Corpora and Language Teaching (Studies in Corpus Linguistics)

Corpora and Language Teaching Studies in Corpus Linguistics (SCL) SCL focuses on the use of corpora throughout langua...

Author: Karin Aijmer

308 downloads 2064 Views 4MB Size Report

This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!
Report copyright / DMCA form

DOWNLOAD PDF

Corpora and Language Teaching

Studies in Corpus Linguistics (SCL) SCL focuses on the use of corpora throughout language study, the development of a quantitative approach to linguistics, the design and use of new tools for processing language texts, and the theoretical implications of a data-rich discipline.

General Editor

Consulting Editor

Elena Tognini-Bonelli

Wolfgang Teubert

The Tuscan Word Center/ The University of Siena

Advisory Board Michael Barlow

Graeme Kennedy

Douglas Biber

Geoffrey N. Leech

Marina Bondi

Anna Mauranen

Christopher S. Butler

Ute Römer

Sylviane Granger

Michaela Mahlberg

M.A.K. Halliday

Jan Svartvik

Susan Hunston

John M. Swales

Stig Johansson

Yang Huizhong

University of Auckland Northern Arizona University University of Modena and Reggio Emilia University of Wales, Swansea University of Louvain University of Sydney University of Birmingham Oslo University

Volume 33 Corpora and Language Teaching Edited by Karin Aijmer

Victoria University of Wellington University of Lancaster University of Helsinki University of Hannover University of Liverpool University of Lund University of Michigan Jiao Tong University, Shanghai


Edited by

Karin Aijmer University of Gothenburg

John Benjamins Publishing Company Amsterdam / Philadelphia

8

TM

The paper used in this publication meets the minimum requirements of American National Standard for Information Sciences – Permanence of Paper for Printed Library Materials, ansi z39.48-1984.

Library of Congress Cataloging-in-Publication Data Corpora and language teaching / edited by Karin Aijmer. p. cm. (Studies in Corpus Linguistics, issn 1388-0373 ; v. 33) Includes bibliographical references and index. 1. Language and languages--Study and teaching. 2. Grammar, Comparative and general-Study and teaching. 3. Corpora (Linguistics) 4. Second language acquisition. I. Aijmer, Karin. P53.C67 2009 418'.0071--dc22 isbn 978 90 272 2307 4 (Hb; alk. paper)

2008045267

© 2009 – John Benjamins B.V. No part of this book may be reproduced in any form, by print, photoprint, microfilm, or any other means, without written permission from the publisher. John Benjamins Publishing Co. · P.O. Box 36224 · 1020 me Amsterdam · The Netherlands John Benjamins North America · P.O. Box 27519 · Philadelphia pa 19118-0519 · usa

Table of contents

List of contributors Introduction: Corpora and language teaching Karin Aijmer

vii 1

Part I. Corpora and second-language acquisition The contribution of learner corpora to second language acquisition and foreign language teaching: A critical evaluation Sylviane Granger Some thoughts on corpora and second-language acquisition Stig Johansson

13 33

Part II. The direct corpus approach Who benefits from learning how to use corpora? Solveig Granath

47

Oslo Interactive English: Corpus-driven exercises on the Web Signe Oksefjell Ebeling

67

Corpus research and practice: What help do teachers need and what can we offer? Ute Römer

83

Part III. The indirect corpus approach Themes in Swedish advanced learners’ writing in English Jennifer Herriman and Mia Boström Aronsson

101

vi


Thematic choice and expressions of stance in English argumentative texts by Norwegian learners Hilde Hasselgård

121

The usefulness of corpus-based descriptions of English for learners: The case of relative frequency Susan Hunston

141

Part IV. New types of corpora Income/interest/net: Using internal criteria to determine the aboutness of a text Winnie Cheng

157

New types of corpora for new educational challenges: Collecting, annotating and exploiting a corpus of textbook material Fanny Meunier and Céline Gouverneur

179

The grammar of conversation in advanced spoken learner English: Learner corpus data and language-pedagogical implications Joybrato Mukherjee

203

Index

231

List of contributors

Karin Aijmer English Department Göteborg University Box 200 40530 Göteborg Sweden [email protected] Mia Boström Aronsson Skövde University College Högskolevägen Box 408 541 28 Skövde Winnie Cheng Research Centre for Professional Communication in English Department of English The Hong Kong Polytechnic University Hung Hom Hong Kong [email protected] Signe Oksefjell Ebeling Department of Literature, Area Studies, and European Languages University of Oslo P.O. Box 1003, Blindern 0315 OSLO Norway [email protected]

Cécile Gouverneur Centre for English Corpus Linguistics Université Catholique de Louvain Collège Erasme Place Blaise Pascal 1 1348 Louvain-la-neuve Belgium [email protected] Solveig Granath English Department Karlstad University 65188 Karlstad Sweden [email protected] Sylviane Granger Centre for English Corpus Linguistics Université Catholique de Louvain Collège Erasme Place Blaise Pascal 1 1348 Louvain-la-neuve Belgium [email protected] Hilde Hasselgård Department of Literature, Area Studies, and European Languages University of Oslo P.O. Box 1003, Blindern 0315 OSLO Norway [email protected]

viii List of contributors

Jennifer Herriman English Department Göteborg University Box 200 40530 Göteborg Sweden [email protected] Susan Hunston Department of English The University of Birmingham Edgbaston Birmingham B15 2TT United Kingdom [email protected] Stig Johansson Department of Literature, Area Studies, and European Languages University of Oslo P.O. Box 1003, Blindern 0315 OSLO Norway [email protected]

Fanny Meunier Centre for English Corpus Linguistics Université Catholique de Louvain Collège Erasme Place Blaise Pascal 1 1348 Louvain-la-neuve Belgium [email protected] Joybrato Mukherjee Justus Liebig University Giessen Department of English Chair of English Linguistics Otto-Behaghel-Str. 10B 35394 Giessen Germany [email protected] Ute Römer University of Michigan English Language Institute 500 E. Washington Street Ann Arbor, MI 48104-2028 USA [email protected]

Introduction Corpora and language teaching Karin Aijmer 1.

Introduction

Corpora have changed our views on language and language use and we can also expect to find them in the class-room. It is not only ‘raw’ corpora which are of interest but corpora come with user-friendly programs and software which makes them suitable for the use by learners. However it is clear that there are also problems and that we do not know enough about how learners and teachers experience the use of corpora in the classroom. When should corpora be used as part of the teaching of a language? How should they be used? What should be the proper balance between the corpus-based approach and more traditional classroom methods? Are corpora good for all kinds of students? These questions are of great concern to teachers and scholars who share an interest in corpora and dedication to using corpora in the classroom. In December 2005 a symposium was organised at the University of Gothenburg in order to discuss such questions. A number of scholars with extensive experience of using corpora in their teaching and for applied linguistics research were invited to review the state of the art and to discuss the role and effectiveness of corpora and corpus-linguistic techniques for language teaching. The volume Corpora and Language Teaching contains a selection of the contributions from the symposium as well as some commissioned articles. Corpus linguists are generally enthusiastic about what they have to offer the teaching profession. However the use of corpora in the EFL classroom is a rare occurrence and teachers are still unwilling to or lack the skill to use corpora as an aid to get new insights into English. On a pessimistic note Mukherjee and Rohrbach (2006: 205) write that ‘we have the impression that in EFL countries like Germany there is a widening gap and a widening lag between on-going and intensive corpus-linguistic research on the one hand and classroom teaching on the other’. The problem is how to reach students and teachers with information about corpora and what they can do. Although courses in corpus linguistics are

Karin Aijmer

sometimes included in the university curriculum the direct exploitation of corpora in the EFL classroom is unusual and the impact of corpora on syllabus and materials design has been slight. The articles in this volume are geared towards the applications of corpora in the classroom and for pedagogical research. They also deal with broader issues such as the relationship between corpora and second language acquisition or foreign language teaching. Applied corpus linguists and the average EFL teacher have different perspectives on language teaching and it is important to clarify how these perspectives differ. For example, teachers (and learners) look for simple answers to grammatical problems in terms of what is right and wrong and shy away from the fuzzy picture of language as used in the corpus concordance. Discussions of the pedagogical implications of corpora can take two forms. They focus on the use of corpora in the classroom. Moreover they deal with the use of corpora for applied linguistics research in particular the use of learner corpora to get a better picture of how advanced learners write and speak.

2.

The use of corpora and second language acquisition

Two of the papers deal with the relationship between corpora and second language acquisition. In Sylviane Granger’s state of the art paper she discusses several of the divisive issues in the field of learner corpora and in particular how we can establish a better link to second language acquisition research and more cooperation with second language acquisition practitioners. Corpus linguistics and second language acquisition need each other and learner corpora could be a meeting place for SLA practitioners and linguists interested in learner corpora. There are several reasons why learner corpus research needs to join forces with SLA. As Granger writes (this work) ‘a wide range of social, cognitive, and psychological factors that play a role in language learning have been extensively studied in SLA and familiarity with SLA findings will greatly help LC (learner corpus analysis) with a focus on learner production’. As a result there are now clear signs of the two fields coming closer together. Both sides have been quick to recognize the advantage of broadening the empirical basis for SLA analysis. It is for instance clear that learner corpora provide a wealth of empirical material making it possible to examine a number of different variables which have an effect on learner output. Differences between learners and native speakers can for example reflect a transfer effect which can be traced back to contrastive differences and be studied on the basis of multilingual corpora. There is a mutual give and take. If there are differences between the target and source language shown by the translations we can hypothesise that these will affect the way learners use L1. On the other hand,

Introduction

deviations from the native speaker norm can at least sometimes be explained as transfer and be traced back to the contrastive analysis. The description of the results of learner corpus investigations can lead to improved language teaching materials which pay more attention to forms and structures which pose special problems for learners. The priorities for the future include both a wider range of learner corpora, standardized annotation and ‘realistic’ learner tools. Corpus annotation is important and PoS tagging is available for the written corpus. Stig Johansson stresses the importance of arranging controlled experiments where teaching based on different theories of foreign language teaching can be compared. The starting-point for Johansson’s paper is an early experiment on explicit and implicit teaching methods which showed that the implicit methods were most successful at least with adult learners. The experiment also raised some questions on corpora and second language acquisition. To what extent is the use of corpora in language teaching tested experimentally? To what extent can the use of corpora be grounded in theories of language acquisition? The experiment reported by Johansson supports the idea that the learner is involved in hypothesis formation and hypothesis testing. There is just a short step from this to the idea that the same processes are involved when the corpus is used for language teaching. The frequency information in the corpus also supports a view of language acquisition where the learner goes to work on finding out the lexico-grammatical patterns helped by repetition and entrenchments of form-meaning links.

3.

