Academic Vocabulary in Learner Writing

Corpus and DiscourseSeries editors: Wolfgang Teubert, University of Birmingham, and Michaela Mahlberg, University of Liverpool.

Editorial Board: Paul Baker (Lancaster), Frantisek Cermák (Prague), Susan Conrad (Portland), Geoffrey Leech (Lancaster), Dominique Maingueneau (Paris XII), Christian Mair (Freiburg), Alan Partington (Bologna), Elena Tognini- Bonelli (Siena and TWC), Ruth Wodak (Lancaster), Feng Zhiwei (Beijing).

Corpus linguistics provides the methodology to extract meaning from texts. Taking as its starting point the fact that language is not a mirror of reality but lets us share what we know, believe and think about reality, it focuses on language as a social phenomenon, and makes visible the attitudes and beliefs expressed by the members of a discourse community.

Consisting of both spoken and written language, discourse always has historical, social, functional, and regional dimensions. Discourse can be monolingual or multilingual, interconnected by translations. Discourse is where language and social studies meet.

The Corpus and Discourse series consists of two strands. The fi rst, Research in Corpus and Discourse, features innovative contributions to various aspects of corpus linguistics and a wide range of applications, from language technology via the teaching of a second language to a history of mentalities. The second strand, Studies in Corpus and Discourse, is comprised of key texts bridging the gap between social studies and linguistics. Although equally academically rigorous, this strand will be aimed at a wider audience of academics and postgraduate students working in both disciplines.

Research in Corpus and Discourse

Conversation in ContextA Corpus-driven ApproachWith a preface by Michael McCarthyChristoph Rühlemann

Corpus-Based Approaches to English Language TeachingEdited by Mari Carmen Campoy, Begona Bellés-Fortuno and Ma Lluïsa Gea-Valor

Corpus Linguistics and World EnglishesAn Analysis of Xhosa EnglishVivian de Klerk

Evaluation and Stance in War NewsA Linguistic Analysis of American, British and Italian television news reporting of the 2003 Iraqi warEdited by Louann Haarman and Linda Lombardo

Evaluation in Media DiscourseAnalysis of a Newspaper CorpusMonika Bednarek

Historical Corpus StylisticsMedia, Technology and ChangePatrick Studer

Idioms and CollocationsCorpus-based Linguistic and Lexicographic StudiesEdited by Christiane Fellbaum

Meaningful TextsThe Extraction of Semantic Information from Monolingual and Multilingual CorporaEdited by Geoff Barnbrook, Pernilla Danielsson and Michaela Mahlberg

Rethinking IdiomaticityA Usage-based ApproachStefanie Wulff

Working with Spanish CorporaEdited by Giovanni Parodi

Studies in Corpus and Discourse

Corpus Linguistics and The Study of LiteratureStylistics In Jane Austen’s NovelsBettina Starcke

English Collocation StudiesThe OSTI ReportJohn Sinclair, Susan Jones and Robert DaleyEdited by Ramesh KrishnamurthyWith an introduction by Wolfgang Teubert

Text, Discourse, and Corpora. Theory and Analysis Michael Hoey, Michaela Mahlberg, Michael Stubbs and Wolfgang TeubertWith an introduction by John Sinclair

This page intentionally left blank

Academic Vocabulary in Learner Writing

From Extraction to Analysis

Magali Paquot

Continuum International Publishing Group

The Tower Building 80 Maiden Lane11 York Road Suite 704, New YorkLondon SE1 7NX NY 10038

www.continuumbooks.com

© Magali Paquot 2010

All rights reserved. No part of this publication may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopying, recording, or any information storage or retrieval system, without prior permission in writing from the publishers.

British Library Cataloguing-in-Publication Data

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

ISBN: 978-1-4411-3036-5 (hardcover)

Library of Congress Cataloging-in-Publication Data

A catalog record for this book is available from the Library of Congress.

Typeset by Newgen Imaging Systems Pvt Ltd, Chennai, IndiaPrinted and bound in Great Britain by the MPG Books Group

Contents

Acknowledgements xiList of abbreviations xiiiList of fi gures xvList of tables xvii

Introduction 1

Part I: Academic vocabulary

Chapter 1 What is academic vocabulary? 91.1. Academic vocabulary vs. core vocabulary and technical terms 9

1.1.1. Core vocabulary 101.1.2. Academic vocabulary 111.1.3. Technical terms 131.1.4. Fuzzy vocabulary categories 13

1.2. Academic vocabulary and sub-technical vocabulary 171.3. Vocabulary and the organization of academic texts 221.4. Is there an ‘academic vocabulary’? 251.5. Summary and conclusion 27

Chapter 2 A data-driven approach to the selection of academic vocabulary 29

2.1. Corpora of academic writing 312.2. Corpus annotation 34

2.2.1. Issues in annotating corpora 342.2.2. The software 36

2.3. Automatic extraction of potential academic words 442.3.1. Keyness 462.3.2. Range 482.3.3. Evenness of distribution 502.3.4. Broadening the scope of well-represented semantic categories 53

2.4. The Academic Keyword List 552.5. Summary and conclusion 61

viii Contents

Part II: Learners’ use of academic vocabulary

Chapter 3 Investigating learner language 673.1. The International Corpus of Learner English 673.2. Contrastive Interlanguage Analysis 703.3. A comparison of learner vs. expert writing 723.4. Summary and conclusion 78

Chapter 4 Rhetorical functions in expert academic writing 814.1. The Academic Keyword List and rhetorical functions 814.2. The function of exemplication 88

4.2.1. Using prepositions, adverbs and adverbial phrases to exemplify 904.2.2. Using nouns and verbs to exemplify 954.2.3. Discussion 106

4.3. The phraseology of rhetorical functions in expert academic writing 1084.4. Summary and conclusion 122

Chapter 5 Academic vocabulary in the International Corpus of Learner English 125

5.1. A bird’s-eye view of exemplifi cation in learner writing 1255.2. Academic vocabulary and general interlanguage features 142

5.2.1. Limited lexical repertoire 1425.2.2. Lack of register awareness 1505.2.3. The phraseology of academic vocabulary

in learner writing 1545.2.4. Semantic misuse 1685.2.5. Chains of connective devices 1745.2.6. Sentence position 177

5.3. Transfer-related effects on French learners’ use of academic vocabulary 1815.4. Summary and conclusion 192

Part III: Pedagogical implications and conclusions

Chapter 6 Pedagogical implications 2016.1. Teaching-induced factors 2016.2. The role of the fi rst language in EFL learning and teaching 2036.3. The role of learner corpora in EAP materials design 206

Contents ix

Chapter 7 General Conclusion 2117.1. Academic vocabulary: a chimera? 2117.2. Learner corpora, interlanguage and second language acquisition 2157.3. Avenues for future research 216

Appendix 1: Expressing cause and effect 219Appendix 2: Comparing and contrasting 226Notes 235References 240Author index 257Subject index 261

Acknowledgements

There are several people without whom this book would never have been written. First and foremost, I want to express my deepest and most sincere gratitude to my PhD supervisor, Professor Sylviane Granger, for her infectious enthusiasm, her intellectual perceptiveness and her unfailing expert guidance. I am greatly indebted to you, Sylviane, for giving me the opportunity to join the renowned Centre for English Corpus Linguistics seven years ago now! I have been lucky enough to undertake research in an environment where writing a PhD also means collaborating with many fellow researchers on up-and-coming projects, attending thought-provoking conferences, organizing seminars, conferences and summer schools, as well as lecturing and offering guidance to undergraduate students.

I am also very grateful to my colleagues and friends at the Centre for English Corpus Linguistics - Céline, Claire, Fanny, Gaëtanelle, Jennifer, Marie-Aude, Suzanne and Sylvie – for making the Centre for English Corpus Linguistics such an inspiring and intellectually stimulating research centre. I also wish to thank them for their moral and intellectual support and for all the entertaining lunchtimes we spent together talking about everyday life . . . and work.

I am indebted to a great number of colleagues not only for supplying me with corpora, corpus-handling tools and references, but also for providing helpful comments on earlier versions and stimulating ideas for my research. I would like to thank Yves Bestgen, Liesbet Degand, Jean Heiderscheidt, Sebastian Hoffmann, Scott Jarvis, Jean-René Klein, Fanny Meunier, Hilary Nesi, John Osborne and JoAnne Neff van Aertselaer. I am also grateful to an anonymous reviewer for recommendations on the fi rst draft of the text.

I gratefully acknowledge the support of both the Communauté française de Belgique, which funded my doctoral dissertation out of which this book has grown, and the Belgian National Fund for Scientifi c Research (F.N.R.S).

xii Acknowledgements

On a more personal note, I would like to express my deepest thanks to my parents and friends for everything they have done to help me while I was working on this book. And last, but not least, Arnaud: thank you for making it all worthwhile.

Magali PaquotLouvain-la-NeuveNovember, 2009

List of abbreviations

AKL Academic Keyword List (my own list)AWL Academic Word List (Coxhead, 2000)BAWE British Academic Written English (BAWE) Pilot

CorpusBNC British National CorpusB-BNC Baby BNC Academic CorpusBNC-AC British National Corpus – academic sub-corpusBNC-AC-HUM British National Corpus – academic sub-corpus

(discipline: humanities and arts)BNC-SP British National Corpus – spoken sub-corpusCALL Computer-assisted language learningCECL Centre for English Corpus Linguistics, Université

catholique de LouvainCIA Contrastive Interlanguage AnalysisCLAWS Constituent Likelihood Automatic Word-tagging

systemCODIF Corpus de Dissertations FrançaisesEAP English for academic purposesEFL English as a foreign languageESL English as a second languageESP English for specifi c purposesGSL General Service List (West, 1953)ICLE International Corpus of Learner English (Granger

et al., 2002)ICLEv2 International Corpus of Learner English (version 2)

(Granger et al., 2009)IL interlanguageL1 First languageL2 Foreign languageLDOCE4 Longman Dictionary of Contemporary English

(4th edition)

LOCNESS Louvain Corpus of Native Speaker EssaysLogL Log-likelihood statistical testMC Micro-Concord Corpus Collection BMED2 Macmillan English Dictionary for Advanced Learners

(second edition)MLD Monolingual learners’ dictionaryNS Native speakerNNS Non-native speakerpmw Per million wordsPOS Part-of-speechSLA Second language acquisitionUCREL University Centre for Computer Corpus Research on

Language, Lancaster UniversityWST4 WordSmith Tools (version 4)

xiv List of abbreviations

List of fi gures

Figure 1.1: The relationship between academic and sub-technical vocabulary 21

Figure 2.1: A three-layered sieve to extract potential academic words 45

Figure 2.2: WordSmith Tools – WordList option 49Figure 2.3: Distribution of the words example and law in the

15 sub-corpora 50Figure 2.4: WordSmith Tools Detailed Consistency Analysis 51Figure 2.5: Distribution of the noun ‘solution’ 53

Figure 3.1: ICLE task and learner variables (Granger et al., 2002: 13) 68

Figure 3.2: Contrastive Interlanguage Analysis (Granger 1996a) 70Figure 3.3: BNCweb Collocations option 77

Figure 4.1: Exemplifi cation in the BNC-AC-HUM 89Figure 4.2: The distribution of the adverb ‘notably’

across genres 93Figure 4.3: The distribution of ‘by way of illustration’

across genres 94Figure 4.4: The distribution of ‘to name but a few’

across genres 95Figure 4.5: The distribution of the verbs ‘illustrate’ and

‘exemplify’ across genres 103Figure 4.6: The phraseology of rhetorical functions

in academic prose 121

Figure 5.1: Exemplifi ers in the ICLE and the BNC-AC-HUM 127Figure 5.2: The use of the prepositions ‘like’ and ‘such as’

in different genres 131Figure 5.3: The use of the adverb ‘notably’ in different genres 131

xvi List of fi gures

Figure 5.4: Distribution of the adverbials ‘for example’ and ‘for instance’ across genres in the BNC 132

Figure 5.5: The treatment of ‘namely’ on websites devoted to English connectors 140

Figure 5.6: The use of ‘despite’ and ‘in spite of’ in different genres 145

Figure 5.7: The frequency of speech-like lexical items in expert academic writing, learner writing and speech (based on Gilquin and Paquot, 2008) 153

Figure 5.8: Phraseological cascades with ‘in conclusion’ and learner-specifi c equivalent sequences 161

Figure 5.9: Collocational overlap 165Figure 5.10: A possible rationale for the use of ‘according to me’

in French learners’ interlanguage 187Figure 5.11: A possible rationale for the use of ‘let us in

French learners’ interlanguage 191Figure 5.12: Features of novice writing - Frequency in expert

academic writing, native-speaker and EFL novices’ writing and native speech (per million words of running text) 195

Figure 6.1: Connectives: contrast and concession ( Jordan 1999:136) 202

Figure 6.2: Comparing and contrasting: using nouns such as ‘resemblance’ and ‘similarity’ (Gilquin et al., 2007b: IW5) 208

Figure 6.3: Reformulation: Explaining and defi ning: using ‘i.e.’, ‘that is’ and ‘that is to say’ (Gilquin et al., 2007b: IW9) 209

Figure 6.4: Expressing cause and effect: ‘Be careful’ note on ‘so’ (Gilquin et al., 2007b: IW13) 210

List of tables

Table 1.1: Composition of the Academic Corpus (Coxhead 2000: 220) 12

Table 1.2: Chung and Nation’s (2003: 105) rating scale for fi nding technical terms, as applied to the fi eld of anatomy 14

Table 1.3: Word families in the AWL 17

Table 2.1: The corpora of professional academic writing 31Table 2.2: The re-categorization of data from the professional

corpus into knowledge domains 32Table 2.3: The corpora of student academic writing 33Table 2.4: Examples of essay topics in the BAWE pilot corpus 34Table 2.5: An example of CLAWS vertical output 39Table 2.6: CLAWS horizontal output [lemma + POS] 40Table 2.7: CLAWS horizontal output [lemma + simplifi ed

POS tags] 40Table 2.8: Simplifi cation of CLAWS POS-tags 41Table 2.9: CLAWS tagging of the complex preposition

‘in terms of’ 41Table 2.10: Semantic fi elds of the UCREL Semantic

Analysis System 42Table 2.11: USAS vertical output 43Table 2.12: USAS horizontal output 44Table 2.13: The fi ction corpus 47Table 2.14: Number of keywords 47Table 2.15: Automatic semantic analysis of potential

academic words 54Table 2.16: Distribution of grammatical categories in the

Academic Keyword List 55Table 2.17: The Academic Keyword List 56Table 2.18: The distribution of AKL words in the GSL

and the AWL 60

xviii List of tables

Table 3.1: Breakdown of ICLE essays 69Table 3.2: BNC Index – Breakdown of written BNC genres

(Lee 2001) 74

Table 4.1: Ways of expressing exemplifi cation found in the BNC-AC-HUM 89

Table 4.2: The use of ‘for example’ and ‘for instance’ in the BNC-AC-HUM 91

Table 4.3: The use of ‘example’ and ‘for example’ in the BNC-AC-HUM 95

Table 4.4: Signifi cant verb co-occurrents of the noun ‘example’ in the BNC-AC-HUM 96

Table 4.5: Adjective co-occurrents of the noun ‘example’ in the BNC-AC-HUM 100

Table 4.6: The use of the lemma ‘illustrate’ in the BNC-AC-HUM 103Table 4.7: The use of the lemma ‘exemplify’ in the BNC-AC-HUM 105Table 4.8: The use of imperatives in academic writing (based

on Siepmann, 2005: 119) 107Table 4.9: Ways of expressing a concession in the

BNC-AC-HUM 109Table 4.10: Ways of reformulating, paraphrasing and clarifying

in the BNC-AC-HUM 109Table 4.11: Ways of expressing cause and effect

in the BNC-AC-HUM 110Table 4.12: Ways of comparing and contrasting found in

the BNC-AC-HUM 112Table 4.13: Co-occurrents of nouns expressing cause or effect

in the BNC-AC-HUM 115Table 4.13a: reason 115Table 4.13b: implication 115Table 4.13c: effect 116Table 4.13d: outcome 116Table 4.13e: result 117Table 4.13f: consequence 117Table 4.14: Co-occurrents of verbs expressing possibility and

certainty in the BNC-AC-HUM 119Table 4.14a: suggest 119Table 4.14b: prove 120Table 4.14c: appear 120Table 4.14d: tend 120

List of tables xix

Table 5.1: A comparison of exemplifi ers based on the total number of running words 128

Table 5.2: A comparison of exemplifi ers based on the total number of exemplifi ers used 129

Table 5.3: Two methods of comparing the use of exemplifi ers 130Table 5.4: Signifi cant adjective co-occurrents of the noun

‘example’ in the ICLE 133Table 5.5: Adjectives co-occurrents of the noun ‘example’

in ICLE not found in the BNC 133Table 5.6: Signifi cant verb co-occurrents of the noun ‘example’

in the ICLE 134Table 5.7: Verb co-occurrent types of the noun ‘example’

in ICLE not found in BNC 134Table 5.8: The distribution of ‘example’ and ‘be’ in the ICLE

and the BNC-AC-HUM 135Table 5.9: The distribution of ‘there + BE + example’ in ICLE

and the BNC-AC-HUM 135Table 5.10: The distribution of AKL words in the ICLE 143Table 5.11: Examples of AKL words which are overused and

underused in the ICLE 144Table 5.12: Two ways of comparing the use of cause and effect

markers in the ICLE and the BNC 146Table 5.13: The over- and underuse by EFL learners of specifi c

devices to express cause and effect (based on Appendix 1) 147

Table 5.14: The over- and underuse by EFL learners of specifi c devices to express comparison and contrast (based on Appendix 2) 149

Table 5.15: Speech-like overused lexical items per rhetorical function 151

Table 5.16: The frequency of ‘maybe’ in learner corpora 154Table 5.17: The frequency of ‘I think’ in learner corpora 154Table 5.18: Examples of overused and underused clusters

with AKL words 156Table 5.19: Clusters of words including AKL verbs which

are over- and underused in learners’ writing, by comparison with expert academic writing 158

Table 5.20: Examples of overused clusters in learner writing 159Table 5.21: Verb co-occurrents of the noun conclusion

in the ICLE 162

xx List of tables

Table 5.22: Adjective co-occurrents of the noun conclusion in the ICLE 167

Table 5.23: The frequency of sentence-initial position of connectors in the BNC-AC-HUM and the ICLE 178

Table 5.24: Sentence-fi nal position of connectors in the ICLE and the BNC-AC-HUM 181

Table 5.25: Jarvis’s (2000) three effects of potential L1 infl uence 183Table 5.26: Jarvis’s (2000) unifi ed framework applied to

the ICLE-FR 184Table 5.27: A comparison of the use of the English verb

‘illustrate’ and the French verb ‘illustrer’ 188Table 5.28: ‘let us’ in learner texts 189Table 5.29: The transfer of frequency of the fi rst person

plural imperative between French and English writing 191

Table 6.1: Le Robert & Collins CD-Rom (2003–2004): Essay writing 205

Introduction

That English has become the major international language for research and publication is beyond dispute. As a result, university students need to have good receptive command of English if they want to have access to the literature pertaining to their discipline. As a large number of them are also required to write academic texts (e.g. essays, reports, MA dissertations, PhD theses, etc.), they also need to have a productive knowledge of academic language. As noted by Biber, ‘students who are beginning university studies face a bewildering range of obstacles and adjustments, and many of these diffi culties involve learning to use language in new ways’ (2006: 1). Several studies have shown that the distinctive, highly routinized, nature of academic prose is problematic for many novice native-speaker writers (e.g. Cortes, 2002), but poses an even greater challenge to students for whom English is a second (e.g. Hinkel, 2002) or foreign language (e.g. Gilquin et al., 2007b).

Studies in second language writing have established that learning to write second-language (L2) academic prose requires an advanced linguistic com-petence, without which learners simply do not have the range of lexical and grammatical skills required for academic writing (Jordan, 1997; Nation and Waring, 1997; Hinkel, 2002; 2004; Reynolds, 2005). A questionnaire survey of almost 5,000 undergraduates showed that students from all 26 depart-ments at the Hong Kong Polytechnic University experienced diffi culties with the writing skills necessary for studying content subjects through the medium of English (Evans and Green, 2006). Almost 50 per cent of the students reported that they encountered diffi culties in using appropriate academic style, expressing ideas in correct English and linking sentences smoothly. Mastering the subtleties of academic prose is, however, not only a problem for novice writers. International refereed journal articles are regarded as the most important vehicle for publishing research fi ndings and non-native academics who want to publish their work in those top jour-nals often fi nd their articles rejected, partly because of language problems.

2 Academic Vocabulary in Learner Writing

These problems include the fact that they have less facility of expression and a poorer vocabulary; they fi nd it diffi cult to ‘hedge’ appropriately and the structure of their texts may be infl uenced by their fi rst language (see Flowerdew, 1999).

Because it causes major diffi culties to students and scholars alike, academic discourse has become a major object of study in applied linguis-tics. Flowerdew (2002) identifi ed four major research paradigms for investigating academic discourse, namely (Swalesian) genre analysis, contrastive rhetoric, ethnographic approaches and corpus-based analysis. While the fi rst three approaches to English for Academic Purposes (EAP) emphasize the situational or cultural context of academic discourse, corpus-linguistic methods focus more on the co-text of selected lexical items in academic texts.

Corpus linguistics is concerned with the collection in electronic format and the analysis of large amounts of naturally occurring spoken or written data ‘selected according to external criteria to represent, as far as possible, a language or language variety as a source of linguistic research’ (Sinclair, 2005: 16). Computer corpora are analysed with the help of software pack-ages such as WordSmith Tools 4 (Scott, 2004), which includes a number of text-handling tools to support quantitative and qualitative textual data anal-ysis. Wordlists give information on the frequency and distribution of the vocabulary – single words but also word sequences – used in one or more corpora. Wordlists for two corpora can be compared automatically so as to highlight the vocabulary that is particularly salient in a given corpus, i.e., its keywords. Concordances are used to analyse the co-text of a linguistic feature, in other words its linguistic environment in terms of preferred co-occurrences and grammatical structures. The research paradigm of corpus linguistics is ideally suited for studying the linguistic features of academic discourse as it can highlight which words, phrases or structures are most typical of the genre and how they are generally used.

Corpus-based studies have already shed light on a number of distinctive linguistic features of academic discourse as compared with other genres. Biber’s (1988) study of variation across speech and writing has shown that academic texts typically have an informational and non-narrative focus; they require highly explicit, text-internal reference and deal with abstract, conceptual or technical subject matter (Biber, 1988: 121–60). The Longman Grammar of Spoken and Written English (Biber et al., 1999) provides a compre-hensive description of the range of distinctive grammatical and lexical features of academic prose, compared to conversation, fi ction and newspa-per reportage. Common features of this genre include a high rate of

Introduction 3

occurrence of nouns, nominalizations, noun phrases with modifi ers, attributive adjectives, derived adjectives, activity verbs, verbs with inanimate subjects, agentless passive structures and linking adverbials. By contrast, fi rst and second person pronouns, private verbs, that-deletions and contrac-tions occur very rarely in academic texts.

In addition, studies of vocabulary have emphasized the importance of a ‘sub-technical’ or ‘academic’ vocabulary alongside core words and techni-cal terms in academic discourse (Nation, 2001: 187–216). Hinkel (2002: 257–65) argues that the exclusive use of a process-writing approach, the relative absence of direct and focused grammar instruction, and the lack of academic vocabulary development contribute to a situation in which non-native students are simply not prepared to write academic texts. She pro-vides a list of priorities in curriculum design and writes that, among the top priorities, ‘NNSs [non-native students] need to learn more contextualized and advanced academic vocabulary, as well as idioms and collocations to develop a substantial lexical arsenal to improve their writing in English’ (Hinkel, 2002: 247). The Academic Word List (Coxhead, 2000) was compiled on the basis of corpus data to meet the specifi c vocabulary needs of stu-dents in higher education settings.

But what is ‘academic vocabulary’? Despite its widespread use, the term has been used in various ways to refer to different (but often overlapping) vocabulary categories. This book aims to provide a better description of the notion of ‘academic vocabulary’. It takes the reader full circle, from the extraction of potential academic words through their linguistic analysis in expert and learner corpus data, to the pedagogical implications that can be drawn from the results. Recent corpus-based studies have emphasized the specifi city of different academic disciplines and genres. As a result, research-ers such as Hyland and Tse (2007) question the widely held assumption that students need a common core vocabulary for academic study. They argue that the different disciplinary literacies undermine the usefulness of such lists and recommend that lecturers help students develop a discipline-based lexical repertoire.

This book is an attempt to resolve the tension between the particularizing trend which advocates the teaching of a more restricted, discipline-based vocabulary syllabus, and the generalizing trend which recognizes the existence of a common core ‘academic vocabulary’ that can be taught to a large number of learners in many disciplines. I fi rst argue that, to resolve this tension, the concept of ‘academic vocabulary’ must be revisited. I demonstrate, on the basis of corpus data, that, as well as discipline-specifi c vocabulary, there is a wide range of words and phraseological patterns that


are used to refer to activities which are characteristic of academic discourse, and more generally, of scientifi c knowledge, or to perform important dis-course-organizing or rhetorical functions in academic writing.

A large proportion of this lexical repertoire consists of core vocabulary, a category which has so far been largely neglected in EAP courses but which is usually not fully mastered by English as a foreign language (EFL) learn-ers, even those at the high-intermediate or advanced levels. I make use of Granger’s (1996a) Contrastive Interlanguage Analysis to test the working hypothesis that upper-intermediate to advanced EFL learners, irrespective of their mother tongue background, share a number of linguistic features that characterize their use of academic vocabulary. The learner corpus used is the fi rst edition of the International Corpus of Learner English (ICLE), which is among the largest non-commercial learner corpora in existence. It contains texts written by learners with different mother tongue back-grounds. Ten ICLE sub-corpora representing different mother tongue backgrounds (Czech, Dutch, Finnish, French, German, Italian, Polish, Russian, Spanish, Swedish) are compared with a subset of the academic component of the British National Corpus (texts written by specialists in the Humanities) to identify ways in which learners’ use of academic vocabulary differs from that of more expert writers. A comparison of the ten sub- corpora then makes it possible to identify linguistic features that are shared by learners from a wide range of mother tongue backgrounds, and therefore possibly developmental. The EFL learners are all learning how to write in a foreign language, and they are often novice writers in their mother tongue as well.

However, not all learner specifi c-features can be attributed to develop-mental factors. The comparison of several ICLE sub-corpora helps to pinpoint a number of patterns that are characteristic of learners who share the same fi rst language, and which may therefore be transfer-related. I made use of Jarvis’s (2000) unifi ed framework to investigate the potential infl uence of the fi rst language on French learners’ use of academic vocabu-lary in English.

The book is organized in three sections. The fi rst scrutinizes the concept of ‘academic vocabulary’, reviewing the many defi nitions of the term and arguing that, for productive purposes, academic vocabulary is more use-fully defi ned as a set of options to refer to those activities that characterize academic work, organize scientifi c discourse, and build the rhetoric of academic texts. It then proposes a data-driven procedure based on the criteria of keyness, range, and evenness of distribution, to select academic words that could be part of a common core academic vocabulary syllabus.

Introduction 5

The resulting list, called the Academic Keyword List (AKL), comprises a set of 930 potential academic words. One important feature of the methodology is that, unlike Coxhead’s (2000) Academic Word List, the AKL includes the 2,000 most frequent words of English, thus making it possible to appreciate the paramount importance of core English words in academic prose.

The AKL is used in Section 2 to explore the importance of academic vocabulary in expert writing and to analyse EFL learners’ use of lexical devices that perform rhetorical or organizational functions in academic writing. This section offers a thorough analysis of these lexical devices as they appear in the International Corpus of Learner English, describing the fac-tors that account for learners’ diffi culties in academic writing. These factors include a limited lexical repertoire, lack of register awareness, infelicitous word combinations, semantic misuse, sentence-initial positioning of adverbs and transfer effects.

The fi nal section briefl y comments on the pedagogical implications of these results, summarizes the major fi ndings, and points the way forward to further research in the area.

Part I

Academic vocabulary

‘Academic vocabulary’ is a term that is widely used in textbooks on English for academic purposes and Second Language Acquisition (SLA) reference books. Nevertheless, it can be understood in a variety of ways and used to indicate different categories of vocabulary. In this section, my objectives are to clarify the meaning of ‘academic vocabulary’ by critically examining its many uses, and to build a list of words that fi t my own defi nition of the term. Chapter 1 therefore tries to identify the key features of academic vocabu-lary and to clear up the confusion between academic words and other vocabulary. Chapter 2 proposes a data-driven methodology based on the criteria of keyness, range and evenness of distribution, and uses this to build a new list of potential academic words, viz. the Academic Keyword List (AKL). This list is very different from Coxhead’s Academic Word List and has already been used to inform the writing sections in the second edition of the Macmillan English Dictionary for Advanced Learners (see Gilquin et al., 2007b). The AKL is used in Section 2 to analyse EFL learners’ use of lexical devices that perform rhetorical or organizational functions in academic writing.

Chapter 1

What is academic vocabulary?

Academic vocabulary is in fashion, as witnessed by the increasing number of textbooks on the topic. Recent titles include Essential Academic Vocabulary: Mastering the Complete Academic Word List (Huntley, 2006) and Academic Vocab-ulary in Use (McCarthy and O’Dell, 2008). But what is academic vocabulary? The term often refers to a set of lexical items that are not core words but which are relatively frequent in academic texts. Examples of academic words include adult, chemical, colleague, consist, contrast, equivalent, likewise, parallel, transport and volunteer (cf. Coxhead, 2000). Unlike technical terms, they appear in a large proportion of academic texts, regardless of the disci-pline. Academic vocabulary is also sometimes used as a synonym for sub-technical vocabulary (e.g. mouse, bug, nuclear, solution) or discourse-organizing vocabulary (e.g. cause, compare, differ, feature, hypothetical, and identify). In this chapter, I set out to review the many defi nitions of academic vocabulary that have been given and to clear up the confusion between academic words, core words, technical terms, sub-technical words and discourse-organizing words. I will show why a defi nition of academic vocabulary that excludes the top 2,000 words of English is not very useful for productive purposes in higher education settings and argue for a function-based defi nition of the term. The very existence of academic words has recently been challenged by several researchers in English for Specifi c Purposes (ESP) who advocate that teachers help students develop a more restricted, discipline-specifi c lexical repertoire. I will round off this chapter by situat-ing the book in ongoing debates over generality vs. disciplinary specifi city in teaching vocabulary for academic purposes.

1.1. Academic vocabulary vs. core vocabulary and technical terms

Numerous second language acquisition studies have investigated whether there is a threshold which marks the point at which vocabulary knowledge


becomes suffi cient for adequate reading comprehension. Laufer (1989; 1992) has shown that at least 95 per cent coverage is needed to ensure reasonable comprehension of a text. To achieve this coverage, it is com-monly believed that students in higher education settings need to master three lists of vocabulary: a core vocabulary of 2,000 high-frequency words, plus some academic words, and technical terms. Some researchers, however, do not agree that vocabulary categories can be described as if they were clearly separable. In this section, the notions of core vocabulary, academic vocabulary and technical terms are described and illustrated. The criticisms levelled at the division of vocabulary into mutually exclusive lists are then reviewed.

1.1.1. Core vocabulary

A core (or basic or nuclear) vocabulary consists of words that are of high frequency in most uses of the language. It comprises the most useful func-tion words (e.g. a, about, be, by, do, he, I, some and to) and content words like bag, lesson, person, put and suggest. Stubbs describes nuclear words as an essential common core of ‘pragmatically neutral words’ (1986: 104) and lists fi ve main reasons for their pragmatic neutrality:

1. Nuclear words have a ‘purely conceptual, cognitive, logical or proposi-tional meaning, with no necessary attitudinal, emotional or evaluative connotations’ (ibid.).

2. They have no cultural or geographical associations.3. They give no indication of the fi eld of discourse from which a text is

taken, i.e. its domain of experience and social settings. 4. They are also neutral with respect to tenor and mode of discourse: they

are not restricted to formal or informal usage or to a specifi c medium of communication, e.g. written or spoken language.

5. They are used in preference to non-nuclear words in summarizing tasks.

The best-known list of core words is West’s (1953) General Service List of English Words (GSL),1 which was created from a fi ve-million word corpus of written English and contains around 2,000 word families. Percentage fi gures are given for different word meanings and parts of speech of each head-word. In a variety of studies, the GSL provided coverage of up to 92 per cent of fi ction texts (e.g. Hirsh and Nation, 1992), and up to 76 per cent of academic texts (Coxhead, 2000). Next to frequency and coverage, other

What is academic vocabulary? 11

criteria such as learning ease, necessity and style were also used in making the selection (West 1953: ix–x). West also wanted the list to include words that are often used in the classroom or that would be useful for understand-ing defi nitions of vocabulary outside the list. The GSL has had a wide infl u-ence for many years and served as a resource for writing graded readers and other material.

A number of criticisms have, however, been levelled at the GSL, most particularly at its coverage and age. Engels (1968) criticized the low cover-age of the second 1,000 word families. While the fi rst 1,000 word families covered between 68 and 74 per cent of the words in the ten texts of 1,000 running words he analysed, the second set of word families in the GSL provided coverage of less than 10 per cent. In addition, because of changes in the English language and culture, the GSL includes many words that are considered to be of limited utility today (e.g. crown, coal, ornament and vessel) but does not contain very common words such as computer, astronaut and television (see Nation and Hwang, 1995: 35–6; Leech et al., 2001: ix–x; Carter, 1998: 207). However, several researchers have pointed out that, for educa-tional purposes, it still remains the best of the available lists because of ‘its information on frequency of each word’s various meanings, and West’s careful application of criteria other than frequency and range’ (Nation and Waring 1997:13).

1.1.2. Academic vocabulary

A number of academic word lists have been compiled to meet the specifi c vocabulary needs of students in higher education settings (e.g. Campion and Elley, 1971; Praninskas, 1972; Lynn, 1973; Ghadessy, 1979; Xue and Nation, 1984). The Academic Word List (Coxhead, 2000) is the most widely used today in language teaching, testing and the development of pedagogi-cal material. It is now included in vocabulary textbooks (e.g. Schmitt and Schmitt, 2005; Huntley, 2006), vocabulary tests (e.g. Schmitt et al., 2001), computer-assisted language learning (CALL) materials, and dictionaries (e.g. Major, 2006).

The Academic Word List (AWL) was created from a corpus of 414 academic texts by more than 400 authors and totals around 3.5 million words. The Academic Corpus includes journal articles, chapters from university textbooks and laboratory manuals. It is divided into four sub-corpora of approximately 875,000 words representing broad academic disciplines: arts, commerce, law and science. Each sub-corpus is further subdivided into seven subject areas as shown in Table 1.1.


Like the General Service List, the Academic Word List is made up of word families. Each family consists of a headword and its closely related affi xed forms according to Level 6 of Bauer and Nation’s (1993) scale, which includes all the infl ections and the most frequent and productive deriva-tional affi xes. For example, the words presumably, presume, presumed, presumes, presuming, presumption, presumptions and presumptuous are all members of the same family.

Coxhead (2000) selected word families to be included in the AWL on the basis of three criteria:

1. Specialized occurrence: a word family could not be in the fi rst 2,000 most frequent words of English as listed in West’s (1953) General Service List.

2. Range: a word family had to occur in all four academic disciplines with a frequency of at least 10 in each sub-corpus and in 15 or more of the 28 subject areas.

3. Frequency: a word family had to occur at least 100 times in the Academic Corpus.

The resulting list consists of 570 word families and covers at least 8.5 per cent of the running words in academic texts. By contrast, it accounts for a very small percentage of words in other types of texts such as novels, suggesting that the AWL’s word families are closely associated with academic writing (Coxhead, 2000: 225). It is divided into 10 sublists ordered accord-ing to decreasing word-family frequency. Some of the most frequent word families included in Sublist 1 are headed by the word forms analyse, benefi t, context, environment, formula, issue, labour, research, signifi cant and

Table 1.1 Composition of the Academic Corpus (Coxhead 2000: 220)

Running words Texts Subject areas

Arts 883,214 122 education; history; psychology; politics; psychol-ogy; sociology

Commerce 879,547 107 accounting; economics; fi nance; industrial rela-tions; management; marketing; public policy

Law 874,723 72 constitutional law; criminal law; family law and medico-legal; international law; pure commer-cial law; quasi-commercial law; rights and remedies

Science 875,846 113 biology; chemistry; computer science; geography; geology; mathematics; physics

Total 3,513,330 414


vary. Examples of the least frequent word families in Sublist 10 are assemble, colleague, depress, enormous, likewise, persist and undergo.

Academic words are likely to be problematic for native as well as non-native students as a large proportion of them are Graeco-Latin in origin and refer to abstract ideas and processes, thus introducing additional prop-ositional density to a text (cf. Corson, 1997). Scarcella and Zimmerman (2005: 127) have also shown that mastery of derivative forms makes aca-demic words particularly diffi cult for foreign language learners who often fail to analyse the different parts of complex words.

1.1.3. Technical terms

Domain-specifi c or technical terms are words whose meaning requires scientifi c knowledge. They are typically characterized by semantic special-ization, resistance to semantic change and absence of exact synonyms (cf. Mudraya, 2006: 238–9). As explained by Nation (2001: 203), some prac-titioners consider that it is not the English teacher’s job to teach technical terms. These words are best learned through the study of the body of knowledge that they are attached to. Language teachers are not specialists in chemistry, computer science, law or economics and may have a great deal of diffi culty with technical words. By contrast, learners who specialize in the fi eld may have little diffi culty in understanding these words (Strevens, 1973: 228).

Since technical terms are highly subject-specifi c, it is possible to identify them on the basis of their frequencies of occurrence, range and distribu-tion (see Section 2.3) and to use them as a way of characterizing text types (Yang, 1986). Technical terms occur with very high or at least moderate frequency within a very limited range of texts (Nation and Hwang, 1995). In biology, for example, we fi nd words such as alleles, genotype, chromatid, cyto-plasm and abiotic. These words are very unlikely to occur in texts from other disciplines or subject areas. Technical vocabulary is diffi cult to quantify. According to Coxhead and Nation (2001), technical dictionaries contain probably 1,000 headwords or less per subject area. Research suggests that knowledge of domain-specifi c or technical terms allows learners to under-stand an additional 5 per cent of academic texts in a specifi c discipline.

1.1.4. Fuzzy vocabulary categories

Although core words, academic words and technical terms are described as if they were clearly separable, the boundaries between them are fuzzy


(cf. Yang, 1986; Mudraya, 2006; Beheydt, 2005). As Nation and Hwang remark, ‘any division is based on an arbitrary decision on what numbers represent high, moderate or low frequency, or wide or narrow range, because vocabulary frequency, coverage and range fi gures for any text or group of texts occur along a continuum’ (1995: 37). Chung and Nation (2003) investigate what kinds of words make up technical vocabulary in anatomy and applied linguistics texts. They classify technical terms on a four-level scale designed to measure the strength of the relationship of a word to a particular specialized fi eld. Results for vocabulary in anatomy texts are given in Table 1.2. Chung and Nation consider items at Steps 3 and 4 to be technical terms, but not items at Steps 1 and 2. A large pro-portion of technical words belong to the 2,000 most frequent word families of English as given in the GSL or to the AWL. In the anatomy texts, 16.3 per cent of the word types at Step 3 are from the GSL or AWL (e.g. cage, chest, neck, shoulder). This increases to 50.5 per cent in the applied linguistics texts (e.g. acquisition, input, interaction, meaning, review). A major result of this study is that a word can only be described as general service, academic or technical in context.

Table 1.2 Chung and Nation’s (2003: 105) rating scale for fi nding technical terms, as applied to the fi eld of anatomy

Step 1

Words such as function words that have a meaning that has no particular relationship with the fi eld of anatomy, that is, words independent of the subject matter. Examples are: the, is, between, it, by, adjacent, amounts, common, commonly, directly, constantly, early and especially

Step 2

Words that have a meaning that is minimally related to the fi eld of anatomy in that they describe the positions, movements, or features of the body. Examples are: superior, part, forms, pairs, structures, surrounds, supports, associated, lodges, protects.

Step 3

Words that have a meaning that is closely related to the fi eld of anatomy. They refer to parts, structures and functions of the body, such as the regions of the body and systems of the body. Such words are also used in general language. The words may have some restrictions of usage depending on the subject fi eld. Examples are: chest, trunk, neck, abdomen, ribs, breast, cage, cavity, shoulder, skin, muscles, wall, heart, lungs, organs, liver, bony, abdominal, breathing. Words in this category may be technical terms in a specifi c fi eld like anatomy and yet may occur with the same meaning in other fi elds where they are not technical terms.

Step 4

Words that have a specifi c meaning to the fi eld of anatomy and are not likely to be used in general language. They refer to structures and functions of the body. These words have clear restrictions of usage depending on the subject fi eld. Examples are: thorax, sternum, costal, vertebrae, pectoral, fascia, trachea, mammary, periosteum, hematopoietic, pectoralis, viscera, intervertebral, demifacets, pedicle.


Similarly, it has been shown that the GSL contains words that appear with particularly high range and frequency in academic texts (e.g. example, reason, argument, result, use, fi nd, show) (cf. Martínez et al., 2009: 192). These words may be used differently in academic discourse. For example, Partington (1998: 98) has shown that a claim in academic or argumentative texts is not the same as in news reporting or a legal report. On the other hand, the AWL includes words that are extremely common outside academia (e.g. adult, drama, sex, tape) (Paquot, 2007a). Hancioglu et al. argue that ‘the assump-tion that any high frequency word outside the GSL coverage in the academic corpus would be a de facto academic item perhaps accounts for the distinctly “un-academic” texture of some of the items on the list’ (Hancioglu et al., 2008: 462). They also comment that the fact that ‘items such as study appear in the GSL (but not in the AWL) and items such as drama in the AWL (but not in the GSL), suggests that the division of vocabulary into mutually exclusive lists is likely to be an activity that for all its initial convenience may prove inherently problematic in the long run’ (ibid.: 463).

Originating from research on vocabulary needs for reading comprehen-sion and text coverage, the division between core words and academic words is very practical for assessing text diffi culty and targeting words that are worthy of explanation when reading an academic text in the classroom. Most English for Academic Purposes (EAP) students recognize core words but are not familiar with the meaning of academic words such as amend, concept, implement, normalize, panel, policy, principle and rationalize, which are not very common in everyday English. These words are, however, relatively frequent in academic texts and students will most probably encounter them quite often while reading. They should therefore be the focus of an aca-demic reading course.

The division of vocabulary into three mutually exclusive lists becomes problematic, however, when it is transposed to academic writing courses and the need arises to distinguish between knowing a word for receptive and productive purposes. As early as 1937, West argued that ‘both as regards Selection and still more as regards detailed Itemization, there is a need of a divorce between receptive and productive work’ (West, 1937: 437) and regretted that teachers were giving

composite lessons aiming at teaching reading and speaking simultane-ously, whereas reading and speaking are the Hare and the Tortoise. Reading and speech bear the same relation to each other as musical appreciation and actual execution on the piano. The one is Recognition of a lot; the other is Skill in using a little. (ibid.)


Learning vocabulary for productive purposes has been found to be much more diffi cult than learning for receptive uses. Knowing a word produc-tively involves, for example, being able to pronounce and/or spell it correctly, produce it to express the intended meaning in the appropriate context, and use it with words that commonly occur with it (Nation, 2001: 27–8). Selection is thus a key issue in teaching vocabulary for academic writing and speaking. It is questionable whether all the words from the AWL should be the focus of productive learning. And yet this strategy lies at the heart of several recent textbooks (e.g. Schmitt and Schmitt, 2005; Huntley, 2006) and CALL materials (see, e.g., Gillett’s website about vocab-ulary in EAP < http://www.uefap.com/vocab/vocfram.htm>; Luton’s Exercises for the Academic Word List < http://www.academicvocabularyex-ercises.com> and Haywood’s AWL Gapmaker <)

Several scholars have suggested replacing separate lists of general service words, academic vocabulary and technical terms by a single list, either a more specialized list or a larger common core vocabulary. Ward (1999), for example, built an engineering word list of 2,000 word families which con-tains both technical terms and all the general words necessary for reading comprehension and shows that it provides 95 per cent coverage of many basic engineering texts (see also Mudraya, 2006). Others, by contrast, have tried to revise the General Service List, to ensure maximum utility for any learner, regardless of specialization. Billuroglu and Neufeld (2007) combined into one list all the words from: (1) the GSL, (2) the AWL, (3) the fi rst 2,000 words of the Brown corpus, (4) the fi rst 5,000 words of the British National Corpus, (5) the revised version of the GSL, (6) the Longman Wordwise of commonly used words and (7) the Longman Defi ning Vocabulary. The resulting Billuroglu-Neufeld-List (BNL) consists of 2,709 word families categorized according to the number of lists in which they were represented. This procedure led to the emergence of only 176 word families that were not in either the GSL or the AWL, thus confi rming that ‘if the GSL was enlarged by even a relatively small degree, [. . . ] much of the AWL would be absorbed into it’ (Hancioglu et al., 2008: 466). See Stein (2008) for a similar approach.

A fi nal criticism that can be levelled at the AWL is related to the notion of a word family. The AWL, as well as most word lists for learners of English, groups words into families. Other examples include the GSL, the University Word List (Xue and Nation, 1984) and recent domain-specifi c lists such as those developed by Ward (1999) and Mudraya (2006). Coxhead (2000: 218) argues that this practice is supported by psycholinguistic evidence suggesting that morphological relations between words are represented in


the mental lexicon. This may well be true and may justify the use of word families for receptive purposes. However, not all members of a word family are likely to be equally helpful in academic writing. For example, under the headword item, which has a relative frequency of 134.29 occurrences per million words in the academic part of the British National Corpus (see Section 3.3), we fi nd the noun itemisation and word forms of the verb itemise. However, these two lemmas are quite rare in academic writing, with relative frequencies of 0.06 and 1.17 occurrences per million words respectively. A related problem is that parts-of-speech are not differenti-ated. Table 1.3 shows several word families taken from the AWL: the only information provided is that the words in italics are the most frequent form of their family. This, however, does not tell us whether the word forms issue and issues (under the headword issue) are more often used as nouns or verbs in EAP.

1.2. Academic vocabulary and sub-technical vocabulary

Like Coxhead (2000), Nation (2001: 187–96) uses the term ‘academic vocab-ulary’ to refer to words that are not in the top 2,000 words of English but which occur reasonably frequently in a wide range of academic texts. Unlike Coxhead, however, he also uses it to label a whole set of lexical items also known as ‘sub-technical vocabulary’ (Cowan 1974; Yang, 1986; Baker, 1988; Mudraya, 2006), ‘semi-technical vocabulary’ (Farrell, 1990), ‘non-technical terms’ (Goodman and Payne, 1981), and ‘specialised non-technical lexis’

Table 1.3 Word families in the AWL

link proceed issue evident item stress utilize

linkagelinkageslinkedlinkinglinks

proceduralprocedureproceduresproceededproceedingproceedingsproceeds

issuedissuesissuing

evidencedevidenceevidentialevidently

itemisationitemiseitemiseditemisesitemisingitems

stressedstressesstressfulstressingunstressed

utilisationutilisedutilisesutilisingutiliserutilisersutilityutilitiesutilizationutilizeutilizedutilizesutilizing


(Cohen et al, 1988). However, all these terms have been used quite differ-ently in the literature. Cowan defi nes sub-technical vocabulary as ‘context independent words which occur with high frequency across disciplines’ and comments that,

Clearly some of what I am calling sub-technical vocabulary would be encompassed in the existing word frequency counts like Thorndike Lorge, Michael West’s General Service List and the recent one million word com-puter analysis by Henry Kucera and Nelson Francis. (Cowan, 1974: 391)

Cowan’s defi nition of sub-technical vocabulary applies to those words that have the same meaning in several disciplines. Trimble (1985) extends Cowan’s (1974) usage to include ‘those words that have one or more “general” English meanings and which in technical contexts take on extended meanings’ (Trimble 1985: 129). Trimble’s defi nition thus encompasses words such as junction, circuit, wage and cage that would be categorized as technical terms according to Chung and Nation’s (2003) four-level rating scale of tech-nicality or fi eld-specifi city (see Table 1.2) (see also Farrell, 1990: 37).

Cohen et al. (1988) regard the extended meanings of what they call ‘non-technical’ words as a major area of diffi culty for non-native readers who may only be aware of one of their meanings. In biology, for example, the adjective specifi c may also be used with reference to the genetic notion of specifi city, which is a characteristic of enzymes. A second area of diffi culty arises because non-technical words may be used in contextual paraphrases to refer to the same concept (e.g. repair processes and repair mechanism in a genetics text), thus causing problems of lexical cohesion at the level of synonymy. Cohen et al. (1988) identify a subset of non-technical vocabulary as a third area of diffi culty, viz. ‘specialized non-technical lexis’. They do not offer a precise defi nition of the term, but explain that this lexis includes vocabulary items indicating, for example, time sequence, measurement, or truth validity. They show that a large proportion of vocabulary items which indicate time sequence or frequency in a genetics text are unknown to their informants (e.g. ensuing, alternatively, consecutively, intermittently, subsequent and successive).

In Li and Pemberton’s (1994) view, sub-technical vocabulary as defi ned by Trimble (1985) is an important subset of academic vocabulary. They showed that fi rst-year computer science students are better able to recog-nize the technical meanings of sub-technical words than their non-technical meanings. For example, they are quite familiar with the technical meaning of the verb compile in computer science and tend to interpret it as ‘convert


or translate a language into a machine code’ or ‘translate’ regardless of the context in which the word occurs. This is problematic as the non-technical meaning of a sub-technical word is often more common than its technical meaning (see Mudraya, 2006). For example, the word solution is more fre-quently used in its non-technical sense in engineering textbooks, even in a chemical engineering thermodynamics textbook.

Baker (1988) has argued that this middle area between core and techni-cal vocabulary is itself made up of several different types of vocabulary:

1. Items which express notions shared by all or several specialized disciplines. Examples include factor, method and function.

2. Items which have a specialized meaning in a particular fi eld, in addition to a different meaning in general language (e.g. bug in computer science, solution in mathematics and chemistry).

3. Items which are not used in general language but which have different technical meanings in different disciplines (e.g. morphological in linguis-tics, botany and biology).

4. General language items which have restricted meanings in one or more disciplines. In botany, ‘genes which are expressed have observable effects, i.e. are more apparent physically, as opposed to being masked. Expressed in botany is therefore not associated with emotional or verbal behaviour as is the case in general language’ (Baker, 1988: 92).

5. General language items which are used, in preference to other semanti-cally equivalent items, to describe or comment on technical processes and functions. For example, an examination of biology textbooks showed that photosynthesis does not happen but takes place or occasionally occurs. Baker thus comments that take place and occur can be regarded as sub-technical words.

6. Items which are used in academic texts to perform specifi c rhetorical functions. These are ‘items which signal the writer’s intentions or his evaluation of the material presented’ (Baker, 1988: 92).

Martin uses the term academic vocabulary as a synonym for sub-technical vocabulary to refer to words that ‘have in common a focus on research, analysis and evaluation – those activities which characterize academic work’ (1976: 92). The vocabulary of the research process consists primarily of verbs, nouns and their co-occurrences (e.g. state the hypothesis and expected results; present the methodology; plan or design the experiment; develop a model). The vocabulary of analysis includes high-frequency verbs and two-word verbs that are ‘often overlooked in teaching English to foreign students but


which graduate students need in order to present information in an organized sequence’ (ibid: 93), e.g. consist of, group, result from, derive, bring about, cause, base on, be noted for. Adjectives and adverbs make up a large proportion of the vocabulary of evaluation.

In summary, the many defi nitions of sub-technical vocabulary proposed in the literature cover very different sets of lexical items, which are of various sizes and may share certain characteristics. Sub-technical vocabu-lary is generally defi ned as a category of words which are frequent across disciplines and account for a signifi cant proportion of word tokens in academic texts. Farrell (1990), for example, found that out of 508 lemmas occurring more than fi ve times in a corpus of electronic texts, 44 per cent were sub-technical. Defi nitions of sub-technical vocabulary also differ widely, referring to words that take on extended meanings in specifi c academic disciplines (Trimble, 1985), or to words that allow scholars to conduct research, analyse data and evaluate results (Martin 1976). Baker (1988) uses the term as a broad category for different types of lexical sets including both Trimble’s (1985) sub-technical vocabulary and Martin’s (1976) academic vocabulary.

Figure 1.1 shows that the various defi nitions of sub-technical vocabulary and academic vocabulary as defi ned in Section 1.1.2 partially overlap. Coxhead’s (2000) Academic Word List includes a large proportion of the words that take on extended meanings in specialised fi elds (cf. Trimble’s defi nition of ‘sub-technical vocabulary’). For example, according to the Oxford English Dictionary (OED), the adjective nuclear has extended senses in astronomy, biology, medicine, psychoanalysis, sociology, linguistics and phonetics; the verb enable has a specialized meaning in computer science (‘to make (a device) operational; to turn on’). The noun error refers to ‘the quantity by which a result obtained by observation or by approximate calculation differs from an accurate determination’ in mathematics. The AWL also contains several sub-technical words as defi ned by Martin (1976) (e.g. hypothesis, signifi cant, method, function) but a large number of them do not fall within Coxhead’s defi nition of academic vocabulary. Many of these are general service words (e.g. cause, develop, group, model, plan, result). The same is true of Baker’s (1988) category of words that perform rhetorical functions: case, cause, compare, describe, explanation, observe, report, and study are among the top 2,000 most frequent words of English. This category will be the focus of the next section as it is itself made up of various sets of lexical items and Baker (1988) suggested that it is the most diffi cult type of sub-technical vocabulary to teach and acquire.


Baker's (1988) sub-technical vocabulary

Coxhead's (2000) academic vocabulary Martin's (1976)academic vocabulary

psychology hypothesiscause, develop,

group, model,interesting, show,experiment,remarkable,result, plan,present, observe

explanationincrease, case

result, study

compile

transport,journal,civil, nuclear,

decade, texterror

enable

morphological

Trimble's (1985) sub-technical vocabulary

base

fastmousedogbugsolution

'expressed' (genes)'masked' (genes)

appropriate

functionderive

consistfactor,

significant,methodcolleague

nevertheless

enormous

thereby

briefly

welfare

hence

widespreadparticipant

Figure 1.1 The relationship between academic and sub-technical vocabulary


1.3. Vocabulary and the organization of academic texts

Baker (1988) gave the following examples of sub-technical words that are used to perform rhetorical functions: ‘One explanation is that…’; ‘Others have said …’; and ‘It has been pointed out by . . . that . . .’. These words bear a strong relationship with what Winter (1977) called ‘Vocabulary 3 items’ and Widdowson (1983) ‘procedural vocabulary’. Winter (1977: 14–23) dis-tinguished between three types of words that are commonly used to create cohesion or structure in discourse and that are essential to the understand-ing of clause relations. Each group is distinguished by its clause-relating functions. Vocabulary 1 consists of ‘subordinators’ which either connect clauses together (e.g. although, as far as, except that, unless, whereas) or embed one clause within another (e.g. not so much . . . as . . .; not . . . let alone . . .). Vocabulary 2 comprises ‘sentence connectors’ which ‘make explicit the clause relation between the matrix clause and the preceding clause or sen-tence’ (Winter, 1977: 15), e.g. therefore, anyway, hence, for example, indeed, thus, that is to say, in other words. Vocabulary 3 items serve to establish semantic relations in the connection of clauses or sentences in discourse. They behave grammatically like subject, verb, object or complement and can be pre- or post-modifi ed. Examples include addition, affi rm, alike, analogous, cause, compare, connect, consequence, contradict, differ, explanation, feature, hypo-thetical, identify, method, reason, result, specify and subsequent. These words ‘may be used to make the relation explicit by saying what the relation is’ (Winter, 1977: 22). As such, they are part of what Widdowson (1983) called ‘proce-dural vocabulary’, i.e. highly context-dependent items with very little lexical content which serve to do things with the content-bearing words and draw attention to the function that a stretch of discourse is performing (see also Harris, 1997; Luzón Marco, 1999).

Vocabulary 3 items include a large proportion of nouns that are inher-ently unspecifi c and require lexical realization in their co-text, either beforehand or afterwards. Francis (1994) refers to this type of lexical cohe-sion as ‘advance’ and ‘retrospective labelling’: labels2 allow the reader to predict the precise information that will follow when they occur before their lexical realization and they encapsulate and package a stretch of dis-course when they occur after their realization (e.g. approach, area, aspect, case, matter, move, problem, and way). Labels have traditionally been described as content words. However, when we encounter them in a text, we often need to do ‘something similar to what we do when we encounter words like it, he and do in texts: we either refer to the bank of knowledge built up with the author, look back in the text to fi nd a suitable referent, or forward,


anticipating that the writer will supply the missing content’ (Carter and McCarthy, 1988: 206–7). As explained by McCarthy, ‘the language learner who has trouble with such words may be disadvantaged in the struggle to decode the whole text as effi ciently as possible and as closely as possible to the author’s designs’ (McCarthy, 1991: 76).

Within the category of labels, Francis identifi ed a set of nouns which are ‘metalinguistic in the sense that they label a stretch of discourse as being a particular type of language’ (1994: 89). Metalinguistic labels are of four types, although there is some overlap between them:

1. Illocutionary nouns are nominalizations of verbal processes, e.g. advice, answer, argument, assertion, claim, observation, recommendation, remark, reply, response, statement, suggestion.

2. Language-activity nouns refer to language activities and the results thereof, e.g. comparison, contrast, defi nition, description, detail, exam-ple, illustration, instance, proof, reasoning, reference, summary, etc.

3. Mental process nouns refer to cognitive states and processes and the results thereof, e.g. analysis, assumption, attitude, belief, concept, conviction, fi nding, hypothesis, idea, insight, interpretation, opinion, position, theory, thesis, view, etc.

4. Text nouns refer to the formal textual structure of discourse, e.g. phrase, words, quotation, excerpt, section, term, etc.

As pointed out by Nation, the strength of labels as discourse organizing vocabulary is that ‘they have a referential function and variable meaning like pronouns but, unlike pronouns, they can be modifi ed by demonstra-tive pronouns, numbers, and adjectives, they can occur in various parts of a sentence and they have a signifi cant constant meaning’ (2001: 212). As well as representing text segments, labels ‘additionally give us indications of the larger text-patterns the author has chosen, and build up expectations concerning the shape of the whole discourse’ (McCarthy, 1991: 76). The following words typically cluster round the elements of problem-solution patterns: concern, diffi culty, dilemma, hinder, obstacle, respond, consequence, effect and result (see also Jordan, 1984; Hoey, 1993; 1994; Flowerdew, 2008; Nation, 2001: 211). Labels not only cluster around elements of macro patterns, they are also characterized by their specifi c collocational environment as shown by Francis:

there is a tendency for the selection of a label to be associated with common collocations. Many labels are built into a fi xed phrase or ‘idiom’


(in the widest sense of the word), representing a single choice. Frequent collocations include, for example, ‘the move follows . . . ’, ‘. . . rejected/denied the allegations’, ‘. . . to solve a problem’, and ‘. . . to reverse the trend’, where the retrospective label is found in predictable company (. . . ). Even where the collocations are less fi xed, the label occurs in a compatible lexical environment. (1994: 100–1)

More generally, Baker comments that sub-technical words which perform specifi c rhetorical functions and structure the writer’s argument ‘should not be taught in isolation but in context and as central elements in typical collocations’ (Baker, 1988: 103).

Labels are not the only indicators of text patterns in academic discourse. The claim-counterclaim pattern, for example, is often organized with verbs such as assert and state, adjectives like false and likely, the preposition according to and adverbs such as apparently and arguably. In Building Academic Vocabu-lary, Zwier focused on lexical items that are ‘particularly useful in the kinds of writing most common in EAP writing classes – general description, description of processes (especially those involving changes), comparison/contrast, and cause/effect’ (Zwier, 2002: xiii) and described the way in which words such as consist of, comprise, parallel, alike, likewise, distinguish, raise, rise, link, stem from, and yield are used to perform specifi c rhetorical functions in academic discourse. He devoted particular attention to verbs because ‘accu-rate verb use is especially diffi cult for academic writers’ (Zwier, 2002: xi) (see also Swales and Feak, 2004). Similarly, Meyer (1997) commented that non-technical words ‘provide a semantic-pragmatic skeleton for the text. They determine the status of the (more or less technically phrased) propositions that are laid down in it, and the relations between them’ (Meyer, 1997: 9). For example, these words express temporal deixis (e.g. original, currently), modality (e.g. may, obviously, likely), epistemic relations between the subject matter and the scholar (e.g. indicate, seem), quantitative changes of entities (e.g. increase, fl uctuation), classifi ers of entities (e.g. problem, method, theory, characteristic), relations between entities (e.g. arising from, follow, since, involve), scholarly speech acts (e.g. suggest, proposal, show, defi ne), and textual deixis (e.g. above, later) (ibid: 10–11). Meyer’s ‘non-technical vocabulary’ is there-fore closely related to the notion of metadiscourse, i.e. ‘a specialized form of discourse which allows writers to engage with and infl uence their inter-locutors and assist them to interpret and evaluate the text in a way they will see as credible and convincing’ (Hyland, 2005: 60).

As shown in the previous sections, the term ‘academic vocabulary’ has been used extensively in the literature to refer to various sets of lexical


items. However, its very existence has recently been challenged by several ESP researchers.

1.4. Is there an ‘academic vocabulary’?

In an article entitled ‘Is there an “academic vocabulary”?’, Hyland and Tse questioned the widely held assumption that ‘a single inventory can represent the vocabulary of academic discourse and so be valuable to all students irrespective of their fi eld of study’ (Hyland and Tse, 2007: 238). They made use of Coxhead’s (2000) Academic Word List and showed that the coverage of AWL items in a corpus of 3.3 million words from a range of academic disciplines is not evenly distributed. The disciplines that make up the corpus are biology, physics and computer science (sciences sub-corpus); mechanical and electronic engineering (engineering sub-corpus), and sociology, business studies and applied linguistics (social sciences sub- corpus). Of the 570 AWL families, 534 (94%) have irregular distributions across the sciences, engineering, and social sciences sub-corpora, with, in many cases, a majority of the occurrences located in just one domain. Of these, 227 (40%) have at least 60 per cent of all occurrences concen-trated in just one sub-corpus. Overall, only 36 word families were found to be relatively evenly distributed across the sub-corpora. By contrast, 78 families were extremely infrequent in one sub-corpus, 63 in two sub-corpora and 6 in all three.

Hyland and Tse further argued that ‘all disciplines shape words for their own uses’ (ibid: 240) as demonstrated by their clear preferences for particular meanings and collocations. They gave the example of the word process which is far more likely to be encountered as a noun by science and engineering students than by social scientists. They also showed that the verb analyse tends to refer to ‘methods of determining the constituent parts or composition of a substance’ in engineering, while in the social sci-ences it often simply means ‘considering something carefully’ (ibid: 244). An investigation of a set of potential homographs in the AWL revealed a considerable amount of semantic variation across fi elds. Science and engi-neering students, for example, are very unlikely to come across the noun volume in the meaning of ‘a book or journal series’ unless they are reading book reviews. In addition, words may take on additional discipline-specifi c meanings as a result of their regular co-occurrence with other items. The noun strategy, for example, often appears in the multi-word unit marketing strategy in business, learning strategy in applied linguistics and coping strategy


in sociology. The authors concluded that ‘By considering context, cotext, and use, academic vocabulary becomes a chimera’ (ibid: 250).

These fi ndings pose a tremendous challenge to the growing number of students who enrol in interdisciplinary programmes and to English teachers who are regularly faced with mixed groups of students, most nota-bly in international EAP programmes (cf. Bhatia, 2002; Huckin, 2003; Eldridge, 2008). Two decades ago, in an article on second language teach-ing for academic achievement, Saville-Troike insisted that ‘vocabulary knowledge is the single most important area of second language compe-tence when learning content through that language is the dependent variable’ (1984: 199). EAP courses need to ensure that suffi cient attention is given to vocabulary development (cf. Sutarsyah et al, 1994: 37). That being the case, if academic vocabulary is a chimera, the problem is to deter-mine what words EAP tutors should teach a mixed groups of students.

Granger and Paquot (2009a) advocate a ‘happy medium’ approach which concurs with Hyland and Tse’s rejection of approaches of EAP as ‘an undifferentiated unitary mass’ (Hyland and Tse, 2007: 247) while also sub-scribing to Eldridge’s claim that ‘though one function of research is to unravel what distinguishes different fi elds and genres, another function is to fi nd similarities and generalities that will facilitate instruction in an imperfect world’ (Eldridge, 2008: 111). This balanced approach aims to reconcile research fi ndings and the reality of EAP teaching practice. An investigation of the verb analyse in a corpus of 1,701,351 words of business, linguistics and medicine articles has shown that it is possible to identify both the common core features of an academic word and its discipline-specifi c characteristics in terms of meaning, lexico-grammar and phraseological patterns (Granger and Paquot, 2009a). Wang and Nation commented that ‘learners should be encouraged to look for the central concept behind a variety of uses’ (2004: 310). As regards the verb analyse, this central concept can be defi ned as ‘to examine data using specifi c methods or tools in order to make sense of it’, with the ‘data’ and the ‘methods or tools’ varying across academic fi elds. It is only by invoking more general defi nitions of this type that EAP tutors will help L2 learners deal with the various uses of verbs that they may come across even within a single discipline. In linguistics, for example, analyse is also often used in the sense of carrying a statistical analysis, submitting data to computer-aided analyses or distinguishing the constituents of a word, phrase or sentence. The lexico-grammatical environment of the verb will help differentiate its ‘distinct (though not unconnected)’ (Hoey, 2005: 105) senses (see also Sinclair, 1987; 1991).


