laufer, B. 1994. The lexical profile of second language writing: Does it change over time? RELC Journal, 25 (1), 21–33. Laufer, B. & Nation P. 1995. Vocabulary size and use: Lexical richness in L2 written production. Applied Linguistics, 16 (3), 307–22.

98 Lautamatti, L. 1978. Observations on the development of the topic in simplified discourse. In V. Kohenen & N. E. Enkvist (Eds), Text linguistics, cognitive learning and language teaching (pp. 71–104). Turku: AFinLa. Reprinted in U. Connor & R. B. Kaplan (Eds), Writing across languages: Analysis of L2 text (pp. 87–113). Reading MA: Addison-Wesley, 1987. 1980. Subject and theme in English discourse. In K. Sajavaara & J. Lehtonen (Eds), Pa- pers in discourse and contrastive discourse analysis (pp. 188–202). Jyväskylä: Department of English, University of Jyväskylä. 1990. Coherence in spoken and written discourse. In U. Connor & A. M. Johns (Eds), Co- herence in writing: Research and pedagogical perspectives (pp. 29–42). Alexandria VA: Teachers of English to Speakers of Other Languages. Li, D. C. S., Lin, A. & Tsang, W. K. 2000. Language and education in postcolonial Hong Kong. Hong Kong: Linguistic Society of Hong Kong. Lieber, P. E. 1979. Cohesion in ESL students’ expository writing: A descriptive study. PhD, New York University. Lindeberg, A.-C. 1985a. Abstraction levels in student essays. Text, 5(4), 327–45. 1985b. Cohesion, coherence patterns, and EFL essay evaluation. In N. E. Enkvist (Ed.), Coherence and composition: A symposium (pp. 62–92). Åbo: Research Institute of the Åbo Akademi Foundation. 1988. Cohesion, coherence, and coherence patterns in expository and argumentative stu- dent essays in EFL: An exploratory study. Åbo: Research Institute of the Åbo Akademi Foundation. 1994. Towards rhetorical structure in EFL: Strategies of development in expository and argumentative student writing. In P. Skyum-Nielsen & H. Schröder (Eds), Rhetoric and sty- listics today: An international anthology (pp. 138–48). Frankfurt am Main: Lang. Longacre, R. E. 1992. The discourse strategy of an appeals letter. In W. C. Mann & S. A. Thompson (Eds), Discourse description: Diverse linguistic analyses of a fund-raising text (pp. 109–30). Amsterdam: Benjamins. Lord, R. 1974. English – How serious a problem for students in Hong Kong? English Bulletin, 6 (3), 1–10. Luke, K. K. & Richards, J. C. 1982. English in Hong Kong: Functions and status. English World- wide, 3(1), 47–64. Lukmani, Y. 1995. Linguistic accuracy versus coherence in assessing examination answers in content subjects. In M. Milanovic & N. Saville (Eds), Performance testing, cognition and as- sessment: Selected papers from the 15th Language Testing Research Colloquium (LTRC) (pp. 130–50). Cambridge: Cambridge University Press. Macquarie University. 1999. Macquarie University Annual Report 1998. Sydney. 2000. Macquarie University Annual Report 1999. Sydney. Malcolm, I. & Honjio, P. 1988. Argumentation patterns in contemporary Chinese: Implications for English teaching. In V. C. Bickley (Ed.), Languages in education in a bilingual or multi- lingual setting (pp. 321–33). Hong Kong: Institute of Language in Education. Mann, W. C. & Thompson, S. A. 1986. Rhetorical Structure Theory: Description and construc- tion of text structures. ISI Reprint Series 86–174. Marina del Rey CA: Information Sciences Institute, University of Southern California. 1987. Rhetorical Structure Theory: A theory of text organization. In L. Polanyi (Ed.), The structure of discourse (pp. 85–96). Norwood NJ: Albex. Page numbers cited from ISI Reprint Series 87–190. Marina del Rey CA: Information Sciences Institute, University of Southern California. 1988. Rhetorical Structure Theory: Towards a functional theory of text organization. Text, 8 (3), 243–81. Matthiessen, C. & Thompson, S. A. 1988. The structure of discourse and ‘subordination’. In J. Haiman & S. Thompson (Eds), Clause combining in grammar and discourse (pp. 275–329). Amsterdam: Benjamins. McArthur, T. 1996. The Oxford companion to the English language. Oxford: Oxford University Press.

99 McCarthy, M. 1991. Discourse analysis for language teachers. Cambridge: Cambridge Univer- sity Press. McCutchen, D. & Perfetti, C. A. 1982. Coherence and connectedness in the development of dis- course production. Text, 2, 113–9. McNamara, J. 1973. Nurseries, streets and classrooms: Some comparisons and deductions. Mod- ern Language Journal, 57, 250–5. Meiland, J. W. 1989. Argument as inquiry and argument as persuasion. Argumentation, 3, 185– 96. Meyer, B. J. F. 1992. An analysis of a plea for money. In W. C. Mann & S. A. Thompson (Eds), Discourse description: Diverse linguistic analyses of a fund-raising text (pp. 79–108). Am- sterdam: Benjamins. Michell, M. 1999. Wither ESL? Post-literacy prospects for English as a second language pro- grams in Australian schools. Prospect, 14 (2), 4. Milanovic, M., Saville, N. & Shen, S. 1995. A study of the decision-making behaviour of com- position markers. In M. Milanovic & N. Saville (Eds), Performance testing, cognition and assessment: Selected papers from the 15th Language Testing Research Colloquium (LTRC) (pp. 92–114). Cambridge: Cambridge University Press. Milavsky, J. R. 1988. Television and aggression once again. In Oskamp, S. (Ed.), Television as a social issue (pp. 163–70). London: Sage. Milton, J. 2001. Elements of a written interlanguage: A computational and corpus-based study of institutional influences on the acquisition of English by Hong Kong Chinese students. Re- search Reports, Volume Two. Hong Kong: Language Centre, Hong Kong University of Sci- ence and Technology. Milton, J. & Tsang, E. S. C. 1993. A corpus-based study of logical connectors in EFL students’ writing: Directions for future research. In R. Pemberton & E. S. C. Tsang (Eds), Studies in lexis (pp. 215–46). Hong Kong: Language Centre, Hong Kong University of Science and Technology. Mullins, G., Quintrell, N. & Hancock, L. 1995. The experiences of international and local stu- dents in three Australian universities. Higher Education Research and Development, 14, 201–32. Neilson, L. & Piche, G. 1981. The influence of headed nominal complexity and lexical choice on teachers’ evaluation of writing. Research in the Teaching of Writing, 15, 65–74. Newcomb, J. S. 1977. The influence of readers on the holistic grading of essays. PhD, University of Michigan. Nixon, U. 1996. Adjusting to Australia: Insights from student journals in a TESOL program. Prospect, 11 (3), 29–44. Nold, E. & Freedman, S. 1977. An analysis of readers’ responses to essays. Research in the Teaching of Writing, 11, 164–74. O’Donnell, R. C 1976. A critique of some indices of syntactic maturity. Research in the Teach- ing of English, 10, 31–8. Oller, J. W., Jr. 1976. Evidence for a general proficiency factor: An expectancy grammar. Die Neueren Sprachen 2: 165–74. 1979. Language tests at school: A pragmatic approach. London: Longman. Pennington, M. C. (Ed.). 1998. Language in Hong Kong at century’s end. Hong Kong: Hong Kong University Press. Peters, P. & Gassmann, B. 1995. Acquiring standard English. In P. Peters (Ed.), Australian Eng- lish in a pluralist Australia: Proceedings of Style Council 95 (pp. 135–46). Sydney: Mac- quarie University Dictionary Research Centre. Purves, A. & Takala, S. (Eds). 1982. An international perspective on the evaluation of written composition. London: Pergamon. Reid, I., Kirkpatrick, A. & Mulligan, D. 1998. Framing reading. Sydney: National Centre for English Language Teaching and Research, Macquarie University. Reid, J. 1984a. The radical onliner and the radical brainstormer: A perspective on composing processes. TESOL Quarterly, 18, 529–33.