The use of corpora in the classroom – the learner as a researcher

The classroom provides the framework within which we can expect a lot of direct interaction between learner and corpus. The direct or data-driven application of corpora in the classroom implies that learners get their hands on authentic corpus material and are encouraged to discover things about language without any previous preconception about what they will find (Johns 1991; Bernardini 2004). The corpora can for instance be used to provide concordances or to select examples for learning activities. The articles by Granath, Ebeling and Römer in this volume give examples of different types of corpus-based learning in the classroom. Solveig Granath shows how corpora can be an integral part of courses in grammar and in spoken and written proficiency. Corpora can for example be used to create exercises, demonstrate variation in grammar, show how syntactic structures can signal differences in meaning, to discuss near-synonyms and collocations. The corpus techniques are especially relevant when the grammatical rules are very general and do not capture the way in which language is actually used. By

Karin Aijmer

consulting the corpus students and teachers can for instance get a more varied picture of the use of concord with different types of collective nouns than you get by consulting a grammar. Corpora can also find an answer to ‘what teachers always wanted to know’ and give informed answers to student questions (cf. Tsui 2004). Students sometimes ask questions about phenomena which are not mentioned in the grammar book and where corpora need to be consulted. In Granath’s experience it takes time for students to become skilful corpus users. Many learners were unused to the inductive methods and therefore found this way of working difficult. It is also possible that corpus activities do not suit all types of learners (cf. Estling Vannestål & Lindquist 2007). Signe Oksefjell Ebeling’s paper describes an interactive web-based learning platform at the University of Oslo (Oslo Interactive English or OIE) with the aim to encourage more flexible learning by means of Information and Communication Technology. The OIE is also intended to serve as an introductory course to the use of corpora. The corpus which is used together with OIE consists of 7 million words of fiction and non-fiction and is a slightly slimmed version of the Longman/Lancaster English Language Corpus. The corpus is the basis for a large number of exercises for example multiple choice tasks, gap-filling exercises, error correction and ‘open’ choices (the students are free to write full answers). By using corpus evidence the OIE makes students reflect on language and critically examine the rules of grammar books. The statistics on how often the interactive web platform had been used seemed to show that the OIE is popular among learners and that its popularity is increasing. However a closer look indicated there were many more people visiting the OIE’s web pages than actually doing the exercises. The way to improve this situation will be to integrate OIE into on-campus teaching. There are a number of useful corpora and corpus tools waiting to be used in the classroom but we need to know if they give the information teachers and students want and what they are looking for. Ute Römer reports on a survey among qualified English language teachers at secondary schools in Germany in order to learn more about the teachers’ working situation and to collect comments on their problems and experiences. Together with an experienced practising teacher Römer devised a questionnaire designed to find out for example if the existing teaching material and handbooks gave sufficient support to the teaching of vocabulary and grammar and what the teachers’ attitudes were towards corpora. Several informants pointed out that it took too long time to look up words in dictionaries and that they often wanted to consult a native speaker. However a majority of teachers did not see the consultation of one of the major language corpora as an alternative or supplement to the dictionary or grammar. The at-

Introduction

titudes towards existing coursebooks were often negative which suggests that this is an area where corpora can make a difference. Teachers in general thought that the existing course books offered too few exercises or lacked interesting, authentic material. This was especially the case for spoken language.

4.

The use of corpora for applied linguistics research

Learner corpora consist of the learners’ own written or spoken production. They have been mainly used for research (what Granger refers to as delayed pedagogical research). An impressive amount of research has been carried out on the basis of the International Corpus of Learner English (ICLE) (see the articles in Granger 1998 and 2002). The ICLE project incorporates a number of national groups with different L1s collecting their own corpora according to shared design criteria such as the level of the learners (Granger 1998, 2002). The studies under the umbrella of the ICLE project have paid special attention to advanced learners. Advanced learners make few morphosyntactic errors but they may use forms and structures in a non-native way. By comparing the essays produced within the ICLE Corpus with a native speaker norm we can discover subtle features such as overuse and underuse which account for the impression of near-nativeness of learners’ essays. In this volume we find two examples of how the study of thematic variation has pedagogical implications for improving advanced learners’ competence. Jennifer Herriman & Mia Boström Aronsson’s paper presents the findings from a comparison of how Swedish advanced learners of English and native speakers of English organize the information in argumentative writing using the Swedish component of the ICLE Corpus. The main focus is on the selection of theme and thematic variation. The comparison showed that the learners tended to thematize their opinions and attitudes to a much higher extent than native speakers. In particular learners overused I think and what I want to say. Swedish learners also overused clefts which were used to thematize new information and to express evaluative comments. The thematic choices made by learners may lead to a persuasive and emphatic style which is not characteristic of the argumentation of native speakers. The reasons for the overuse of certain types of theme include transfer of the native language structures or cultural conventions, general learner strategies such as the use of formulas and lack of knowledge about the conventions for argumentative writing in English. Hilde Hasselgård explores the extent to which Norwegian learners apply Norwegian patterns in their choice of thematic structure on the basis of the Norwegian component of ICLE. It is shown that word order patterns are transferred

Karin Aijmer

from Norwegian to English and that learners do not seem to have acquired the grammatical and stylistic norms in the relevant genres of English. In particular initial adverbials are overused in the Norwegian learner data compared to authentic English data. Norwegian learners overused extraposition but unlike the Swedish learners they didn’t overuse clefts. Like the Swedish learners they often referred to themselves using I think and related structures. Another similarity is the overuse of (some) disjuncts such as of course and probably. The conclusion we can draw from these two studies is that learners have a good mastery of thematic structures and of thematic variation but they do not know in what styles or registers it is appropriate to use them. Many of the features overused by non-native speakers were for example characteristic of speech rather than of writing. Other studies focus on learners’ problems in the area of phraseology and in particular how phraseology should be presented to learners. Susan Hunston discusses how information about corpus frequency can be linked to phraseology and how this information can be presented to learners. The suggested methodology (the ‘corpus-driven’ approach) gives priority to lexis and the phraseology associated with words. The focus of Hunston’s study is on multiword units which are often ranked in the same way as single words with regard to frequency. An alternative view is to recognize that not only overall frequency plays a role but the strength of the collocation between the elements in the units. This results in fairly long phrases representing information about what is relatively often said rather than information about what is grammatically correct. Also when we consider the different forms subsumed under a lemma the probability of occurrence is affected. What we find is semantic sequences where a word form can be related to the complementation pattern and modal meaning by means of the probability of occurrence of each linguistic feature. However such sequences are hardly useful for prescription in the classroom. On the other hand, this type of sequences are important in the devising of teaching materials focusing on the functions, vocabulary and grammar items most needed by learners. Corpora of learners’ production are now also collected by teachers for use in the classroom. The object of Winnie Cheng’s paper is to use such corpora to study the phraseology typical of the fields of economics and financial services in order to describe the content of the text in terms of ‘aboutness’. Aboutness is especially clear when a genre is domain-specific. Central to the study of ‘aboutness’ is the establishment of collocational profiles on the basis of keywords and their collocates. The text-specific patterns which are found are then compared and their ‘about distance’ can be calculated by comparing it with a reference corpus such as the British National Corpus. The search engine ConcGram@ makes it possible to handle not only contiguous word combinations but variations in the patterns with

Introduction

regard to position and constituency. The method and its findings have important implications both for ESP and LSP for example in raising language awareness and increasing knowledge about the ‘aboutness’ in discipline-specific discourse.

5.

New types of pedagogical corpora

5.1

Textbook corpora

Corpora and corpus-based research can have an impact on syllabus design and on the preparation of textbooks, dictionaries, grammars and course-books. Dictionaries are for example generally corpus-based and oriented towards the learners’ communicative needs and more and more grammars now base themselves on genuine corpus examples. In the new Cambridge Grammar of English (Carter & McCarthy 2006) all the examples represent natural English taken from a variety of written or spoken texts. However, with a few exceptions (Broadhead 2003; McCarthy et al. 2005) textbooks still shy away from corpora. This is shown by a survey of English for General Purposes textbooks carried out by Fanny Meunier and Céline Gouverneur. Meunier and Gouverneur have compiled a new type of textbook material corpus (the TeMa Corpus) which contains over 700,000 words of textbooks which are popular on the international ELT market. The corpus was tagged pedagogically with tags referring to the type of exercise. The corpus can for instance be used to provide a list of all the words/expressions practised in the exercises at one level and compare it with other levels. Another use is to investigate the type of metalanguage used in the textbooks to see if the terms are used consistently.

5.2 Spoken learner corpora Research on the basis of spoken corpora has shown that there is a ‘grammar of conversation’ with different forms and syntactic structures from what we find in written language. However the teaching of forms and structures which are typical of spoken language is still a neglected area. Textbook dialogues are generally stilted and written-like and lack the features which make texts come alive. Joybrato Mukherjee focuses on the importance of syntactic features of conversation in advanced German learners’ speech and discusses their language-pedagogical implications. The paper presents some findings from a number of case studies based on the German component of the Louvain International Database of Spoken English Interlanguage (LINDSEI). The first case study is concerned

Karin Aijmer

with the differences between spoken and written (learner) language with regard to the number, kind and range of collocations. It is shown that learners are more restricted with regard to the range of verb-noun collocations they use in speech than in writing. Moreover many of the collocations used by the (German) learners were deviant. Discourse markers such as you know, well, sort of are another relevant area for the teaching of spoken grammar. They occur frequently in spoken communication where they are used for interactional and interpersonal functions. Discourse markers are used by both learners and native speakers but they are used in different ways and with different frequencies (cf. also Hasselgren 2002; Müller 2005; Aijmer 2004). In the third case study the comparison involves spoken performance phenomena such as repetition or pauses which have important roles in speech production. The different trends which can be observed when comparing such phenomena in the speech of learners and native speakers show that the planning pressure is higher for learners and can explain why learners’ speech appears to be less fluent and spontaneous. The upshot of the case studies is that it is necessary to pay more attention in the classroom to forms and structures which are typical of spoken language. These forms include discourse markers but also preconstructed phrases that can help learners to become more fluent (cf. De Cock 2004; Rühlemann 2006). For spontaneous spoken language we can envisage various DDL (data driven learning) scenarios with the purpose to raise the learners’ awareness about how a particular marker such as you know or routinised phrases are used (cf. also Edmondson & House 1981). In order to become fluent speakers learners also need to practice dialogue techniques which force them to produce speech under real-time online production constraints.

6.

Avenues for the future

The picture of the future for corpora in teaching is bright although tempered by what we know about attitudes of teachers and learners. As Römer points out (in this volume and in Römer 2006), corpus linguists have a tough job to meet the challenges from teachers and students who are used to more traditional methods. Corpora draw attention to complex patterns and phraseology rather than regularities and supports the view of language learning as a complex process involving hypothesis formation and testing. Corpora have an obvious place in the classroom but cannot replace the teacher or language teaching. However the teacher has an important role to guide the stu-

Introduction

dents to the use of corpora in the classroom. The use of corpora should therefore have an important role in teacher education. Granath (this work) writes that ‘if training in how to use corpora were integrated into university level courses such as syntax, written proficiency and translation, in time it could become just as natural to consult a corpus as to look up an item in a dictionary or a grammar book’. Corpora provide a rich resource for the language teacher or applied linguist but they do not necessarily make life easier. In the words of Susan Hunston, Our own roles may change. In some ways, our lives become simpler: questions such as ‘Under what circumstances is this expression preferred to that one?’ become much easier to answer. On the other hand, our lives become more complex, simply because it is much harder to ignore the endless intricacy of language itself. (Hunston 2002: 220)

References Aijmer, K. 2004. Pragmatic markers in spoken interlanguage. In Worlds of Words. A Tribute to Arne Zettersten [Special Issue of Nordic Journal of English Studies], C. Dollerup (ed.), 173–190. Broadhead, A. 2003. Advance your English. Coursebook. Camebridge: CUP. Bernardini, S. 2004. Corpora in the classroom: An overview and some reflections on future developments. In How to Use Corpora in Language Teaching, J. M. Sinclair (ed.), 15–36. Amsterdam: John Benjamins. Carter, R. & McCarthy, M. 2006. The Cambridge Grammar of English. A Comprehensive Guide. Spoken and Written English Grammar and Usage. Camebridge: CUP. De Cock, S. 2004. Preferred sequences of words in NS and NNS speech. Belgian Journal of English Language and literatures (BELL) New series 2: 225–246. Edmondson, W. & House, J. 1981. Let’s Talk and Talk About it. A Pedagogic Interactional Grammar of English. München: Urban & Schwarzenberg. Estling Vannestål, M. & Lindquist, H. 2007. Learning English grammar with a corpus: Experimenting with concordancing in a university grammar course. ReCALL 19(3): 329–350. Granger, S. 1998. Learner English on Computer. London: Longman. Granger, S. 2002. A bird’s eye view of learner corpus research. In Computer Learner Corpora, Second Language Acquisition and Foreign Language Teaching, S. Granger, J. Hung & S. Petch-Tyson (eds), 3–33. Amsterdam: John Benjamins. Hasselgren, A. 2002. Learner corpora and language testing: Smallwords as markers of language fluency. In Computer Learner Corpora, Second Language Acquisition and Foreign Language Teaching, S. Granger, J. Hung & S. Petch-Tyson (eds), 143–173. Amsterdam: John Benjamins. Hunston, S. 2002. Corpora in Applied Linguistics. Camebridge: CUP. Johns, T. 1991. Should you be persuaded: Two examples of data-driven learning. In Classroom Concordancing [ELR Journal 4], T. Johns & P. King (eds), 1–16. Birmingham: CELS University of Birmingham.