1.5. Summary and conclusion

There have been several studies that have investigated the vocabulary needed for academic study. Some of them have assumed that learners already knew the 2,000 most frequent words of English and looked at aca-demic texts to see what words not in the core vocabulary occur frequently across a range of academic disciplines. The Academic Word List (Coxhead, 2000) consists of 570 word families that are not in West’s (1953) General Service List but which have wide range and occur reasonably frequently in a 3,500,000 word corpus of academic texts. This list is very useful for students entering university, as well as being an excellent resource for preparing for the reading test in International English Certifi cates such as TOEFL and IELTS. It proves helpful in setting feasible learning goals and assessing vocabulary learning.

Defi ning academic vocabulary in opposition to core words, however, is of limited use when the role words play in academic discourse is examined. As shown by Martínez et al. (2009: 192), the verbs show, fi nd and report are not presented as academic words because they are part of the GSL. They perform, however, the same rhetorical function of reporting research as establish, conclude, and demonstrate and are often more frequent than these three AWL verbs in academic texts. These GSL verbs therefore also deserve careful attention in the academic writing classroom.

I agree with Hanciog lu and her colleagues that EAP practitioners should ‘avoid taking the GSL as any kind of “given” in the compilation of more specialized wordlists’ (Hancioglu et al. 2008: 464). I do not, however, subscribe to the idea according to which we ‘should seriously consider put-ting aside the idea of a distinct discrete-item Academic Word List’ (Hancioglu et al. 2008: 468). The construct of academic vocabulary remains a useful one which is, nevertheless, in need of a more precise defi nition (cf. Beheydt, 2005). That defi nition should rely on the work of researchers such as Martin (1976) and Meyer (1997) who focused on the nature and role of words that occur across subject-oriented texts, irrespective of the disci-plines. Martin (1976) discussed words that are useful instruments in the description of activities that characterize academic work, that is, research, analysis and evaluation. Meyer (1997) focused on words that provide a semantic-pragmatic skeleton for academic texts and identifi ed a number of lexical subsets that fulfi l important rhetorical and organizational functions in academic discourse (e.g. expressing modality, textual deixis, scholarly speech acts). More recently, Martínez et al. commented that academic words should serve ‘to build the rhetoric of a text, providing words useful


for the construction of the argument of science’ (2009: 193). All in all, it seems reasonable to argue that, for productive purposes, academic vocabulary would be more usefully defi ned as a set of options to refer

to those activities that characterize academic work, organize scientifi c

discourse and build the rhetoric of academic texts.The next step is to build a list of academic words according to this

defi nition and it remains to be seen whether this can be done automati-cally. Academic words in their ‘functional’ sense should be useful to biolo-gists, agronomists, physicists, historians, sociologists, lawyers, economists, linguists and computer scientists writing in higher education settings but not to novelists, poets or playwrights. Following Coxhead (2000), this lexi-cal set should therefore be reasonably frequent in a wide range of academic texts but relatively uncommon in other kinds of texts. There are, however, two major differences between this proposal and Coxhead’s work:

– The 2,000 most frequent words of English may be part of a list of aca-demic vocabulary;

– Words that are reasonably frequent in a wide range of academic texts but relatively uncommon in other kinds of texts will not be granted the sta-tus of academic words automatically. This frequency-based criterion is not regarded as a defi ning property of academic words but as a way of operationalizing a function-based defi nition of academic vocabulary.

In the next chapter, I will investigate whether academic words can be automatically extracted from corpora. To weed out those words that are not specifi c to academic texts, I will use a number of corpus linguistics tech-niques, and more particularly, keyword analysis.

Chapter 2

A data-driven approach to the selection of academic vocabulary

In this chapter, I describe the data-driven approach used to extract potential academic words from corpora. The term ‘potential academic words’ is used to refer to words that are reasonably frequent in a wide range of academic texts but relatively uncommon in other kinds of texts and which, as such, might be used to refer to those activities that characterize academic work, organize scientifi c discourse and build the rhetoric of academic texts, and so be granted the status of academic vocabulary. The method used to extract potential academic words is based on Rayson’s (2008) data-driven approach, which draws on both the ‘corpus-based’ and the ‘corpus-driven’ paradigms in corpus linguistics. Rayson (2008) identifi ed two general kinds of research question that can be investigated using a corpus-based paradigm. The majority of corpus-based studies tend to focus on a particular linguistic feature, possibly a word, lemma, multi-word expression or a grammatical construction. They examine ‘linguistic (lexical or grammatical associations of the feature), and non-linguistic aspects (distribution of the feature across different types of texts or speech)’ (Rayson 2008: 520). Other corpus-based studies invert this relationship and investigate the characteristics of whole texts or language varieties, by examining how certain linguistic features appear in a text (e.g. Biber, 1988). Common to all corpus-based studies is the prior selection of which linguis-tic features to study.

Rayson proposed a different approach: ‘decisions on which linguistic features are important or should be studied further are made on the basis of information extracted from the data itself; in other words, it is data-

driven’ (2008: 521). This model is set out in fi ve main steps:

1. Build: corpus design and compilation.2. Annotate: manual or automatic analysis of the corpus.


3. Retrieve: quantitative and qualitative analyses of the corpus.4. Question: devise a research question or model (iteration back to

Step 3).5. Interpret: interpretation of the results or confi rmation of the accuracy of

the model.

Studies that make use of the data-driven approach fi rst focus on whole texts (Step 3) and then refi ne the research question or suggest specifi c linguistic features to study in further detail (Step 4). Research questions emerge from iterative analyses of the corpus data.

The model bears some similarity to corpus-driven linguistics as pre-sented by Tognini-Bonelli (2001: 85), in which the corpus is the main informant. However, Rayson (2008) uses the term ‘data-driven’ to distin-guish this approach from the corpus-driven paradigm. Corpus-driven linguists question the ‘underlying assumptions behind many well estab-lished theoretical positions’ (Tognini-Bonelli, 2001: 48), stating that pre-corpus theories need to be re-examined in the light of evidence from corpora. They also have strong objections to corpus annotation (see McEnery et al., 2006: 9–10). Rayson’s (2008) data-driven method thus combines elements of both the corpus-based and the corpus-driven approaches. It relies on pre-existing part-of-speech tagsets but considers corpus data as ‘the starting point of a path-fi nding expedition that will allow linguists to uncover new grounds, new categories and formulate new hypotheses on the basis of the patterns that were observed’ (De Cock, 2003: 197). The model is also testimony to the fact that the ‘distinction between the corpus-based vs. corpus-driven approaches to language studies is overstated. In particular the latter approach is best viewed as an idealized extreme’ (McEnery et al., 2006: 8).

Following Rayson’s (2008) data-driven approach, I fi rst detail the cor-pora used (Step 1) and the type of annotation adopted (Step 2). I then focus on the different steps undertaken to retrieve potential academic words. The keyword procedure is fi rst used to retrieve a set of words which are distinctive of academic writing. The advantages and disadvan-tages of a keyword list are discussed and the criteria of range and even-ness of distribution are proposed to refi ne the list of potential academic words (Steps 3 and 4). In the last part of the chapter, I give a description of the fi nal list of potential academic words, the Academic Keyword List, and investigate whether its constituents fi t my defi nition of academic vocabulary. This provides a check on the accuracy of the retrieval proce-dure (Step 5).

Selection of academic vocabulary 31

2.1. Corpora of academic writing

Corpus-based studies of vocabulary in academic discourse (e.g. Johansson, 1978; Coxhead, 2000; Mudraya, 2006) have principally considered book sections, journal articles and textbooks. Academic writing, however, includes other kinds of text than professionally edited articles and books, notably student essays. As Nesi et al. (2004: 440) comment, ‘novice writers do not (. . . ) begin by writing for publication, or for a readership of strangers. Their early attempts at academic writing are more likely to be assessed texts produced in the context of a course study.’ The automatic selection of potential academic words for this study was therefore made on the basis of an analysis of both professional and student writing.

The professional academic corpora used are the Micro-Concord Corpus Collection B (MC) and the Baby BNC Academic Corpus (B-BNC). The corpora contain about a million words of published academic prose each. The MC comprises 33 book sections and the B-BNC is made up of 30 book sections and extracts from scientifi c journals. Texts in the B-BNC were written by British scholars while the MC also includes texts written by American research-ers. As shown in Table 2.1, both corpora consist of fi ve sections of about 200,000 words each, corresponding to fi ve broad academic domains (e.g. arts, social science, science). This division into ‘knowledge domains’ (Hyland, 2009: 62–5) is particularly well suited to extracting words that are used by all members of the ‘academic discourse community’ (Swales, 1990).

Table 2.1 The corpora of professional academic writing

Corpus Variety of English Text type Number of words

MCArts

Belief and religionScience

Applied scienceSocial science

mainly British English

books 1,005,060180,496199,612219,596203,316202,302

B-BNCHumanities

Politics, education and lawSocial science

ScienceTechnology and engineering

British English books and periodicals

1,021,007262,476196,322132,678283,490146,041

TOTAL 2,026,067


For centuries the traditional dividing line in the history of academia has been between the natural sciences and technology (‘hard sciences’), and humanities and the social sciences (‘soft sciences’). It is across this dividing line that ‘we tend to see the clearest discoursal variation and rhetorical distinctiveness’ (Hyland, 2009: 63). There seem to be good reasons for taking knowledge domains as the point of departure for identifying poten-tial academic words. As shown in Table 2.2, two corpora were compiled from the MC and the B-BNC: a corpus of professional ‘soft science’ (ProfSS) and a corpus of professional ‘hard science’ (ProfHS).

Two corpora of student writing were also used: part of the Louvain Corpus of Native Speaker Essays (LOCNESS) and a selection of texts from the British Academic Written English (BAWE) Pilot Corpus. Together they constitute the Student Writing Corpus.

LOCNESS totals 323,304 words and consists of argumentative and literary essays written by British A-level students (60,209 words), British uni-versity students (95,695 words) and American university students (168,400 words) (see Granger, 1996a; 1998a for further details). The part used for this study consists of argumentative essays written by university students and totals 168,593 words. Argumentative essay titles include, among others, ‘The death penalty’, ‘Euthanasia’, ‘Fox hunting’, ‘The National Lottery’, ‘Nuclear power’, ‘Crime does not pay’ and ‘Money is the root of all evil’. The BAWE Pilot Corpus1 contains about one million words of profi cient assessed student writing, in the form of 500 assignments ranging from 1,000 to 5,000 words in length (Nesi et al., 2004). 27 per cent of the contributors

Table 2.2 The re-categorization of data from the professional corpus into knowledge domains

Corpus Number of words

ProfSSMC Arts

MC Belief and religionMC Social scienceBNC Humanities

BNC Politics, education and lawBNC Social science

1,173,886180,496199,612202,302262,476196,322132,678

ProfHSMC Science

MC Applied scienceBNC Science

BNC Technology and engineering

852,443219,596203,316283,490146,041


were not native speakers of English. Nesi et al. (2004: 444) comment that ‘the University of Warwick is a multicultural, multilingual environment, and in their departments students are assessed on merit, without regard for their language background’, and add that ‘all contributors are profi cient users of English, given that their assignments have been awarded high grades’. For the purpose of identifying potential academic words, I decided to only make use of assignments written by British students as Hinkel (2003) has shown that even English as a Second Language (ESL) students ‘continue to have a restricted repertoire of syntactic and lexical features common in the written academic genre’ (Hinkel, 2003: 1066).

Unlike the fi nal BAWE corpus2, disciplines are not equally represented in the pilot corpus and the majority of student assignments come from the humanities and social sciences. As shown in Table 2.3, the texts were grouped into four sub-corpora which represent a discipline or a set of disciplines. The ‘Language studies’ sub-corpus consists of essays produced for courses in English studies, French studies, Italian studies, theatre, and literature. Texts in business, law, politics, sociology and economics were grouped together as social sciences as there were not enough texts per discipline to build separate corpora. Essay topics in the BAWE pilot corpus are very diverse and seldom repeated (see Table 2.4 for examples).

The Student Writing Corpus is thus quite representative of university students’ writing in that it comprises different types of writing tasks (skills-based writing and content-based writing; argumentative and expository writing). It is, however, skewed towards humanities and social sciences. It could be argued that the Academic Keyword List might therefore not fully represent academic vocabulary used in the ‘hard sciences’. It will be seen in Section 2.3 that the procedure used to extract potential academic words largely overcomes this limitation.

Table 2.3 The corpora of student academic writing

Corpus Variety of English Text type Number of words

BAWELanguage studies

Social sciencesPsychology

History

British English assignments 845,344221,841163,300201,946258,257

LOCNESS mainly American English

argumentative essays

168,593

Student Writing Corpus 1,013,937


2.2. Corpus annotation

The Academic Word List (Coxhead, 2000) is a list of word forms that were manually classifi ed into 570 word families (cf. Section 1.1.2.). Most studies of vocabulary in the fi eld of English for specifi c purposes (ESP) are based on raw corpora (e.g. Coxhead and Hirsh, 2007; Martínez et al., 2009; Mudraya, 2006; Wang et al., 2008; Ward, 2009). None of them discuss issues arising from the format of the corpus. However, any corpus-based study that aims to identify a specifi c set of vocabulary items should consider the advantages and disadvantages of annotating corpora.

2.2.1. Issues in annotating corpora

As Leech put it, ‘corpora are useful only if we can extract knowledge or information from them. The fact is that to extract information from a corpus, we often have to begin by building information in’ (1997: 4). Corpus annotation refers to the practice of adding linguistic information to an electronic corpus of language data. Various levels of annotation can be distinguished, starting from the addition of lemma information to each word in the corpus. A lemma is used to group together infl ected forms of a word, such as the singular and the plural forms of a noun, or the different conjugated forms of a verb. A second type of annotation is the morphosyn-tactic level of annotation, which concerns the labelling of the part-of-speech (POS) or grammatical category of each word in the corpus. POS tagging is

Table 2.4 Examples of essay topics in the BAWE pilot corpus

Language studies[98 essays]

– Visual arts in Britain– Prince Arthur portrayed in books– Rise of aestheticism– Modes of writing essays

Social sciences[64 essays]

– Housing policy– Teachers as professionals– Would you agree that subordination was inscribed in the life of domestic

servant?

Psychology [103 essays]

– Clinical depression– Psychology as a science– Expressing attitude– Is attention merely a matter of selection?

History [136 essays]

– Absolutism in early modern Europe– Why did America dominate the world fi lm market by the 1920s?– Who was to blame for the Boxer rising?


the most popular kind of linguistic annotation applied to text. By providing information about the grammatical nature of a word, it makes it possible to extract information about its various meanings and uses. Thus, it distinguishes between left as the past tense or past participle of leave (‘I left early’), and left as a word meaning the opposite of right, either as an adjective (‘my left hand’), an adverb (‘turn left’) or a noun (‘on your left’). Other levels of annotation are syntactic annotation or parsing (the analysis of sentences into their constituents), semantic annotation (the labelling of semantic fi elds) and discourse tagging (the annotation of discourse rela-tions within the texts). For more information on the different levels of annotation, see McEnery et al. (2006: 33–43).

A number of criticisms have been directed at corpus annotation, notably by distinguished contributors to corpus-driven linguistics. One of the most widespread criticisms is that annotation refl ects, at least to a certain extent, some theoretical perspective. Although the sets of categories and features used in annotating a corpus are generally chosen to be as uncontroversial as possible, the interpretative nature of corpus annotation has been perceived as a way of imposing pre-existing models of language on corpus data (Tognini-Bonelli, 2001: 73–4). These models of language date from a ‘pre-corpus’ time and some of them derive from descriptions which ignore empirical evidence altogether (Sinclair, 2004a: 52). The argument, although valid, is certainly not strong enough to counterbalance all the advantages of corpus annotation, but it should be taken as a warning against the naive assumption that using annotating software is a neutral act.

Another argument against annotation is that it may introduce errors. While it is inevitable that annotation systems will sometimes get things wrong, the various levels of annotation distinguished above are performed with varying degrees of accuracy. POS tagging is a well-researched kind of linguistic annotation and taggers perform with very high levels of accuracy. However discourse annotation systems, for example, are more recent and still need to be substantially refi ned.

Although annotated data is often described as ‘enriched’ data (Leech and Smith, 1999; Aarts, 2002; Bowker and Pearson, 2002), annotation has also sometimes been criticized for resulting in a loss of information (Sinclair, 1992; 2004a). The argument can be summarized as follows:

It could be argued that in a tagged text no information is lost because the words of the text are still there and available, but the problem is that they are bypassed in the normal use of a tagged text. The actual loss of information takes place when, once the annotation of the corpus is completed and the tagsets are attached to the data, the linguist processes


the tags rather than the raw data. By doing this the linguist will easily lose sight of the contextual features associated with a certain item and will accept single, uni-functional items —tags — as the primary data. What is lost, therefore, is the ability to analyse the inherent variability of language which is realised in the very tight interconnection between lexical and grammatical patterns. This is the price paid for simplifi cation; a process that is so useful – but it is argued here that the interconnection between lexis and grammar is crucial in determining the meaning and function of a given unit: any processing that loses out on this is bound to lose out in accuracy. (Tognini-Bonelli 2001: 73–4)

The data-driven methodology adopted in this book aims to preserve the best of both worlds, by fi rst extracting potential academic words from annotated corpora and then returning to raw data to analyse their use in context.

Finally, it is worth stressing that this chapter does not attempt to meet a theoretical objective. Rather it is content with an applied aim. Even the linguists who have directed the most severe criticisms at annotated data acknowledge that the ‘good point of annotation lies in its value in applica-tions’ (Tognini-Bonelli, 2001: 73). The main objective of this chapter is to select nouns, verbs, adjectives, adverbs and other function words that are commonly used in academic texts. Part-of-speech tagged corpora will thus facilitate the extraction of specifi c word classes. Elsewhere (Paquot, 2007b), I have tested the extraction procedure described in this chapter on two corpus formats, (word form + morphosyntactic tag, and lemma + morpho-syntactic tag), and shown that using lemmatised corpora makes it possible to identify some 31 per cent more lexical verbs that are typical of academic texts than using unlemmatised corpora.. If lemmas are used, the different infl ectional forms of a verb (e.g. consist, consists, consisted, and consisting) are merged and so a better frequency distribution for the lemma across texts is obtained. Word forms of lexical items that have two alternative spellings (e.g. analyse /analyze; characterise/characterize; centre/center; behaviour/behav-ior) were lemmatized under the same headword (either the British variant or the most frequently used option).

2.2.2. The software

The analysis was carried out using Wmatrix, a web-based corpus processing environment which gives researchers access to several corpus annotation and retrieval tools developed at the University Centre for Computer Corpus Research on Language (UCREL) at Lancaster University. The tools


available in Wmatrix include the Constituent Likelihood Automatic Word-tagging System (CLAWS) and the UCREL Semantic Analysis System (USAS) (see Rayson, 2003).

The Constituent Likelihood Automatic Word-tagging System

A corpus uploaded to the Wmatrix environment is fi rst grammatically tagged with the Constituent Likelihood Automatic Word-tagging System (CLAWS) (Garside and Smith, 1997). The tagger makes use of a detailed set of 146 tags3 (CLAWS C7 tagset). It also uses two lexicons: (a) a lexicon of single words with all their possible parts of speech and associated lemmas; and (b) a multiword expression lexicon. Multiword expressions include adverbs such as a bit, all the same, and so forth, at least, and by and large; prepo-sitions such as as opposed to, because of, contrary to, in the light of, and in compari-son with; compounds such as tabula rasa, brand new, matter-of-fact and grown up; and conjunctions such as even though, as if, provided that, and so that.

Part-of-speech tagging is essentially a disambiguation task. Many words are part-of-speech homographs, i.e. they are spelt the same but belong to different word classes. A tagger needs to determine which part-of-speech is most probable, given the immediate syntactic and semantic con-text of a homograph. Although close to 90 per cent of English types4 can only be one part-of-speech (e.g. abound can only be a verb and kindness is always a noun), over 40 per cent of the running words (or tokens) in a cor-pus are morphosyntactically ambiguous (DeRose, 1988: 31). This is largely due to the ambiguity of a number of high-frequency words such as that, which can be a determiner (Do you remember that nice Mr. Hoskins who came to dinner?)5, a relative pronoun (The people that live next door), a conjunction (I can’t believe that he is only 17) or an adverb (I hadn’t realized the situation was that bad!). Another very common source of ambiguity in English is homog-raphy between verbs and nouns, e.g. use, issue, cause, abandon, craft, etc. (see Ide, 2005).

Most current part-of-speech taggers use an approach to disambiguation which is at least partly probabilistic: they rely on co-occurrence probabili-ties between neighbouring tags. Co-occurrence probabilities are often automatically derived by training the software on manually disambiguated texts. For example, given that x is a determiner, the probability that the item to its immediate right is a noun or an adjective can be calculated. Non-probabilistic or rule-based taggers have also been making a comeback with systems such as that proposed by Brill (1992). Typical rule-based taggers use context frame rules to assign tags to unknown or ambiguous words. An example of a context frame rule is ‘if an ambiguous or unknown


word is preceded by a determiner and followed by a noun, tag it as an adjective’. Voutilainen (1999) has surveyed the history of the different approaches to word class tagging.

CLAWS is a hybrid tagger, combining both probabilistic and rule-based approaches. This hybrid approach allows CLAWS to assign POS-tags with a very high degree of accuracy – 97–98 per cent for written texts (Rayson, 2003: 63). The tagger is commonly described as going through fi ve major stages (Garside, 1987):

1. A pre-editing or tokenisation phase: This stage prepares the text for the tagging process by segmenting it into words and sentence units, a task which is not trivial. A sentence is generally described as a string of words followed by a full stop. A full stop does not, however, always signal the end of a sentence (e.g. in fi gures (5.8 or 14.28), title nouns (Mr., Dr.), and other types of abbreviations (i.e., viz., fi g.). Similarly, a word is generally considered as an orthographic word, i.e. a string of letters surrounded by white spaces. However, words are not always separated by blanks (e.g. in contractions such as don’t, it’s, they’re).

2. An initial part-of-speech assignment: Once a text has been tokenised, the tagger assigns part-of-speech tags to all the word tokens in the text with-out considering the context. If a word is unambiguous, i.e. belongs to only one part-of-speech category or word class (e.g. boat, person, belong), it is assigned a single tag. If a word is ambiguous, that is, if it can belong to more than one word class (e.g. use, cause, fi re), it is assigned several tags listed in decreasing likelihood. Thus, fi re is fi rst tagged as a noun and then as a verb, because the probability of it being a noun is higher than that of it being a verb. If a particular word is not found in the tagger’s lexicon, it is assigned a tag based on various sets of rules, e.g. morpho-logical rules, for tagging unknown items. Thus, a word ending in *ness will be classifi ed as a noun; a word ending in *ly will be classifi ed as an adverb, etc.

3. A rule-based contextual part-of-speech assignment: This stage assigns a single ‘ditto-tag’ to two or more orthographic words which function as a single unit or multiword expression (e.g. as well as is tagged as a conjunc-tion, and in situ as an adverb (see below for more details on ditto-tags and their advantages)).

4. The probabilistic tag-disambiguation program: The task of the probabi-listic tag-disambiguation program is to inspect all the cases where a word has been assigned two or more tags and choose a preferred tag by considering the context in which the word appears and assessing the


probability of any particular sequence of tags. The probability of a tag sequence is typically a function of,

– the probability that one tag follows another; and – the probability of a word being assigned a particular tag from the list

of all its possible tags (Garside and Smith, 1997: 104).

If, for example, the word run has been assigned both a noun and a verb tag, it is less likely to be classifi ed as a verb if it appears in the vicinity of another verb, despite the fact that run is more often a verb than a noun.

5. Output: The output data can be presented in intermediate format (vertical output for manual post-editing) or fi nal format (horizontal and encoded in SGML). Table 2.5 shows a typical CLAWS vertical output: each line rep-resents a running word in the corpus and gives its POS-tag and lemma.

The intermediate format has the advantage of allowing researchers to select the information needed. I have written a Perl program which takes this intermediate format as its input and creates a corpus with lemmas followed by their POS-tags (Table 2.6).

The problem with the format shown in Table 2.6 is that the word forms are replaced by their lemmas, while the POS-tags are too specifi c for our purposes. Redundant information includes, for example, that on number given by the tags NN1 (singular common noun) or DD1 (singular

Table 2.5 An example of CLAWS vertical output

POS-tag Word form Lemma

ATJJNN1IOATNN1VVZTOVBIAT1NN1IIATNN1.

ThewholepointoftheplayseemstobeanattackontheChurch.

thewholepointofthe playseemtobeanattackonthechurchPUNC


determiner) and that on verbal forms given by the tags VVZ (-s form of a lexical verb) or VVG (-ing form of a lexical verb). As a result, frequency lists based on this format generate different frequencies for ‘example_NN1’ and ‘example_NN2’. POS-tags were therefore simplifi ed by a Perl program to match the level of specifi city of the lemmas. Table 2.7 shows the same sentence in Table 2.6 after simplifi cation of the POS-tags. The simplifi ca-tion routines are presented in Table 2.8.

Finally, each CLAWS7 tag can be modifi ed by the addition of a pair of digits to show that it occurs as part of a sequence of similar tags, represent-ing a group of graphemic words which, for grammatical purposes, are best treated as a single unit. The expression ahead of is an example of a sequence of two graphemic words treated as a single preposition. It receives the tags: ahead_II21 of_II22, where II stands for a general preposition. The fi rst of the two digits indicates the number of graphemic words in the sequence, and the second digit the position of each graphemic word within that sequence. Such ‘ditto tags’ are not included in the lexicon but the program assigns them via an algorithm which is applied after initial part-of-speech assignment and before disambiguation by looking for a range of multiword expressions included in a pre-established list.

Ditto tags are very useful as they make it possible to extract complex prep-ositions, and complex conjunctions as well as single words that are typical of academic discourse. However, the annotation format has to be slightly modifi ed to do this. Table 2.9 shows the CLAWS vertical output for the com-plex preposition in terms of. Each graphemic word of the complex preposi-tion is tagged and lemmatised independently. A word list based on CLAWS horizontal output would thus distinguish between the preposition in (in_II) and the preposition in used as the fi rst word of three-word sequences (such as in terms of ) (in_II31). It would not be able to retrieve the complex

Table 2.7 CLAWS horizontal output [lemma + simplifi ed POS tags]

the_AT whole_JJ point_NN of_IO the_AT play_NN seem_VV to_TO be_VB an_AT attack_NN on_II the_AT Church_NN ._PUNC

Table 2.6 CLAWS horizontal output [lemma + POS]

the_AT whole_JJ point_NN1 of_IO the_AT play_NN1 seem_VVZ to_TO be_VBI an_AT1 attack_NN1 on_II the_AT Church_NN1 ._PUNC

Where AT: article; JJ: adjective; NN1: singular common noun; IO: of (as preposition); VVZ: -s form of lexical verb; TO: infi nitive marker ‘to’; VBI: be, infi nitive; AT1: singular article; II: general preposition; PUNC: punctuation


Table 2.8 Simplifi cation of CLAWS POS-tags

Simplifi ed POS tags CLAWS7 POS tags

Singular vs. plural forms

MC (cardinal number) MC1, MC2

NN (common nouns) NN1, NN2

NNL (locative nouns, e.g. island, street) NNL1, NNL2

NNO (numeral nouns, e.g. hundred) NNO, NNO2

NNT (temporal nouns, e.g. day, week) NNT1, NNT2

NNU (units of measurement, e.g. inch) NNU1, NNU2

NP (proper nouns) NP1, NP2

NPD (weekday noun) NPD1, NPD2

NPM (month noun) NPM1, NPM2

Comparative and superlative forms

DA (after-determiners, e.g. little, much, few) DAR (more, less), DAT (most, fewest)

JJ (adjective) JJR, JJT

Verb forms

VB (be) VB0 (be, base form), VBDR (were), VBDZ (was), VBI (be, infi nitive), VBM (am), VBN (been), VBR (are), VBZ (is)

VD (do) VD0 (do, base form), VDD (did), VDG (doing), VDI (do, infi nitive), VDN (done), VDZ (does)

VH (have) VH0 (have, base form), VHD (had), VHG (having), VHI (have, infi nitive), VHN (had), VHZ (has)

VV (lexical verbs) VV0 (base form of lexical verb), VVD (past tense), VVG (-ing participle), VVGK (-ing participle catenative, e.g. be going to), VVI (infi nitive), VVN (past participle), VVNK (past participle catenative, e.g. be bound to), VVZ (-s form)

Table 2.9 CLAWS tagging of the complex preposition ‘in terms of’

POS-tag Word form Lemma

II31II32II33

intermsof

intermof


preposition itself. Another Perl program was therefore used to replace any sequence of words with ditto tags (e.g. in_II31 terms_II32 of_II33) by the component words, separated by a hyphen and followed by their POS-tag (e.g. in-terms-of II).

The UCREL Semantic Analysis System

A second layer of annotation was applied by the UCREL Semantic Analysis System (USAS). This tool assigns tags representing the general semantic fi eld of words from a lexicon of single words and multiword expressions. A semantic fi eld is a theoretical construct which groups together ‘words that are related by virtue of their being connected – at some level of gener-ality – with the same mental concept’ (Wilson and Thomas, 1997: 54). This includes not only synonyms and antonyms of a word but also its hypernyms and hyponyms, and any other words that are linked in other ways with the concept concerned. For example, the category ‘language and commu-nication’ (Q) includes words such as answer, reply, response, question, query, statement, message, feedback, anecdote, explain, and explanation.

The USAS tagset includes 21 major semantic fi elds (see Table 2.10), which, in turn, expand into 232 categories (see Archer et al., 2002). Letters

Table 2.10 Semantic fi elds of the UCREL Semantic Analysis System

A General and abstract termsB The body and the individualC Arts and craftsE Emotional actions, states and processesF Food and farmingG Government and publicH Architecture, house and the homeI Money and commerce in industryK Entertainment, sports and gamesL Life and living thingsM Movement, location, travel and transportN Numbers and measurementO Substances, materials, objects and equipmentP Education in generalQ Language and communicationS Social actions, states and processesT Time W World and environmentX Psychological actions, states and processesY Science and technologyZ Names and grammar


are used to denote the major semantic fi elds while numbers indicate fi eld subdivisions. For example, the semantic tag A2.2 represents a word in the category ‘general and abstract words’ (A), the subcategory ‘affect’ (A2) and more precisely the sub-subcategory ‘cause / connected’ (A2.2). The seman-tic annotation does not apply to proper names and closed classes of words such as prepositions, conjunctions and pronouns. These categories are all marked with a Z-tag.

Like part-of-speech tagging, semantic tagging can be subdivided broadly into a tag assignment phase and a tag disambiguation phase. First, a set of potential semantic tags are attached to each lexical unit. The next stage consists of selecting the contextually appropriate semantic tag from the set of potential tags provided by the tag assignment algorithm. The program makes use of a number of sources of information in the disambiguation phase, notably POS-tags, domain of discourse, and contextual rules (Rayson, 2003: 67–8). It assigns a semantic fi eld tag to every word in the text with about 92 per cent accuracy. Table 2.11 shows that in the sentence ‘This chapter deals with the approach of the criminal law to behaviour which causes or risks causing death’, the word chapter has been assigned the tags Q4.1 (‘language and communication – media – books’), S5 (‘social actions, states and pro-cesses – groups and affi liation’), S9 (‘social actions, states and processes – religion and the supernatural’) and T1.3. (‘time-period’). The program

Table 2.11 USAS vertical output

POS-tag Word form Semantic tag

DD1NN1VVZIWATNN1IOATJJNN1IINN1DDQVVZCCVVZVVGNN1.

Thischapterdealswiththeapproachofthe criminallawtobehaviourwhichcausesor riskscausingdeath.

M6 Z5 Z8Q4.1 T1.3 S9/S5 S5+A1.1.1 I2.2 I2.1 A9- K5.2 F3/I2.2Z5Z5X4.2 M1 E1 S1.1.1Z5Z5G2.1[i1.2.1 G2.1- A5.1-G2.1[i1.2.2 G2.1 S6+ Y1Z5S1.1.1 A1.1.1Z8 Z5A2.2Z5A15-A2.2L1-


ranked these semantic tags and chose Q4.1 as the semantic tag with the highest correctness probability. This is displayed in the fi nal output format (see Table 2.12).

The same occurrence of a word in a text may simultaneously signal more than one semantic fi eld. The word chapter in the sense of ‘an ecclesiastical assembly of priests or monks’ is a case in point. It belongs equally to the semantic fi elds of ‘groups and affi liation’ and ‘religion’. The two semantic tags are thus assigned in the form of a single tag S9/S5 (see Table 2.11).

In the USAS lexicon of multiword expressions, phrasal verbs (e.g. break out, take off), compounds (e.g. academic year, advisory committee, bank account), and idioms (e.g. at the drop of a hat, to bark up the wrong tree, by the skin of one’s teeth) are described as regular expressions or templates, i.e. sequences of words, parts of words and grammatical categories used to match similar patterns of text and extract them. Thus, the template ‘ma[kd]*_V* {JJ, D*, AT*} sense_NN1’ identifi es all occurrences of the verb make directly followed by an optional adjective (JJ), determiner (D*) or article (AT*) and the singular noun (NN1) sense. It thus retrieves all instances of the expression make sense and its variants make no sense, makes little sense, made more sense, etc. Multiword expressions are analysed as if they were single words, using ditto-tags similar to those used in part-of-speech tagging. For example, criminal law is tagged as: criminal G2.1[i1.2.1 law G2.1[i1.2.2 (see Table 2.12).

2.3. Automatic extraction of potential academic words

Coxhead (2000) made use of the Range corpus analysis program (Heatley & Nation, 1996) to select words that met three frequency-based criteria:

1. The word families included had to be outside the fi rst 2,000 most fre-quent words in English, as represented by West’s (1953) General Service List.

Table 2.12 USAS horizontal output

This_M6 chapter_Q4.1 deals_A1.1.1 with_Z5 the_Z5 approach_X4.2 of_Z5 the_Z5 criminal_G2.1[i1.2.1 law_G2.1[i1.2.2 to_Z5 behaviour_S1.1.1 which_Z8 causes_A2.2 or_Z5 risks_A15 causing_A2.2 death_L1- ._PUNC


2. A member of a word family had to occur in all 4 disciplines represented in the Academic Corpus, with a frequency of at least 10 occurrences in each sub-corpus and in 15 or more of the 28 subject areas.

3. Members of a word family had to occur at least 100 times in the corpus (cf. Section 1.1.2.).

If Criterion 1 had not been used, the resulting list would have included a large number of function words and other high-frequency words that tend to be frequent in the English language as a whole (cf. Section 1.1.1) but which are not necessarily the most representative lexical items in the Academic Corpus. On the other hand, applying Criterion 1 makes it impos-sible to identify high-frequency words that are particularly prominent in academic texts.

To address this limitation, the procedure described in this book is primar-ily based on keyness (Scott, 2001), a fully data-driven method that is often used in corpus linguistics to fi nd salient linguistic features in texts (e.g. Archer, 2009) and which does not require the use of a stop list to fi lter out function words. These words and other high-frequency words will only occur in a keyword list “if their usage is strikingly different from the norm established by the reference text” (Archer, 2009a: 3).

Two quantitative fi lters, namely range and evenness of distribution, are subsequently used to narrow down the resulting list of potential academic words (Figure 2.1).

1. Keyness

2. Range

3. Evenness of distribution

Potential academic words

Figure 2.1 A three-layered sieve to extract potential academic words


2.3.1. Keyness

Keyword analysis has been used in a variety of fi elds to extract distinctive words or keywords, e.g. business English words (Nelson, 2000), words typically used by men and women with cancer in interviews and online cancer support groups (Seale et al., 2006), and terminological items typical of specifi c sub-disciplines of English for information science and techno-logy (Curado Fuentes, 2001). As emphasized by Scott and Tribble (2006: 55–6), ‘keyness is a quality words may have in a given text or set of texts, suggesting that they are important, they refl ect what the text is really about, avoiding trivia and insignifi cant detail. What the text ‘boils down to’ is its keyness, once we have steamed off the verbiage, the adornment, the blah blah blah’.

The procedure to identify keywords of a particular corpus involves fi ve main stages (see Scott and Tribble, 2006: 58–60):

1. Frequency-sorted word lists are generated for a reference corpus and the research corpus.

2. A minimum frequency threshold is usually set at 2 or 3 occurrences in the research corpus. Thus, ‘for a word to be key, then it (a) must occur at least as frequently as the threshold level, and (b) be outstandingly fre-quent in terms of the reference corpus’ (Scott and Tribble, 2006: 59).

3. The two lists of word types and their frequencies are compared by means of a statistical test, usually the log-likelihood ratio.

4. Words that occur less frequently than the threshold in the research corpus, or are not signifi cantly more frequent in the research corpus than in the reference corpus, are fi ltered out.

5. The word list for the research corpus is reordered in terms of the keyness of each word type. Software tools usually list positive keywords, i.e. words that are statistically prominent in the research corpus, as well as negative keywords, i.e. words that have strikingly low frequency in the research corpus in comparison to the reference corpus.

For the purposes of this research, the two corpora of professional writing and the corpus of student academic writing described in Section 2.1 were each compared with a large corpus of fi ction on the grounds that academic words would be particularly under-represented in this literary genre. Thus, the reference corpus was not chosen to represent all the varieties of the language6 but to serve as a ‘strongly contrasting reference corpus’ (Tribble, 2001: 396). The K (general fi ction), L (mystery and detective fi ction),


M (science fi ction), N (adventure and western fi ction) and P (romance and love story) categories of the LOB (Lancaster-Oslo/Bergen) corpus, the FLOB (Freiburg Lancaster-Oslo/Bergen) corpus, the BROWN corpus and the FROWN (Freiburg-Brown) corpus7 were combined with the Baby BNC fi ction corpus (Table 2.13) to form the reference corpus for this study.

Keyness values were calculated with the Keyness module of WordSmith Tools 4 (Scott, 2004). The signifi cance of the log-likelihood test was set at 0.01 with a critical value of 15.13 (see Rayson et al., 2004), which means that there is less than 1 per cent danger of mistakenly claiming a signifi cant difference in frequency. Keywords were extracted for the ProfSS corpus, the ProfHS corpus and the Student Writing Corpus, in lemma + POS-tag format. Table 2.14 gives the number of positive and negative keywords for each corpus.

Positive keywords are more numerous than negative keywords for each aca-demic corpus. This can be explained by the large amount of specialized vocabulary present in academic texts, e.g. formula, cell and species in biology (hard science), law, offence and policy in law (soft science) and theory, factor and participant in student writing. However, not all of the keywords meet the defi -nition of academic vocabulary in Section 1.5. The keyword procedure selects all words that occur with unusual frequency in a given text/corpus, com-pared to a reference corpus. The resulting list is therefore likely to include technical words that do not occur in all types of academic texts, simply because they are under-represented in fi ction writing (e.g. bacterium, methane, DNA, penicillin, chromosome, enzyme, jurisdiction, rape, archbishop, martyr, etc.).

Table 2.13 The fi ction corpus

Corpora Number of words

LOB (categories K, L, M, N, P)FLOB (categories K, L, M, N, P)BROWN (categories K, L, M, N, P)FROWN (categories K, L, M, N, P)

946,337

Baby BNC fi ction 999,688

TOTAL 1,946,025

Table 2.14 Number of keywords

Corpus Positive keywords Negative keywords

ProfHS corpus 4,322 837ProfSS corpus 4,656 1,201Student Writing corpus 4,492 956


The keyword procedure relies on the conception of a corpus as one big text rather than as a collection of smaller texts. Statistical measures such as the log-likelihood ratio are computed on the basis of absolute frequencies and cannot account for the fact that ‘corpora are inherently variable inter-nally’ (Gries, 2007: 110). As a consequence, the procedure cannot distin-guish between ‘global’ and ‘local’ keywords (Katz, 1996). Global keywords are dispersed more or less evenly through the corpus while local keywords appear repeatedly in some parts of the corpus only, a phenomenon which Katz (1996: 19) has referred to as ‘burstiness’8. For example, in a keyword analysis of gay male vs. lesbian erotic narratives, Baker shows that wuz (used as a non-standard spelling of was) appears to be a keyword of gay male erotic narratives, when in fact its use is restricted to one single text, ‘which suggests that this word is key because of a single author’s use of a word in a specifi c case, rather than being something that indicates a general differ-ence in language use’ (2004: 350). In other words, the keyword status of wuz is more a function of the sampling decision to include one particular narrative in the corpus than evidence of the distinctiveness of the word in gay male erotic narratives (see also Oakes and Farrow, 2007: 91).

As a fi rst step to overcome this inherent limitation of the keyword procedure, I wrote a Perl program which automatically compares keywords for several corpora and creates a list of positive keywords that are shared in the ProfHS corpus, the ProfSS corpus and the Student Writing corpus (cf. Scott’s (1997) notion of ‘key keywords’). Although the resulting number of keywords fell by more than 60 per cent, 2,048 shared keywords were still identifi ed. The criteria of range and evenness of distribution were subsequently used to refi ne the list of potential academic words still further.

2.3.2. Range

Range (i.e. the number of texts in which a word appears) is used to deter-mine whether a word appears to be a potential academic keyword because it occurs in most academic disciplines or because of a very high usage in a limited subset of texts. It is calculated on the basis of the 15 sub-corpora described in Section 2.1 with the WordList option of WordSmith Tools (Scott, 2004). This tool can take several corpus fi les as input and range comes automatically with any word list it produces. Figure 2.2 shows that there is a column headed ‘Texts’ which shows the number of texts each word occurred in. The words ability, able and about, for example, are shown to appear in all 15 sub-corpora, that is, in 100 per cent of the corpora


analysed. For the purposes of this study, only words appearing in all 15 academic sub-corpora were retained as potential academic words.

Used alone, range has an important limitation in that it gives no informa-tion on the frequency of a word in each sub-corpus. Thus, the criterion of range excludes the words sector, paradigm and variance as they only appear in 11 sub-corpora but includes both the word example, which we intuitively regard as an academic word, and the word law, the meaning of which is more discipline or topic-dependent (e.g. canon law, criminal law, the law of gravity). The frequencies of these two words in each sub-corpus are shown in Figure 2.3 and accurately refl ect the difference between them. The fre-quency of the word example ranges from 26 to 226 in the 15 sub-corpora, while that of the word law varies between 11 and 812. The large variation in the range of law can be explained by the peak frequency of occurrence of the noun in the professional soft science sub-corpora.

Figure 2.2 WordSmith Tools – WordList option


2.3.3. Evenness of distribution

Differences in range can be highlighted by a measure of the evenness of the distribution of words in a corpus. This is the last criterion I applied to restrict the list of potential academic words. The evenness of the distribu-tion or dispersion of a word is ‘a statistical coeffi cient of how evenly distrib-uted a word is across successive sectors of the corpus’ (Rayson, 2003: 93). This measure takes into account ‘not only the presence or absence of a word in each subsection of the corpus, but the exact number of times it appears’ (Oakes and Farrow 2007: 91). A number of studies have used a measure of dispersion to defi ne a core lexicon on the basis that ‘if a word is commonly used in a language, it will appear in different parts of the corpus. And if the word is used commonly enough, it will be well-distributed’ (Zhang et al., 2004).

One such measure is Juilland’s D statistical coeffi cient. Juilland’s D was fi rst used in the Frequency Dictionary of Spanish Words (Juilland and Rodriguez, 1964) and is calculated as D = 1 – V / √n-1 where n is the number of sectors (i.e. the number of sub-corpora or texts) in the corpus. The variation coef-fi cient V is given by V = s / x where x is the mean sub-frequency of the word in the corpus and s is the standard deviation of these sub-frequencies. Its values range from 0 (most uneven distribution possible) to 1 (perfectly even distribution across the sectors of the corpus) (see Oakes (1998: 189–92) and Gries (2008) for more information on dispersion measures).

Juilland’s Ds were calculated for each word using the output list from WordSmith Tools Detailed Consistency Analysis. Figure 2.4 gives an example of a

900

LAWEXAMPLE

800700600500400300200100

MC A

RTS

MC B

EL

BNC HUM

BNC POL

BNC SOC

MC S

C

MC A

PPSC

BNC SC

BNC TECH

BAWE A

RTS

BAWE H

ISTORY

BAWE P

SYCHO

BAWE S

OC

LOCNESS

MC S

OC

0

Figure 2.3 Distribution of the words example and law in the 15 sub-corpora

Selection of academ

ic vocabulary 51

Figure 2.4 WordSmith Tools Detailed Consistency Analysis


detailed consistency analysis: the second column (Total) gives the total frequency of each word in the whole corpus, the third column (Texts) gives the range of each word and the following columns show its frequencies in each sub-corpus. These frequencies were copied into an Excel fi le and normalized per 100,000 words as the 15 sub-corpora are of different sizes. The measures necessary to calculate Juilland’s D values (i.e. the variation coeffi cient, the mean sub-frequency and the standard deviation) were computed in Excel and Juilland’s D values were then calculated for each word.9

For a word to be selected as a potential academic word, its Juilland’s D value had to be higher than 0.8. The noun example was thus identifi ed as a potential academic word as its dispersion value was 0.83, whereas the noun law, with a Juilland’s D value of 0.69, was not selected. Dispersion values make it possible to avoid the mistaken conclusion that these two words behave similarly in academic writing, and confi rm that only example is of widespread and general use in this genre, while the noun law is over-represented in the professional soft science corpus, and more specifi -cally in the social science sub-corpus. Evenness of distribution is the only criterion used that could perhaps favour keywords that are more promi-nent in the different parts of the ProfSS corpus and the Student Writing corpus. However, a relatively high minimum threshold of 0.8 reduces the possibility of giving too much weight to words that would be particularly frequent in the ‘soft science’ sub-corpora but much less common in the ‘hard science’ sub-corpora.

The resulting list of 599 potential academic words includes nouns such as conclusion, difference, extent, signifi cance, and consequence; verbs such as prove, appear, provide, discuss, show, result and illustrate; adjectives such as signifi cant, effective, similar and likely and adverbs such as particularly, conversely, highly and above. Examples of words that have D values lower than 0.8 and were therefore not selected include the nouns health, employment, personality, and treatment; and the verbs label, perceive and isolate. At times, the cut-off point of 0.8 appears to be too restrictive and words that would intuitively be considered as academic words are excluded. Some words have skewed Juilland’s D values because of their polysemy. The noun solution is a case in point, as it has both a general meaning (‘a way of solving a problem’) and a technical meaning (‘a liquid in which a solid or gas has been mixed’) with different frequencies and distributional behaviours. Its general meaning is found in all academic sub-corpora while its technical meaning is restricted to scientifi c writing and accounts for its much higher frequency in the two professional scientifi c sub-corpora (MC-SC and BNC-SC in Figure 2.5).


These two peak frequencies of occurrence are responsible for the relatively low D value (54.6) of the noun solution.

2.3.4. Broadening the scope of well-represented semantic categories

More generally, the 15 sub-corpora are relatively small and the frequencies of occurrence of words may be skewed by a particular topic or author’s pre-ferred turn of phrase. I therefore made use of a semi-automatic procedure to identify words that did not pass the dispersion criterion but were seman-tically related to the 599 potential academic words.

Section 2.2.2 discussed how a text uploaded to the web-based environ-ment Wmatrix is morphosyntactically and semantically tagged. The seman-tic analysis was conducted with the UCREL System which classifi es words and multiword units into 21 major semantic categories. Table 2.15 shows the distribution of the 599 potential academic words across these semantic classes. Some words were automatically classifi ed into more than one cate-gory but the fi gures given are based on the semantic tag most frequently attributed to each word. It is notable that 87 per cent of the 599 potential academic words fall into just six of the categories; in particular, the category ‘general and abstract terms’ includes almost half the potential academic words. Examples include the nouns activity, circumstance, and limitation as well as the verbs perform and cause, the adjectives detailed and particular and the adverbs similarly and conversely. The category ‘numbers and measure-ment’ accounts for more than 10 per cent of the potential academic words and includes nouns (e.g. degree, measure, amount, extent), adjectives (e.g. high,

300

250

200

150

100

50

MC A

RTS

MC B

EL

MC S

OC

BNC HUM

BNC POLI

BNC SOC

MC S

CIEN

MC A

PL SC

BNC SCI

BNC TECH

BAWE A

RTS

BAWE H

IST

BAWE P

SCHY

BAWE S

C

LOCNESS

0

Figure 2.5 Distribution of the noun ‘solution’


large, wide), verbs (e.g. extend, increase, reduce), adverbs (e.g. frequently, subsequently, also) and prepositions (e.g. in addition to). The categories ‘psychological actions, states and processes’ (e.g. assumption, analyse, inter-pretation, conclusion, attempt), ‘names and grammar’ (mainly consisting of connective devices such as conjunctions (or, whether), prepositions (such as, according to, since, during) and adverbs (moreover, thus, therefore)), ‘social actions, states and processes’ (e.g. social, encourage, facilitate, impose), and ‘language and communication’ (e.g. argue, claim, defi ne, suggest) represent 9.2 per cent, 8.3 per cent, 7.7 per cent and 5.7 per cent of the potential academic words respectively.

On this basis, 331 keywords that did not have Juilland’s D values higher than 0.8 but which formed part of one of the six semantic categories described above were added to the list of potential academic words. Many of the words that were retrieved by this additional criterion are morphologi-cally related to words that had already been automatically selected. For example, the noun analysis, which is morphologically related to the poten-tial academic verb analyse, was retrieved by the semantic criterion although

Table 2.15 Automatic semantic analysis of potential academic words

Semantic categories Number of words Percentage of words

A. General and abstract terms 267 44.6B. The body and the individual 2 0.3C. Arts and crafts 2 0.3E. Emotion 4 0.7F. Food and farming 0 0.0G. Government and public 4 0.7H. Architecture, house and the home 2 0.3I. Money and commerce in industry 7 1.2K. Entertainment, sports and games 0 0.0L. Live and living things 0 0.0M. Movement, location, travel and transport 12 2.0N. Numbers and measurement 74 12.4O. Substances, materials, objects and equipment 7 1.2P. Education in general 4 0.7Q. Language and communication 34 5.7S. Social actions, states and processes 47 7.7T. Time 26 4.3W. World and environment 2 0.3X. Psychological actions, states and processes 55 9.2Y. Science and technology in general 2 0.3Z. Names and grammar 50 8.3

TOTAL 599 100


its Juilland’s D was below 0.8. However this is not an argument for using word families instead of lemmas. The criteria of minimum frequency and range still apply: the noun analysis was retrieved only because it is very frequent in academic prose and appears in a wide range of academic texts. Other morphologically related words such as analyst or analysable were still excluded from the list.

2.4. The Academic Keyword List

The (semi-)automatic extraction procedure described in Section 2.3 identifi ed 930 potential academic words on the basis of four criteria. First, the words had to be keywords in professional (both ‘hard’ and ‘soft’ disciplines) and student academic writing. Second, they had to be charac-terized by wide range, i.e. to appear in all 15 sub-corpora representing different academic disciplines. Third, they had to be well-distributed across the corpora and have a Juilland’s D value higher than 0.8. Keywords that did not match this last criterion but belonged to one of the six best represented semantic categories – ‘general and abstract’, ‘numbers and measurement’, ‘psychological actions, states and processes’, ‘names and grammar’, ‘social actions, states and processes’ and ‘language and communication’ – were also included. The resulting list of potential academic words has been named the Academic Keyword List (AKL) to emphasize the fact that it is the output of a data-driven set of criteria, the fi rst of which is keyness, and not a list of aca-demic vocabulary in its functional sense. Table 2.16 presents a breakdown of the AKL by grammatical category. The complete list is given in Table 2.17.

Nouns make up 38.17 per cent of all potential academic words. This is consistent with Biber et al.’s (1999) fi nding that nouns are particularly frequent in academic prose. A large proportion of the nouns in the list are

Table 2.16 Distribution of grammatical categories in the Academic Keyword List

Number Percentage

Nouns 355 38.17Verbs 233 25.05Adjectives 180 19.35Adverbs 87 9.35Others 75 8.06

Total 930 100

56 A

cademic Vocabulary in L

earner Writing

Table 2.17 The Academic Keyword List

355 nouns

ability, absence, account, achievement, act, action, activity, addition, adoption, adult, advance, advantage, advice, age, aim, alternative, amount, analogy, analysis, application, approach, argument, aspect, assertion, assessment, assistance, association, assumption, attempt, attention, attitude, author, awareness, balance, basis, behaviour, being, belief, benefi t, bias, birth, capacity, case, category, cause, centre, challenge, change, character, characteristic, choice, circumstance, class, classifi cation, code, colleague, combination, commitment, committee, communication, community, comparison, complexity, compromise, concentration, concept, conception, concern, conclusion, condition, conduct, confl ict, consensus, consequence, consideration, constraint, construction, content, contradiction, contrast, contribution, control, convention, correlation, country, creation, crisis, criterion, criticism, culture, damage, data, debate, decision, decline, defence, defi nition, degree, demand, description, destruction, determination, development, difference, diffi culty, dilemma, dimension, disadvantage, discovery, discrimination, discussion, distinction, diversity, division, doctrine, effect, effectiveness, element, emphasis, environment, error, essence, establishment, evaluation, event, evidence, evolution, examination, example, exception, exclusion, existence, expansion, experience, experiment, explanation, exposure, extent, extreme, fact, factor, failure, feature, female, fi gure, fi nding, force, form, formation, function, future, gain, group, growth, guidance, guideline, hypothesis, idea, identity, impact, implication, importance, improvement, increase, indication, individual, infl uence, information, insight, instance, institution, integration, interaction, interest, interpretation, intervention, introduction, investigation, isolation, issue, kind, knowledge, lack, learning, level, likelihood, limit, limitation, link, list, literature, logic, loss, maintenance, majority, male, manipulation, mankind, material, means, measure, medium, member, method, minority, mode, model, motivation, movement, need, network, norm, notion, number, observation, observer, occurrence, operation, opportunity, option, organisation, outcome, output, parallel, parent, part, participant, past, pattern, percentage, perception, period, person, personality, perspective, phenomenon, point, policy, population, position, possibility, potential, practice, presence, pressure, problem, procedure, process, production, programme, progress, property, proportion, proposition, protection, provision, publication, purpose, quality, question, range, rate, reader, reality, reason, reasoning, recognition, reduction, reference, relation, relationship, relevance, report, representative, reproduction, requirement, research, resistance, resolution, resource, respect, restriction, result, review, rise, risk, role, rule, sample, scale, scheme, scope, search, section, selection, sense, separation, series, service, set, sex, shift, signifi cance, similarity, situation, skill, society, solution, source, space, spread, standard, statistics, stimulus, strategy, stress, structure, subject, success, summary, support, survey, system, target, task, team, technique, tendency, tension, term, theme, theory, tolerance, topic, tradition, transition, trend, type, uncertainty, understanding, unit, use, validity, value, variation, variety, version, view, viewpoint, volume, whole, work, world

Selection of academ

ic vocabulary 57

233 verbs

accept, account (for), achieve, acquire, act, adapt, adopt, advance, advocate, affect, aid, aim, allocate, allow, alter, analyse, appear, apply, argue, arise, assert, assess, assign, associate, assist, assume, attain, attempt, attend, attribute, avoid, base, be, become, benefi t, can, cause, characterise, choose, cite, claim, clarify, classify, coincide, combine, compare, compete, comprise, concentrate, concern, conclude, conduct, confi ne, conform, connect, consider, consist, constitute, construct, contain, contrast, contribute, control, convert, correspond, create, damage, deal, decline, defi ne, demonstrate, depend, derive, describe, design, destroy, determine, develop, differ, differentiate, diminish, direct, discuss, display, distinguish, divide, dominate, effect, eliminate, emerge, emphasize, employ, enable, encounter, encourage, enhance, ensure, establish, evaluate, evolve, examine, exceed, exclude, exemplify, exist, expand, experience, explain, expose, express, extend, facilitate, fail, favour, fi nance, focus, follow, form, formulate, function, gain, generate, govern, highlight, identify, illustrate, imply, impose, improve, include, incorporate, increase, indicate, induce, infl uence, initiate, integrate, interpret, introduce, investigate, involve, isolate, label, lack, lead, limit, link, locate, maintain, may, measure, neglect, note, obtain, occur, operate, outline, overcome, participate, perceive, perform, permit, pose, possess, precede, predict, present, preserve, prevent, produce, promote, propose, prove, provide, publish, pursue, quote, receive, record, reduce, refer, refl ect, regard, regulate, reinforce, reject, relate, rely, remain, remove, render, replace, report, represent, reproduce, require, resolve, respond, restrict, result, retain, reveal, seek, select, separate, should, show, solve, specify, state, stimulate, strengthen, stress, study, submit, suffer, suggest, summarise, supply, support, sustain, tackle, tend, term, transform, treat, undermine, undertake, use, vary, view, write, yield

180 adjectives

absolute, abstract, acceptable, accessible, active, actual, acute, additional, adequate, alternative, apparent, applicable, appropriate, arbitrary, available, average, basic, central, certain, clear, common, competitive, complete, complex, comprehensive, considerable, consistent, conventional, correct, critical, crucial, dependent, detailed, different, diffi cult, distinct, dominant, early, effective, equal, equivalent, essential, evident, excessive, experimental, explicit, extensive, extreme, far, favourable, fi nal, fi xed, following, formal, frequent, fundamental, future, general, great, high, human, ideal, identical, immediate, important, inadequate, incomplete, independent, indirect, individual, inferior, infl uential, inherent, initial, interesting, internal, large, late, leading, likely, limited, local, logical, main, major, male, maximum, mental, minimal, minor, misleading, modern, mutual, natural, necessary, negative, new, normal, obvious, original, other, overall, parallel, partial, particular, passive, past, permanent, physical, positive, possible, potential, practical, present, previous, primary, prime, principal, productive, profound, progressive, prominent, psychological, radical, random, rapid, rational, real, realistic, recent, related, relative, relevant, representative, responsible, restricted, scientifi c, secondary, selective, separate, severe, sexual, signifi cant, similar, simple, single, so-called, social, special, specifi c, stable, standard, strict, subsequent, substantial, successful, successive, suffi cient, suitable, surprising, symbolic, systematic, theoretical, total, traditional, true, typical, unique, unlike, unlikely, unsuccessful, useful, valid, valuable, varied, various, visual, vital, wide, widespread

(Continued)

58 A


earner Writing

87 adverbs

above, accordingly, accurately, adequately, also, approximately, at best, basically, clearly, closely, commonly, consequently, considerably, conversely, correctly, directly, effectively, e.g., either, equally, especially, essentially, explicitly, extremely, fairly, far, for example, for instance, frequently, fully, further, generally, greatly, hence, highly, however, increasingly, indeed, independently, indirectly, inevitably, initially, in general, in particular, largely, less, mainly, more, moreover, most, namely, necessarily, normally, notably, often, only, originally, over, partially, particularly, potentially, previously, primarily, purely, readily, recently, relatively, secondly, signifi cantly, similarly, simply, socially, solely somewhat, specifi cally, strongly, subsequently, successfully, thereby, therefore, thus, traditionally, typically, ultimately, virtually, wholly, widely

75 others

according to, although, an, as, as opposed to, as to, as well as, because, because of, between, both, by, contrary to, depending on, despite, due to, during, each, even though, fewer, fi rst, former, from, for, given that, in, in addition to, in common with, in favour of, in relation to, in response to, in terms of, in that, in the light of, including, its, itself, latter, less, little, many, most, of, or, other than, per, prior to, provided, rather than, same, second, several, since, some, subject to, such, such as, than, that, the, their, themselves, these, third, this, those, to, unlike, upon, versus, whereas, whether, whether or not, which, within

Table 2.17 (Cont’d)


abstract terms which belong to Francis’s (1994) category of metalinguistic labels: illocutionary nouns (e.g. argument, question), language-activity nouns (e.g. comparison, contrast, defi nition, and description), mental process nouns (e.g. analysis, concept, hypothesis, theory, and view) and text nouns (e.g. section, term). Verbs account for 25 per cent of the list and include verbs that Hinkel (2004) classifi ed as activity verbs (e.g. deal, use, show, provide), reporting verbs (e.g. argue, discuss, emphasize, explain, respond), mental verbs (e.g. assess, examine, interpret, note), linking verbs (e.g. appear, be, become, prove) and logico-semantic relationship verbs (e.g. arise, cause, contrast, differ, follow, imply, illustrate, include).

Although usually disregarded in academic textbooks and teaching mate-rials, adjectives represent 19.35 per cent of the potential academic words in this list. The syntactic function of adjectives is to modify nouns and noun phrases. So it is logical that, if nouns are frequent in academic writing, numerous adjectives will be used to qualify them. The majority of adjectives express value judgements, either positive or negative (e.g. adequate, appropriate, clear, inadequate, incomplete, interesting, prime, useful), possibility (e.g. possible, potential, likely, unlikely) and logico-semantic relationships (e.g. additional, alternative, different, equivalent, fi nal, following, parallel, similar). As Conrad (1999) pointed out, the category of potential academic adverbs (9.35%) consists essentially of linking (e.g. consequently, conversely, hence, however, thus) and evaluative (e.g. adequately, correctly, effectively, highly, increasingly, inevitably, signifi cantly) words. The category ‘other’ includes prepositions, conjunctions, pronouns, articles, determiners and ordinal numbers. Inspection of this list illustrates the value of CLAWS ditto tags (Section 2.2.2): some 40 per cent of the potential academic prepositions are complex (e.g. such as, according to, in terms of, because of). There are also complex conjunctions such as whether or not and given that. Pronouns, arti-cles and determiners are not fully lemmatised by CLAWS. For example, this, these and those are not analysed as word forms of the same determiner and are categorized as separate lemmas.

In order to calculate the percentages of GSL and AWL words in the list of potential academic words, the Academic Keyword List was uploaded to the Web Vocab Profi le developed by Tom Cobb10. This web interface takes a text fi le as its input, analyses the vocabulary in the text, and classifi es the words into four main categories: (1) the fi rst 1,000 most frequent words of English; (2) the second 1,000 most frequent words of English; (3) words of the Academic Word List; and (4) other words. Before uploading the list of potential academic words, it was necessary to remove multiword units (such as the adverbs for example and for instance and the complex prepositions


in addition to, due to, prior to, in the light of, in favour of and because of) since Web Vocab Profi le cannot deal with them and would simply decompose multi-word units into their parts. The comparison is thus based on single words only. The results show that only 40 per cent of the potential academic words in my list also occur in the Academic Word List, while 57 per cent are among the 2,000 most frequent words of English as described in West’s (1953) General Service List (see Table 2.18). These results highlight the important role played by general service vocabulary in academic writing.

The Academic Keyword List is a list of potential academic words. As explained in Section 1.5, frequency-based criteria are not used as a defi ning property of academic words but as a way of operationalizing a function-based defi ni-tion of academic vocabulary. Table 2.17 shows that a high proportion of AKL words fi t the functional defi nition of academic vocabulary (e.g. evidence, approach, result, show, defi ne, measure, degree, extent, condition, experience, result), which provides the retrieval procedure with some evidence of face validity.

An important application of word frequency lists for course design is to uncover ‘functional and notional areas which might be important for the syllabus’ (Flowerdew, 1993: 237). The automatic semantic analysis has shown that a number of semantic sub-categories are particularly well- represented. For example, the semantic sub-category named ‘A2.2. Affect: cause – connected’ consists of the nouns basis, cause, consequence, correlation, effect, factor, impact, implication, infl uence, link, motivation, relation, reason, result and stimulus; the verbs associate, attribute, base, cause, combine, connect, depend, derive, effect, generate, induce; infl uence, lead, link, produce, relate, render, result and stimulate; the adjectives dependent, related and resulting; the adverbs conse-quently, hence, therefore, thereby, and thus; the prepositions depending on, due to, in relation to, in view of, on account of, with respect to, in respect of, in response to, because of, in the light of, in terms of and subject to; and the conjunctions because, given that, provided and since. Other well-represented semantic categories

Table 2.18 The distribution of AKL words in the GSL and the AWL

Lists % Examples

GSL 57% aim, argue, argument, because, compare, comparison, differ, difference, discuss, example, exception, explain, explanation, importance, include, increasingly, likelihood, namely, point, reason, result, therefore, typically

AWL 40% accurately, adequate, analysis, assess, comprise, conclude, conclusion, consequence, emphasize, hypothesis, inherent, method, proportion, relevance, scope, summary, survey, theory, validity, whereas

Other 3% assertion, correlation, criticism, exemplify, proposition, reference, tackle, versus, viewpoint


include ‘A1.1. General actions, making, etc’ (e.g. activity, circumstance, event, arise, perform), ‘A1.8. Inclusion / exclusion’ (e.g. content, scope, consist, exclude, include), ‘A4.1. General kinds, groups, examples’ (e.g. case, category, example, instance, kind, classify, illustrate), ‘A5. Evaluation’ (e.g. progress, enhance, improve, favourable, positive), ‘A6. Comparing’ (e.g. contrast, difference, similar, unlike, conversely), ‘A11.1. Importance’ (e.g. signifi cance, emphasize, fundamen-tal, major, primary); ‘N5. Quantities’ (e.g. amount, extent, fi gure, considerable, limited, widely, several), ‘Q2.2. Speech acts’ (e.g. account, defi nition, discussion, describe, quote, refer, suggest) and ‘X2.1. Thought and belief’ (e.g. view, assume, consider, formulate).