100 1984b. Comments on Vivian Zamel’s “The composing process of advanced ESL students: Six case studies”. TESOL Quarterly, 18, 149–59. Richards, K. & Skelton, J. 1991. How critical can you get? In P. Adams, B. Heaton & P. How- arth (Eds), Socio-cultural issues in English for Academic Purposes (pp. 24–40). London: Modern English Publications / British Council. Rossier, J. 1975. Extroversion-introversion as a significant variable in the learning of English as a second language. PhD, University of Southern California. Schumann, J. 1978a. The acculturation model for second language acquisition. In R. Gingras (Ed.), Second language acquisition and foreign language teaching (pp. 27–50). Arlington VA: Center for Applied Linguistics. 1978b. The acquisition of English negation by speakers of Spanish: A review of the litera- ture. In R. Anderson (Ed.), The acquisition and use of Spanish and English as first and sec- ond languages (pp. 3–32). Washington DC: Teachers of English to Speakers of Other Lan- guages. Schwartz, S. & Li, A. 2001. 40pc of young blame poor language skills on lack of practice. South China Morning Post, 26th November, p. 6. Seliger, H. W. & Shohamy, E. 1989. Second language research methods. Oxford: Oxford Uni- versity Press. Selinker, L. 1972. Interlanguage. IRAL, 10, 219–31. 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). Cambridge: Cambridge University Press. Sinclair, J. M. & Coulthard, R. M. 1975. Towards an analysis of discourse. London: Oxford University Press. Singer, J. L. & Singer, D. G. 1988. Some hazards of growing up in a television environment: Children’s aggression and restlessness. In Oskamp, S. (Ed.), Television as a social issue (pp. 171–88). London: Sage. Slavoff, G. & Johnson, J. 1995. The effects of age on the rate of learning a second language. Studies in Second Language Acquisition, 17, 1–16. Smith, S. N., Miller, R. J. & Crassini, B. 1998. Approaches to studying of Australian and over- seas Chinese university students. Higher Education Research & Development, 17 (3), 261– 76. So, D. W. C. 1998. One country, two cultures and three languages: Sociolinguistic conditions and language education in Hong Kong. In B. Asker (Ed.), Teaching language and culture: Building Hong Kong on education (pp. 152–76). Hong Kong: Addison Wesley Longman. Sparks, J. 1988. Syntactic complexity, error and the holistic evaluation of ESL student essays. ORTESOL Journal, 9, 35–49. Spolsky, B. 1989. Conditions for second language learning. Oxford: Oxford University Press. Stern, H. H. 1983. Fundamental concepts of language teaching. Oxford: Oxford University Press. Stewart, M. & Grobe, C. 1979. Syntactic maturity, mechanics of writing and teachers’ quality rating. Research in the Teaching of Writing, 13, 207–15. Stuart-Smith, V. & Busbridge, R. 1996. A picture paints a thousand words: Rhetorical Structure Theory & tertiary literacy. Paper presented at the Applied Linguistics Association of Austra- lia Conference, Sydney, September. Swain, M. 1985. Communicative competence: Some roles of comprehensible input and compre- hensible output in its development. In S. Gass & C. Madden (Eds), Input in second language acquisition (pp. 235–53). Reprinted in J. Cummins & M. Swain (Eds), Bilingualism in edu- cation (pp. 116–37). New York NY: Longman, 1986. Thompson, S. A. & Mann, W. C. 1987. Rhetorical Structure Theory: A framework for the analy- sis of texts. IPRA Papers in Pragmatics, 1, 79–105. Tirkkonen-Condit, S. 1984. Towards a description of argumentative text structure. In H. Ring- bom & M. Rissanen (Eds), Proceedings from the Second Nordic Conference for English Studies: Hanasaari/Hanaholmen, 19–21 May, 1983 (pp. 221–36). Åbo: Research Institute of the Åbo Akademi Foundation.

101 1985. Argumentative text structure and translation. PhD, University of Jyväskylä. 1989. Topic and structure in argumentation. In J. Niemi (Ed.), Papers from the Eleventh Scandinavian Conference of Linguistics (pp. 416–33). Joensuu: University of Joensuu. Toulmin, S. 1958. The uses of argument. Cambridge: Cambridge University Press. Toulmin, S., Rieke, R. & Janik, A. 1984. An introduction to reasoning (Second edition). New York: Macmillan / London: Collier Macmillan. Vann, R. J., Lorenz, F. O. & Meyer, D. M. 1991. Error gravity: Faculty response to errors in the written discourse of nonnative users of English. In L. Hamp-Lyons (Ed.), Assessing second language writing in academic contexts (pp. 181–96). Norwood NJ: Ablex. Vaughan, C. 1991. Holistic assessment: What goes on in the raters’ minds? In L. Hamp-Lyons (Ed.), Assessing second language writing in academic contexts (pp. 111–26). Norwood NJ: Ablex. Veliz, M. 1999. Syntactic complexity and discourse type. Estudios Filológicos, 34, 181–90. Villegas, C.-A. 1998. The semantic macrostructure and argumentative texts: Instructional strat- egy design. Letras, 56, 23–52. 1999. Text linguistics and the writing process: An evaluation of learner strategies using participatory action research. Lectura-y-Vida, 20 (3), 44–53. Volet, S. & Renshaw, P. 1996. Chinese students at an Australian university: Adaptability and continuity. In D. Watkins & J. Biggs (Eds), The Chinese learner: Cultural, psychological and contextual influences (pp. 205–20). Hong Kong: Comparative Education Research Cen- tre, Faculty of Education, . Watson, C. 1983. Syntactic change: Writing development and the rhetorical context. In M. Martlew (Ed.), The psychology of written language (pp. 127–40). Chichester: Wiley. Werlich, E. 1976. A text grammar of English. Heidelberg: Quelle & Meyer. White, E. 1984. Holisticism. College Composition and Communication, 35, 400–9. Widdowson, H. G. 1973. Directions in the teaching of discourse. In E. Roulet & S. P. Corder (Eds), Theoretical linguistic models in applied linguistics (pp. 65–76). Brussels: AIMAV. Winter, E. O. 1971. Connection in science material: A proposition about the semantics of clause relations. In Papers and Reports. No. 7 (pp. 41–52). London: Centre for Information on Lan- guage Teaching and Research, for the British Association for Applied Linguistics. 1974. Replacement as a function of repetition: A study of its principal features in the clause relations of contemporary English. PhD, University of London. 1976. Fundamentals of information structure: Pilot manual for further development ac- cording to student need. Unpublished document, Hatfield Polytechnic. 1977. A clause-relational approach to English texts: A study of some predictive lexical items in written discourse. Instructional Science (Special Issue), 6 (1), 1–92. 1992. The notion of unspecific versus specific as one way of analysing the information of a fund-raising letter. In W. C. Mann & S. A. Thompson (Eds), Discourse description: Di- verse linguistic analyses of a fund-raising text (pp. 131–70). Amsterdam: Benjamins. Witte, S. P. 1983. Topical structure and revision: An exploratory study. College Com-position and Communication, 34 (3), 313–41. Witte, S. P. & Davis, A. S. 1980. The stability of T-unit length: A preliminary investigation. Re- search in the Teaching of English, 14, 5–17. Wolfe-Quintero, K., Inagaki, S. & Kim, H-Y 1998. Second language development in writing: Measures of fluency, accuracy and complexity. Honolulu HI: Second Language Teaching and Curriculum Center, University of Hawaii. Wong, H. Y. G. 1995. Coherence rating: What goes on in the raters’ minds? In D. Nunan, R. Berry & V. Berry (Eds), Language awareness in language education (pp. 149–68). Hong Kong: Department of Curriculum Studies, University of Hong Kong. 2001a. How do first-year students evaluate their writing skills? Unpublished report. Hong Kong: Language Centre, Hong Kong University of Science and Technology. 2001b. The assessment criteria and the perception of assessors: Match or mismatch? Un- published report. Hong Kong: Language Centre, Hong Kong University of Science and Technology.

102 Wong, K.-W. 1993. An analysis of matriculants’ writing with special reference to communica- tive functions. MEd, University of Hong Kong. Yancey, K. B. 1999. Looking back as we look forward: Historicizing writing assessment. Col- lege Composition and Communication, 50 (3), 483–503. Zamel, V. 1983. The composing processes of advanced ESL students: Six case studies. TESOL Quarterly, 17, 165–87.

103 104 Appendix 1: Taxonomy of Rhetorical Function Analysis

Rhetorical Functions at the Lower Levels

(Functional Unit/Simple Sentence/Complex Sentence and Propositional Segment Level)

Rhetorical Definition Example(s) Function TRUTH & VALIDITY *Assertion Introduces a new topic or an Nowadays, televisions are common in every (Ass) aspect of a new topic which is family and they have become a necessity in not covered in a previous part people’s daily life. (HKH2 S1a–b) of the text Justification Establishes explicitly the ap- Thirdly, Freedman believes that the results (Just) propriateness or acceptability of from the laboratory experiments cannot an idea/assertion already intro- generalise the real life situation. This is be- duced in the text; often used cause he believes that laboratory experi- together with Evalua- ments tend to localise social learning. (AH6 tion/Reported Evaluation S21–S22) Concession While admitting or making al- Many psychologists and sociologists have (Conc) lowance for the truth/validity of carried out a lot of experiments to seek for a fact/statement already intro- the answer. However, there are different duced, or to be introduced, in ideas about the issue and it seems hard to another part of the text, this come to a unique conclusion. (HKH3 S4– rhetorical function expresses a S5) reservation of the fact/statement. *Evaluation Presents writer’s own opinion or After examining those three areas of ex- (Eval) judgment on a fact/statement periments towards the relationship between already introduced, or to be TV viewing and its influences on children’s introduced, in the text aggressive behaviours, it seems that we do not have strong enough evidence to support the conclusion that TV viewing has definite influences on children’s violent behaviour. (HKH2 S28a–b) Reported Quotes or relates the opinion or … [A]s Singer and Singer (1988:180) say, Evaluation judgment of another person on “while experiments in the field present (RepE) a fact/statement already intro- enormous logistical and control problems, duced/ to be introduced in the the bulk of date from such studies again text points to the potential influences of televi- sion on aggressive behaviour”… . (HKH2 S21a) CAUSE & EFFECT Cause Identifies the cause for an event The samples were from diverse socio and (Cau) or assertion presented in an- ethnic groups and from two cities in the other part of the text; often used midwest of United States. These cities were together with Result chosen because they had a longer air time of “adult” programmes. (AH6 S51–S52)