10

Karin Aijmer

McCarthy, M., McCarten, J. & Sandiford, H. 2005. Touchstone Student’s Book 1. Camebridge: CUP. Mukherjee, J. & Rohrbach, J.-M. 2006. Rethinking applied corpus linguistics from a languagepedagogical perspective: New departures in learner corpus research. In Planing, Gluing and Painting Corpora: Inside the Applied Corpus Linguist’s Workshop, B. Ketteman & G. Marko (eds), 205–232. Frankfurt: Peter Lang. Müller, S. 2005. Discourse Markers in Native and Non-native English Discourse. Amsterdam: John Benjamins. Römer, U. 2006. Pedagogical applications of corpora: Some reflections on the current scope and a wish list for future developments. Zeitschrift für Anglistik und Amerikanistik 54(2): 121–134. Rühlemann, C. 2006. Conversation in Context. A Corpus-driven Approach. London: Continuum. Tsui, A. B. M. 2004. What teachers have always wanted to know – and how corpora can help. In How to Use Corpora in Language Teaching, J. Sinclair (ed.), 39–61. Amsterdam: John Benjamins.

part i

Corpora and second-language acquisition

The contribution of learner corpora to second language acquisition and foreign language teaching A critical evaluation Sylviane Granger

Learner corpus research lies at the crossroads between four major disciplines: corpus linguistics, linguistic theory, second language acquisition and foreign language teaching. The first part of the article highlights the contribution of these four disciplines and shows that each of them is necessary for an optimal exploitation of the data. Some of the controversies that learner corpus research is currently giving rise to are tackled, in particular the reference to native speaker norms. The second part is devoted to pedagogical applications of learner corpus research. First, the contribution of two types of corpora – corpora for delayed pedagogical use and corpora for immediate pedagogical use – is compared. Secondly, current learner-corpus-informed pedagogical tools are assessed on the basis of the targeted type of proficiency: accuracy, complexity and fluency (Skehan 1998). The article closes with an agenda of priorities for future research.

1.

Introduction

There has been a tendency over the years for researchers within the fields of Second Language Acquisition (SLA) and foreign language teaching (FLT) to maintain a certain distance. In the face of this situation, those working within the younger field of learner corpus research have always considered it as one way of providing a bridge between the two disciplines. So while Gass & Selinker (2001: 2) explicitly express in the introduction to their SLA volume that “over the years, the study of second language acquisition has become inextricably intertwined with language pedagogy” and that their goal in writing the book is to “disentangle the two fields”, early publications on learner corpus research point to the dual contribution of the

14

Sylviane Granger

field, emphasizing in particular its potential for language teaching (Granger 1993, 1994; Milton & Chowdhury 1994). Undeniably however, learner corpus research has not yet fully realized its stated ambition as its links with Second Language Acquisition have been somewhat weak and it has given rise to relatively few concrete pedagogical applications. This lack of concrete applications can in fact easily be explained by the vast amount of analytic work that needs to precede the design of learner-corpus-informed teaching tools. The weakness of its links with SLA is more difficult to explain. One potential factor is the background of learner corpus researchers: most are corpus linguists and/or language teachers rather than SLA specialists. De Haan (1999) for instance, in his review of the first volume devoted to learner corpus research, states “I am a corpus linguist rather than an SLA specialist, so I do not have more than a basic knowledge of SLA research. Still, I have an idea that this volume is more to do with EFL/ESL than with SLA.” Barlow (2000) points to another crucial reason, viz. the fact that the SLA field has, for several years now, been dominated largely by a single issue, namely the role played by UG (universal grammar) in the learning of a second language. As a result, the data-driven paradigm has held little appeal. This said, there are clear signs that the two fields are now coming closer together, with learner corpus researchers becoming progressively aware of the importance of SLA theory and SLA researchers finally acknowledging the potential value of learner corpora. This period of rapprochement provides an excellent backdrop against which to take a critical look at the work achieved so far from a combined SLA-FLT perspective with a view to identifying the most promising avenues for future research in the field. This chapter is organized as follows. Section 2 highlights the inherently interdisciplinary nature of learner corpus research. Section 3 takes the SLA perspective and addresses one of the main criticisms levelled at learner corpus research, viz. its tendency to fall prey to the so-called ‘comparative fallacy’. Section 4 looks at the pedagogical contributions of learner corpus research and surveys existing and potential learner-corpus-informed tools in relation to three aspects of learner proficiency – accuracy, complexity and fluency – with particular emphasis on accuracy and its hotly debated correlate, the notion of error. Section 5 concludes and describes some priorities for the future.

2.

Core components of learner corpus research

Learner corpora (LC) are electronic collections of foreign or second language learner texts assembled according to explicit design criteria. The fact that they contain data from language learners makes them a very special type of corpus, requir-

The contribution of learner corpora

Corpus Linguistics

Foreign Language Teaching

LEARNER CORPUS RESEARCH

Linguistic Theory

Second Language Acquisition

Figure 1. Core components of learner corpus research

ing from the analyst a wider range of expertise than is necessary for native corpora. Figure 1 highlights the inherently interdisciplinary nature of the LC field. For the process of data collection and automated analysis, the researcher needs to have corpus linguistic expertise, familiarity with corpus design, corpus annotation, automated data extraction and analysis as well as statistical handling of the data. While learner corpus researchers can take advantage of the many off-the-shelf corpus tools and well-tried corpus methods, they have to approach their task with great caution as most of these tools and methods have been developed on the basis of native speaker data and are therefore not fully adapted for processing learner language. The high rate of errors in learner corpora affects the accuracy rate of POS-tagging (Van Rooy & Schäfer 2003) and other quantitative measures such as the type-token ratio (Granger & Wynne 1999). In addition, the very nature of the data calls for the design of new tools, such as error editing software (Izumi et al. 2005; Lüdeling et al. 2005). A good background in linguistic theory is also necessary to provide the necessary underpinning for linguistic analysis of the data. Although functional approaches are particularly well-suited for the analysis of language in use (Meyer 2002: 6), recent studies have shown that other approaches such as those used within cognitive linguistics are in no way incompatible with a corpus approach to language (Gries & Stefanowitsch 2006; Stefanowitsch & Gries 2006). Stefanowitsch & Gries (2003), in particular, use corpus data to investigate the strength of

15

16

Sylviane Granger

association between words and constructions which, in the cognitive framework of Construction Grammar, are viewed as meaningful units. Knowledge of SLA theory is a prerequisite for interpretation of the data. A wide range of social, cognitive and psychological factors that play a key role in language learning have been extensively studied in SLA and familiarity with SLA findings will greatly help LC analysts provide correct interpretations of their results (for an excellent survey of SLA with a focus on the analysis of learner production, see Ellis & Barkhuizen 2005). Some recent SLA developments, such as Ellis’s (2002) description of frequency effects and Jarvis’s (2000) rigorous approach to transfer, are particularly relevant for learner corpus analysis. Finally, a good understanding of foreign language teaching issues is essential if LC research is to lead to effective pedagogical applications. It is particularly important to study the impact of contextual factors, as these will determine whether and to what extent the results of learner corpus research can be integrated into teaching (cf. Section 4.2). In addition, new corpus-based learning methods such as data-driven learning should be carefully assessed in terms of feasibility and efficiency. If LC analysts want to use their data to the full, they ideally need to be able to consider it from all these four perspectives.

3.

Learner corpus research and SLA

3.1

Empirical basis

One of the main assets of learner corpus research is that it brings to the SLA field a much wider empirical basis than has ever previously been available. The difficulty of drawing firm conclusions from a narrow empirical basis is underlined by many SLA specialists, among others Gass & Selinker (2001: 31) who note that “It is difficult to know with any degree of certainty whether the results obtained are applicable only to the one or two learners studied, or whether they are indeed characteristic of a wide range of subjects”. Similarly, Myles (2005) notes in her survey of interlanguage corpora and SLA research that “Time has now come, though, to test some of the current hypotheses on larger and better constructed datasets, as has happened in L1 acquisition”. Learner corpora contain data from hundreds (sometimes thousands) of learners and can therefore lay claim to greater representativeness than previous SLA studies. In addition, the very nature of the data, viz. continuous contextualized discourse, enables researchers to tackle a much wider range of topics and hence brings to light a much more diversified view of learner language. The traditional overemphasis on morphology is progressively


being replaced by a greater attention to lexis, phraseology, genre diversification and many other hitherto neglected aspects of learner language (Cobb 2003: 419). It is important to bear in mind, however, that for the SLA specialist, big is not necessarily beautiful. Unlike the corpus linguist for whom “the whole point of assembling a corpus is to gather data in quantity” (Sinclair 1995: 21), the SLA specialist attaches more importance to control over the many variables that affect learner production than to sheer size. As a result, learner corpora need to be assembled on the basis of very strict design criteria and a wide range of variables should ideally be recorded for each learner production. There are several learner corpora which meet this challenge and which contain a large number of recorded and searchable variables. For example, in the International Corpus of Learner English (ICLE) (Granger et al. 2002; Granger 2003), each of the 3,640 learner texts is accompanied by over 20 learner and task variables, all of which have been stored in a database and can be used by researchers as queries to compile subcorpora that match certain criteria, thus allowing for interesting comparisons (e.g. female vs. male learners, German- vs. Spanish-speaking learners, etc.). One must admit, however, that this facility is still seldom used and LC researchers (myself included) have had a tendency to base their analyses on the whole corpus or on subcorpora distinguished only on the basis of the learners’ mother tongue. In fact, a properly coded learner corpus makes it possible for researchers to study the effect of a much wider range of variables, as illustrated by Ädel (2006) who shows the effect of planning time (timed vs untimed essay writing) or Abe & Tono (2005) who highlight the effect of medium (speech vs writing), task and proficiency level and one can only hope for an increase of this type of study in the future. In spite of some continuing reticence on the part of SLA researchers however, the fact that a recent SLA publication for the very first time devotes a whole chapter to learner corpora (Ellis & Barkhuizen 2005) is a clear sign that this new resource will soon be accepted as a bona fide data type in SLA research. However, for the resource to be maximally useful, it needs to cover a wider range of tasks and include a larger proportion of longitudinal data. To increase corpus interchange, it would also be very useful to design a standardized architecture for learner corpora, with a tailor-made Document Type Definition and a user-friendly interface. As things stand at the moment, the number and format of learner and task variables encoded differs dramatically from one learner corpus to another. The LeaP (Learning Prosody in a Foreign Language) corpus (Gut et al. 2004), for example, includes variables relating to motivation and attitude which are seldom included in learner corpus metadata. While total standardization is neither possible nor desirable, some degree of harmonization would greatly foster both theoretical and practical research in the field.

17

18

Sylviane Granger

Another factor which may slow down the process of assimilation of learner corpus research into SLA is its close association with Contrastive Interlanguage Analysis (CIA), a method highly popular among LC researchers but condemned by a large proportion of the SLA community. The next section gives a brief survey of the CIA method and addresses the criticism it is facing.