The fact that these AKL words are used to serve particular functions in academic prose has ‘to be corroborated by concordancing’ (Flowerdew, 1993: 237). Words such as the noun illustration, the verb illustrate and the preposition like are often employed as exemplifi ers but also have different uses. The verb illustrate, for example, also means ‘to put pictures in a book’, a sense that will not be useful to all scholars who are writing in higher edu-cation settings. Similarly, the Academic Keyword List includes several words that are relatively well distributed keywords in academic prose but should probably not be part of a list of academic vocabulary (e.g. the nouns coun-try, female, male, parent, sex and world). The Academic Keyword List is not a fi nal product and does not in itself ‘carry any guarantee of pedagogical relevance’ (Widdowson, 1991: 20–1). It still needs ‘pedagogic mediation’ (Widdowson, 2003) and is thus better conceived of as a ‘platform from which to launch corpus-based pedagogical enterprises’ (Swales, 2002: 151).


In Chapter 1, academic vocabulary was defi ned as ‘a set of options to refer to those activities that characterize academic work, organize scientifi c discourse and build the rhetoric of academic texts’. The main objective of Chapter 2 has been to operationalize this function-based defi nition using frequency-based criteria, and to retrieve potential academic words (i.e. words that are reasonably frequent in a wide range of academic texts but relatively uncommon in other kinds of texts and which, as such, might count as academic vocabulary) from several corpora of academic prose. The (semi-)automatic method used relies on Rayson’s (2008) data-driven model, which combines elements of both ‘corpus-based’ and ‘corpus-driven’ paradigms in corpus linguistics. The keyword procedure was fi rst used to retrieve a set of words which are distinctive of academic writing.


Second, the criteria of range and evenness of distribution were used as a sieve to refi ne the list of potential academic words. Third, a number of keywords that had failed the dispersion test were selected on the basis of a semantic criterion: they belonged to one of the six semantic categories (‘general and abstract’, ‘numbers and measurement’, ‘psychological actions, states and processes’, ‘names and grammar’, ‘social actions, states and processes’ and ‘language and communication’) that included most potential academic words. The method provides a good illustration of the usefulness of annotation for the development of practical applications such as the Academic Keyword List.

This methodology has limitations. First, the corpora used are relatively small and contain a limited number of text types. However, at present no corpus exists that represents all the varieties of academic discourse. As Krishnamurthy and Kosem (2007: 370) comment, ‘the one thing that EAP seems to lack is a corpus that includes all levels of data – from pre- sessional students’ writing and speech to academic lectures, PhD theses, and published research articles and books. Such a corpus would need to include as many disciplines as possible, with suffi cient detailed categoriza-tion to enable the users (teachers or students) to select a customized set of corpus texts appropriate for their needs’. Second, a limitation inherent in the keyness approach is the use of a reference corpus. A reference corpus is itself characterized by a set of distinctive linguistic features, some of which may be shared with the corpus under study. There is thus a strong case for using ‘strongly contrasting reference corpora’ (cf. Tribble, 2001: 396). However, it is likely that a few potential academic words passed unnoticed because they are also used in fi ction, irrespective of differences in meaning or function. Third, the results are dependent on a number of arbitrary cut-off points: the probability threshold under which log-likelihood ratio values are not considered signifi cant, the minimum number of texts in which a keyword must appear, and the minimum coeffi cient of dispersion (see Oakes and Farrow, 2007: 92; Paquot and Bestgen, 2009). The Academic Keyword List might have looked quite different if other cut-off points had been adopted.

Despite these limitations, a high proportion of the words included in the Academic Keyword List match the defi nition of academic vocabulary, confi rm-ing the accuracy of the retrieval procedure. AKL words have been shown to fall in semantic categories such as cause and effect, inclusion/exclusion, evaluation, comparison, importance, quantities, and speech acts, which clearly relate to academic work, the organization of scientifi c discourse, or the building of the argument of academic texts. The method has also made


it possible to appreciate the paramount importance of general service words in academic prose. A number of general service words take on prominent rhetorical and organizational functions in academic discourse and their absence from lists such as Coxhead’s (2000) Academic Word List may be highly problematic in academic writing courses.

The Academic Keyword List still needs validation. However it unquestion-ably offers ‘a portal into the complex behaviour and intricate relationships of individual lexical items’ (Hanciog lu et al., 2008: 461). Each AKL word should be subject to a careful corpus-based analysis to confi rm its status as an academic word and establish how it is used in academic prose in terms of meaning, phraseology, and sentence position.

Part II

Learners’ use of academic vocabulary

Having clarifi ed what academic vocabulary is by providing a critical over-view of its many defi nitions, and proposed a data-driven methodology to extract potential academic words from corpora, I will now examine EFL learners’ use of academic vocabulary. Chapter 3 offers a detailed descrip-tion of the corpora used and the method adopted to compare them: Granger’s (1996) Contrastive Interlanguage Analysis (CIA). In Chapter 4 the Academic Keyword List is shown to include a large number of lemmas that can be used to organize academic texts and structure their content around logico-semantic relations. The function of ‘exemplifi cation’ is presented in detail. This illustrates the type of data and the results obtained when the whole range of lexical strategies available to expert writers for establishing cohesive links in their texts is examined. Chapter 4 also focuses on the phraseology of academic words. Chapter 5 aims to test the working hypoth-esis that upper-intermediate to advanced EFL learners, irrespective of their mother-tongue background, share a number of linguistic features that characterize their use of academic vocabulary. The learner corpus used is a collection of texts taken from the International Corpus of Learner English. As learner texts are short argumentative essays, I focus on AKL words that are used to organize discourse and build the rhetoric of a text, rather than on words that focus on research, analysis and evaluation. Many of the lin-guistic features that characterize learner writing are shared by learners from a range of mother tongue backgrounds, and can therefore be assumed to be developmental. The EFL learners are all learning how to write in a foreign language, and they are often novice writers in their mother tongue as well. However, not all the features of their writing are shared across the


different language backgrounds, and the differences may be due to transfer from the writers’ mother tongues. Chapter 5 therefore ends with a brief discussion of the potential infl uence of their mother tongue on French speakers’ use of academic vocabulary in English.

Chapter 3

Investigating learner language

Whatever the defi nition adopted, academic vocabulary has generally been described as a major source of diffi culty for EFL learners. In the second part of this book, I will focus on learners’ use of academic words that serve typical organizational or rhetorical functions in academic discourse. I will investigate whether upper-intermediate to advanced EFL learners, irrespec-tive of their mother tongue backgrounds, share characteristic ways of using academic vocabulary. In this chapter, I describe the corpora and methodol-ogy used to pursue these objectives. The learner corpus used is the Interna-tional Corpus of Learner English, which is among the largest non-commercial learner corpora currently existing, and contains texts written by learners with different mother-tongue backgrounds. The method of analysing learners’ use of academic vocabulary and comparing it with that of expert writers relies on Granger’s (1996a) Contrastive Interlanguage Analysis. A subset of the British National Corpus is used as a control corpus and helps bring to light features of learner language.

3.1. The International Corpus of Learner English

The learner data used consist of ten sub-corpora of the International Corpus of Learner English version 1 (ICLE) compiled at the University of Louvain, Belgium, under the supervision of Sylviane Granger (Granger et al., 2002; Granger, 2003). A computer learner corpus is an electronic collection of (near-) natural language learner texts assembled according to explicit design criteria (see Granger, 2009 for the design criteria, and Ellis and Barkhuizen, 2005 for a discussion of different types of learner data). Each learner text is documented with a detailed learner-profi le questionnaire, which all the learners were requested to fi ll in. Learner-profi le question-naires give two types of information: learner characteristics and informa-tion on the type of task.


As shown in Figure 3.1, ICLE texts share a number of learner and task variables, which were used as corpus-design criteria. All the learners who submitted an essay to the ICLE were university undergraduates and were therefore usually in their twenties. They were learners of English as a Foreign Language rather than as a Second Language and were in their third or fourth year of university study. On the basis of these external crite-ria, their level is described as advanced although ‘individual learners and learner groups differ in profi ciency’ (Granger, 2003: 539).1 Learner pro-ductions have quite a few task variables in common, notably in terms of medium (writing), genre (academic essay), fi eld (general English rather than English for Specifi c Purposes) and length (between 500 and 1,000 words).

Other variables differ (e.g. mother tongue background of learners, L2 exposure, essay topic and task settings). The ICLE learners represent 11 different mother tongue backgrounds: Bulgarian, Czech, Dutch, Finnish, French, German, Italian, Polish, Russian, Spanish and Swedish.2 A large proportion of the learner texts are argumentative, but the ICLE essays cover a wide range of topics. Topics include Most university degrees are theoretical and do not prepare students for the real word; In the words of the old song, ‘Money is the root of all evil’ and Feminists have done more harm to the cause of women than good. Essays differ in task conditions: they may have been written in timed or untimed conditions, as part of an exam or not, with reference tools such as

International Corpus of Learner English

Shared features

Learner variables

Age Medium Gender TopicTask setting

TimmingExamReferencetools

Mother tongueRegionOther FLL2 exposure

FieldGenreLength

Learning contextProficiency level

Task variables Learner variables Task variables

Variable features

Figure 3.1 ICLE task and learner variables (Granger et al., 2002: 13)

Investigating learner language 69

grammars and dictionaries or not. Most studies of the ICLE data to date have not taken these task settings into consideration (see Ädel’s (2008) analysis of timed and untimed essays for an exception). However, research-ers in second and foreign language acquisition and teaching insist that the infl uence of task type and condition is important (Shaw, 2004; Kroll, 1990).

Task and learner variables can be used to compile homogeneous sub-corpora. As shown in Table 3.1, this study makes use of ten ICLE sub- corpora, representing different mother tongue backgrounds. Learner essays in each sub-corpus were carefully selected in an attempt to control for the task variables which may affect learner productions: all the texts are untimed argumentative essays, potentially written with the help of reference tools. Although essays written without the help of reference tools would arguably have been more representative of what advanced EFL learners can pro-duce, I chose to select untimed essays with reference tools as they represent the majority of learner texts in ICLE.3

The ICLE sub-corpora compiled for this study were analysed with the Concord tool of the computer software WordSmith Tools 4 (Scott, 2004). The software was used to examine the occurrences of potential academic words in context. The Clusters option proved very useful for identifying the most frequent n-grams or lexical bundles that contained the words being studied. More information on the types of analyses that can be performed with WordSmith Tools can be found on Mike Scott’s webpage (http: //www.lexically.net) and in Scott and Tribble (2006).

Table 3.1 Breakdown of ICLE essays

No. of essays

No. of words Average no. of

words per essay

Czech (ICLE-CZ) 147 130,768 890Dutch (ICLE-DU) 196 162,243 828Finnish (ICLE-FI) 167 125,292 750French (ICLE-FR) 228 136,343 598German (ICLE-GE) 179 109,556 612Italian (ICLE-IT) 79 47,739 604Polish (ICLE-PO) 221 140,521 636Russian (ICLE-RU) 194 165,937 855Spanish (ICLE-SP) 149 99,119 665Swedish (ICLE-SW) 81 48,060 593

TOTAL 1641 1,165,524 697


3.2. Contrastive Interlanguage Analysis

The methodology most frequently used to analyse learner corpora is Contrastive Interlanguage Analysis (CIA) (Granger 1996a; Gilquin 2000/2001). Unlike contrastive analysis, which consists of comparing two or more lan-guages, CIA compares varieties of one language: native and non-native vari-eties (L1/L2), or different non-native varieties (L2/L2) (see Figure 3.2). L1/L2 comparisons bring out the features of non-nativeness in learner pro-ductions, ‘which at an advanced level are as much (if not more) a question of over- and under-use of linguistic items or structures as a question of downright errors’ (Granger et al., 2009: 41). Comparisons of different interlanguages (e.g. the English of French speakers compared to that of Dutch speakers), on the other hand, make it possible to assess whether these features are peculiar to one language group (and thus possibly due to the infl uence of the learner’s mother tongue), or shared by several learner populations (and therefore likely to be developmental or due to other causes such as teaching methods) (cf. Granger, 2002).

In his preface to Learner English on Computer (Granger, 1998), Leech describes the native control corpus as ‘a standard of comparison, a norm against which to measure the characteristics of the learner corpora’ (Leech, 1998: xv). In second language acquisition research, however, L1/L2 comparisons have generally been criticized for being guilty of the ‘comparative fallacy’ (Bley-Vroman, 1983), i.e. for comparing learner language to a native speaker norm and thus failing to analyse interlanguage in its own right (see Lakshmanan and Selinker, 2001; Firth and Wagner, 1997). A strong argument that can be invoked in defence of the CIA model is that the native speaker norm used in learner corpus research is explicit and corpus-based (Mukherjee, 2005) rather than implicit and intuition-based as has been common in second language acquisition (SLA) studies

CIA

L1 > < L2 L2 > < L2

Figure 3.2 Contrastive Interlanguage Analysis (Granger, 1996a)


(see also Granger, 2009: 20). Lakshmanan and Selinker (2001) address the issue of the comparative fallacy and warn against the danger of ‘judging language learner speech utterances as ungrammatical from the standpoint of the target grammar without fi rst having compared the relevant inter language utterances with the related speech utterances in adult native-speaker spoken discourse’ (Lakshmanan and Selinker, 2001: 401). Although they do not dwell on it, Lakshmanan and Selinker’s point may be understood as a plea for more natural language data (i.e. corpus data) and a warning against hasty conclusions based on a single researcher’s intuitions.

Another criticism of L1/L2 comparisons is directed at the idea of the ‘native speaker’ as the target norm (e.g. Piller, 2001; Tan, 2005). Mukherjee, however, argues that ‘nativeness’ remains a useful construct both for lin-guistics and for the ELT community, a ‘useful myth’ in Davies’s (2003) terms. He proposes a usage-based defi nition of the native speaker based on three aspects that he regards as central to native-like performance, i.e. lexico-grammaticality, acceptability and idiomaticity (see Pawley and Syder, 1983):

The term ‘native speaker’ should be used for an abstraction of all language users (1) who have good intuitions about what is lexicogram-matically possible in a given language and speak/write accordingly, (2) who know to a large extent what is acceptable in a given communi-cation situation and speak/write accordingly, (3) whose usage is largely idiomatic in terms of linguistic routines commonly used in a given speech community. If we refer to an individual speaker as a native speaker, this speaker is thus taken to exemplify the abstract native speaker model on grounds of his/her language use. (Mukherjee, 2005: 14)

Mukherjee advocates a corpus-approximation to the native speaker norm and argues that corpus data can be used to describe this norm by ‘general-izing and abstracting from a vast amount of representative performance data’ (ibid: 15). In this book, the corpus-approximation to the native speaker norm is based on British and American English corpora. It should be noted, however, that the existence of a variety of norms is recognized in learner corpus research (see Granger, 2009) and that other varieties of English are sometimes used as control corpora. For example, Gilquin and Granger (2008) compared the Tswana component of the second edition of the International Corpus of Learner English to a corpus of South African English editorials. The control corpora used in this study are described below.


3.3. A comparison of learner vs. expert writing

Carrying out L1/L2 comparisons implies choosing an L1 corpus to be used as some kind of ‘norm’ with which the learner corpus data can be compared. In this study, learner writing was compared to expert academic prose. There is no general agreement, however, on the type of material that is best suited to serve as a control for a learner corpus. Several researchers have criticized the use of professional writing in learner corpus research, arguing that it is ‘both unfair and descriptively inadequate’ (Lorenz, 1999a: 14) and taking a stand against the ‘unrealistic standard of “expert writer” models’ (Hyland and Milton, 1997: 184). Native student writing is arguably a better source of comparable data to EFL learner writing if the aim of the comparison is to describe and evaluate interlanguage(s) as fairly as possible. It is, however, doubtful whether the fi ndings from such comparisons could make their way into the classroom. As Leech puts it, ‘native-speaking students do not necessarily provide models that everyone would want to imitate’ (Leech, 1998: xix). For example, native students have been shown to produce more dangling participles than EFL learners (Granger, 1997) and different types of spelling mistakes (Cutting, 2000).

The question of the norm is best addressed by considering the aim of the comparison. Professional writing has a major role to play in learner corpus research if instruction and pedagogical applications are the goals of the comparison between learner and native-speaker productions. As Ädel put it,

On the one hand, it can be argued that in order to evaluate foreign learner writing by students justly, we need to use native-speaker writing that is also produced by students for comparison. On the other hand, it can also be argued that professional writing represents the norm that advanced foreign learner writers try to reach and their teachers try to promote. In this respect, a useful corpus for comparison is one which offers a collection of what Bazerman (1994: 131) calls ‘expert performances’. (Ädel, 2006: 206–7)

The International Corpus of Learner English, however, consists of argu-mentative texts and ‘argumentative essay writing has no exact equivalent in professional writing’ (De Cock, 2003: 196). It has been suggested that the ICLE might be compared with ‘a corpus of newspaper editorials, a text type which combines the advantages of being argumentative in nature and written by professionals’ (Granger, 1998a: 18, footnote 10). In a number of


studies based on ICLE texts produced by Spanish EFL learners, Neff and her colleagues used both native-speaker student writing and newspaper editorials as control corpora (Neff et al., 2004a; 2004b; Neff van Aertselaer, 2008). General corpora have also been used in learner corpus research. Nesselhauf (2005), for example, made use of the written part of the British National Corpus to determine the degree of acceptability of verb-noun combinations that had been extracted from the German subset of ICLE.

The British National Corpus (BNC) was created to be a balanced reference corpus of late twentieth century British English. The BNC contains both written and spoken material. The written component totals about 90 mil-lion words and includes samples of academic books, newspaper articles, popular fi ction, letters, university essays and many other text types. The text selection procedure has been described as follows:

In selecting texts for inclusion in the corpus, account was taken of both production, by sampling a wide variety of distinct types of material, and reception, by selecting instances of those types which have a wide distri-bution. Thus, having chosen to sample such things as popular novels, or technical writing, best-seller lists and library circulation statistics were consulted to select particular examples of them. (Aston and Burnard, 1998: 28)

The BNC mark-up conforms to the Text Encoding Initiative (TEI) recom-mendations (Burnard, 2007). Mark-ups include rich metadata on a variety of structural properties of texts (e.g. headings, sentences and paragraphs), fi le description, text profi le, as well as linguistic information (morphosyn-tactic tags, lemmas, etc.).

Three criteria were originally used to select written texts to design a balanced corpus: domain, time and medium. Domain refers to the subject fi eld of the texts; time refers to the period when the text was written, and medium refers to the type of publication (books, newspapers, periodi-cals, etc.). Lee (2001) criticized the domain categories for being overly broad and not suffi ciently explicit, and developed a new resource called the BNC Index which contains genre labels for all BNC texts. Table 3.2 gives the breakdown of the genre categories in the BNC written corpus and shows that genre labels are often hierarchically nested. Thus, if we want to analyse texts produced by scholars specializing in natural sciences, we can select all BNC texts classifi ed under ‘W_ac_nat_science’. On the other hand, if we are not interested in discipline-specifi c differences and want to examine


Table 3.2 BNC Index – Breakdown of written BNC genres (Lee 2001)

BNC written No. of

words

% ‘Super genre’ No. of

fi les

W_ac_humanities_arts 3,321,867 3.8%

Academic prose 17.7%

87W_ac_medicine 1,421,933 1.6% 24W_ac_nat_science 1,111,840 1.3% 43W_ac_polit_law_edu 4,640,346 5.3% 186W_ac_soc_science 4,247,592 4.9% 138W_ac_tech_engin 686,004 0.8% 23W_admin 219,946 0.3% 12W_advert 558,133 0.6% 60W_biography 3,528,564 4.0% 100W_commerce 3,759,366 4.3% 112W_email 213,045 0.2% 7W_essay_sch 146,530 0.2% Unpublished

essays 0.3%7

W_essay_univ 65,388 0.1% 4W_fi ct_drama 45,757 0.1%

Fiction 18.6%2

W_fi ct_poetry 222,451 0.3% 30W_fi ct_prose 15,926,677 18.2% 432W_hansard 1,156,171 1.3% 4W_institut_doc 546,261 0.6% 43W_instructional 436,892 0.5% 15W_letters_personal 52,480 0.1% Letters 0.2% 6W_letters_prof 66,031 0.1% 11W_misc 9,140,957 10.5% 500W_news_script 1,292,156 1.5% 32W_news_brdsht_nat_arts 351,811 0.4%

News 7.8%

51W_news_brdsht_nat_commerce 424,895 0.5% 44W_news_brdsht_nat_editorial 101,742 0.1% 12W_news_brdsht_nat_misc 1,032,943 1.2% 95W_news_brdsht_nat_reportage 663,355 0.8% 49W_news_brdsht_nat_science 65,293 0.1% 29W_news_brdsht_nat_social 81,895 0.1% 36W_news_brdsht_nat_sports 297,737 0.3% 24W_news_other_arts 239,258 0.3% 15W_news_other_commerce 415,396 0.5% 17W_news_other_report 2,717,444 3.1% 39W_news_other_science 54,829 0.1% 23W_news_other_social 1,143,024 1.3% 37W_news_other_sports 1,027,843 1.2% 9W_news_tabloid 728,413 0.8% 6W_non_ac_humanities_arts 3,751,865 4.3%

Non-academic prose 19.1%

111W_non_ac_medicine 498,679 0.6% 17W_non_ac_nat_science 2,508,256 2.9% 62W_non_ac_polit_law_edu 4,477,831 5.1% 93W_non_ac_soc_science 4,187,649 4.8% 128W_non_ac_tech_engin 1,209,796 1.4% 123W_pop_lore 7,376,391 8.5% 211W_religion 1,121,632 1.3% 35

TOTAL 87,284,364 100% 3,144


texts produced by professional writers in higher education settings, we can select all texts whose categorizing labels begin with ‘W_ac’.

The BNC sub-corpus of academic prose in humanities and arts (W_ac_humanities_arts; henceforth BNC-AC-HUM) totals 3,321,867 words. It was used as the comparison corpus to ICLE in this study. This sub-corpus was chosen for two main reasons. First, ICLE texts were produced by students of humanities; texts in the BNC-AC-HUM are arguably quite close to the type of text these students might have come across in their fi rst few years at university. They also have the advantage of corresponding to the type of writing that learners will try to produce during their university studies. There are, however, major differences between the two corpora. First, ICLE is a corpus of unpublished university student essays while BNC-AC-HUM consists of samples of published articles and books. Second, student essays rarely total more than 1,000 words while samples in the BNC-AC-HUM are much longer (from 25,000 to 45,000 words).4 Third, the topics in BNC-AC-HUM differ from those in ICLE (described in Section 3.1 above). They include The people’s peace; National liberation; The morality of freedom; Europe in the central middle ages; China’s students; British literature since 1945; What is this thing called science?; Soviet relations with Latin America; Nietzsche on tragedy, etc. Unlike the ICLE, topics in BNC-AC-HUM appear only once. Interpreting the results in the light of genre analysis thus required special care: differences between student essays and expert writing may simply refl ect differences in their communicative goals and settings (Neff et al., 2004a).

The W_fi c and W_news sub-corpora (cf. Table 3.2) were sometimes used to compare the frequency of words and phrases across ‘super genres’. The spoken part of the British National Corpus was also regularly consulted to check whether words and word sequences that were found in learner writing are more typical of speech or academic writing. The BNC spoken corpus (BNC-SP) consists of 10,334,947 words and includes a wide variety of spoken registers, among others, broadcast documentaries and news, interviews and lectures.

The British National Corpus was accessed via the BNCweb (CQP-edition) interface developed by Stefan Evert and Sebastian Hoffmann. This web interface is the result of a ‘marriage of two corpus tools’ (Hoffmann and Evert, 2006), i.e. the BNCweb, a web-based client developed at the University of Zurich which allows users to access the BNC by means of a Web browser (see Lehmann et al., 2000) and the Corpus Query Processor (CQP), a central component of IMS Open Corpus Workbench, which ‘allows sophisticated searches both for individual words (which can be matched against regular


expressions) and for lexico-grammatical patterns (using linear grammars that have access to all levels of annotation)’ (Hoffmann and Evert, 2006: 180). The CQP edition of the BNCweb combines the strengths of both software packages while overcoming their limitations. It is a marriage between the effi ciency and fl exibility of CQP queries, and the user-friendli-ness of BNCweb with its wide range of query options and display facilities. Hoffmann et al. (2008) proved further information on the British National Corpus and the BNCweb interface.

One tool that is particularly useful is Collocations, which picks out signifi cant co-occurrents of the search word on the basis of a number of measures of association. Association measures are the most widely used method of distinguishing between casual and signifi cant co-occurrences. They compute an association score for each pair of words extracted from a corpus, which indicates the strength of the association relative to that expected by chance. Users of the BNCweb can decide to use any of fi ve different measures: mutual information, MI3, z-score, log-likelihood and log-log measures.5 They can also sort co-occurrents by decreasing frequency. A number of other settings are customizable, e.g. maximum window span, minimum frequency of the co-occurrence, minimum frequency of the co-occurrent, inclusion of lemma and part-of-speech information, etc. Figure 3.3 displays a collocation query result. Signifi -cant co-occurrents are sorted by decreasing log-likelihood values (right column). The frequency of the co-occurrence is given together with the number of texts in which it appears.

Mutual information, MI3, z-score, log-likelihood and log-log measures rank co-occurrences in very different ways (Evert, 2004). McEnery et al. (2006) compared the various statistical measures provided by BNCweb and reported that ‘MI and z-scores tend to put too much emphasis on infre-quent words. In contrast, the log-likelihood, log-log and MI3 tests appear to provide more realistic collocation information’ (McEnery et al., 2006: 220). The log-likelihood test was therefore used to study the phraseology of aca-demic words in expert and learner writing. The log-likelihood scores can be directly compared with critical values of a chi-square distribution table (see Oakes, 1998: 176). Rayson et al. (2004), however, focused on the comparison of word frequencies between corpora and suggested that, in order to extend applicability of the frequency comparisons to low expected values, use of a threshold value of 15.13 is preferred at p < 0.01. Co-occurrence frequencies can be quite low and I therefore followed Rayson et al.’s (2004) suggestion. Co-occurrences were analysed in windows of one to three words to both the left (3L-1L) and the right (1R-3R).

Investigating learner language

77

Figure 3.3 BNCweb Collocations option


Log-likelihood measures are strongly dependent on corpus size and word frequencies. Co-occurrence statistics are therefore not comparable across corpora of different sizes such as the British National Corpus and the Interna-tional Corpus of Learner English. The ICLE sub-corpora are in fact too small for a statistical analysis of co-occurrences to be meaningful. Academic words are not high-frequency words such as make, do and take and co-occurrences often appear less than three times. In his study of the statistics of word co-occurrences, Evert argued that ‘data with co-occurrence frequency f < 3, i.e. the hapax and dis-legomena, should always be excluded from the statistical analysis’ (Evert, 2004: 133) as expected frequencies and p values for low frequency words are distorted in unpredictable ways.

A distributional (Evert, 2004) or frequency-based approach (Nesselhauf, 2004) was adopted to examine the phraseology of academic words in learner writing. Word pairs in the ICLE sub-corpora were classifi ed into three groups according to their co-occurrence status in professional academic writing:

– word pairs that are statistically signifi cant co-occurrents in the academic sub-corpus of the BNC (BNC-AC). In a pilot study, I found that learners’ word pairs were sometimes not statistically signifi cant in BNC-AC-HUM just because the co-occurrence was not frequent enough. As soon as more data was used, the co-occurrence proved signifi cant. I therefore decided to use the larger BNC-AC instead of the BNC-AC-HUM to judge the acceptability and typicality of EFL learners’ phraseological sequences (see Section 5.2.3).

– word pairs that appear in the BNC-AC but are not statistically signifi cant co-occurrents;

– word pairs that do not appear in the BNC-AC;

Word pairs that did not appear in the BNC-AC were presented to a native speaker of English for acceptability judgments.


This chapter has described the data and methodology used to investigate the use of academic vocabulary in writing by EFL learners. Special care has been taken to select a set of learner essays from the International Corpus of Learner English that is as homogeneous as possible and to control for a number of variables that have been found to infl uence such writing.


The learner corpus can be compared to the humanities and arts academic sub-corpus of the British National Corpus to identify learner-specifi c features of the use of academic vocabulary. The BNC spoken corpus can also sometimes be useful to check whether specifi c words and phrases that appear in the learner sub-corpora are more typical of speech or writing. The method used to investigate learners’ use of academic vocabulary is based on Contrastive Interlanguage Analysis (CIA) and combines compari-sons of learner and native-speaker writing, and comparisons of different learner interlanguages.

CIA is very popular among researchers in the fi eld of learner corpus research and has helped to highlight an unprecedented number of fea-tures that characterize learner interlanguages. To date, however, most stud-ies have used the technique only to compare a learner corpus and a native reference corpus, rather than to explore different learner corpora in the same target language. The studies that have compared more than one interlanguage have usually focused on learners from one mother-tongue background, and used data from one or two other learner populations only to check whether the features they have identifi ed are L1-specifi c (and thus possibly transfer-related) or are shared by other learners. L2/L2 comparisons involving many different fi rst languages are, however, indispensable if we want to identify the distinguishing features of learner language at a given stage of development (Bartning, 1997). In the following chapters, I try to make the most of CIA by comparing academic vocabulary in ten learner corpora representing different mother-tongue backgrounds.

Chapter 4

Rhetorical functions in expert academic writing

This chapter deals with academic vocabulary that serves specifi c rhetorical and organizational functions in expert academic writing. Section 4.1 focuses on the Academic Keyword List and shows that a high proportion of AKL words can fulfi l these functions in academic prose. It lists several steps which are necessary to turn the AKL into a tool that can be used for curriculum and materials design (most notably a phraseological analysis of AKL words). Section 4.2 presents a detailed analysis of exemplifi catory devices in academic writing. This serves as an illustration of the type of data and results obtained when the whole range of lexical strategies available to expert writers to organize scientifi c discourse are examined. For lack of space it is impossible to describe in similar detail all the functions that were analysed in the BNC-AC-HUM so as to provide a basis for comparison to EFL learner writing. Section 4.3 briefl y comments on the types of lexical devices used by expert writers to serve four additional functions: ‘expressing cause and effect’, ‘comparing and contrasting’, ‘expressing a concession’ and ‘refor-mulating: paraphrasing and clarifying’ and aims to characterize the phrase-ology of rhetorical functions in academic prose.

4.1. The Academic Keyword List and rhetorical functions

The functional syllabus has a long tradition in English language teaching (see Wilkins, 1976; Weissberg and Buker, 1978). Jordan (1997: 165) reports that most of the textbooks that were published in Britain in the 1980s and 1990s that followed a product approach to academic writing were orga-nized according to language functions such as explanation, defi nition, exemplifi cation, classifi cation, cause and effect, and comparison and con-trast (e.g. Jordan, 1999). However, they were rarely based on principled


selection criteria, relying instead on the writers’ perceptions of good practice in academic writing.

Unlike textbooks adopting a functional approach, courses which use vocabulary as the unit of progression, introduce new words according to principles such as frequency and range of occurrence. Nation explains that

such courses generally combine a “series” and a “fi eld” approach to selection and sequencing. In a series approach, the items in a course are ordered according to a principle such as frequency of occurrence, complexity or communicative need. In a fi eld approach, a group of items is chosen and the course covers them in any order that is convenient, eventually checking that all the items are adequately covered. Courses which use vocabulary as the unit of progression tend to break vocabulary lists into manageable fi elds, (. . . ), according to frequency, which are then covered in an opportunistic way. (Nation, 2001: 386)

Most pedagogical applications of Coxhead’s (2000) Academic Word List to date have adopted this particular approach, using the frequency-based AWL sub-lists as fi elds (e.g. Obenda, 2004; Huntley, 2006).

There is a need for teaching materials that merge the two types of syllabus design, thus adopting a ‘functional-product’ approach (Jordan, 1997: 165) to academic writing while introducing new vocabulary according to princi-pled criteria such as frequency and range of occurrence. This is precisely where the Academic Keyword List has a role to play.

As explained in Section 2.4, the Academic Keyword List requires pedagogic mediation: it is a platform which can inform a functional syllabus for academic writing, but it needs to be organized. As argued by Martinez et al. (2009: 193), ‘a list based on semantic and pragmatic criteria would perhaps be more useful than lists built solely on frequency criteria.’ Sinclair, however, warns us that ‘there is no assumption that meaning attaches only to the word’ (Sinclair, 2004b: 160). Similarly, Siepmann (2005: 86) comments that ‘neat compartmentalizing of meanings or functions can do no more than partially capture a complex reality’ in which any word or multi-word sequence may express more than one discourse relation. This being said, the results of the automatic semantic analysis of the Academic Keyword List revealed that a signifi cant proportion of AKL words fall into semantic categories that correspond to the rhetorical functions typical of scientifi c discourse, e.g. A2.2. Affect: Cause – connected, A4. Classifi cation, A5. Evaluation, A6. Comparing, Q2.2. Speech acts (see Section 2.4). A close

Rhetorical functions in expert academic writing 83

examination of the words classifi ed into these semantic categories made it possible to identify twelve rhetorical functions that dovetail with the func-tions typically treated in EAP textbooks adopting a functional approach to academic writing:

1. Exemplifi cation, e.g. example, for example, illustrate 2. Cause and effect, e.g. cause, consequence, result 3. Comparison and contrast, e.g. contrast, difference, same 4. Concession, e.g. although, despite, however 5. Adding information, e.g. fi rst, further, in addition to 6. Expressing personal opinion, e.g. appropriate, essential, major 7. Expressing possibility and certainty, e.g. likely, possibility, unlikely 8. Introducing topics and ideas, e.g. discuss, examine, subject 9. Listing items, e.g. fi rst, second, third10. Reformulating – paraphrasing and clarifying, e.g. namely11. Quoting and reporting, e.g. defi ne, report, suggest12. Summarizing and drawing conclusions, e.g. conclude, conclusion, summary.

The next step is to analyse all words that may serve one of these twelve functions in context, with special emphasis on their phraseology. Multi-word sequences have been shown to provide ‘basic building blocks for constructing spoken and written discourse’ (Biber and Conrad, 1999: 185) and to correlate closely with the complex communicative demands of a particular genre, thus contributing to its lexical profi le (Biber et al, 1999; 2004; Luzón Marco, 2000). A phraseological analysis will also make it possible to investigate how academic vocabulary contributes to this ‘shared scientifi c voice or “phraseological accent” which leads much tech-nical writing to polarise around a number of stock phrases’ (Gledhill, 2000: 204). It will examine phrasemes, i.e. syntagmatic relations between at least two lemmas, contiguous or not, written separately or together, which are typically syntactically closely related and constitute ‘“preferred” ways of saying things’ (Altenberg, 1998: 122). This is because such phrasemes:

– form a functional (referential, textual or communicative) unit (e.g. Burger, 1998);

and

– display arbitrary lexical restrictions (e.g. Mel’cuk, 1998);


and/or

– are characterized by a certain degree of semantic non-compositionality (e.g. Barkema, 1996);

– display arbitrary restrictions on the word forms that can be used to instan-tiate at least one of the lemmas involved;

– display a certain degree of syntactic fi xity.

The phraseological analysis used here is based on Granger and Paquot’s (2008a) classifi cation of phraseological units into three main categories: referential phrasemes, textual phrasemes and communicative phrasemes. Referential phrasemes are used to convey a content message: they refer to objects, phenomena or real-life facts. They include lexical and grammatical collocations, idioms, similes, irreversible bi- and trinomials, compounds and phrasal verbs. Textual phrasemes are typically used to structure and organize the content (i.e. referential information) of a text or any type of discourse; they include grammaticalized sequences such as complex prepo-sitions and complex conjunctions, linking adverbials and textual sentence stems. Communicative phrasemes are used to express feelings or beliefs towards a propositional content or to explicitly address interlocutors, either to focus their attention, include them as discourse participants or infl uence them. They include speech act formulae, attitudinal formulae, common-places, proverbs and slogans.

In this chapter and the next, I focus on the vocabulary of fi ve rhetorical functions — exemplifi cation, cause and effect, comparison and contrast, concession, and reformulating — with occasional forays into other func-tions. Apart from being essential rhetorical functions in academic prose, these functions should be among the least sensitive to the text type differ-ences discussed in Section 3.3. The use of academic words is compared in the BNC-AC-HUM corpus (a corpus of book samples and journal articles written by experts in the fi elds of arts and humanities) and the International Corpus of Learner English (a corpus of short argumentative essays produced by EFL learners of English). As the BNC includes truncated texts, it would not be reasonable to quantitatively compare the words that are used to serve the function of ‘summarizing and drawing conclusions’. This func-tion is typically localized in the last paragraphs of a piece of academic writing and might thus be absent from a number of BNC texts. Nor is it reasonable to focus on functions such as ‘reporting and quoting’ and ‘expressing personal opinion’. Unlike experts writing in their fi eld, the


learners who produced the argumentative essays were not supposed to show that they were familiar with the subject by referring to or quoting from the literature. By contrast, they were explicitly encouraged to give their personal opinions (topics for the essays include: ‘Some people say that in our modern world, dominated by science, technology and industri-alism, there is no longer a place for dreaming and imagination. What is your opinion?’ and ‘In the 19th century, Victor Hugo said: “How sad it is to think that nature is calling out but humanity refuses to pay heed.” Do you think it is still true nowadays?’) (see Section 3.1).

Several researchers in applied corpus linguistics have examined language features in general reference corpora and compared the distributions and patterns found in actual language use with the presentations of the same features in teaching materials such as textbooks or grammars (e.g. Carter, 1998b; Conrad, 2004; Römer, 2004a; 2004b; 2005). They have often found considerable mismatches between naturally-occurring language and the type of language that is presented as a model in teaching materials (Römer, 2008: 4). I therefore consulted several EAP textbooks (Harris Leonhard, 2002; Jordan, 1999; Lonon Blanton, 2001; Oshima and Hogue, 2006; Ruetten, 2003; Zemach and Rumisek, 2005; Zwier, 2002), and listed all the lexical items that are commonly taught to serve rhetorical functions. The textbooks-derived list appeared to be very different from the words found in the Academic Keyword List. For example, the AKL includes a number of words and phrasemes that are commonly used as exemplifi ers: the word-like units for example and for instance, the noun example, the verbs illustrate and exemplify, the preposition such as, the adverb notably and the abbrevia-tion e.g. Other lexical items listed in EAP materials but not found in the AKL are the expressions by way of illustration and to name but a few, the nouns illustration and a case in point and the preposition like.

I decided to include the lexical items found in EAP textbooks in my study of academic vocabulary for two main reasons. First, it is not quite clear why these items are taught to novice writers and EFL learners while other much more frequent lexical items that are used to express the same rhetorial func-tions are missing. Frequency, however, may not be the sole criterion to include lexical items in the curriculum (see Section 1.1.1). Some of these non-AKL words may be used in very specifi c lexico-grammatical environ-ments, have a very restricted meaning or prove particularly diffi cult for learn-ers. It is only by examining their frequency and patterns of use in expert and learner corpora that I shall be able to assess whether these words and phrase-mes should be part of an academic vocabulary and added to the AKL.


Second, their inclusion in the description of a specifi c function in academic writing allows us to approximate as closely as possible what Hoffmann (2004: 190) referred to as ‘conceptual frequency’, so that the frequency of each exemplifi catory lexical item can be calculated as a pro-portion of the total number of exemplifi ers. As Wray stated in her book on formulaic language,

To capture the extent to which a word string is the preferred way of expressing a given idea (for this is at the heart of how prefabrication is claimed to affect the selection of a message form), we need to know not only how often that form can be found in the sample, but also how often it could have occurred. In other words, we need a way to calculate the occurrences of a particular message form as a proportion of the total number of attempts to express that message. (Wray, 2002: 30)

This approach should help us move towards ‘understanding the intersec-tion of form and function’ (Swales, 2002: 163) in academic prose.

The Academic Keyword List is based on native corpora only, which has limitations for an analysis of learner writing, especially if conceptual frequency is to be investigated. EFL learners may use different lexical devices than native writers to serve rhetorical functions. For example, they repeatedly use word-like units such as in a nutshell, in brief and all in all for summarizing and concluding, which are quite rare in academic prose. A keyword procedure such as that described in Section 2.3.1 for the auto-matic extraction of potential academic words was therefore adopted to identify words and word sequences that EFL learners frequently use, but which are not favoured by expert academic writers. The ICLE corpus was compared to the BNC-AC-HUM to extract distinctive words in the learner corpus. The resulting list was analysed to identify words that might serve one of the 12 rhetorical functions listed above. In learner corpus research, positive keywords are often referred to as overused words and negative keywords are said to be underused. These two terms are neutral, and sim-ply refl ect the fact that a word is more/less frequent in learner writing. Examples of overused words which do not belong to the AKL but are employed to serve rhetorical functions in learner writing include like, thing, say, let, I, really, fi rstly, secondly, thirdly, opinion, maybe, say, sure, but, thanks, always, so and why (see De Cock, 2003 for a keyword analysis of the French sub-corpus of ICLE). Learner-specifi c word sequences are discussed in Section 5.2.3.


The fi nal lists of words that may be used to serve one of the fi ve selected rhetorical functions are given below. The words in italics are not part of the Academic Keyword List. They were identifi ed on the basis of a close examination of EAP materials and a keyword analysis of learner corpora. They are included in the corpus-based analyses presented in this chapter and in Chapter 5 to assess the adequacy of the treatment of rhetorical func-tions in EAP textbooks and investigate whether the AKL should be supple-mented with additional academic words.

Exemplifi cation: example, illustration, a case in point, illustrate, exemplify, such as, like, for example, for instance, e.g., notably, to name but a few, by way of illustration

Comparison and contrast: analogy, comparison, (the) contrary, contrast, difference, differentiation, distinction, distinctiveness, (the) opposite, parallel, parallelism, resemblance, (the) reverse, (the) same, similarity, alike, analogous, common, comparable, contrary, contrasting, different, differing, distinct, dis-tinctive, distinguishable, identical, opposite, parallel, reverse, same, similar, unlike, compare, contrast, correspond, differ, distinguish, differentiate, look like, parallel, resemble, analogously, by/in comparison, by/in contrast, by way of contrast, comparatively, contrariwise, contrastingly, conversely, correspond-ingly, differently, distinctively, identically, in the same way, likewise, on the con-trary, on the other hand, *on the other side, *on the opposite, parallely, reversely, similarly, as against, as opposed to, by/in comparison with, contrary to, *in contrary to, like, in contrast to/with, in parallel with, unlike, versus, as, whereas, while, as … as, compared with/to, in the same way as/that

Cause and effect: cause (n.), consequence, effect, factor, implication, origin, outcome, root, reason, result, source, arise from/out of, bring about, cause (v.), contribute to, derive, emerge, follow from, generate, give rise to, induce, lead, make sb/sth do sth, prompt, provoke, result in/from, stem from, trigger, yield, consequent, responsible, as a result of, as a consequence of, because of, due to, in consequence of, in (the) light of, in view of, on account of, on the grounds that, owing to, thanks to, accordingly, as a consequence, as a result, by implication, consequently, hence, in consequence, so, thereby, therefore, thus, as, because, for, on the grounds that, since, so that, is why

Concession: however, nevertheless, nonetheless, though (adv.), yet, although, though (conj.), even though, even if, albeit, despite, in spite of, notwithstanding

Reformulating – paraphrasing and clarifying: i.e., that is, that is to say, in other words, namely, viz., or more precisely, or more accurately, or rather


4.2. The function of exemplication

This section presents a detailed analysis of the academic words that are used by expert writers to serve the rhetorical function of exemplifi cation. Siepmann (2005) showed that exemplifi catory discourse markers occur in all kinds of discursive prose, and are particularly frequent in humanities texts. He argued, however, that as an object of study, ‘exemplifi cation con-tinues to be the poor relation of other rhetorical devices’ and that ‘such neglect has led to a commonly held view in both the linguistic and the pedagogic literature that exemplifi cation is a minor textual operation, sub-ordinate to major discoursal stratagems such as “inferring” and “proving”’ (Siepmann, 2005: 111). Coltier (1988) remarked that examples and exem-plifi cation merit close investigation at two levels: the exemplifi catory strate-gies adopted (i.e. when and why are examples introduced into a text); and the wording of the example (i.e. the choice of exemplifi ers). This section deals with the latter and focuses on the lexical items used by expert writers to give an example. For a rhetorical perspective on exemplifi ers in native writing, see Siepmann (2005: 112–18).

The Academic Keyword List (AKL) includes a number of words and multi-word sequences that are commonly used as exemplifi catory discourse mark-ers: the mono-lexemic or word-like units for example and for instance, the noun example, the verbs illustrate and exemplify, the preposition such as, the adverb notably and the abbreviation e.g. Other lexical items commonly listed in textbooks and EAP/EFL materials, but not found in the AKL, are the expressions by way of illustration and to name but a few, the nouns illustration and a case in point and the preposition like. Table 4.1 gives the absolute fre-quencies of these words in the BNC-AC-HUM as well as their relative fre-quencies per 100,000 words and the percentage of exemplifi catory discourse markers they represent. In Figure 4.1, the lexical items are ordered by decreasing relative frequency in the academic corpus. The most frequent exemplifi ers in professional academic writing are the mono-lexemic phrase-mes such as and for example, plus the noun example, which occur more than 35 times per 100,000 words. Almost half of the exemplifi ers — for instance, like, illustrate, e.g. and notably — occur with a relative frequency of between 5 and 20 occurrences per 100,000 words. The verb exemplify and the noun illustration are less frequent (around 2.3 occurrences per 100,000 words) while the adverbials to name but a few and by way of illustration as well as the noun case in point appear very rarely in the BNC-AC-HUM.

I will now discuss my main fi ndings on the exemplifi catory functions of prepositions, adverbs and adverbial phrases, and then focus on the exemplifi catory use of nouns and verbs.


Table 4.1 Ways of expressing exemplifi cation found in the BNC-AC-HUM

Abs. freq. % Rel. freq.

Nouns

example 1285 21.6 38.7illustration 77 1.3 2.3(BE) a case in point 18 0.3 0.5TOTAL NOUNS 1380 23.2 41.5

Verbs

illustrate 259 4.4 7.8exemplify 79 1.3 2.4TOTAL VERBS 338 5.7 10.2

Prepositions

such as 1494 25.0 45.0like 532 8.9 16.0TOTAL PREP. 2026 34.0 61.0

Adverbs

for example 1263 21.2 38.0for instance 609 10.2 18.3e.g. 259 4.3 7.8notably 77 1.3 2.3to name but a few 4 0.1 0.1by way of illustration 3 0.1 0.1TOTAL ADVERBS 2215 37.2 66.7

TOTAL 5959 100 179.4

50454035302520151050

such

as

exam

ple

for e

xam

ple

for i

nsta

nce

illustr

ate

nota

bly

exem

plify

illustr

ation

BE a ca

se in

poin

t

to n

ame

but a

few

by w

ay o

f illus

tratio

ne.

g.like

Figure 4.1 Exemplifi cation in the BNC-AC-HUM


4.2.1. Using prepositions, adverbs and adverbial phrases to exemplify

As shown in Figure 4.1, the complex preposition such as is the most frequent exemplifi er in the BNC-AC-HUM (see Example 4.1). Unlike in other genres (such as speech and fi ction), it is much more frequent than the preposition like in professional academic writing (Example 4.2).

4.1. This is the arrangement in Holland whereby various institutions such as media, schools, cultural organisations, welfare services, and hospitals are duplicated, and run by the separate catholic and protestant communities.

4.2. Surrealist painting had publicity value, especially when executed by a showman like Salvador Dali, who married the former wife of the poet Paul Éluard.

Similarly, for example is twice as frequent as for instance. These two adverbials are commonly classifi ed as ‘code glosses’ in metadiscourse theory (see Section 1.3) as they are used to ‘supply additional information, by rephrasing, explaining, or elaborating what has been said, to ensure the reader is able to recover the writer’s intended meaning’ (Hyland, 2005: 52). Code glosses are ‘interactive resources’ in Hyland’s typology of metadiscourse: they are features used to ‘organize propositional information in ways that a projected target audience is likely to fi nd coherent and convincing’ (ibid, 50). In a phraseological approach to academic vocabulary, they fall into the category of textual phrase-mes as they are mono-lexemic multi-word units, i.e. multi-word units that are equivalent to single words and which fi ll only one grammatical slot, with an organizational – exemplifi catory – function. In the BNC-AC-HUM, for example and for instance are typically used within the sentence, enclosed by commas, especially after the subject. But they can also follow the subject of the exempli-fying sentence, while remaining essentially cataphoric in nature (i.e. pointing forward to the example) as shown in Examples 4.3 and 4.4.

4.3. Such associations of sexual deviance and political threat have a long history sedimented into our language and culture. The term ‘buggery’, for example, derives from the religious as well as sexual nonconformity of an eleventh-century Bulgarian sect which practised the Manichaean heresy and refused to propagate the species; the OED tells us that it was later applied to other heretics, to whom abominable practices were also ascribed.

4.4. The small mammals living today in many different habitats and climatic zones have been described, so that the associations between faunal types and ecology are well documented [ . . . ]. Woodland faunas, for instance, are distinct from grassland faunas, and tropical faunas distinct from temperate faunas, and when these and more precise distinctions are made it is possible to correlate and even defi ne ecological zones by their small mammal faunas.


For example and for instance can also function as endophoric markers and refer back to an example given before, as illustrated in Example 4.5. This use is, however, much less frequent (see Table 4.2).

4.5. Thirdly, the debates over how far to forge a strategy either for winning power or for promoting economic development in a post-revolutionary society have not been satisfactorily resolved, and indeed perhaps cannot be, given that counter-revolutionary response to any successful formula will ensure that it will be that much more diffi cult to apply the same tactics in another situation. Such is the relation which Nicaragua bears to El Salvador, for example.

In Mieux écrire en anglais, Laruelle (2004: 96–7) writes that for example should be placed in the initial position if the whole sentence has an exem-plifi catory function, while the adverbial should follow the subject, between commas, if only the subject is the example. This statement, however, is not confi rmed by corpus data. Example 4.3 clearly shows that for example need not be placed in the initial position to introduce an exemplifi catory sentence.

Like nouns and verbs, mono-lexemic adverbial phrases can also have their own phraseological patterns. Three verbs, i.e. consider (f[n, c]1 = 13; log-likelihood = 92.5), take (f[n, c] = 7; log-likelihood = 19.1) and see (f[n, c] = 19; log-likelihood = 71.7) are signifi cant left co-occurrents of for example in the BNC-AC-HUM. They are used in the second person of the imperative. The verbs consider and take are typically used with for example to introduce an example that is discussed in further detail over several sentences:

4.6. It is worth pausing here momentarily to observe that such legally provided rem-edies can be morally justifi ed even when applied to people who are not subject to the authority of the government and its laws. Consider for example the law of defamation. Assuming that it is what it should be, it does no more than incor-porate into law a moral right existing independently of the law. The duty to compensate the defamed person is itself a moral duty. Enforcing such a duty against a person who refuses to pay damages is morally justifi ed because it

Table 4.2 The use of ‘for example’ and ‘for instance’ in the BNC-AC-HUM

Cataphoric marker Endophoric marker

for example 1,185 (93.8%) 78 (6.2%)for instance 588 (96.5%) 21 (3.5%)


implements the moral rights of the defamed. One need not invoke the authority of the law over the defamer to justify such action. The law may not have authority over him.

4.7. But the concept of compresence is far from clear. If it implies that no time-lag is detectable between elements of an experienced “complex”, then this is true only in a very limited sense. Take, for example, the perceptual experience that I have while looking at this bunch of carnations arranged in a vase on the table in the middle of the room. I see this “complex” as one whole. But while I am looking at it my eyes constantly wander from one fl ower to the next, pausing at some, ignor-ing others, picking out the details of their shapes and colours. Finally, without taking my eyes off the fl owers, I may move the vase closer, or walk around the table and look at the fl owers from different angles. The scene will keep constantly changing. As a result, I shall experience a succession of different “complexes of qualities” but I shall still be looking at the same bunch of fl owers.

Hyland describes this type of imperative as directives with a rhetorical purpose that ‘can steer readers to certain cognitive acts, where readers are initiated into a new domain of argument, led through a line of reason-ing, or directed to understand a point in a certain way’ (2002a: 217). He categorizes them as interactional resources, and more specifi cally as engagement markers, i.e., ‘devices that explicitly address readers, either to focus their attention or include them as discourse participants’ (Hyland, 2005:53).

The verb see is frequently used in professional academic writing as an endophoric marker to refer to tables, fi gures, or other sections of the article or to someone else’s ideas or publications (Hyland, 1998, 2002a, 2005; Hyland and Tse, 2007). The use of the second person imperative see ‘allow[s] academic writers to guide readers to some textual act, referring them to another part of the text or to another text’ (Hyland, 2002a: 217). In the BNC-AC-HUM, 63 per cent of the occurrences of the sequence see for example appear between brackets as in Example 4.8:

4.8. Afro-Caribbean and Asian children are indeed painfully aware that many teachers view them negatively and some studies have documented reports of routine racist remarks by teachers (see for example Wright in this volume).

Swales et al. (1998) examined a corpus of research articles in ten disciplines (art history, chemical engineering, communication studies, experimental geology, history, linguistics, literary criticism, philosophy, political science and statistics) and found that second person imperative see was the most


frequent imperative form across disciplines. Similarly, in his study of directives in academic writing, Hyland (2002a) analysed a corpus of 240 published research articles, seven textbook chapters and 64 project reports written by fi nal year Hong Kong undergraduates and found that the second person imperative see represented 45 per cent of all imperatives in that corpus. Note that in both studies, the use of the imperative varied across disciplines.

The advantage of adopting a phraseological approach to rhetorical func-tions, and hence metadiscourse resources, appears quite clearly here. The sequences take/consider for example consist of two metadiscourse resources in Hyland’s (2005) categorization scheme: the imperative forms take and con-sider are interactional resources, and more specifi cally engagement mark-ers, while for example is a code gloss. Similarly, see is an endophoric marker in see for example. In our phraseological framework, the sequences take/consider/see for example are textual phrasemes as they form functional — textual — units and display arbitrary lexical restrictions.

The adverb notably can be regarded as a typical academic word: Figure 4.2 shows that it is much more frequent in academic writing than in the other genres. It is typically preceded by a comma (Example 4.9) and is qualifi ed by the adverb most in 15.2 per cent of its occurrences in the BNC-AC-HUM (Example 4.10).

4.9. Some bishops, notably Jenkins of Durham, Sheppard of Liverpool, and Hapgood of York, have spoken out about deprivation in the inner cities, the miners’ strike, and the need for government to show a greater compassion for, and understanding of, the poor.

Figure 4.2 The distribution of the adverb ‘notably’ across genres

50

4040

30

freq

. per

mill

ion

wor

ds

20

10

academicwriting

news fiction speech0


4.10. At leading public schools, most notably Eton, there is a tradition of providing MPs, government ministers, and prime ministers.

The abbreviation e.g. (or less frequently eg) stands for the Latin ‘exempli gratia’ and means the same as for example. It is quite common in the BNC-AC-HUM, in which 65.7 per cent of its occurrences are between brackets:

4.11. Direct curative measures (e.g. fl ood protection) are clearly within the domain of a soil conservation policy.

In contrast to for example and for instance, the great majority of occurrences of e.g. introduce one or more noun phrases rather than full clauses:

4.12. It may help to refer the patient to other agencies (e.g. social services, a psycho-sexual problems clinic, self-help groups).

When e.g. is used without brackets, it is preceded by a comma:

4.13. Primary industries are those which produce things directly from the ground, the water, or the air, e.g. farming.

As shown in Figure 4.1, the textual phrasemes by way of illustration and to name but a few are quite rare in the BNC-AC-HUM. In fact, these expres-sions are very infrequent in all types of discourse. Figures 4.3 and 4.4 show the distribution of the two phrasemes in four main genres of the British

Figure 4.3 The distribution of ‘by way of illustration’ across genres

1

0.5

freq

. per

mill

ion

wor

ds

0

acad

emic

news

fictio

n

spee

ch


National Corpus (BNC), namely academic writing, fi ction, newspaper texts, and speech. Some 36 per cent of the occurrences of by way of illustration in the BNC (i.e. 10 out of a total of 28) appear in academic texts, and only one occurrence comes from speech. The expression to name but a few is more frequent than by way of illustration in the whole BNC, but only 12.8 per cent of its occurrences (10 out of 78) appear in academic writing. No instance of to name but a few was found in speech.

4.2.2. Using nouns and verbs to exemplify

Nouns and verbs are used to give examples in specifi c phraseological patterns. The noun which is most frequently used in this way is example, which is much more common than illustration or a case in point. Table 4.3 shows that it is as frequently used as its connective counterpart, the textual phraseme for example, in the BNC-AC-HUM.

The signifi cant verb co-occurrents of the noun example in the BNC-AC-HUM are listed in Table 4.4. The verb be is the most frequent verb co-occurrent of example in windows of one to three words to both the left

1

0.5

freq

. per

mill

ion

wor

ds

0

acad

emic

news

fictio

n

spee

ch

Figure 4.4 The distribution of ‘to name but a few’ across genres

Table 4.3 The use of ‘example’ and ‘for example’ in the BNC-AC-HUM

example for example

Absolute freq. % Absolute freq. %

BNC-AC-HUM 1285 50.43 1263 49.57


(3L-1L) and the right (1R-3R). Be is, however, twice as frequent in the left window. When example is preceded by the verb be, it mainly functions as a retrospective label, i.e. it refers back to the exemplifying element which is given as the subject. The noun example may refer back directly to a noun phrase (Example 4.14) or to the demonstrative pronoun this which further points to a previous exemplifying sentence (Example 4.15).

4.14. Vision is a better example of a modular processing system. 4.15. The designer at Olympia chose to represent the race by the moment

before it started, as Polygnotos showed the sack of Troy in its aftermath. This is the supreme surviving example of the early classical taste for stillness and indirect narrative.

By contrast, when the noun example is introduced by there + BE (11%) or here + BE (15%), it functions as an advance label which refers forward to a following example (underlined):

4.16. In addition, of course, choices can result from lengthy weighing of odds. Here is a simple example of the complexity at issue. I am driving along a narrow main road, used by fast-moving traffi c, with my children in the back seat. A car some distance ahead strikes a large dog but does not stop, leaving the creature walking-wounded but in obvious distress.

Table 4.4 Signifi cant verb co- occurrents of the noun ‘example’ in the BNC-AC-HUM

Left co-occurrents Right co-occurrents

Verb freq. Verb freq.

be 139 be 84provide 26 illustrate 14take 29 show 21give 12 give 15cite 5 suggest 12consider 12 quote 6illustrate 7 include 7show 9 provide 8see 10 concern 6serve 5

will 16can 15

would 13


My children, seeing what occurred, cry out. I glance in the rear-view mirror to see other cars close behind; slowing down but then speed-ing up again. I do not stop.

When example is the subject of the verb be, it always functions as an advance label. It is often qualifi ed by an adjective (see Examples 4.17 to 4.19) and the exemplifi ed item is generally introduced by the preposition of (Examples 4.18 and 4.19). In Example 4.19, the exemplifi ed item is the pronoun this which refers back to the previous sentences.

4.17. The prime example is the Dada movement, whose nihilistic work is now admired for its qualities of imagination.

4.18. The clearest example of emotive language is poetry, which is entirely concerned with the evocation of feelings or attitudes, and in which the writer’s and reader’s attention is not, or should not be, directed at any of the objective relationships between words and things.

4.19. Until the seventeenth century many, even most, European frontiers were very vague, zones in which the claims and jurisdictions of different rulers and their subjects overlapped and intersected in a complex and confusing way. This was especially true in eastern Europe, where many states were large and central governments were usually less effective at the peripheries of their territories than in the west. The most striking example of this is perhaps the frontier in the Danubian plain between the Ottoman empire and the Habsburg territories in central Europe.

Copular clauses using the noun example consist of textual sentence stems (An example of Y is . . . ) and rhemes (. . . is an example of Y). Textual sentence stems are routinized fragments of sentences which serve specifi c textual or organizational functions. They consist of sequences of two or more clause constituents, and typically involve a subject and a verb, e.g. An example of Y is . . . . They typically have an empty slot for the following object or complement. Rhemes typically consist of a verb and its post-verbal elements, which do not contain any thematic element (e.g. . . . is another issue).

Four other verbs, namely provide, give, illustrate and show (given in italics in Table 4.4), are signifi cant left and right co-occurrents of the noun example. The verbs take, cite, consider, see and serve are only signifi cant left co-occurrents, while the verbs suggest, concern, quote and include and the modals will, can and would are signifi cant right co-occurrents. The verbs provide, take, give, cite, consider, see, serve and include often co-occur with the noun example to form textual — exemplifi catory — phrasemes. The verb


provide can be used in active or passive structures, but active structures in which the subject is the example (Example 4.20) are more frequent. The verb cite is more often found in a passive structure in which example functions as a retrospective label (Example 4.21). The two verbs often form rhemes with the noun example:

4.20. The Magdalen College affair, for example, provides a classic example of passive resistance.

4.21. A famous passage of art criticism can be cited as one example entirely beyond dispute.

The verb take is mainly used with the noun example in sentence-initial exemplifi catory infi nitive clauses (68.9%; Example 4.22). It also occurs in active structures with a personal pronoun subject (13.79%; Example 4.23) and in imperative sentences (13.79%). When used in the imperative, it generally appears in the second person (Example 4.24) and there is only one occurrence in the fi rst person plural in the BNC-AC-HUM. By contrast, the verb consider is mainly used with the noun example in imperative sentences (70%), usually second person imperatives (Example 4.25) and less frequently fi rst person plural. The verb see always co-occurs with the noun example in the second person of the imperative (Example 4.26). It is never used to introduce an example, but always as an endophoric marker to direct the reader’s attention to an example elsewhere in the text.

4.22. To take one example, at the beginning of the project seven committees were established, each consisting of about six people, to investigate one of a range of competing architectural possibilities.

4.23. In accordance with the theme of this chapter, I shall simply use ‘stylistics’ as a convenient label (hence the inverted commas) for the branch of literary studies that concentrates on the linguistic form of texts, and I shall take four different examples of this kind of work as alternatives to the Prague School’s and Jakobson’s approach to the relationship between linguistics and literature.

4.24. Take the example of following an object by eye-movements (so-called ‘tracking’).

4.25. Consider the following example. 4.26. The most important vowel is set to two or more tied notes in a phrase designed

to increase the lyrical expression (see Example 47, above).

The verb include is used with the plural form of the noun example in subject position to introduce an incomplete list of examples in object position:


4.27. The fl oral examples include a large lotus calyx and two ivy leaves joined by a slight fi llet.

Another set of verb co-occurrents of example is used to discuss the examples given in a text. These include quote (Example 4.28), suggest and show (Example 4.29) to talk about conclusions that can be drawn from the examples, and illustrate (Example 4.30) to show what something is like or that something is true.

4.28. Thirdly, in all the examples quoted here, there is a sense in which all observers see the same thing.

4.29. The example shows that the objector’s neat distinction between adjudicative and legislative authorities is mistaken.

4.30. This example clearly illustrates the theory dependence and hence fallibility of observation statements.

These signifi cant co-occurrences illustrated in Examples 4.28 to 4.30 do not qualify as collocations as the meaning of the verb is not restricted by the noun example, and the combinations are fully explicable in semantic and syntactic terms. However, these co-occurrences are frequently used in adverbial clauses (e.g. as this example suggests . . . ) and sentence stems (e.g. this example [adv.] illustrates . . .) which describe examples, give more detail about them, and make suggestions on their basis.

The advantage of using the noun example rather than the adverbials for example or for instance is that it allows the writer to evaluate the example in terms of its suitability, e.g. good, outstanding, fi ne, excellent (Example 4.31) or typicality, e.g. classic, typical, prime (Example 4.32). The adjectives above and following are used to situate the example in the text (Example 4.33): used with the noun example, they function as endophoric markers in Hyland’s (2005) typology of metadiscourse features. Table 4.5 gives the 24 adjectives that signifi cantly co-occur with the noun example in the BNC-AC-HUM.

4.31. An outstanding example of this type of narrative is Vargas Llosa’s Conversa-tion in the Cathedral, which pivots around a four-hour conversation between two characters, the whole novel being made up of dialogue and narrative units generated in waves by the central conversation, as the two men’s review of their past lives sparks off inner thoughts and recollections and conjures up other conversations and dramatised episodes.

4.32. The prime example is the Dada movement, whose nihilistic work is now admired for its qualities of imagination.


4.33. Consider the following example.

There is a case for considering the co-occurrence classic example as a free combination, i.e. a word combination that is semantically fully composi-tional, syntactically fully fl exible and collocationally open: the adjective clas-sic is used with a meaning that is listed as its fi rst sense in the Longman Dictionary of Contemporary English (LDOCE4) (1. TYPICAL: having all the features that are typical or expected of a particular thing or situation) and the Oxford English Dictionary Online2 (1. of the fi rst class, of the highest rank or importance; approved as a model; standard, leading). However, the adjective is only commonly used in this sense with a very limited number of nouns— example, mistake and case3. This is clearly an illustration of the diffi culty of separating the senses that a word has in isolation from those that it acquires in context (see Barkema, 1996). Following Granger and Paquot (2008a: 43), I classifi ed co-occurrences of this type as collocations, i.e. usage-determined or preferred syntagmatic relations between two lexemes in a specifi c syntactic pattern. Both lexemes make a separate semantic contribution to the word combination but they do not have the same status. The ‘base’ of a collocation is semantically autonomous and is selected fi rst by a language user for its independent meaning. The second element, i.e. the ‘collocate’ or ‘collocator’, is selected by and semantically dependent on the ‘base’. The co-occurrence prime + example is a clear example of a collocation: the adjective prime has two core meanings – ‘most important’ and ‘of the very best quality or kin’ – but a prime example is ‘a very typical example of sth’. Collocations represent 8.3 per cent of the types and 6.87 per cent of the tokens of adjective + example co-occurrences.