105

Rhetorical Definition Example(s) Function Result Indicates an effect, out- If experiments are conducted under subjective (Resu) come, or consequence of conditions, then the findings could be biased. a) an event or asser- (AH6 S25a–b) tion b) a hypothetical or conditional case. This rhetorical function is often used together with Result/Condition. Condition States the situation, includ- If experiments are conducted under subjective (Conc) ing a hypothetical situation, conditions, then the findings could be biased. under which an assertion (AH6 S25a–b) holds true; often introduced by ‘when’, ‘as’ or ’if’ Purpose Indicates what intention an a) In this paper, I would like to introduce (Purp) action mentioned in another various ideas and experiments concerned part of the text aims at to our readers so we can make a fair judge about them. (HKH3 S6a–b)

b) In the past decades, many thousands of laboratory experiments had been carried out to see whether children exposing to violent programs would increase their ag- gressive behaviour. (HKH2 S7a–b) DISCOURSE DEVELOP- MENT *Elaboration a) Expands on the specific a) Although laboratory experiments have (Elab) details of an proved very consistent results, Freedman idea/object/event already (1988) has expressed his concerns about introduced in the text; the reliability of the findings. Firstly, he has b) Makes a general questioned the objectivity of the experi- proposition more menters. (AH6 S15–S16) precise; b) The last type of experiment is naturalistic c) Defines a term experiment. It is a field experiment but it is not intentionally carried out. (HKH3 S28– S29) Exemplification Illustrates an idea already However, some researchers have found no (Exem) introduced in the text which evidence or very weak evidence in laboratory is more general in nature studies … .For examples, Lovaas, 1961 and Hapkiewiez and Roden, 1971, found neither no impact of television violence on aggression or only in some conditions. (AH4 S26–S27) Comparison a) Points out the similar- Detailed examinations … showed no effects (Comp) ity/similarities between on the crime rates but only a little influence on two or more the amount of petty theft … . Other re- facts/statements searches showed that children who watch b) Points out the differ- television frequently behave more aggressive ence(s) between two or than those who watch less did. (HKH1 S28– more facts/statements S29)

106

Rhetorical Definition Example(s) Function Coupling The second member adds This is essential in interpreting the results be- (Coup) at least one new proposi- cause if anger is a precondition in the test, tion to the first and the then the increase in children’s aggressive be- members are not con- haviour after viewing violent TV programmes nected in an elective, a is just a release of impulses because of anger comparative or a sequential but not the consequences of viewing violent way TV programmes. Besides, Freedman also states that viewing TV programmes can arouse any dominant responses, not just ag- gression. (HKH2 S14–S15)

Alternation Involves a choice between It was not clear from the analysis whether ag- (Alt) two or more antithetical or gressive behaviour was caused by the TV non-antithetical choices violence or was the influence of other factors such as cultural background, physical pun- ishment at home etc. (AH6 S57) Summary Repeats the main points or In a nutshell, both the arguments of Freedman (Sum) an idea/event already intro- and that of Singer and Singer were based on duced in the text in a more laboratory experiments, field experiments and concise manner, without natural experiments. (HKH6 S30) additional information Restatement Rephrases or repeats an … [W]e can conclude that television watching (ReS) idea already introduced in have some negative effects on children’s be- the text on approximately havior. … Violent programs have caused the same level of general- some bad influences on many children. ity, where the repeated idea (HKH1 S31c; S39b) and the original idea are comparable Logical Makes a logical deduction In real life, children are exposed to various Conclusion reached by reasoning on kinds of TV programmes not solely to the vio- (LCon) the basis of some previous lent ones. Therefore a generalised statement observation (This can be an is not acceptable. (AH6 S23–S24) interim conclusion.) Chronological Expresses the point in time a) In addition, a study by Singer and Singer Frame when an event takes place. (1981) has indicated that children showed (ChF) This may include: less aggression when parents controlled a) two events over- their television viewing and encouraged lapping, either more imaginative or cognitive play with wholly or partly, in them. time b) one event following b) After examining those three areas of ex- the other in time periments towards the relationship be- tween TV viewing and its influences on children’s aggressive behaviours, it seems that we do not have strong enough evi- dence to support the conclusion that TV viewing has definite influences on chil- dren’s violent behaviour. (HKH2 S28a–b)

107

Rhetorical Definition Example(s) Function Metatextual a) Makes a statement a) In this paper, I will compare these three Statement about the text itself; types of experiments one by one. (MeS) b) The writer comments on his/her own treat- b) and then make a judgement at the end. ment of the content (HKH4 S10a–b) OTHERS *Unidentified This refers to a rhetorical a) The public might think it is the basic fun- Rhetorical unit whose rhetorical damental for children to prepare the Function function cannot be pretection in dealing effectively with the (UNID) worked out from the con- completitiure world. (AL1 S56) text, owing to different kinds of expression prob- b) Freedman’s second criticism of lab. re- lems, such as search is that it may not be possible to ag- a) wrong spelling gressive what happens in the lab. to what b) vocabulary choice or happens in the real world. (AL1 S18) c) confusing reference. c) Milavsky point out that cable-services, video stories growing rapidly, provide op- portunity to obtained easily available for rent or buy. He clarify if T.V. violent affect people how come violent crime rate delined in these areas. (AL1 S37–S38) *Unfulfilled This refers to a rhetorical A person in authority, has chosen to show a Rhetorical Func- unit which signals to the violent programs, in the position is more re- tion (UNFU) reader through contextual sponsible than of industry representative. clues that it serves a cer- (AL1 S77a–b) tain rhetorical function. Yet the rhetorical function S77b fails to serve as the Comparison of signalled has not been S77a because the reader is not told who the fulfilled, owing to different “person in authority” and the “industry kinds of expression prob- representative” are, what position the former lems is in, and why he is more responsible than the “industry representative”.

Key: * indicates that the same Rhetorical Function can be found at both lower and upper levels ( ) the abbreviated term used in the graphic representation of RFA

108 Rhetorical Functions at Paragraph Level

Rhetorical Definition Example Function TRUTH & VALIDITY *Evaluation Presents writer’s own Upon reading all the evidence presented, I (Eval) opinion or judgment on would think the long term effect of violent TV a fact/statement al- programmes has been overstated by some pro- ready introduced, or to fessionals. … I would think the most important be introduced in the thing is for the parents to set up a good model text behaviour for the children. I also think that it is irresponsible to blame everything on television. (AH6 S69–S74, i.e. Para. 13: EVALUATION OF S1–S4) DISCOURSE DEVELOPMENT Background Provides information Television programs play an important role in (Back) which is essential to the the way society perceives their world. Studies adequate comprehen- suggest that programs containing violent sion of a previous sec- themes will often influence a child’s aggressive tion of a text or a forth- behaviours. On the other hand, there are also coming section studies that suggest there is only a tentative relationship between viewing violent television programs and aggressive behaviours. These differing results in various studies have in turn influenced varied opinions amongst researchers of the topic. (AH4 S1–S4) *Metatextual a) Makes a statement a) In this paper, I would like to introduce vari- Statement about the text itself; ous ideas and experiments concerned to (MeS) b) The writer comments our readers. on his/her own b) so we can make a fair judge about them. treatment of the con- (HKH3 S6a–b) tent *Elaboration a) Expands on the spe- And I will contrast both parties’ findings in these (Elab) cific details of an three areas: the laboratory experiments, the idea/object/event al- field experiments and the natural experiments. ready introduced in (HKH2 S6) the text; A natural experiment conducted by Hennigan et b) Makes a general al. in 1982, which was comparing the freeze on proposition more the introduction of television into American cities precise between 1949 to 1952 and the crime rate, showed that there was no evidence that diet of television affected aggressive crimes. But later, in 1986, Williams carried out another experi- ment which examined the relationship between in the introduction of television and observations of children’s aggressive behaviour. … However, no other experiment in this aspect can be found to support this conclusion (HKH2 S23–S27, i.e. Para. 4)

This paragraph is the Elaboration of the natural experiments, introduced in S6.