3.2 Contrastive Interlanguage Analysis under fire Contrastive Interlanguage Analysis (CIA) involves quantitative and qualitative comparisons between native language and learner language (L1 vs L2) and between different varieties of interlanguage (L2 vs L2) (Granger 1996). The first type of comparison plays an important role in uncovering the distinctive features of learner language, while the second makes it possible to assess the degree of generalizability of interlanguage features across learner populations and language situations. The CIA approach has been used in a wide range of learner corpus studies and has helped uncover the frequency and linguistic context of a wide range of interlanguage features in the field of morphology (Housen 2002), lexis (Källkvist 1999), syntax (Smits 2002), discourse (Ädel 2006) and phraseology (Nesselhauf 2005). While L2 vs L2 comparisons have never aroused any criticism from SLA specialists, L1 vs L2 comparisons have 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 analyze interlanguage in its own right. The first possible answer to this criticism is that CIA is not a necessary component of learner corpus research. It is perfectly possible to do learner corpus analysis without comparing L1 and L2: learner language can be analyzed in its own right either cross-sectionally or longitudinally. Housen (2002) is a perfect example of this type of approach. It should be pointed out, however, that L1-L2 comparisons are extremely powerful heuristic techniques which help bring to light features of learner language which have not been focused on before and which, once uncovered, can be analyzed from a strictly L2 perspective. In addition, a look at the non-corpus-based SLA literature shows that the comparative fallacy is in fact pervasive but in a hidden undercover way, to the point that the term ‘comparative hypocrisy’ comes to mind. For example, all the studies that compare learners of . For recent discussions of the notion of comparative fallacy, see Purdy (2004) and Tenfjord et al. (2006). . These are just a few examples. For a comprehensive learner corpus bibliography, see http:// cecl.fltr.ucl.ac.be/learner%20corpus%20bibliography.html.


different proficiency levels are in fact based on an underlying L1 norm as proficiency is usually assessed with an L1 target in mind. The same can be said of SLA studies reporting the results of grammaticality judgment tests. I therefore agree with Sung Park (2004) that “any SLA study implicitly has a built-in notion of interlanguage with the target language lurking in the background”. The difference is that in traditional SLA studies, the native speaker norm is often implicit and intuition-based while in learner corpus research, it is explicit and corpus-based (Mukherjee 2005). Corpus linguistics provides us with a wide diversity of possible norms, in terms of both dialects and diatypes. In the case of languages like English, researchers have an extremely wide choice of reference corpora. They can choose between the geographical varieties of English covered in the International Corpus of English (Greenbaum 1996), several of which are available in electronic format, or opt for a corpus of competent L2 user data instead, as suggested by proponents of English as a Lingua Franca (Seidlhofer 2004). In addition, the wide coverage of mega-corpora like the British National Corpus will ensure that researchers compare like with like: learner interview data with native interview data (De Cock 2004), learner academic writing with native speaker academic writing (Paquot 2008). Researchers should make full use of this rich diversity rather than restrict themselves to one monolithic and monocultural norm as has all too often been the case in the past and indeed is still the rule where the norm is implicit rather than explicit. From a pedagogical point of view, the interest of L1-L2 comparisons is even more obvious as they help teachers identify the lexical, grammatical and discourse features that differentiate learners’ production from the targeted norm and are therefore a very rich source of data for pedagogical applications.

4.

Learner corpus research and foreign language teaching

From the very early days of learner corpus research, researchers have focused much attention on the potential pedagogical applications of their research and yet there is undeniably very little evidence of fully-fledged up-and-running applications. In relation to corpus-based studies, Flowerdew (1998: 550) notes that “the implications for pedagogy are not developed in any great detail with the consequence that the findings have had little influence on ESP syllabus and materials design”. Her comment, made in relation to ESP, applies to syllabus and materials design as a whole. In this section the relationship between learner corpus research and FLT is examined from three different angles: a pedagogically-oriented classification of corpora, a warning against direct linking between learner corpus analysis

19

20 Sylviane Granger

and teaching applications and a survey of existing LC-informed tools and those researchers would like to see developed.

4.1 Learner corpora for delayed or immediate pedagogical use Learner corpora can be put into two categories according to the way in which they are collected and their ensuing pedagogical use. Up to very recently, all learner corpora were corpora for delayed pedagogical use (DPU). As shown in Figure 2, DPU corpora are not used directly as teaching/learning materials by the learners who have produced the data (learner population x). They are compiled by academics or publishers with a view to providing a better description of one specific interlanguage and/or designing tailor-made pedagogical tools which will benefit similar-type learners (learner population y), i.e. learners with the same profile as the students who have produced the corpus data (same mother tongue background, same level of proficiency, etc.). More recently, a new type of corpus has begun to emerge. As shown in Figure 3, corpora for immediate pedagogical use (IPU) are collected by teachers as part of their normal classroom activities (Franca 1999) and the learners are at the same time producers and users of the corpus data (learner group x). As indicated by the dotted arrow, nothing prevents the data from being subsequently used by other similar-type learner groups (learner group y). DPU corpora are usually bigger and therefore have wider generalizability. A good example of this type of corpus is the Longman Learners’ Corpus or the Cambridge Learners’ Corpus, which contain several million words of data from learners with a wide range of L1s. They are the ideal resources for designing generic pedagogical tools like EFL dictionaries or grammars (Gillard & Gadsby 1998). IPU corpora are usually much smaller and therefore, as pointed out by Ragan (2001: 210), not representative in the usual sense of the word: they only represent themselves “providing specific information and a basis for generalizations concerning the limited range of the variety of language”. Their advantage over DPU corpora is that they are arguably more relevant as learners work on their own productions. Mukherjee & Rohrbach (2006: 228) consider that this kind of “localisation of learner corpus compilation” is one of the most promising avenues in learner corpus research. They argue that “firstly, the focus on their own students’ output will involve many more teachers in corpus-based activities and that, secondly, the exploration of learner data by the learners themselves will motivate many more learners to reflect on their language use and thus raise their foreign language awareness”.


LC collection

LC analysis

LC-informed pedagogical application

Learner population x

Academics Publishers

Learner population y

Figure 2. Learner corpora for delayed pedagogical use

Learner group x

LC collection Teachers

LC analysis LC application

Learner group x

Learner group y

Figure 3. Learner corpora for immediate pedagogical use

With the advent of web-based learning platforms, it is becoming possible to combine the advantages of DPU and IPU corpora. The IWiLL language learning environment described by Wible et al. (2001) allows both students and teachers to create and use an online database of Taiwanese learners’ essays and teachers’ error annotations. The learner corpus serves as a dual IPU and DPU resource: the students’ productions are directly integrated in classroom activities but also help to feed the Taiwan Learners’ Corpus, which currently contains over one million words and continues to grow as the writing environment is used by the students and their teachers. Another project which demonstrates the integration of learner corpus collection and analysis in an e-learning environment is described by Smrz (2004). A vast amount of learner data is currently being produced in electronic format and is therefore analyzable using corpus analytic techniques. Matsuda et al. . IWiLL stands for Intelligent Web-based Interactive Language Learning.

21

22

Sylviane Granger

(2003: 165) mention corpus studies of technology-based writing as one key area for future research: “The heightened accessibility of both first- and second-language electronic writing – whether carried out in school or as part of the public domain – will allow researchers to explore topics such as cohesion, collocation, coherence, lexis, and syntax through large scale analyses and comparisons of online writing samples in a variety of modes, genres and contexts.” Kushner et al. (2001) have collected a large database of learner corpora generated from asynchronous email and postings to a listserv and synchronous chat and highlight the interest of analyzing this kind of data: “These analyses allow us to correlate specific kinds of activity (e.g., moderated debate, free discussion), within specific communicative modalities (e.g., asynchronous threaded discussion, synchronous chat), to gains in specific skills (e.g., grammatical complexity, contextual uses of idiomatic expressions, morphological accuracy)”. One major advantage of collecting data from computer-mediated communication is that these tasks give learners the opportunity to practise a wide range of interactional speech acts (Kung 2004: 169) and thereby give researchers and teachers access to features of learner language that have traditionally been very difficult to collect.

4.2 From learner corpus analysis to language teaching Several statements in the learner corpus literature may give the impression that there should be a direct link between uncovering differences between learner and native data and designing remedial pedagogical tools and methods. Leech (1998: xix–xx) warns against the danger of interpreting terms like ‘overuse’ and ‘underuse’ prescriptively, insisting that “it is a matter for debate whether quantitative findings of this kind “should feed into the development of teaching materials”. Although I may not have made this sufficiently clear in my writings, it has always been clear to me that features of learner language uncovered by LC research need not necessarily lead to targeted action in the classroom. As shown in Figure 4, whether the features are selected for pedagogical action or ignored depends on a variety of features, including learner needs, teaching objectives and teachability. Features such as the overuse of analytic causative constructions in lieu of synthetic verbs (Altenberg & Granger 2001) or the overuse of the all-round intensifier very and associated underuse of other collocationally restricted -ly intensifiers like closely or highly (Granger 1998) are clearly useful targets for exercises in a translator- or teacher-training context or with any learner who wants to achieve near-native proficiency but can clearly be neglected or simply presented as useful strategies for learners whose language learning aims are less ambitious.


L1

L2

Overuse / Underuse / Misuse

Learner needs Teaching objectives Teachability

SELECT

IGNORE

Figure 4. From learner corpus analysis to language teaching

4.3 The contribution of LC-informed pedagogical tools to learner proficiency It is interesting to survey the contribution of learner corpus research to pedagogical tools in relation to the three dimensions of language proficiency distinguished by Skehan (1998): accuracy, complexity and fluency.

4.3.1 Accuracy 4.3.1.1 Error analysis revisited Learner corpora tend to contain a much higher error rate than native corpora. In addition to the instances of over- and underuse described above, they also contain numerous instances of misuse. The emergence of learner corpus research has therefore quite naturally been accompanied by the development of a new type of error analysis (EA), similar to traditional error analysis in that it aims to detect, correct and analyze learner errors, and yet different from it in several major respects (cf. Dagneaux et al. 1998). One important difference is that unlike many traditional EA studies, it does not focus exclusively on errors. It is therefore fully compatible with ‘obligatory occasion analysis’ (Brown 1973) which analyzes “what learners get right as well as what they get wrong” (Ellis & Barkhuizen 2005: 70). In this connection, having access to a POS-tagged corpus is a great help as it is possible to count the number of auxiliary errors, for example, out of the total number of auxiliaries used by the learners. It must be acknowledged, however, that not all learner corpus researchers avail themselves of this possibility and a number of

23

24

Sylviane Granger

corpus-based error analyses suffer from the same weaknesses as earlier EA studies. Reading of the vast EA and SLA literature on the subject is clearly a must for any researcher embarking on computer-aided error analysis. To annotate errors, researchers have designed error-tagging systems with varying degrees of sophistication. Flat annotation systems (Milton & Chowdhury 1994; Dagneaux et al. 1998; Nicholls 2003) are progressively being replaced by more sophisticated systems, such as the Falko multi-level annotation system which makes it possible to encode competing analyses of learner errors on several independent levels (Lüdeling et al. 2005). Whatever the system used, all researchers highlight the need for detailed documentation on the system and tests to assess inter- and intra-rater reliability. A thoroughly error-analyzed learner corpus is an invaluable resource which can inform most pedagogical tools. The few up-and-running applications that have been based on learner corpus analysis give some idea of their value but it is quite clear that major developments are still to come. Learners’ dictionaries have been the first to integrate insights drawn from learner corpora in the form of explicit notes meant to warn learners against attested pitfalls (De Cock & Granger 2005). EFL grammars are lagging behind, probably because of the difficulty of collecting a large representative learner corpus, the daunting amount of work needed to analyze it and, if the targeted grammar is a generic one, the challenge of distinguishing between difficulties shared by a large number of learners and those typical of a given learner population. However, the reference to learner errors extracted from the Cambridge Learners’ Corpus in Carter & McCarthy’s (2006) recent grammar of English shows that change is clearly underway. More rapid development can be expected in the field of specific applications, such as remedial exercises targeting error-prone items or structures for a specific learner population. Cowan et al.’s (2003) ESL Tutor program is an error correction courseware tool that contains units targeting persistent grammatical errors produced by Korean ESL students, such as the overpassivization of unaccusative verbs. In the same vein, Chuang & Nesi (2006) describe a web-based remedial resource called GrammarTalk which tackles recurring errors made by Chinese students studying through the medium of English.