Table 4.5 Adjective co-occurrents of the noun ‘example’ in the BNC-AC-HUM

Adjective freq. Adjective freq.

good 38 fi ne 9above 15 notable 8following 18 isolated 8well-known 10 interesting 9obvious 16 known 7classic 11 excellent 6typical 13 prime 7outstanding 10 trivial 5extreme 12 previous 6clear 16 remarkable 5simple 13 numerous 5striking 9 single 6


Other adjectives form semantically and syntactically fully compositional sequences with the noun example. Thus, the meaning of an outstanding exam-ple is composed of the meanings of the adjective outstanding and the noun example. This does not mean, however, that they are pedagogically uninter-esting. First, they constitute ‘preferred ways’ of qualifying example as they are repeatedly used with this noun. Second, in her study of verb + noun combinations, Nesselhauf (2005) has shown that free combinations are prone to erroneous or, at least, unidiomatic use in learners’ writing. Similarly, Lorenz (1998; 1999a) has pointed out that German learners’ use of adjectives, irrespective of their phraseological status, differs from that of native students.

The added value of using statistics, and more specifi cally association measures, to analyse the common co-occurrences of a word in a large corpus is made clear by comparing the signifi cant adjective co-occurrents of the noun example (listed in Table 4.5) with attested adjectival collocates (as given by Siepmann, 2005: 137). In addition to most of the adjectives given in Table 4.5, Siepmann listed a number of adjectives that do not appear even once in the 87-million word written part of the BNC (beguiling, consummate, eminent, apposite, anodyne, happy, alarming, crass, cautionary), and adjectives which occur only once or twice in the corpus (exquisite, well-worn, edifying, emotive, awe-inspiring, glittering, hideous). To use Sinclair’s (1999: 18) words, these co-occurrents are best described as ‘singularities’ and do not represent ‘the habitual usages of the majority of users’.

Apart from verbs and adjectives, other signifi cant co-occurrents of the noun example are found in professional academic writing. Left co-occurrents include determiners and the pronoun this. Indefi nite determin-ers (a, another and one) are more frequent than the defi nite article the with example. The is mainly used when the noun is qualifi ed by a superlative adjective or preceded by ordinals such as fi rst, next and last (Example 4.34).

4.34. The fi rst two examples discussed below illustrate different ways in which the linguistic model is used to develop a narrative model, and (. . . ).

The pronoun this is typically used as a subject with the verb be to refer back to an example given in a previous sentence (see Example 4.15 above). Right co-occurrents include the preposition of and the pronoun this. In 40 per cent of its occurrences, the noun example is directly followed by the preposition of which introduces the idea, class or event exemplifi ed, which in turn is often determined by a demonstrative (Example 4.31 above) or pronominalized to refer back to a previous sentence.


These fi ndings support Gledhill’s (2000) view that there may be a very specifi c phraseology and set of lexico-grammatical patterns for function words in academic discourse. Function words seem to display co-occurrence preferences just as content words do (also see Renouf and Sinclair’s (1991) notion of a ‘collocational framework’). These fi ndings also provide strong evidence against the use of stopword lists when extracting co-occurrences from corpora as there is a serious danger of missing a whole set of phraseo-logical patterns (Clear, 1993).

The verbs illustrate and exemplify can also be used as exemplifi ers. The verb illustrate is used with the meaning of ‘to be an example which shows that something is true or that a fact exists’ (Example 4.35) or ‘to make the meaning of something clearer by giving examples’ (Example 4.36) (LDOCE4). The verb exemplify is used with the meanings of ‘to be a very typical example of something’ and ‘to give an example of’.

4.35. The narratives of the Passio Praeiecti and of the Vita Boniti both have their peculiarities, and it is possible that the appointment of Praeiectus and the retirement of Bonitus were less creditable than their hagiographers claim. Nevertheless they do illustrate the complexities of local ecclesiastical politics.

4.36. My aim will be to illustrate different ways of approaching literature through its linguistic form, ways involving the direct application of linguistic theory and linguistic methods of analysis in order to illuminate the specifi cally literary character of texts.

Both verbs are more frequent in academic writing than in any other genre. Figure 4.5 compares the relative frequencies of the two verbs in academic writing with three main genres represented in the British National Corpus. The verb illustrate is not uncommon in news but a quick look at its concor-dances shows that a signifi cant proportion of its occurrences are used not to introduce an example, but with the meaning of ‘to put pictures in a book, article, etc’ (Example 4.37). Exemplify is very rarely used in other genres.

4.37. Also in the pipeline is an Australian children ‘s TV series based on Gumnut Factory Folk Tales (written, illustrated and published by Chris Trump). (BNC-NEWS)

Figure 4.5 also shows that the verb illustrate is more frequent than exemplify in professional academic writing. The frequencies of the two verb lemmas, their word forms and tenses in the BNC-AC-HUM4 were computed in the way described by Granger (2006). Table 4.6 shows that there is no


major difference in proportion between the verb forms illustrate, illustrated and illustrates. When used in active structures, the verb is often preceded by a non-human subject such as example, fi gure, table, case or approach (Example 4.38). Almost all occurrences of the past participle appear in the passive construction BE illustrated by/in (Example 4.39).

140

120

100

80

60

freq

uenc

y pe

r m

illio

n w

ord

40

20

Academic News Fiction Speech0

illustrate exemplify

Figure 4.5 The distribution of the verbs ‘illustrate’ and ‘exemplify’ across genres

Table 4.6 The use of the lemma ‘illustrate’ in the BNC-AC-HUM

The lemma illustrate BNC-AC-HUM

illustrate

simple presentinfi nitive

97

3661

37.45%

13.89%23.55%

illustrated

simple pastpresent/past perfectpast participle

84

70

77

32.43%

2.7%0%

29.73%

illustrates 63 24.32%

illustrating

continuous tense-ing clause

15

213

5.79%

0.77%5%

Total 259 100%

Nr of words 3,321,867

Relative freq. per 100,000

words

7.8


4.38. This example clearly illustrates the theory dependence and hence fallibility of observation statements.

4.39. The contrast between the conditions on the coast and in the interior is illus-trated by the climatic statistics for two stations less than 30 km (18.5 miles) apart.

The sentence-initial adverbial clause To illustrate this/the point/X, . . . (Example 4.40) represents 2.7 per cent of the occurrences of the lemma illustrate in the BNC-AC-HUM.

4.40. How many observations make up a large number? (. . . ) Whatever the answer to such a question, examples can be produced that cast doubt on the invariable necessity for a large number of observations. To illustrate this, I refer to the strong public reaction against nuclear warfare that followed the dropping of the fi rst atomic bomb on Hiroshima towards the end of the Second World War.

In the BNC-AC-HUM, illustrate signifi cantly co-occurs with the noun exam-ple [LogL = 112] in a 3L-1L window, and with the nouns point [LogL = 168.78], example [LogL = 49.65] and fi g. [LogL = 45.08] in a 1R-3R window. The noun point is used as an object of illustrate which refers back to an idea put forward in a previous sentence:

4.41. For most of this century it is those disorders gathered together under the head-ing of ‘schizophrenia’ that have been used as the paradigm for trying to describe and understand psychosis. Yet even in this form, or forms – for many would prefer to talk of ‘the schizophrenias’ – there is still no universally accepted set of criteria for diagnosis. To illustrate the point, one of the present authors was recently asked to review a paper submitted to a prominent psychiatric journal, proposing a new set of rules for diagnosing schizophrenia. In the course of their analysis the authors determined the extent to which their proposed criteria agreed with those contained in other existing diagnostic schemes – some ten or twelve of them. Correlations varied over a very wide range.

The noun fi gure (and the abbreviation fi g.) is used either as the subject of the verb illustrate or in the passive structure illustrated in Figure x. This co-occurrence is even more marked in academic genres such as social sciences, natural sciences and medicine which rely extensively on fi gures, tables and diagrams (see Examples 4.42 and 4.43).

4.42. Figure 1 illustrates the spread of results for the alcoholics and the controls. (W_ac_medicine BNC sub-corpus, see Table 3.2)


4.43. The advantages of the system are illustrated in Fig. 8.2 and, like the Peruvian example discussed above, the fallow stage is contributing to crop productivity as well as providing protection against soil erosion. (W_ac_soc_science BNC sub-corpus, see Table 3.2)

The adverbs well, better, best and clearly are sometimes used with illustrate to evaluate the typicality or suitability of the example (Example 4.44). The verb illustrate also co-occurs signifi cantly with how to introduce a clause (Example 4.45), with the verb serve (Example 4.46), and with the modals will, can and may (Example 4.47).

4.44. The history of the English monarchy well illustrates both the importance and the unimportance of war.

4.45. We recently did a simple experiment which happens to illustrate how children’s knowledge of where an object is determines their behaviour.

4.46. While our discussion in this chapter is of the doctrine of neutrality as such, Rawls ‘ treatment of it will serve to illustrate the problems involved.

4.47. This prejudice against close involvement with the secular government may be illustrated by an anecdote related in the about Molla Gurani.

Table 4.7 shows that the lexico-grammatical preferences of the verb exemplify differ from those of illustrate. A large proportion of the occurrences

Table 4.7 The use of the lemma ‘exemplify’ in the BNC-AC-HUM

The lemma exemplify BNC-AC-HUM

exemplify

simple presentinfi nitive

9

54

11.4%

6.33%5%

exemplifi ed

simple pastpresent/past perfectpast participle

53

81

44

67%

10%1.26%55.7%

exemplifi es 15 19%

exemplifying

continuous tense-ing clause

2

02

2.53%

0%2.53%

Total 79 100%

Nr of words 3,321,867

Relative freq. per 100,000

words

2.38


of the lemma exemplify are –ed forms, and more precisely past participle forms, of the verb. In the BNC-AC-HUM, the verb signifi cantly co-occurs with the verb be and the conjunction as in a 3L-1L window, and with the prepositions by and in in a 1R-3R window. These signifi cant co-occurrents highlight the preference of the verb for the passive structure BE exemplifi ed by/in (Example 4.48) and the lexico-grammatical pattern as exemplifi ed by/in (Example 4.49). Exemplify is also often used after a noun phrase, preceded by a comma (Example 4.50). Unlike illustrate, the verb exemplify does not co-occur signifi cantly with nouns.

4.48. The association of this material with the clerk is clearly exemplifi ed by Chaucer’s wife of Bath’s fi fth husband, the clerk Jankyn, who, in the Wife of Bath’s Prologue, reads antifeminist material to her from his book Valerie and Theofraste.

4.49. He assumed, without argument, that science, as exemplifi ed by physics, is supe-rior to forms of knowledge that do not share its methodological characteristics.

4.50. Piaget’s claim that thinking is a kind of internalised action, exemplifi ed in the assimilation-accommodation theory of infant learning mentioned above, is really a global assumption in search of some refi ned, detailed and testable expression.

4.2.3. Discussion

The description of exemplifi ers presented here does not aim at exhaustive-ness in professional academic writing but at typicality. The corpus-based methodology adopted has highlighted a number of lexical items that are repeatedly used as exemplifi ers in academic writing. The function of exem-plifi cation can be fulfi lled by a whole spectrum of single words (the preposi-tion like, the adverb notably, the abbreviation e.g.) and word combinations, i.e. word-like units or mono-lexemic phrasemes (the preposition such as, the adverbials for example and for instance), sentence stems (An example of Y is X; Examples include . . .) and rhemes (… is an example of . . .; . . . provides a classic example of . . .), imperative clauses (Consider, for example . . .) and sentence-initial infi nitive clauses (To take one example, . . .). A large majority of these word combinations are semantically and syntactically fully compo-sitional; the exceptions are a few collocations such as prime example and classic example. They are, however, characterized by their high frequency of use and can be described as ‘preferred ways’ (Altenberg, 1998) of giving an example in professional academic writing.

Siepmann (2005) analysed a 9.5-million word corpus of academic writing, but did not make use of statistical methods. He enumerated every


single occurrence of word sequences used to give an example and listed rare events such as the infi nitive clauses to paint an extreme example and to pick just one example (a single occurrence in his corpus), the co-occurrence example + is afforded by and the expression for the sake of example. It may be argued that privileging exhaustiveness over typicality in corpus linguistic research is counter-productive, and that such an approach results in too much — unreliable — information. Siepmann, for example, wrote that

English authors have a large range of exemplifi catory imperatives at their disposal, using the direct second-person imperative VP ~ as well as the less imposing hortative let us + VP and the inclusive let me + VP. Of these last two, the former is around fi ve times more frequent than the latter, showing a high degree of audience sensitivity among authors. (Siepmann, 2005:120)

A closer look at his frequency data (reprinted in Table 4.8) shows, however, that the co-occurrences see/take/consider + for example account for 89.4 per cent of the imperatives Siepmann found. First person imperatives are extremely rare and let me + VP only appeared three times in the 9.5-million word corpus of professional academic writing he used. Although a large range of exemplifi catory imperatives may be available to language users, only a very limited set of these are widespread in professional academic writing.

Table 4.8 The use of imperatives in academic writing (based on Siepmann, 2005: 119)

Imperatives in academic writing Frequency %

(for example/for instance) see (for example/for instance) NP(for example) consider (for example) NPtake, for (another) example, NPConsider a(n) (ADJ) example/instancetake the example of (as examples of NP)consider (as an example) NPtake, as an example, NPas an illustration (of this)/ by way of (brief) illustration,

consider NP (2)Take (even) NP (2)Let us (now) take + (as) + DET + ADJ + example(s)Let us consider + DET + ADJ + example(s)Let me give (you) (but) one example Let me offer + DET (+ ADJ+) example Let us consider, for the sake of illustration, NP

200541675312244211

66.217.95.32.31.71

0.30.70.71.31.30.70.30.3

Total 302 100


The analysis of exemplifi ers presented here also validates the method used to design the Academic Keyword List. The exemplifi catory lexical items which were extracted are of two types:

— the most frequent exemplifi ers in academic writing (such as, example, for example and for instance) (see Figure 4.1 discussed earlier in this chapter);

— lexical items which are not as frequent as such as, example, for example and for instance, but which are more frequent in academic prose than in any other genres (illustrate, exemplify, e.g. and notably).

The preposition like can be used to fulfi l an exemplifi catory function in academic writing but it is much more common in other genres. The nouns illustration and case in point are quite characteristic of formal textual genres, but they are infrequent. The expressions to name but a few and by way of illustration are rare in all types of discourse.

4.3. The phraseology of rhetorical functions in expert academic writing

This section briefl y comments on the types of lexical devices used by expert writers to serve the functions of ‘expressing cause and effect’, ‘comparing and contrasting’, ‘expressing a concession’ and ‘reformulating’ in an attempt to give a wider overview of the way academic vocabulary is used to serve specifi c rhetorical functions. It aims to characterize the phraseology of these rhetorical functions in academic prose.

Table 4.9 shows that the lexical means of expressing a concession consist of single word adverbs (e.g. however, nevertheless, yet), (complex) conjunc-tions (e.g. although, even though) and (complex) prepositions (e.g. despite, in spite of). Similarly, reformulation is most frequently achieved by means of the mono-lexemic units that is and in other words, the abbreviation i.e. and the adverb namely (Table 4.10).

Adverbs, prepositions and conjunctions also represent a large proportion of the lexical devices used by expert writers to serve the functions of ‘expressing cause and effect’ (Table 4.11) and ‘comparing and contrasting’ (Table 4.12). However, these two functions can also be realized by means of nouns, verbs and adjectives in specifi c phraseological or lexico-grammatical patterns. As shown in Table 4.11, nouns account for 32.5 per cent of the lexical means used to express a cause or an effect in academic writing, e.g. cause, factor, source, effect, result, consequence, outcome and implication.


Table 4.9 Ways of expressing a concession in the BNC-AC-HUM


Adverbs

however 3,353 28.6 100.9nevertheless 676 5.8 20.3nonetheless 66 0.6 2.0though ADV 144 1.2 4.3yet 1,817 15.5 54.7

TOTAL ADVERBS 6,056 51.6 182.3

Conjunctions

although 2,292 19.5 69.0though CONJ 1,721 14.7 51.8even though 248 2.1 7.5(even if) 451 3.8 13.6albeit 80 0.7 2.4

TOTAL CONJ. 4,792 40.86 144.26

Prepositions

despite 681 5.8 20.5in spite of 159 1.4 4.8notwithstanding 39 0.3 1.2

TOTAL PREP. 879 7.5 26.46

TOTAL 11,727 100 353

Table 4.10 Ways of reformulating, paraphrasing and clarifying in the BNC-AC-HUM


i.e. 330 25.1 9.9that is 375 28.5 11.3that is to say 81 6.2 2.4in other words 210 16.0 6.3namely 187 14.2 5.6viz. 21 1.6 0.6or more precisely 12 0.9 0.4or more accurately 7 0.5 0.2or rather 91 6.9 2.7

TOTAL 1,314 100 39.6


Verbs are also common: cause, bring about, contribute to, lead to, result in, derive, emerge, and stem. Patterns involving nouns (e.g. contrast, comparison, difference and distinction) and verbs (e.g. contrast, differ, distinguish and differentiate) are often used to compare and contrast but adjectives (e.g. different, distinct, differing and distinctive) play a more prominent role and account for 29.2 per cent of the lexical means used by expert writers (Table 4.12).

Table 4.11 Ways of expressing cause and effect in the BNC-AC-HUM


nouns

cause 755 2.8 22.7factor 550 2.1 16.6source 1,175 4.4 35.4origin 500 1.9 15.0root 183 0.7 5.5reason 1,802 6.8 54.2consequence 450 1.7 13.6effect 1,830 6.9 55.0result 813 3.1 24.5outcome 143 0.5 4.3implication 411 1.7 12.4

TOTAL NOUNS 8,612 32.52 259.25

Verbs

cause 570 2.2 17.2bring about 125 0.5 3.8contribute to 276 1.0 8.3generate 227 0.8 6.8give rise to 101 0.4 3.0induce 67 0.2 2.0lead to 671 2.5 20.2prompt 115 0.4 3.5provoke 161 0.6 4.9result in 327 1.2 9.8yield 129 0.5 3.9make sb/sth do sth 171 0.6 5.2arise from/out of 145 0.6 4.4derive 476 1.8 14.3emerge 466 1.8 14.0follow from 74 0.3 2.2trigger 56 0.2 1.7stem 95 0.4 2.9

TOTAL VERBS 4,252 16.06 128.0



Adjectives

consequent 53 0.2 1.6responsible (for) 344 1.3 10.4

TOTAL ADJ. 397 1.49 12

Prepositions

because of 599 2.3 18.0due to 195 0.7 5.9as a result of 196 0.7 5.9as a consequence of 22 0.1 0.7in consequence of 1 0.0 0.0in view of 66 0.3 2.0owing to 52 0.2 1.6in (the) light of 109 0.4 3.3thanks to 35 0.1 1.0on the grounds of 22 0.1 0.7on account of 24 0.1 0.7

TOTAL PREP. 1,321 4.99 39.8

Adverbs

therefore 1,412 5.3 42.5accordingly 130 0.5 3.9consequently 143 0.5 4.3thus 1,767 6.7 53.2hence 283 1.1 8.5so 1,894 7.2 57.0thereby 182 0.7 5.5as a result 101 0.4 3.0as a consequence 20 0.1 0.6in consequence 14 0.0 0.4by implication 35 0.1 1.1

TOTAL ADVERBS 5,981 22.59 180.04

Conjunctions

because 2,207 8.3 66.4since 955 3.6 28.7As 5 883 3.3 26.6for 1,036 3.9 31.2so that 696 2.6 21.0PRO is why

that is whythis is why

which is why

52221812

0.20.10.10.0

1.60.70.50.4

on the grounds that 83 0.312 2.5

TOTAL CONJ. 5,912 22.33 177.97

TOTAL 26,475 100 796.99


Table 4.12 Ways of comparing and contrasting found in the BNC-AC-HUM


Nouns

resemblance 116 0.4 3.5similarity 212 0.7 6.4parallel 147 0.5 4.4parallelism 19 0.1 0.6analogy 175 0.6 5.3contrast 522 1.8 15.7comparison 311 1.1 9.4difference 1,318 4.5 39.7differentiation 76 0.3 2.3distinction 595 2.0 17.9distinctiveness 10 0.0 0.3(the) same 559 1.9 16.8(the) contrary 28 0.1 0.8(the) opposite 85 0.3 2.6(the) reverse 56 0.2 1.7TOTAL NOUNS 4,229 14.46 127.3

Adjectives

same 2,580 0.9 77.7similar 1,027 3.5 30.9analogous 55 0.2 1.7common 1055 3.6 31.8comparable 223 0.8 6.7identical 137 0.5 4.1parallel 52 0.2 1.6alike 98 0.3 2.9contrasting 63 0.2 1.9different 2,496 8.5 75.1differing 72 0.3 2.2distinct 278 0.9 8.4distinctive 163 0.6 4.9distinguishable 33 0.1 1.0unlike 43 0.1 1.3contrary 27 0.1 0.8opposite 127 0.4 3.8reverse 23 0.1 0.7

TOTAL ADJECTIVES 8,552 29.24 257.44

Verbs

resemble 138 0.5 4.1correspond 137 0.5 4.1look like 102 0.3 3.1compare 278 0.9 8.4parallel 56 0.2 1.7contrast 137 0.5 4.1



differ 242 0.8 7.3distinguish 404 1.4 12.2differentiate 74 0.3 2.2

TOTAL VERBS 1,568 5.36 47.2

Adverbs

similarly 394 1.3 11.9analogously 2 0.0 0.1identically 2 0.0 0.1correspondingly 29 0.1 0.9parallely 0 0.0 0.0likewise 118 0.4 3.5in the same way 56 0.2 1.7contrastingly 3 0.0 0.1differently 97 0.3 2.9by/in contrast

by contrastin contrast

185116

69

0.6 5.6

by way of contrast 0 0.0 0.0by/in comparison

by comparisonin comparison

23149

0.1 0.70.40.3

comparatively 69 0.2 2.1contrariwise 4 0.0 0.1distinctively 25 0.1 0.7on the other hand 372 1.3 11.2(on the one hand) 136 0.5 4.1on the contrary 95 0.3 2.9quite the contrary 2 0.0 0.1conversely 62 0.2 1.9

TOTAL ADVERBS 1,674 5.72 50.39

Prepositions

like6 2,812 9.6 84.6unlike 244 0.8 7.3in parallel with 8 0.0 0.2as opposed to 121 0.4 3.6as against 46 0.2 1.4in contrast to/with

in contrast toin contrast with

8273

9

0.3 2.52.20.3

versus 53 0.2 1.6contrary to 66 0.2 2.00by/in comparison with

in comparison within comparison to

by comparison within comparison with

5214

42114

0.2 1.60.40.10.60.4

TOTAL PREP. 3,484 11.91 104.88

(Continued)


Table 4.13 shows a co-occurrence analysis of several nouns that are used to express cause or effect in academic prose: reason, implication, effect, outcome, result and consequence. Most of the co-occurrents listed form quite fl exible and compositional textual sentence stems with their nominal node, as illustrated in the following examples:

4.51. Another direct result of conquest by force of arms was the development of slavery, which was widespread up to the beginning of the nineteenth century.

4.52. This may be an effect of the uncertainty around television’s textuality; but it is now an extremely limiting effect for the development of theory.

4.53. Health for women was held to be synonymous with healthy motherhood. This had important implications for the debate over access to birth control informa-tion and abortion – rarely were demands for freer access to birth control information devoid of maternalist rhetoric.

4.54. The reason is that with Van Gogh art and life are not merely conditioned by each other to a greater degree than with any other artist, but actually merge with each other.

4.55. However it is fi rst necessary to consider another important consequence of the view of psychosis being presented here.


Conjunctions

as 5,045 17.2 151.9while 1264 4.3 38.0whereas 442 1.5 13.3

TOTAL CONJ. 6,751 23.08 203.23

Other expressions

as . . . as 2,766 9.5 83.3in the same way as/that 38 0.1 1.1compared with/to

compared withcompared to

155113

42

0.5 4.73.41.3

CONJ compared to/withas compared to/with

when compared to/withif compared to/with

321120

1

0.1 1.00.30.60.0

TOTAL 29,249 100 880.5

Table 4.12 Cont’d


Table 4.13 Co-occurrents of nouns expressing cause or effect in the BNC-AC-HUM

Table 4.13a: reason

Adjective + reason Verb + reason Determiner + reason

good have thismain give anothersuffi cient see (no) reason to + verb

obvious base on believe other provide supposedifferent fi nd doubtalleged examine prefersimple Auxiliary verb + reason thinktactical be fearpolitical seem acceptmajor reason + verb reason(s) for . . . .additional be supposingright justify believingvalid reason + preposition thinkingsimilar for acceptingfundamental against rejectingreal Preposition (2L) + reason adoptingindependent for There + verb + reasonspecial reason + conjunction There is (no) reason topossible why There seems no reasonhistorical which There are (DET/ADJ) reasonsparticular that

Table 4.13b: implication

Adjective + implication Auxiliary verb + implication

important bepractical implication + preposition

political ofserious forsocial Preposition + implicationVerb + implication withhave Determiner + implicationcarry thisimplication + verb implication + conjunction

be that


Table 4.13c: effect

Adjective + effect Verb + effect

adverse haveoverall producegood achieveprofound createknock-on causeindirect Auxiliary verb + effectfar-reaching bedamaging effect + verb

cumulative bedramatic depend onimmediate occurexcellent effect + preposition

long-term ofpractical onparticular uponpowerful Determiner + effectspecial thisfull effect + conjunction

general Thatimportant Noun and effectother cause

Table 4.13d: outcome

Adjective + outcome Verb + outcome

logical infl uenceeventual determinelikely representdifferent affectinevitable outcome + verb

fi nal beoutcome + preposition Auxiliary verb + outcomeof beDeterminer + outcomethis


Table 4.13e: result

Adjective + result Verb + result

inevitable producedirect achieveimmediate yieldbenefi cial giveeventual bringinteresting lead topractical showmain presentsimilar interpretresult + preposition obtainof havefrom result + verb

Preposition (3L) + result bewith Auxiliary verb + resultDeterminer + result bethis

Table 4.13f: consequence

Adjective + consequence Verb + consequence

inevitable haveunintended suffer (from)unfortunate avoiddirect considerimportant outweighnecessary discusspolitical consequence + verb

natural bebad followpractical ensuesocial Auxiliary verb +

consequencelikely bemajor consequence + preposition

possible ofDeterminer + consequence forthis Preposition (3L) +

consequenceanother withconsequence + conjunction ofthat -


The word combinations illustrated in Examples 4.51 to 4.55 are good illus-trations of what Sinclair and his followers have called ‘extended units of meaning’ where lexical and grammatical choices are ‘intertwined to build up a multi-word unit with a specifi c semantic preference, associating the formal patterning with a semantic fi eld, and an identifi able semantic pros-ody, performing an attitudinal and pragmatic function in the discourse’ (Tognini-Bonelli, 2002: 79). These extended units of meaning are catego-rized as textual phrasemes in Granger and Paquot’s (2008a) typology as they function as sentence stems to organize the propositional content at a metadiscoursal level.

A few co-occurrences are collocations as illustrated by Example 4.56. The verb carry is used in a delexical sense in the collocation carry implica-tions, which basically means have implications.

4.56. We may certainly talk of animals, in the absence of speech, “consciously intend-ing” or being compassionate, both of which carry implications of understand-ing to some degree.

The variety of adjectives used with the nouns reason, implication, effect, outcome, result and consequence is also worthy of note and bears testimony to their prom-inent role in argumentation (Soler, 2002; Tutin, forthcoming). A large pro-portion of those are evaluative adjectives (e.g. fundamental, good, important, inevitable, major, serious, suffi cient) and are used to express the ‘writer’s atti-tude or stance towards, viewpoint on, or feelings about the entities or propo-sitions that he or she is talking about’ (Hunston and Thompson, 2000: 5).

Like nouns, verbs that serve specifi c rhetorical or organizational func-tions in academic prose generally enter compositional and fl exible sequences. Table 4.14 gives the most frequent lexical bundles containing one of the four verbs suggest, appear, prove and tend typically used to express possibility or certainty. Most clusters are lexico-grammatical patterns which function as textual sentence stems (e.g. it has been suggested that, it appears that), sentence-initial adverbial clauses (e.g. as suggested above, . . . ) or rhemes (e.g. . . . proved a complete failure). It is worth noting that each verb form has its own ‘distinctive collocational relationship’ (Sinclair, 1999: 16), and that these constitute different form/meaning pairings, and thus differ-ent complete units of meaning. For example, the –ed form of suggest (unlike that of appear, tend or prove) is mainly used in passive constructions. It is often used to report suggestions made by other people in impersonal structures introduced by it (e.g. it has been suggested, it is sometimes suggested), and in phrases introduced by the conjunction as (e.g. as already suggested by).


As-phrases are also used with an endophoric marker (e.g. as suggested above) and/or the fi rst person pronoun I (e.g. as I have suggested) to refer to a sug-gestion previously made. Suggested is also used in impersonal structures introduced by it followed by a modal verb (e.g. it may/might be suggested that) to make a tentative suggestion. By contrast, the verb form suggests is typically used to make it clear that the suggestion offered is made on the basis of who/whatever is the subject of the sentence:

4.57. More recent evidence suggests, however, that while it lives in woodland it actually hunts over nearby open areas.

4.58. Sinclair Hood (1971) suggests that woollen cloth and timber were sent to Egypt in exchange for linen or papyrus.

In summary, results indicate that the phraseology of rhetorical or organi-zational functions in academic prose does not consist of idioms, similes, phrasal verbs, idiomatic sentences, proverb fragments and the like (see also Pecman, 2004 and Gledhill, 2000).7 Referential phrasemes that serve to organize scientifi c discourse mainly consist of lexical and grammatical collocations. Results also confi rm Howarth’s (1996; 1998) conclusion that a large proportion of the lexical collocations found in academic discourse consist of a verb in a fi gurative sense and an abstract noun denoting a recurrent concept in academic discussion (e.g. adopt an approach/a method;

Table 4.14 Co-occurrents of verbs expressing possibility and certainty in the BNC-AC-HUM

Table 4.14a: suggest

suggested suggest

– it has been suggested that– it is (sometimes, commonly) suggested

that– it was (fi rst, also, even) suggested that– it can / could / may be suggested that– this is suggested by – as (already) suggested by– as suggested above– (as) I (have) (already) suggested

– NP / it / this might / may / would suggest (that)

– NP does suggest (that)– there is evidence to suggest– I (would / want to) suggest– NP / it / this seems to suggest (that)

suggests suggesting

– NP / it / this (ADV: strongly, also) suggests (that)

– . . ., which suggests (that)– as NP suggests

– … , (ADV: strongly) suggesting (that)– I am (not) suggesting that


Table 4.14b: prove

proved prove

– NP / it / this proved to– NP / it / this proved (ADV) ADJ (to) with

ADJ: diffi cult, unable, abortive, impossible, inadequate, successful, possible

– NP / it / proved to be (ADV) ADJ– NP proved NP

– ADJ (likely, diffi cult, easy, possible) to prove

. . . may / might / would prove ADJ to– NP was to prove ADJ– attempt to prove– seek to prove

proves proving

– NP proves ADJ (impossible, necessary, inadequate, successful)

– NP proves that

– BE proving– . . ., proving that– . . . by proving– . . . of proving

Table 4.14c: appear

appeared appears

– it appeared (ADJ) that– there appeared to be – this appeared to V– . . . which appeared ADJ/ to V

– NP / it / this appears to V– which appears to V– what appears to V– there appears to V– it appears that – as appears from/in

appear appearing

NP would/might/may appear to be/V /

draw an analogy/a comparison/a distinction; reach a conclusion/a consensus/a point; develop an idea/a method/a model; carry out a task/a test/a study).

In academic prose, the category of textual phrasemes consists of three types of phraseme (cf. Figure 4.6). The fi rst is complex prepositions (e.g. with respect to, in addition to) and complex conjunctions (e.g. so that, as if, even

Table 4.14d: tend

tended tend

– NP tended to V (be, favour, take, see) – NP tend to V (be, see, look, regard)

tends tending

– NP tends to V– . . . which tends to V– it tends to VV: be, confi rm, ignore, obscure, become,

support, conclude

/


though) which are used to establish grammatical relations (cf. Burger’s (1998) category of structural phrasemes). The second is multiword linking adverbials, used to connect two stretches of discourse. Although the major-ity of linking adverbials are single adverbs, and are therefore not part of the phraseological spectrum, prepositional phrases functioning as adverbs (e.g. for example, in other words, in addition, in conclusion, as a result) and clausal linking adverbials (e.g. that is, that is to say, what is more, to conclude) are also common in academic prose (Conrad, 1999: 11–12). These fi rst two catego-ries of textual phraseme broadly correspond to Moon’s (1998) set of orga-nizational fi xed expressions and idioms. Textual sentence stems and rhemes constitute the third type of textual phrasemes, which I refer to as ‘textual formulae’. Textual sentence stems are multiple clause elements involving a subject and a verb, which ‘form the springboard of utterances leading up to the communicatively most important — and lexically most variable — ele-ment’ (Altenberg, 1998: 113). Examples include It has been suggested; Another reason is . . . ; and It is argued that. . . . Rhemes typically consist of a verb and its post-verbal elements (e.g. . . . is another issue). They also sometimes func-tion as textual phrasemes but are less frequent than sentence stems, possi-bly because rhemes are ‘usually tailored to expressing the particular new information the speakers want to convey to their listeners, and are there-fore, as Altenberg (1998: 111) points out, “composed of variable items drawn from an open set”’ (De Cock, 2003: 269). Textual formulae are par-ticularly prominent in academic writing and display different degrees of fl exibility, from fl exible fragments such as ‘DET (a, another) ADJ (typical, classic, prime, good, etc.) example of [NP] is . . .’ to more infl exible phrasemes such as ‘to be a case in point’.

Referential functionReferential phrasemes

(Lexical) collocationsGrammatical collocations

Attitudinal formulaeComplex prepositionscomplex conjuctionsLinking adverbials

Textual formulae (includingtextual sentence stems and

rhemes)

Textual phrasemes Communicative phrasemesTextual function

Phrasemes

Communicative function

Figure 4.6 The phraseology of rhetorical functions in academic prose


Attitudinal formulae make up a large proportion of communicative phrasemes in academic prose. They largely consist of sentence stems such as it is important/necessary that, it seems that or it is noteworthy that. This group is similar to Biber et al.’s (2004) category of stance bundles that ‘provide a frame for the interpretation of the following proposition, conveying two major kinds of meaning: epistemic and attitude/modality’ (Biber et al., 2004: 389).

The frequency-based approach adopted to study the phraseology of rhetorical functions has also helped uncover a whole range of word combi-nations that do not fi t traditional phraseological categories. Co-occurrences such as direct result, evidence suggests, fi nal outcome, and outstanding example have traditionally been considered as peripheral or falling outside the lim-its of phraseology (Granger and Paquot, 2008a: 29) but results suggest that they are essential for effective communication and are also part of the pre-ferred lexical devices used to organize scientifi c discourse.


In this chapter, I have shown that a high proportion of words in the Academic Keyword List (AKL) fi t my defi nition of academic vocabulary and serve rhetorical or organizational functions in academic prose. The analysis of exemplifi ers presented in Section 4.2 has also validated the method used to select AKL words: the lexical items which were automatically extracted included the most frequent exemplifi ers in academic writing (such as, example, for example and for instance) and lexical items which are not as frequent but which are more common in academic prose than in other genres (illustrate, exemplify, e.g., notably). The AKL could be very useful for curriculum and materials design as it includes a high number of words that serve rhetorical functions in academic prose.

The list, however, still needs to be refi ned in various ways. To be useful to apprentice writers, it should include the word combinations (frequent co-occurrences, collocations, textual phrasemes, etc.) in which each AKL word is commonly found in academic prose, together with information on the word’s frequency (see Coxhead et al. (forthcoming) for a similar project for Coxhead’s (2000) Academic Word List). This means that each AKL word has to be described in context, as was done above (Section 4.2) for the function of exemplifi cation. Such a contextual analysis will also make it possible to decide whether each word fi ts my defi nition of academic vocabulary and deserves to be retained in the Academic Keyword List.


The type of data analysis presented in this chapter has also offered valuable insights into the distinctive nature of the phraseology of rhetorical functions in scientifi c discourse. Most notably, results have shown that textual phrasemes make up the lion’s share of multiword units that ensure textual cohesion in academic prose. This type of phraseme, however, has often been neglected in theories of phraseology (cf. Granger and Paquot, 2008a: 34–5). Attitudinal formulae serve a major role in a restricted number of functions such as ‘expressing personal opinion’ and ‘expressing possibility and certainty’. Results have also pointed to the prominent role of free combinations to build the rhetoric of academic texts. My fi ndings thus support Gledhill’s call for a rhetorical or pragmatic defi nition of phraseology:

Phraseology is the ‘preferred way of saying things within a particular discourse’. The notion of phraseology implies much more than invento-ries of idioms and systems of lexical patterns. Phraseology is a dimension of language use in which patterns of wording (lexico-grammatical patterns) encode semantic views of the world, and at a higher level idi-oms and lexical phrases have rhetorical and textual roles within a specifi c discourse. Phraseology is at once a pragmatic dimension of linguistic analysis, and a system of organization which encompasses more local lexi-cal relationships, namely collocation and the lexico-grammar. I claim that the phraseological analysis of a text should not only involve the identifi ca-tion of specifi c collocations and idioms, but must also take account of the correspondence between the expression and the discourse within which it has been produced. (Gledhill, 2000: 202)

In line with this call, the functions of all AKL words and their preferred phraseological and lexico-grammatical patterns should be identifi ed by examining them in context.

Another objective of this chapter has also been to assess the adequacy of the treatment of rhetorical functions in EAP textbooks and investigate whether the AKL should be supplemented with additional academic words. To do so, I listed the words and phrases given in academic writing textbooks as typical lexical devices to perform the fi ve rhetorical functions analysed in detail in this book and compared them with the AKL. I identifi ed the words that were not part of the AKL and examined their use in the BNC-AC-HUM. Some of these lexical items turned out not to be typical of academic prose or to be extremely rare (e.g. to name but a few, by way of illustration) and should therefore not deserve the attention they have been given in


pedagogical materials. By contrast, a large proportion of AKL words were not found in textbooks in spite of their relatively high frequency and major discourse functions in academic prose. These fi ndings show the power of a data-driven approach to the selection of academic vocabulary and clearly call for a revision of the treatment of rhetorical functions in academic writing textbooks.

A pedagogically-oriented investigation of academic vocabulary cannot rest solely on native speaker data. It is essential to examine what learners actually do with lexical devices that serve rhetorical functions. For example, do they use exemplifi ers? Do they rely on words and phrasemes that are typical of academic prose? Do they use the expressions to name but a few and by way of illustration? If so, do they use them correctly? And do they use them sparingly or do they make heavy use of these infrequent exemplifi ers? These questions can only be answered by an analysis of learner corpus data. Such an analysis is presented in the next chapter.

Chapter 5

Academic vocabulary in the International Corpus of Learner English

This chapter is devoted to academic vocabulary in learner writing. Section 5.1 presents a detailed comparison of exemplifi catory devices in native and learner writing. This illustrates the type of results obtained when the range of lexical strategies available to EFL learners is compared to that of expert writers. Differences between learner and native writing are highlighted by means of log-likelihood tests. The UCREL log-likelihood calculator website (http://ucrel.lancs.ac.uk/llwizard.html) was used to compute log- likelihood values; 6.64 (p < 0.01) was taken as the threshold value. The whole learner corpus was compared to the BNC-AC-HUM but the results are only reported if they are common to learners from a majority of the mother tongue backgrounds considered. The same methodology was used to examine learners’ use of words that serve the rhetorical functions of ‘expressing cause and effect’, ‘comparing and contrasting’, ‘expressing a concession’ and ‘reformulating: paraphrasing and clarifying’. However these analyses are not presented in as much detail as for exemplifi cation, both for reasons of space and because the presentation would soon become cumbersome.

Instead, the focus of Section 5.2 is on the general interlanguage features that emerge from these analyses. These fall into six broad categories: limited lexical repertoire, lack of register awareness, learner-specifi c phraseological patterns, semantic misuse, clusters of connectives and unmarked position of connectors. However not all learner specifi c-features can be attributed to devel-opmental factors. The learner’s fi rst language also plays a considerable part in his or her use of academic vocabulary. In Section 5.3, I focus on transfer effects on French learners’ use of multiword sequences with rhetorical functions.

5.1. A bird’s-eye view of exemplifi cation in learner writing

A general fi nding of the comparison between the International Corpus of Learner English (ICLE) and the British National Corpus – Academic Humanities


(BNC-AC-HUM) subcorpus is that exemplifi catory lexical items are signifi -cantly more frequent in learner writing than in professional academic prose. This result highlights the importance of analysing several learner populations and comparing them so as to avoid faulty conclusions about EFL learner writing in general. Siepmann (2005) fi nds that the adverbials for example and for instance are less frequent in German learner writing than in native and non-native professional writing and argues that ‘under-use of exemplifi cation as a rhetorical strategy in student writing may (. . .) bespeak a general lack of concern for comprehensibility’ (Siepmann, 2005: 255). This explanation for German learners’ underuse of exemplifi ers is not entirely satisfactory, and does not apply to EFL learner writing in general: most L1 learner populations overuse exemplifi catory discourse markers.

The bar chart in Figure 5.1 shows the frequencies per 100,000 words of exemplifi ers in the ICLE and the BNC-AC-HUM. The lexical items are ordered by decreasing relative frequency in the ICLE. The bar chart shows that EFL learners’ use of exemplifi ers differs from that of professional writers in at least two ways. First, they do not choose the same exemplifi ers. Thus the most frequent exemplifi er in the ICLE is the adverbial for example, whereas the most frequent one in the BNC-AC-HUM is such as.

The frequencies of individual items also differ widely. Figures and log-likelihood values for each corpus comparison are given in Table 5.1. This shows that EFL learners’ overuse of the function of exemplifi cation is largely explained by their massive overuse of the adverbials for example and for instance, the noun example 2 and the preposition like. The overuse of for instance has already been reported by Granger and Tyson (1996) for French learners and Altenberg and Tapper (1998) for Swedish learners. Overuse of for example has also been found in other learner populations such as Japanese and Taiwanese learners (Narita and Sugiura, 2006; Chen, 2006). By contrast, learners tend to make little use of the verbs illustrate and exem-plify and the adverb notably, which are underused in the ICLE. There is no signifi cant difference in the use of the preposition such as, the abbreviation e.g., the nouns illustration and case in point and the expressions to name but a few and by way of illustration when comparisons are based on the total num-ber of running words in each corpus. Except for the preposition such as and the abbreviation e.g., these lexical items are quite infrequent in both native-speaker and learner writing.

As explained in Section 4.1, the frequency of each exemplifi catory lexical item can also be calculated as a proportion of the total number of exempli-fi ers. Corpus comparisons based on the total number of running words have shown that exemplifi cation is used signifi cantly more in the ICLE than

A

cademic vocabulary in the IC

LE

127

80

70

60

50

4030

20

10

ICLE BNC-AC-HUM

for exampleexample

such as like

for instance

illustrate

illustration

BE a case in pointnotably

exemplify

to name but a few

by way of illustratione.g.

0

Figure 5.1 Exemplifi ers in the ICLE and the BNC-AC-HUM


Table 5.1 A comparison of exemplifi ers based on the total number of running words

ICLE BNC-AC-HUM LogL

Abs. Rel. Abs. Rel.

Nouns

exampleexample

examples*exemple1

*exampl*examle

713477230

411

61.1740.919.7

0.30.10.1

1285665620

38.6820

18.7

91.6 (++)134 (++)

0.5

illustrationillustration

illustrations

1716

1

1.51.40.1

776314

2.32

0.4

3.3

(BE) a case in point 10 0.86 18 0.5 1.3

TOTAL NOUNS 740 63.5 1380 41.5 83.6 (++)

verbs

illustrateillustrate

illustratesillustrated

illustrating

512914

80

4.382.51.20.7

0

25997638415

7.82.91.92.50.5

16.1 (− −)0.62.6

17.7 (− −)9

exemplifyexemplify

exemplifi esexemplifi ed

exemplifi ed*examplifi edexemplifying

622

2110

0.430.20.2

0.180.10.1

0

799

1553

2

2.380.30.51.6

0

20.32 (− −)0.42.1

20.09 (− −)

1.2

TOTAL VERBS 57 4.8 338 10.2 32.1 (− −)

prepositions

such as 489 42 1494 45 1.8like 468 40.2 532 16 199.6 (++)TOTAL PREP. 957 82.1 2026 61 55.3 (++)

Adverbs

for examplefor example

*for exemple

857854

3

73.5 1263 38.00 209.9 (++)

for instance 344 29.5 609 18.3 47.3 (++)e.g. 94 8 259 7.8 0.1notably 5 0.4 77 2.3 22.1 (− −)to name but a few 3 0.3 4 0.1 0.9by way of illustration 1 0.1 3 0.1 0TOTAL ADVERBS 1304 111.9 2215 66.7 208.3 (++)

TOTAL 3058 262.4 5959 179.4 279.2 (++)

Legend: (++) signifi cantly more frequent (p < 0.01) in ICLE than in BNC-AC-HUM; (− −) signifi cantly less frequent (p < 0.01) in ICLE than in BNC-AC-HUM

Academic vocabulary in the ICLE 129

in the BNC-AC-HUM, and that the four lexical items discussed above are largely responsible for this overuse (Table 5.1). Comparisons based on the total number of exemplifi ers allow us to ask and answer different research questions. They give information about which lexical item(s) EFL learners prefer to use when they want to give an example, and in what proportions. Thus Table 5.2 shows that EFL learners select for example on 28 per cent of the occasions when they introduce an example, whereas native-speaker academics only use it to introduce 21 per cent of their examples.

Both methods indicate that EFL learners overuse the preposition like and the adverbial for example. As shown in Table 5.3, however, the two methods may also give different results. The noun example appears to be overused in the ICLE when comparisons are based on the total number of running

Table 5.2 A comparison of exemplifi ers based on the total number of exemplifi ers used

ICLE BNC-AC-HUM LogL

Abs. % Abs. %

Nouns

example 713 23.3 1285 21.6 2.8illustration 17 0.6 77 1.3 11.7 (− −)(BE) a case in point 10 0.3 18 0.3 0TOTAL NOUNS 740 24.2 1380 23.2 0.9

Verbs

illustrate 51 1.7 259 4.4 47.7 (− −)exemplify 6 0.2 79 1.3 35 (− −)TOTAL VERBS 56 1.8 338 5.7 77.3 (− −)

Prepositions

such as 489 16 1494 25 80 (− −)like 468 15.3 532 8.9 70.7 (++)TOTAL PREP. 957 31.3 2026 34 4.5

Adverbs

for example 854 28 1263 21.2 39 (++)for instance 344 11.3 609 10.2 2e.g. 94 3 259 4.3 8.1 (− −)notably 5 0.2 77 1.3 36.9 (− −)to name but a few 3 0.1 4 0.1 0.2by way of illustration 1 0 3 0 0.2TOTAL ADVERBS 1301 42.6 2215 37.2 15.3 (++)

TOTAL 3054 100 5959 100

Legend: (++) signifi cantly more frequent (p < 0.01) in ICLE than in BNC-AC-HUM; (− −) signifi cantly less frequent (p < 0.01) in ICLE than in BNC-AC-HUM


words in each corpus. However, a comparison based on the total number of exemplifi ers suggests that the learners choose the noun example about as often as professional academics when they want to introduce an example (23.3% vs. 21.6%). More lexical items are signifi cantly underused when fi gures are based on the total number of exemplifi ers. In addition to illustrate, exemplify and notably, the noun illustration and the preposition such as are selected proportionally less often by EFL learners than by professionals to introduce an example.

This fi rst broad picture of the use of exemplifi ers in the ICLE points to EFL learners’ limited repertoire of lexical items used to serve this specifi c EAP function. This characteristic of learner writing is discussed in more detail in Section 5.2.1.

By comparison with academics, EFL learners overuse the preposition like and underuse such as. Figure 5.2 shows the relative frequencies per 1,000,000 words of like and such as in four sub-corpora of the British National Corpus representing different ‘super genres’ (see Section 3.3): academic writing, fi ction, newspaper texts and speech (BNC-SP) as well as in the ICLE. The

Table 5.3 Two methods of comparing the use of exemplifi ers

Lexical item Comparison based on total

number of running words

Comparison based on total

number of exemplifi ers

example ++ //illustration // − −(be) a case in point // //

TOTAL NOUNS ++ //

illustrate − − − −exemplify − − − −

TOTAL VERBS − − − −

Such as // − −Like ++ ++

TOTAL PREPOSITIONS ++ //

for example ++ ++for instance ++ //e.g. // //notably − − − −to name but a few // //by way of illustration // //

TOTAL ADVERBS ++ ++

Legend: ++ signifi cantly more frequent (p < 0.01) in ICLE than in BNC-AC-HUM; − − signifi cantly less frequent (p < 0.01) in ICLE than in BNC-AC-HUM; // no signifi cant difference between the frequencies in the two corpora


preposition like is much more frequent than such as in speech3, fi ction, news and learner writing but is less frequent in academic prose. By contrast, such as is more frequently used in academic prose. Learners’ use of these exemplifi catory prepositions thus differs from academic expert writing, but resembles more informal genres such as speech. Learners’ underuse of the adverb notably in their academic writing is another illustration of the same point (Figure 5.3).

2500

2000

1500

1000

500

Academicwriting

Fiction News

like such as

Speech Learnerwriting

0

Figure 5.2 The use of the prepositions ‘like’ and ‘such as’ in different genres

50454035

freq

. per

mill

ion

wor

ds

30252015105

speech Fiction Learnerwriting

Academicwriting

News0

Figure 5.3 The use of the adverb ‘notably’ in different genres


A large proportion of EFL learner populations make repeated use of the word-like unit for instance. The use of this adverbial by native-speakers, however, differs signifi cantly from that of for example, both in terms of frequency and register. Figure 5.4 shows that 77 per cent of all instances of for example in the BNC are found in the academic sub-corpus. However only 59 per cent of the occurrences of for instance appear in academic prose while 30 per cent are found in more informal genres such as speech and fi ction. Lee and Swales (2006: 64) also showed that the use of these two adverbials differs across academic disciplines: for instance is more frequent in the social sciences and humanities while in natural sciences, technology and engineering, for example is strongly favoured to clarify a diffi cult or complex point through exemplifi cation.

Lack of register awareness manifests itself in a number of ways in learner academic writing. This will be the focus of Section 5.2.2.

The phraseology of academic words is also a major source of diffi culties to EFL learners. One of the main advantages of using a noun rather than the adverbials for example and for instance is that the use of a noun allows the writer to qualify the example with an adjective (see Section 4.2.2). However only 18 per cent of the adjective co-occurrents (types) of the noun example in the ICLE are signifi cant co-occurrents in the BNC-AC-HUM (Table 5.4). A quarter of the adjective co-occurrents of example in the ICLE do not appear at all in the 100-million word British National Corpus (Table 5.5). A large proportion of these adjectives have been described by our

for example

14%20%

59%

10%

11%77%

2%

7%

for instance

Academic writing News Fiction Speech

Figure 5.4 Distribution of the adverbials ‘for example’ and ‘for instance’ across genres in the BNC


native-speaker informant as forming awkward co-occurrences with example as illustrated in the following sentences:

5.1. The story of Cinderella is one more impermissible example. Cinderella is a neglected child, and once again the step-family is the guilty party. (ICLE-DU)

5.2. For example a disliked politician will be shot through such a zoom as to expose his ugly bits. Which may most probably infl uence our feeling towards him. We all know thousands of such manipulative examples. (ICLE-PO)

5.3. This mere example proves that the ideal union people dream of is not yet a total reality: national confl icts are still at work, every nation defends its own interests before fi ghting for those of “the group” they joined. (ICLE-FR)

5.4. The opposite example is (the former?) USSR, where the union was imposed by a central power without real approbation of the states and against people’s will. (ICLE-FR)

5.5. Of course, that was an overstated example, extreme, so to speak. (ICLE-RU)

Table 5.4 Signifi cant adjective co-occurrents of the noun ‘example’ in the ICLE


good 77 excellent 4extreme 12 typical 3above 8 classic 2clear 8 interesting 2striking 7 numerous 2simple 6 outstanding 1Well-known 5

Table 5.5 Adjectives co-occurrents of the noun ‘example’ in ICLE not found in the BNC


big 2 manipulative 1warning 2 mere 1absolute 1 model 1bright 1 opposite 1cruel 1 overstated 1present day 1 polemic 1evident 1 hair raising 1frightening 1 stirring 1impermissible 1 upsetting 1


Similarly, only 23 per cent of the verb types that are used with example in the ICLE are signifi cant co-occurrents of the noun in the BNC-AC-HUM (see Table 5.6). Some 27 per cent of the verb co-occurrents (types) of the noun example in the ICLE do not appear with example in the whole BNC. They are listed in Table 5.7. Like adjective co-occurrents, several of these verbs form awkward co-occurrences with the noun example:

5.6. In a new society made with less inequality, less poverty and more social justice we would not fi nd the same quantity of crime that we fi nd in our society. I can make the example of Naples: here there is everyday an incredible lot of crimes. (ICLE-IT)

Table 5.6 Signifi cant verb co- occurrents of the noun ‘example’ in the ICLE



be 162 be 119take 36 show 31give 28 illustrate 15fi nd 10 concern 2show 10 suggest 1serve 4 Suffi ce 1illustrate 3provide 2cite 2consider 1

TOTAL 258 TOTAL 169

Table 5.7 Verb co-occurrent types of the noun ‘example’ in ICLE not found in BNC



culminate into 1 say 1glide into 1 reinforce 1state 1 criticize 1plaster with 1 point out 1derive 1 express 1write 1help as 1appear 1

TOTAL 8 TOTAL 5


5.7. Their understanding of the outside world differs. It originates in dissimilar climate, life-style, social organization, political and economical stability of the country. To glide into an extreme example, unequality appears even between people living in towns and villages. (ICLE-CZ)

5.8. The rules of the road you have to learn to pass your driving license are plastered with examples of children who cross the road unexpectedly, running after a ball. (ICLE-GE)

The copular be is the most frequent left and right co-occurrent of the noun example in learner writing. Textual sentence stems and rhemes with the verb be are signifi cantly more frequent in learner writing than in profes-sional academic writing (Table 5.8). These results differ markedly from those reported in Paquot (2008a) in which French, Spanish, Italian and German learners were shown to underuse stems and rhemes with the verb be. This difference may be explained by the fact that the reference corpus used for comparison in Paquot (2008a) was a collection of native-speaker student essays.

Table 5.9 shows that the structure there + be + example is more frequently used in learner writing than in professional academic writing. It appears in all 10 learner corpora (i.e. irrespective of the learner’s mother tongue) as illustrated by the following sentence:

5.9. There is the example of Great Britain where a professional army costs less than, for example, the French army based on conscription. (ICLE-RU)

Table 5.8 The distribution of ‘example’ and ‘be’ in the ICLE and the BNC-AC-HUM

be + example example + be TOTAL Rel. freq. LogL

ICLE 162 (57.7%) 119 (42.3%) 281 24.1 199.76 (++)BNC-AC-HUM 139 (62.3%) 84 (37.7%) 223 6.71

Table 5.9 The distribution of ‘there + BE + example’ in ICLE and the BNC-AC-HUM

there + BE + example LogL

Abs. freq. Rel. freq.34.52 (++)ICLE 31 2.66

BNC-AC-HUM 15 0.45


In professional academic writing, the verb take is mainly used in sentence-initial exemplifi catory infi nitive clauses with the noun example (Example 5.10). This pattern is very infrequent in ICLE. EFL learners prefer to use the verb take in active structures introduced by the personal pronoun I (Example 5.11) or in fi rst person plural imperative sentences (Example 5.12).

5.10. To take one example, at the beginning of the project seven committees were established, each consisting of about six people, to investigate one of a range of competing architectural possibilities. (BNC-AC-HUM)

5.11. I can take the example of the ‘Société Générale de Belgique’ which is directed by ‘Suez’. (ICLE-FR)

5.12. Let’s take the example of painting. (ICLE-FR)

As illustrated by Examples 5.13 and 5.14, learners often use the verb have in the same structures as take to introduce an example. The imperative sentence, however, was judged to be awkward by our native-speaker informant.

5.13. Let us have an example — an extract out of the famous Figaro’s soliloquy: There is a liberty of the press in Madrid now, so that I can write about anything I like, providing I will have it checked by two or three censors and an condition that I will not write against the government and religion. (ICLE-CZ)

5.14. I have a good example in my family. (ICLE-PO)

Interestingly, the verb have and the fi rst person plural imperative let’s are not signifi cant left co-occurrents of example in the BNC-AC but they are in the BNC-SP corpus of spoken language. The verb have is often used in speech with an inclusive we as subject (Example 5.15); let’s is typically used with the verb take + example (Example 5.16).

5.15. Er in relation to existing employment sites er and Mr Laycock referred to National Power, erm there we have an example of the attitude that the the council is taking towards the the re-use of employment sites. (BNC-SP)

5.16. Let’s take the example of a cooker. (BNC-SP)

The verb give is the most signifi cant co-occurrent of the noun example in the BNC-SP. It is used in questions and fi rst person plural imperative sentences (Examples 5.17 and 5.18), two patterns that are not found in the BNC-AC-HUM despite the fact that the verb is also a signifi cant co-occurrent of


example in academic prose. By contrast, fi rst person plural imperative sentences with the verb give do appear in the ICLE (Example 5.19).4

5.17. Can you give an example when you say that the law is designed? (BNC-SP)

5.18. Let me give you some examples. (BNC-SP)5.19. Let me give you one example – appaling shots from the war in ex-Yugoslavia

that we can see nearly every day. (ICLE-CZ)

In summary, verb co-occurrents of the noun example provide further evidence for the genre-bound nature of phrasemes: the preferred phraseo-logical environment of the noun differs in academic writing and speech (see Biber et al. 1999; 2004; Luzón Marco, 2000). Results suggest that EFL learners sometimes select co-occurrences that are more typical of speech, which can be interpreted as further indication of their lack of register awareness.

Differences in phraseological or lexico-grammatical preferences are often revealed by patterns of overuse and underuse of word forms. Thus, the different forms of the verbs illustrate and exemplify are not all under-used in learner writing. Table 5.1 above shows that the two verbs are under-used in their –ed form only. This underuse corresponds to an underuse of the passive constructions BE illustrated by/in (Example 5.20) and BE exem-plifi ed by/in (Example 5.21), the past participle exemplifi ed following a noun phrase (Example 5.22) and the patterns as illustrated/exemplifi ed by/in (Example 5.23):

5.20. The contrast between the conditions on the coast and in the interior is illus-trated by the climatic statistics for two stations less than 30 km (18.5 miles) apart. (BNC-AC-HUM)

5.21. The association of this material with the clerk is clearly exemplifi ed by Chau-cer’s Wife of Bath’s fi fth husband, the clerk Jankyn, who, in the Wife of Bath’s Prologue, reads antifeminist material to her from his book Valerie and Theofraste. (BNC-AC-HUM)

5.22. Piaget’s claim that thinking is a kind of internalized action, exemplifi ed in the assimilation-accommodation theory of infant learning mentioned above, is really a global assumption in search of some refi ned, detailed and testable expression. (BNC-AC-HUM)

5.23. He assumed, without argument, that science, as exemplifi ed by physics, is superior to forms of knowledge that do not share its methodological character-istics. (BNC-AC-HUM)


The verb illustrate is more often used with human subjects (11.76%) in learner writing, and more specifi cally with the personal pronoun I:

5.24. I would like to illustrate that by means of some examples which, as you will see, are very diverse; . . . (ICLE-DU)

5.25. In the worst cases people decide to suicide. I can illustrate that by a real example. (ICLE-CZ)

It is also frequently used in sentence-initial infi nitive clauses (13.72%):

5.26. To illustrate the truth of this, one has only to mention people’s disappointment when realizing how little value has the time spent at university. (ICLE-SP)

5.27. To illustrate this point, it would be interesting to compare our situation with the U.S.A.’s. (ICLE-FR)

As in professional academic writing, the noun case in point is very rarely used in learner writing. When used, however, it sometimes appears in lexico-grammatical patterns that are not found in expert academic writing, e.g. in an infi nitive clause with the verb take (Example 5.28) or determined by a defi nite article and followed by the verb be and a that-clause (Example 5.29).

5.28. However, wars always break out for economical reasons; For example, the fi rst world war, to take a case in point, did not start because the murder of archduke Frank Ferdinand, heir of Autro-Hungary; that was only the straw that broke the camel’s back. (ICLE-SP)

5.29. Professional observers see some even deeper danger in the emerging situation. A great number of children spend more and more time watching television. They take into consideration the behaviour patterns of fi lm stars, they want to be like them. The case in point is that little children learn how to smoke how to drink how to be cunning and clever and get round the adults. Film stars are usually very attractive and it’s not a surprise that children want to follow them. (ICLE-RU)

EFL learners’ phraseological and lexico-grammatical specifi cities will be discussed in detail in Section 5.2.3 below.

EFL learners may also experience diffi culty with the meaning of single words and phrasemes. For example, they sometimes use the abbreviation i.e. instead of e.g. as an exemplifi catory discourse marker (Examples 5.30


to 5.32). The abbreviation i.e., however, is a synonym of ‘that is’ used to reformulate by paraphrasing or clarifying, and not an exemplifi er at all.

5.30. The states mostly tend to solve their politic problems in a peaceful way (*i.e. [e.g.] the split of Czech federation or the unifi cation of Germany). (ICLE-CZ)

5.31. One of the examples that makes this point is related to children’s toys, because nowadays children play with technological toys (*i.e.: [e.g.] video games), and these toys do not let the children develop their imagination and, in many cases, they are so inactive that playing with these toys does not permit physical exercise. (ICLE-SP)

5.32. It might seem absurd, but many progressive social changes (*i.e. [e.g.] an increase of individual liberty) may lead to further increase of crime. (ICLE-RU)

Learners also sometimes use as in lieu of the complex preposition such as (Examples 5.33 to 5.37). It should be noted, however, that this erroneous use is more frequently found in learner populations with Romance mother tongue backgrounds.

5.33. Thus soldiers learned mostly bad habits *as [such as] smoking, drinking (if possible) and being lazy in their leisure time. (ICLE-CZ)

5.34. In addition to the familiar subjects *as [such as] reading, writing and math-ematics, time should be reserved for making children conscious of the fact that there is more to life than the things we see. (ICLE-DU)

5.35. There should be particular institutions for those who are mentally alienated *as [such as] the rapists, others for the young people, etc. (ICLE-FR)

5.36. In this essay I would like to show how, in my opinion, crime is caused by a pre-disposition of the individuals and how, of course, other factors *as [such as] society, culture and politics can infl uence this natural inclination. (ICLE-IT)

5.37. Another proof will be the role that imagination plays in all the Arts *as [such as] Literature, Music and Painting. (ICLE-SP)

As illustrated in Example 5.38, the adverb namely is also sometimes misused by EFL learners who use it instead of notably or another exemplifi er.

5.38. This new wave of revolting trivial events is all the more worrying since it is linked to a rise of the small delinquance, implying a generalized climate of ter-ror and a total mistrust of the citizens towards the police forces and the law, both accused of all vices and *namely [(most) notably] of being too lax with those evils. (ICLE-FR)


This confusion is relatively common, which is not surprising as it is even found on websites supposed to help learners master English connectors (Figure 5.5).

More generally, namely is very often misused in learner writing and it is not always clear what learners mean when they use this adverb:

5.39. Because the campus consists of modern buildings, built closely together, it is no more than a ten minute’s walk to get where you need to be for lectures and semi-nars. All the academic facilities are ?namely located on the main campus. (ICLE-DU)

5.40. Why, then, so many people object to gay marriages and, at the same time, yearn for equality? It is ?namely just equality what gay marriages are about, isn’t it? (ICLE-FI)

5.41. The efforts made by the fi rms are obvious. They ?namely create replacement products: they replace the gas in the aerosols and so we have ozone-friendly aerosols, . . . (ICLE-FR)

5.42. Reluctance to eventually join The Common Market is ?namely caused by fear, disbelieves, inferiority complex, short-sightedness or even nationalistic and xenophobic tendencies. (ICLE-PO)

More examples of semantic misuse are illustrated and discussed in Section 5.2.4.

Another explanation for the general overuse of the function of exempli-fi cation in learner writing may be that exemplifi ers are repeatedly used when they are superfl uous, redundant or even when other rhetorical func-tions should be made explicit. In Example 5.43, the logical relation between the two sentences is a causal link that is left implicit while an unnecessary exemplifi er is used:

5.43. I described there only some examples from the great number of criminal offences. After some years many of those criminals will be set free because of their

Figure 5.5 The treatment of ‘namely’ on websites devoted to English connectors

Pour donner des exemples

Example: for example, for instance, just as, in particular, namely, one example, such as, toillustrate

for instance, for example, such as, likenamely (c' est-à-dire)above all (surtout)

http://page sperso-orange fr/frat. st.paul/BACK itde Survie.pdf

http://fr.wikibooks.org/wiki/utilisateur:Jean-Francois_Gagnon/Anglais:Connective_words


relatively mild punishment. They had for example youthful age. (Youthful age – by the way in contrast to the punishment of 16 years old boys in our country, who got off with the light punishment, in England were recently sentenced two 10 years old boys for murder of a 3 years old boy to the lifelong punishment!) (ICLE-CZ)

Section 5.2.5 will focus on the unnecessary use of lexical items that serve rhetorical or organizational functions as well as on learners’ tendency to clutter up their texts with too many logical devices.

EFL learners’ use of exemplifi ers also differs from that of expert writers with respect to positioning. A sentence-initial position for the adverbials for example and for instance is clearly favoured in the ICLE, compared to the BNC-AC-HUM:

5.44. But there are actually a number of things we all can do that make a difference. For example, there ought to be information about different ways to save elec-tricity. (ICLE-SW)

5.45. There were a lot of wars due to the religion. For instance, England has always been divided according to the kind of religion in which a person believed. (ICLE-SP)

The two adverbials are also repeatedly found at the end of a sentence in the learner subcorpora (7.14% of the occurrences of for example and 8.4% of the occurrences of for instance), although this position is rare in academic professional writing (1.6% for for example; 1.3% for for instance):

5.46. Let us have a good look at television for example. (ICLE-PO)5.47. They only want an easy to operate camera, a Single Use Camera for instance.