109

Rhetorical Definition Example Function OTHERS *Unidentified This refers to a rhetori- Singer & Singer point that experimental studies Rhetorical cal unit at Paragraph/ in field settings … have started the negative Function (UNID) Macrostructure Level effects of television viewing. There is no doubt whose rhetorical func- that violent events occur on television pro- tion cannot be worked grammes can produce more extreme perceptual out from the context, or behavioral reactions for children. Although owing to different kinds these studies are difficult to control, overall they of expression problems, show a potential influence. They also indicate such as that prosocial programmes may increase posi- a) faulty sentence tive influences. Because of logistical and ethical structure, considerations, the results of the studies are not b) erroneous choice of often affirmative. (AL4 Paragraph 8: S25–S29) vocabulary or c) inappropriate use of This paragraph was categorised as “UNID” as conjunctions. the reader cannot tell whether it contains mainly Singer & Singer’s Reported Evaluation or the writer’s evaluation. *Unfulfilled This refers to a rhetori- Laboratory experiments are the clearest re- Rhetorical cal unit at Paragraph/ search among all the others. Over 50 well or- Function (UNFU) Macrostructure Level ganized laboratories shows that there is a rela- which signals the tion between aggressive behaviour and TV vio- reader through contex- lence where children do increase aggression tual clues that it serves after watching violent TV programs. This ex- a certain rhetorical periment is supported by many well known ex- function. Yet the rhe- perts included Andison, 1977; Bandura, 1973; torical function sig- Hearold 1979. This experiment also shows the nalled has not been possible outcomes of violent TV viewing in- fulfilled, owing to differ- cluded reduce inhibitation of aggression (Singer ent kinds of expression & Singer 1988). (AL2 Paragraph 4: S10–S13) problems (e.g. faulty sentence structure, er- This paragraph was categorised as “UNFU” for roneous vocabulary 2 reasons: choice or inappropriate a) Wrong use of vocabulary “clearest research” use of conjunctions). in S10 blurs the rhetorical function of the whole paragraph in relation to S7–8 (“Re- search shows that TV violence … increases children aggressive behaviour … .However, some researchers argue … that TV violence would only be a short-term effect … .)

b) Unclear references: "This experiment” blurs the meaning of the Elaboration in S12–S13.

Key: * indicates that the same Rhetorical Function can be found at both lower and upper levels ( ) the abbreviated term used in the graphic representation of RFA

110 Rhetorical Functions at Macrostructure Level

Rhetorical Definition Example Function TRUTH & VALIDITY/ DISCOURSE DEVELOP- MENT *Assertion Introduces a new topic which The effect of violent TV programme on chil- (Ass) is not covered in a previous dren especially aggressive behaviour has part of the text; often used been debated by academics over time. A lot together with Evaluation of experiments have been done by sociolo- gists and psychologists. However, they have different point of views on the matter. Psychologists tend to believe that the effect could be long term whereas sociologists believe that the impact is only short lived and they claim that the effect has been overstated. (AH6 S1–S4) *Evaluation Presents writer’s own opinion Upon reading all the evidence presented, I (Eval) or judgment on a would think the long term effect of violent fact/statement already intro- TV programmes has been overstated by duced, or to be introduced in some professionals. … I would think the the text; often used together most important thing is for the parents to set with Assertion up a good model behaviour for the children. I also think that it is irresponsible to blame everything on television. (AH6 S69–S74, i.e. Para. 13: EVALUATION OF S1–4) Question Presents a question (includ- The question whether TV viewing has a (Ques) ing an embedded question negative effect on children’s behaviour ahs and a rhetorical question); already been discussed for several years. In often used together with An- this paper, it will focus on examining swer whether TV viewing has any effect on chil- dren’s violent behaviour. (HKH2 S3–S4) Answer Provides a reply to a question After examining those three areas of ex- (Ans) already presented in the text; periments towards the relationship between often used together with TV viewing and its influences on children’s Question aggressive behaviours, it seems that we do not have strong enough evidence to support the conclusion that TV viewing has definite influences on children’s aggressive behav- iour. Although the natural experiment in 1986 do support this conclusion, there is no other experiment that can support it strongly, consistently and objectively. Therefore, we conclude that TV viewing do have potential effect on children’s violent behaviour but it is not necessarily intensified their aggression behaviour. (HKH2 S28– S30)

111

Rhetorical Definition Example Function Problem a) Expresses unfavourable Modern technology has, undoubtedly, im- (Pro) conditions or unwelcome proved our living standard and made our situations; life easier. Nevertheless, it has been fre- b) Expresses needs and de- quently argued that the misuse of those sires technologies may bring adverse effects to This rhetorical function is of- us. One topic that aroused much public ten used together with Solu- concern is whether ‘TV viewing will have tion any effects on children’s violent behavior’. (HKH3 S1–S3) Solution Puts forth a means or method Although we cannot come to a definite con- (Sol) to tackle a problem already clusion on whether the statement in the presented in the text; often introduction is appropriate or not, we can used together with Problem generalize from what we have learnt that as children is easily influenced by the things around them, we should take much care of them. (HKH3 S39a–c)

This sentence presents the Solution of the problem described in HKH3 S1–S3 (cf. above). OTHERS *Unidentified This refers to a rhetorical unit Singer & Singer point that experimental Rhetorical at Paragraph/ Macrostructure studies in field settings … have started the Function Level whose rhetorical func- negative effects of television viewing. There (UNID) tion cannot be worked out is no doubt that violent events occur on from the context, owing to television programmes can produce more different kinds of expression extreme perceptual or behavioral reactions problems, such as for children. Although these studies are dif- a) faulty sentence structure, ficult to control, overall they show a poten- b) erroneous choice of vo- tial influence. They also indicate that proso- cabulary or cial programmes may increase positive in- c) inappropriate use of con- fluences. Because of logistical and ethical junctions. considerations, the results of the studies are not often affirmative. (AL4 Paragraph 8: S25–S29)

This paragraph was categorised as “UNID” as the reader cannot tell whether it contains mainly Singer & Singer’s Reported Evalua- tion or the writer’s evaluation.

112

Rhetorical Definition Example Function *Unfulfilled This refers to a rhetorical unit Laboratory experiments are the clearest Rhetorical at Paragraph/Macrostructure research among all the others. Over 50 well Function Level which signals the organized laboratories shows that there is a (UNFU) reader through contextual relation between aggressive behaviour and clues that it serves a certain TV violence where children do increase ag- rhetorical function. Yet the gression after watching violent TV pro- rhetorical function signalled grams. This experiment is supported by has not been fulfilled, owing many well known experts included Andison, to different kinds of expres- 1977; Bandura, 1973; Hearold 1979. This sion problems (e.g. faulty experiment also shows the possible out- sentence structure, erroneous comes of violent TV viewing included re- vocabulary choice or inap- duce inhibitation of aggression (Singer & propriate use of conjunc- Singer 1988). (AL2 Paragraph 4: S10–S13) tions). This paragraph was categorised as “UNFU” for 2 reasons: a) Wrong use of vocabulary “clearest re- search” in S10 blurs the rhetorical func- tion of the whole paragraph in relation to S7-8 (“Research shows that TV violence … increases children aggressive behav- iour … .However, some researchers ar- gue … that TV violence would only be a short-term effect … .)

b) Unclear references: "This experiment” blurs the meaning of the Elaboration in S12–S13.

Key: * indicates that the same Rhetorical Function can be found at both lower and upper levels ( ) the abbreviated term used in the graphic representation of RFA

113 114 Appendix 2: HKH4 – Text and Rhetorical Function Analysis

HKH4 - Text Paragraph 1 1 The question whether TV viewing has a negative effect on children’s behaviour is controversial. 2 There are two points of view. 3 One of these is that television viewing of violent programming leads to aggres- sive behaviour of children. 4 This is extracted from ‘Some Hazards Of Growing Up In A Television Environ- ment’ by Jerome L. Singer & Dorothy G. Singer. 5 And the other is that there is no direct relationship between TV viewing and ag- gressive behaviour. 6 This is extracted from ‘Television Violence And Aggression’ by Jonathan L. Freedman. 7 They are two opposite opinions. 8 However, they all use similar experiments to do the research. 9 They are laboratory experiments, field experiments and natural experiments. 10a In this paper, I will compare these three types of experiments one by one b and make a judgement at the end.

Paragraph 2 11 The first one is laboratory experiment. 12a Singer & Singer (1986) said that many laboratory studies indicate children will imitate or increase aggressive behaviour b after watching violent programming. 13 This shows that there is a causal relationship between heavy television viewing and children’s aggression.

Paragraph 3 14 However, Freedman (1986) has a different opinion. 15 He said different laboratory experiments have different outcomes. 16 Some of them are based on some factors, like anger. 17a The aggression may occur b just because the person is angry. 18 Anger is apparently necessary in some studies and not in others. 19a Therefore, he thought that we can’t make sure the aggressive behaviour of chil- dren is due to anger or actually the heavy TV viewing. b so there is no direct causal link between TV violence and aggressive behaviour.

Paragraph 4 20 The second one is field experiment. 21 It is more natural than laboratory experiment. 22a Singer & Singer (1986) said that field experiments indicate that there is an in- crease in aggressive behaviour b after watching violent TV programming. 23a However, Freedman (1986) said that the field experiments require much time and money,

115 b and the most important of all, the kind of personality that can organize complex arrangements. 24 For instance, in some studies which assuming boys in each cottage are independ- ent of each other. 25a But when boys live together in the cottage, b they will become dependent on each other. 26 Therefore, the analysis cannot be appropriate and reliable.

Paragraph 5 27 The third one is natural experiment. 28 It is bigger in size than field experiment. 29 It takes consideration of many factors, like parents’ viewing, no. of TV sets, pur- chase of cable, as well as children’s sleep patterns or parental use of physical force. 30 From this experiment, direct observations of the aggressive behaviour of children made it clear that such reactions increased after the introduction of television (Singer & Singer (1986)) 31a However, Freedman stated one study which was held in the years 1949 to 1952 (Hennigan et al., 1982), b there was a freeze on the introduction of television into American cities. 32a Hennigan et al. looked at the relationship between crimes rates and the freeze of TV programming, b and stated that if television affected crime, c there would be a increase in crime rates in cities with television compared to those without it. 33a Freedman thinks that what people actually watched is a mixture of television programming, b there is no evidence that diet of television affected aggressive crimes.