4.3.1.2 Errors in the hot seat The focus on errors which characterizes much of present-day learner corpus research may seem surprising at a time when the very notion of error has become somewhat taboo. It is particularly being challenged by proponents of English as a Lingua Franca (ELF), viz. the variety of English spoken by proficient non-native speakers of English who use English as a means of communication with speakers of different first languages. Jenkins (2005) criticizes the terms ‘learner language’,


‘interlanguage’ and ‘errors’ to refer to what are in her view “alternative, but legitimate ELF varieties”. Here too, but for reasons quite different from those invoked by SLA researchers in relation to the notion of comparative fallacy (cf. Section 3.2), comparison with the native speaker norm is rejected: “Given the shift in the nature of English, it is time to recognize the multilingual context of English use and put aside a native speaker model of research and pedagogy” (McKay 2003). Instead it is suggested that recurrent lexico-grammatical features in non-native varieties such as the conversion of uncount nouns into count nouns (informations, an advice) or divergent use of prepositions should be considered not as errors but as new ‘normative features’ and ‘nativized norms’ (Lowenberg 2002). The main criterion is no longer accuracy but mutual intelligibility. Jenkins (2000) has identified a phonological ELF core which ensures minimal standards of intelligibility and on which all teaching efforts should concentrate. There is no denying that the ELF debate is a healthy one which reminds us that accuracy is not the main teaching objective in all learning contexts. Many of the subtle differences uncovered by the CIA method can and indeed should be disregarded in some teaching contexts. This is the meaning of the ‘select’ and ‘ignore’ arrows in Figure 4. It is therefore quite inappropriate to stigmatize the notion of error in a general way. In addition, it is clearly too early to speak of new norms. In order to discover the distinguishing features of ELF, it is necessary to carry out an in-depth corpus-based study of a wide variety of ELF exchanges. Jonghak Baik & Jinyoung Shim (2002) have produced an entire web-based course to raise students’ awareness of the existence of varieties of English but insist that the only way to identify the level of pronunciation, lexicon and grammatical variation that is acceptable for worldwide communication is to carry out an in-depth survey of the various varieties of English in Kachru’s three concentric circles: inner, outer and expanding. Research on learner corpus data, which are very close to ELF data, points to a very high degree of L1-specificness, especially as regards lexis and lexico-grammar. Like Mackenzie (2003) I therefore think that a description of ELF features is unlikely to reveal a uniform ELF: a Romance ELF will probably turn out to be quite different from a Germanic ELF, which in turn will be quite different from an Asian or South African ELF (see also Mollin 2006 on the controversial notion of Euro-English). One should not forget, however, that accuracy is only one facet of language proficiency. In the next section we turn to the possible contribution of learner corpora to two other equally important aspects: complexity and fluency.

4.3.2 Complexity and fluency Very few applications of learner corpus research aim to improve learners’ proficiency in terms of complexity and fluency. Yet these aspects of language

25

26 Sylviane Granger

proficiency are largely amenable to automated corpus analytic methods and should therefore not be disregarded.

4.3.2.1 Complexity Learner corpora provide an excellent basis for assessing the level of lexical and grammatical complexity displayed by learner language. For lexis, a function like Keywords in WordSmith Tools (Scott 2004) allows researchers to have immediate access to all words or phrases that are significantly over- or underused by learners and are therefore potential targets for vocabulary expansion exercises. For example, the LC-attested overuse of the adjective important can easily be turned into a useful vocabulary enhancement exercise where learners are requested to replace important by other more appropriate adjectives such as critical, crucial, major, serious, significant, vital which they tend to underuse (cf. Granger & Tribble 1998: 205–207). Improving one’s vocabulary is not simply a question of acquiring new words. It is equally if not more important to learn new meanings for old words, i.e. extend the knowledge of words that are already partially known. Liu & Shaw’s (2001) investigation of the high frequency verb make in learner and native writing shows that EFL learners do not exploit the full range of meanings displayed by the verb and suggests that “L2 vocabulary teaching be aimed at raising awareness of word potential”. Similarly, Kaneko’s (2005) investigation of the use of four vertical axis prepositions (above, over, under and below) in a developmental corpus of Japanese learners of English shows how the semantic network gradually expands across proficiency levels. Such findings can help teachers and syllabus designers identify which words and/or meanings of words need to be studied at a particular proficiency level. While researchers can greatly benefit from automated methods of analyzing lexical complexity, they have to rely much more on manual analysis when it comes to syntactic complexity as learner corpora are not usually annotated or, if they are, they are POS-tagged but practically never parsed. This should not deter researchers from analyzing learner syntax. Relying on a combination of manual and automatic retrieval, Delier’s (2006) investigation of relative clauses in learner writing shows an interesting difference between French- (FR) and Spanish-speaking (SP) learners: while the FR learners significantly underuse relative clauses in general, SP learners significantly overuse them. However, closer analysis of the data highlights an interesting similarity between the two groups: they both overuse non-restrictive relative clauses. Another feature that the two learner groups share is the overuse of the pronoun which with non-personal antecedents in subject position and the underuse of the alternative pronoun that in such contexts. Other studies such as my own studies of passives and non-finite clauses (Granger


1997a & b) show a high degree of similarity among learner populations: all four groups studied significantly underuse non-finite clauses and passives. There is no doubt that studies of this kind provide useful information on the topics that need to be addressed in advanced EFL grammars.

4.3.2.2 Fluency Wolfe-Quintero et al. (1998: 4) define the notion of fluency by contrasting it with accuracy and complexity: “while complexity reveals the scope of expanding or restructured second language knowledge, accuracy shows the conformity of second language knowledge to target language norms. In comparison, fluency is a function of the control in accessing that knowledge, with control improving as the learner automatizes the process of gaining access”. One important aspect of fluency is phraseology: “Multiword units provide language that is “ready to speak”, thus facilitating fluency” (Harris 1997). Powerful automated corpus methods like concordancing, collocate display or n-gram extraction have brought out the pervasiveness of what Sinclair (1991: 110) calls the ‘idiom principle’ in language, i.e. the fact that “a language user has available to him or her a large number of semipreconstructed phrases that constitute single choices, even though they might appear to be analysable into segments”. Learner corpus data can be submitted to these same methods and therefore constitute a unique resource to bring out (dis)fluency features in learner language. De Cock’s (2004) investigation of recurrent sequences in native and learner speech has identified a range of multi-word units that learners either fail to use or use significantly less than native speakers. These include hesitation markers, discourse markers and a wide range of “routinized building blocks” which, if acquired, could greatly increase learners’ fluency in speech. Similarly, using corpora of native and learner academic writing, Paquot (2008) and Gilquin et al. (2007) demonstrate that learners lack the largely conventionalized ways of dealing with crucial functions, such as exemplifying, summarizing, contrasting or expressing personal opinion. They overuse a limited number of frequent English collocations and discourse markers (e.g. for example, on the other hand, on the contrary) but underuse a whole set of typical EAP multiword sequences, e.g. is an example of, as discussed, etc. This research has resulted in learner-corpus-informed writing aids that show typical examples of native and learner writing and provide a rich semantic, syntactic, phraseological and stylistic description of the ways particular functions and concepts are lexicalized in academic English.

27

28

Sylviane Granger

5.

Conclusion

Although the field of learner corpus research cannot claim to have reached full maturity yet, it is clear that after the tremendous excitement and partially unrealistic expectations of the pioneering years, researchers are now getting a progressively more realistic view of what learner corpora can and cannot be expected to contribute to language learning and teaching. The future of the field looks bright. On the one hand, it is slowly but surely being integrated into SLA, a movement which is due both to recognition among SLA researchers of the value of the LC approach and a corresponding recognition among LC researchers of the importance of SLA findings. On the other hand, it is progressively bridging the gap with the teaching world thanks to the localization of learner corpora and the production of useful pedagogical tools. As the field is still very young, there is still a lot of work to do. Among the many avenues for future research, I would place the following three quite high on the list of priorities. First, we need more learner corpora – particularly longitudinal ones – representing a much wider range of genres, tasks and learners in a wider range of languages. Secondly, it is time to start thinking seriously of a standardized markup and annotation system and a purpose-built architecture for storing, annotating and searching learner corpora, which should be both powerful and user-friendly (see Tenfjord et al. 2004 for a description of one such tool). Thirdly, there is a need both for thorough analyses of learner data based on solid theoretical underpinnings and to design pedagogical tools which meet the realities of the teaching and learning context. This will take time as there is no such thing as quick and dirty learner corpus analysis (or, at least, there should not be!). In view of the interdisciplinary nature of the field, this programme can only be carried out by multi-disciplinary teams made up of corpus linguists, computer scientists, SLA specialists and teaching practitioners. As in many fields but perhaps more so in learner corpus research than in some others, the motto ‘United we stand, divided we fall’ has a strong resonance.

References Abe, M. & Tono, Y. 2005. Variations in L2 spoken and written English: Investigating patterns of grammatical errors across proficiency levels. The Corpus Linguistics Conference Series 1.1. Corpus Linguistics 2005. Downloaded from http://www.corpus.bham.ac.uk/PCLC/ Ädel, A. 2006. Metadiscourse in L1 and L2 English. Amsterdam: John Benjamins. Altenberg, B. & Granger, S. 2001. The grammatical and lexical patterning of make in native and non-native student writing. Applied Linguistics 22(2): 173–194.


Barlow, M. 2000. Review of Learner English on Computer, S. Granger (ed.). London: Longman. 1998. Language Learning and Technology 3(2): 15–17. Downloaded from http://llt.msu. edu/vol3num2/review/review1/index.html Bley-Vroman, R. 1983. The comparative fallacy in interlanguage studies: The case of systematicity. Language Learning 33: 1–17. Brown, R. 1973. A First Language: The Early Stages. Cambridge MA: Harvard University Press. Carter, R. & McCarthy, M. 2006. Cambridge Grammar of English. A Comprehensive Guide. Cambridge: Cambridge University Press. Chuang, F.-Y. & Nesi, H. 2006. An analysis of formal errors in a corpus of Chinese student writing. Corpora 1(2): 251–271. Cobb, T. 2003. Analyzing late interlanguage with learner corpora: Québec replications of three European studies. The Canadian Modern Language Review/La Revue canadienne des langues vivantes 59(3): 393–423. Cowan, R., Choi, H. E. & Kim, D. H. 2003. Four questions for error diagnosis and correction in CALL. CALICO 20(3): 451–463. Dagneaux, E., Denness, S. & Granger, S. 1998. Computer-aided Error Analysis. System. An International Journal of Educational Technology and Applied Linguistics 26(2): 163–174. De Cock, S. 2004. Preferred sequences of words in NS and NNS speech. Belgian Journal of English Language and Literatures (BELL), New Series 2: 225–246. De Cock, S. & Granger, S. 2005. Computer learner corpora and monolingual learners’ dictionaries: The perfect match. Lexicographica 20: 72–86. De Haan, P. 1999. Review of Learner English on Computer, S. Granger (ed.). London: Longman. 1998. The Clarion 5: 28–31. Delier, S. 2006. Relative Pronouns in Native and Non-native English Academic Writing. A Corpus-based Study. MA thesis. Université catholique de Louvain, Louvain-la-Neuve. Ellis, N. 2002. Frequency effects in language processing. A review with implications for theories of implicit and explicit language acquisition. Studies in Second Language Acquisition 24: 143–188. Ellis, R. & Barkhuizen, G. 2005. Analysing Learner Language. Oxford: Oxford University Press. Flowerdew, L. 1998. Corpus-linguistic techniques applied to textlinguistics. System 26: 541– 552. Franca, V. 1999. Using student-produced corpora in the L2 classroom. In IATEFL 1999 Edinburgh Conference Selections, P. Grundy (ed.), 116–117. Whitstable: IATEFL. Gass, S. M. & Selinker, L. 2001. Second Language Acquisition. An Introductory Course. Mahwah NJ: Lawrence Erlbaum. Gillard, P. & Gadsby, A. 1998. Using a learners’ corpus in compiling ELT dictionaries. In Learner English on Computer, S. Granger (ed.), 159–171. London: Addison Wesley Longman. Gilquin, G., Granger, S. & Paquot, M. 2007. Learner corpora: The missing link in EAP pedagogy. In Corpus-based EAP Pedagogy, P. Thompson (ed.). Special issue of Journal of English for Academic Purposes 6(4): 319–335. Granger, S. 1993. The International Corpus of Learner English. In English Language Corpora: Design, Analysis and Exploitation, J. Aarts, P. de Haan & N. Oostdijk (eds), 57–69. Amsterdam: Rodopi. Granger, S. 1994. The learner corpus: A revolution in applied linguistics. English Today 39(10/3): 25–29. Granger, S. 1996. From CA to CIA and back: An integrated approach to computerized bilingual and learner corpora. In Languages in Contrast. Text-based cross-linguistic studies [Lund