(ICLE-DU)

Aspects of sentence position are dealt with in Section 5.2.5. In Section 4.4, I argued that Academic Keyword List (AKL) lexical items

and their phraseological patterns should be taught to EFL learners. Learner corpus data support this claim as all the AKL words that are used to give examples in academic prose present one or more learner-specifi c diffi culties. The adverb notably and the abbreviation e.g. are semantically misused; the adverbials for example and for instance are predominantly used in sentence-initial position; and the noun example and the verbs illustrate and exemplify are used in learner-specifi c phraseological patterns. It was also argued that the pedagogical relevance of non-AKL items – the preposition


like, the nouns illustration and case in point and the expressions to name but a few and by way of illustration – depended on whether learners already used these exemplifi ers and how they used them. The analysis of the ICLE corpus suggests that:

– A word of caution is needed against excessive reliance on the preposition like;

– The noun illustration should be specifi cally taught to upper-intermediate and advanced learners as it is underused in the ICLE;

– The specifi c lexico-grammatical patterns of case in point should also be taught as this phraseme is repeatedly used in ‘unidiomatic’ patterns.

The pedagogical implications of learner corpus-based fi ndings will be further considered in Chapter 6.

5.2. Academic vocabulary and general interlanguage features

A comparison of words that serve the rhetorical functions of ‘giving examples’, ‘expressing cause and effect’, ‘comparing and contrasting’, ‘expressing a concession’ and ‘reformulating: paraphrasing and clarifying’ in learner and expert academic writing has made it possible to identify six specifi c areas of where learner English varies from native-speaker academic English. Section 5.2.1 focuses on learners’ limited lexical reper-toire by examining aspects of over- and underuse. In Section 5.2.2, the char-acteristics of learner’s lack of register awareness are presented. Section 5.2.3 explores the type of phraseological and lexico-grammatical patterns that are found in most learner sub-corpora. Section 5.2.4 discusses patterns of semantic misuse of connectors and abstract nouns. Learners’ tendency to clutter their texts with unnecessary connectives is the focus of Section 5.2.5 and Section 5.2.6 illustrates their preference for placing connectors at the beginning of sentences.

5.2.1. Limited lexical repertoire

Several studies based on one or more ICLE subcorpora have argued that ‘these EFL writers are not equipped with the type of lexical knowledge necessary for the type of writing task they are undertaking’ (Petch-Tyson, 1999: 60). An analysis of learners’ use of potential academic words from the


Academic Keyword List (AKL) supports this view. Table 5.10 shows that almost 50 per cent of the words in the AKL are underused in the ICLE, a percent-age that rises to 52.1 per cent for nouns and 56.3 per cent for adverbs. By contrast, the proportion of words in the AKL that are overused in learner academic writing is only 21.4 per cent . The largest percentages of overused items are found in nouns and in the ‘other’ category which includes prepo-sitions, conjunctions, determiners, etc.

Table 5.11 gives examples of overused and underused AKL words in the ICLE. It could be argued that ‘learner usage tends to amplify the high frequencies and diminish the low ones’ (Lorenz 1999b: 59). For example, overused items such as the nouns idea and problem, the verbs be and become and the adjectives diffi cult and important are very frequent words in general English (relative frequencies of more than 200 occurrences per million words in the whole BNC). Conversely, underused items such as the nouns hypothesis and validity, the verbs exemplify and advocate, the adverbs conversely and ultimately and the prepositions as opposed to and in the light of are much less frequent in English (relative frequencies of less than 30 occurrences per million words in the whole BNC).

The picture, however, appears to be more complex than Lorenz’s quote suggests. Not all high frequencies are amplifi ed in EFL learner writing. Many AKL words that appear with a relative frequency of more than 100 occurrences per million words in the whole BNC are underused in the ICLE, e.g. the nouns argument, difference and effect, the verbs argue and explain, the adjectives likely and signifi cant and the adverbs generally and par-ticularly (in bold in Table 5.11). Key function words such as between, in, by, and of are quite representative of the nominal style of academic texts, where 60 per cent of all noun phrases have a modifi er (Biber, 2006). However, these highly frequent prepositions are underused in the ICLE, a fact that can be related to EFL learners’ tendency to avoid prepositional noun phrase postmodifi cation (Aarts and Granger, 1998; Meunier, 2000: 279).

Table 5.10 The distribution of AKL words in the ICLE

overused no statistical

difference

underused

nouns 86 [24.2%] 84 [23.7%] 185 [52.1%]verbs 40 [17.2%] 93 [39.9%] 100 [42.9%]adjectives 34 [18.9%] 59 [32.8%] 87 [48.3%]adverbs 16 [18.4%] 22 [25.3%] 49 [56.3%]other 21 [28.0%] 21 [28.0%] 33 [44.0%]

TOTAL 199 [21.4%] 277 [29.8%] 454 [48.8%]


The preposition despite is underused, while its much less frequent synonym, the complex preposition in spite of, is overused in learner writing (Figure 5.6), irrespective of genre. In addition, words such as the noun disadvantage, the verbs participate and solve, and the adverbs consequently and moreover (underlined in Table 5.11) are overused although they appear with frequencies of less than 50 per million words in the BNC.

The amplifi cation of a restricted set of low frequency words in learner writing may be partly explained by teaching-induced factors. Words such as consequently, moreover and secondly usually appear in the long and

Table 5.11 Examples of AKL words which are overused and underused in the ICLE

overused underused

nouns advantage, aim, benefi t, change, choice, conclusion, consequence, degree, disadvantage, example, fact, idea, infl uence, possibility, problem, reality, reason, risk, solution, stress

addition, argument, assumption, basis, bias, comparison, concept, contrast, criterion, difference, effect, emphasis, evidence, extent, form, hypothesis, issue, outcome, perspective, position, scope, sense, summary, theme, theory, validity

verbs aim, allow, avoid, be, become, cause, choose, concern, consider, consist, contribute, create, deal, depend, develop, exist, improve, increase, infl uence, participate, prove, solve, study, treat, use

adopt, advocate, argue, assert, assess, assume, cite, comprise, conduct, contrast, defi ne, derive, describe, emphasise, enhance, ensure, examine, exemplify, explain, highlight, indicate,

note, propose, refl ect, reveal, specify, suggest, view, yield

adjectives common, different, diffi cult, important, interesting, main, necessary, obvious, possible, practical, real, special, true, useful

adequate, appropriate, comprehensive, critical, detailed, explicit, extensive, inherent, likely, major, misleading, parallel, particular, prime, relative, representative, signifi cant, similar, subsequent, substantial, unlikely

adverbs also, consequently, especially, extremely, however, mainly, more, moreover, often, only, secondly, successfully, therefore

adequately, conversely, effectively, essentially, generally, hence, increas-ingly, largely, notably, originally, particularly, potentially, previously, primarily, readily, relatively, similarly, specifi cally, subsequently, ultimately

other according to, because, due to, during, each, for, less, many, or, same, several, some, than, this

although, an, as opposed to, between, by,

despite, from, given that, in, in relation to, in response to, in terms of, in the light of, including, its, latter, of, prior to, provided, rather than, subject to, the, to, unlike, upon, which


undifferentiated lists of connectors provided in EFL/EAP teaching materi-als (see Section 6.1). This situation may be compounded by problems of semantic misuse as will be discussed in Section 5.2.4. The underuse of some frequent, but semantically specialized, words probably stems from learners’ tendency to rely on all-purpose, general, and vague words where more pre-cise vocabulary should be used (Granger and Rayson, 1998; Petch-Tyson, 1999). Another tentative explanation may be that EFL learners do not amplify any high frequencies words except those that are common in speech. As argued by Baayen et al. (2006), ‘the complexity of the frequency variable has been underestimated’ and it may be that more emphasis should be placed on the explanatory potential of spoken frequency counts. Underused words such as argument, issue, assume, indicate, appropriate, and particularly are quite frequent in general English (as represented by the whole BNC), but their frequencies are signifi cantly less when the conversa-tion component is analysed separately.

In Section 5.1, it was shown that, although they generally overuse exem-plifi ers, EFL learners make little use of a number of EAP-specifi c lexical devices such as the verbs illustrate and exemplify or the adverb notably. They rely instead on a restricted lexical repertoire mainly composed of the adver-bials for example and for instance, the noun example and the prepositions like and such as. The same conclusion holds for learners’ use of cause and effect lexical items, which is compared with that of expert writers in Appendix 1. Broadly speaking, learners overuse logical links signifying cause and effect in their argumentative essays. This overuse does not, however, affect all grammatical categories. When corpus comparisons are based on the total

Figure 5.6 The use of ‘despite’ and ‘in spite of’ in different genres

250

200

150

100

50

0Academic

writingNews Fiction

despite in spite of

Speech Learnerwriting


number of running words in each corpus, the overuse seems to be generally attributable to adverbs, prepositions and conjunctions. The categories of nouns, verbs and adjectives do not display signifi cant patterns of over- or underuse. By contrast, when frequencies are compared to the total number of cause and effect lexical items, only prepositions and conjunctions are signifi cantly overused, while nouns and verbs are underused (Table 5.12). This means that, compared to expert writers, EFL learners prefer to use prepositions, conjunctions and, to a lesser extent, adverbs to express a cause or an effect, and tend to avoid nouns and verbs.

Table 5.13 shows that, even though EFL learners prefer to use preposi-tions, conjunctions and adverbs to express cause and effect, not all indi-vidual connectors are overused in learner writing. The overuse of conjunctions largely stems from learners’ marked preference for because, which represents 19.9 per cent of all cause and effect markers in the ICLE. Lorenz (1999b) examined the use of causal links in essays written by 16-to-18-year-old German learners and described the marked overuse of the conjunction because as ‘wild-card use’. He argued that ‘if a linguistic element is used as an all-purpose wild card, that usage is bound to include a number of instances of over-extension. In other words, it can be expected that learners may disregard target-language restrictions which are not that obvious, or even accounted for in the standard grammars, but which are nev-ertheless observed by the native speakers. Such “simplifi cation” is one of the most frequently cited features of learner language’ (Lorenz, 1999b: 60–1).

Several of the overused lexical items are massively overused in learner writing. The adverb so represents 11.5 per cent of the ‘cause and effect’

Table 5.12 Two ways of comparing the use of cause and effect markers in the ICLE and the BNC

Absolute frequency / total

number of words

Absolute frequency / total

number of ‘cause and

effect’ markers

nouns // − −verbs // − −adjectives // //adverbs ++ //prepositions ++ ++conjunctions ++ ++

Legend: ++ signifi cantly more frequent (p < 0.01) in ICLE than in BNC-AC-HUM;− − signifi cantly less frequent (p < 0.01) in ICLE than in BNC-AC-HUM;// no signifi cant difference between the frequencies in the two corpora


lexical items used by learners while it only accounts for 7.2 per cent of those in expert writing. Other examples of ‘lexical teddy bears’ (Hasselgren, 1994) or ‘pet’ discourse markers (Tankó, 2004) are the prepositions because of and due to. In their study of expressions of doubt and certainty, Hyland and Milton (1997) reported similar fi ndings: Cantonese learners used a more limited range of epistemic modifi ers, with the ten most frequently used items (will, may, think, would, always, usually, know, in fact, actually, and probably) accounting for 75 per cent of the total.5

Table 5.13 The over- and underuse by EFL learners of specifi c devices to express cause and effect (based on Appendix 1)

overuse no statistical

difference

underuse TOTAL

nouns

2 [19%] 4 [37%] 5 [45%]11

[100%]root, consequence cause, factor, reason, result

source, origin, effect, outcome, implication

verbs

1 [6%] 3 [18%] 13 [76%]

17[100%]

cause bring about, contribute to, lead to

generate, give rise to, induce, prompt, stem, provoke, result in, yield, arise, derive, emerge, follow, trigger

adjectives0 1 [50%] 1 [50%] 2

[100%]responsible (for) consequent

adverbs

4 [40%] 2 [20%] 4 [40%]

10[100%]

consequently, as a result, as a consequence, so

therefore, in consequence

accordingly, thus, hence, thereby

prepositions

3 [27%] 6 [54%] 2 [18%]

11[100%]

because of, due to, thanks to

as a result of, owing to, as a consequence of, on the grounds of, in consequence of, on account of

in view of, in (the) light of

conjunctions

2 [40%] 0 3 [60%]5

[100%]because, this/that is why

for, so that, on the grounds that

TOTAL 12 [21%] 16 [29%] 28 [50%] 56

[100%]


On the other hand, 50 per cent of the lexical devices which serve to express cause or effect in expert writing are underused by learner writers. While underuse was found in all grammatical categories, the proportions varied signifi cantly. Nouns and verbs constitute a large proportion of the possible ways of expressing a cause or an effect in academic prose, but 64.3 per cent of them are underused in the ICLE (e.g. the nouns source, effect and implication; the verbs induce, result in, yield, arise, emerge and stem from). As will be discussed in Section 6.1, this may be explained by teaching-induced factors, as lexical cohesion has been largely neglected in teaching materials (textbooks and especially grammars), where the focus has gener-ally been on adverbial connectors.

An analysis of the lexical items which serve to express a comparison or a contrast in academic prose shows that the rate of underuse is also quite high in this function. Table 5.14 shows that almost half of all comparison and contrast markers are underused. As with cause and effect lexical items, the degree of underuse varies signifi cantly. Nouns and adjectives (e.g. resem-blance, similarity, contrast, similar, distinct, and unlike) account for 59 per cent of all underused lexical items in the comparison and contrast category. The rate of overuse is relatively low, but once again overused items include words and phrasemes that are more frequent in speech (e.g. look like, in the same way) (see Section 5.2.2) as well as commonly misused expressions such as on the contrary (see Section 5.2.4). Unlike the cause and effect lexical items, overused comparison and contrast word do not compensate for the underused ones. Comparisons and contrasts are generally underused in learner writing.

In summary, EFL learners tend to rely heavily on a restricted set of greatly overused adverbs, prepositions or conjunctions to establish textual cohe-sion. Logical links can also be provided by nouns (cf. the concept of ‘label-ling’ explained in Section 1.3), verbs and adjectives, which often account for a large proportion of the lexical strategies used to serve a specifi c rhe-torical or organizational function in expert academic prose. These cohesive devices, however, do not seem to be readily accessible to upper-intermedi-ate/advanced EFL learners. This is not particularly surprising, as lexical cohesion has generally been neglected in teaching materials. These fi ndings are not restricted to EFL learners: although they may become fl uent in English conversational discourse, English as a Second Language (ESL) speakers have also been reported to ‘continue to have a restricted repertoire of syntactic and lexical features common in the written academic genre’ (Hinkel, 2003: 1066). Tables 5.13 and 5.14 provide useful

Table 5.14 The over- and underuse by EFL learners of specifi c devices to express comparison and contrast (based on Appendix 2)

overuse no statistical difference underuse TOTAL

nouns

0 5 [33%] 10 [67%]

15 [100%]parallelism, difference, distinctiveness, the contrary, the opposite

resemblance, similarity, parallel, analogy, contrast, comparison, differentiation, distinction, the same, the reverse

verbs

2 [22%] 5 [56%] 2 [22%]9 [100%]look like, compare resemble, correspond, differ,

distinguish, differentiate parallel, contrast

adjectives

2 [11%] 4 [22%] 12 [67%]

18 [100%]same, different alike, contrary, opposite, reverse similar, analogous, common, comparable, identical, parallel, contrasting, differing, distinct, distinctive, distinguishable, unlike

adverbs

4 [19%] 10 [48%] 7 [33%]

21 [100%]in the same way, on the other hand, on the one hand, on the contrary,

+ erroneous expressions

analogously, differently, identically, parallely, reversely, contrariwise, by way of contrast, contrastingly,quite the contrary, comparatively

similarly, likewise, correspondingly, by/in comparison, conversely, by/in contrast, distinctively

prepositions

2 [22%] 3 [33%] 4 [44%]9 [100%]like, by/in comparison with

+ erroneous expressionsin parallel with, in contrast to/with, contrary to

unlike, as opposed to, as against, versus

conjunctions0 1 [33.33%] 2 [66.67%]

3 [100%]whereas as, while6

other expressions

1 [25%] 3 [75%] 0

4 [100%]as … as, in the same way as/ that, compared with/to, CONJ compared with/to

TOTAL 11 [13.9%] 31 [39.2%] 37 [46.8%] 79


information about learners’ particular needs. Section 6.3 will discuss how they can be used to inform pedagogical material.

In this section, the breadth of EFL learners’ lexical repertoire has been examined in terms of the proportion of over- and underused AKL single words and mono-lexemic units used to perform specifi c rhetorical func-tions. In Section 5.2.3, it will be shown that the limited nature of EFL learn-ers’ lexical repertoire also stems from a restricted use of the phrasemes and lexico-grammatical patterns typically found in expert academic prose.

5.2.2. Lack of register awareness

Many learner corpus-based studies have reported on EFL learners’ lack of register awareness (e.g. Granger and Rayson, 1998; Lorenz, 1999b; Altenberg and Tapper, 1998; Meunier, 2000; Ädel, 2006). These studies, however, have often focused on learners with the same mother tongue background. The large-scale study undertaken here allows for a more sys-tematic description of register awareness, by exploring the way EFL learn-ers with different mother tongue backgrounds use academic vocabulary.

In the ICLE, most rhetorical functions are characterized by the overuse of at least one lexical item that is more typical of speech than of expert writ-ing (Table 5.15). Examples 5.48 to 5.52 illustrate overused lexical items that are more frequent in the BNC spoken component than in the BNC-AC-HUM: the adverb so to express an effect, the adverb though to introduce a concession, the adverbial of course to express certainty, the stem I am going to talk about to introduce a new topic, and the adverbial all in all which is used to ‘show that you are considering every part of a situation’ (Longman Dictionary of Contemporary English (LDOCE4)).

5.48. Many people who are in this situation think that this is a waste of time: you lose an entire year. So they want to get rid of the military service. (ICLE-DU)

5.49. Spanish holds an important position in South America and increasingly so in the United States, too. According to Crystal it has little further potential ouside Spain, though. (ICLE-FI)

5.50. But practically everybody is able to dream. Of course, there are different people with different concepts of happiness, different thoughts and emotions. (ICLE-RU)

5.51. In this essay I am going to talk about the link between crime and politics; what I want to demostrate is that a good way of making politics can cut the roots to crime. (ICLE-IT)


5.52. Thanks to them anyone willing to broaden his/her general knowledge of the world has an easy access to useful information. All in all, there are many ways in which mass media affect our approach to reality and they are, by no means, all positive or good for us. (ICLE-PO)

Gilquin and Paquot (2008) examined the use of some of the lexical items listed in Table 5.15 in the ten learner corpora used here as well as in four L1 sub-corpora (Norwegian, Japanese, Chinese, and Turkish) from the second version of the International Corpus of Learner English (ICLEv2) (Granger et al., 2009). The corpus totalled around 1.5 million words.

Table 5.15 Speech-like overused lexical items per rhetorical function

Rhetorical function Speech-like overused lexical item

Exemplifi cation like

Cause and effect thanks tosobecausethat/this is why

Comparison and contrast look likelike

Concession the (sentence-fi nal) adverb though

Adding information sentence-initial andthe adverb besides

Expressing personal opinion I thinkto my mindfrom my point of viewit seems to me

Expressing possibility and certainty reallyof courseabsolutely maybe

Introducing topics and ideas I would like to/want/am going to talk about

thingby the way

Listing items fi rst of all

Reformulation: paraphrasing and clarifying

Quoting and reporting say

Summarizing and drawing conclusions all in all


We compared the frequencies of speech-like lexical items in learner writing with their frequencies in the 10-million word spoken component (BNC-SP) and the 15-million word academic sub-corpus of the British National Corpus. Our fi ndings support Lorenz’s (1999b: 64) statement that there is ‘mounting evidence that text-type sensitivity does indeed lie at the heart of the NS/NNS numerical contrast.’ They show that the relative frequency of these speech-like lexical items in learner writing is often situated between their frequency in academic prose and in speech (see the bar charts for maybe, I would like/want/am going to talk about, really, absolutely, defi nitely, by the way and though in Figure 5.7). However some of these items (so expressing effect, it seems to me, of course and certainly) are even more frequent in learner writing than in speech.

The overuse of several of these speech-like lexical items has been high-lighted in a number of studies focusing on specifi c L1 learner populations. For example, Chen (2006) reports on the overuse of besides in Taiwanese student writing; Lorenz (1999b) discusses the marked overuse of the conjunction because and the adverb so in German learner writing; French, Spanish and Swedish learners’ heavy reliance on I think to express their personal opinion is reported by Granger (1998b), Neff et al. (2007) and Aijmer (2002); Japanese, French and Swedish learners’ overuse of of course is highlighted by Narita and Sugiura (2006), Granger and Tyson (1996) and Altenberg and Tapper (1998). Using the ICLE, my results suggest that these features are often shared by a large proportion of the learners investigated, irrespective of their mother tongues, and are therefore likely to be develop-mental or teaching-induced. It remains to be seen, however, whether lack of register awareness is a typical feature of EFL learner writing or whether it is a more general characteristic of novice writing. This issue will be touched upon in Section 5.4.

Different EFL learner populations, however, do not use speech-like lexical items similarly. Although all L1 learner populations overuse the adverb maybe when compared to the BNC-AC-HUM, Table 5.16 shows that relative frequencies differ widely across L1 populations. Another example is EFL learners’ use of I think, which is overused by all L1 learner populations while showing marked differences across learner L1 sub-corpora. As shown in Table 5.17, relative frequencies range from 17.29 occurrences per 100,000 words in the Polish learner sub-corpus (ICLE-PO) to 143.57 occurrences per 100,000 words in the Swedish one (ICLE-SW). This huge difference may be partly explained by L1 infl uence. Studies in contrastive rhetoric (e.g. Connor, 1996; Vassileva, 1998) have shown that features of writer visibility in academic prose may differ markedly across languages.


350 1200

1000

800

600

400

200

0

300

250

200

150

100

Frequency of maybe (pmw)

Frequency of it seems to me (pmw)

Frequency of by the way (pmw) Frequency of through at the end of a sentence(pmw)

Frequency of amplifying adverbs (pmw)

Frequency of I would like/want/am going totalk about (pmw)

Frequency of so expressing effect (pmw)

50

40 20

15

10

5

0

35

30

2520

15

10

2000

1500

1000

really of course certainly absolutely definitely

500

120

100

80

60

40

20

0

40

30

20

10

0

0

50

0

Academic writing: British National Corpus, academic component (15million words)Learner writing: ICLEv2 (14 L1s; 1.5million words)Speech: British National Corpus, spoken component (10million words)

Figure 5.7 The frequency of speech-like lexical items in expert academic writing, learner writing and speech (based on Gilquin and Paquot, 2008)


5.2.3. The phraseology of academic vocabulary in learner writing

In this section, I fi rst present the major results of an analysis of recurrent word sequences in EFL learner writing. I focus on aspects of over- and underuse of word sequences that include AKL words before discussing learner-specifi c clusters that are not found in professional academic prose. Learner writing is also typically recognizable by a whole range of co-occurrences that differ from academic prose in quantitative and qualitative

Table 5.16 The frequency of ‘maybe’ in learner corpora

relative freq. per

100,000 words

ICLE-IT 48.18 ++ICLE-GE 38.34 ++ICLE-DU 35.13 ++ICLE-CZ 32.88 ++ICLE-SP 32.28 ++ICLE-SW 31.21 ++ICLE-FI 24.74 ++ICLE-FR 20.34 ++ICLE-PO 16.37 ++ICLE-RU 13.26 ++BNC-AC-HUM 1.93

Legend: ++ frequency signifi cantly higher (p < 0.01) than in the BNC-AC-HUM

Table 5.17 The frequency of ‘I think’ in learner corpora

relative freq. per

100,000 words

ICLE-SW 143.57 ++ICLE-IT 134.06 ++ICLE-RU 121.13 ++ICLE-CZ 101.7 ++ICLE-FR 94.61 ++ICLE-GE 72.11 ++ICLE-SP 66.59 ++ICLE-FI 55.87 ++ICLE-DU 51.77 ++ICLE-PO 17.79 ++BNC-AC-HUM 6.14

Legend: ++ frequency signifi cantly higher (p < 0.01) than in the BNC-AC-HUM


terms. I illustrate this with a comparison of the co-occurrents of the noun conclusion in academic and learner writing and examine EFL learners’ phra-seological infelicities and lexico-grammatical errors.

An analysis of word sequences in EFL learner writing

The results presented in this section are based on an analysis of 2-to-5 word sequences that are over- or underused in learner writing. The compari-son between the ICLE and the BNC-AC-HUM was performed with the Keywords option of the software tool WST4. The results show that learner writing is characterized by a marked underuse of a large proportion of the 2-to-5 word sequences that include AKL words and that are typically used to serve specifi c rhetorical and/or organizational functions in academic prose. EFL learners rely instead on a restricted set of clusters which they massively overuse (e.g. for example, main reason, it depends, more and more, in order to, the problem is that). Granger (1998b) suggests that the use of these sequences ‘could be viewed as instances of what Dechert (1984: 227) calls “islands of reliability” or “fi xed anchorage points”, i.e. prefabricated formulaic stretches of verbal behaviour whose linguistic and paralinguistic form and function need not be “worked upon”’ (Granger, 1998b: 156). This is also consistent with the author’s statement that ‘while the foreign-soundingness of learners’ productions has generally been related to a lack of prefabs, it can also be due to an excessive use of them’ (Granger, 1998b: 155). The foreign-soundingness of EFL learner writing also stems from learners’ overuse of AKL words in clusters that are not typical of the particular genre of academic prose but are more frequently used in speech or more informal types of writing (e.g. people claim that, I will discuss, from my point of view, because of the fact7).

Table 5.18 shows that EFL learners overuse adjective + noun sequences with ‘nuclear’ adjectives (see Section 1.1.1) such as main (e.g. main reason, main cause, main problem), real (e.g. real problem, real value), important (e.g. important role, important question, important factor), great (e.g. great num-ber, great importance), different (e.g. different points, different problems, different reasons) and big (e.g. big problem) to the detriment of more EAP-like phrase-mes such as extensive use, crucial importance, central issue, signifi cant number, integral part, lesser extent and wide variety. Similarly, they overuse adverb + adjective/adverb /conjunction sequences with highly frequent adverbs such as mainly (e.g. mainly because), quite (e.g. quite clear) and very (e.g. very important) but make little use of phrasemes such as readily available, relatively few, signifi cantly different, almost entirely, closely associated, particularly interesting, more generally, highly signifi cant and precisely because.

Table 5.18 Examples of overused and underused clusters with AKL words

Overused clusters Underused clusters2-

wor

d cl

uste

rsfor example, for instance, important to, main reason, opportunity of, therefore I, and therefore, have problems, are concerned, another important, mainly because, only because, quite clear, different reasons, totally different, different way, more diffi cult, great importance, very important, main cause, main problem, absolutely necessary, because I, negative consequences, real problem, great amount, good idea, I consider, great part, important part, big problem, best solution, allows us, conclusion I, different points, we can, can choose, it depends, good use, good example, real value, important question, important factor, I can, different problems

by contrast, in particular, was probably, a similar, the view, suggestion that, described as, suggested above, was effectively, still further, more generally, readily available, relatively few, more signifi cantly, is ultimately, he concludes, on average, central issue, certain respects, radically different, consistent with, crucial importance, signifi cantly different, extensive use, fi nal analysis, they suggest, inferred from, listed above, general principles, inherent in, major source, particular attention, highly signifi cant, by comparison, considerable degree, perhaps because, much emphasis, he cites, provide evidence, little evidence, central fi gure, in practice, reports that, allowing for, what appears, discussed in, may suggest, reported by, precisely because, crucial role, integral part, wide variety, they argued, partly because, somewhat different, almost entirely, he remarks, his method

3-w

ord

clus

ters as a result, as a consequence, in my view, more and more, more or less, take into

account, advantages and disadvantages, aim of this, pay attention to, as a conclusion, take into consideration, of great importance, it means that, affect our approach, people claim that, I will discuss, may say that, prevents us from, provides us with,

in terms of, the absence of, the view that, extent to which, the implications of, an account of, a theory of, in relation to, an attempt to, closely associated with, a considerable degree, as distinct from, high degree of, high proportion of, it seems likely, various forms of, a concern with, to this extent, despite the fact, the hypothesis that, the issue of, this need not, at any rate, by reference to, in certain respects, were subject to, in his view, in view of, it was claimed, it follows that, by showing that, this suggests that, be ascribed to, when compared with, as noted above, is described in

4-w

ord

clus

ters

the problem is that, it is very diffi cult, the fact is that, is the fact that, it is also true, there are also people, a great number of, it is high time, it is obvious that, as much as possible, it is true that, to a great extent, because of the fact, to answer this question, in order to achieve, it is necessary for,

it may be that, may well have been, to the effect that, are likely to be, would seem to be, to the extent that, with the exception of, it does not follow, it seems likely that, in the presence of, the edge of the, it was diffi cult to, the immediate aftermath of, it is possible that, can be related to, similar to that of, the total number of, it is unlikely that, a wide variety of, in the absence of, to the advantage of, on the assumption that, as an attempt to, on the basis of, in the belief that, might have been expected, with the exception of, the extent to which, was by no means, in the presence of, no reason to suppose, with the result that, it would appear that, it is assumed that, may have been used

5-w

ord from my point of view, far as I am concerned, there are more and more, it is very

diffi cult to, but it is true that, this is not the case, as a matter of fact, it is very important to, it is a fact that, one of the most important

as in the case of, it has been suggested that, it could be argued that, in so far as they, it is more likely that, it is hardly surprising that, be defi ned in terms of, it is worth noting that, be explained in terms of


The results also seem to support the widely held view that EFL learners’ academic writing is characterized by ‘fi rmer assertions, more authoritative tone and stronger writer commitments when compared with native speaker discourse’ (Hyland and Milton, 1997: 193) (see also Petch-Tyson, 1998; Lorenz, 1998; Neff et al., 2004a). EFL learners state propositions more force-fully and make a more overt persuasive effort: they overuse communicative phrasemes that serve as attitude markers (e.g. it is very diffi cult to, it is very important to) and boosters (e.g. but it is true that, it is a fact that, it is obvious that). By contrast, they underuse hedges such as it is (more) likely that, it may be that, it seems likely that, it is possible that, it is unlikely that, and it would appear that.

Word sequences used as self mentions are also much more frequent in learner writing than in academic prose (Aijmer, 2002; De Cock, 2003; Ädel, 2006). Examples include therefore I, because I, I consider, we can, I can, in my view, I will discuss, provides us with, and from my point of view. Conversely, academic writers use more clusters with third person pronouns with an evidential function, e.g. he remarks, she cites, his method, they suggest, a differ-ence which can be related to the more intertextual nature of professional academic texts.

EFL learners also underuse a whole set of word sequences involving the –ed form of verbs, and more precisely, their past participle form. For example, they underuse the 2-word clusters described as, suggested above, inferred from, listed above, discussed in and reported by, the 3-word clusters closely associated with, it was claimed, be ascribed to, when compared with, as noted above, is described in; the 4-word clusters can be related to, might have been expected, it is assumed that, may have been used; and the 5-word clusters it has been suggested that, it could be argued that, be defi ned in terms of, and be explained in terms of. This is consistent with Granger and Paquot’s (2009b) fi nding that past participles are the most frequent verb forms in academic prose, but are highly underused in learner writing.

Verbs may have similar frequencies as lemmas in learner writing and academic prose, but still display over- or underuse of some forms (Granger and Paquot, 2009b). Examples of AKL verbs following this pattern are differ and discuss. The lemmas do not differ signifi cantly in their use. However, differ is underused in its –ing form while discuss is overused in its unmarked form (discuss) and underused in its –ed form. Similarly, some verbs are under- or overused as lemmas without this affecting all forms of the verb. For example, the lemma provide is underused in learner writing compared to expert writing, but an analysis of word forms indicates that this only applies to provided; use of other forms of the verb does not differ signifi -cantly in the two corpora. Table 5.19 shows that the picture can even be more complex: verb forms may be overused in some specifi c lexical bundles,


Table 5.19 Clusters of words including AKL verbs which are over- and underused in learners’ writing, by comparison with expert academic writing

Lemmas and their

word forms

Overused clusters Underused clusters

affect (++)affect (++)

affects (++)

affect our, media affect, affect us, affects the, affect our approach, media affect our, mass media affect, affect our approach, media affect our approach, mass media affect our, affect our approach to reality, media affect our approach to

was affected, not affect the

allow (++)allowed (++)

allowed to, not allowed, are allowed, not allow, be allowed, allow them, it allows, allows us, are not allowed, are allowed to, not allowed to, allow them to, be allowed to, allows us to, are not allowed to

allow them, allowed him, by allowing, allow it, allows for, to allow, allowing for, allow that, which allowed, allowed him, to allow for

concern (++)concerning (++)

is concerned, are concerned, am concerned, concerned about, it concerns, concerning the, I am concerned, as I am concerned, far as I am concerned

was concerned, been concerned, concerned to, concerned with, we are concerned, been concerned with, was concerned with, concerned with the, is concerned with the

depend (++)depends (++)

depending (++)depended (− −)

depends on, it depends, depending on, much depends, it depends on, depends on the, depending on the

depending upon, depended on, depended upon, will depend, depends upon the, depend upon the, will depend on

differ (//)differing (− −)

- differed from, differs from the

discuss (//)discuss (++)

discussed (− −)

will discuss, to discuss, I will discuss was discussed, already discussed, and discussed, discussed below, in discussing, discussed in, discussed in chapter

tend (//)tend (++)

tended (− −)

tend to, people tend, we tend, they tend, people tend to, we tend to, they tend to

they tended to, and tended to, has tended to, have tended to, tended to be

provide (− −-)provided (− −)

provides us, provide us, provide them, can provide, provide us with, provides us with, provide them with

might provide, provide the, provides that, provide an, provide evidence, to provide, provide a, provides an, provides a, was to provide, to provide an, to provide a

Legend: ++ signifi cantly more frequent (p < 0.01) in ICLE than in BNC-AC-HUM;− − signifi cantly less frequent (p < 0.01) in ICLE than in BNC-AC-HUM; // no signifi cant difference between the frequencies in the two corpora


while being underused in others. For example, the verb form concerned is overused in as I am concerned and concerned about but underused in been concerned with or we are concerned. Similarly, EFL learners overuse the sequences it allows and allows us to and underuse the EAP sequence allows for.

A keyword analysis of recurrent word sequences is indispensable if we want to build up a full picture of all the possible lexical realizations of rhe-torical functions in learner writing. It makes it possible to uncover a whole range of words and word sequences that are not typical of academic prose but which are nevertheless used by EFL learners to organize scientifi c dis-course and build the argument of academic texts. Examples of learner- specifi c sequences that do not include an AKL word are given in Table 5.20. They include:

– word sequences that are more frequently used in speech, e.g. of course, I think that, there are a lot of (see Section 5.2.2);

– sequences that are not used in English to establish the logical link intended by the EFL learner, e.g. on the other side (see Section 5.2.4 on semantic misuse);

– sequences that exist in English but are very rare in all types of discourse, e.g. the sequence as far as I am concerned which is repeatedly used to express a personal opinion in the ICLE;

Table 5.20 Examples of overused clusters in learner writing

Examples

2-word clusters in sum, of course, in fact, is why, let us, I think, instead of, look at, we must, or maybe, really think, there are, my opinion, if you, but I, if we, there is, thanks to, we want, sure that, I believe, people say, people think, when I, said that, I agree, many things, no matter, means that, opinion is, I want, everybody knows, people often, let them, we look, I hope, at all, people believe, even worse, I really, so why, we think, people feel, we get, I guess, just imagine, think twice, quite sure, why we, I must, very serious, helps us

3-word clusters in my opinion, in spite of, to sum up, fi rst of all, I think that, in order to, I would like, that is why, on the contrary, I believe that, to my mind, we have to, all kinds of, I would say, we all know, people think that, if we want, it means that, by the way, a look at, on one hand, I am convinced, people believe that, I will try, I agree that, and of course, everybody knows that, many people think

4-word clusters on the one hand, last but not least, I would like to, some people say that, we can say that, in this essay I, are more and more, I am sure that, there are a lot, it is impossible to, I don’t agree with, I want to say, but if we look, I am afraid that, it is easy to

5-word clusters I do not think that, as a matter of fact, from my point of view, I would like to say, far as I am concerned, it seems to me that, I do not agree with, but at the same time, due to the fact that, I do not think so


– ‘unidiomatic’ sequences such as as a conclusion used as a textual phraseme to introduce a conclusion (see below for a co-occurrence analysis of the noun conclusion in the ICLE);

– erroneous sequences such as in contrary, by the contrary, in the contrary, in contrary to that are used to express a contrast in EFL learner writing.

EFL learners’ overuse of sequences that are rarely used by native-speakers (such as as far as I am concerned or last but not least) or ‘unidiomatic’ sequences (such as as a conclusion) may be partly explained by poor teaching materials and/or the infl uence of their mother tongue. For example, Les fi ches essentielles du Baccalauréat en anglais (published in 2008 by Clairefontaine) give a list of linking words that French students are encouraged to use in the English test of the ‘Baccalauréat’ (the fi nal secondary school examina-tion which gives successful students the right to enter university) to ‘enrich their essay and give more clarity to their argumentation’. This includes as a conclusion but not in conclusion.8 The rare expression as far as I am concerned is also given as a key expression for voicing one’s own opinion.

Preferred co-occurrences in EFL learner writing

In Section 5.2.1, it was shown that EFL learners manifest a marked prefer-ence for a restricted set of single words and mono-lexemic phrasemes to express logical links. They also use learner-specifi c functional equivalents of these markers such as the sequence as a conclusion instead of in conclusion. This learner-specifi c word combination represents 39.2 per cent of the concluding textual phrasemes involving the noun conclusion in the ICLE. In a longitudinal study of German learners’ use of the noun conclusion, Mukherjee and Rohrback (2006) commented that the sequence as a conclu-sion is gaining ground in learner writing to the extent that it is even more frequent than in conclusion in the more recent corpus they use:

Interestingly, the most frequent phrase is no longer in conclusion, but as a conclusion. This certainly is a problematical development because in con-clusion is much more frequent and idiomatic than as a conclusion, the lat-ter being notoriously overused by German learners of English at university level as well. (Mukherjee and Rohrback, 2006: 224)

This development may be related to the increasing use of the internet for study purposes and of the type of teaching materials available on this channel, as discussed above. Another example of a learner-specifi c logical


marker is on the other side which they use instead of on the other hand to compare and contrast (see Section 5.2.4 below for more details of learners’ use of on the other side).

In Section 4.2.1, it was shown that mono-lexemic phrasemes such as for example have their own phraseological patterns in academic prose. However, these do not seem to be readily available to EFL learners, who tend to produce their own phraseological ‘cascades’, ‘collocational patterns which extend from a node to a collocate and on again to another node (in other words, chains of shared collocates)’ (Gledhill, 2000: 212).9 Figure 5.8 shows that the textual phraseme in conclusion (or one of its learner- specifi c functional equivalents) is very often directly followed by the personal pronoun I in the ICLE. This is consistent with Ädel’s (2006) fi nd-ing that personal metadiscourse, i.e. metadiscourse items that refer explic-itly to the writer and/or reader, serves a wide range of rhetorical functions (including exemplifying, arguing, anticipating the reader’s reaction, and concluding) in Swedish learner writing. The sequence in conclusion, I is generally followed by the modal would to produce the word sequence in conclusion, I would, which, in turn, very often introduces the sequence like to. The sequence in conclusion, I would like to either introduces the verb say or another verb of saying such as tell or mention.

EFL learners use AKL nouns and verbs in different lexico-grammatical or phraseological patterns than professional writers. This has already been illustrated by learners’ use of the noun example and the verbs illustrate and exemplify in Section 5.1. Another example is learners’ use of the noun conclusion. Table 5.21 lists the verb co-occurrents of the noun conclusion in the ICLE. Some 30.8 per cent of verb co-occurrent types are signifi cant co-occurrents of the noun conclusion in the BNC-AC . However almost half of the verb co-occurrent types (46.2%) used in the ICLE do not appear in the BNC-AC. When tokens are analysed, the percentage of verb co-occurrents

In conclusion 59 I 37 21would like to

say 6emphasize 2

mention 1speak about 1reiterate 1quote 1

tell 113(36%) (20.6%) (12.7%)As a conclusion 40

As conclusion 3

Figure 5.8 Phraseological cascades with ‘in conclusion’ and learner-specifi c equivalent sequences

162 A


earner Writing

Table 5.21 Verb co-occurrents of the noun conclusion in the ICLE

Verb + conclusion as

object

Freq. in

ICLE

Statistically signifi cant

co-occurrent in

BNC-AC

Appearance in

the BNC-AC

conclusion as

subject + verb

Freq. in

ICLE

Signifi cant

co-occurrent

in BNC-AC

Appearance

in the

BNC-AC

add up to 1 − x emerge 1 ** √apply 1 − x arise 1 − √approach 1 − x contain 1 − Xarrive at 5 ** √ be 23 ** √bring 1 − x come 1 − Xbring sb to 2 − √ need 1 − √come to 52 ** √ bring sb to 1 − X*come into 1 − xconfi rm 1 ** √contain 1 − xdraw 25 ** √*draw up 1 − xend with 1 − xescape 1 ** √express 1 ** √fi nd 2 − xgather 1 − xget 1 − √give 1 − √have 1 − √infl uence 1 − √jump to 2 ** √lead to 4 ** √

A

cademic vocabulary in the IC

LE

163leave sb with 1 − xlook for 1 − xmake 11 − √point to 1 * √put 1 − xput forward 1 − xreach 3 ** √write as 1 − x

TOTAL 128 tokens (32 types) TOTAL 29 tokens (7 types)

Legend: ** signifi cant co-occurrent in the BNC-AC (p < 0.01);− not signifi cant co-occurrents in the BNC-AC; √ the co-occurrent appears in the BNC-AC;x the co-occurrent is not found in the BNC-AC


that are signifi cant co-occurrents of the noun conclusion in the BNC-AC rises to 75.8 per cent as several of the verbs are repeatedly used in learner writing (e.g. come to and draw). Conversely, the percentage of verb co-occurrents that are not found in the BNC-AC falls to 12 per cent as ‘non-native’ co-occurrences are rarely repeated.

EFL learners use the collocations arrive at + conclusion, come to + conclu-sion, draw + conclusion, lead + conclusion and reach + conclusion. However, they do not always use them in native-like lexico-grammatical patterns. In Examples 5.53 and 5.54, the indefi nite article a is used instead of the defi nite article the, which is always used in the BNC-AC when the conclu-sion (underlined in the examples) is introduced by a that-clause. In Exam-ple 5.55, the frequent phraseme lead to the conclusion that is used with the personal pronoun us, a pattern which is very rarely found in academic prose. In the context of EFL teaching/learning, these fi ndings support Nesselhauf’s (2005: 25) argument that collocations should not be viewed as involving only two lexemes; other elements closely associated with them should also be taught.

5.53. However, when we consider all the pros and cons of fast food we will certainly arrive at a conclusion that it is not an ideal way of eating. (ICLE-PO)

5.54. And taking into consideration that Marx was a materialist we can come to a conclusion that he himself would be attracted by the advantages of television, and religion for him would remain the opium of the masses. (ICLE-RU)

5.55. To sums up, all I have mentioned before lead us to the conclusion that if our lifes were a little ‘easier’ and we wouldn’t be dominated by a world that is constantly changing, due to new techniques and industrialization, we could enjoy doing things as dream and imagine more frequently. (ICLE-SP)

The collocation escape + conclusion appears in two phraseological patterns in academic prose: ‘it is diffi cult to escape the conclusion that’ and ‘we cannot escape the conclusion that’. The single occurrence of the collocation that appears in the ICLE is used in the native-like lexico-grammatical pattern ‘cannot escape the conclusion that’ but its subject is a nominal phrase headed by the noun evaluation:

5.56 However, a more objective evaluation of the problem cannot escape the conclusion that, drug use and abuse have occurred in all civilizations all over the world, and that it is the criminalization of drugs that has created a much heavier burden on society. (ICLE-DU)


In the collocation express + conclusion, the verb express has acquired a semi-technical sense and means ‘make something public’. It is mainly used in legal discourse and thus conveys a rather formal tone as illustrated in Example 5.57. Its single occurrence in the ICLE can be qualifi ed as ‘non-native like’ as it appears with the fi rst person singular pronoun I as subject and the possessive determiner my (Example 5.58). It may be hypothesized that the learner who wrote this sentence has been infl uenced by the native-like co-occurrence ‘express one’s opinion/view’.

5.57. The Divisional Court expressed its conclusion in the following terms: . . . (BNC-AC-HUM)

5.58. Finally, I wanted to express my conclusions. (ICLE-SP)

There are many other examples of EFL learners’ attempts at using native-like collocations, which result in crude approximations. In Example 5.59, the phrasal verb draw up is used in place of draw and in Example 5.60, the preposition into replaces to, and no article is used in an attempt to produce the native-speaker sequence ‘came to the conclusion that’.

5.59. Finally, a conclusion can be drawn up emphasizing our fi rst statement, that is: technology, science and industrialization have not killed dreams and imagination. (ICLE-SP)

5.60. The woman started to think about the price of progress and came into conclu-sion that automation causes more problems than it solves. (ICLE-PO)

In Example 5.61, the verb put forward is used with the noun conclusion. This verb is commonly used with the abstract nouns plan and proposal, two nouns that, like conclusion, combine with the verb draw to form collocations. However, the verb put forward is not used with the noun conclusion in English (see Figure 5.9). Howarth (1996; 1998) refers to this phenomenon as a collocational overlap, i.e. a set of nouns which have partially shared collocates (see also Lennon, 1996).

draw

put forward

plan

proposal

conclusion

Figure 5.9 Collocational overlap


5.61. Without putting forward premature conclusions, we can nevertheless notice that a certain importance is granted to them. (ICLE-FR)

The semantic incongruity of the co-occurrence ‘put forward a conclusion’ is made apparent by contrasting the defi nitions of put forward and conclusion. The verb put forward means ‘to suggest an idea, explanation etc, especially one that other people later consider and discuss’ (LDOCE4) while a conclusion is ‘something you decide after considering all the infor-mation you have’ (LDOCE4). Thus, a conclusion can hardly be put forward as it is supposed to be more than a suggestion and the result of serious con-sideration and discussion.

As already pointed out by Nesselhauf (2005), EFL learners also produce deviant verb + noun free combinations. The noun conclusion enters into combinations that are not found in academic prose and which are semanti-cally awkward:

5.62. Looking for the conclusion I would like to say that every person is individual and each has his or her own character. (ICLE-RU)

5.63. Having considered the various aspects of capitalism a conclusion must be gathered: the system cannot provide for the basic needs of the population; consequently it needs to take steps in order to prevent combativity which will endangered their interests. (ICLE-SP)

The same remark can be made about several adjective + conclusion co- occurrences (Example 5.64). More importantly, however, adjective co-occurrents of the noun conclusion in learner writing are not the most typical ones in academic prose even though a large proportion of them occur in the BNC-AC (see Table 5.22). The fi rst ten most signifi cant adjective co-occurrents of the noun conclusion in the BNC-AC are general, logical, tentative, similar, foregone, main, fi rm, different, opposite, and defi nite. None of these appear in learner writing except for logical. This reveals learn-ers’ weak sense of native speakers’ ‘preferred ways of saying things’.

5.64. Looking at this idea from the Polish point of view, also brings double standard conclusions. (ICLE-PO)

The phraseology of EFL learner writing is also characterized by a number of lexico-grammatical infelicities and errors. Learners sometimes use the preposition about after the abstract noun account (e.g. an account *about a murder (ICLE-RU)) or the preposition of instead of for after the noun demand (e.g. the demand *of raw material (ICLE-GE)). They also use


Table 5.22 Adjective co-occurrents of the noun conclusion in the ICLE

Adjectives Frequency Signifi cant co-occurrents of

conclusion in the BNC-AC

Appearance in the

BNC-AC

absolute 1 − √

awful 1 − xcertain 3 ** √clear 1 ** √clever 1 − xconcrete 1 − √depressing 1 ** √double standard 1 − xfair 1 − √

false 1 − √fi nal 5 ** √frightening 1 − xinteresting 1 − √liberal 1 − xlogical 4 ** √long-searched for 1 − xobvious 1 ** √overall 1 ** √only 1 − √own 4 ** √personal 1 − √premature 1 − √private 1 − xradical 1 − √right 2 − √sad 1 − xsame 2 ** √satisfactory 2 ** √satisfying 1 − xsensible 2 − √successful 1 ** √terrifying 1 − xunderstated 1 − xunequivocal 1 − √wrong 1 − √

TOTAL 51 tokens (35 types)

Legend: ** signifi cant co-occurrent in the BNC-AC (p < .01);− not signifi cant co-occurrents in the BNC-AC; √ the co-occurrent appears in the BNC-AC;x the co-occurrent is not found in the BNC-AC.


a to-infi nitive structure after the noun possibility instead of an –ing form (e.g. the possibility *to learn a good job (ICLE-FR)). Other examples of colliga-tional errors include suggest *to, related *with, attempt *of, and discuss *about. Example 5.65 illustrates learners’ confusion between the prepositions despite and in spite of, which results in the blend *despite of (cf. Dechert and Lennon, 1989):

5.65. Despite *of [Despite] the absence of such professionalism our nation overcame fascists. (ICLE-RU)

Learners also have a tendency to use the impersonal pronoun it in the subject position after as:

5.66. It is a matter of fact that these ‘things’ cannot be bought and sold like shares on the stockmarket. Luckily, I would say because otherwise only the rich would be able to posses them as *it is [as is] unfortunately the case with many prod-ucts in other areas of living. (ICLE-GE)

5.67. Because of the ambition for the power, their rivalry made them hold continuous battles, as *it was [as was] the case of Catholics and Protestants. (ICLE-SP)

Another source of error is the adjective same which is sometimes preceded by the indefi nite article in the ICLE:

5.68. The negative image of feminism makes it twice as hard for women to rise above it than it would be if men were facing *a [the] same kind of dilemma. (ICLE-FI)

5.69. When different people read *a [the] same book they have probably various imaginations while reading. (ICLE-CZ)

It should be noted that very few of these errors are widespread in learner writing and that some of them may be partly L1-induced. For example, French learners use the erroneous colligation discuss *about as a translation of the French discuter de (Granger and Paquot, 2009b).

5.2.4. Semantic misuse

In Section 5.2.1, the function of comparing and contrasting was shown to be generally underused in learner writing. An analysis of individual lexical


items, however, reveals that the adverbials on the contrary and on the other hand are overused in the ICLE. As Lorenz (1999b: 72) has demonstrated, overuse is often accompanied by patterns of non-native usage. EFL learn-ers’ semantic misuse of the phraseme on the contrary has already been reported for different learner populations:

In Hong Kong, we are all familiar with students who use ‘on the contrary’ for ‘however/on the other hand’, thus adding an unintended ‘corrective’ force to the merely ‘contrastive’ function sought. (Crewe, 1990: 317)

Granger and Tyson (1996) report the same conceptual problems for French learners. Lake (2004) states that a large proportion of EAP non-native speakers who use on the contrary do so inappropriately. This is confi rmed by our corpus-based analysis of EFL learners from different mother tongue backgrounds (see also Celce-Murcia and Larsen-Freeman, 1999: 534-535). EFL learners typically use on the contrary erroneously (instead of a contras-tive discourse marker such as on the other hand or by contrast) to contrast the qualities of two different subjects (underlined in Examples 5.70 to 5.72). Thus, in Example 5.70, the fact that Onasis had everything is contrasted with the fact that Raskolnikov had nothing and the phraseme by contrast would have been more appropriate.

5.70. Raskolnikov differs from Onasis, of course. Onasis had everything but he wanted to have more. Raskolnikov, *on the contrary [by contrast], had nothing. (ICLE-RU)

5.71. The young like crazy driving, overtaking and leading on the roads. Sports cars are created for this use and this may be the reason why their price is so high and use is expensive. *On the contrary [By contrast], station wagons are not expensive in maintenance. The main users of this kind of vehicles are families. (ICLE-PO)

5.72. For instance, most Americans have moved to the USA from different coun-tries as immigrants. *On the contrary [By contrast], Europeans have lived in their countries for hundreds of years. (ICLE-FI)

The semantic inappropriacy of on the contrary in EFL learner writing has been attributed to teaching practices. Teaching materials often provide lists of connectors in which the adverbial on the contrary is described as a phrase of contrast, that is, as an equivalent alternative to on the other hand, by contrast, etc. (cf. Crewe, 1990). For pedagogical purposes, Lake (2004)


proposes a checklist of contextual features that should be present when on the contrary is employed:

As for the implications for learners, it now becomes possible to consult a checklist of contextual features that should be present in order for on the contrary to be appropriate:

one subject;two contrasting qualities;one positive statement and one negative statement open to similar interpretations;an argument, either genuinely present or implied, to which the two statements, adjacent to the phrase both form a refutation.

Such a checklist may be simplistic in that it does not cover all the possible lexico-syntactical environments in which the phrase might be encoun-tered; but as a guideline for production, it ought to prove a useful starting point from which EAP teachers can devise their own practice materials. (Lake, 2004: 142)

Lake (2004) rules out the possibility of an L1 infl uence on EFL learners’ semantic misuse of on the contrary on the basis that over 70 per cent of international students from widely different mother tongue backgrounds produced two distinctly separate L1-equivalent items in a cloze test in which they were required to insert on the contrary or on the other hand and provide an equivalent phrase for both adverbials. It is, however, probable that misguided teaching practices and L1 interact here. The L1 equivalent forms to on the contrary and on the other hand may be characterized by different patterns of usage and thus be the source of negative transfer. Granger and Tyson (1996), for example, argued that French learners’ overuse and misuse of on the contrary is probably due to an over-extension of the semantic properties of the French ‘au contraire’, which can be used to express both a concessive and an antithetic link. The potential infl uence of the fi rst language on French learners’ use of on the contrary is discussed in Section 5.3 below.

Lake (2004) considers EFL learners’ misuse of on the contrary to be ‘some-thing of an exception’ and writes that ‘in the EAP context, such functional phrases [connectives] are usually familiar to learners from an early stage, and do not pose great problems of usage’ (Lake, 2004: 137). This view, however, is over-optimistic and is clearly not refl ected in our corpus-based learner data. In Section 5.1, EFL learners’ inappropriate use of the abbre-viation i.e. (in lieu of e.g.), the preposition as (instead of such as) and the


adverb namely was discussed. Other examples of semantically misused lexical items in learner writing include on the other hand, on the other side, moreover, besides, and even if.

Field and Yip (1992: 25) reported that on the other hand is frequently used by Cantonese speakers to make an additional point, with no implied contrast. They suggested that this semantic misuse might be L1 induced: the Chinese equivalent of on the other hand is often misused by novice L1 writers, who use it to mean ‘another side or aspect’. Although L1 infl uence may play a part in Hong Kong Chinese students’ inappropriate use of the adverbial, erroneous uses of on the other hand are found in most ICLE sub-corpora, which suggests that there are other contributing factors to this learner diffi culty. The following extracts are examples of the use of on the other hand in the ICLE where it would have been more appropriate to use no connector or an additive marker:

5.73. I strongly believe that there is still a place for dreaming and imagination in our modern society. [P]10 Firstly, where there is a child, there are always dreams and imagination. Everybody knows that children like inventing funny stories and amusing plays by using their wide fantasy. This is one reason why children always bring happiness and awake the adults’ childish part. *On the other hand, fantasy is [also] a useful mean used by teachers in primary schools to teach school subjects to their little students. So, it is children who keep dreams and imagination alive! (ICLE-IT)

5.74. The re-introduction of the death penalty may have positive sides, too. Criminality would be limited, because criminals would be afraid of the severe punishment. [P] This might be an illusion, because *on the other hand [ø] the death penalty develops violence and is incompatible with the basic laws of humanity. (ICLE-GE)

5.75. The function of punishment is to show that crimes are not acceptable or that they can solve any problems. *On the other hand the aim of punishments is [also] to make the criminals obey the laws and show example to other’s so that they will not follow the bad example and commit the same crime. (ICLE-FI)

The word combination on the other side sometimes appears in the ICLE in places where a contrast seems to be the logical link intended by EFL learn-ers. This does not occur in academic prose. It is illustrated in the following examples:

5.76. Poland cannot reply with isolation as the unifi cation still remains the best solution to its problems. On the other side, all countries should understand


that history and its consequences cannot divide the continent. The successful process of unifi cation should be carried out with respect to nations’ rights and without special privileges given to the powerful. (ICLE-PO)

5.77. Another big problem is our environment. There is pollution wherever you look. We can no longer enjoy the sun in summer because of the hole in the ozone layer. This hole is caused by technical improvements in the last decades. But on the other side it is sometimes hard to live without car or aerosols. (ICLE-GE)

5.78. Europe 92 means well a loss of identity since we’ll be no longer Belgians, Italians, English ... but Europeans. But on the other side we will form a new nation with new hopes, new ideas . . . (ICLE-FR)

There is also some confusion between the conjunctions even if and even though in EFL learner writing. Learners often use even if in lieu of even though to introduce a concession:

5.79. However,*even if [even though] I agree that the American public school system is defective, home schooling to me is no real alternative, as I feel that parents are not the best teachers for their own children. (ICLE-GE)

5.80. We must forget about refrigerators containing CFC-11 and CFC-12, *even if [even though] they are cheaper. (ICLE-PO)

5.81. We are as much a part of Europe as any other country here, *even if [even though] we are not in the European Union. (ICLE-SW)

Even if should be used to introduce a condition, not a concession. Compare:

5.82. Even if these descriptions are valid they still leave open a number of questions, particularly why the same mechanisms do not operate with girls.

5.83. Even though these descriptions are valid they still leave open a number of questions, particularly why the same mechanisms do not operate with girls.

In the second sentence, the writer knows and accepts that the descriptions are valid. In the fi rst sentence, he or she does not.

Semantic misuse has often been discussed in the literature in relation to logical connectives. However, EFL learners also experience diffi culty with the semantic properties of other types of cohesive devices, and more specifi cally, labels, i.e. abstract nouns such as issue, argument, and claim that are inherently unspecifi c and require lexical realization in their co-text, either beforehand or afterwards (Flowerdew, 2006). In addition to phraseo-logical and lexico-grammatical infelicities, EFL learners’ use of labels is


characterized by semantic infelicity or lack of semantic precision. Learners, for example, use the noun problem as an ‘all purpose wild card’ (cf. Lorenz, 1999b) in lieu of more specifi c nouns such as issue or question as illustrated in the following sentences:

5.84. This short discussion of the main points linked to the problem [issue] of capital punishment leads to the fi nal question. (ICLE-PO)

5.85. The most important question concerning genetic engineering is the problem [that] of gen manipulation with humans. (ICLE-GE)

5.86. If we are aware of the fact that such time-tables are very common for people living in a modern society like ours, the problem [question] of the place of imagination and dreaming is not even worth examining. Industrialisation has transformed dreaming into a waste of time which is now “cleverly” linked to money. (ICLE-FR)

The noun argument also seems to cause diffi culty to EFL learners. In Example 5.87, the rather unidiomatic expression ‘familiar arguments about’ should be rephrased as ‘widespread or popular beliefs about’. In Example 5.88, the sentences that follow the label argument would be better described as ‘reasons’ why Big Tobacco did not depart from pre-pared statements.

5.87. Female participation in making decisions concerning war and peace, economy and environmental protection would be to the benefi t of all. However it will not be possible until males re-think and, hopefully, reject familiar arguments [widespread/popular beliefs] about women being unreliable, irrational and dependent on instincts. (ICLE-PO)

5.88. There are two main arguments [?reasons] that help us understand why Big Tobacco stuck to their statements for so long. [P] First, the companies feared the consequences that would follow a confession. They feared that there was going to be even more legislation and regulation if they would ever admit to lying. . . . . (ICLE-DU)

Other problematic labels include, among many others, aspect and issue. In Example 5.89, another aspect introduces a second example (about the unemployed and housewives) of the fact that you are judged by what you do rather than by what you are, contrasting it with the fi rst example (about physicists and mathematicians). In Example 5.90, in certain aspects stands for in some respects and the aspect of money probably refers to the ‘money issue’ or the ‘money question’ in Example 5.91.


5.89. Our modern western society puts a lot of pressure on people as far as work is concerned. Your job is your “trademark”. Or, in other words, you are judged by what you do rather than by what you are. Sad, but true. For example, according to popular opinion you must be very intelligent if you are a physicist or a mathematician. And another aspect is that [?by contrast,] the unem-ployed or housewives are sometimes treated as social outcasts. (ICLE-GE)

5.90. Actually, bits of information from the remotest parts of the globe reach us in an instant. Human beings can eventually feel as one great family, but only *in certain aspects [in some respects], for as far as real good relations among countries are concerned, it is still a matter of distant future. (ICLE-PO)

5.91. A legend exists that money was invented by the devil to tempt the mankind. The aspect [?issue/question] of money includes the problem of equality. There were and there are different ideas about making all people equal, because it was considered that this would lead to common happiness. (ICLE-RU)

In Example 5.92, it is not quite clear what her issues refer to and in Example 5.93, issue most probably stands for ‘product’:

5.92. Uta Ranke-Heinemann, the most famous woman in the fi eld of Catholic theology, tries to provide answers to them. Her issues [?] lies on the verge of theology, philosophy and fi rst of all, religion. She is employed in defi ning the relation between faith and the mind. (ICLE-PO)

5.93. The picture I draw from my dear old houseman admittedly is nothing but a mere cliché, a hyperbolic issue [product] of my vivid imagination. (ICLE-GE)

5.2.5. Chains of connective devices

EFL learners’ texts are sometimes characterized by the use of too many connective devices (Crewe, 1990; Chen, 2006; Narita and Sugiura, 2006). The following text is an excerpt from an essay written by a French-speaking EFL learner. Each sentence contains at least one connective device – typically an adverbial connector or a sentence stem – which is often found in sentence-initial position (see Section 5.2.6 below).

5.94. [1] But what about these prestigious institutions today? [2] To caricature them rapidly one could say that universities consist of courses given by profes-sors (competent in their fi elds) in front of a silent audience who is conscien-tiously taking notes. [3] So one can wonder if a university degree really prepare students for real world and what his value is nowadays. [4] I think it is true that lectures in themselves are theoretical. [5] Firstly because students spend


most of their time sitting in big classrooms which do not allow practical exercises but only ex cathedra lectures. [6] Secondly because the subjects of the lectures are theoretical. [7] For example: during a general methodology course (which, we think, could be more practical) different theories as Krashen’s, Lado’s are studied in detail but practical points are hardly ever considered. [8] However is it true that this formation does not prepare students for real world? [9] I am of the opinion that the answer is no. [10] First I think that university degrees are theoretical on purpose (as opposed to high schools which are more practical.). [11] The reason is that, thanks to the theoretical back-ground they have learned, university students are able to build up their own way to achieve their aim. [12] Moreover they are also able to adapt or to modify their method according to the situation. [13] To take the example of a teacher again, I could say that a teacher in front of a classroom does not think about particular methodological theories again but that he has created his own methodology. [14] Secondly, I think that academic studies develop a critical mind. [15] The students are indeed trained to analyse pieces of information coming from different horizons from a critical point of view, which means that they have to dissect them, to confront them and then to be able to pass judgment on them. [16] That is the way they should create a personal opinion for them-selves. [17] Nevertheless, I do not want to go too far. [18] I really think that theory is essential but I am convinced that practice should also be present. [19] Let’s take the example of a student in economics who has his certifi cate in his pocket and proudly goes working in a big fi rm for the fi rst time. [20] I would compare this business man to a gentleman who perfectly knows the highway code and who knows how to start and how to run through the gears but who fi nds himself in the center of Paris at the peak hours the fi rst time he really drives! [21] By this example, I want to show that theory must always be accompagnied by practical applications, which is not often the case at univer-sity. [22] I think that this is a fully justifi ed criticism against this institution.

Some EFL learners use many logical connectives between sentences simply to indicate to the reader that they are adding another point (e.g. fi rstly, secondly, for example, fi rst, moreover, to take the example of). Several of these connectors are superfl uous and sometimes wrongly used (e.g. moreover in sentence [12], indeed in sentence [15]). Crewe (1990) attributed EFL learners’ massive overuse of connective devices to their attempt at imposing ‘surface logicality on a piece of writing where no deep logicality exists’ (Crewe, 1990: 320). He added that ‘over-use at best clutters up the text unnecessarily, and at worst causes the thread of the argument to zigzag about, as each connective points it in a different direction’ (ibid: 324). The following excerpt from an EFL learners’ essay is a good example


of EFL learners’ use of logical connectors as ‘stylistic enhancers’, i.e. ‘words or expressions that may be sprinkled over a text in order to give it an “educated” or “academic” look’ (Crewe, 1990: 316) but whose presence will not make the text coherent.