Paragraph 6 35 In conclusion, Singer & Singer suggested that heavy television viewing puts chil- dren at risk of increased aggression and restlessness at least in the short run. 35a On the contrary, Freedman suggested that all the experiments have many vari- ables, b so we cannot conclude that there is a direct relationship between TV violence and aggressive behaviour. 36 Above all, I will agree with Freedman. 37 It is because in different experiments, they have different outcomes. 38 Sometimes, the outcomes contradict our assumption. 39 Moreover, increased aggressive behaviour may be due to, let say, insufficient sleep. 40a Therefore, we can’t even convince ourselves b to draw a fair conclusion c and prove that there is a causal link between TV violence and aggressive behav- iour.

116

HKH4 – Rhetorical Function Analysis

Lev el PARAGRAPH 1

Ques M

P Back + MeS

PS Ass MeS

Ass Elab (1) MeS

Elab Elab CompComp Elab (2) ((22-–44)) (2–6) MeS Ass (1)

CS MeS CS

FU/SS Ass Elab Elab Elab Comp Elab Eval Conc Elab MeS Coup (7) (8) (1) (2) (3) (2–4) (5) (2–6) (10a) Sent. 1 2 3 4 5 6 7 8 9 10a 10b

HKH4 – Rhetorical Function Analysis

Level PARAGRAPH 2

M Ans (1)

Elab P (1–10 : Para. 1)

PS Elab RepE (1–10) (11)

RepE CS (11)

Elab RepE ChF LCon FU/SS (1–10) (11) (12a) (12)

Sent. 11 12a 12b 13

117

HKH4 – Rhetorical Function Analysis

Level PARAGRAPH 3

M Ans (1)

P Comp = Elab (12–13) (1–10 : Paras 1)

PS Comp Elab (12–13) (14)

CS Elab LCon (15–17) (16)

Comp RepE Elab Elab Cau RepE LCon LCon FU/SS (12–13) (14) (15) (16) (17a) (17) (15–17) (19a)

Sent. 14 15 16 17a 17b 18 19a 19b

HKH4 – Rhetorical Function Analysis

Level PARAGRAPH 4

M Ans (1)

P Elab (1–10 : Para. 1)

PS Elab RepE Comp (1–10) (20) (22)

Elab RepE Comp Exem LCon (1–10) (20) (22) (23) (23–25)

CS RepE Comp Elab (20) (22) (24)

FU/SS Elab Eval RepE ChF Comp Coup Exem Conc Resu LCon (1–10) (20) (20) (22a) (22) (23a) (23) (24) (25a) (23–25)

Sent. 20 21 22a 22b 23a 23b 24 25a 25b 26

118

HKH4 – Rhetorical Function Analysis

Level PARAGRAPH 5

M Ans (1)

P Elab (1–10 : Para. 1)

PS Elab RepE Comp (1–10) (27–29) (28–30)

Elab Eval RepE Comp (1–10) (27) (27–29) (28–30)

CS Comp Elab RepE (28–30) (31) (32)

FU/SS Elab Eval Elab RepE Comp Elab Elab Cond Resu RepE LCon (1–10) (27) (28) (27–29) (28–30) (31a) (31) (31c) (31b) (32) (31–32)

Sent. 27 28 29 30 31a 31b 32a 32b 32c 33a 33b

HKH4 – Rhetorical Function Analysis

Level PARAGRAPH 6

M Ans (1)

P EvalEval (1(1-35–35: :P Parara.as1 1-–55))

PS SumSum EvalEval (1(1-–33)33) ((1-1–35)35)

CS Comp LCLConon (34) (3(366-–3939))

FU/SS SumSum Just LCon EvalEval Just Elab Coup LConLcon Purp Coup (1(1–35)-35) (35b) (35a) ((11–-3535)) (36) (37) (38) ((3366–-3939)) (40a) (40b)

Sent. 34 35a 35b 36 37 38 39 40a 40b 40c

119 Abbreviations for Rhetorical Functions used in RFA of HKH4:

Ans Answer Eval Evaluation Ass Assertion Exem Exemplification Back Background Just Justification Cau Cause LCon Logical Conclusion ChF Chronological MeS Metatextual Statement Frame Comp Comparison Purp Purpose Conc Concession Ques Question Cond Condition RepE Reported Evaluation Coup Coupling Resu Result Elab Elaboration Sum Summary

120 Appendix 3: Taxonomy of Grammatical Accuracy Analysis

Category Definition Example ① Derivational Wrong word form, which a) This is a violence program. Morphology changes the word class b) “I work hardly.” For “I work hard.” ② Spelling Wrong spelling … the dipiction of violence in television programs (AL4 S3) ③ Tenses Mistakes in tense form Thus heavy television viewing especially viewing violent programs is positively cor- related with concurrent and subsequent aggressive behavior. Consequently, I be- lieved heavy exposure to the mass media would have effects on children. (AL5 S24–S25) ④ Number Mistakes in a) There are three sources of informa- Agreement a) Singular/plural form tion … : experimental studies in labora- b) Subject-verb agreement tory, field experiment, and correlational studies … . (HKL5 S8) b) … to keep an eyes on the production of television programmes (AL1 S15c) ⑤ Nonfinite Mistakes in forms of a) There is no doubt that violent event Forms a) Gerunds occur in television programmes can pro- b) Infinitives duce more extreme perceptual or behav- c) Participles ioral reactions for children. (AL4 S26) b) … the boys in the studies were initially aggressive, although watching violent program may gave them new idea … . (AL6 S40) c) … if the aggressive behavior is regard as the effect of TV violence, how do we … measure aggression? (HKL2 S16) ⑥ Voice Mistakes in active/passive It found that T.V. had no effect on violent voice crime or other serious crimes, but only on petty theft. (AL1 S29) ⑦ Wrong Word Wrong choice of content or a) … children confronted too much time function word, i.e. a para- watching T.V. (AL1 S4) digmatic error, which con- b) … it is more supportive that violent fuses the meaning, but the television really cause increase in ag- reader can still understand gression of children. (HKL4 S32) what the writer is trying to say. ⑧ Missing This includes any obligatory The social scientists ^ that T.V. viewing, Word word (except articles, collo- especially of more violent content may cation and conjunctive de- influence children overt behavior. (AL1 vices) which is missing. S49) ⑨ Redundant A word which is redundant, Since the relationships between violent Word or unnecessary, in the con- TV programs and aggressive behaviour text were examined in many different ways, therefore measurement errors might oc- cur. (AL2 S3)

121

Category Definition Example ⑩ Collocation Wrong choice of, redun- The third town had no access to televi- dant, or missing, word, i.e. sion of all until 1973. (AL3 S46) a syntagmatic error, which is against conventional id- iom or usage ⑪ Articles Wrong choice of, redun- In the real world, there are many condi- dant, or missing, articles tions that are out of ^ researcher’s con- trol ... (HKL2 S17a) ⑫ Conjunctive Wrong choice of, redun- From the past researches, long time in Devices dant, or missing, linking TV viewing can cause shorts sighted, but words (except articles) people concerns much about it ... (HKL2 S2) ⑬ Anaphoric/ Wrong choice of pronouns Even though there can carry out a field Cataphoric to refer to a person/object experiment, many technical problems Pronouns already/to be mentioned in should be solved. It should contact with a the text large of ‘sample’…(AKL2 S18–S19) ⑭ Sequencing The sequencing of (part of) Singer & Singer point out theoretically a phrase/clause is not cor- that television viewing of violent pro- rect. grams definitely increases children’s ag- gression. (AL4 S11) ⑮ Word Wrong punctuation within a … in todays television programs … (AL5 Punctuation word, e.g. apostrophe S3) ⑯ Internal Wrong or missing punctua- … the introduction of new media, such Punctuation tion within a phrase/clause, as cable T.V. makes viewing television a within & e.g. commas, semicolons, more common and more frequent enter- between quotation marks tainment (AL6 S2) Phrases or Clauses ⑰ Sentence-final Wrong punctuation at the Needless to say, every such scheme has Punctuation end of a sentence its advantages and disadvantages, The point is that it is essential for the gov- ernment and authorities concerned to keep an eye on the production of televi- sion programs. (HKL1 S15) ⑱ Tangled The phrase/clause is tan- The one talked by Jonathan L. Freed- Construction gled, which makes it im- man is not done by himself but the quot- possible to count discrete ing of the experiments done by Henni- mistakes. Yet despite the gan et al in 1982. (HKL4 S25) entanglement, the reader can still understand what the writer is trying to say. ⑲ Total Commu- The whole phrase/clause is The opposite of Jonathan L. Freedman nication tangled, which makes it are Jerome L. Singer and Dorothy G. Breakdown impossible to count discrete Singer who said that children’s aggres- mistakes, nor can the sive behavior permit more difinitive reader understand what the causal. (HKL4 S14) writer is trying to say.