29

30

Sylviane Granger

Studies in English 88], K. Aijmer, B. Altenberg & M. Johansson (eds), 37–51. Lund: Lund University Press. Granger, S. 1997a. Automated retrieval of passives from native and learner corpora: Precision and recall. Journal of English Linguistics 25(4): 365–374. Granger, S. 1997b. On identifying the syntactic and discourse features of participle clauses in academic English: Native and non-native writers compared. In Studies in English Language and Teaching, J. Aarts, I. de Mönnink & H. Wekker (eds), 185–198. Amsterdam: Rodopi. Granger, S. 1998. Prefabricated patterns in advanced EFL writing: Collocations and formulae. In Phraseology: Theory, Analysis and Applications, A. Cowie (ed.), 145–160. Oxford: Oxford University Press. 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): 538–546. Granger, S., Dagneaux, E. & Meunier, F. 2002. The International Corpus of Learner English. Handbook and CD-ROM. Louvain-la-Neuve: Presses Universitaires de Louvain. Available from http://www.i6doc.com. Granger, S. & Tribble, C. 1998. Learner corpus data in the foreign language classroom: Formfocused instruction and data-driven learning. In Learner English on Computer, S. Granger (ed.), 199–209. London: Addison Wesley Longman. Granger, S. & Wynne, M. 1999. Optimising measures of lexical variation in EFL learner corpora. In Corpora Galore, J. Kirk (ed.), 249–257. Amsterdam: Rodopi. Greenbaum, S. (ed.). 1996. Comparing English Worldwide. The International Corpus of English. Oxford: Clarendon Press. Gries, S. & Stefanowitsch, A. (eds). 2006. Corpora in Cognitive Linguistics. Corpus-Based Approaches to Syntax and Lexis. Berlin: Mouton de Gruyter. Gut, U., Milde, J.-T., Voormann, H. & Heid, U. 2004. Querying annotated speech corpora. In Speech Prosody 2004, B. Bel & I. Marlien (eds), 569–572. ISCA Archive, 56. Available from http://www.isca-speech.org/archive/sp2004/sp04_569.pdf Harris, C. 1997. Psycholinguistic studies of entrenchment. In Conceptual Structure, Discourse and Language, J. P. Koenig (ed.), 55–70. Stanford CA: CSLA. Housen, A. 2002. A corpus-based study of the L2-acquisition of the English verb system. In Computer Learner Corpora, Second Language Acquisition and Foreign Language Teaching, S. Granger, J. Hung & S. Petch-Tyson (eds), 77–116. Amsterdam: John Benjamins. Izumi, E., Uchimoto, K. & Isahara, H. 2005. Error Annotation for Corpus of Japanese Learner English. Proceedings of the Sixth International Workshop on Linguistically Interpreted Corpora (LINC-2005). Downloaded from http://acl.ldc.upenn.edu/I/I05/I05-6009.pdf Jarvis, S. 2000. Methodological rigor in the study of transfer: Identifying L1 influence in the interlanguage lexicon. Language Learning 50: 245–309. Jenkins, J. 2000. The Phonology of English as an International Language. Oxford: Oxford University Press. Jenkins, J. 2005. ELF at the gate: the position of English as a Lingua Franca. Humanising Language Teaching 7(2). Downloaded from http://www.hltmag.co.uk/mar05/idea.htm Jonghak Baik, M. & Jinyoung Shim, R. 2002. Teaching world Englishes via the Internet. World Englishes 21(3): 427–430. Källkvist, M. 1999. Form-class and Task-type Effects in Learner English. A Study of Advanced Swedish Learners. Lund: Lund University Press.


Kaneko, T. 2005. Semantic network makes things problematic: Use of vertical axis prepositions by Japanese learners of English. Proceedings of the NICT JLE Corpus Symposium, Kyoto, 7–10. Kung, S.-C. 2004. Synchronous electronic discussions in an EFL reading class. ELT Journal 58(2): 164–173. Kushner, J., Lantolf, J., Thorne, S., Jimenez-Jimenez, A. & Ross, B. 2001. Spanish acquisition: Analysis of learner corpora generated through inter-cultural telecollaboration. Procesamiento del Lenguaje Natural 27: 301. http://www.sepln.org/revistaSEPLN/revista/27/ 27-proyecto6.pdf Leech, G. 1998. Learner corpora: What they are and what can be done with them. In Learner English on Computer, S. Granger (ed.), xiv–xx. London: Addison Wesley Longman. Liu, E. T. K. & Shaw, P. M. 2001. Investigating learner vocabulary: A possible approach to looking at EFL/ESL learners’ qualitative knowledge of the word. International Review of Applied Linguistics in Language Teaching (IRAL) 39(3): 171–194. Lowenberg, P. H. 2002. Assessing English proficiency in the expanding circle. World Englishes 21(3): 431–435. Lüdeling A., Maik, W., Kroymann, E. & Adolphs, P. 2005. Multi-level error annotation in learner corpora. The Corpus Linguistics Conference Series 1, 1: Corpus Linguistics 2005. Downloaded from http://www.corpus.bham.ac.uk./PCLC Mackenzie, I. 2003. English as a Lingua Franca. The European Messenger 12(1): 59–61. Matsuda P. K., Canagarajah, A. S., Harklau L., Hyland, K. & Warschauer, M. 2003. Changing currents in second language writing research: A colloquium. Journal of Second Language Writing 12: 151–179. McKay, S. L. 2003. Toward an appropriate EIL pedagogy: Re-examining common ELT assumptions. International Journal of Applied Linguistics 13(1): 1–22. Meyer, C. 2002. English Corpus Linguistics. An Introduction. Cambridge: Cambridge University Press. Milton, J. & Chowdhury, N. 1994. Tagging the interlanguage of Chinese learners of English. In Entering Text, L. Flowerdew & K. K. Tong (eds), 127–143. Hong Kong: The Hong Kong University of Science and Technology. Mollin, S. 2006. Euro-English. Assessing Variety Status. Tübingen: Narr. Mukherjee, J. 2005. The native speaker is alive and kicking – Linguistic and language-pedagogical perspectives. Anglistik 16(2): 7–23. Mukherjee, J. & Rohrbach, J.-M. 2006. Rethinking applied corpus linguistics from a languagepedagogical perspective: New departures in learner corpus research. In Planing, Painting and Gluing Corpora. Inside the Applied Corpus Linguist’s Workshop, B. Kettemann & G. Marko (eds), 205–232. Frankfurt: Lang. Myles, F. 2005. Review article: Interlanguage corpora and second language acquisition research. Second Language Acquisition Research 21(4): 373–391. Nesselhauf, N. 2005. Collocations in a Learner Corpus. Amsterdam: John Benjamins. Nicholls, D. 2003. The Cambridge Learner Corpus – error coding and analysis for lexicography and ELT. In Proceedings of the Corpus Linguistics 2003 Conference, D. Archer, P. Rayson, A. Wilson & T. McEnery (eds), 572–581. Lancaster: UCREL, Lancaster University. Paquot, M. 2008. Exemplification in learner writing: A cross-linguistic perspective. In Phraseology in Foreign Language Learning and Teaching, F. Meunier & S. Granger (eds). Amsterdam: John Benjamins.

31

32

Sylviane Granger

Purdy, J. D. (ed.) 2004. Commentaries on the Comparative Fallacy in second language research. Working Papers in TESOL and Applied Linguistics 4(1). Available from http://journals. tc-library.org/index.php/tesol/article/view/172/169 Ragan, P. H. 2001. Classroom use of a systemic functional small learner corpus. In Small Corpus Studies and ELT: Theory and Practice, M. Ghadessy, A. Henry & R. L. Roseberry (eds), 207–236. Amsterdam: John Benjamins. Scott, M. 2004. WordSmith Tools, version 4. Oxford: Oxford University Press. Seidlhofer, B. 2004. Research perspectives on teaching English as a Lingua Franca. Annual Review of Applied Linguistics 24: 209–239. Sinclair, J. 1991. Corpus, Concordance, Collocation. Oxford: Oxford University Press. Sinclair, J. 1995. Corpus typology – a framework for classification. In Studies in Anglistics, G. Melchers & B. Warren (eds), 17–33. Stockholm: Almqvist & Wiksell International. Skehan, P. 1998. A Cognitive Approach to Language Learning. Cambridge: Cambridge University Press. Smits, A. 2002. How Writers Begin Their Sentences. Complex Beginnings in Native and Learner English. Utrecht: LOT. Smrz, P. 2004. Integrating Natural Language Processing into e-learning – A case of Czech. Paper presented at COLING 2004. Downloaded from http://acl.ldc.upenn.edu/coling2004/ W6/ps/1.ps Stefanowitsch, A. & Gries, S. T. 2003. Collostructions: Investigating the interaction between words and constructions. International Journal of Corpus Linguistics 8(2): 209–243. Stefanowitsch, A. & Gries, S. T. 2006. Corpus-based Approaches to Metaphor and Metonymy. Berlin: Mouton de Gruyter. Sung Park, E. 2004. The Comparative Fallacy in UG Studies. Working Papers in TESOL and Applied Linguistics 4(1). Available from http://www.tc.columbia.edu/academic/tesol/ Webjournal/forum2004.htm Tenfjord, K., Meurer, P. & Hofland, K. 2004. The ASK corpus – a language learner corpus of Norwegian as a second language. Paper presented at the TALC 2004 conference, Granada – Spain, 6–9 July 2004. Tenfjord, K., Hagen, J. E. & Johansen, H. 2006. The hows and whys of coding categories in a learner corpus (or ‘how and why an error tagged learner corpus is not ipso facto one big comparative fallacy’). Rivista di Psicolinguistica Applicata (RiPLA) VI(3): 93–108. Van Rooy, B. & Schäfer, L. 2003. Automatic POS tagging of a learner corpus: The influence of learner error on tagger accuracy. In Proceedings of the Corpus Linguistics 2003 Conference (CL 2003) [Technical Papers 16], D. Archer, P. Rayson, A. Wilson & T. McEnery (eds), 835–844. Lancaster University: University Centre for Computer Corpus Research on Language. Wible, D., Huo, C.-H., Chien, F.-Y., Liu, A. & Tsao, N.-L. 2001. A web-based EFL writing environment: Integrating information for learners, teachers, and researchers. Computers and Education 37: 297–315. Wolfe-Quintero, K., Inagaki, S. & Kim, H.-Y. 1998. Second Language Development in Writing: Measures of Fluency, Accuracy and Complexity. Honolulu HI: University of Hawai’i Press.

Some thoughts on corpora and second-language acquisition* Stig Johansson

Using experiments on explicit vs. implicit teaching methods as a starting-point, this paper raises some questions on corpora and second-language acquisition: To what extent are controlled experiments arranged to test the effectiveness of the use of corpora in language teaching? To what extent can the use of corpora be grounded in theories of language acquisition and of language in general? Attention is drawn to the connection between ideas of learning as hypothesis testing as expressed by proponents of error analysis and the more recent ideas connected with corpora and the student as a researcher. Reference is made to the usage-based model, which may account both for the nature of language and language acquisition. The usage-based model and the relevance of corpora deserve to be recognised in work on second-language acquisition. It is claimed that corpora have many uses of relevance for language teaching and language learning. Special attention is drawn to three types of corpora: textbook corpora, learner corpora, and multilingual corpora. There is a need, however, for more systematic studies on the effectiveness of corpora in language teaching. Their use is vindicated to the extent that it agrees with what we know about language and language acquisition, and can be shown to be an effective learning tool.

1.