5.95. Furthermore, Hobbes is a stern determinist. He regards man, like nature, as subject to the chain of cause and effect. Therefore a concept like ‘free will’ is impossible. Hobbes even considers people as artifi cial creatures, not belonging to nature, because they are not able to live together in harmony, something which animals like bees and ants are capable of, because they are natural. Of course, these ideas were as much an insult to man’s estimation of himself as Darwin’s allegation, two hundred years later, that our ancestors used to live in trees. As a consequence, Hobbes was accused of being an atheist and forbid-den to publish any more books. (ICLE-DU)

As Aijmer (2001) showed in a study of Swedish EFL student writing, learn-ers use I think to make their claims more persuasive rather than to express a tentative degree of commitment. They often use I think or an equivalent expression (e.g. I am of the opinion that, I am convinced that) when it is communicatively unnecessary in the fl ow of argumentation. For example, Sentence [18] in Example 5.94 could be rephrased as ‘Theory is essential but practice should also be present’. The sequence I think it is true in Sentence [4] corresponds to what Aijmer (2001) described as a ‘rhetorical overstatement’, which the author regards as typical of non-native-speaker argumentative essays. The clusters To me, I think and as far as I am concerned, I think that in Examples 5.96 and 5.97 respectively are two more instances of rhetorical overstatement.

5.96. To me I think technology and imagination are very much interrelated, and then on the other hand I understand that they also can be seen as separate. (ICLE-SW)

5.97. I agree with George Orwell, because as far as I am concerned I think that in every country there are few people which are rich and many people which are poor. (ICLE-IT)

The pedagogical implication of these fi ndings is that, ‘important as these links are, learning when not to use them is as important as learning when to do so. In other words, students need to be taught that excessive use of linking devices, one for almost every sentence, can lead to prose that sounds both artifi cial and mechanical’ (Zamel, 1983: 27).


5.2.6. Sentence position

Linking adverbials occur in different sentence positions. They often occur initially, as does however in Example 5.98. They can also occur in a medial position, i.e. within the sentence, often immediately after the subject, as shown in Example 5.99. The fi nal position is also possible, but is more typi-cal of speech as illustrated in Example 5.100.

5.98. In practice, the Red Army units did nothing to conciliate the Ukrainian Left or the peasants. Agriculture was brutally collectivized and no conces-sions were made in the use of the Ukrainian language and culture. However, Denikin’s White armies counter-attacked and after seven months the Red Army was obliged to withdraw. (BNC-AC-HUM)

5.99. Coysevox’s bust of Lebrun repeats – again with a certain restraint – the general outlines of Bernini’s bust of Louis XIV. The face, however, shows a realism and subtlety of characterization that are Coysevox’s own. (BNC-AC-HUM)

5.100. It’d be worth asking him fi rst, though. (BNC-SP)

EFL learners’ marked preference for the sentence-initial position has been reported in various studies focusing on one L1 learner populations (Field and Yip, 1992; Lorenz, 1999b; Zhang, 2000; Narita and Sugiura, 2006). Granger and Tyson (1996: 24) commented that ‘it is likely that this tendency for learners to place connectors in initial position is not language-specifi c’. Our analysis of connectors in the ICLE supports this hypothesis. Table 5.23 shows that the total proportion of sentence-initial connectors in learner writing is much higher than that found in academic prose (13.2% compared to 6%). Examples include the preposition despite which appears in sentence-initial position in 52 per cent of its occurrences in the ICLE but only in 34.5 per cent in the BNC-AC-HUM (see Example 5.101), and sentence-initial due to which is repeatedly used in learner writing but hardly ever occurs in academic prose (Example 5.102).

5.101. Despite its commercial character Christmas still means a lot to me. (ICLE-FI)

5.102. Due to these developments the production expanded enormously, which meant that a greater number of people could be fed. (ICLE-DU)

Another example is the adverb therefore which often appears in sentence-initial position in the ICLE but is not often used in that position in the BNC-AC-HUM:


5.103. Scientifi c research as well as individual observations prove that eating habits have a great impact on the condition of the human body and soul and, con-sequently, on rest, sleeping and even dreams. Therefore people should pay more attention to what they consume. (ICLE-PO)

These fi ndings provide evidence for EFL learners’ lack of knowledge of the preferred syntactic positioning of connectors in English.11 This lack has often been attributed to L2 writing instruction. Flowerdew (1993) argued that teaching materials do not provide students with authentic descriptions of syntactic patterns of words. He showed that, contrary to what is often

Table 5.23 The frequency of sentence-initial position of connectors in the BNC-AC-HUM and the ICLE

ICLE BNC-AC-HUM

S-I Total

freq.

% Rel. freq.

pmw

S-I Total

freq.

% Rel. freq.

pmw

although 263 522 50.4 225.6 676 2,276 29.7 203.5and 1456 32,236 4.5 1249 1374 91,306 1.5 413.6as a result 71 103 68.9 60.9 65 102 63.7 19.6as a result of 24 79 30.4 20.6 22 194 11.3 6.6as far as X is

concerned96 167 57.5 82.4 31 59 52.5 9

because 107 2,493 4.3 91.8 151 2,207 6.79 45.4because of 62 530 11.7 53.2 46 599 7.67 13.8consequently 103 179 57.5 88.4 60 143 42 18despite 50 96 52 42.9 235 681 34.5 70.7due to 29 246 11.8 24.9 3 195 1.5 0.9even if 83 274 30.3 71.2 94 451 20.8 28.2even though 46 127 36.2 39.5 28 248 11.3 8.4for example 235 854 27.5 201.6 233 1263 18.4 70for instance 93 344 27 79.8 86 609 14.1 25.9furthermore 113 127 96.6 96.9 176 217 81.1 53however 673 1,128 59.7 577.4 882 3,353 26.3 265.5in spite of 47 106 44.3 40 42 159 26.4 12.6moreover 255 292 87.3 218.8 365 495 73.7 109.9nevertheless 170 250 68 145.8 392 676 58 118on the

contrary92 164 56.1 78.9 48 95 50.5 14.4

on the other hand

228 418 54.5 195.6 155 372 41.7 46.7

so 805 1,436 56 690 675 1,894 35.6 203.2thanks to 68 199 34.2 58.3 5 35 14.3 1.5therefore 340 689 49.3 291.7 75 1,412 5.3 22.5thus 221 446 49.5 189.6 756 1,767 42.8 227.6

TOTAL 5,730 43,505 13.2 4,916.24 6,675 110,808 6 2009


taught in course books, the adverbial connector then rarely occurs in sentence-initial position, but is more usually found in a medial position. Similarly, Milton (1999: 225) discussed the problematic aspects of teaching connectors by means of lists of undifferentiated items, and suggested that one way in which instruction may skew EFL learners’ style is ‘by the presentation of these expressions as if they occurred in only sentence-initial position’ (see also Narita and Sugiura, 2006). Thus, EFL learners’ tendency to place connectors in unmarked sentence-initial position seems to be reinforced by teaching (see Granger, 2004: 135).

Unmarkedness provides another possible explanation for EFL learners’ massive overuse of sentence-initial connectors. Conrad (1999) studied variation in the use of linking adverbials across registers. She showed that, in both conversation and academic prose, the highest percentage of linking adverbials appear in sentence-initial position and concluded that ‘initial position seems the unmarked position for linking adverbials’ (Conrad, 1999:13) (see also Biber et al., 1999 and Quirk et al., 1985). EFL learners seem to use the unmarked sentence-initial position as a safe bet.

Contrary to our expectations, the proportion of sentence-initial because is lower in learner writing than in professional writing. However, sentence-initial because is signifi cantly more frequent (relative frequencies of 9.18 in learner writing and 4.54 in academic prose). It is also used to serve different functions in learner writing. In academic prose, sentence-initial because-clauses are attached to a main clause. As shown in the following examples, they introduce the cause of something that is described in the main clause:

5.104. Because these changes were worldwide, Europe’s history is inseparable from world history between 1880 and 1945. (BNC-AC-HUM)

5.105. Because the death-rate was high, marriages were usually short-term. (BNC-AC-HUM)

Unlike expert writers, EFL learners sometimes use sentence-initial because to introduce new information in independent segments and give the cause of something that was referred to in the previous sentence:

5.106. The crime rate would also strongly reduce and this is of course the main objec-tive of all this measures. Because everybody wants to live in a safe society. (ICLE-DU)

5.107. To directly try to change people with ‘experience of life’ would, at best, only be to win Pyrrhic-victories, compared to this effective investment. Because deep inside every man’s heart lies the ‘Indian’-insight that we are only borrowing the earth from our children. (ICLE-SW)


5.108. In my opinion it is useful only for them, for their trial. Because their sorrow is found as the extenuating circumstance. (ICLE-CZ)

EFL learners share this characteristic with ESL writers. In a comparison of strategies for conjunction in spoken English and English as a Second Language (ESL) writing, Schleppegrell (1996) found that students who had spent most of their lives in the US and learnt English primarily through oral interaction, transferred conjunction strategies from speech to essay writing. They employed both an ‘afterthought’ because (Altenberg, 1984) to add information in independent segments, and other types of speech-like clause-combining strategies.

Conrad (1999) reported that, in academic prose, most linking adverbials are placed in sentence-initial or medial position. Three types of medial position are particularly frequent (Conrad, 1999: 14–15):

1. Linking adverbials which occur immediately after the subject as illustrated in Example 5.99 above.

2. Linking adverbials which occur between an auxiliary and the main verb, such as: All estimates of population size must therefore allow for a large measure of conjecture, a fact stressed by all reputable modern historians who have worked on this intractable subject. (BNC-AC-HUM)

3. Linking adverbials which occur between the main verb and its comple-ment, e.g.: It is diffi cult to believe therefore that one of these mosaics was not infl uenced by the other. (BNC-AC-HUM)

A medial position for connectors is quite typical of academic prose. However, it is clearly less favoured by EFL learners. As indicated above, teaching materials tend to focus on sentence-initial position, and EFL learners probably feel unsafe about other syntactic positionings for connectors.

Table 5.24 shows that several connectors are also repeatedly used in sentence-fi nal position in the ICLE, which is quite uncommon in the BNC-AC-HUM. The fi nal position is frequent in conversation, but rare in academic prose. Conrad (1999) found that three highly frequent items – then, anyway and though — account for the relatively high proportion of sentence-fi nal linking adverbials in native-speaker’s conversation. She argued that these linking adverbials are commonly found in sentence-fi nal position as they serve important interpersonal functions:

Adverbials in conversation, in addition to showing a link with previous discourse, can also play important roles in the interpersonal interaction


that takes place. These roles are often particularly noticeable for the common adverbials in fi nal position. (. . . ) [A] fi nal though often occurs when speakers are disagreeing or giving negative responses, fi nal anyway is often associated with expressions of doubt or confusion, and (. . . ) then typically indicates that a speaker is making an inteference (sic) based on another speaker’s utterance. The placement of these adverbials in fi nal position is consistent with previous corpus analysis of conversation that has found that elements with particular interpersonal importance are often placed at the end of a clause (. . . ). It may be, then, that in some cases in conversation there is a tension between placing the linking adver-bial at the beginning of the clause, due to its linking function, and at the end of the clause, due to its interactional function. (Conrad 1999:14)

The type of interpersonal interaction that takes place in conversation is not typical of academic prose. Thus, none of the linking adverbials commonly associated with the fi nal position in conversation are common in formal writing. These fi ndings suggest that the positioning of linking adverbials in native discourse is directly infl uenced by the register in which they appear, and the textual and/or interpersonal functions they serve.

5.3. Transfer-related effects on French learners’ use of academic vocabulary

The focus of Section 5.2 was on interlanguage features that are shared by most learner populations when compared to expert academic writing, and which are therefore likely to be developmental. Multiple factors, however, may combine to infl uence learners’ use of academic vocabulary. It has, for example, been suggested that learners’ preference for the sentence-initial

Table 5.24 Sentence-fi nal position of connectors in the ICLE and the BNC-AC-HUM

ICLE BNC-AC-HUM

S-F Tot. freq. % Rel. freq. S-F Tot. freq. % Rel. freq.

anyway 25 132 18.9 2.1 20 71 28.2 0.6for example 63 854 7.4 5.4 20 1263 1.6 0.6for instance 31 344 9.0 2.7 8 609 1.3 0.2indeed 15 257 5.8 1.3 18 1413 1.3 0.5of course 34 750 4.5 2.9 14 863 1.6 0.4then 35 1054 3.3 3.0 17 3062 0.5 0.5though 11 256 4.3 0.9 7 178 0.9 0.2


position for connectors may be attributed to the infl uence of instruction or ‘transfer of training’ (Selinker, 1972). The marked difference in frequency of I think across the learner sub-corpora may be partly explained by different academic writing conventions in the different mother tongues. As Granger (1998b: 158) put it, ‘learners clearly cannot be regarded as “phraseologically virgin territory”: they have a whole stock of prefabs in their mother tongue which will inevitably play a role – both positive and negative – in the acquisition of prefabs in the L2’.

Claims made about the nature of L1 infl uence and its interaction with other factors, however, have often been built on shaky methodological foundations and suffer from what Jarvis (2000: 246) referred to as a ‘you-know-it-when-you-see-it’ syndrome. To remedy this situation, Jarvis (2000) incorporated three types of L1 observable effects into a unifi ed framework for the study of L1 infl uence and proposed the following working defi nition of L1 infl uence, which is intended as a methodological heuristic to be used by transfer researchers:

L1 infl uence refers to any instance of learner data where a statistically signifi cant correlation (or probability-based relation) is shown to exist between some features of learners’ IL performance and their L1 back-ground. (Jarvis, 2000: 252)

Jarvis translated his working defi nition of L1 infl uence into a list of specifi c types of L1 observable effects that should be examined when investigating transfer. He argued that transfer studies should minimally consider at least three potential effects of L1 infl uence when presenting a case for or against L1 infl uence:

1. Intra-L1-group homogeneity in learners’ IL performance is found when learners who speak the same fi rst language behave as a group with respect to a specifi c L2 feature. To illustrate this fi rst L1 effect, Jarvis used Selinker’s (1992) fi nding according to which Hebrew-speaking learners of English as a group tend to produce sentences in which adverbs are placed before the object (e.g. I like very much movies).12 Intra-L1-group homogeneity is verifi ed by comparing the interlanguage of learners sharing the same mother tongue background.

2. Inter-L1-group heterogeneity in learners’ IL performance is found when ‘comparable learners of a common L2 who speak different L1s diverge in their IL performance’ (Jarvis, 2000: 254). To illustrate this effect, Jarvis referred to a number of studies reported by Ringbom (1987) that


have shown that Finnish-speaking learners are more likely than Swedish-speaking learners to omit English articles and prepositions. Jarvis argued that ‘this type of evidence strengthens the argument for L1 infl uence because it essentially rules out developmental and universal factors as the cause of the observed IL behaviour. In other words, it shows that the IL behaviour in question (omission of function words) is not something that every learner does (to the same degree or in the same way) regard-less of L1 background’ (Jarvis, 2000: 254–5). Inter-L1-group heterogene-ity is identifi ed by comparing the interlanguage of learners from different mother tongue backgrounds.

3. Intra-L1-group congruity between learners’ L1 and IL performance is found where ‘learners’ use of some L2 feature can be shown to parallel their use of a corresponding L1 feature’ (Jarvis, 2000: 255). Selinker (1992) uses this type of evidence to show that Hebrew-speaking learners’ positioning of English adverbs parallels their use of adverbs in the L1. The added value of this third L1 effect is that it also has explanatory power by showing ‘what it is in the L1 that motivates the IL behavior’ (Jarvis, 2000: 255). Intra-L1-group congruity is confi rmed by an IL/L1 comparison.

These three effects can emerge in circumstances in which transfer is not at play and can thus be misleading when considered in isolation. As shown in Table 5.25, Jarvis concluded that, despite differences in the degree of reliability, none of the three effects is suffi cient by itself to verify or charac-terize L1 infl uence. The identifi cation of two simultaneous L1 effects is necessary to present a convincing case for L1 infl uence. Identifying the three L1 effects would be even more convincing if it were not that ‘the ubiquity of conditions that can obscure L1 effects renders the three-effect requirement unrealistic in many cases’ ( Jarvis, 2000: 255).

Table 5.25 Jarvis’s (2000) three effects of potential L1 infl uence

L1 effect reliability suffi cient criterion

Intra-L1-group homogeneity in learners’ IL performance

poor no

Inter-L1-group heterogeneity in learners’ IL performance

strong no

Intra-L1-group congruity between learners’ L1 and IL performance

strongest no


I made use of Jarvis’s (2000) unifi ed framework to investigate the poten-tial infl uence of the fi rst language on multiword sequences that serve rhe-torical functions in French learners’ argumentative writing. The International Corpus of Learner English appears to be ideally suited to analysing the three potential effects of L1 infl uence described by Jarvis (2000). Table 5.26 lists the three steps needed to investigate the infl uence of French on recurrent word sequences in the ICLE-FR. Intra-L1-group homogeneity in learners’ performance is investigated by comparing all the essays written by French learners to verify whether they behave as a group with respect to a specifi c L2 feature. Inter-L1-group heterogeneity in learners’ IL performance is verifi ed by a comparison of the number of texts in which a specifi c lexical item is used in the ICLE-FR and in other L1 sub-corpora. Unlike Jarvis (2000), I made use of comparison of means tests and post hoc tests such as Ryan’s procedure and Dunnett’s test to confi rm this second L1 effect.13 To establish intra-L1-group congruity between learners’ L1 and IL perfor-mance, French EFL learners’ use of a specifi c lexical item is compared to the use of its equivalent form in a 225,174-word comparable corpus of essays written by French-speaking students collected at the University of Louvain, i.e. the Corpus de Dissertations Françaises (CODIF).

Applying Jarvis’s (2000) framework to the ICLE texts reveals the potential infl uence of transfer on French learners’ use of multiword sequences that serve specifi c rhetorical functions in English. For example, the three trans-fer effects are found in French learners’ use of on the contrary, indicating that L1 infl uence most probably reinforces the conceptual problems and misguided teaching practices that were identifi ed in Section 5.2.4 as poten-tial explanations for the frequent misuse of the adverbial. This strongly supports Granger and Tyson’s (1996) suggestion that French learners’ overuse and misuse of the connector is probably due to an over-extension

Table 5.26 Jarvis’s (2000) unifi ed framework applied to the ICLE-FR

L1 effect Corpus comparisons

Intra-L1-group homogeneity in learners’ performance

A comparison of the use of a specifi c lexical item in all the essays written by French learners

Inter-L1-group heterogeneity in learners’ IL performance

A comparison of the use of a specifi c lexical item in the ICLE-FR against other L1 sub-corpora

Intra-L1-group congruity between learners’ L1 and IL performance

A comparison of a specifi c lexical item in the ICLE-FR to the use of its equivalent form in a comparable corpus of French native student writing


of the semantic properties of the French au contraire, which can be used to express both a concessive and an antithetic link.

Most transfer studies have focused on what we can call ‘transfer of form’ (e.g. borrowing), ‘transfer of meaning’ (cf. semantic transfer, seman-tic extension) or ‘transfer of form/meaning mapping’ (e.g. cognates) (see Jarvis and Pavlenko, 2008; Odlin, 1989; 2003 and Ringbom, 2007 for excellent syntheses on lexical and semantic transfer). Next to knowledge of form and meaning, however, knowing a word also involves knowing in what patterns, with what words, when, where and how to use it (Nation, 2001: 27). These other types of knowledge can also give rise to transfer. For example, research into learners’ use of cognates has highlighted transfer effects on style and register (cf. Granger and Swallow 1988; Van Roey 1990; Granger 1996b). Studies focusing on learners’ use of phrasemes have brought to light transfer effects on collocational restrictions and lexico-grammatical patterns (e.g. Biskup, 1992; Granger, 1998b; Nesselhauf, 2003). However, much remains to be done regarding ‘transfer of use’.

Applying Jarvis’s (2000) unifi ed framework on learner corpus data brings to light interesting fi ndings relating to L1 infl uence on word use. It helps to identify a number of transfer effects that remain largely undocu-mented in the SLA literature: transfer of function, transfer of the phraseo-logical environment, transfer of style and register, and transfer of L1 frequency. These four transfer effects often accompany transfer of form and meaning and may also reinforce each other. They are illustrated in the remaining of this section.

Multiword sequences with a pragmatic anchor seem to be quite easily transferred. French learners’ use of the idiosyncratic expression *according to me is a good example of transfer of function. This sequence is repeatedly used in the ICLE-FR; it does not appear in other learner sub-corpora except for the ICLE-DU and the ICLE-SW, where it is extremely rare. Moreover, there is congruity between French learners’ use of according to me in English and selon moi in French, which are probably regarded as translation equivalents by French EFL learners. The English preposition according to and the French selon both mean ‘as shown by something or stated by someone’ (e.g. According to George Heard Hamilton, Rodin became “a fi gure of international signifi cance, the most admired, prolifi c, and infl uential sculptor since Bernini”, BNC-AC-HUM). However, they differ in one signifi cant way: accord-ing to me is usually not accepted as a correct English phraseme. By contrast, selon moi is perfectly fi ne in French and is, in fact, quite frequent in French native-speaker students’ writing. This may explain why French EFL learners are keen to use what they regard as a direct translation of a common French expression.


The following examples illustrate French students’ use of selon moi and French EFL learners’ use of according to me:

5.109. Selon moi, la chanson est un vecteur de culture parce qu’elle est un art qui impose l’engagement des différents acteurs. (CODIF)

5.110. Selon moi, tout le monde pense ce qu’il veut et comme il veut, agit comme il l’entend en respectant la loi et les codes établis. (CODIF)

5.111. According to me, the real problem now is not that man refuses to pay heed but that man refuses to make some sacrifi ces for the sake of ecology and to understand that the values that we have chosen are the wrong ones. (ICLE-FR)

5.112. According to me, the prison system is not outdated: it has never been a solution per se. (ICLE-FR)

Figure 5.10 represents graphically how the misleading translation equiva-lent may be created by French EFL learners.

Transfer effects are also detectable in French learners’ use of lexico-grammatical and phraseological patterns. The English verb illustrate is a case in point. Although it is not found in many texts written by French learners, it is much more frequent in ICLE-FR overall than in any other learner sub-corpus. Table 5.27 shows that French EFL learners frequently use the verb illustrate in its infi nitive form. The percentage of use of this form (40%) is quite similar to that of the infi nitive form of the French cog-nate verb illustrer in CODIF, but differs signifi cantly from the proportion of infi nitive forms of the English verb illustrate that were found in the BNC-AC-HUM (23.6%) (cf. Table 4.6 in Section 4.2.2). A closer look at the occur-rences of the infi nitive form of illustrate in ICLE-FR reveals that it is repeatedly used in sentence-initial to-infi nitive structures (Examples 5.113 and 5.114), a pattern that is also the preferred lexico-grammatical environ-ment of illustrer in the corpus of French essays (Example 5.115).

5.113. To illustrate this, we can mention the notion of culture and language in the north of Belgium. (ICLE-FR)

5.114. To illustrate this point, it would be interesting to compare our situation with the U.S.A.’s. (ICLE-FR)

5.115. Pour illustrer cela, prenons l’exemple des pâtes alimentaires italiennes. (CODIF)

French learners’ knowledge of the verb illustrer in their mother tongue probably infl uences the type of word combinations and lexico-grammatical patterns in which they use the English verb illustrate.


FRENCH

FRENCH LEARNERS' INTERLANGUAGE

'according to'

'according to' + [+HUM]

'according to' + [+HUM] 'according to' + [-HUM]

'according to X' *'accordingto me'

'according to'

ENGLISH

'according to'

+ [+HUM]

'selon'

'selon' + [+HUM]

'selon X'

e.g. lui, Hugo,monsieur Bernanos,certains, etc.

e.g. Civil Liberty Members,supporters, Judge Kamins, XavierFlores, etc.

e.g. idea, article, theory, argument,situation, etc.

e.g. idée, loi, principe, philosophie,argument, théorie, norme, etc.

'selon' + [-HUM]

'selon moi'

Figure 5.10 A possible rationale for the use of ‘according to me’ in French learners’ interlanguage

Similarly, French EFL learners almost always use the verb conclude in the sentence-initial discourse marker To conclude followed by an active structure introduced by a fi rst person pronoun + modal verb. This pattern is less fre-quent in the writing of EFL learners with other mother tongue backgrounds and parallels a very frequent way of concluding in French, viz. sentence-initial Pour conclure. The following examples show that longer sequences


and phraseological cascades (see Section 5.2.3) may also be transfer-related.

5.116. Pour conclure, nous pouvons dire que les deux stades sont aussi importants l’ un que l’ autre : il est nécessaire que l’ homme soit membre d’ un groupe mais il est tout aussi primordial qu’ il s’ en détache pour construire son iden-tité propre. (CODIF)

5.117. To conclude, we can say that many people are today addicted to television. (ICLE-FR)

5.118. Pour conclure, je dirais que chaque individu est unique, différent et qu’il est facile de vouloir ressembler aux autres plutôt que de s’accepter tel qu’on est. (CODIF)

5.119. To conclude, I would say that science, technology and industrialisation cer-tainly stand in the way of human relationships but not in people’s dreams and imagination. (ICLE-FR)

My fi ndings also point to a transfer of style and register. In Section 5.2.3, the fi rst person plural imperative form let us was shown to be overused by all L1 learner populations when compared to expert academic writing. As shown in Table 5.28, the two-word sequence occurs in 25.9 per cent of the texts produced by French learners and is much more frequent in the ICLE-FR than in any other learner sub-corpus. This difference in use between ICLE-FR and the other ICLE sub-corpora proved to be statistically signifi cant.

An analysis of concordance lines for let us shows that this sequence is repeatedly used by French speaking students to serve a number of rhetori-cal and organisational functions. For example, it is used as a code gloss to

Table 5.27 A comparison of the use of the English verb ‘illustrate’ and the French verb ‘illustrer’

En. ‘illustrate’

in ICLE-FR

Fr. ‘illustrer’ in

CODIF

simple present 10 50% 8 31%infi nitive 8 40% 13 50%past participle 2 10% 3 12%imperative 0 0% 1 4%past 0 0% 1 4%

TOTAL 20 100% 26 100%

Rel. freq. per 100,000 words 14.67 11.55


introduce an example (Example 5.120), a transition marker to change topic (Examples 5.121 and 5.122), and an attitude marker (Example 5.123).

5.120. To illustrate the truth of this, let us take the example of Britain which was already fi ghting its corner alone after Mrs Thatcher found herself totally isolated over the decision that Europe would have a single currency. (ICLE-FR)

5.121. Let us then focus on the new Europe as a giant whose parts are striving for unity. (ICLE-FR)

5.122. Let us now turn our attention to the students who want to apply for a job in the private sector. (ICLE-FR)

5.123. Let us be clear that we cannot let countries tear one another to pieces and if we closed our eyes to such an atrocity, our behaviour would be cowardly. (ICLE-FR)

As explained in Section 4.2.3, the fi rst person plural imperative form let us is found in professional academic writing, but it is not frequent (relative frequency of 5.46 occurrences per 100,000 words). It is also restricted to a limited set of verbs (see Swales et al., 1998; Hyland, 2002). In the BNC-AC-HUM, there are only eight signifi cant verb co-occurrents of let us: consider, say, suppose, return, begin, look, take and have.

There is no lexically equivalent form to En. let us in French. Equivalence is however found at the morphological level as French makes use of an infl ec-tional suffi x to mark the fi rst imperative plural form. Thus, to investigate the third L1 effect, i.e. intra-L1 group congruity between learners’ L1 and IL

Table 5.28 ‘let us’ in learner texts

Rel. freq. of

‘let us’ and

‘let’s’ per

100,000 words

Number of

texts including

‘let us’ or ‘let’s’

Number

of texts

%

French 71.88 59 228 25.9%Czech 25.24 19 147 12.9Dutch 12.33 19 196 9.7Finnish 8.78 10 167 6German 13.69 14 179 7.8Italian 20.95 10 79 12.7Polish 19.21 23 221 10.4Russian 38.57 47 194 24.2Spanish 26.23 14 149 9.4Swedish 18.73 9 81 11.11

TOTAL 26.85 224 1641 13.65


performance, I compared the use of let us in ICLE-FR with that of fi rst person plural imperative verbs in CODIF. The rhetorical and organisational func-tions fulfi lled by let us in French EFL learner writing can be paralleled with the very frequent use of fi rst person plural imperative verbs in French student writing to organize discourse and interact with the reader (Paquot, 2008a):

5.124. Prenons l’exemple des sorciers ou des magiciens au Moyen Age. (CODIF)5.125. Ajoutons qu’une partie plus spécifi que de la population est touchée.

(CODIF)5.126. Comparons cela à la visite de la cathédrale d’Amiens. (CODIF)5.127. Envisageons tout d’abord la question économique. (CODIF)5.128. Examinons successivement le problème de l’abolition des frontières d’un point

de vue économique, juridique et enfi n culturel. (CODIF)5.129. Imaginons un monde ou règne une pensée unique. (CODIF)5.130. Considérons un instant le cinéma actuel. (CODIF)

First person plural imperative verbs serve specifi c discourse strategies in French formal types of writing, and more specifi cally in academic writing. French EFL learners seem to transfer their knowledge of French academic writing conventions (Connor, 1996) and make use of imperatives in English academic writing in the same way as in French academic writing. Impera-tive forms that are repeated in the ICLE-FR often have formal equivalents that are found in CODIF (e.g. let us take the example of ‘prenons l’exemple de’; let us consider ‘considérons’; let us hope ‘espérons’; let us examine ‘exami-nons’; let us take ‘prenons’; let us (not/never) forget ‘oublions/n’oublions pas que’; let us think ‘pensons’). This generalized overuse of the fi rst person plural imperative in EFL French learner writing as a rhetorical strategy does not conform to English academic writing conventions but rather to French academic style. In English, let us (and more precisely its contracted form let’s) is much more typical of speech (relative frequency of 42.5 occur-rences per 100,000 words in the BNC-SP but only 5.3 per 100,000 in the BNC-AC). As a result, the speech-like nature of let us in French EFL learner writing leads to an overall impression of stylistic inappropriateness.

This example points to yet another type of transfer effect, namely transfer of L1 frequency. As shown in Table 5.29, the frequency of let us in the ICLE-FR is much closer to the frequency of fi rst person plural imperative verbs in student writing in French, than in English expert or novice writing. Other examples of sequences that have French-like frequencies in the ICLE-FR include on the contrary, on the other hand, let us take the example, to illustrate this, to conclude and *according to me.


Table 5.29 The transfer of frequency of the fi rst person plural imperative between French and English writing

Corpus Relative frequency per

100,000 words of fi rst person

plural imperative verbs

French L1 students (CODIF) 95.5French EFL learners (ICLE-FR) 71.9English expert writers (BNC-AC-HUM) 5.7English novice writers (LOCNESS) 3

FRENCH

Fr. 1st plural imperative

Fr. ‘prenons’ example de’Fr. ‘n’ oublions pas’

Fr. ‘examinons’

FREQUENCYFR

REGISTERFR

FUNCTIONFR

PHRASEOLOGYFR

FREQUENCYFR

REGISTERFR

FUNCTIONFR

PHRASEOLOGYFR

FREQUENCYEN

REGISTEREN

FUNCTIONEN

PHRASEOLOGYEN

En. ‘let us take the example of’En ‘let us not forget’En. ‘let us examine’

En. ‘let us take the example of’En ‘let us not forget’En. ‘let us examine’

En. 1st plural imperative

En. 1st plural imperative

FRENCH EFL LEARNERS'INTERLANGUAGE

ENGLISH

. . .

Figure 5.11 A possible rationale for the use of ‘let us’ in French learners’ interlanguage


Transfer effects often interact in learners’ use of English lexical devices. Thus, French EFL learners use English fi rst person plural imperatives in academic writing with the frequency of French imperative verbs in the corre-sponding register, in French-like phraseological patterns and to serve the same organizational and interactional functions. As illustrated in Figure 5.11, French EFL learners’ use of textual phrasemes such as let’s take the example of, let’s examine or let us not forget mirror the stylistic profi le of the French sequences prenons l’exemple de .., examinons et n’oublions pas in French academic writing.

The transfer effects identifi ed in this section – transfer of function, trans-fer of lexico-grammatical and phraseological patterns, transfer of style and register, and transfer of frequency – make up what, following Hoey (2005: 183), I refer to as ‘transfer of primings’. EFL learners’ knowledge of words and word combinations in their mother tongue includes a whole range of information about their preferred co-occurrences and sentence position, stylistic or register features, discourse functions and frequency. Primings for collocational and contextual use of (at least a restricted set of frequent or core) L1 lexical devices are particularly strong in the mental lexicon of adult EFL learners. They are the result of many encounters with these lexi-cal items in L1 speech and writing. Mental primings in the L1 lexicon prob-ably infl uence EFL learners’ knowledge of English words and word sequences by priming the lexico-grammatical preferences of an L1 lexical item to its English counterpart.


The data presented in this chapter support the idea that the ‘English of advanced learners from different countries with a relatively limited varia-tion of cultural and educational background factors share a number of fea-tures which make it differ from NS language’ (Ringbom, 1998: 49). The focus of the analysis has been on the lexical means available to learners to perform specifi c rhetorical and organizational functions in academic writ-ing, and more precisely in argumentative essays. This textual dimension is particularly diffi cult to master and has been described by Perdue (1993) as the last developmental stage before bilingualism in second language acqui-sition. My results show that the expression of rhetorical and organizational functions in EFL writing is characterized by:

A limited lexical repertoire: EFL learners tend to massively overuse a restricted set of words and phrasemes to serve a particular rhetorical


function and to underuse a large proportion of the lexical means avail-able to expert writers. They also seem to prefer to use conjunctions, adverbs and prepositions rather than phraseological patterns with nouns, verbs and adjectives.

A lack of register awareness: texts produced by EFL learners often ‘give confusing signals of register’ (Field and Yip, 1992: 26) as they display mixed patterns of formality and informality. The frequency of informal words and phrases in learner writing is often closer to their frequency in native-speakers’ speech than in their academic prose.

Lexico-grammatical and phraseological specifi cities: EFL learners’ writing is distinguishable by a whole range of lexico-grammatical patterns and co-occurrences that differ from academic prose in both quantitative and qualitative terms. Preferred co-occurrences in the ICLE are often not the same as in academic prose, which reveals learners’ weak sense of native speakers’ ‘preferred ways of saying things’. Learners’ attempts at using collocations are not always successful and sometimes result in crude approximations and lexico-grammatical infelicities. My results also support Lorenz’s (1999b) remark that ‘advanced learn-ers’ defi cits are most resilient in the area of lexico-grammar, where lexi-cal items are employed to signal grammatical and textual relations’ and that ‘a lack of coherence in advanced learners’ writing must at least partly be attributable to lexico-grammatical defi cits’ (Lorenz, 1999b: 56).

Semantic misuse: As Crewe (1990: 317) commented, ‘the misuse of logical connectives is an almost universal feature of ESL students’ writing’. What is less well-documented in the literature, however, is that EFL learners also experience diffi culty with the semantics of other types of cohesive devices, and specifi cally, with labels, i.e. abstract nouns that are inherently unspecifi c and require lexical realization in their co-text, either beforehand or afterwards.

Chains of connective devices: EFL learners’ texts are sometimes charac-terized by the use of superfl uous (and sometimes semantically inconsis-tent) connective devices.

A marked preference for sentence-initial position of connectors: con-nectors are often used in the unmarked sentence-initial position in learner writing. A medial position is not favoured by EFL learners, although it is typical of academic prose.

The methodology used in the fi rst part of this chapter has made it possi-ble to draw a general picture of the writing of upper-intermediate to advanced EFL learners from different mother tongue backgrounds. Most


of these features have already been mentioned in the literature, but they have always been reported on the basis of only one or two L1 learner popu-lations. My methodology makes it possible to avoid hasty interpretations in terms of L1 infl uence. Consider the following quotations by Zhang (2000), who attributes a number of features to the infl uence of the learners’ mother tongue, in this case Chinese:

The overuse of this expression [more and more] was most probably due to language transfer since a familiar expression in the Chinese language ye lai yue was popularly used. (Zhang, 2000: 77);

The reason for the initial positioning of conjunctions was again due to the transfer of the Chinese language where conjunction devices with sim-ilar meaning are mostly used at the beginning of a sentence. (Zhang, 2000: 83).

As explained above, sentence-initial positioning of conjunctions is common to most learner populations. The mother tongue may reinforce learners’ preference for sentence-initial position but cannot be regarded as a complete explanation for this learner-specifi c feature. In Section 5.2.6, teaching-induced factors have been identifi ed as a possible cause for learn-ers’ preference for sentence-initial position. Syntactic positioning of connectors is rarely taught and EFL learners often consider the sentence-initial position to be a safe strategy. As for the overuse of the expression more and more, although it is indeed very signifi cant in the Chinese component of the second edition of the ICLE (Granger et al., 2009), this feature is actu-ally common to all learner populations represented in the corpus. This suggests that, while transfer may be at work in Chinese learners’ use of more and more, it is probably not the only explanation. It is not always possible to attribute learner-specifi c features to a single factor, as developmental, teaching-induced and transfer-related effects can reinforce each other (Granger, 2004: 135–6).

Another advantage of the method I used is that, once linguistic features of upper-intermediate to advanced EFL learner writing have been identi-fi ed, we can check to what extent they are specifi c to EFL learners or just typical of novice writing. This is precisely where a corpus of essays written by English native university students such as LOCNESS (see Section 2.1) has a role to play. Tripartite comparisons between professional writing, foreign learner writing and native student writing make it possible to distinguish between learner-specifi c and developmental features (e.g. Neff et al., 2008).


Whether a feature is learner-specifi c or developmental varies from lexical item to lexical item, but as a general rule, the fi ndings suggest that the main feature shared by native and non-native novice writers is a lack of register-awareness. Figure 5.12 shows that a whole range of lexical items that Gilquin and Paquot (2008) found to be overused in EFL learners’ writing – maybe, so expressing effect, it seems to me, really, sentence-fi nal though, this/that is why, I think and fi rst of all –are also more frequently used by native-speaker student writers than in expert academic prose.14 The overuse of I think in both EFL learner and native-speaker student writing has already been reported by Neff et al. (2004a) who described it as a general ‘novice-writer characteristic of excessive visibility’ (Neff et al, 2004a: 152).

Figure 5.12 also shows that not all learner-specifi c speech-like lexical items are overused in the writing of native-speaking students. Thus, the lexical items of course, certainly, absolutely, by the way and I would like/want/am going to talk about are quite rare in LOCNESS. They are even less frequent in native-speaker students’ writing than in academic prose, which suggests that native novice writers do not transfer all spoken features to their

200

200

150

100

50

0

3000

2500

2000

1500

1000

500

0

150

100

50

0Freq. of PRO (this, that, which) is why

(pmw)

Freq. of first of all (pmw)

Freq. of I think (pmw)

Expert academic writing: BritishNational Corpus, academiccomponent (15m words)Native-speaker student writing:Sub-corpus of LOCNESS (100,702words)EFL learners' writing: ICLEv2(14L1s; around 1.5m. words)Native speech: British Nationalcorpus, spoken component(10m words)

Figure 5.12 Features of novice writing – Frequency in expert academic writing, native-speaker and EFL novices’ writing and native speech (per million words of running text)


350 1200

1000

800

600

400

200

0

300250

200150100

50

40

2000

1500

1000

500

really of course certainly

Freq. of amplifying adverbs (pmw)

absolutely definitely0

181614121086420

35302520151050

Freq. of maybe (pmw)

Freq. of it seems to me (pmw)

Freq. of by the way (pmw) Freq. of sentence-final though (pmw)

Freq. of I would like / want / am going to talkabout (pmw)

Freq. of so expressing effect (pmw)0

45 120

100

80

60

40

20

0

40353025201510

50

Figure 5.12 Continued


academic writing. It seems that lexical items which are not particularly frequent in speech and are rare in academic prose (e.g. I want/would like/am going to talk about) are less likely to be overused by native novice writers. By contrast, lexical items that are very frequent in speech, and acceptable in academic prose, are very likely to be overused (e.g. maybe, so expressing effect).

Other linguistic features are limited to EFL learners. These include lexico-grammatical errors (*a same, possibility *to, despite *of, discuss *about), the use of non-native-like sequences (e.g. according to me and as a conclusion), and the overuse of relatively rare expressions such as in a nutshell. As Gilquin, Granger and Paquot (2007a: 323) have argued, the issue of the degree of overlap between novice native writers and non-native writers has far-reaching methodological and pedagogical implications and is clearly in need of further empirical study.

Developmental factors in L1 and L2 acquisition cannot, however, be held responsible for all learner specifi c-features. In addition to teaching-induced factors and profi ciency, the fi rst language also plays a part in EFL learners’ use of academic vocabulary. In the last part of this chapter, I focused on the potential infl uence of the fi rst language on multiword sequences that serve rhetorical functions in French learner writing. I made use of Jarvis’s (2000) framework for assessing transfer and identifi ed a number of transfer effects – transfer of function, of the phraseological environment, of style and register, and of L1 frequency – that I referred to as ‘transfer of primings’. These results support Kellerman’s claim that the ‘hoary old chestnut’ according to which transfer does not affl ict the more advanced learner ‘should fi nally be squashed underfoot as an unwarranted overgeneralization based on very limited evidence’ (Kellerman, 1984: 121). However, they also suggest that the main effect of the students’ mother tongue on higher-intermediate to advanced learner writing is not errors, but more subtle transfer effects, especially at higher levels of profi ciency.

Part III

Pedagogical implications and conclusions

In the fi rst two sections I defi ned the concept of ‘academic vocabulary’, built a list of academic keywords from corpora of expert writing, and analysed their use in ten sub-corpora of the International Corpus of Learner English. In Chapter 6, I discuss some of the important pedagogical implications of this research. There are three key aspects: the infl uence of teaching on learners’ writing; the role of the fi rst language in EFL learning and teaching; and the use of corpora, and more specifi cally, learner cor-pora, in the development of EAP teaching materials. Chapter 7 then briefl y summarizes the major results, discusses some of their implications, and sug-gests several remaining issues and avenues for future research.

Chapter 6

Pedagogical implications

This chapter considers three areas where my fi ndings have major pedagogi-cal implications: teaching-induced factors; the role of the fi rst language in EFL learning and teaching; and the role of corpora in EAP material design. The ways in which corpus data, in particular data from learner corpora, have been used to inform the academic-writing sections of the second edition of the Macmillan English Dictionary for Advanced Learners (MED2) (Rundell, 2007) are also discussed.

6.1. Teaching-induced factors

Factors linked to teaching have repeatedly been denounced in the litera-ture as being responsible for a number of learners’ inappropriate uses of connectors (see Zamel, 1983; Hyland and Milton, 1997; Flowerdew, 1998; Milton, 1999). Connectors are often presented in long lists of undifferenti-ated and supposedly equivalent items, classifi ed in broad functional catego-ries. This can cause semantic misuse (Crewe, 1990; Lake, 2004). For example, Jordan (1999) describes the adverbial on the contrary as a phrase of contrast equivalent to on the other hand and by contrast (see Figure 6.1). The same is said about conversely. However this adverb should only be used to indicate that one situation is the exact opposite of another:

6.1. American consumers prefer white eggs; conversely, British buyers like brown eggs. (LDOCE4)

Also problematic are the categorization of besides as a marker of concession, and the misleading presentation of the conjunctions even if and even though as synonyms.

Overuse of connectors such as nevertheless, in a nutshell, as far as I am con-cerned, on the one hand, and on the other hand can also be attributed to the long


lists of connectors found in most textbooks (Granger, 2004: 135) as no information is given about their frequency or semantic properties. Milton (1999) has shown that there is a strong correlation between the words and phrases overused by Hong Kong students and the functional lists of expres-sions distributed by tutorial schools (private institutions which prepare most high school students in Hong Kong for English examinations). The selection of connectors to be taught may also lend itself to criticism. It was shown in Section 5.2.3 that sequences that are rarely used by native speak-ers (e.g. as far as I am concerned or last but not least) or ‘unidiomatic’ sequences (e.g. as a conclusion) are sometimes found in teaching materials, especially in the lists of connectors freely available on the Internet.1 By contrast, the connectors most frequently used to serve rhetorical functions are sometimes missing from these lists.

Another direct consequence of these lists is EFL learners’ stylistic inappropriateness, as Milton explains:

Students are drilled in the categorical use of a short list of expressions – often those functioning as connectives or alternatively those which are

A. Contrast, with what has preceded:

insteadconverselythenon the contraryby (way of ) contrast in comparison(on the one hand) . . . on the other hand . . .

B. Concession indicates the unexpected, surprising nature of what is being said in view ofwhat was said before:

besides yet(or) else in any casehowever at any ratenevertheless for all that nonetheless in spite of/despite thatnotwithstanding after allonly at the same timestill on the other handwhile all the same(al)though even if/though

Figure 6.1 Connectives: contrast and concession (Jordan, 1999: 136)

Pedagogical implications 203

colourful and complicated (and therefore impressive) – regardless of whether they are used primarily in spoken or written language (if indeed at all), or to which text types they are appropriate (1998: 190).

Thus, the spoken-like expression all the same is given as an equivalent alter-native to more formal connectors such as on the other hand or notwithstanding in Jordan (1999) (see Figure 6.1). This example also illustrates the fact that no information about the connectors’ grammatical category or syntactic properties is made available to the learners. The preposition notwithstand-ing is listed together with adverbs and adverbial phrases (e.g. however, yet) as well as conjunctions (e.g. although, while). Learners’ marked preference for the sentence-initial positioning of connectors has also been related to L2 instruction (see Flowerdew, 1998; Milton, 1999; Narita and Sugiura, 2006). Positional variation of connectors is usually not taught, and learners use the sentence-initial position as a safe bet.2

Another problem of teaching practices (which has not often been documented) is that too much emphasis tends to be placed on connectors, that is, on grammatical cohesion (see Halliday and Hasan, 1976), to the detriment of lexical cohesion.3 However I have shown in this book that nouns, verbs and adjectives all have prominent rhetorical functions in academic prose. Labels, have also been found to fulfi l a prominent cohesive role in this particular genre. It is most probable that lexical cohesion has been neglected in EFL teaching because ‘there have been no good descriptions of the forms and functions of this phenomenon’ (Flowerdew, 2006: 345).

6.2. The role of the fi rst language in EFL learning and teaching

My fi ndings have at least two important pedagogical implications relating to the role of the fi rst language in EFL learning and teaching. Transfer of primings means that words or word sequences in the foreign language may be primed for L1 use in terms of discourse function, colloca-tional and lexico-grammatical preferences, register and frequency. One of the many roles of teaching should thus be to counter these ‘default’ and sometimes misleading primings in EFL learners’ mental lexicons. Aware-ness-raising activities focusing on similarities and differences between the mother tongue and the foreign language are clearly needed to achieve this. These activities should not be restricted to ‘helping learners focus on errors typically committed by learners from a particular L1’ (Hegelheimer and


Fisher, 2006: 259). They should also raise learners’ awareness of more subtle differences such as the register differences and collocational prefer-ences of similar words in the two languages. This recommendation stands in sharp contrast to Bahns’s (1993: 56) claim that collocations that are direct translation equivalents do not need to be taught. Learners have no way of knowing which collocations are congruent in the mother tongue and the foreign language; moreover, the differences between the colloca-tions in L1 and L2 may lie in aspects of use rather than form or meaning.

However, as Odlin commented, it is not always possible to make use of the fi rst language in the classroom and to rely on contrastive data:

Whatever the merits of contrastive materials in some contexts, it is clear that such materials are not always feasible. For example, when an ESL class consists of speakers of Chinese, Persian, Spanish, Tamil, and Yoruba, there is not likely to be any textbook that contrasts English verb phrases with verb phrases in all of those languages – and even if there were, teachers could not profi tably spend the class time necessary to illu-minate so many contrasts. Yet even in such classes, one type of contrastive information is frequently available: bilingual dictionaries. Although the comparisons are sometimes restricted to words in the native and target languages, the most carefully prepared dictionaries often provide some comparisons of pronunciation and grammar as well. If the class size allows it, teachers can help individual students in using any contrastive informa-tion that their dictionaries provide. (1989: 162)

Bilingual dictionaries should ideally facilitate the teacher’s task in multilin-gual as well as monolingual classrooms. However, it is questionable whether the type of contrastive information they provide is fully adequate. For exam-ple, the Robert & Collins CD-Rom (Version 1.1) includes an essay-writing sec-tion in which fi rst person plural imperatives in French are systematically translated by structures employing let us in English (Granger and Paquot, 2008b). In Section 5.3, however, I showed that fi rst person plural impera-tives are not the best way of organizing discourse and interacting with the reader in English academic writing. Table 6.1 lists examples of infelicitous translation equivalents. Similarly, a web-page devoted to linking words and hosted by the ‘Académie de Lille (Anglais BTS Informatique)’ lists accord-ing to me as a direct translation equivalent of the French ‘à mon avis’, and as a conclusion as a possible equivalent of the French ‘pour conclure / pour résumer ’.4


Table 6.1 Le Robert & Collins CD-Rom (2003–2004): Essay writing

Essay writing: function French sentence Proposed English equivalence

Developing the argument

Prenons comme point de départ le rôle que le gouvernement a joué dans l’élaboration de ces programmes

= ‘let us take ... as a starting point’

En premier lieu, examinons ce qui fait obstacle à la paix

= ‘fi rstly, let us examine’

The other side of the argument

Après avoir étudié la progres-sion de l’action, considérons maintenant le style

= ‘after studying ... let us now consider’

Venons-en maintenant à l’analyse des retombées politiques

= ‘now let us come to’

Assessing an idea Examinons les origines du problème ainsi que certaines des solutions suggérées

= ‘let us examine ... as well as’

Sans nous appesantir or nous attarder sur les détails, notons toutefois que le rôle du Conseil de l’ordre a été déterminant

= ‘without dwelling on the details, let us note, however, that’

Nous reviendrons plus loin sur cette question, mais signalons déjà l’absence totale d’émotion dans ce passage

= ‘we shall come back to this question later, but let us point out at this stage’

Avant d’aborder la question du style, mentionnons brièvement le choix des métaphores

= ‘before tackling ... let us mention briefl y’

Adding or detailing Ajoutons à cela or Il faut ajouter à cela or À cela s’ajoute un sens remarqua-ble du détail

= ‘let us add to this or added to this’

Introducing an example Prenons le cas de Louis dans «le Nœud de vipères»

= ‘(let us) take the case of’

Stating facts Rappelons les faits. Victoria l’Américaine débarque à Londres en 1970 et réussit rapidement à s’imposer sur la scène musicale

= ‘let’s recall the facts’

Emphasizing particular points

N’oublions pas que, sur Terre, la gravité pilote absolument tous les phénomènes

= ‘let us not forget that’


These fi ndings are quite representative of a general lack of good contras-tive studies on which pedagogical materials can be based. Multilingual corpora clearly have an important role to play here by providing an empir-ically-based source of translation equivalents (Bowker, 2003; King, 2003).

6.3. The role of learner corpora in EAP materials design

While teaching materials designed to help undergraduate students improve their academic writing skills are legion (e.g. Bailey 2006; Hamp-Lyons and Heasley 2006), few make use of authentic texts and very few are informed by the use of corpora. Even when they are corpus-informed, EAP resources tend to be based on data from native-speakers only. Thus, Thurstun and Candlin’s (1997) Exploring Academic English, which uses concordance lines to introduce new words in context and familiarize learners with phraseological patterns, relies exclusively on data from a native-speakers’ academic corpus. Although this is one of the most innova-tive EAP textbooks to date, it is arguably less useful for non-native learners, despite Thurstun and Candlin’s (1998) claim that it is equally appropriate for native and non- native writers. As shown in Section 5.2, EFL writing is characterized by a number of linguistic features that differ from novice native-speakers’ writing.

The value of pedagogical tools for non-native speakers of English would be greatly increased if fi ndings from learner corpus data were also used to select what to teach and how to teach it. As Flowerdew (1998) put it, ‘when choosing which markers to teach, decisions made should also be based on fi ndings from a parallel student corpus to ascertain where students’ main defi ciencies lie. If not, there is a danger that the emphasis on teaching the most frequent markers may focus on ones already familiar to and correctly used by students, or in this case, exacerbate the problem with their overuse’ (Flowerdew, 1998: 338). By showing, in context, the types of infelicities EFL learners produce and the types of errors they make, as well as the items they tend to under- or overuse, learner corpora are the most valuable resources for designing EAP materials which address the specifi c problems that EFL learners encounter (see also Flowerdew, 2001; Granger, 2009). Yet, such corpora have very rarely been used systematically to inform EAP materials (see Milton, 1998 and Tseng and Liou, 2006 for two exceptions in Computer-Assisted Language Learning).5

The only type of resource in which learner corpus data have been relatively successfully implemented up to now is the monolingual learners’ dictionary (MLD). For example, the Longman Dictionary of Contemporary


English and the Cambridge Advanced Learner’s Dictionary include a number of learner corpus-informed usage notes which warn against common learner errors (e.g. the confusion between the adjectives actual and current, the countable use of the noun information). Yet, if MLDs are to take further ‘proactive steps to help learners negotiate known areas of diffi culty’ (Rundell, 1999: 47), learner corpora should not only be exploited to compile error notes but also to improve other aspects of the dictionary.

As put by Cook (1998: 57) referring to Carter’s (1998b) standpoint, however, ‘materials should be infl uenced by, but not slaves to, corpus fi ndings’ (see also Swales, 2002; Widdowson, 2003). The method used in Chapter 5 has made it possible to identify a number of common features of EFL learners’ expression of rhetorical and organizational functions. A selected list of features were used to inform a 30-page writing section which I and two other members of the Centre for English Corpus Linguistics (CECL), Gaëtanelle Gilquin and Sylviane Granger, designed for the second edition of the Macmillan English Dictionary for Advanced Learners (Gilquin et al. 2007b: IW1–IW29). The writing section includes 12 functions that EFL learners need to master in order to write well-structured academic texts. These were identifi ed in Section 4.1 as typically appearing in EAP textbooks which adopt a functional approach to academic writing: (1) adding infor-mation; (2) comparing and contrasting: describing similarities and differ-ences; (3) exemplifi cation: introducing examples; (4) expressing cause and effect; (5) expressing personal opinions; (6) expressing possibility and certainty; (7) introducing a concession; (8) introducing topics and related ideas; (9) listing items; (10) reformulation: paraphrasing or clarifying; (11) reporting and quoting; (12) summarizing and drawing conclusions.

Each writing section includes a detailed ‘corpus-based rather than corpus-bound’ description (Summers, 1996: 262) of the many lexical means that are available to expert writers to perform a specifi c function. Special emphasis is placed on AKL nouns, adjectives and verbs and their phraseo-logical patterns. As shown in Figures 6.2 and 6.3, the sections provide infor-mation about how to use these words appropriately by focusing on their:

– semantic properties,– syntactic positioning,– collocations,– frequency,– style and register differences.

All the examples come from the academic component of the British National Corpus.


Evidence from learner corpora was used in several ways to inform the writing sections. The sections specifi cally address the types of problems discussed in Chapter 5 — limited lexical repertoire, lack of register aware-ness, phraseological infelicities, semantic misuse, overuse of connective

You can use the nouns resemblance, similarity, parallel, and analogy to show that twopoints, ideas, or situations are similar in certain ways:

If there is a resemblance or similarity between two or more points, ideas, situations, orpeople, they share some characteristics but are not exactly the same:

There is a striking resemblance between them.

He would have recognized her from her strong resemblance to her brother.

There is a remarkable similarity of techniques, of clothes and of weapons.

The noun similarity also refers to a particular characteristic or aspect that is shared by two ormore points, ideas, situations, or people:

The orang-utan is the primate most closely related to man; its lively facial expressions showstriking similarities to those of humans.

These theories share certain similarities with biological explanations.

Collocation

Adjectives frequently used with resemblance andsimilarity.

Certain, close, remarkable, striking, strong, superficial

The distribution of votes across the three parties in1983 bears a close resemblance to the elections of1923 and of 1929.

Collocation

Adjectives frequently used with analogy and parallelclose, interesting, obvious

A close analogy can be drawn between cancer ofthe cell and a society hooked on drugs

You can also use the noun parallel to refer to the way in which points, ideas, situations, orpeople, are similar to each other:

Scientists themselves have often drawn parallels between the experience of a scientificvocation and certain forms of religious experience.

There are close parallels here with anti-racist work in education.

An analogy is a comparison between two situations, processes, etc which are similar in someways, usually made in order to explain something or make it easier to understand:

A usefull analogy for understanding Piaget's theory is to view the child as a scientists who isseeking a 'theory' to explain complex phenomena.

Figure 6.2 Comparing and contrasting: using nouns such as ‘resemblance’ and ‘similarity’ (Gilquin et al., 2007b: IW5)


devices and syntactic positioning. Our treatment of these problems is mainly explicit, in that we draw learners’ attention to error-prone items and we provide negative feedback in the form of ‘Be careful!’ notes which focus on problems of frequency (over- and underuse), register confusion and atypical positioning. These notes are typically supported by frequency data, in the form of graphs which help the reader visualize the differences between learners’ language and that of native writers. Thus, in the section on ‘Expressing cause and effect’, a graph is used to show that learners have a strong tendency to use the adverb so, which is relatively rare in academic prose and much more typical of speech (see Figure 6.4). There are also ‘Get it right’ boxes which are intended to give guidance on how to avoid common errors. Numerous authentic examples are provided to illustrate

Figure 6.3 Reformulation: explaining and defi ning: using ‘i.e.’, ‘that is’ and ‘that is to say’ (Gilquin et al., 2007b: IW9)

Academic writing

160140120

Fre

q. p

er m

illio

n w

ord

s

10080604020

i.e. that is that is to say0

When you want to explain or define exactly what you mean by something, you can use theabbreviation i.e. (short for 'id est', the Latin equivalent of 'that is') or the expressions that isand that is to say:

The police now have up to ninety-six hours, i.e. four days and nights, to detain people withoutcharge.

Descartes was obsessed by epistemological questions, that is, questions about what we canknow and how we can know it.

First, it excludes the public sector, that to say, the nationalized industries.

Network emergencies (i.e. network failures) should be reported immediately.

That is and that is to say are usually enclosed by commas. The abbreviation i.e. follows acomma or is used between brackets:

Note that, in academic writing and professional reports, i.e. and that is are much morefrequent than that is to say.


all the points we make. The reader is referred to Gilquin et al. (2007a) for more detailed information on the principles that guided the design of these writing sections.

My investigation of academic vocabulary has shown that the use of learner corpus data, and their systematic comparison with native corpora, can bring to light a wide range of learner-specifi c features, not limited to grammatical or lexical errors, but also including over-reliance on a limited set of lexical devices and under-representation of a wide range of typical academic words and phraseological patterns. While Gilquin et al. (2007a; 2007b) have shown how these fi ndings can be integrated into a learner’s dictionary, other writing resources, such as textbooks or electronic writing aids,6 could equally benefi t from the use of learner corpus data.

Learners often use so to express an effect. This use is correct, but it is more typical of speechand should therefore not be used too often in academic writing and professional reports.

Be careful!

so expressing effect

Academic writing Learner writing Speech

1200

1000

800600

400

200

Fre

q. p

er m

illio

n w

ord

s

0

Figure 6.4 Expressing cause and effect: ‘Be careful’ note on ‘so’ (Gilquin et al., 2007b: IW13)

Chapter 7

General conclusion

This book lies at the intersection of three areas of research: English for academic purposes, learner corpus research and second language acquisition. In this fi nal chapter, I take stock of the main fi ndings of the present study and bring out its major contributions to these three research areas. The chapter concludes with some avenues for future research.

7.1. Academic vocabulary: a chimera?

The status and usefulness of EAP has been questioned by Hyland who believes that ‘academic literacy is unlikely to be achieved through an orientation to some general set of trans-disciplinary academic conven-tions and practices’ (Hyland, 2000: 145). This book, however, supports and substantiates the concept of ‘English for (General) Academic Purposes’ both as a macro-genre which subsumes a wide range of text types in academic settings (Biber et al., 1999), and as a teaching practice that deals with ‘the teaching of the skills and language that are common to all disciplines’ (Dudley-Evans and St Johns, 1998: 41) and focuses on ‘a general academic English register, incorporating a formal, academic style, with profi ciency in the language use’ (Jordan, 1997: 5). My own contribution to legitimizing EAP has been to demonstrate – on the basis of corpus data – that ‘it is possible to delimit a procedural vocabulary of such words that would be useful for readers/writers over a wide range of academic disci-plines involving varied textual subject matters and genres’ (McCarthy, 1991: 78).

Academic texts are characterized by a wide range of words and phrase-mes that refer to activities which are typical of academic discourse, and more generally, of scientifi c knowledge. These lexical items also contribute to discourse organization and cohesion, from topic introduction to concluding statements. I have therefore argued in favour of a functional


defi nition of ‘academic vocabulary’ (Martínez et al., 2009) and proposed the following defi nition:

academic vocabulary consists of a set of options to refer to those activities that characterize academic work, organize scientifi c discourse and build the rhetoric of academic texts.

Unlike Coxhead’s (2000) defi nition of the term, a large proportion of what has been referred to as academic vocabulary in this book consists of core words, a category which has so far largely been neglected in EAP courses. Following researchers such as Hanciog lu et al. (2008), I have therefore questioned the fuzzy but well-established frequency-based distinction between general service words and academic words.

Teachers should not assume that EAP students know the fi rst 2,000 words of English. Numerous so-called general service words are not mastered productively by L2 learners, even at upper-intermediate to advanced levels of profi ciency. However, these words serve important discourse-organizing functions in academic writing; this suggests that they should be the target of teaching, particularly teaching aimed at productive activities. My fi nd-ings call into question the systematic use of Coxhead’s Academic Word List as the exclusive vocabulary syllabus in a number of recent productivity- oriented vocabulary textbooks. Another fact that stands out is that a clear distinction should be made between vocabulary needs for academic reading and writing.

As a result, I have derived a productive counterpart to the Academic Word List, and have developed a rigorous and empirically-based procedure to select potential academic words for this list. The methodology makes use of the criteria of keyness, range and evenness of distribution, and provides a good illustration of the usefulness of POS-tagged corpora for applied pur-poses. One important feature of the methodology adopted here is that it includes the 2,000 most frequent words in English, thus making it possible to appreciate the paramount importance of core English words in academic prose. The outcome of this procedure is the Academic Keyword List.

This list should not, however, be regarded as an end product. In its current form (see Table 2.17), the list is the raw result of the application of purely quantitative criteria to native-speaker corpus data. As such, it is not a list of academic vocabulary in a functional sense. Each word still needs ‘pedagogic mediation’ (Widdowson, 2003): its different meanings, lexico-grammatical patterning and phraseology in expert academic prose needs to be carefully described and learner corpus data should be used to

General conclusion 213

complement these descriptions. This procedure has already been applied to the study of words that serve discourse functions (such as exemplifying, expressing cause and effect, comparing and contrasting) in academic prose. I have shown that a phraseological approach to the description of academic vocabulary provides a mine of valuable information for pedagogical tools.

The fi rst result of this method has been to dethrone adverbs from their dominant position as default cohesive markers. Adverbs do not have a monopoly on lexical cohesion and discourse organization in academic writing. My results have provided ample evidence for the prominent discur-sive role of nouns, verbs and adjectives and their phraseological patterns, a role which is hardly ever mentioned in EFL/EAP teaching. These part-of-speech categories, however, serve organizational functions as diverse as exemplifi cation, comparing and contrasting, and expressing cause and effect. Second, the method has helped to demonstrate that an essential set of phrasemes in academic prose consists of ‘lexical extensions’ (Curado Fuentes, 2001: 115) of academic words (e.g. conclusion, issue, claim, argue). These words acquire their organizational or rhetorical function in specifi c word combinations that are essentially semantically and syntactically compositional (e.g. as discussed below, an example of . . . is . . ., the aim of this study, the next section aims at . . ., it has been suggested) (Oakey, 2002; Biber et al., 2004) and contribute to push ‘the boundary that roughly demarcates the “phraseological” more and more into the zone previously thought of as free’ (Cowie, 1998: 20).

The focus has been on words that are reasonably frequent in a wide range of academic texts and their preferred lexicogrammatical and phraseologi-cal patterns, irrespective of discipline. As well as their common core features, these words may also have a discipline-specifi c phraseology (Granger and Paquot, 2009a). Different disciplines may also have their pre-ferred ways of performing rhetorical or organizational functions. A decade ago, Milton (1999: 223) commented that ‘a great deal of research [was] still necessary to describe with any empirical rigour the lexis that is characteris-tic of particular purposes, genres, and registers’. Since then, there has been a huge increase in the number of corpus-based studies highlighting the specifi city of vocabulary and phraseology in different academic disciplines and genres.

The primary motivation of these studies, however, has not been pedagogical. As a result, their fi ndings do not easily lend themselves to being used in general EAP courses and it is now essential to fi nd ways of reconciling research fi ndings and the reality of EAP teaching practice. EAP tutors are left wondering how they can possibly meet the needs of all their


students in classes which are ‘often composed of students from different disciplines and/or language backgrounds with different purposes for taking the class’ (Huckin, 2003: 6). They do not know either what should be taught, for example, to law students who also have to take courses in economics, history, sociology or psychology. With the emergence of a wide range of interdisciplinary curricula, the problem is likely to become even more acute in the future, not only for students but also for their teachers as ‘it seldom happens, especially in mixed classes, that the LSP [Language for Specifi c Purposes] teacher has the disciplinary knowledge needed to pro-vide reliably accurate instruction in technical varieties of language’ (Huckin, 2003: 8).