122 Appendix 4: Frequency of rhetorical functions identified at the lower and upper levels of Australian/Hong Kong high-/low-rated essays

Table A4.1a: Frequency of rhetorical functions at the lower levels in Australian high-rated essays. Rhetorical Function FU/SS Level CS Level PS Level Total frequency TRUTH & VALIDITY 11 5 8 24 Assertion (1.68%) (2.83%) (5.20%) (2.43%) 19 1 4 24 Justification (2.90%) (0.57%) (2.60%) (2.43%) 53 8 11 72 Concession (8.09%) (4.52%) (7.14%) (7.30%) 44 21 19 84 Evaluation (6.72%) (11.86%) (12.34%) (8.52%) Reported 27 9 14 50 Evaluation (4.12%) (5.09%) (9.09%) (5.07%) CAUSE & EFFECT 15 3 1 19 Cause (2.29%) (1.70%) (0.65%) (1.93%) 87 26 16 129 Result (13.28%) (14.69%) (10.39%) (13.08%) 23 23 Condition 0 0 (3.51%) (2.33%) 32 1 33 Purpose 0 (4.89%) (0.57%) (3.35%) DISCOURSE DEVELOPMENT 130 55 45 230 Elaboration (19.85%) (31.07%) (29.22%) (23.33%) 15 7 2 24 Exemplification (2.29%) (3.96%) (1.30%) (24.34%) 44 13 13 70 Comparison (6.72%) (7.35%) (8.44%) (70.99%) 73 15 15 103 Coupling (11.15%) (8.48%) (9.74%) (10.45%) 10 10 Alternation 0 0 (1.53%) (1.01%) Summary 0 0 0 0 9 3 3 15 Restatement (1.37%) (1.70%) (1.95%) (1.52%) 32 10 3 45 Logical Conclusion (4.89%) (5.65%) (1.95%) (4.56%) Chronological 27 27 0 0 Frame (4.12%) (2.74%) Metatextual 4 4 0 0 Statement (0.61%) (0.41%) OTHERS Unidentified 0 0 0 0 Rhetorical Function Unfulfilled 0 0 0 0 Rhetorical Function Total frequency 655 177 154 986 at each level (100%) (100%) (100%) (100%)

123 Table A4.1b: Frequency of rhetorical functions at Paragraph Level in Australian high-rated essays. Rhetorical Function Paragraph Level TRUTH & VALIDITY Evaluation 19 (25.68%) DISCOURSE DEVELOPMENT Background 5 (6.76%) Metatextual Statement 4 (5.41%) Elaboration 46 (62.16%) OTHERS Unidentified 0 Rhetorical Function Unfulfilled Rhetorical Function 0 Total frequency at each level 74 (100%)

Table A4.1c: Frequency of rhetorical functions at Macrostructure Level in Australian high- rated essays. Rhetorical Function Macrostructure Level TRUTH & VALIDITY/ DISCOURSE DEVELOPMENT Assertion 6 (8.11%) Evaluation 68 (91.89%) Question 0 Answer 0 Problem 0 Solution 0 OTHERS Unidentified 0 Rhetorical Function Unfulfilled 0 Rhetorical Function Total frequency at each level 74 (100%)

Key: The numbers in parentheses are the percentages of each rhetorical function at the respective level.

124 Table A4.2a: Frequency of rhetorical functions at the lower levels in Australian low-rated essays. Rhetorical Function FU/SS Level CS Level PS Level Total frequency TRUTH & VALIDITY 15 8 6 29 Assertion (2.58%) (5.06%) (3.75%) (3.22%) 8 2 2 12 Justification (1.38%) (1.27%) (1.25%) (1.33%) 26 3 1 30 Concession (4.47%) (1.90%) (0.63%) (3.33%) 12 6 3 21 Evaluation (2.06%) (3.80%) (1.88%) (2.33%) Reported 25 7 9 41 Evaluation (4.30%) (4.43%) (5.63%) (4.56%) CAUSE & EFFECT 20 2 22 Cause 0 (3.44%) (1.27%) (2.44%) 44 13 10 67 Result (7.56%) (8.23%) (6.25%) (7.44%) 6 6 Condition 0 0 (1.03%) (0.67%) 22 22 Purpose 0 0 (3.78%) (2.44%) DISCOURSE DEVELOPMENT 131 55 28 214 Elaboration (22.51%) (34.81%) (17.50%) (23.78%) 9 2 3 14 Exemplification (1.55%) (1.27%) (1.88%) (1.56%) 31 6 5 42 Comparison (5.33%) (3.80%) (3.13%) (4.67%) 46 7 4 57 Coupling (7.90%) (4.43%) (2.50%) (6.33%) 14 14 Alternation 0 0 (2.41%) (1.56%) 3 1 4 Summary 0 (0.52%) (0.63%) (0.44%) 5 5 Restatement 0 0 (0.86%) (0.56%) 19 8 1 28 Logical Conclusion (3.27%) (5.06%) (0.63%) (3.11%) Chronological 12 12 0 0 Frame (2.06%) (1.33%) Metatextual 0 0 0 0 Statement OTHERS Unidentified 44 12 33 89 Rhetorical Function (7.56%) (7.60%) (20.63%) (9.89%) Unfulfilled 90 27 54 171 Rhetorical Function (15.46%) (17.09%) (33.75%) (19%) Total frequency 582 158 160 900 at each level (100%) (100%) (100%) (100%)

125 Table A4.2b: Frequency of rhetorical functions at Paragraph Level in Australian low-rated essays. Rhetorical Function Paragraph Level TRUTH & VALIDITY Evaluation 2 (2.04%) DISCOURSE DEVELOPMENT Background 8 (8.16%) Metatextual Statement 4 (4.08%) Elaboration 36 (36.73%) OTHERS Unidentified 15 (15.31%) Rhetorical Function Unfulfilled 33 (33.67%) Rhetorical Function Total frequency at each level 98 (100%)

Table A4.2c: Frequency of rhetorical functions at Macrostructure Level in Australian low- rated essays.

Rhetorical Function Macrostructure Level TRUTH & VALIDITY/ DISCOURSE DEVELOPMENT *Assertion 5 (5%) *Evaluation 12 (12%) Question 7 (7%) Answer 27 (27%) Problem 0 Solution 1 (1%) OTHERS *Unidentified 15 (15%) Rhetorical Function *Unfulfilled 33 (33%) Rhetorical Function Total frequency at each level 100 (100%)

Key: The numbers in parentheses are the percentages of each rhetorical function at the respective level.

126 Table A4.3a: Frequency of rhetorical functions at the lower levels in Hong Kong high-rated essays. Rhetorical Function FU/SS Level CS Level PS Level Total frequency TRUTH & VALIDITY 3 1 4 8 Assertion (0.99%) (1.39%) (5.48%) (1.79%) 7 4 2 13 Justification (2.32%) (5.56%) (2.74%) (2.91%) 22 1 1 24 Concession (7.29%) (1.39%) (1.37%) (5.37%) 18 7 7 32 Evaluation (5.96%) (9.72%) (9.59%) (7.16%) Reported 27 9 7 43 Evaluation (8.94%) (12.50%) (9.59%) (9.62%) CAUSE & EFFECT 10 1 11 Cause 0 (3.31%) (1.39%) (2.46%) 31 6 4 41 Result (10.26%) (8.33%) (5.48%) (9.17%) 9 9 Condition 0 0 (2.98%) (2.01%) 12 12 Purpose 0 0 (3.97%) (2.69%) DISCOURSE DEVELOPMENT 71 23 33 127 Elaboration (23.51%) (31.94%) (45.21%) (28.41%) 2 1 1 4 Exemplification (0.66%) (1.39%) (1.37%) (0.89%) 31 10 8 49 Comparison (10.26%) (13.89%) (10.96%) (10.96%) 25 4 29 Coupling 0 (8.28%) (5.48%) (6.49%) Alternation 0 0 0 0 2 1 3 Summary 0 (0.66%) (1.37%) (0.67%) 1 1 2 Restatement 0 (0.33%) (1.37%) (0.45%) 16 8 24 Logical Conclusion 0 (5.30%) (11.11%) (5.37%) Chronological 12 12 0 0 Frame (3.97%) (2.69%) Metatextual 3 1 4 0 Statement (0.99%) (1.39%) (0.89%) OTHERS Unidentified 0 0 0 0 Rhetorical Function Unfulfilled 0 0 0 0 Rhetorical Function Total frequency 302 72 73 447 at each Level (100%) (100%) (100%) (100%)

127 Table A4.3b: Frequency of rhetorical functions at Paragraph Level in Hong Kong high-rated essays. Rhetorical Function P Level TRUTH & VALIDITY *Evaluation 7 (14.89%) DISCOURSE DEVELOPMENT Background 5 (10.64%) *Metatextual Statement 6 (12.77%) *Elaboration 29 (61.70%) OTHERS *Unidentified 0 Rhetorical Function *Unfulfilled 0 Rhetorical Function Total frequency at each level 47 (100%)

Table A4.3c: Frequency of rhetorical functions at Macrostructure Level in Hong Kong high- rated essays. Rhetorical Function Macrostructure Level TRUTH & VALIDITY/ DISCOURSE DEVELOPMENT Assertion 0 Evaluation 0 Question 6 (14.63%) Answer 33 (80.49%) Problem 0 Solution 2 (4.88%) OTHERS Unidentified 0 Rhetorical Function Unfulfilled 0 Rhetorical Function Total frequency at each level 41 (100%)

Key: The numbers in parentheses are the percentages of each rhetorical function at the respective level.