Introduction

The starting-point for my paper is a project that was going on at Göteborg University about 35 years ago. This was at a time when corpora were scarce and difficult to use. Few researchers were using corpora, and even fewer had probably even thought of using them in language teaching. The project did not have anything to

* I am grateful for comments on the role of corpora in language teaching from Kirsten Haastrup, Copenhagen Business School, and my colleague Kay Wikberg, University of Oslo, but neither of them can be held responsible for the views expressed in this paper.

34

Stig Johansson

do with corpora, so why am I taking it up here? I hope this will become clear in the course of my discussion. In the 1960s and early 1970s there was a lively debate in Sweden on methods of foreign language teaching. How does one learn a foreign language? What is the best way of teaching a foreign language? On the one hand, there were those who believed in a mechanistic learning model and advocated teaching by exposure to, and use of, the foreign language and by structural drills without explanation. Use of the mother tongue was banned. Then there were those who favoured a mentalistic learning model and thought that there was a need for explanation, including a comparison with the mother tongue. The main spokesman for the mentalistic approach was Alvar Ellegård, professor of English at Göteborg University. He put forward his views in a book published in 1971 and, with a colleague from Malmö, he published a collection of articles from the lively debate on foreign language teaching, where both sides were represented (Ellegård and Lindell 1970). Significantly, the book had the title: Direkt eller insikt? That is: Direct or insight?

2.

The explicit vs. the implicit method

The debate in Sweden was the background for a research project on the relative effectiveness of an implicit vs. an explicit method of teaching English grammatical structures to Swedish learners. The project was carried out by some of Ellegård’s students, and Ellegård himself gave a brief account in his book from 1971. Two of the participants in the project, Tibor von Elek and Mats Oskarsson, presented some results of their work in an article in the International Review of Applied Linguistics in 1972 and later in a longer report from 1975. Briefly, what was done was the following: a number of experiments were carried out testing the explicit vs. the implicit method. I will refer to the experiment reported in the journal article from 1972. Groups of adult learners were presented with one of two lesson series in an elementary English course, either a series designed according to the explicit method or one using the implicit method. The average age of the learners was 33, and the total number was 125. To eliminate the teacher variable, the lessons were pre-recorded and run from tape-recorders. The lesson series were preceded and followed by a number of tests. The results supported the explicit method and, by implication, the mentalistic learning model; see Figure 1. The main findings were summarised in this way by von Elek and Oskarsson:

Some thoughts on corpora and second-language acquisition Pre-test % correct

Post-test % correct

80

80

70

70

60

60

50

50

Explicit method Implicit method

Figure 1. The explicit vs. implicit experiment: percentage scores in three proficiency tests before and after the lesson series (based on Ellegård 1971: 127)

[…] we venture to conclude that adult learners profit more from a cognitive than from a habit-forming approach to the teaching of grammar. […] The fact that the EX method led to superior results suggests that explanations clarifying language patterns are efficient in internalizing the grammar of a foreign language even when supplied at the expense of practice. […] Further, the fact that the EX group did better on oral tests as well, suggests that the development of aural-oral skills is not entirely dependent upon aural-oral practice; it may well be that the cognitive command of language structures is a short-cut towards the acquisition of such skills. The attitude tests revealed that the cognitive lessons were somewhat more favourably received than the IM series. […] The more effective method will inevitably result in better motivation and a more positive attitude. (von Elek and Oskarsson 1972: 72)

The authors are careful to point out possible weaknesses of the study. They stress that the results cannot be automatically generalised to other age groups and proficiency levels or to learning other aspects of the language. There are no doubt numerous points that could be debated with respect to the explicit vs. implicit experiment, but I would like to stress what seems fundamentally right. It was a good idea to arrange a controlled experiment. It was good that the two methods were based on different theories of how foreign language learning proceeds, so that the results could say something not just about teaching methods but also about language acquisition. The questions that I would like

35

36

Stig Johansson

to raise with respect to corpora are: To what extent are controlled experiments arranged to test the effectiveness of the use of corpora in language teaching? To what extent can the use of corpora be grounded in theories of language acquisition and of language in general? But first we need to look more generally at second language acquisition.

3.

Language learning as hypothesis testing

The time when the teaching methods experiment was carried out was also the period of the breakthrough of error analysis. Pit Corder published his important articles on ‘The significance of learners’ errors’ (Corder 1967) and ‘Idiosyncratic dialects and error analysis’ (Corder 1971). Jan Svartvik started the Swedish-English Contrastive Studies project at Lund University, with an international conference in 1972 which led to a book with the intriguing title Errata (Svartvik 1973). A couple of years later Claus Færch and his colleagues in Copenhagen launched the PIF project, Prosjekt i fejlanalyse or Project In Foreign Language Pedagogy. And many more names, projects, and publications could be mentioned. A fundamental insight of error analysis was that errors provide a window into the learner’s mind. They are not just unfortunate breaches of the language code, to be combated by language teachers. For the researcher, they are a way of uncovering processes of language learning. What the learner produces is not just what he or she has been taught or unsuccessful attempts at reproducing what has been taught. It represents a system, the learner’s interlanguage, to use a term that was introduced at this time (Selinker 1972). Following Corder, the learner is actively involved in a process of hypothesis testing. Figure 2 reproduces a diagram from Pit Corder’s Introducing Applied Linguistics (1973: 337). This seems to me to be in line with the mentalistic model advocated by Ellegård. The idea of hypothesis formation and hypothesis testing was taken up by Claus Færch and his team, though they stress that the hypotheses are not necessarily conscious: The model we present sees foreign language learning as primarily a cognitive process of hypothesis formation and hypothesis testing, supplemented by pro(Færch et al. 1984: 190) cesses of automatization and consciousness raising.

And they continue: One important source for hypothesis formation is input, interpreted by means of inferencing strategies which utilize the learner’s L1 or other languages different from the relevant foreign language, as well as the learner’s existing interlanguage (p. 192) knowledge.

Some thoughts on corpora and second-language acquisition

T eacher’s function:

Learner’s function:

Provision of data and examples

Explanation, description and correction (verification)

Hypothesis formation by induction from data and exs examples

Hypothesis testing

Figure 2. Teaching and learning activities in the classroom (based on Corder 1973: 337)

The idea of inferencing was developed in particular by Kirsten Haastrup, one of the members of the Copenhagen team, in her thesis on lexical inferencing procedures (Haastrup 1991).

4.

The student as researcher

If we accept that the language learner is engaged in hypothesis formation and hypothesis testing, there is just a short step to the use of corpora in language teaching. The most exciting ideas on the use of corpora in language teaching, to my mind, have to do with the notion of the student as researcher, as developed by Tim Johns and others: The assumption that underlies this approach [the data-driven learning approach] is that effective language learning is itself a form of linguistic research, and that the concordance printout offers a unique resource for the stimulation of inductive learning strategies – in particular the strategies of perceiving similarities and (Johns 1991b: 30) differences and of hypothesis formation and testing. With the appropriate tool-kit, a corpus affects not only the syllabus: it also affects the role of the student. […] with a corpus, the student can actually test the conventional wisdom of the textbooks, and find out what really happens in connected texts. In this way the distinction between teaching and research becomes (Knowles 1990: 47) blurred and irrelevant. […] the use of text analysis tools and corpora allows everyone to become researchers: from the theoretical linguist to the student learning a second language. (Barlow 1996: 2)

37

38

Stig Johansson

Intuitively, this use of corpora seems attractive. Given appropriate search tools, corpora are a way of enhancing and focusing the input to the student. They provide authentic data. They encourage reflection. They are well suited for consciousness-raising activities and for the training of inferencing. They stimulate the student to work actively and independently, and in this way they probably increase both the motivation of the student and the learning effect.

5.

Corpora and language acquisition

The use of corpora has given an impetus to new ideas on the nature of language and language acquisition. One of the main new findings is that texts are to a large extent made up of phrasal units, some of them fixed and others semi-fixed and variable, units which did not fit comfortably into traditional language descriptions. John Sinclair claims that texts are interpreted primarily with reference to what he calls the idiom principle (Sinclair 1991: 110ff.), and a separation of grammar and lexis is rejected. Building on his ideas, Susan Hunston and Gill Francis (1999) have proposed their pattern grammar as an alternative to traditional accounts. This view is also compatible with construction grammar, as developed by Charles Fillmore and others (e.g. Fillmore et al. 1988). With this new view of language, how do we account for language acquisition? Michael Barlow (1996: 14ff.) suggests that form-meaning links, so-called schemas, are acquired through instances of use in a discourse context and are further entrenched through frequency of use. Acquisition is seen primarily as an inductive process. Barlow adds, however, that induction is insufficient: Having argued for the role of induction, I should note that in my view language learning is not based totally on induction. There are two complications. One is the existence of innate structures relevant to language. The second is the fact that certain cognitive distinctions are prime targets for coding in a language and so in a sense there is an impetus to code certain dimensions. Thus, the learner is not seen as just a passive pattern extractor, but is, in addition, a cognizer with the ability to make numerous cognitive distinctions, some of which will be able to be (Barlow 1996: 17f.) linked with formal distinctions made in the grammar.

In this context, I would also like to quote from Joan Bybee’s Presidential Address to the Linguistic Society of America (2005), recently also given as a guest lecture at the University of Oslo: While all linguists are likely to agree that grammar is the cognitive organization of language, a usage-based theorist would make the more specific proposal that grammar is the cognitive organization of one’s experience with language. […]


The proposal presented here is that the general cognitive capabilities of the human brain, which allow it to categorize and sort for identity, similarity and difference, go to work on the language events a person encounters, categorizing and entering in memory these experiences. Grammar cannot be thought of as pure abstract structure that underlies language use: just as there can be no discrete separation of grammar and lexicon because there are so many cases in which specific lexical items go with / require certain grammatical structures, there can be no strict separation of grammar and usage. Grammar is built up from specific instances of use which marry lexical items with constructions; it is routinized and entrenched by repetition and schematised by the categorization of exemplars. Language can be viewed as a complex system in which the processes that occur in individual language events, […] with high levels of repetition, not only lead to the establishment of a system within the individual, but also lead to the creation of grammar, its change, and its maintenance in society.

This usage-based approach (see also Barlow and Kemmer 2000) thus accounts both for the nature of language and for language acquisition. If this view is correct, the use of corpora will strengthen natural processes of language acquisition in that they make it easier for learners to notice and experience repeated instances of use. It seems to me that the usage-based model and the relevance of corpora deserve to be recognised in works on second-language acquisition. I have looked at some recent works to see if there was anything to find on the role of corpora. I found precious little. And yet there is a discussion of notions which are central in the use of corpora for teaching and learning, such as attention and awareness, input, hypothesis formation. Here is a task for the future for those who believe in the validity of the usage-based model and the corpus-based approach.1

6.

The many uses of corpora

So far, I have referred to corpora in general, without saying anything about the various types of corpora and their uses in connection with language teaching. Corpora have many applications which are relevant to language teaching (see Figure 3). They can be used in the preparation of textbooks, grammars, dictionaries, and other teaching material. They can be used in syllabus design. They can be 1. We find a beginning in a recent book on Analysing Learner Language, which includes a paper by Michael Barlow on ‘Computer-based analyses of learner language’ (Barlow 2005).

39

40 Stig Johansson

Dictionaries Grammars

Textbooks Teaching material Syllabus design

Corpus

Classroom activities Testing

Basic research

Figure 3. Uses of corpora of relevance for language teaching

used in the training of teachers. They can be used in testing. And they can be used in the classroom. The types of corpora will vary with the type of use (see, for example, Hunston 2002: 198ff.). Corpora are of course also important for basic research of relevance for language teaching and language learning. Three types should be singled out especially: – textbook corpora, used for the study of the language of textbooks, as in the work of Dieter Mindt (1987) and the recently published monograph by Ute Römer (2005); – learner corpora, used for research on material produced by language learners (Granger 1998, Granger et al. 2002, etc.);2 – multilingual corpora, used for contrastive analysis and translation studies (Aijmer et al. 1996, Johansson and Oksefjell 1998, etc.). The study of textbook corpora focuses on one of the most important sources of input in language teaching. Research on learner corpora gives insight into the state of knowledge of language learners, which is of great relevance for the planning of teaching and for the understanding of language learning. Research on multilingual corpora reveals similarities and differences between languages and is crucial for understanding problems that language learners and translators may face in coming to terms with a new language. Combining contrastive analysis and error analysis was a dream I had long ago when I wrote a paper for the Swedish-English Contrastive Studies project (Johansson 1975). Then we had no multilingual corpora, and the learner corpus existed in the form of heaps of papers which were cumbersome and time-consuming to use. 2. As pointed out by Sylviane Granger in her contribution to this book, learner corpora can be used both for research and for immediate classroom use.