Faced with this diffi culty, we have advocated elsewhere (Granger and Paquot, 2009a) a balanced approach which concurs with Hyland’s (2002b) plea for more specifi city in EAP teaching while also subscribing to Eldridge’s view that an essential function of research is to identify ‘similarities and generalities that will facilitate instruction in an imperfect world’ (Eldridge, 2008: 111). We have shown that it is possible to identify both the common core features of an academic word and its discipline-specifi c characteristics in terms of meaning, lexico-grammar, phraseological patterns, etc. One way of implementing this ‘happy medium’ approach in the classroom is to apply a data-driven learning methodology, which consists of making use of corpus data as a source of learning materials for language students (Johns, 1994). The study of ‘individualized’ examples derived from specialized corpora can be of considerable benefi t in helping learners to appreciate the possi-ble linguistic realisations of rhetorical and organizational functions in their own disciplines. As Charles put it,

although it may not be possible in all teaching situations to provide mate-rials that are specifi cally tailored to the disciplines of the students taught, the process of investigation is itself of great value in raising students’ awareness of the patterned nature of academic discourse. With this understanding, students are better equipped to examine the ways in which grammatical patterns and lexical choices combine to perform rhe-torical functions within their own disciplines and hence to apply this knowledge to their own academic writing. (2007: 216)

In a heterogeneous EAP class, where disciplinary variability constitutes a serious problem, this approach allows teachers to emphasize general academic words and phrasemes which ‘are not likely to be glossed by the content teacher’ (Flowerdew 1993: 236), while also empowering learners by giving them the tools to investigate authentic texts and practices


in their own disciplines, ‘thereby allowing considerations of subject specifi city and disciplinary variation to inform classroom discussion’ (Groom, 2005: 273).

My journey into academic vocabulary – from the extraction of potential academic words through their linguistic analysis in expert and learner corpus data, to the pedagogical implications that can be drawn from the results – has contributed to fl eshing out this concept and has convincingly demonstrated that academic vocabulary is anything but a chimera.

7.2. Learner corpora, interlanguage and second language acquisition

Contrastive Interlanguage Analysis (CIA) (Granger, 1996) involves two types of comparison. One compares native with non-native (or inter-) language, for example native English and the English produced by French-speaking learners. The other type of analysis compares two (or more) interlanguages, for example the English produced by French-speaking learners and the English produced by Italian-speaking learners. Although the CIA method has become quite popular, most studies using the method have been of the fi rst type. Studies comparing more than one IL usually focus on learners from one mother tongue background and use data from one or two other learner populations only to check whether the features they have high-lighted in one corpus are common to other learners, or are L1-specifi c (and so possibly transfer-related). In this book, I have tried to make the most of CIA by systematically exploiting the two types of comparison it allows to examine EFL learners’ use of academic vocabulary.

The results show that academic, and more precisely argumentative, essays written by upper-intermediate to advanced EFL learners share a number of linguistic features irrespective of the learners’ mother tongue backgrounds or language families. The common core of interlanguage features that characterize the expression of rhetorical and organizational functions in EFL writing includes a limited lexical repertoire and a lack of register aware-ness as well as lexico-grammatical and phraseological specifi cities, the semantic misuse of connectors and labels, the extensive use of chains of connective devices and a marked preference for placing connectors in the sentence-initial position. Several of these linguistic features, and more specifi cally, the lack of register awareness, may also be found in novice native-speaker writing. However, other features such as lexico-grammatical errors, the use of non-native-like sequences and the overuse of relatively rare expressions seem to be largely learner-specifi c.


A systematic analysis of several interlanguages is necessary to analyse the potential infl uence of developmental, teaching-induced and transfer- related factors on EFL learner writing. By focusing on shared features across L1 learner populations, I have highlighted the important role played by developmental and teaching-induced factors in learners’ written production. I have also shown that it is not always possible to attribute learner-specifi c features to a single factor, because developmental, teach-ing-induced and transfer-related effects can reinforce each other (Granger, 2004:135–6). Applying Jarvis’s (2000) methodological framework to learner corpus data has helped identify a number of transfer effects that until now have been largely undocumented in the SLA literature. Lexical transfer has too often been narrowed down to transfer of form/meaning mappings and the third aspect of word knowledge, i.e. use, has rarely been investigated. My study has helped to identify a number of transfer effects relating to word use that make up what, following Hoey (2005), I refer to as ‘transfer of primings’. Transfer of primings includes L1 infl uence on collocational use, lexico-grammatical and phraseological patterns, discourse function, style and register preferences, and frequency of use.

The valuable theoretical insights provided by a learner-corpus based approach to the study of L1 infl uence bring to the fore the potential contri-bution of learner corpora for SLA studies. Learner corpora are probably the best – if not the sole – type of learner interlanguage samples which can be used to investigate these transfer effects. In addition, they arguably pro-vide a good account of the complexity and versatility of L1 infl uence. With its focus on frequency, register differences and phraseology, corpus linguis-tics clearly has numerous resources and specifi c tools to offer SLA research-ers who wish to further investigate the manifestations of L1 infl uence on learners’ interlanguage. There are many other variables that interact in learners’ interlanguage which are also in need of careful operationaliza-tion. Learner corpora can clearly act as a test bed for studies that aim to provide empirical evidence for theories of second language acquisition. They are not the exclusive preserve of learner corpus researchers, and should feature prominently in the battery of data types used by all SLA specialists.

7.3. Avenues for future research

A promising area of research which has only been touched upon in this book lies in the investigation of patterns of diffi culty shared by


mother-tongue English-speaking students and EFL learners. Such research would enable linguistic features that are characteristic of novice writing to be separated from those features that have commonly been attributed to EFL writing. Novice native-speaker writers have been shown to have diffi -culty with academic language, and more particularly with its highly conven-tionalized phraseology. Howarth postulated the existence of a continuum of phraseological competence that would ‘encompass mature NS writers at one extreme and weak NNS writers at the other, with NS and NNS students of varying levels of profi ciency in between, and some overlap between native and non-native writers’ (Howarth, 1999: 151). Hoey (2005) insisted that primings are constrained by register and genre. He gave the example of the word research which is primed in the mind of academic language users to occur with recent in academic discourse and news reports on research. The collocation is not primed to occur in other text types or other contexts. A direct implication of Hoey’s theory of lexical priming is that academic phraseology cannot be assumed to be primed in the mental lexicon of novice native-speaker writers who have had little contact with academic disciplines. Further research is clearly needed to shed more light on the similarities and differences between EFL learners’ use of academic words and phrasemes and that of novice native-speaker writers.

All in all, I have shown that the research paradigm of corpus linguistics is ideally suited to studying the lexical specifi cities of academic discourse in native-speaker and learner writing. The many corpora already available make it possible to examine a wide range of genres and text types. However, much more could be achieved in the fi eld if other types of corpora were collected. In particular, longitudinal corpora of learner language are sorely lacking.1 L1 writing skills also need to fi gure more prominently in future research. It does not make sense to expect learners to write properly in English, and produce coherent and cohesive texts in a foreign language, if they cannot already perform this task in their mother tongue. Learner corpus research would greatly benefi t from the design of comparable corpora of L1 and L2 writing produced by the same learners. There is also an urgent need for learner corpora which represent academic text types other than argumentative essays. New corpora such as the British Academic Written Corpus and the Michigan Corpus of Upper-level Student Papers are thus particularly welcome, as they consist of ESP texts produced by writers at different stages of undergraduate and graduate level study, both native and non-native speakers, in a variety of disciplines. A new corpus currently under development at Louvain, the Varieties of English for Specifi c Purposes dAtabase (VESPA) learner corpus, has been designed as the ESP


counterpart of the International Corpus of Learner English. It includes English for specifi c purposes texts written by L2 writers from various mother tongue backgrounds.

New avenues of research can now be explored by SLA specialists, corpus learner researchers and teachers alike. Not only have a number of largely unrecognized transfer effects been brought to light, but the potential infl u-ence of L1 frequency on learner interlanguage has also been highlighted. The role of frequency is a key issue in second language acquisition. However, it has generally been conceived of in terms of L2 frequency.2 Not a single article in the special issue of Studies in Second Language Acquisition (2002, Volume 24/2) is devoted to L1 frequency effects and their implica-tions for second language acquisition. The volume largely focuses on input frequency, and its relation with language processing, intake3, and implicit vs. explicit learning. Similarly, in a state-of-the-art article on SLA theory, Gregg (2003) only addresses the issue of frequency in relation to the role of input, thus restricting his discussion to the question of ‘how often does input of X need to be provided in order for X to be acquired?’ (Gregg, 2003: 846). The role of L1 frequency is particularly interesting, and can be expected to be the object of much attention in the next few years.

My journey into academic vocabulary has led me to explore a large number of fascinating fi elds of research, and experiment with a wide range of tools and methods. Navigating my way through the complexity of each of these research areas, I have sought to unify several aspects of English for academic purposes, learner corpus research, and second language acquisition into a coherent whole. The challenges presented by such a cross-disciplinary position have quickly been proved worthwhile by the fresh light the approach has shed on key issues such as the nature of academic vocabulary, the relative infl uence of developmental features and transfer effects, and the methodological aspects of interlanguage studies. I hope that this book will serve as a starting block for further research into the many issues raised. There is still so much to explore.

Appendix 1: Expressing cause and effect

Comparisons based on total number of running words

ICLE BNC−AC−HUM LogL

Abs. Rel. Abs. Rel.

nouns

causecause

causes*causae

314127186

1

26.9 755492263

−

22.7 6.3 (++)

factorfactor

factors

229100129

19.7 550244306

16.6 4.6

sourcesource

sources*sourse

27419478

2

23.5 1,175577598

−

35.4 40.2 (− −)

originorigin

origins*origine

604811

1

5.2 500286214

−

15 81.2 (− −)

rootroot

roots

17311261

14.8 18372

111

5.5 83.3 (++)

reasonreason

reasons*reaons

*reasongs

939563374

11

80.6 1,8021105

697−−

54.3 92.2 (++)

consequenceconsequence

consequences*consecvencies*consecuence

*consecuences*consecuenses

*consequencies*consequense

*consequenses

31976

2271232413

27.4 450223269

−−−−−−−

13.6 87.1 (++)

(Continued)

220 Appendix 1


Abs. Rel. Abs. Rel.

effecteffect

effectsefect

395214179

2

33.9 1,8301249581

55 84.8 (− −)

resultresult

results*resut

381167213

1

32.7 813502311

−

24.5 20.9 (++)

outcomeoutcome

outcomes

2821

7

2.4 143135

8

4.3 9.03 (− −)

implicationimplication

implications

1248

1 41193

318

12.4 170.4 (− −)

TOTAL NOUNS 3,124 268 8,612 259.3 2.5

verbs

causecause

causescaused

causing

499140106220

33

42.8 57013366

31754

17.2 211 (++)

bring aboutbringsbrings

broughtbrining

51251014

2

4.4 12544

66411

3.8 0.8

contribute tocontribute

contributescontributed

contributing*contribuates

11661202113

1

10 27652188226

−

8.3 2.6

generategenerate

generatesgenerated

generating

143290

1.2 2276323

11922

6.8 67.4 (− −)

give rise togive

givesgave

givengiving

2084350

1.7 101232132187

3 6.2

Appendix 1 221


Abs. Rel. Abs. Rel.

induceinduce

inducesinduced

inducing

157800

1.3 67195

358

2 2.7

lead tolead

leadsled

leading

356184

837217

30.5 671161105334

71

20.2 31.8 (++)

promptprompt

promptsprompted

prompting

124233

1 1151413826

3.5 22.1 (− −)

provokeprovoke

provokesprovoked

provokingprovocate

provocatedprovoqued

5014

816

8121

4.3 1613811

10210

−−−

4.9 0.6

result in/fromresult

resultsresulted

resulting

114303332

5

8.6 327104

18138

67

9.8 0

yieldyield

yieldsyielded

yielding

22000

0.2 88311634

7

2.7 39.1 (− −)

make sb/sth do sth# 489 42 171 5.2 666.1 (++)

arise from/out ofarise

arisesarose

arisenarising

842200

0.7 145312830

452

4.4 46 (− −)

derivederive

derivesderived

deriving*derivated

3912

815

31

3.4 4767768

29734

−

14.3 115.2 (− −)

(Continued)

222 Appendix 1


Abs. Rel. Abs. Rel.

emergeemerge

emergesemerged

emerging

3311

615

1

2.8 466107

95221

43

14 126.2 (− −)

follow fromfollow

followsfollowed

following

41021

0.3 743335

51

2.2 23.7 (− −)

triggertrigger

triggerstriggered

triggering

85030

0.7 56143

2712

1.7 7 (− −)

stem fromstem

stemsstemmed

stemming

71501

0.6 681422239

2.9 13.3 (− −)

TOTAL VERBS 1,847 158.5 4,174 125.7 66.8 (++)

adjectives

consequent 10 0.9 53 1.6 3.7

responsible (for) 171 14.7 344 10.4 13.3 (++)

TOTAL ADJ. 181 15.5 397 12 4.89

prepositions

because of 531 45.6 599 18 229.6 (++)

due to 246 21.1 195 5.9 175.1 (++)

as a result of 79 6.8 196 5.9 1.1

as a consequence of 7 0.6 22 0.7 0.1

in consequence of 5 0.4 1 0 8.7 (++)

in view of 8 0.7 66 2 10.6 (− −)

owing to 17 1.5 52 1.6 0.1

in (the) light of 7 0.6 109 3.3 31.6 (− −)

thanks to 199 17 35 1 360.1 (++)

on the grounds of 3 0.3 22 0.7 3

on account of 7 0.6 24 0.7 0.19

TOTAL PREP. 1,109 95 1,321 39.8 433.4 (++)

Appendix 1 223


Abs. Rel. Abs. Rel.

Adverbs

thereforetherefore*therefor

70168912

60.1 1,412 42.5 54.1 (++)

accordingly 26 2.2 130 3.9 7.7 (−)

consequentlyconsequently

*consecuently

183179

4

15.7 143 4.3 132.4 (++)

thus 446 38.3 1,767 53.2 41.2 (− −)

hence 42 3.6 283 8.5 33.3 (− −)

so 1,436 123.2 1,894 57 457.8 (++)

thereby 15 1.3 182 5.5 43.8 (− −)

as a result 103 8.8 101 3 55.7 (++)

as a consequence 35 3 20 0.6 34.3 (++)

in consequence 11 0.9 14 0.4 3.8

by implication 0 0 35 1.1 21.1 (− −)

TOTAL ADVERBS 2,998 257.2 5,981 180 243.3 (++)

conjunctions

becausebecause

*becausae*becaus

2,4952,493

11

214 2,207 66.4 1553.8 (++)

since## 428 36.7 955 28.74 17.1 (++)

as ## 331 28.4 883 26.6 1

for 58 5 1,036 31.2 325.9 (− −)

so that 273 23.4 696 21 2.4

PRO is whythat is whythis is why

which is why

220189

1812

18.916.2

1.51

52221812

1.560.70.50.4

359 (++)381.7 (++)

24.7 (++)0.3

on the grounds that 5 0.4 83 2.5 25 (− −)

TOTAL CONJ. 3,810 326.9 5,912 178 809.1 (++)

TOTAL 13,066 1121 26,407 794.9 989.9 (++)

224 Appendix 1

Comparisons based on total number of ‘cause and effect’ lexical items


Abs. % Abs. %

nouns

cause 314 2.4 755 2.9 6.9 (− −)factor 229 1.8 550 2.1 4.9source 274 2.1 1,175 4.5 145.3 (− −)origin 60 0.55 500 1.9 153.3 (− −)root 173 1.3 183 0.7 36.4 (++)reason 939 7.2 1,802 6.8 1.7consequence 319 2.4 450 1.7 23.5 (++)effect 395 3 1,830 6.9 263.8 (− −)result 381 2.9 813 3.1 0.8outcome 28 0.2 143 0.5 24.4 (− −)implication 12 0.1 446 1.7 274 (− −)

TOTAL NOUNS 3,124 23.9 8,612 32.6 231.2 (− −)

Verbs

cause 499 3.8 570 2.2 84.4 (++)bring about 51 0.4 125 0.5 1.4contribute to 116 0.9 276 1.1 2.2generate 14 0.1 227 0.9 106.6 (− −)give rise to 20 0.2 101 0.4 16.9 (− −)induce 15 0.1 67 0.3 9 (− −)lead to 356 2.7 671 2.5 1.1prompt 12 0.1 115 0.4 39.5 (− −)provoke 50 0.4 161 0.6 8.9 (− −)result in 114 0.9 327 1.2 10.9 (− −)yield 2 0.0 88 0.3 56 (− −)make sb/sth do sth# 489 3.7 171 0.7 463.6 (++)arise from/out of 8 0.1 145 0.6 71.5 (− −)derive 39 0.3 476 1.8 192.7 (− −)emerge 33 0.3 466 1.8 201 (− −)follow from 4 0.0 74 0.3 36.8 (− −)trigger 8 0.1 56 0.2 14.5 (− −)stem 7 0.1 68 0.3 23.6 (− −)

TOTAL VERBS 1,847 14.1 4,174 15.8 16.2 (− −)

adjectives

consequent 10 0.1 53 0.2 9.6 (− −)responsible (for) 171 1.3 344 1.3 0

TOTAL ADJ. 181 1.4 397 1.5 0.9

prepositions

because of 531 4.1 599 2.3 93.3 (++)due to 246 1.9 195 0.7 95.3 (++)as a result of 79 0.6 196 0.7 2.4as a consequence of 7 0.1 22 0.1 1.1

Appendix 1 225


Abs. % Abs. %

in consequence of 5 0.0 1 0 6.45in view of 8 0.1 66 0.3 20.1 (− −)owing to 17 0.1 52 0.2 2.4in (the) light of 7 0.1 109 0.4 50.2 (− −)thanks to 199 1.5 35 0.1 270.7 (++)on the grounds of 3 0.0 22 0.1 6on account of 7 0.1 24 0.1 1.7

TOTAL PREP. 1109 8.5 1321 5 164.1 (++)

adverbs

therefore 701 5.4 1,412 5.4 0.0accordingly 26 0.2 130 0.5 21.4 (− −)consequently 183 1.4 143 0.5 72.6 (++)thus 446 3.4 1,767 6.7 182.7 (− −)hence 42 0.3 283 1.1 70.2 (− −)so 1,436 11 1,894 7.2 144.9 (++)thereby 15 0.1 182 0.7 73.4 (− −)as a result 103 0.8 101 0.4 26.2 (++)as a consequence 35 0.3 20 0.1 21.4 (++)in consequence 11 0.1 14 0.1 1.3by implication 0 0 35 0.1 28.1 (− −)

TOTAL ADVERBS 2,998 23 5,981 22.7 0.3

Conjunctions

because 2,495 19.1 2,207 8.4 790.6 (++)since## 428 3.3 955 3.6 2.9as## 331 2.6 883 3.3 19.3 (− −)for 58 0.4 1,036 3.9 507.59 (− −)so that 273 2.1 696 2.6 10.92 (−)PRO is why

that is whythis is why

which is why

22018928

3

1.71.50.20.0

52221812

0.20.10.10.0

262.8 (++)294.5 (++)

14.8 (++)1.3

on the grounds that 5 0.0 83 0.3 39.4 (− −)

TOTAL CONJ. 3,810 29.2 5,912 22.4 158.3 (++)

TOTAL 13,066 100 26,407 100

# Estimations based on Gilquin (2008).## Estimations based on an analysis of the fi rst 200 occurrences of the word in each corpus.

Appendix 2: Comparing and contrasting

Comparisons based on total number of running words


Abs. Rel. Abs. Rel.

nouns

resemblanceresemblance

resemblances

220

0.20.2

0

116100

16

3.493

0.5

54.9 (− −)

similaritysimilarity

similarities*similarieties

*similiraty

257

1611

2.10.61.40.10.1

212106106

−−

6.383.193.19

35.2 (− −)

parallelparallel

parallels

660

0.50.5

0

1477671

4.42.32.1

54 (− −)

parallelismparallelism

parallelisms*paralelism*parallelim

31011

0.30.1

00.10.1

19109−−

0.60.30.3

2

analogyanalogy

analogies

330

0.30.3

0

175133

42

5.34

1.3

82.9 (− −)

contrastcontrast

contrasts

2518

7

2.11.50.6

522470

52

15.714.2

1.6

178.3 (− −)

comparisoncomparison

comparisons*comparaison*comparision

3836011

3.33.1

00.10.1

311249

62−−

9.47.51.9

49.3 (− −)

differencedifference

differences*differencies

*difference

394187191

63

33.816

16.40.50.3

1,318802516

−−

39.724.115.5

8 (− −)

Appendix 2 227


Abs. Rel. Abs. Rel.

*diference*difference*difference*differency*differene

*difference*diffrences

1111111

0.10.10.10.10.10.10.1

−−−−−−−

differentiationdifferentiation

differentiations*differenciation

3201

0.30.2

00.1

76724−

2.32.20.1

28.3 (− −)

distinctiondistinction

distinctions

4738

9

4.13.30.8

595498

97

17.915

2.9

148.4 (− −)

distinctiveness 2 0.2 10 0.3 0.6

(the) same*similars

2461

21.10.1

559−

16.8 8.5 (+)

(the) contrarycontrary

contraries

17161

1.51.40.1

28271

0.8 3

(the) oppositeopposite

opposites

44404

3.783.40.3

855827

2.6 4.2

(the) reverse 5 0.4 56 1.7 12.6 (− −)

TOTAL NOUNS 860 73.8 4,229 127.3 283.7 (− −)

Adjectives

same 1,058 90.8 2,580 77.7 18.8 (++)similar

similar*similiar

*simmilar

160157

21

13.713.5

0.20.1

1,027 30.930.9

110.5 (− −)

analogous 1 0.1 55 1.7 25.8 (− −)common 275 23.6 1055 31.8 20.4 (− −)comparable 16 1.4 223 6.7 59.8 (− −)identical 12 1 137 4.1 31.3 (− −)parallel 5 0.4 52 1.6 10.9 (− −)alike 23 2 98 3 3.3contrasting 1 0.1 63 1.9 30.3 (− −)different

different*differents*differrent

*diffrent

1,5151510

212

130129.6

0.20.10.1

2,4962496

−−−

75.175.1

268 (++)

differing 4 0.3 72 2.17 22.75 (− −)

(Continued)

228 Appendix 2


Abs. Rel. Abs. Rel.

distinctdistinct*distinc

972

0.80.60.2

278278

−

8.48.4

111.4 (− −)

distinctive 13 1.1 163 4.9 40.3 (− −)distinguishable 2 0.2 33 1 9.9 (− −)unlike 2 0.2 43 1.3 14.9 (− −)contrary 7 0.6 27 0.8 0.5opposite 53 4.6 127 3.8 1.1reverse

reverse*reversed

734

0.60.30.3

23 0.7 0.1

TOTAL ADJECTIVES 3,163 271.4 8,552 257.4 6.4

verbs

resembleresemble

resembledresembles

resembling

3116

311

1

2.71.40.30.90.1

13851184623

4.21.50.51.40.7

5.5

correspondcorrespond

correspondedcorresponds

corresponding

4127

347

3.522.30.30.30.6

13773164828

4.12.20.51.40.8

0.8

look likelook like

looks likelooked like

looking like

106722112

1

9.16.21.81.00.1

1024238193

3.11.31.10.60.1

58.9 (++)

comparecompare

comparedcompares

comparing

1297536

216

11.16.43.10.21.4

278140

711750

8.44.22.10.51.5

6.6 (+)

parallelparallel

parallelsparalleled

paralleling

21001

0.20.1

00

0.1

5694

385

1.70.30.11.10.2

21.7 (− −)

contrastcontrast

contrastedcontrasts

contrasting

73400

0.60.30.3

00

13731474217

4.1 45.3 (− −)6.4

11 (− −)

Appendix 2 229


Abs. Rel. Abs. Rel.

differdiffer

differsdiffered

865729

0

7.44.92.5

0

242112

7357

7.293.42.21.7

0.015

0.334.3 (− −)

distinguishdistinguish

distinguisheddistinguishes

distinguishing*distinquish*distingush

107701612

621

9.26

1.41.00.50.20.1

404164116

3688−−

12.164.93.42.21.7

7.1 (− −)1.8

15.4 (− −)0.0

24.5 (− −)

differentiatedifferentiate

differentiatesdifferentiated

differentiating*differenciate

1812

1212

1.51.00.10.20.10.2

74226

3115−

2.20.70.20.90.5

2.091.40.6

9 (− −)4.2

TOTAL VERBS 527 45.2 1,568 47.2 0.7

adverbs

similarlysimilarly

*similarely*similarily

*similary

3126

113

2.72.20.10.10.3

394 11.9 98.6 (− −)

analogously 1 0.1 2 0.1 0.1identically 0 0 2 0.1 1.2correspondingly 0 0 29 0.9 17.4 (− −)parallely 1 0.1 − −likewise 9 0.8 118 3.6 30.3 (− −)in the same way 38 3.3 56 1.7 9.3 (+)contrastingly 0 0 3 0.1 1.8differently 42 3.6 97 2.9 1.3by/in contrast


927

0.80.20.6

185116

69 5.6

62.7 (− −)54.9 (− −)13.7 (− −)

by way of contrast 1 0.1 0 3.5 2.7 by/in comparison


000

000

23149

0.70.40.3

13.8 (− −)

comparativelycomparatively

*comparitively

1413

1

1.21.10.1

69

−

2.1 3.9

contrariwise 0 0 4 0.1 2.4distinctively 1 0.1 25 0.8 9.3 (− −)on the other hand 418 35.9 372 11.2 258.3 (++)

(Continued)

230 Appendix 2


Abs. Rel. Abs. Rel.

(on the one hand) 100 8.6 136 4.1 29.8 (++)*on the other side 23 2 0 0 62 (++)*on the opposite 3 0.3 0 0 8.1 (++)on the contrary

on the contrary*on the contray

*on the contrairy

164160

31

14 9595

−−

2.9 158.9 (++)

Other expressions with contrary

*in contrary*by the contrary*to the contrary

quite the contrary*in the contrary

rather the contrary*quite contrary

*contrary

13

11242111

1.1 2

00020000

0.1

000

0.10000

4.4

reversely 1 0.1 0 0conversely 6 0.5 62 1.9 12.9 (− −)

TOTAL ADVERBS 875 76.7 1,250 38.7 231.7 (++)

Prepositions

like# 1,435 123.1 2,812 84.7 127.5 (++)unlike 26 2.2 244 7.3 45.8 (− −)in parallel with 0 0 8 0.2 4.8as opposed to 7 0.6 121 3.6 37.4 (− −)as against 0 0 46 1.4 27.7 (− −)in contrast to/with

in contrast toin contrast with

2315

8

21.30.7

82739

2.52.20.3

0.94

3.5versus 7 0.6 53 1.6 7.5 (− −)contrary to 18 1.5 66 2 0.9*in contrary to 2 0.2 0 0*opposite to 3 0.3 0 0 8.1 (++)by/in comparison with

in comparison within comparison to

by comparison within comparison with

392811

00

3.42.40.9

00

5214

42114

1.60.40.10.60.4

12.1 (+)30.5 (++)

14.7 (+)12.6 (− −)

8.4 (− −)

TOTAL PREP. 1,560 133.8 3,484 104.9 62 (++)

Conjunctions

as # 1,157 99.3 5,045 151.9 185.4 (− −)while # 206 17.7 1264 38 124.4 (− −)

Appendix 2 231


Abs. Rel. Abs. Rel.

whereaswhereaswheras

137135

2

11.811.6

0.2

442 13.3 1.6

TOTAL CONJ. 1,500 128.7 6,751 203.2 281.3 (− −)

Other expressions

as . . . as 1,287 110.4 2,766 83.26 67.5 (++)in the same way as/that 19 1.6 38 1.14 1.5compared with/to

compared withcompared to

491237

4.1.03.2

155113

42

4.673.4

1.26

0.421.3 (− −)

15.8 (+)

CONJ compared to/withas compared to/with

when compared to/withif compared to/with

14536

1.20.40.30.5

3211201

10.30.60.0

0.50.22.3

11 (++)

TOTAL 9,854 845.5 29,249 880.5 12.24 (− −)

232 Appendix 2

Comparisons based on total number of ‘comparison and contrast’ lexical items


Abs. % Abs. %

Nouns

resemblance 2 0.0 116 0.4 52.6 (− −)similarity 25 0.3 212 0.7 32.3 (− −)parallel 6 0.1 147 0.5 51.3 (− −)parallelism 3 0.0 19 0.1 1.8analogy 3 0.0 175 0.1 79.5 (− −)contrast 25 0.3 522 1.8 168.9 (− −)comparison 38 0.4 311 1.1 45.1 (− −)difference 394 4 1,318 4.5 4.4differentiation 3 0.0 76 0.3 26.9 (− −)distinction 47 0.5 595 2.0 138.9 (− −)distinctiveness 2 0.0 10 0.0 0.5(the) same 246 2.5 559 1.9 11.8 (− −)(the) contrary 17 0.2 28 0.1 3.5(the) opposite 44 0.5 85 0.3 5.1(the) reverse 5 0.1 56 0.2 11.7 (− −)

TOTAL NOUNS 860 8.7 4,229 14.5 202.8 (− −)

Adjectives

same 1,058 10.7 2,580 0.9 28.2(++)similar 160 1.6 1,027 3.5 98.8(− −)analogous 1 0.0 55 0.2 24.7(− −)common 275 2.8 1055 3.6 15.1(− −)comparable 16 0.2 223 0.8 56.2 (− −)identical 12 0.1 137 0.5 29.2 (− −)parallel 5 0.1 52 0.2 10.1 (−)alike 23 0.2 98 0.3 2.6contrasting 1 0.0 63 0.2 29 (− −)different 1,515 15.4 2,496 8.5 307.7 (++)differing 4 0.0 72 0.3 21.5 (− −)distinct 9 0.1 278 1 106.2 (− −)distinctive 13 0.1 163 0.6 37.7 (− −)distinguishable 2 0.0 33 0.1 9.3 (− −)unlike 2 0.0 43 0.2 14.1 (− −)contrary 7 0.1 27 0.1 0.4opposite 53 0.5 127 0.4 1.7reverse 7 0.1 23 0.1 0.1

TOTAL ADJECTIVES 3,163 32.1 8,552 29.2 19.82 (++)

Verbs

resemble 31 0.3 138 0.5 4.5correspond 41 0.4 137 0.5 0.5look like 106 1.1 102 0.4 63.2 (++)compare 129 1.3 278 1 8.7 (+)parallel 2 0.0 56 0.2 20.6 (−)

Appendix 2 233


Abs. % Abs. %

contrast 7 0.1 137 0.5 42.9 (− −)differ 86 0.9 242 0.8 0.2distinguish 107 1.1 404 1.4 5.1differentiate 18 0.2 74 0.3 1.6

TOTAL VERBS 527 5.4 1,568 5.4 0

Adverbs

similarly 31 0.3 394 1.4 92.3 (− −)analogously 1 0.0 2 0.0 0.1identically 0 0 2 0.0 1.2correspondingly 0 0 29 0.1 16.8 (− −)parallely 1 0.0 0 0 2.8likewise 9 0.1 118 0.4 28.3 (− −)in the same way 38 0.4 56 0.2 10.4 (++)contrastingly 0 0 3 0.0 1.7differently 42 0.4 97 0.3 1.8by/in contrast


927

0.1 185116

69

0.6 59.4 (− −)

by way of contrast 1 0.0 0 0 2.8by/in comparison


000

0 2314

9

0.1 13.4 (− −)

comparatively 14 0.1 69 0.2 3.3contrariwise 0 0 4 0.0 2.3distinctively 1 0.0 25 0.1 8.8 (− −)on the other hand 418 4.2 372 1.3 275.8 (++)(on the one hand) 100 1.0 136 0.5 33 (++)*on the other side 23 0.2 0 0 63.4 (++)*on the opposite 3 0.0 0 0 8.3 (++)on the contrary 164 1.7 95 0.3 166.8 (++)Other expressions with

contrary 13 0.1 2 0.0 25.2 (++)

reversely 1 0.0 0 0 2.8conversely 6 0.1 62 0.2 12 (−)

TOTAL ADVERBS 875 8.9 1,250 4.3 258.6 (++)

Prepositions

like# 1,435 14.6 2,812 9.6 155.8 (++)unlike 26 0.3 244 0.8 42.3 (− −)in parallel with 0 0 8 0.0 4.7as opposed to 7 0.1 121 0.4 35.3 (− −)as against 0 0 46 0.2 26.7 (− −)in contrast to/with 23 0.2 82 0.3 0.6versus 7 0.1 53 0.2 6.9 (− −)contrary to 18 0.2 66 0.2 0.7*in contrary to 2 0.0 0 0 5.5

(Continued)

234 Appendix 2


Abs. % Abs. %

*opposite to 3 0.0 0 0 8.3 (+)by/in comparison with 39 0.4 52 0.2 13.4 (+)

TOTAL PREP. 1,560 15.8 3,484 11.9 83.9 (++)

Conjunctions

as# 1,157 11.7 5,045 17.3 150.5 (− −)while# 206 2.1 1264 4.3 110.6 (− −)whereas 137 1.4 442 1.5 0.7

TOTAL CONJ. 1,500 15.2 6,751 23.1 231.6 (− −)

Other expressions

as … as 1,287 13.1 2,766 9.5 87.8 (++)in the same way as/that 19 0.2 38 0.1 2.7compared with/to 49 0.5 155 0.5 0.2CONJ compared to/

with14 0.1 32 0.1 0.6

TOTAL 9,854 100 29,249 100

# Estimations based on an analysis of the fi rst 200 occurrences of the word in each corpus.

Notes

Chapter 1

1 See Stein (2008) for a review of major twentieth-century projects aimed at developing a controlled vocabulary for foreign language learners.

2 This specifi c set of abstract nouns has variously been referred to as ‘signalling words’ (Jordan, 1984), ‘anaphoric nouns’ (Francis, 1986), ‘carrier nouns’ (Ivanic , 1991), ‘shell nouns’ (Schmid, 2000) and ‘discourse-organising words’ (McCarthy, 1991).

Chapter 2

1 The BAWE Pilot Corpus was a pilot for the ESRC funded project ‘An investigation of genres of assessed writing in British higher education (RES-000-23-0800). It was created in 2001 under the directorship of Hilary Nesi, with support from the University of Warwick Teaching Development Fund.

2 The British Academic Written English (BAWE) corpus was developed at the Universi-ties of Warwick, Reading and Oxford Brookes under the directorship of Hilary Nesi and Sheena Gardner (formerly of the Centre for Applied Linguistics at Warwick University), Paul Thompson (Department of Applied Linguistics, Reading) and Paul Wickens (Westminster Institute of Education, Oxford Brookes), with funding from the ESRC (RES-000-23-0800). The BAWE corpus contains 2761 pieces of profi cient assessed student writing. Holdings are fairly evenly distributed across four broad disciplinary areas (Arts and Humanities, Social Sciences, Life Sciences and Physical Sciences). Thirty-fi ve disciplines are represented.

3 See http://ucrel.lancs.ac.uk/claws7tags.html for a list of tags used in CLAWS C7 tagset (accessed 2 August 2009).

4 If a text is 75,000 words long, it has 75,000 tokens. But a lot of these words will be repeated, and there may be only 2,000 different words (called types) in the text.

5 Sentence examples are taken from the Longman Dictionary of Contemporary English (2005)

6 See the defi nition of a reference corpus proposed by the Expert Advisory Group on Language Engineering Standards (EAGLES96) at http://www.ilc.cnr.it/EAGLES96/corpustyp/node18.html (accessed 2 August 2009).

7 Each of these corpora consists of one million words of British or American written English. The four corpora are equivalent in the sense that they were compiled using the same corpus design and sampling methods. For more information about these corpora, see http://khnt.hit.uib.no/icame/manuals (accessed 2 August 2009).

236 Notes

8 Katz (1996: 19) distinguishes between ‘document-level burstiness’, i.e. ‘multiple occurrences of a content word or phrase in a single-text document, which is contrasted with the fact that most other documents contain no instances of this word or phrase at all’; and ‘within-document burstiness’ or ‘burstiness proper’, i.e. the ‘close proximity of all or some individual instances of a content word or phrase within a document exhibiting multiple occurrences’.

9 Scott’s (2004) WordSmith Tools 4 can compute Juilland’s D values, but only for words in a single fi le, based on an arbitrary division of a text into 8 segments of equal size.

10 Available at http://www.lextutor.ca/vp/eng/ (accessed 2 August 2009).

Chapter 3

1 A random sample of 20 essays from each of the 16 L1 sub-corpora available in the second version of ICLE were submitted to a professional rater who was asked to rate them on the basis of the Common European Framework of Reference for Languages (CEF) descriptors for writing. While 60 per cent of the sample essays were rated as advanced (C1 or C2), the proportion was much higher in some sub-corpora, reaching 100 per cent for students with Swedish mother tongue, but falling as low as 40 per cent for Spanish speakers (Granger et al., 2009: 11–12).

2 ICLEv2 now also includes texts written by students with Chinese, Japanese, Norwegian, Turkish and Tswana mother tongue backgrounds (cf. Granger et al., 2009).

3 ICLE also comprises a Bulgarian sub-corpus. However, essays written by Bulgarian-speaking learners were mainly written without the help of reference tools and were therefore not included in the analysis.

4 Texts longer than 45,000 words were sampled so as to allow for a wider coverage of text types and avoid over-representation of idiosyncratic uses. This design criterion, however, causes problems for certain types of linguistic enquiries. A number of studies in the fi eld of English for academic purposes have shown that words may behave differently and display different preferred lexico- grammatical environments in different sections of a text (see, for example, Gledhill, 2000). Quantitative comparisons between the BNC and ICLE thus have to be treated with caution, especially when the lexical items under study are closely linked to specifi c parts of texts (e.g. words and phrasemes used to intro-duce the main topic or a conclusion).

5 See Stefan Evert’s webpage (http: //www.collocations.de/index.html (accessed 2 August 2009)) for a comprehensive list of measures of association and their mathematical interpretation.

Chapter 4

1 In f[n, c], f is the frequency, n the node and c the collocate. 2 http://www.oed.com (accessed 2 August 2009). 3 These three nouns are listed under the fi rst sense of ‘classic’ in LODCE4.

Notes 237

4 These fi gures are based on disambiguated data. The instances of illustrate used in the sense of ‘to put pictures in a book, article, etc’ are not included.

5 Estimations based on an analysis of the fi rst 200 occurrences of the conjunction in the BNC-AC-HUM.

6 Estimations based on an analysis of the fi rst 200 occurrences of the preposition in the BNC-AC-HUM.

7 This does not mean, however, that there are no idioms, similes, compounds, phrasal verbs, commonplaces and allusions to proverbs and quotations in academic prose. As shown by Gläser, ‘authors of scientifi c writing are prone to modify idioms, proverbs, and quotations for intellectual punning and sophisti-cated allusions’ (1998: 143). Studies focusing on terminological terms used in English for Specifi c Purposes have also revealed the pervasiveness of compounds (e.g. Bourigault et al., 2004) in specialized texts.

Chapter 5

1 The ‘word list’ option of WST4 was used to search for any misspelt form of the words under study in the ICLE.

2 The relative frequencies of for instance and example are higher in most learner cor-pora than in the BNC-AC-HUM in most learner corpora. When the learner corpora for different mother tongues are analysed separately, however, the differences in use are only signifi cant for a few groups. Aggregated frequencies thus also help to reveal general, though moderate, overuse in learner corpora in general.

3 See Miller and Weinert (1995), Siegel (2002) and Biber et al. (1999: 562) for specifi c functions of like in speech. See Müller (2005: 197–228) for an analysis of like as a discourse marker.

4 Other verb co-occurrents that are quite frequent in the BNC-SP but not found in the BNC-AC-HUM are the verbs get and think.

So we’ve got some examples here of some patterns that we want to learn using the N tuple method and tuple and tuple. (BNC-SP)

Again think of the example of erm erm a social club you know, relationships between members, although they may be close and intimate and friendly and all that, are not the same as a relationship between members of a family. (BNC-SP)

5 The noun root is overused in the ICLE largely because it appears in an essay title given to some of the EFL learners, ‘In the words of the old song: “Money is the root of all evil”’, which learners then tend to work into their essays.

6 The underuse of the conjunctions as and while reported here must be treated with caution as it results from estimations based on an analysis of only the fi rst 100 occurrences of each conjunction in each corpus.

7 AKL words are printed in bold in these examples. 8 John Osborne (Université de Haute Savoie, France) kindly pointed out to me

that the sequence according to me also appeared in published textbooks such as Ok! (Lacoste and Marcelin, Nathan 1984), which was widely used in French colleges throughout the 1980s and early 1990s.

238 Notes

9 Gledhill (2000) uses the term ‘collocational cascade’ but, following Granger and Paquot (2008a), I prefer to avoid using the adjective ‘collocational’ to refer to sequences of co-occurrents.

10 [P] indicates a new paragraph in learner writing11 A related problem is that of punctuation. EFL learners sometimes omit commas

after sentence-initial subordinate clauses or connectors or before and after appositives such as that is and that is to say (e.g. According to von Mayer, however, what matters is relative poverty* that is to say* the sudden decrease of wealth, ICLE-IT). By contrast, they sometimes erroneously use a comma after the conjunctions although or (even) though (e.g. When I compare these languages I do not consider English as an easy language, although, I do admit that I have noticed some things that are easier about English than about the other languages that I had the chance to learn, ICLE-PO).

12 Osborne (2008) compared adverb placement in the various interlanguages represented in the fi rst version of the International Corpus of Learner English and found that ‘V-Adv-O order is most frequent in the productions of learners whose L1 has verb-raising (French, Italian and Spanish), and least frequent with speakers of V2 languages (Dutch, German and Swedish), with speakers of non-raising languages (Russian, Polish, Czech and Bulgarian) in between’ (Osborne, 2008: 77).

13 See Paquot (2008b and in preparation) for details on the corpus linguistics meth-ods and statistical measures used to operationalize Jarvis’s (2000) framework on learner corpus data.

14 The results reported here are only preliminary. The fi gures should be treated with caution as the LOCNESS corpus is quite small.

Chapter 6

1 The quality of the teaching material on the use of connectors in English that is freely available on the Internet is generally quite alarming, especially given that students increasingly use the Internet for study purposes.

2 As shown in Section 4.2.1, when the preferred sentence position of individual connectors is taught, the information is often neither corpus-based nor con-fi rmed by corpus data.

3 Cohesion is often dealt with in grammars, where the focus is always on connec-tors. It is noteworthy that, in the new corpus-based Cambridge Grammar of English (Carter and McCarthy, 2006), no attention is given to lexical cohesion, although there is a chapter on textual cohesion (‘Grammar across turns and sentences’, pp. 242–62) as well as a full chapter on ‘Grammar and Academic English’ (pp. 266–94).

4 http://www2c.ac-lille.fr/malraux-bethune/FORMAT/super/anglaisinfo/methodes/ Expressions_et_mots_de_liaison.htm (last accessed: 30 July 2009).

5 See Gilquin et al.(2007a) for a detailed discussion of the role of corpora, and more specifi cally, learner corpora in the design of EAP materials and for possible explanations of the relatively modest role that corpora have played so far.

6 The ‘Improve your writing skills’ section in the MED2 shows how a rigorous corpus-based method can help users achieve higher levels of accuracy and fl uency

Notes 239

in academic writing. However, to achieve maximum effi ciency, it is essential to explore ways of integrating this type of description into the microstructure of dictionaries rather than inserting it as a separate middle section. The Centre for English Corpus Linguistics (Université catholique de Louvain) has therefore recently launched a new dictionary project which consists of a web-based EAP dictionary-cum-writing aid tool, the Louvain EAP Dictionary (LEAD) (see Granger and Paquot, 2008b and 2010). This project is innovative in two main respects: it allows for both onomasiological (via the lexeme) and semasiological (via the con-cept) access and is customizable according to the learner’s mother tongue and the fi eld in which he or she is specializing (business, medicine, etc.).

Chapter 7

1 The Centre for English Corpus Linguistics launched the LONGDALE project in January 2008, with the intention of building a large longitudinal database of learner English containing data from learners with a wide range of mother tongue backgrounds. In the LONGDALE project, the same students will be fol-lowed over a period of two to three years.

2 The major role of L1 frequency has been identifi ed in a few transfer studies focusing on phonology and syntax (Selinker, 1992: 211; Kamimoto et al., 1992).

3 De Bot et al. distinguish between input and intake as follows: ‘“Input” is every-thing around us we may perceive with our senses, and “uptake” or “intake” is what we pay attention to and notice’ (2005: 8).

References1

Aarts, J. (2002), ‘Does corpus linguistic exist? Some old and new issues’, in L. Breivik and A. Hasselgren (eds), From the COLT’s Mouth . . . and Others. Amsterdam: Rodopi, pp. 1–17.

Aarts, J. and Granger, S. (1998), ‘Tag sequences in learner corpora: a key to interlanguage grammar and discourse’, in Granger S. (ed.), Learner English on Com-puter. London and New-York: Addison Wesley Longman, pp. 132–41.

Ädel, A. (2006), Metadiscourse in L1 and L2 English. Amsterdam: John Benjamins.Ädel, A. (2008), ‘Involvement features in writing: do time and interaction trump register

awareness’, in Gilquin, G., Papp, S. and Diez-Bedmar B. M. (eds), Linking up Contras-tive and Learner Corpus Research. Amsterdam and Atlanta: Rodopi, pp. 35–53.

Aijmer, K. (2001), ‘I think as a marker of discourse style in argumentative Swedish student writing’, in Aijmer, K. (ed.), A Wealth of English. Studies in Honour of Göran Kjellmer. Göteborg: Acta Universitatis Gothoburgensis, pp. 247–57.

Aijmer, K. (2002), ‘Modality in advanced Swedish learners’ written interlanguage’, in Granger, S., Hung, J. and Petch-Tyson, S. (eds), Computer Learner Corpora, Second Language Acquisition and Foreign Language Teaching. Language Learning and Language Teaching 6. Amsterdam and Philadelphia: John Benjamins, pp. 55–76.

Altenberg, B. (1984), ‘Causal linking in spoken and written English’. Studia Linguistica, 38, 20–69.

Altenberg, B. (1998), ‘On the phraseology of spoken English: the evidence of recurrent word-Combinations’, in Cowie, A. P. (ed.), Phraseology: Theory, Analysis and Applica-tions. Oxford: Oxford University Press, pp. 101–22.

Altenberg, B. and Tapper, M. (1998), ‘The use of adverbial connectors in advanced Swedish learners’ written English’, in Granger S. (ed.), Learner English on Computer. London and New York: Addison Wesley Longman, pp. 80–93.

Archer, D. (ed.) (2009), What’s in a Word-list? Investigating Word Frequency and Keyword Extraction. Farnham: Ashgate.

Archer, D. (2009a), ‘Does frequency really matter?’, in Archer, D. (ed.), What’s in a Word-list? Investigating Word Frequency and Keyword Extraction. Farnham: Ashgate, pp. 1–15.

Archer, D., Wilson, A. and Rayson, P. (2002), Introduction to the USAS category system. Avail-able from http://www.comp.lancs.ac.uk/computing/research/ucrel/usas/usas%20guide.pdf.

Aston, G. and Burnard, L. (1998), The BNC Handbook. Edinburgh: Edinburgh University Press.

Baayen, R. H., Feldman, L. F. and Schreuder, R. (2006), ‘Morphological infl uences on the recognition of monosyllabic monomorphemic words’. Journal of Memory and Language, 53, 496–512.

Bahns, J. (1993), ‘Lexical collocations: a contrastive view’. ELT Journal, 47 (1), 56–63.Bailey, S. (2006), Academic Writing: A Handbook for International Students (2nd edition).

London and New York: Routledge.

References 241

Baker, M. (1988), ‘Sub-technical vocabulary and the ESP teacher: an analysis of some rhetorical items in medical journal articles’. Reading in a Foreign Language, 4, 91–105.

Baker, P. (2004), ‘Querying keywords: questions of difference, frequency and sense in keyword analysis’. Journal of English Linguistics, 32 (4), 346–59.

Barkema, H. (1996), ‘Idiomaticity and terminology: a multi-dimensional descriptive model’. Studia Linguistica, 50 (2), 125–60.

Bartning, I. (1997), ‘L’apprenant dit avancé et son acquisition d’une langue étrangère: tour d’horizon et esquisse d’une caractérisation de la variété avancée’. AILE, 9, 9–50.

Bauer, L. and Nation, I. S. P. (1993), ‘Word families’. International Journal of Lexicography, 6 (4), 253–79.

Bazerman, C. (1994), ‘Systems of genres and the enactment of social intentions’, in Freedman, A. and Medway, P. (eds), Genre and the New Rhetoric. London: Taylor and Francis, pp. 79–101.

Beheydt, L. (2005), ‘The development of an academic vocabulary’, in Battaner, P. and DeCesaris, J. (eds), De lexicografi a. Actes del I Symposium Internacional de Lexicografi a. Série actvitats 15. Barcelona: Institut universitari de linguistica applicada, pp. 241–50.

Bhatia, V. (2002), ‘A generic view of academic discourse’, in Flowerdew, J. (ed.) Academic discourse. Harlow: Longman, pp. 21–39.

Biber, D. (1988), Variation across Speech and Writing. Cambridge: Cambridge University Press.

Biber, D. (2006), University Language: A Corpus-based Study of Spoken and Written Registers. Amsterdam and Philadelphia: John Benjamins.

Biber, D. and Conrad, S. (1999), ‘Lexical bundles in conversation and academic prose’, in Hasselgård, H. and Oksefjell, S. (eds), Out of Corpora: Studies in Honour of Stig Johans-son. Amsterdam: Rodopi, pp. 181–90.

Biber D., Conrad, S. and Cortes, V. (2004), ‘If you look at . . . .: lexical bundles in university teaching and textbooks’. Applied Linguistics, 25 (3), 371–405.

Biber, D., Johansson, S., Leech, G., Conrad, S. and Finegan, E. (1999), Longman Grammar of Spoken and Written English. Harlow: Longman.

Billurog lu, A. and Neufeld, S. D. (2007), BNL 2709: The most commonly used words in Eng-lish. Fourth Edition. Nicosia: Rüstem Kitabevi.

Biskup, D. (1992), ‘L1 infl uence on learners’ renderings of English collocations: a Polish/ German empirical study’, in Arnaud, P. and Béjoint, H. (eds), Vocabulary and Applied Linguistics. London: Macmillan, pp. 85–93.

Bley-Vroman, R. (1983), ‘The comparative fallacy in interlanguage studies: the case of systematicity’. Language Learning, 33, 1–17.

Bourigault, D., Aussenac-Gilles, N. and Charlet, J. (2004), ‘Construction de ressources terminologiques ou ontologiques à partir de textes: un cadre unifi cateur pour trois études de cas’. Revue d’Intelligence Artifi cielle, 18 (1), 87–110. Available from http://w3.univ-tlse2.fr/erss/textes/pagespersos/bourigault/RIA-bourigault-aussenac-charlet.doc

Bowker, L. (2003), ‘Corpus-based applications for translator training: exploring the possibilities’, in Granger, S., Lerot, J. and Petch-Tyson, S. (eds), Corpus-based Approaches to Contrastive Linguistics and Translation Studies. Amsterdam and Philadelphia: John Benjamins, pp. 169–83.

Bowker, L. and Pearson, J. (2002), Working with Specialized Text: A Practical Guide to Using Corpora. London and New York: Routledge.

242 References

Brill, E. (1992), ‘A simple rule-based part of speech tagger’. Proceedings of ANLP-92, 3rd Conference on Applied Natural Language Processing. Available from http://citeseer.ist.psu.edu/brill92simple.html

Burger, H. (1998), Phraseologie. Eine Einführung am Beispiel des Deutschen. Berlin: Erich Schmidt VercClag.

Burnard, L. (2007), Reference Guide for the British National Corpus (XML edition). Available from http://www.natcorp.ox.ac.uk/XMLedition/URG/

Campion, M. E. and Elley, W. B. (1971), An Academic Vocabulary List. Wellington: NZCER.

Carter, R. (1998 [1987]) Vocabulary: Applied Linguistic Perspectives (2nd edition). London: Routledge.

Carter, R. (1998b), ‘Orders of reality: CANCODE, communication, and culture’. ELT Journal, 52 (1), 43–56.

Carter R. and McCarthy, M. (1988), ‘Lexis and discourse: vocabulary in use’, in Carter, R. and McCarthy, M. (eds), Vocabulary and Language Teaching. New York: Longman, pp. 201–20.

Carter, R. and McCarthy, M. (2006), Cambridge Grammar of English: A Comprehensive Guide. Spoken and Written English Grammar and Usage. Cambridge: Cambridge University Press.

Celce-Murcia, M. and Larsen-Freeman, D. (1999), The Grammar Book: An ESL/EFL Teacher’s Course (2nd edition). Boston: Heinle and Heinle.

Charles, M. (2007), ‘Argument or evidence? Disciplinary variation in the use of the noun “that” pattern in stance construction’. English for Specifi c Purposes, 26 (2), 203–18.

Chen, C. W. (2006), ‘The use of conjunctive adverbials in the academic papers of advanced Taiwanese EFL learners’. International Journal of Corpus Linguistics, 11 (1), 113–30.

Chung, T. and Nation, P. (2003), ‘Technical vocabulary in specialised texts’. Reading in a Foreign Language, 15 (2). Available from http://www.nfl rc.hawaii.edu/rfl /

Clear, J. (1993) ‘From Firth principles: computational tools for the study of collocation’, in Baker M., Francis, G. and Tognini-Bonelli, E. (eds), Text and Technology: In Honour of John Sinclair. Amsterdam: John Benjamins, pp. 271–92.

Cohen, A. D., Glasman, H., Rosenbaum-Cohen, P. R., Ferrera, J. and Fine, J. (1988), ‘Reading English for specialised purposes: discourse analysis and the use of student informants’, in Carrrell P., Devine, J. and Eskey, D. E. (eds), Interactive Approaches to Second Language Reading. Cambridge: Cambridge University Press, pp. 152–67.

Coltier, D. (1988), ‘Introduction et gestion des exemples dans les textes à thèse’. Pra-tiques, 58, 23–41.

Connor, U. (1996), Contrastive Rhetoric: Cross-Cultural Aspects of Second-Language Writing. New York: Cambridge University Press.

Conrad, S. (1999), ‘The importance of corpus-based research for language teachers’. System, 27, 1–18.

Conrad, S. (2004), ‘Corpus Linguistics, Language Variation, and Language Teaching’, in Sinclair, J. McH. (ed.) How to Use Corpora in Language Teaching. Amster-dam: John Benjamins, pp. 67–85.

Cook, G. (1998), ‘The uses of reality: a reply to Ronald Carter’. ELT Journal, 52 (1): 57–63.

Corson, D. (1997), ‘The learning and use of academic English words’. Language Learning, 47 (4), 671–718.

References 243

Cortes, V. (2002), ‘Lexical bundles in Freshman composition’, in Reppen, R., Fitzmaurice, S. M. and Biber, D. (eds), Using Corpora to Explore Linguistic Variation. Amsterdam: John Benjamins, pp. 131–45.

Council of Europe (2001), Common European Framework of Reference for Languages: Learn-ing, Teaching, Assessment. Cambridge: Cambridge University Press.

Cowan, J. R. (1974), ‘Lexical and syntactic research for the design of EFL reading materi-als’. TESOL Quarterly, 8 (4), 389–400.

Cowie, A. P. (1998), ‘Phraseological dictionaries: some east-west comparisons’, in Cowie, A. P. (ed.), Phraseology: Theory, Analysis and Applications. Oxford: Oxford University Press, p. 209–28.

Coxhead, A. (2000), ‘A new Academic Word List’. TESOL Quarterly, 34 (2), 213–38. Coxhead, A. and Nation, P. (2001), ‘The specialised vocabulary of English for

Academic Purposes’, in Flowerdew, J. and Peacock, M. (eds), Research Perspectives on English for Academic Purposes. Cambridge: Cambridge University Press, pp. 252–67.

Coxhead, A. and Hirsh, D. (2007), ‘A pilot science-specifi c word list’. Revue française de Linguistique Appliquée, 12 (2), 65–78.

Coxhead, A., Bunting, J., Byrd, P. and Moran, K. (forthcoming), The Academic Word List: Collocations and Recurrent Phrases. Boston: University of Michigan Press.

Crewe, W. (1990), ‘The illogic of logical connectors’. ELT Journal, 44 (4), 316–25. Curado Fuentes, A. (2001), ‘Lexical behaviour in academic and technical corpora: impli-

cations for ESP development’. Language Learning and Technology, 5 (3), 106–29.Cutting, J. (2000), ‘Written errors of international students and English native speaker

students’, in Blue, G. M., Milton, J. and Saville, J. (eds), Assessing English for Academic Purposes. Frankfurt am Main: Peter Lang, pp. 97–113.

Davies, A. (2003), The Native Speaker: Myth and Reality. Clevedon: Multilingual Matters.De Bot, K., Lowie, W. and Verspoor, M. (2005), Second Language Acquisition: An Advanced

Resource Book. London and New York: Routledge.De Cock, S. (2003), ‘Recurrent sequences of words in native speaker and advanced

learner spoken and written English: a corpus-driven approach’. Unpublished PhD thesis. Louvain-la-Neuve: Université catholique de Louvain.

De Cock, S. and Granger, S. (2004), ‘Computer learner corpora and monolingual learn-ers’ dictionaries: the perfect match’, in Teubert, W. and Mahlberg, M. (eds), The Corpus Approach to Lexicography. Special issue of Lexicographica, 20, 72–86.

Dechert, H. (1984), ‘Second language production: six hypotheses’, in Dechert, H., Möhle, D. and Raupach, M. (eds), Second Language Productions. Tübingen: Gunter Narr, pp. 211–30.

Dechert, H. and Lennon, P. (1989), ‘Collocational blends of advanced language learn-ers: a preliminary analysis’, in Olesky, W. (ed.), Contrastive Pragmatics. Amsterdam: John Benjamins, pp. 131–68.

DeRose, S. (1988), ‘Grammatical category disambiguation by statistical optimization’. Computational Linguistics, 14, 31–9.

Dudley-Evans, T. and St Johns, M. J. (1998), Developments in English for Specifi c Purposes. Cambridge: Cambridge University Press.

Eldridge, J. (2008), ‘“No, there isn’t an ‘academic vocabulary,’ but . . . ” A reader responds to K. Hyland and P. Tse’s “Is there an ‘academic vocabulary’?”’. TESOL Quarterly, 42 (1), 109–13.

Ellis, R. and Barkhuizen, G. (2005), Analysing Learner Language. Oxford: Oxford Univer-sity Press.

244 References

Engels, L.K. (1968), ‘The fallacy of word counts’. International Review of Applied Linguis-tics, 10, 213–31.

Evans, S. and Green, C. (2006), ‘Why EAP is necessary: a survey of Hong Kong tertiary students’. Journal of English for Academic Purposes, 6, 3–17.

Evert, S. (2004), ‘The statistics of word cooccurrences: word pairs and collocations’. Ph.D. thesis, Institut für maschinelle Sprachverarbeitung, University of Stuttgart. Available from http://www.collocations.de/phd.html

Farrell, P. (1990), ‘Vocabulary in ESP: a lexical analysis of the English of electronics and a study of semi-technical vocabulary’. CLCS Occasional Paper, 25, 1–83.

Field, Y. and Yip, L.M.O. (1992), ‘A comparison of internal conjunctive cohesion in the English essay writing of Cantonese and native speakers of English’. RELC Journal, 23 (1), 15–28.

Firth, A. and Wagner, J. (1997), ‘On discourse, communication and (some) fundamental concepts in SLA research’. Modern Language Journal, 81, 285–300.

Flowerdew, J. (1993), ‘Concordancing as a tool in course design’. System, 21 (2), 231–44.Flowerdew, J. (1999), ‘Problems in writing for scholarly publication in English: the case

of Hong-Kong’. Journal of Second Language Writing, 8 (3), 243–64.Flowerdew, J. (2002), ‘Introduction: approaches to the analysis of academic discourse in

English’, in Flowerdew, J. (ed.) Academic Discourse. Harlow: Longman, pp. 1–17.Flowerdew, J. (2006), ‘Signalling nouns in a learner corpus’. International Journal of

Corpus Linguistics, 11 (3), 345–62.Flowerdew, L. (1998), ‘Integrating ‘expert’ and ‘interlanguage’ computer corpora fi nd-

ings on causality: discoveries for teachers and students’. English for Specifi c Purposes, 17 (4), 329–345.

Flowerdew, L. (2001), ‘The exploitation of small learner corpora in EAP materials design’, in Ghadessy, M. and Roseberry, R. (eds), Small corpus studies and ELT. Amsterdam: John Benjamins, pp. 363–79.

Flowerdew, L. (2008) Corpus-based Analyses of the Problem-Solution Pattern: A Phraseological Approach. Studies in Corpus Linguistics 29. Amsterdam and Philadelphia: John Benjamins.

Francis, G. (1986) Anaphoric Nouns. Discourse Analysis Monograph 11. Birmingham: English Language Research, University of Birmingham.

Francis, G. (1994), ‘Labelling discourse: an aspect of nominal-group lexical cohesion’, in Coulthard, M. (ed.), Advances in Written Text Analysis. London and New-York: Routledge, pp. 82–101.

Garside, R. (1987), ‘The CLAWS word-tagging system’, in Garside, R., Leech, G. and Sampson, G. (eds), The Computational Analysis of English. London and New York: Longman, pp. 30–41.

Garside, R. and Smith, N. (1997), ‘A hybrid grammatical tagger: CLAWS4’, in Garside, R., Leech, G. and McEnery, A. (eds), Corpus Annotation: Linguistics Information from Computer Text Corpora. New York: Addison Wesley Longman, pp. 102–21.

Ghadessy, M. (1979), ‘Frequency counts, word lists and materials preparation: a new approach’. English Teaching Forum,17, 24–7.

Gilquin, G. (2000/2001), ‘The integrated contrastive model. Spicing up your data’. Lan-guages in Contrast, 3 (1), 95–123.

Gilquin, G. (2008), ‘Combining contrastive and interlanguage analysis to apprehend transfer’, in Diez-Bedmar, B. M., Gilquin, G. and Papp, S. (eds), Linking Up Contrastive and Learner Corpus Research. Amsterdam and Atlanta: Rodopi, pp. 3–33.

Gilquin, G. and Granger, S. (2008), ‘From EFL to ESL: evidence from the International Corpus of Learner English’. Paper presented at the First Triennial Conference of the

References 245

International Society for the Linguistics of English, Albert-Ludwigs-Universität Freiburg, 8–11 October 2008.

Gilquin, G., Granger, S. and Paquot, M. (2007a) ‘Learner corpora: the missing link in EAP pedagogy, in Thompson, P. (ed.), Corpus-based EAP pedagogy. Special issue of the Journal of English for Academic Purposes, 6 (4), 319–35.

Gilquin, G., Granger, S. and Paquot, M. (2007b), ‘Improve your writing skills: writing sections’, in Rundell, M. (editor in chief) Macmillan English Dictionary for Advanced Learners (2nd edition). Oxford: Macmillan Education, pp. IW4–IW28.

Gilquin, G. and Paquot, M. (2008), ‘Too chatty: learner academic writing and register variation’. English Text Construction, 1 (1), 41–61.

Gläser, R. (1998), ‘The stylistic potential of phraseological units in the light of genre analysis’, in Cowie, A. P. (ed.), Phraseology: Theory, Analysis and Applications. Oxford: Oxford University Press, pp. 125–43.

Gledhill, C. (2000), Collocations in Science Writing. Language in Performance 22. Tuebingen: Gunter Narr Verlag.

Goodman, A. and Payne, E. (1981), ‘A taxonomic approach to the lexis of science’, in Selinker L., Tarone, E. and Hnazeli, V. (eds), English for Academic and Technical Purposes: Studies in Honour of Louis Trimble. Rowley MA: Newbury House, pp. 23–39.

Granger, S. (1996a), ‘From CA to CIA and back: an integrated approach to computer-ized bilingual and learner corpora’, in Aijmer K., Altenberg, B. and Johansson, M. (eds), Languages in Contrast: Text-based Cross-linguistic Studies. Lund Studies in English 88. Lund: Lund University Press, pp. 37–51.

Granger, S. (1996b), ‘Romance words in English: from history to pedagogy’, in Svartvik, J. (ed.), Words. Proceedings of an International Symposium. Stockholm: Almqvist and Wiksell International, pp. 105–21.

Granger, S. (1997), ‘On identifying the syntactic and discourse features of participle clauses in academic English: native and non-native writers compared’, in Aarts, J., de Mönnink, I. and Wekker, H. (eds), Studies in English Language and Teaching. Amsterdam and Atlanta: Rodopi, pp. 185–98.

Granger, S. (ed.) (1998), Learner English on Computer. London and New York: Addison Wesley Longman.

Granger, S. (1998a), ‘The computer learner corpus: a versatile new source of data for SLA research’, in Granger, S. (ed.), Learner English on Computer. London and New York: Addison Wesley Longman, pp. 3–18.

Granger, S. (1998b), ‘Prefabricated patterns in advanced EFL writing: collocations and formulae’, in Cowie, A. P. (ed.), Phraseology: Theory, Analysis and Applications. Oxford: Oxford University Press, pp. 145–60.

Granger, S. (2002), ‘A bird’s-eye view of learner corpus research’, in Granger, S., Hung, J. and Petch-Tyson, S. (eds), Computer Learner Corpora, Second Language Acquisition and Foreign Language Teaching. Language Learning and Language Teaching 6. Amsterdam and Philadelphia: John Benjamins, pp. 3–33.

Granger, S. (2003), ‘The International Corpus of Learner English: a new resource for foreign language learning and teaching and second language acquisition research’. TESOL Quarterly, 37 (3), 538–46.

Granger, S. (2004), ‘Computer learner corpus research: current status and future prospects’, in Connor, U. and Upton, T. (eds), Applied Corpus Linguistics: A Multidi-mensional Perspective. Amsterdam and Atlanta: Rodopi, pp. 123–45.

Granger, S. (2006), ‘Lexico-grammatical patterns of EAP verbs: how do learners cope?’ Paper presented at Exploring the Lexis-Grammar Interface, 5–7 October 2006, University of Hanover, Germany.

246 References

Granger, S. (2009), ‘The contribution of learner corpora to second language acquisition and foreign language teaching: a critical evaluation’, in Aijmer, K. (ed.), Corpora and Language Teaching. Amsterdam and Philadelphia: John Benjamins, pp. 15–32.