128 Table A4.4a: Frequency of rhetorical functions at the lower levels in Hong Kong low-rated essays. Rhetorical Function FU/SS Level CS Level PS Level Total frequency TRUTH & VALIDITY 9 3 5 17 Assertion (3.69%) (4.41%) (9.80%) (4.68%) 3 1 3 7 Justification (1.23%) (1.47%) (5.88%) (1.93%) 12 3 15 Concession 0 (4.92%) (4.41%) (4.13%) 6 1 1 8 Evaluation (2.46%) (1.47%) (1.96%) (2.20%) Reported 11 2 3 16 Evaluation (4.51%) (2.94%) (5.88%) (4.41%) CAUSE & EFFECT 3 3 Cause 0 0 (1.23%) (0.83%) 17 4 21 Result 0 (6.97%) (5.88%) (5.79%) 10 10 Condition 0 0 (4.10%) (2.76%) 12 12 Purpose 0 0 (4.92%) (3.31%) DISCOURSE DEVELOPMENT 48 17 7 72 Elaboration (19.67%) (25%) (13.73%) (19.83%) 4 2 6 Exemplification 0 (1.64%) (2.94%) (1.65%) 14 3 2 19 Comparison (5.74%) (4.41%) (3.92%) (5.23%) 16 3 1 20 Coupling (6.56%) (4.41%) (1.96%) (5.51%) 1 1 Alternation 0 0 (0.41%) (0.28%) 1 1 2 Summary 0 (0.41%) (1.96%) (0.55%) 3 1 2 6 Restatement (1.23%) (1.47%) (3.92%) (1.65%) 3 3 Logical Conclusion 0 0 (1.23%) (0.83%) Chronological 4 4 0 0 Frame (1.64%) (1.10%) Metatextual 1 2 3 0 Statement (0.41%) (3.92%) (0.83%) OTHERS Unidentified 18 9 5 32 Rhetorical Function (7.38%) (13.24%) (9.80%) (8.82%) Unfulfilled 48 19 19 86 Rhetorical Function (19.67%) (27.94%) (37.25%) (23.69%) Total frequency 244 68 51 363 at each level (100%) (100%) (100%) (100%)

129 Table A4.4b: Frequency of rhetorical functions at Paragraph Level in Hong Kong low-rated essays. Rhetorical Function Paragraph Level TRUTH & VALIDITY Evaluation 2 (3.85%) DISCOURSE DEVELOPMENT Background 5 (9.62%) Metatextual Statement 6 (11.54%) Elaboration 12 (23.08%) OTHERS Unidentified 5 (9.62%) Rhetorical Function Unfulfilled 22 (42.31%) Rhetorical Function Total frequency at each level 52 (100%)

Table A4.4c: Frequency of rhetorical functions at Macrostructure Level in Hong Kong low- rated essays. Rhetorical Function Macrostructure Level TRUTH & VALIDITY/ DISCOURSE DEVELOPMENT *Assertion 0 *Evaluation 1 (2.5%) Question 5 (12.5%) Answer 6 (15%) Problem 0 Solution 2 (5%) OTHERS *Unidentified 3 (7.5%) Rhetorical Function *Unfulfilled 23 (57.5%) Rhetorical Function Total frequency at each level 40 (100%)

Key: The numbers in parentheses are the percentages of each rhetorical function at the respective level.

130 Appendix 5: Frequency of grammatical errors identified in Australian/Hong Kong high-/low-rated essays

Table A5.1: Frequency of grammatical errors identified in Australian high-rated essays. Essay Error Total: category AH1 AH2 AH3 AH4 AH5 AH6 Each category Derivational 2 5 2 2 1 12 0 Morphology (5.56%) (14.29%) (5.13%) (8.33%) (3.13%) (6.03%) 1 4 1 5 2 13 Spelling 0 (3.03%) (11.11%) (2.86%) (12.82%) (6.25%) (6.53%) 2 2 Tenses 0 0 0 0 0 (6.06%) (1.01%) Number 4 7 5 8 3 9 36 Agreement (12.12%) (19.44%) (14.29%) (20.52%) (12.50%) (28.13%) (18.09%) Nonfinite 1 4 2 7 0 0 0 Forms (3.03%) (11.11%) (5.13%) (3.52%) 1 1 Voice 0 0 0 0 0 (3.13%) (0.50%) 3 1 1 2 7 Wrong Word 0 0 (9.09%) (2.78%) (2.56%) (6.25%) (3.52%) 1 3 1 1 2 8 Missing Word 0 (2.78%) (8.57%) (2.56%) (4.17%) (6.25%) (4.02%) Redundant 4 3 1 2 10 0 0 Word (12.12%) (8.33%) (2.56%) (8.33%) (5.03%) 2 4 4 2 3 2 17 Collocation (6.06%) (11.11%) (11.43%) (5.13%) (12.50%) (6.25%) (8.54%) 3 1 10 3 4 2 23 Articles (9.09%) (2.78%) (28.56%) (7.69%) (16.67%) (6.25%) (11.56%) Conjunctive 3 1 2 3 9 0 0 Devices (9.09%) (2.86%) (8.33%) (9.38%) (4.52%) Anaphoric/ Cataphoric 0 0 0 0 0 0 0 Pronouns 1 2 1 2 6 Sequencing 0 0 (3.03%) (5.56%) (2.86%) (8.33%) (3.02%) Word 1 3 1 5 0 0 0 Punctuation (2.86%) (7.69%) (3.13%) (2.51%) Internal 8 7 4 10 5 7 41 Punctuation (24.24%) (19.44%) (11.43%) (25.64%) (20.83%) (21.88%) (20.60%) Sentence-final 1 1 2 0 0 0 0 Punctuation (9.09%) (2.56%) (1.01%) Tangled 0 0 0 0 0 0 0 Construction Total Communication 0 0 0 0 0 0 0 Breakdown Total: 33 36 35 39 24 32 199 Each essay (100%) (100%) (100%) (100%) (100%) (100%) (100%)

131 Table A5.2: Frequency of grammatical errors identified in Australian low-rated essays. Essay Error Total: category AL1 AL2 AL3 AL4 AL5 AL6 Each category Derivational 17 4 1 15 37 0 0 Morphology (6.85%) (3.33%) (1.89%) (13.27%) (5.83%) 5 4 2 3 4 7 25 Spelling (2.02%) (3.33%) (3.77%) (6.25) (7.55%) (6.20%) (3.94%) 6 2 5 1 2 2 18 Tenses (2.42%) (1.67%) (9.43%) (2.08%) (3.77%) (1.77%) (2.83%) Number 37 24 11 7 11 30 120 Agreement (14.92%) (20%) (20.75%) (14.58%) (20.75%) (26.55%) (18.90%) Nonfinite 14 5 2 2 3 13 39 Forms (5.65%) (4.17%) (3.77%) (4.17%) (5.66%) (11.50%) (6.14%) 3 1 2 6 Voice 0 0 0 (1.21%) (1.89%) (1.77%) (0.94%) 15 10 4 2 2 33 Wrong Word 0 (6.05%) (8.33%) (8.33%) (3.77%) (1.77%) (5.20%) 27 8 4 5 7 0 51 Missing Word (10.89%) (6.67%) (7.55%) (10.42%) (13.21%) (0%) (8.03%) Redundant 27 10 2 4 2 45 0 Word (10.89%) (8.33%) (3.77%) (8.33%) (1.77%) (7.09%) 15 4 2 2 3 1 27 Collocation (6.05%) (3.33%) (3.77%) (4.17%) (5.66%) (0.89%) (4.25%) 6 7 7 2 5 12 39 Articles (2.42%) (5.83%) (13.21%) (4.17%) (9.43%) (10.62%) (6.14%) Conjunctive 2 2 3 1 3 11 0 Devices (0.81%) (1.67%) (6.25%) (1.89%) (2.66%) (1.73%) Anaphoric/ 15 13 3 10 1 42 Cataphoric 0 (6.05%) (10.83%) (5.66%) (20.83%) (1.89%) (6.61%) Pronouns 2 2 2 1 7 Sequencing 0 0 (0.81%) (1.67%) (4.17%) (0.89%) (1.10%) Word 7 1 1 1 1 1 12 Punctuation (2.82%) (0.83%) (1.89%) (2.08%) (1.89%) (0.89%) (1.89%) Internal 21 14 12 2 10 12 71 Punctuation (8.47%) (11.67%) (22.64%) (4.17%) (18.87%) (10.62%) (11.18%) Sentence-final 6 6 1 7 20 0 0 Punctuation (2.42%) (5%) (1.89%) (6.20%) (3.15%) Tangled 8 3 1 12 0 0 0 Construction (3.23%) (2.50%) (0.89%) (1.89%) Total 15 1 2 2 20 Communication 0 0 (6.05%) (0.83%) (3.78%) (1.78%) (3.15%) Breakdown Total: 248 120 53 48 53 113 635 Each essay (100%) (100%) (100%) (100%) (100%) (100%) (100%)

132 Table A5.3: Frequency of grammatical errors identified in Hong Kong high-rated essays. Essay Error Total: category HKH1 HKH2 HKH3 HKH4 HKH5 HKH6 Each category Derivational 2 2 2 6 0 0 0 Morphology (5.71%) (10%) (8.33%) (4.65%) 2 2 1 2 2 9 Spelling 0 (5.71%) (14.29%) (5%) (8.33%) (11.76%) (6.98%) 2 1 1 1 1 6 Tenses 0 (5.71%) (7.14%) (5.26%) (4.17%) (5.88%) (4.65%) Number 11 6 6 1 9 2 35 Agreement (3.14%) (42.86%) (30%) (5.26%) (37.50) (11.76%) (27.13%) Nonfinite 2 1 1 4 0 0 0 Forms (5.71%) (5.26%) (4.17%) (3.10%) 1 1 2 Voice 0 0 0 0 (2.86%) (7.14%) (1.55%) 1 2 1 3 7 Wrong Word 0 0 (2.86%) (10%) (5.26%) (17.65%) (5.43%) 3 1 2 2 1 9 Missing Word 0 (8.57%) (7.14%) (10.53%) (8.33%) (5.88%) (6.98%) 3 1 2 1 7 Redundant Word 0 0 (8.57%) (7.14%) (10.53%) (5.88%) (5.43%) 4 1 2 4 3 1 15 Collocation (11.43%) (7.14%) (10%) (21.05%) (12.50%) (5.88%) (11.63%) 1 4 2 4 11 Articles 0 0 (7.14%) (20%) (10.53%) (23.53%) (8.53%) Conjunctive 2 2 0 0 0 0 0 Devices (10.53%) (1.55%) Anaphoric/ Cataphoric 0 0 0 0 0 0 0 Pronouns Sequencing 0 0 0 0 0 0 0 Word 0 0 0 0 0 0 0 Punctuation Internal 3 1 2 4 2 12 0 Punctuation (8.57%) (5%) (10.53%) (16.67%) (11.76%) (9.30%) Sentence-final 1 2 1 4 0 0 0 Punctuation (5.71%) (10%) (5.26%) (3.10%) Tangled 0 0 0 0 0 0 0 Construction Total Communication 0 0 0 0 0 0 0 Breakdown Total: 35 14 20 19 24 17 129 Each essay (100%) (100%) (100%) (100%) (100%) (100%) (100%)

133 Table A5.4: Frequency of grammatical errors identified in Hong Kong low-rated essays. Essay Error Total: category HKL1 HKL2 HKL3 HKL4 HKL5 HKL6 Each category Derivational 2 2 4 0 0 0 0 Morphology (1.98%) (2.22%) (1.27%) 2 3 1 6 Spelling 0 0 0 (5.56%) (2.97%) (2.70%) (1.90%) 1 3 6 1 5 16 Tenses 0 (2.78%) (2.97%) (6.67%) (2.78%) (13.51%) (5.06%) Number 3 18 8 18 6 7 60 Agreement (8.33%) (17.82%) (50%) (20%) (16.67%) (18.92%) (18.99%) 1 7 4 1 13 Nonfinite Forms 0 0 (2.78%) (6.93%) (4.44%) (2.70%) (4.11%) 4 4 1 9 Voice 0 0 0 (3.96%) (11.11%) (2.70%) (2.85%) 1 4 11 2 1 19 Wrong Word 0 (2.78%) (3.96%) (12.22%) (5.56%) (2.70%) (6.01%) 3 7 4 4 18 Missing Word 0 0 (2.97%) (7.78%) (11.11%) (10.81%) (5.70%) Redundant 5 18 16 3 3 45 0 Word (13.89%) (17.82%) (17.78%) (8.33%) (8.11%) (14.24%) 6 4 3 8 2 2 25 Collocation (16.67%) (3.96%) (18.75%) (8.89%) (5.56%) (5.41%) (7.91%) 4 5 4 2 5 20 Articles 0 (11.11%) (4.95%) (4.44%) (5.56%) (13.51%) (6.33%) Conjunctive 1 3 1 0 3 0 8 Devices (2.78%) (2.97%) (6.25%) (0%) (8.33%) (0%) (2.53%) Anaphoric/ 1 12 1 2 1 0 17 Cataphoric (2.78%) (11.88%) (6.25%) (2.22%) (2.78%) (0%) (5.38%) Pronouns 2 2 4 Sequencing 0 0 0 0 (1.98%) (5.41%) (1.27%) Word 2 1 3 0 0 0 0 Punctuation (5.56%) (0.99%) (0.95%) Internal 5 3 1 9 4 1 23 Punctuation (13.89%) (2.97%) (6.25%) (10%) (11.11%) (2.70%) (7.28%) Sentence-final 2 5 1 0 2 2 12 Punctuation (5.56%) (4.95%) (6.25%) (0%) (5.56%) (5.41%) (3.80%) Tangled 1 1 2 1 5 0 0 Construction (2.78%) (0.99%) (2.22%) (2.78%) (1.58%) Total 1 3 1 1 1 2 9 Communication (2.78%) (2.97%) (6.25%) (1.11%) (2.78%) (5.41%) (2.85%) Breakdown Total: 36 101 16 90 36 37 316 Each essay (100%) (100%) (100%) (100%) (100%) (100%) (100%)

134 Appendix 6: Comparison of frequencies and relative rankings of error types: High-rated essays versus low- rated essays/Australian essays versus Hong Kong essays

Table A6.1: Comparison of frequencies, percentages and relative rankings of error types: high-rated essays versus low-rated essays. Essay groups High-rated essays Low-rated essays Error types (AH+HKH) (AL+HKL) Frequency and Relative Frequency and Relative percentage ranking percentage ranking Derivational Morphology 18 (5.49%) 6th 41 (4.31%) 10th Spelling 22 (6.71%) 5th 31 (3.26%) 13th Tenses 8 (2.44%) 12th 34 (3.58%) 11th Number Agreement 71 (21.65%) 1st 180 (18.93%) 1st Nonfinite Forms 11 (3.35%) 10th 52 (5.47%) 7th Voice 3 (0.91%) 16th 15 (1.58%) 17th Wrong Word 14 (4.27%) 9th 52 (5.47%) 7th Missing Word 17 (5.18%) 7th 69 (7.26%) 4th Redundant Word 17 (5.18%) 7th 90 (9.46%) 3rd Collocation 32 (9.76%) 4th 52 (5.47%) 7th Articles 34 (10.37%) 3rd 59 (6.20%) 5th Conjunctive Devices 11 (3.35%) 10th 19 (2.0%) 15th Anaphoric/Cataphoric 0 17th 59 (6.20%) 5th Pronouns Sequencing 6 (1.83%) 13th 11 (1.16%) 19th Word Punctuation 5 (1.52%) 15th 15 (1.58%) 17th Internal Punctuation 53 (16.16%) 2nd 94 (9.88%) 2nd Sentence-final Punctuation 6 (1.83%) 13th 32 (3.36%) 12th Tangled Construction 0 17th 17 (1.79%) 16th Total Communication 0 17th 29 (3.05%) 14th Breakdown Totals 328 (100%) 951 (100%)

Key: The numbers in parentheses are the percentage of each error type in the respective grouping.

135 Table A6.2: Comparison of frequencies, percentages and relative rankings of error types: Australian essays versus Hong Kong essays. Essay groups Australian L2 Hong Kong Error types (AH+AL) (HKH+HKL) Frequency and Relative Frequency and Relative percentage ranking percentage ranking Derivational Morphology 49 (5.88%) 6th 10 (2.25%) 14th Spelling 38 (4.56%) 11th 15 (3.37%) 12th Tenses 20 (2.40%) 13th 22 (4.94%) 8th Number Agreement 156 (18.71%) 1st 95 (21.35%) 1st Nonfinite Forms 46 (5.52%) 7th 17 (3.82%) 9th Voice 7 (0.84%) 19th 11 (2.47%) 13th Wrong Word 40 (4.80%) 10th 26 (5.84%) 7th Missing Word 59 (7.07%) 4th 27 (6.07%) 6th Redundant Word 55 (6.59%) 5th 52 (11.69%) 2nd Collocation 44 (5.28%) 8th 40 (8.99%) 3rd Articles 62 (7.43%) 3rd 31 (6.97%) 5th Conjunctive Devices 20 (2.40%) 13th 10 (2.25%) 14th Anaphoric/Cataphoric 42 (5.04%) 9th 17 (3.82%) 9th Pronouns Sequencing 13 (1.56%) 17th 4 (0.90%) 18th Word Punctuation 17 (2.04%) 16th 3 (0.67%) 19th Internal Punctuation 112 (13.43%) 2nd 35 (7.87%) 4th Sentence-final Punctuation 22 (2.64%) 12th 16 (3.60%) 11th Tangled Construction 12 (1.44%) 18th 5 (1.12%) 17th Total Communication 20 (2.40%) 13th 9 (2.02%) 16th Breakdown Totals 834 (100%) 445 (100%)

Key: The numbers in parentheses are the percentage of each error type in the respective grouping.

136