Granger, S., Dagneaux, E. and Meunier, F. (eds) (2002), The International Corpus of Learner English. CD-ROM and Handbook. Presses universitaires de Louvain: Louvain-la-Neuve.

Granger S., Dagneaux, E., Meunier, F. and Paquot, M. (2009), The International Corpus of Learner English. Version 2. Handbook and CD-ROM. Louvain-la-Neuve: Presses univer-sitaires de Louvain.

Granger, S. and Paquot, M. (2008a), ‘Disentangling the phraseological web’, in Granger, S. and Meunier, F. (eds), Phraseology: An Interdisciplinary Perspective. Amsterdam and Philadelphia: John Benjamins, pp. 27–49.

Granger, S. and Paquot, M. (2008b), ‘From dictionary to phrasebook?’, in Bernal, E. and DeCesaris, J. (eds), Proceedings of the XIII EURALEX International Congress, Barcelona, Spain, 15–19 July 2008, pp. 1345–55.

Granger, S. and Paquot, M. (2009a), ‘In search of General Academic English: a corpus-driven study’, in Katsampoxaki-Hodgetts, K. (ed.), Options and Practices of L.S.P practitioners Conference Proceedings. University of Crete Publications, E-media, pp. 94–108. Available from http://cecl.fl tr.ucl.ac.be/Downloads/In_search_of_a_general_academic_english.pdf

Granger, S and Paquot, M. (2009b), ‘Lexical verbs in academic discourse: a corpus-driven study of learner use’, in Charles, M., Pecorari , D. and Hunston, S. (eds.), Academic Writing: At the Interface of Corpus and Discourse. Continuum, pp. 193–214.

Granger, S. and Paquot, M. (2010), ‘Customising a general EAP dictionary to learner needs’, in Granger, S. and Paquot, M. (eds) eLexicography in the 21st century: new chal-lenges, new applications. Proceedings of the eLex2009 Conference. Cahiers du Cental, 6. Louvain-la-Neuve: Presses universitaires de Louvain.

Granger, S. and Rayson, P. (1998), ‘Automatic profi ling of learner texts’, in Granger, S. (ed.), Learner English on Computer. London and New York: Addison Wesley Longman, pp. 119–31.

Granger, S. and Swallow, H. (1988), ‘False friends: a kaleidoscope of translation diffi culties’. Langage et l’Homme, 23, 108–20.

Granger, S. and Tyson, S. (1996), ‘Connector usage in the English essay writing of native and non-native EFL speakers of English’. World Englishes, 15, 19–29.

Gregg, K. R. (2003), ‘SLA theory: construction and assessment’, in Doughty, C. and Long, M. H. (eds), Handbook of Second Language Research. London: Blackwell, pp. 831–65.

Gries, S. (2007), ‘Exploring variability within and between corpora: some methodologi-cal considerations’. Corpora, 1 (2), 109–51.

Gries, S. (2008), ‘Dispersions and adjusted frequencies in corpora’. International Journal of Corpus Linguistics, 13 (4), 403–37.

Groom, N. (2005), ‘Pattern and meaning across genres and disciplines: an exploratory study’. Journal of English for Academic Purposes, 4, 257–77.

Halliday, M. and Hasan, R. (1976), Cohesion in English. London: Longman.Hamp-Lyons, L. and Heasley, B. (2006), Study Writing: A Course in Writing Skills for Aca-

demic Purposes. Cambridge: Cambridge University Press.Hancioglu, N., Neufeld, S. and Eldridge, J. (2008), ‘Through the looking glass and into

the land of lexico-grammar’. English for Specifi c Purposes, 27 (4), 459–79.Harris, S. (1997), ‘Procedural vocabulary in law case reports’. English for Specifi c

Purposes, 16 (4), 289–308.

References 247

Harris Leonhard, B. (2002), Discoveries in Academic Writing. Bonston: Heinle and Heinle.Hasselgren, A. (1994), ‘Lexical teddy bears and advanced learners: a study into the ways

Norwegian students cope with English vocabulary’. International Journal of Applied Lin-guistics, 4 (2), 237–60.

Heatley, A. and Nation, P. (1996), Range [Computer software]. Wellington, New Zealand: Victoria University of Wellington. Available from http://www. victoria.ac.nz/lals/resources/range.aspx

Hegelheimer, V. and Fisher, D. (2006), ‘Grammar, writing, and technology: a sample technology-supported approach to teaching grammar and improving writing for ESL learners’. CALICO Journal, 23 (2), 257–79.

Hinkel, E. (2002), Second Language Writers’ Text: Linguistic and Rhetorical Features. London: Lawrence Erlbaum Associates.

Hinkel, E. (2003), ‘Adverbial markers and tone in L1 and L2 students’ writing’. Journal of Pragmatics, 35 (7), 1049–68.

Hinkel, E. (2004), Teaching Academic ESL Writing: Practical Techniques in Vocabulary and Grammar. Mahwah, NJ: Lawrence Erlbaum Associates.

Hirsh, D. and Nation, P. (1992), ‘What vocabulary size is needed to read unsimplifi ed texts for pleasure?’ Reading in a Foreign Language, 8, 689–96.

Hoey, M. (1993), ‘A common signal in discourse: how the word reason is used in texts’, in Sinclair, J. M., Hoey, M. and Fox, G. (eds), Techniques of Description. London and New York: Routledge, pp. 67–82.

Hoey, M. (1994), ‘Signalling in discourse: a functional analysis of a common discourse pattern in written and spoken English’, in Coulthard, M. (ed.), Advances in Written Text Analysis. London: Routledge, pp. 26–45.

Hoey, M. (2005), Lexical Priming: A New Theory of Words and Language. London and New-York: Routledge.

Hoffmann, S. (2004), ‘Are low-frequency complex prepositions grammaticalized? On the limits of corpus data – and the importance of intuition’, in Lindquist, H. and Mair, C. (eds), Corpus Approaches to Grammaticalization in English. Amsterdam and Philadelphia: John Benjamins, pp. 171–210.

Hoffmann, S. and Evert, S. (2006), ‘BNCweb (CQP-edition): The marriage of two corpus tools’, in Braun, S., Kohn, K. and Mukherjee, J. (eds), Corpus Technology and Language Pedagogy: New Resources, New Tools, New Methods, volume 3 of English Corpus Linguis-tics. Peter Lang, Frankfurt am Main, pp. 177–95. Available from http://purl.org/stefan.evert/PUB/HoffmannEvert2006.pdf

Hoffmann, S., Evert, S., Smith, N., Lee D. and Berglund Prytz, Y. (2008), Corpus Linguis-tics with BNCweb – a Practical Guide. Frankfurt am Main: Peter Lang.

Howarth, P. (1996), Phraseology in English Academic Writing: Some Implications for Language Learning and Dictionary Making. Tübingen: Max Niemeyer Verlag.

Howarth, P. (1998), ‘The phraseology of learners’ academic writing’, in Cowie, A. P. (ed.), Phraseology: Theory, Analysis and Applications. Oxford: Oxford University Press, pp. 161–86.

Howarth, P. (1999), ‘Phraseological standards in EAP’, in Bool H. and Luford, P. (eds)

Academic Standards and Expectations: The Role of EAP. Nottingham: Nottingham Univer-sity Press, 149–58.

Huckin, T. (2003), ‘Specifi city in LSP’. Ibérica, 5, 3–17.Hunston, S. and Thompson, G. (eds) (2000), Evaluation in Text: Authorial Stance and the

Construction of Discourse. Oxford: Oxford University Press.Huntley, H. (2006), Essential Academic Vocabulary: Mastering the Complete Academic Word

List. Boston: Houghton Miffl in Company.

248 References

Hyland, K. (1998), ‘Persuasion and context: the pragmatics of academic metadiscourse’. Journal of Pragmatics, 30, 437–55.

Hyland, K. (2000) Disciplinary Discourses: Social Interactions in Academic Writing. Harlow: Pearson Education Limited.

Hyland, K. (2002a), ‘Directives: argument and engagement in academic writing’. Applied Linguistics, 23, 215–39.

Hyland, K. (2002b), ‘Specifi city revisited: how far should we go now?’ English for Specifi c Purposes, 21, 385–95.

Hyland, K. (2005), Metadiscourse. London and New York: Continuum.Hyland, K. (2009), Academic Discourse. London and New York: Continuum.Hyland, K. and Milton, J. (1997), ‘Qualifi cations and certainty in L1 and L2 students’

writing’. Journal of Second Language Writing, 6 (2), 183–205.Hyland, K. and Tse, P. (2007), ‘Is there an “academic vocabulary”?’ TESOL Quarterly, 41

(2), 235–53.Ide, N. (2005), ‘Preparation and analysis of linguistic corpora’, in Schreibman, S.,

Siemens, R. and Unsworth, J. (eds), A Companion to Digital Humanities. Oxford: Blackwell, pp. 289–306.

Ivanic, R. (1991), ‘Nouns in search of a context: a study of nouns with both open- and closed-system characteristics’. International Review of Applied Linguistics in Language Teaching, 29, 93–114.

Jarvis, S. (2000), ‘Methodological rigor in the study of transfer: identifying L1 infl uence in the interlanguage lexicon’. Language Learning, 50 (2), 245–309.

Jarvis, S. and Pavlenko, A. (2008), Crosslinguistic Infl uence in Language and Cognition. New York and London: Routledge.

Johansson, S. (1978), Some Aspects of the Vocabulary of Learned and Scientifi c English. Göteborg: Acta Universitatis Gothoburgensis.

Johns, T. (1994), ‘From printout to handout: grammar and vocabulary teaching in the context of data-driven learning’, in Odlin, T. (ed.), Approaches to Pedagogic Grammar. Cambridge: Cambridge University Press, pp. 293–313.

Jordan, M. P. (1984), Rhetoric of Everyday English Texts. London: Allen and Unwin.Jordan, R. R. (1997), English for Academic Purposes. A Guide and Resource Book for Teachers.

Cambridge: Cambridge University Press.Jordan, R. R. (1999), Academic Writing Course. Study Skills in English. Harlow: Pearson

Education Limited.Juilland, A. and Rodriguez, E. C. (1964), Frequency Dictionary of Spanish Words. La Haye:

Mouton.Kamimoto, T., Shimura, A. and Kellerman, E. (1992), ‘A second language classic recon-

sidered: the case of Schachter’s avoidance’. Second Language Research, 8 (3): 251–77.Katz, S. (1996), ‘Distribution of common words and phrases in text and language model-

ling’, Natural Language Engineering, 2 (1), 15–59.Kellerman, E. (1984), ‘The empirical evidence for the infl uence of the L1 in interlan-

guage’, in Davies A., Criper, C. and Howatt, A. (eds), Interlanguage. Edinburgh: Edinburgh University Press, pp. 98–122.

King, P. (2003), ‘Parallel concordancing and its applications’, in Granger, S., Lerot, J. and Petch-Tyson, S. (eds), Corpus-based Approaches to Contrastive Linguistics and Transla-tion Studies. Amsterdam and Philadelphia: John Benjamins, pp. 157–68.

Krishnamurthy, R. and Kosem, I. (2007), ‘Issues in creating a corpus for EAP pedagogy and research’. Journal of English for Academic Purposes, 6, 356–73.

Kroll, B. (1990), ‘What does time buy? ESL student performance on home vs. class com-positions’, in Kroll, B. (ed.) Second Language Writing. Cambridge: Cambridge University Press, 140–53.

References 249

Lake, J. (2004), ‘Using ‘on the contrary’: the conceptual problems for EAP students’. ELT Journal, 58 (2), 137–44.

Lakshmanan, U. and Selinker, L. (2001), ‘Analysing interlanguage: how do we know what learners know?’ Second Language Research, 17, 393–420.

Laruelle, P. (2004), Mieux écrire en anglais. Paris: Presses Universitaires de France.Laufer, B. (1989), ‘What percentage of text-lexis is essential for comprehension?’, in

Lauren, C. and Nordman, M. (eds), Special Language: From Humans Thinking to Think-ing Machines. Clevedon: Multilingual Matters, pp. 316–23.

Laufer, B. (1992), ‘How much lexis is necessary for reading comprehension?’, in Arnaud, P. J. and Béjoint, H. (eds), Vocabulary and Applied Linguistics. London: Macmillan, pp. 126–132.

Lee, D. (2001), ‘Genres, registers, text types, domains, and styles: clarifying the concepts and navigating a path through the BNC jungle’. Language Learning and Technology, 5 (3), 37–72.

Lee, D. and Swales, J. (2006), ‘A corpus-based EAP course for NNS doctoral students: Moving from available specialized corpora to self-compiled corpora’. English for Specifi c Purposes, 25, 56–75.

Leech, G. (1997), ‘Introducing corpus annotation’, in Garside, R., Leech, G. and McEnery, A. (eds), Corpus Annotation: Linguistics Information from Computer Text Corpora. New York: Addison Wesley Longman, pp. 1–18.

Leech G. (1998) ‘Preface’, in Granger, S. (ed.), Learner English on Computer. London and New York: Addison Wesley Longman.

Leech, G., Rayson, P. and Wilson, A. (2001), Word Frequencies in Written and Spoken English. London: Longman.

Leech, G. and Smith, N. (1999), ‘The use of tagging’, in van Halteren, H. (ed.), Syntactic Wordclass Tagging. Dordrecht: Kluwer Academic Publishers, pp. 23–36.

Lehmann, H.-M., Schneider, P. and Hoffmann, S. (2000), ‘BNCweb’, in Kirk, J. (ed.), Corpora Galore: Analysis and Techniques in Describing English. Amsterdam: Rodopi, pp. 259–66.

Lennon, P. (1996), ‘Getting ‘easy’ verbs wrong at the advanced level’. IRAL, 34 (1), 23–36.

Li, E. S.-L. and Pemberton, R. (1994), ‘An investigation of students’ knowledge of academic and subtechnical vocabulary’, in Flowerdew, L. and Tong, A. K. K. (eds), Entering Text. Hong Kong: The Hong Kong University of Science & Technology, pp. 183–96.

Lonon Blanton, L. (2001), Composition Practice 3. Boston: Heinle and Heinle.Lorenz, G. (1998), ‘Overstatement in advanced learners’ writing: stylistic aspects of

adjective intensifi cation’, in Granger, S. (ed.), Learner English on Computer. London and New York: Addison Wesley Longman, pp. 53–66.

Lorenz, G. (1999a), Adjective Intensifi cation – Learners versus Native Speakers. A Corpus Study of Argumentative Writing. Language and Computers: Studies in Practical Linguistics 27. Amsterdam and Atlanta: Rodopi.

Lorenz, G. (1999b), ‘Learning to cohere: causal links in native vs. non-native argumenta-tive writing’, in Bublitz, W., Lenk, U. and Ventola, E. (eds), Coherence in Spoken and Written Discourse. How to Create it and How to Describe it. Amsterdam and Philadelphia: John Benjamins, pp. 55–75.

Luzón Marco, M. J. (1999), ‘Procedural vocabulary: lexical signalling of conceptual rela-tions in discourse’. Applied Linguistics, 20 (1), 1–21.

Luzón Marco, M. J. (2000), ‘Collocational frameworks in medical research papers: a genre-based study’. English for Specifi c Purposes, 19 (1), 63–86.

250 References

Lynn, R. W. (1973), ‘Preparing word lists: a suggested method’. RELC Journal, 4 (1), 25–32.

Major, M. (ed.) (2006), The Longman Exams Dictionary. Harlow: Longman.Martin, A. (1976), ‘Teaching academic vocabulary to foreign graduate students’. TESOL

Quarterly, 10 (1), 91–7.Martínez, I., Beck, S. and Panza, C. (2009), ‘Academic vocabulary in agricultural research

articles: a corpus-based study’. English for Specifi c Purposes, 28, 183–98.McCarthy, M. (1991), Discourse Analysis for Language Teachers. Cambridge: Cambridge

University Press.McCarthy, M. and O’Dell, F. (2008), Academic Vocabulary in Use. Cambridge: Cambridge

University Press.McEnery, A., Xiao, R. and Tono, Y. (2006), Corpus-based Language Studies: An Advanced

Resource Book. London and New-York: Routledge.Mel’cuk, I. (1998), ‘Collocations and lexical functions’, in Cowie, A. P. (ed.), Phraseology:

Theory, Analysis and Applications. Oxford: Oxford University Press, pp. 23–53.Meunier, F. (2000), ‘A computer corpus linguistics approach to interlanguage grammar:

noun phrase complexity in advanced learner writing’. Unpublished PhD thesis. Université catholique de Louvain: Louvain-la-Neuve.

Meyer, P.G. (1997), Coming to Know: Studies in the Lexical Semantics and Pragmatics of Aca-demic English. Tübingen: Gunter Narr Verlag Tübingen.

Miller, J. and Weinert, R. (1995), ‘The function of like in dialogue’. Journal of Pragmatics, 23, 365–93.

Milton, J. (1998). ‘Exploiting L1 and interlanguage corpora in the design of an elec-tronic language learning and production environment’, in Granger, S. (ed.), Learner English on Computer. London and New York: Addison Wesley Longman, 186-198.

Milton, J. (1999), ‘Lexical thickets and electronic gateways: making text accessible by novice writers’, in Candlin, C. N. and Hyland, K. (eds), Writing: Texts, Processes and Practices. London and New York: Longman, pp. 221–43.

Moon, R. (1998), Fixed Expressions and Idioms in English. Oxford: Clarendon Press.Mudraya, O. (2006), ‘Engineering English: a lexical frequency instructional model’.

English for Specifi c Purposes, 25 (2), 235–56.Mukherjee, J. (2005), ‘The native speaker is alive and kicking – linguistic and language-

pedagogical perspectives’. Anglistik, 16 (2), 7–23.Mukherjee, J. and Rohrback, J-M. (2006), ‘Rethinking applied corpus linguistics from a

language-pedagogical perspective: new departures in learner corpus research’, in Kettemann, B. and Marko, G. (eds), Planning, Gluing and Painting Corpora: Inside the Applied Corpus Linguist’s workshop. Frankfurt am Main: Peter Lang, pp. 205–32. Available from http://www.uni-giessen.de/anglistik/LING/Staff/mukherjee/

Müller, S. (2005), Discourse Markers in Native and Non-native English Discourse. Amsterdam and Philadelphia: John Benjamins.

Narita, M. and Sugiura, M. (2006), ‘The use of adverbial connectors in argumentative essays by Japanese EFL college students’. English Corpus Studies, 13, 23–42.

Nation, P. (2001), Learning Vocabulary in another Language. Cambridge: Cambridge University Press.

Nation, P. and Hwang, K. (1995), ‘Where would general service vocabulary stop and special purposes vocabulary begin?’ System, 23 (1), 35–41.

Nation, P. and Waring, R. (1997), ‘Vocabulary size, text coverage and word lists’, in Schmitt, N. and Nation, P. (eds), Vocabulary: Description, Acquisition and Pedagogy. Cambridge: Cambridge University Press, pp. 6–19.

References 251

Neff, J., Ballesteros, F., Dafouz, E., Martínez, F. and Rica, J. P. (2004a), ‘The expression of writer stance in native and non-native argumentative texts’, in Facchinetti, R. and Palmer, F. (eds), English Modality in Perspective. Frankfurt am Main: Peter Lang, pp. 141–61.

Neff J., Dafouz, E., Díez, M., Prieto, R. and Chaudron, C. (2004b), ‘Contrastive discourse analysis: argumentative text in English and Spanish’, in Moder, C. and Martinovic-Zic, A. (eds), Discourse across Languages and Cultures. Amsterdam and Philadelphia: John Benjamins, pp. 267–83.

Neff J., Ballesteros, F., Dafouz, E., Martínez, F. and Rica, J-P. (2007), ‘A contrastive func-tional analysis of errors in Spanish EFL university writers’ argumentative texts: corpus-based study’, in Fitzpatrick, E. (ed.), Corpus Linguistics beyond the Word: Corpus Research from Phrase to Discourse (Language and Computers 23). Amsterdam: Rodopi, pp. 203–25.

Neff, J., Ballesteros, F., Dafouz, E., Martínez, F., Rica, J-P., Díez, M. and Prieto, R. (2008), ‘Formulating writer stance: a contrastive study of EFL learner corpora’, in Gerbig, A. and Mason, O. (eds), Language, People, Numbers. Corpus Linguistics and Society. Amsterdam and New York: Rodopi, pp. 73–89.

Neff van Aertselaer, J. (2008), ‘Contrasting English-Spanish interpersonal discourse phrases: a corpus study’, in Meunier, F. and Granger, S. (eds), Phraseology in Language Learning and Teaching. Amsterdam: John Benjamins, pp. 85–100.

Nelson, M. (2000), ‘A corpus-based study of business English and business English teaching materials’. Unpublished PhD Thesis. Manchester: University of Manchester.

Nesi, H., Sharpling, G. and Ganobcsik-Williams, L. (2004), ‘Student papers across the curriculum: designing and developing a corpus of British student writing’. Computers and Composition, 21, 439–50.

Nesselhauf, N. (2003), ‘Transfer at the locutional level: an investigation of German-speaking and French-speaking learners of English’, in Tschichold, C. (ed.) English Core Linguistics. Essays in Honour of D. J. Allerton. Bern: Lang, pp. 269–86.

Nesselhauf, N (2004), ‘What are collocations?’, in Allerton, D. J., Tschichold, C. and Wieser, J. (eds), Phraseological Units: Basic Concepts and their Application. Basel: Schwabe, pp. 1–22.

Nesselhauf, N. (2005), Collocations in a Learner Corpus. Amsterdam: John Benjamins.Oakes, M. P. (1998), Statistics for Corpus Linguistics. Edinburgh: Edinburgh University

Press.Oakes, M. and Farrow, M. (2007), ‘Use of the Chi-Squared Test to examine vocabulary

differences in English language corpora representing seven different countries’. Literary and Linguistic Computing, 22 (1), 85–99.

Oakey, D. (2002), ‘Formulaic language in English academic writing: a corpus-based study of the formal and functional variation of a lexical phrase in different academic disci-plines’, in Reynolds, D. (2005), ‘Linguistic correlates of second language literacy development: evidence from middle-grade learner essays’. Journal of Second Language Writing,14 (1), 19–45.

Obenda, D. (ed.) (2004), Academic Word Power (1 – 4). Boston and New York: Houghton Miffl in Company.

Odlin, T. (1989), Language Transfer: Cross-linguistic Infl uence in Language Learning. Cambridge: Cambridge University Press.

Odlin, T. (2003), ‘Cross-linguistic infl uence’, in Doughty, C. J. and Long, M. H. (eds), The Handbook of Second Language Acquisition. Oxford: Blackwell, pp. 436–86.

252 References

Osborne, J. (2008), ‘Phraseology effects as a trigger for errors in L2 English: the case of more advanced learners’, in Meunier, F. and Granger, S. (eds), Phraseology in Language Learning and Teaching. Amsterdam and Philadelphia: John Benjamins, pp. 67–84.

Oshima, A. and Hogue, A. (2006), Writing Academic English. New York: Pearson Education.

Paquot, M. (2007a), ‘Towards a productively-oriented academic word list’, in Walinski, J., Kredens, K. and Gozdz-Roszkowski, S. (eds), Corpora and ICT in Language Studies. PALC 2005. Lodz Studies in Language 13. Frankfurt am Main: Peter Lang, pp. 127–40.

Paquot, M. (2007b), ‘EAP vocabulary in native and learner writing: from extraction to analysis. A phraseologically-oriented approach’. Unpublished PhD thesis, Université catholique de Louvain.

Paquot, M. (2008a), ‘Exemplifi cation in learner writing: a cross-linguistic perspective’, in Granger, S. and Meunier, F. (eds), Phraseology in Language Learning and Teaching. Amsterdam: John Benjamins, pp. 101–19.

Paquot, M. (2008b), ‘Lifting the “methodological fog” that covers transfer studies: a combination of Granger’s (1996) integrated contrastive model and Jarvis’s (2000) unifi ed framework for transfer research’. Paper presented at the ASKeladden Opening Conference, 24–25 June 2008, Bergen, Norway. Available from http://cecl.fl tr.ucl.ac.be/publications.html

Paquot, M. (in preparation), ‘Unveiling L1-induced effects with the help of learner corpora: Transfer of lexical priming’.

Paquot, M. and Bestgen, Y. (2009), ‘Distinctive words in academic writing: a comparison of three statistical tests for keyword extraction’, in Hundt, M., Schreier, D. and Jucker, A. H. (eds), Corpora: Pragmatics and Discourse. Amsterdam: Rodopi, pp. 243–65.

Partington, A. (1998), Patterns and Meanings. Using Corpora for English Language Research and Teaching. Amsterdam and Philadelphia: John Benjamins.

Pawley, A. and Syder, F. H. (1983), ‘Two puzzles for linguistic theory: nativelike selection and nativelike fl uency’, in Richards, J. C. and Schmidt, W. (eds), Language and Communication. London and New York: Longman, pp. 29–59.

Pecman, M. (2004), ‘Phraséologie contrastive anglais-français: analyse et traitement en vue de l’aide à la rédaction scientifi que’. Unpublished PhD Thesis, Université de Nice, Sophia Antipolis.

Perdue, C. (1993), ‘Comment rendre compte de la “logique” de l’acquisition d’une langue étrangère par l’adulte?’ Etudes de Linguistique Appliquée, 92, 8–22.

Petch-Tyson, S. (1998), ‘Reader/writer visibility in EFL persuasive writing’, in Granger, S. (ed.), Learner English on Computer. London and New York: Addison Wesley Longman, pp. 107—118.

Petch-Tyson, S. (1999), ‘Demonstrative expressions in argumentative discourse – a com-puter-based comparison of non-native and native English’, in Botley, S. P. and McEnery, A. M. (eds), Corpus-based and Computational Approaches to Discourse Anaphora. Amsterdam and Philadelphia: John Benjamins, pp. 43–64.

Piller, I. (2001), ‘Who, if anyone, is a native speaker?’ Anglistik, 12 (2), 109–21.Praninskas, J. (1972), American University Word List. London: Longman.Quirk, R., Greenbaum, S., Leech, G. and Svartvik, J. (1985), A Comprehensive

Grammar of the English Language. London: Longman.Rayson, P. (2003), ‘Matrix: a statistical method and software tool for linguistic analysis

through corpus comparison’. Unpublished PhD thesis, Lancaster University. Avail-able from http://www.comp.lancs.ac.uk/~paul/public.html

Rayson, P. (2008), ‘From key words to key semantic domains’. International Journal of Corpus Linguistics, 13 (4), 519–49.

References 253

Rayson, P., Berridge, D. and Francis, B. (2004), ‘Extending the Cochran rule for the com-parison of word frequencies between corpora’, in Purnelle, G., Fairon, C. and Dister, A. (eds), Le Poids des Mots: Proceedings of the 7th International Conference on Statistical Analy-sis of Textual Data (JADT 2004), Louvain-la-Neuve, Belgium, March 10–12, 2004. Louvain-la-Neuve: Presses universitaires de Louvain, pp. 926–36.

Renouf, A. and Sinclair, J. (1991), ‘Collocational frameworks in English’, in Aijmer, K. and Altenberg, B. (eds), English Corpus Linguistics: Studies in Honour of Jan Svartvik. London and New York: Longman, pp. 128–43.

Reynolds, D. W. (2005), ‘Linguistic correlates of second language literacy development: evidence from middle-grade learner essays’. Journal of Second Language Writing, 14, 19–45.

Ringbom, H. (1987), The Role of the First Language in Foreign Language Learning. Clevedon and Philadelphia: Multilingual Matters.

Ringbom, H. (1998), ‘Vocabulary frequencies in advanced learner English: a crosslinguistic approach’, in Granger, S. (ed.), Learner English on Computer. London and New York: Addison Wesley Longman, pp. 41–52.

Ringbom, H. (2007), Cross-linguistic Similarity in Foreign Language Learning. Clevedon: Multilingual Matters.

Römer, U. (2004a), ‘A corpus-driven approach to modal auxiliaries and their didactics’, in Sinclair, J. McH. (ed.), How to Use Corpora in Language Teaching. Amsterdam: John Benjamins, pp. 185–99.

Römer, U. (2004b), ‘Comparing real and ideal language learner input: The use of an EFL textbook corpus in corpus linguistics and language teaching’, in Aston, G., Bernardini, S. and Stewart, D. (eds), Corpora and Language Learners. Amsterdam: John Benjamins, pp. 151–68.

Römer, U. (2005), Progressives, Patterns, Pedagogy. A Corpus-driven Approach to English Progressive Forms, Functions, Contexts and Didactics. Amsterdam: John Benjamins.

Römer, U. (2008), ‘Corpora and language teaching’, in Lüdeling, A. and Kytö, M. (eds), Corpus Linguistics. An International Handbook (volume 1). [HSK series]. Berlin: Mouton de Gruyter, pp. 112–30.

Ruetten, M. (2003), Developing Composition Skills. Rhetoric and Grammar. Boston: Heinle.Rundell, M. (1999), ‘Dictionary use in production’. International Journal of Lexicography,

12 (1), 35–53.Rundell, M. (ed.) (2007), Macmillan English Dictionary for Advanced Learners. (2nd

Edition). Oxford: Macmillan Education. Saville-Troike, M. (1984), ‘What really matters in second language learning for academic

achievement?’ TESOL Quarterly 18 (2), 199–219.Scarcella, R. C. and Zimmerman, C. B. (2005), ‘Cognates, cognition and writing: an

investigation of the use of cognates by university second-language learners’, in Tyler, A., Takada, M., Kim, Y. and Marinova, D. (eds), Language in Use: Cognitive and Discourse Perspectives on Language and Language Learning. Washington: Georgetown University Press, pp. 123–36.

Schleppegrell, M. J. (1996), ‘Conjunction in spoken English and ESL writing’. Applied Linguistics, 17 (3), 271–85.

Schmid, H-J. (2000), English Abstract Nouns as Conceptual Shells: From Corpus to Cognition. Berlin: Mouton de Gruyter.

Schmitt, N. and Schmitt, D. (2005), Focus on Vocabulary: Mastering the Academic Word List. London: Longman.

Schmitt, N., Schmitt, D. and Clapham, C. (2001), ‘Developing and exploring the behaviour of two new versions of the Vocabulary Levels Test’. Language Testing, 18 (1), 55–88.

254 References

Scott, M. (1997), ‘PC analysis of keywords and key keywords’. System, 25 (2), 233–45.Scott, M. (2001), ‘Comparing corpora and identifying key words, collocations, frequency

distributions through the WordSmith Tools suite of computer programs’, in Ghadessy, M., Henry, A. and Roseberry, L. (eds), Small Corpus Studies and ELT. Amsterdam: John Benjamins, pp. 47–67.

Scott, M. (2004), WordSmith Tools 4. Oxford: Oxford University Press.Scott, M. and Tribble, C. (2006), Textual Patterns: Key Words and Corpus Analysis in Lan-

guage Education. Amsterdam: John Benjamins.Seale, C., Ziebland, S. and Charteris-Black, J. (2006), ‘Gender, cancer experience and

internet use: a comparative keyword analysis of interviews and online cancer support groups’. Social Science and Medicine, 62, 2577–90.

Selinker, L. (1972), ‘Interlanguage’. IRAL, X (3), 209–31.Selinker, L. (1992), Rediscovering Interlanguage. London and New York: Longman.Shaw, P. (2004), ‘The development of Swedish university students’ written English,

appropriacy, scope and coherence’, in Proceedings of the Ninth Nordic Conference for Eng-lish Studies, Aarhus, Denmark, May 27–29. Available from http://www.hum.au.dk/engelsk/naes2004/papers.html

Siegel, M. (2002), ‘Like: the discourse particle and semantic’. Journal of Semantics, 19 (1), 35–71.

Siepmann, D. (2005), Discourse Markers across Languages: A Contrastive Study of Second-level Discourse Markers in Native and Non-native Text with Implications for General and Pedagogic Lexicography. London and New York: Routledge.

Sinclair, J. M. (1987), ‘The nature of the evidence’, in Sinclair, J. M. (ed.), Looking up. London: Collins, pp. 150–9.

Sinclair, J. M. (1991), Corpus, Concordance, Collocation. Oxford: Oxford University Press.Sinclair, J. M. (1992), ‘The automatic analysis of corpora’, in Svartvik, J. (ed.), Directions

in Corpus Linguistics. Berlin and New York: Mouton de Gruyter, pp. 378–97.Sinclair, J. M. (1999), ‘The lexical item’, in Weigand, E. (ed.), Contrastive Lexical

Semantics. Current Issues in Linguistic Theory 17. Amsterdam and Philadelphia: John Benjamins, pp. 1–24.

Sinclair, J. M. (2004a), ‘Intuition and annotation – the discussion continues’, in Aijmer, K. and Altenberg, B. (eds), Advances in Corpus Linguistics. Papers from the 23rd Interna-tional Conference on English Language Research on Computerized Corpora. Amsterdam and New York: Rodopi, pp. 39–59.

Sinclair, J. M. (2004b), ‘The empty lexicon’, in Sinclair, J. (ed.), Trust the Text: Language, Corpus and Discourse. New York: Routledge, pp. 149–63.

Sinclair, J. M. (2005), ‘Corpus and text–basic principles’, in Wynne, M. (ed.), Developing Linguistic Corpora: A Guide to Good Practice. Oxford: Oxbow Books, 1–16. Available from http://ahds.ac.uk/creating/guides/linguistic- corpora/chapter1.htm.

Soler, V. (2002), ‘Analysing adjectives in scientifi c discourse: an exploratory study with educational applications for Spanish speakers at advanced university level’. English for Specifi c Purposes, 21, 145–65.

Stein, G. (2008), Developing your English Vocabulary. A Systematic New Approach. Tübingen: Stauffenburg Verlag.

Strevens, P. (1973), ‘Technical, technological, and scientifi c English’. ELT Journal, 27 (3), 223–34.

Stubbs, M. (1986), ‘Language development, lexical competence and nuclear vocabu-lary’, in Stubbs, M. (ed.), Educational Linguistics. Oxford and New York: Blackwell, pp. 98–115.

References 255

Summers, D. (1996), ‘Computer lexicography: the importance of representativeness in relation to frequency’, in Thomas, J. and Short, M. (eds), Using Corpora for Language Research: Studies in Honour of Geoffrey Leech. London: Longman, pp. 260–6.

Sutarsyah, C., Nation, P. and Kennedy, G. (1984), ‘How useful is EAP vocabulary for ESP? A corpus based case study’. RELC Journal, 25, 34—50.

Swales, J.M. (1990) Genre Analysis: English in Academic and Research Settings. Cambridge: Cambridge University Press.

Swales, J. M. (2002), ‘Integrated and fragmented worlds: EAP materials and corpus linguistics’, in Flowerdew, J. (ed.), Academic Discourse. Harlow: Pearson Education, pp. 150–64.

Swales, J.M., Ahmad, U., Chang, Y., Chavez, D., Dressen, D. and Seymour, R. (1998), ‘Consider this: the role of imperatives in scholarly writing’. Applied Linguistics, 19, (1), 97–121.

Swales, J.M. and Feak, C. B. (2004), Academic Writing for Graduate Students: Essential Tasks and Skills (2nd edition). Ann Arbor, MI: University of Michigan Press.

Tan, M. (2005), ‘Authentic language or language errors? Lessons from a learner corpus’. ELT Journal, 59 (2), 126–34.

Tankó, G. (2004), ‘The use of adverbial connectors in Hungarian university students’ argumentative essays’, in Sinclair, J. (ed.), How to Use Corpora in Language Teaching. Amsterdam and Philadelphia: John Benjamins, pp. 157–81.

Thurstun, J. and Candlin, C. (1997), Exploring Academic English. A Workbook for Student Essay Writing. Sydney: NCELTR Publications.

Thurstun, J. and Candlin, C. (1998), ‘Concordancing and the teaching of the vocabulary of Academic English’. English for Specifi c Purposes, 17 (3), 267–80.

Tognini-Bonelli, E. (2001), Corpus Linguistics at Work. Amsterdam and Philadephia: John Benjamins.

Tognini-Bonelli, E. (2002), ‘Functionally complete units of meaning across English and Italian: towards a corpus-driven approach’, in Altenberg, B. and Granger, S. (eds) Lexis in Contrast: Corpus-based Approaches. Amsterdam and Philadephia: John Benjamins, pp. 73–95.

Tribble, C. (2001), ‘Small corpora and teaching writing: towards a corpus-informed pedagogy of writing’, in Ghadessy, M., Henry, A. and Roseberry, R. (eds), Small Corpus Studies and ELT: Theory and Practice. Amsterdam and Philadelphia: John Benjamins, pp. 381–408.

Trimble, L. (1985), English for Science and Technology. Cambridge: Cambridge University Press.

Tseng, Y-C. and Liou, H-C. (2006), ‘The effects of online conjunction materials on college EFL students’ writing’. System, 34, 270–83.

Tutin, A. (forthcoming), ‘Evaluative adjectives in academic writing in the humanities and social sciences’. Paper presented at Interpersonality in Written Academic Discourse: Perspectives across Languages and Cultures, Jaca, 11–13 December 2008. Available from http://w3.u-grenoble3.fr/lidilem/labo/fi le/evaluative_adjectives_interlae_2008_tutin.pdf

Van Roey, J. (1990), French-English Contrastive Lexicology: An Introduction. Louvain-la-Neuve: Peeters.

Vassileva, I. (1998), ‘Who am I/who are we in academic writing? A contrastive analysis of authorial presence in English, German, French, Russian and Bulgarian’. International Journal of Applied Linguistics, 8 (2), 163–90.

256 References

Voutilainen, A. (1999), ‘A short history of tagging’, in van Halteren, H. (ed.), Syntactic wordclass tagging. Dordrecht: Kluwer Academic Publishers, pp. 9–21.

Wang, J., Liang, S., and Ge, G. (2008), ‘Establishment of a Medical Academic Word List’. English for Specifi c Purposes, 27 (4), 442–58.

Wang, K. and Nation, P. (2004), ‘Word meaning in academic English: homography in the Academic Word List’. Applied Linguistics, 25 (3), 291–314.

Ward, J. (1999), ‘How large a vocabulary do EAP engineering students need?’ Reading in a Foreign Language, 12 (2), 309–24.

Ward, J. (2009), ‘A basic engineering English word list for less profi cient foundation engineering undergraduates’. English for Specifi c Purposes, 28, 170–82.

Weissberg, R. and Buker, S. (1978), ‘Strategies for teaching the rhetoric of written English for Science and Technology’. TESOL Quarterly,12 (3), 321–9.

West, M. (1937), ‘The present position in vocabulary selection for foreign language teaching’. The Modern Language Journal, 21 (6), 433–7.

West, M. (1953), A General Service List of English Words. London: Longman. Widdowson, H. G. (1983), Learning Purpose and Language Use. Oxford: Oxford

University Press.Widdowson, H. G. (1991), ‘The description and prescription of language’, in Alatis, J. E.

(ed.), Linguistics and Language Pedagogy: The State of the Art. Washington, D.C.: Georgetown University Press, pp. 11–24.

Widdowson, H. G. (2003), Defi ning Issues in English Language Teaching. Oxford: Oxford University Press.

Wilkins, D. A. (1976), Notional Syllabuses. Oxford: Oxford University Press.Wilson, A. and Thomas, J. (1997), ‘Semantic annotation’, in Garside, R., Leech, G. and

McEnery, A. (eds), Corpus Annotation: Linguistics Information from Computer Text Corpora. New York: Addison Wesley Longman, pp. 53–65.

Winter, E. (1977), ‘A clause relational approach to English texts: a study of some predic-tive lexical items in written discourse’. Instructional Science, 6, 1–92.

Wray, A. (2002), Formulaic Language and the Lexicon. Cambridge: Cambridge University Press.

Xue, G. and Nation, P. (1984), ‘A University Word List’. Language Learning and Communication, 3 (2), 215–29.

Yang, H. (1986), ‘A new technique for identifying scientifi c / technical terms and describ-ing science texts’. Literary and Linguistic Computing, 1 (2), 93–103.

Zamel, V. (1983), ‘Teaching those missing links in writing’. ELT Journal, 37 (1), 22–9.Zemach, D. and Rumisek, L. (2005), Academic Writing: From Paragraph to Essay. Oxford:

Macmillan.Zhang, M. (2000), ‘Cohesive features in the expository writing of undergraduates in two

Chinese universities’. RELC Journal, 31, 61–95.Zhang, H., Huang, C. and Yu, S. (2004), ‘Distributional consistency: A general method

for defi ning a core lexicon’, in Proceedings of the Fourth International Conference on Lan-guage Resources and Evaluation, Lisbon, Portugal, 26–28 May 2004. Available from http://cwn.ling.sinica.edu.tw/churen/LA040524CT01.pdf

Zwier, L. J. (2002), Building Academic Vocabulary. Michigan: The University of Michigan Press.

Note

1 All cited internet sources were correct as of 2 August 2009.

Author index

Aarts, J. 35, 143Ädel, A. 69, 72, 150, 157, 161Aijmer, K. 152, 157, 176Altenberg, B. 83, 106, 121, 126, 150,

152, 180 Archer, D. 42, 45Aston, G. 73

Baayen, R. H. 145 Bahns, J. 204 Bailey, S. 206Baker, M. 17, 19, 20, 21, 22, 24Baker, P. 48Barkema, H. 84, 100 Barkhuizen, G. 67Bartning, I. 79Bauer, L. 12Bazerman, C. 72Beheydt, L. 14, 27Bestgen, Y. 62Bhatia, V. 26Biber, D. 2, 29, 55, 83, 122, 137, 143,

179, 211, 237n. 3Billuroglu, A. 16Biskup, D. 185 Bley-Vroman, R. 70Bourigault, D. 237n. 7 (Ch. 4)Bowker, L. 35, 206 Brill, E. 37Buker, S. 81 Burger, H. 83, 121 Burnard, L. 73

Campion, M. E. 11Candlin, C. 206 Carter, R. 11, 23, 85, 207, 238n. 3Celce-Murcia, M. 169

Charles, M. 214 Chen, C. W. 126, 152, 174Chung, T. 14, 18Clear, J. 102 Cohen, A. D. 18Coltier, D. 88 Connor, U. 152, 190 Conrad, S. 59, 83, 85, 121, 179, 180 Cook, G. 207 Corson, D. 13Cortes, V. 1Cowan, J. R. 17, 18Cowie, A. P. 213 Coxhead, A. 3, 5, 9, 10, 11, 12, 13, 16, 17,

20, 21, 25, 27, 28, 31, 34, 44, 63, 82, 122, 212

Crewe, W. 169, 174, 175, 176, 193, 201 Curado Fuentes, A. 46, 213 Cutting, J. 72

Davies, A. 71De Bot, K. 239n. 3 Dechert, H. 155, 168 De Cock, S. 30, 72, 86, 121, 157 DeRose, S. 37Dudley-Evans, T. 211

Eldridge, J. 26, 214 Elley, W. B. 11Ellis, R. 67Engels, L. K. 11Evans, S. 1Evert, S. 75, 76, 78

Farrell, P. 17, 18, 20Farrow, M. 48, 50, 62Feak, C. B. 24

Note: Page numbers in italics denote illustrations.

258 Author index

Field, Y. 171, 177, 193 Firth, A. 70Fisher, D. 204 Flowerdew, J. 2, 60, 61, 172, 178, 201,

214Flowerdew, L. 23, 199, 201, 204Francis, G. 22, 23, 59, 235n. 2 (Ch. 1)

Garside, R. 37, 38, 39Ghadessy, M. 11Gilquin, G. 1, 7, 70, 71, 151, 153, 195,

197, 207, 207, 208, 209, 210, 225, 238n. 5

Gläser, R. 237n. 7 (Ch. 4)Gledhill, C. 83, 102, 119, 123, 161,

236n. 4, 238n. 9Goodman, A. 17Granger, S. 4, 26, 32, 65, 67, 68, 70, 71,

72, 84, 100, 102, 118, 122, 123, 126, 143, 145, 150, 151, 152, 155, 157, 168, 169, 170, 177, 179, 182, 184, 185, 194, 197, 202, 204, 206, 213, 214, 215, 216, 236n. 1, 236n. 2, 238n. 9

Green, C. 1Gregg, K. R. 218Gries, S. 48, 50Groom, N. 215

Halliday, M. 203Hamp-Lyons, L. 206Hanciog lu, N. 15, 16, 27, 63, 212Harris, S. 22Harris Leonhard, B. 85Hasan, R. 203Hasselgren, A. 147Heasley, B. 206Heatley, A. 44Hegelheimer, V. 203Hinkel, E. 1, 3, 33, 59, 148Hirsh, D. 10, 34Hoey, M. 23, 26, 192, 216, 217Hoffmann, S. 75, 76, 86Hogue, A. 85Howarth, P. 119, 165, 217Huckin, T. 26, 214Hunston, S. 118

Huntley, H. 9, 11, 16, 82Hwang, K. 11, 13, 14Hyland, K. 3, 24, 25, 26, 31, 32, 72, 90,

92, 93, 99, 147, 157, 189, 201, 211, 214

Ide, N. 37Ivanic, R. 235n. 2 (Ch. 1)

Jarvis, S. 4, 182, 183, 184, 185, 197, 216, 238n. 13

Johansson, S. 31Johns, T. 214Jordan, M. P. 23, 235n. 2 (Ch. 1)Jordan, R. R. 1, 81, 82, 85, 201, 202,

203, 211Juilland, A. 50

Kamimoto, T. 239n. 2Katz, S. 48, 236n. 8Kellerman, E. 197King, P. 206Kosem, I. 62Krishnamurthy, R. 62Kroll, B. 69

Lake, J. 169, 170, 201Lakshmanan, U. 70, 71Larsen-Freeman, D. 169Laruelle, P. 91Laufer, B. 10Lee, D. 73, 74, 132Leech G. 11, 34, 35, 70, 72Lennon, P. 165, 168Li, E. S.-L. 18Liou, H.-C. 206Lonon Blanton, L. 85Lorenz, G. 72, 101, 143, 146, 150, 152,

157, 169, 173, 177, 193Luzón Marco, M. J. 22, 83, 137Lynn, R. W. 11

Major, M. 11Martin, A. 19, 20, 21, 27Martínez, I. 15, 27, 34, 82, 212McCarthy, M. 9, 23, 211, 235n. 2

(Ch.1), 238n. 3,

Author index 259

McEnery, A. 30, 35, 76Mel’cuk, I. 83Meunier, F. 143, 150Meyer, P. G. 24, 27Miller, J. 237n3Milton, J. 72, 147, 157, 179, 201, 202,

203, 206, 213Moon, R. 121Mudraya, O. 13, 14, 16, 17, 19, 31, 34Mukherjee, J. 70, 71, 160Müller, S. 237,

Narita, M. 126, 152, 174, 177, 179, 202Nation, I. S. P. 12Nation, P. 1, 3, 10, 11, 13, 14, 16, 17, 18,

23, 26, 44, 82, 185Neff, J. 73, 75, 152, 157, 194, 195Neff van Aertselaer, J. 73Nelson, M. 46Nesi, H. 31, 32, 33Nesselhauf, N. 73, 78, 101, 164, 166,

185Neufeld, S. D. 16

Oakes, M. P. 48, 50, 62, 76Oakey, D. 213Obenda, D. 82O’Dell, F. 9Odlin, T. 185, 204Osborne, J. 238n. 12Oshima, A. 85

Paquot, M. 15, 26, 36, 62, 84, 100, 118, 122, 123, 135, 151, 153, 157, 168, 190, 195, 197, 204, 213, 214, 238nn. 9,13, 239n. 6

Partington, A. 15Pavlenko, A. 185Pawley, A. 71Payne, E. 17Pearson, J. 35Pecman, M. 119Pemberton, R. 18Perdue, C. 192Petch-Tyson, S. 142, 145, 157Piller, I. 71Praninskas, J. 11

Quirk, R. 179

Rayson, P. 29, 30, 37, 38, 43, 47, 50, 61, 76, 145, 150

Renouf, A. 102Reynolds, D. W. 1Ringbom, H. 182, 185, 192Rodriguez, E. C. 50Rohrback, J.-M. 160Römer, U. 85Ruetten, M. 85Rumisek, L. 85Rundell, M. 201, 207

St Johns, M. J. 211Saville-Troike, M. 26Scarcella, R. C. 13Schleppegrell, M. J. 180Schmid, H.-J. 235n. 2 (Ch.1)Schmitt, D. 11, 16Schmitt, N. 11, 16Scott, M. 2, 45, 46, 47, 48, 69, 236n. 9Seale, C. 46Selinker, L. 70, 71, 182, 183, 239n. 2Shaw, P. 69Siegel, M. 237n. 3Siepmann, D. 82, 88, 101, 107, 126Sinclair, J. M. 2, 26, 35, 82, 101, 102,

118Smith, N. 35, 37, 39Soler, V. 118Stein, G. 16, 235n. 1 (Ch.1)Strevens, P. 13Stubbs, M. 10Sugiura, M. 126, 152, 174, 177, 179,

203Summers, D. 207Sutarsyah, C. 26Swales, J. M. 24, 31, 61, 86, 92, 132, 189,

207Swallow, H. 185Syder, F. H. 71

Tan, M. 71Tankó, G. 147Tapper, M. 126, 150, 152Thomas, J. 42

260 Author index

Thompson, G. 118Thompson, P. 255n. 2Thurstun, J. 206Tognini-Bonelli, E. 30, 35, 36, 118Tribble, C. 46, 62, 69Trimble, L. 18, 20, 21Tse, P. 3, 25, 26, 92Tseng, Y.-C. 206Tutin, A. 118Tyson, S. 126, 152, 169, 170, 177, 184

Van Roey, J. 185Vassileva, I. 152Voutilainen, A. 38

Wagner, J. 70Wang, J. 34Wang, K. 26Ward, J. W. 16, 34Waring, R. 1, 11

Weinert, R. 237n. 3Weissberg, R. 80West, M. 10, 11, 12, 15, 27, 44, 60Widdowson, H. G. 22, 61, 207, 212Wilkins, D. A. 81Wilson, A. 42Winter, E. 22Wray, A. 86

Xue, G. 11, 16

Yang, H. 13, 14, 17Yip, L. M. O. 171, 177, 193

Zamel, V. 176, 201Zemach, D. 85Zhang, H. 50Zhang, M. 177, 194Zimmerman, C. B. 13Zwier, L. J. 24, 85

Subject index

Academic Corpus 11–12academic discourse 2, 3–4, 15, 24,

27–8, 31, 40, 63, 102, 119, 214, 217

academic discourse community 31Academic Keyword List (AKL) 5, 7, 55–61,

122academic vocabulary and 60automatic semantic analysis of 82–3distribution, in ICLE 143exemplifi catory discourse markers

in 88grammatical distribution categories

in 55need for concordancing in 61need for pedagogic mediation 61, 82nouns and 56overused and underused clusters

with 156and rhetorical functions 81–7words distribution, in GSL and

AWL 60words, overused and underused in

ICLE 144academic literacy 231academic vocabulary 7

vs. core vocabulary and technical terms 10–13

defi nition of 212fuzzy vocabulary categories 13–17meaning of 9, 24–5, 28and sub-technical vocabulary 17–21

Academic Word List (AWL) 3, 5, 11, 12, 15, 16, 17, 20, 25, 27, 34, 59, 60, 63, 82, 122, 212

activity verbs 59

adjectives 101, 118in the Academic Keyword List 57as co-occurrents of academic

nouns 100, 133, 167potential academic 57, 59

adverbials/adverbs 93, 213in the Academic Keyword List 58mono-lexemic 91multiword linking 121potential academic 58, 59semantic misuse and 139–40sentence position 179

annotation 34–6, 37–42, 43part-of-speech annotation 30, 34–5,

36, 37, 38, 40, 41, 43semantic annotation 35, 43–4, 53

association measures 76, 101attitudinal formulae 84, 122, 123automatic semantic analysis, of

AKL 82–3

Baby BNC Academic corpus (B-BNC) 31, 32, 47

bilingual dictionaries 204Billuroglu-Neufeld-List (BNL) 16blend 168BNC-AC-HUM 75, 78, 90, 95, 100, 102

comparing and contrasting in 112–14

expressing cause and effect in 110–11, 114–18

expressing a concession in 109expressing possibility and

certainty 118–20reformulating in 109see also British National Corpus (BNC)

Note: Page numbers in italics denote illustrations.

262 Subject index

BNCweb 75, 76, 77booster 157British Academic Written Corpus 217British Academic Written English

(BAWE) Pilot Corpus 32–3, 34, 235n. 1,2 (ch2)

British National Corpus (BNC) 4, 16, 17, 31, 67, 73, 74, 75–6, 77, 78, 79, 84, 95, 102, 125, 130, 132, 133, 134, 146, 152, 207

Baby BNC Academic Corpus (B-BNC) 31

BNC-AC 78BNC-AC-HUM see BNC-AC-HUMIndex 73–5mark-ups 73

BROWN corpus 47burstiness 48, 236n. 8

Cambridge Advanced Learner’s Dictionary 207

cataphoric markers 90–1cause and effect markers 87, 210, 219–25

in BNC-AC-HUM 110–11EFL learners’ use of 146–8, 147

Centre for English Corpus Linguistics (CECL) 207

ClairefontaineLes fi ches essentielles du Baccalauréat en

anglais 160CLAWS 37–42, 59code gloss 90, 93, 188–9cohesion 22, 123, 211, 213, 238n. 3

advance and retrospective labelling 22

grammatical 203lexical 18, 22, 148, 203, 213non-technical words 18textual 123, 148

colligation 168colligational errors 166, 168collocation 23–4, 76, 77, 100, 102, 118,

119, 161, 164, 165, 192, 204, 217, 238n. 9

collocational framework 102collocational overlap 165

Common European Framework of Reference for Languages (CEF) 236n. 1 (ch3)

communicative phrasemes 84, 121–2 comparative fallacy 70, 71comparison and contrast markers 87,

202, 208, 226–34in BNC-AC-HUM 112–14EFL learners’ use of 148, 149

conceptual frequency 86concession markers 87

in BNC-AC-HUM 109conjunctions

complex 40, 59, 84, 119, 120overuse of 146sentence initial position of 194

connectors 140, 169, 170, 172, 174–6, 177, 178–9, 178, 180, 181, 193, 201–2, 202, 203, 215, 238nn. 1–3

medial position for 180overuse of 201–2semantic misuse 201sentence 22sentence position 141, 174–82, 193–4,

203Constituent Likelihood Automatic Word-

tagging System (CLAWS) 37–42, 59content words 10, 22, 102, 236n. 8Contrastive Interlanguage Analysis

(CIA) 4, 65, 70, 79, 85, 87, 215contrastive rhetoric 2, 152control corpus 67, 70, 71, 73co-occurrence 37, 76, 78, 95, 96, 99,

100, 101, 114, 115–17, 119–20, 133–4, 137, 160, 162, 166, 167, 193

preferred co-occurrences in EFL writing 160–8

core vocabulary 3, 4, 10–11, 15Corpus de Dissertations Françaises

(CODIF) 184, 186, 188, 190, 191

corpus-based approach 2, 3, 29, 30, 31, 61, 87, 106, 150, 216, 238n. 6

corpus-driven approach 29, 30, 35Corpus Query Processor (CQP) 75, 76co-text 2, 22, 172, 193, 203

Subject index 263

data-driven approach 4, 29, 30, 36data-driven learning 214derivation 12, 13, 16, 19developmental factor 4, 125, 181, 183,

197, 216directives

see imperativesdiscipline 3, 18, 25, 26, 27, 33, 55, 211,

213, 214, 215, 235n. 2see also knowledge domain

discourse marker 88, 126, 138cataphoric marker 90, 91endophoric marker 91, 92, 93, 98,

99, 119engagement marker 92, 93

discourse-organizing vocabulary 9, 23dispersion see distributiondistribution 29, 45, 50–3, 55, 60, 78, 93,

94, 95, 103, 132, 135, 143ditto-tag 38, 40, 44, 59document-level burstiness 236n. 8

EAP material design 221EAP teaching 26, 213, 214endophoric markers 91, 93, 98,

99, 119English as a Second Language

(ESL) 33, 148, 180, 204English for Academic Purposes

(EAP) 15, 26, 27, 62, 85English for Specifi c Purposes (ESP) 9,

217epistemic modifi ers 147evenness of distribution see distribution

and Juilland’s D statistical coeffi cient

exemplifi ers 85–8in BNC-AC-HUM 88–108learners’ use of 125–42, 189–91

‘extended units of meaning’ 118

fi ction 46, 47fi eld approach 82fi xed phrase 23–4, 121FLOB corpus 47formulae 12, 47, 61, 84

attitudinal formulae 84, 122, 123textual formulae 121

free combination 100, 101, 123FROWN corpus 47functional-product approach 82functional syllabus 81–2, 83function words 10, 45, 102, 143fuzzy vocabulary categories 13–17

General Service List of English Words (GSL) 10–11, 12, 14, 15, 16, 18, 27, 44, 59, 60

general service word 16, 20, 63, 212genre 2, 73, 74, 75, 93, 94, 95, 102, 103,

130, 131, 132, 145global keywords 48grammatical cohesion see cohesiongraphemic words 40

hedge 2, 157high-frequency word 5, 10, 14, 15, 20,

27, 28, 37, 45, 60, 212homographs 25, 37

idiom 3, 23, 44, 71, 84, 119, 123, 237n. 7illocutionary nouns 23imperatives

in academic writing 93, 107as directives with rhetorical

purpose 92fi rst person plural 136, 137, 188,

189–90, 191, 192, 204second person 91–2, 98

IMS Open Corpus Workbench 75International Corpus of Learner

English (ICLE) 4, 5, 65, 67–9, 71, 72–3, 75, 78, 84, 86, 125, 236n. 3

Juilland’s D statistical coeffi cient 50–3

keyness 4, 45, 46–8, 55, 62, 159, 212keyword 30, 46–8, 47, 55, 61, 62, 86, 159

global keyword 48local keyword 48negative keyword 47, 86positive keyword 47, 86

264 Subject index

keyword analysis see keynessknowledge domain 31, 32

L1 frequency 185, 190, 218, 239n2L1-induced factor see transferL1 infl uence 182–92

Jarvis’s unifi ed framework 182–4labeling 22–4, 73, 173, 203

advance labeling 96, 97retrospective labeling 22, 96, 98

labels 22–3semantic misuse 172–3

language-activity nouns 23learners’ dictionary 206–7lexical bundle 69, 118, 177lexical cohesion see cohesionlexical extension 213lexical priming see priminglexical repertoire 3, 4, 5, 9, 125,

142–50, 192–3lexical teddy bear 147lexical transfer see transferlexico-grammar 26, 71, 85, 105, 118,

123, 137, 138, 161, 164, 186, 193, 197, 214, 215

lexico-grammatical error 155, 197, 215

linking word 3, 84, 121, 160, 177, 179, 180, 181, 204

LOB corpus 47local keywords 48logico-semantic relationship verbs 59log-likelihood 47, 48, 62, 76, 78, 125log-likelihood calculator, UCREL 125LONGDALE project 239n. 1Longman Dictionary of Contemporary

English 206Louvain Corpus of Native Speaker

Essays (LOCNESS) 32, 194, 195

Macmillan English Dictionary for Advanced Learners 7, 201, 207, 238n. 6

meaning 10, 13, 14, 18–19, 20, 25, 35, 49, 52, 100–1, 102, 118, 185

delexical meaning 118fi gurative meaning 119over-extension 146, 170, 184

non-technical meaning 18, 19technical meaning 18, 19, 52

mental process nouns 23mental verbs 59metadiscourse 24, 90, 93, 99, 118, 161metalinguistic labels 23, 59Michigan Corpus of Upper-level

Student Papers 217Micro-Concord Corpus Collection

B (MC) 31, 32monolingual learners’ dictionary

(MLD) 206–7morphosyntactic annotation 34–5multiword expression 37, 44, 53, 59, 60,

83, 88, 121, 184, 185, 197

native control corpus 70native speaker norm, corpus-

approximation to 70, 71native student writing 72negative keywords 86n-gram 69non-technical term 17, 18non-technical words 18–19, 24nouns 22–3, 108, 138

in the Academic Keyword List 56adjectives as co-occurrents of

academic 100, 133, 167verbs as co-occurrents of

academic 95–9, 134, 137, 162–3novice writing 1, 4, 31, 65, 85, 95–6,

152, 190, 194, 195–6, 197, 206, 215, 217

nuclear vocabulary 9, 10, 14, 20nuclear words and pragmatic

neutrality 14

organizational function see rhetorical function

overuse 86, 126, 129, 130, 140, 143, 144, 145–6, 148, 150, 151, 152, 155, 156, 157, 158, 159, 160, 194, 195, 201, 237n. 5

Oxford English Dictionary (OED) 20

paraphrasing and clarifying see reformulation markers

Subject index 265

parsing see syntactic annotationpart-of-speech (POS) tagging 34–5, 37–8

see also annotationpedagogic mediation 61, 82, 212Perl program 48personal metadiscourse 161personal pronoun 98, 136, 138, 161, 164phraseme 83, 84, 93, 94, 106, 118, 119,

121, 137, 138, 148, 164, 185, 211, 214

communicative phraseme 84, 122, 157mono-lexemic phraseme 88, 90, 106,

160, 161referential phraseme 84, 119structural phraseme 121textual phraseme 84, 90, 94, 95, 118,

120, 121, 123, 161phraseological accent 83phraseological analysis 83–4, 90, 93,

123, 213phraseological ‘cascade’ 161, 188phraseological competence 217phraseological infelicity 155phraseology of rhetorical functions 65,

76, 78, 81, 102, 108, 109, 110–11, 112–14, 115–17, 119–20, 121, 123, 132, 154, 166, 213, 217

frequency-based approach to 122positive keywords 86potential academic words 29, 44–55preferred co-occurrence 2, 160, 192, 193preferred ways of saying things 83, 123,

166, 193preposition 97, 101, 108, 143

complex 40, 41, 84, 90, 120, 139, 144priming 192, 197, 203, 216, 217

mental 192transfer of 192, 203

procedural vocabulary 22, 211production 1, 4, 9, 15–16, 33, 68, 69,

70, 142, 155, 212pronoun 23, 101

demonstrative 96as exemplifi ed item 97impersonal 168personal 98, 136, 138, 161, 164third person 157

range 1, 4, 11, 12, 13, 14, 30, 45, 48–50, 62, 212

Range corpus analysis program 44reception 1, 15, 17, 73reference corpus 46, 47, 62, 73, 135referential phrasemes 84, 119reformulation markers 87, 108, 139, 209

in BNC-AC-HUM 109register awareness 5, 125, 132, 142,

150–2, 193, 208, 215reporting verbs 59retrospective labelling 22rhemes 97, 98, 121, 135rhetorical function 5, 7, 9, 10, 20, 22,

24, 26, 27, 60, 61, 63, 81, 125, 141, 142, 148, 150, 151, 155, 161, 184, 188, 190, 192–3, 197, 202, 203, 207, 213, 214, 215

rhetorical overstatement 176Robert & Collins CD-Rom 204, 205

semantic annotation 35semantic misuse 5, 139–40, 145,

168–74, 170, 172, 193, 201semantic tagging 43semantic transfer see transfersemi-technical vocabulary 17sentence connectors 22sentence stem 97, 99, 106, 114, 118,

121, 122, 135specialised non-technical lexis 17, 18speech 2, 62, 71, 84, 95, 131, 136, 145,

151, 152, 153, 177, 190, 195, 197, 213

speech-like lexical item 151–2, 153, 195

spoken frequency counts 145Student Writing Corpus 32, 33sub-technical vocabulary 17–21, 21, 22,

24syntactic annotation 35

tagging see annotationteaching material 82, 85, 148, 160, 169,

178, 180, 206, 238n. 1technical terms 3, 9, 13, 14, 18technical vocabulary 13, 17–21

266 Subject index

text coverage 10, 11, 15, 16, 236n. 4text nouns 23textual formulae 121textual phrasemes 84, 93, 94, 118, 119,

121, 123, 160, 161textual sentence stems 97, 121, 135tokenisation 38transfer 4, 168, 171, 181, 190, 191, 194,

197, 203, 216, 218lexical transfer 216transfer effects 182–5transfer of form 185transfer of form/meaning

mapping 185, 216transfer of L1 frequency 185, 190–1transfer of meaning 185transfer of the phraseological

environment 185, 197transfer of primings 192, 197, 203,

216transfer of style and register 185,

188, 192transfer of training 144, 182, 194, 201–3transfer of use 185

transfer-related factor see transfertypicality 106, 107

underuse 86, 126, 130, 131, 135, 137, 143, 144, 145, 146, 147, 148, 149, 155, 156, 157, 158, 159

underused words see negative keywordsUniversity Word List 16USAS (UCREL Semantic Analysis

System) 37, 42–4

Varieties of English for Specifi c Purposes dAtabase (VESPA) 217

verbs 24, 26–7, 59, 91, 118–20, 136, 157–8activity 59co-occurrents

of academic nouns 95–9, 134, 137, 162–3

forming rhemes with noun 98lexical 36linking 59mental 59potential academic 57reporting 59in sentence-initial infi nitive

clauses 138Vocabulary 3 items 22

Web Vocab Profi le 59, 60within-document burstiness 236n. 8Wmatrix 36–7, 53word families 12, 16, 17, 45

in AWL 12, 16–17, 17in GSL 11

word form 12, 17, 34, 36, 39, 102, 157word list 2, 16, 27, 40, 46

in the Academic Keyword List 57Word Smith Tools 2, 47, 48, 49, 51, 69

Concord tool 69Detailed Consistency Analysis 51Keywords option 155WordList option 49

‘you-know-it-when-you-see-it’ syndrome 182

Academic Vocabulary in Learner Writing

Documents

Transcript of Academic Vocabulary in Learner Writing