Two decades later, Sylviane Granger presented a case for an integrated contrastive model, combining research on bilingual and learner corpora (Granger 1996), and we find a very good example of this kind of study in a paper by Gaëtanelle Gilquin (2000/2001). I believe this approach is very promising. With the access to multilingual corpora, where a number of languages are represented, and learner corpora, including texts by learners with a range of different mother-tongue backgrounds, we are now in a good position to tackle one of the fundamental questions on second-language acquisition: to what extent is this process guided by, or independent of, the mother tongue?

7.

Corpora in language teaching: concluding remarks

In my paper I have not given any specific examples of how corpora can be used in the classroom. There is a lot to find in publications by Tim Johns (1991a, 1991b), Guy Aston (1997), Aston et al. (2004), Silvia Bernardini (2004), and many others. What I miss are systematic studies testing the benefits of the approach. According to Susan Hunston, “[t]he hypotheses about the benefits of DDL [data-driven learning] have not yet been adequately tested”, though she mentions a couple of small-scale studies (Hunston 2002: 170), including an investigation by Cobb and Horst (2001). In a recent paper reviewing corpus consultation in language studies we read: Alongside developments in language research, the potential of corpora as a resource in language learning and teaching has been evident to researchers since the late 1960s [sic]. Despite publications which emphasise the benefits of corpus consultation for language learners […], there is little evidence to suggest that direct corpus consultation is coming to be seen as a complement or alternative to consultation of a dictionary, course book, or grammar by the majority of learners. There is thus a need for research to underpin the integration of corpora and concordancing in the language-learning environment. (Chambers 2005: 111)

We need studies of the use of corpora along the lines of Kennedy and Miceli (2001), who discuss their students’ problems with corpus investigation and suggest how they could be better equipped to be corpus researchers. There is also a need for controlled experiments, as in the comparison of the explicit vs. the implicit method, which I mentioned at the outset. Is the use of corpora to be grouped with the explicit or the implicit method? The term ‘data-driven learning’ suggests that it is an inductive approach and therefore comparable with the implicit method, though the emphasis is on gaining insight rather than establishing habits, and in this sense it is mentalistic. I believe

41

42

Stig Johansson

that the dichotomy explicit-implicit is far too simple. In the case of corpora in language teaching, I would favour a guided inductive approach or a combination of an inductive and a deductive approach where the elements of explanation and corpus use are tailored according to the needs of the student.3 I found support for this view in the discussion of formal instruction and second language acquisition in the important book by Rod Ellis (1994). After discussing a range of relevant studies, including the Gothenburg implicit vs. explicit experiments, Ellis comes out in favour of a combinatory approach. Formal instruction is seen as facilitating natural language development: Instruction works by helping learners to pay selective attention to form and form-meaning connections in the input. It provides learners with tools that help them to recognize those features in their interlanguages which are in need of (p. 656) modification. Facilitating selective attention by devising instructional activities that equip learners with conscious rules, or that help them interpret the meanings of specific forms in the input, is both psycholinguistically feasible and possible in practical (p. 657) terms.

Corpora should be a good way of devising such instructional activities. But Ellis wrote his book before the breakthrough of corpus methods in language teaching, and as far as I can see, there is no mention of corpora in the book.4 The point of my paper has been to argue for the relevance of corpora in language teaching and at the same time sound a note of caution. Corpora are important in basic research, and they have a role to play in the classroom as well. But let’s not exaggerate. Corpora are no replacement for natural communication. They cannot replace the teacher. And, finally, they should not be used in language teaching just because we now have this wonderful tool and would like to apply it in language teaching as well. Their use is vindicated to the extent that it agrees with what we know about language and language acquisition, and can be shown to be an effective learning tool.

3. To return to Figure 2, the corpus could come in either in connection with the provision of data/examples or following the stage of explanation. 4. It is worth noting, however, that Ellis is one of the editors of a recent book which contains a paper on learner corpora; cf. Note 2 above.


References Aijmer, K., Altenberg, B. & Johansson, M. (eds). 1996. Languages in Contrast. Papers from a Symposium on Text-based Cross-linguistic Studies, Lund 4–5 March 1994 [Lund Studies in English 88]. Lund: Lund University Press. Aston, G. 1997. Enriching the learning environment: Corpora in ELT. In Teaching and Language Corpora, A. Wichmann, S. Fligelstone, T. McEnery & G. Knowles (eds), 51–64. London: Longman. Aston, G., Bernardini, S. & Stewart, D. (eds). 2004. Corpora and Language Learners. Amsterdam: John Benjamins. Barlow, M. 1996. Corpora for theory and practice. International Journal of Corpus Linguistics 1(1): 1–37. Barlow, M. 2005. Computer-based analyses of learner language. In Analysing Learner Language, R. Ellis & G. Barkhuizen (eds), 335–357. Oxford: OUP. Barlow, M. & Kemmer, S. (eds). 2000. Usage-based Models of Language. Stanford CA: CSLI. Bernardini, S. 2004. Corpora in the classroom: An overview and some reflections on future developments. In Sinclair (2004), 15–36. Bybee, J. 2005. From usage to grammar: The mind’s response to repetition. LSA Presidential Address 2005. Accessed from: http://www.unm.edu/~jbybee/. Chambers, A. 2005. Integrating corpus consultation in language studies. Language Learning & Technology 9(2): 111–125. Cobb, T. & Horst, M. 2001. Reading academic English: Carrying learners across the lexical threshold. In Research Perspectives on English for Academic Purposes, J. Flowerdew & A. Peacock (eds), 315–329. Cambridge: CUP. Corder, S. P. 1967. The significance of learners’ errors. International Review of Applied Linguistics 5: 161–170. Corder, S. P. 1971. Idiosyncratic dialects and error analysis. International Review of Applied Linguistics 9: 147–159. Corder, S. P. 1973. Introducing Applied Linguistics. Harmondsworth: Penguin. Elek, T. von & Oskarsson, M. 1972. An experiment assessing the relative effectiveness of two methods of teaching English grammatical structures to adults. International Review of Applied Linguistics 10: 60–72. Elek, T. von & Oskarsson, M. 1975. Comparative Method Experiments in Foreign Language Teaching: The Final Report of The GUME/Adults Project. Research Bulletin 19. Mölndal: Pedagogiska institutionen, Lärarhögskolan i Mölndal. Ellegård, A. 1971. Språk, språkvetenskap, språkinlärning. Aldus No. 303. Stockholm: Bonniers. Ellegård, A. & Lindell, E. (eds). 1970. Direkt eller insikt? Lund: CWK Gleerup. Ellis, R. 1994. The Study of Second Language Acquisition. Oxford: OUP. Faerch, C., Haastrup, K. & Phillipson, R. 1984. Learner Language and Language Learning. Clevedon: Multilingual Matters. Fillmore, C. J., Kay, P. & O’Connor, M. C. 1988. Regularity and idiomaticity in grammatical constructions: The case of let alone. Language 64(3): 501–538. Gilquin, G. 2000/2001. The Integrated Contrastive model: Spicing up your data. Languages in Contrast 3(1): 95–123. Granger, S. 1996. From CA to CIA and back: An integrated approach to computerized bilingual and learner corpora. In Aijmer et al. (1996): 37–51.

43

44 Stig Johansson

Granger, S. (ed.). 1998. Learner English on Computer. London: Longman. Granger, S., Hung, J. & Petch-Tyson, S. (eds). 2002. Computer Learner Corpora, Second Language Acquisition and Foreign Language Learning. Amsterdam: John Benjamins. Haastrup, K. 1991. Lexical Inferencing Procedures or Talking about Words. Tübingen: Narr. Hunston, S. 2002. Corpora in Applied Linguistics. Cambridge: CUP. Hunston, S. & Francis, G. 1999. Pattern Grammar: A Corpus-driven Approach to the Lexical Grammar of English. Amsterdam: John Benjamins. Johansson, S. 1975. The uses of error analysis and contrastive analysis. In Papers in Contrastive Linguistics and Language Testing [Lund Studies in English 50], S. Johansson, 9–23. Lund: CWK Gleerup. Johansson, S. & Oksefjell, S. (eds). 1998. Corpora and Cross-linguistic Research: Theory, Method, and Case Studies. Amsterdam: Rodopi. Johns, T. 1991a. Should you be persuaded: Two examples of data-driven learning materials. English Language Research Journal 4: 1–16. Johns, T. 1991b. From printout to handout: Grammar and vocabulary learning in the context of data-driven learning. English Language Research Journal 4: 27–45. Kennedy, C. & Miceli, T. 2001. An evaluation of intermediate students’ approaches to corpus investigation. Language Learning & Technology 5(3): 77–90. Knowles, G. 1990. The use of spoken and written corpora in the teaching of language and linguistics. Literary and Linguistic Computing 5(1): 45–48. Mindt, D. 1987. Sprache – Grammatik – Unterrichtsgrammatik. Futurischer Zeitbezug im Englischen I. Frankfurt: Moritz Diesterweg. Römer, U. 2005. Progressives, Patterns, Pedagogy. A Corpus-driven Approach to English Progressive Forms, Functions, Contexts and Didactics. Amsterdam: John Benjamins. Selinker, L. 1972. Interlanguage. International Review of Applied Linguistics 10: 219–231. Sinclair, J. M. 1991. Corpus, Concordance, Collocation. Oxford: OUP. Sinclair, J. M. 2004 (ed.). How to Use Corpora in Language Teaching. Amsterdam: John Benjamins. Svartvik, J. (ed.). 1973. Errata. Papers in Error Analysis. Lund: CWK Gleerup.

part ii

The direct corpus approach

Who benefits from learning how to use corpora? Solveig Granath

Whereas researchers have long seen the benefits of using corpora to enhance the description of language, the regular use of corpora in the EFL classroom is still a rare occurrence. One reason is likely to be that learning how to use corpora is seldom part of teacher training courses. As a result, teachers themselves – both at university level and at lower levels – lack the skills needed to use this “native-speaker consultant”. If training in how to use corpora were integrated into university level courses such as syntax, written proficiency and translation, in time it could become just as natural to consult a corpus as to look up an item in a dictionary or a grammar book. The present paper, which is based on almost a decade’s experience of using corpora in the teaching of EFL syntax at the university level, outlines how teachers can utilize this resource both to design exercises and to make grammar “come alive” in the classroom. Furthermore, it summarizes students’ attitudes regarding hands-on exercises using corpora, and discusses the problems as well as the benefits of involving students directly in interpreting edited and unedited corpus data.

1.

Introduction

The use of computerized corpora in linguistic research is relatively recent; although the first computerized reference corpus in English, the Brown corpus of American English, was compiled in the 1960’s, it took almost three decades for the use of corpora to spread beyond the inner circle of corpus linguists. At about the same time, the realization that corpora could be used to enhance language learning and teaching took hold, and at the second TaLC (Teaching and Language Corpora) conference in Lancaster in 1996, there was a lot of optimism about the benefits of using corpora in the classroom. A quote from the conference announcement (cited by Stewart et al. 2004: 1) attests to this:

48 Solveig Granath

While the use of computer text corpora in research is now well established, they are now being used increasingly for teaching purposes. This includes the use of corpus data to inform and create teaching materials; it also includes the direct exploration of corpora by students, both in the study of linguistics and of foreign languages.

A decade later, it appears that although the use of corpora in teaching has continued to spread, it is not as pervasive today as many linguists in the late 1990’s had expected it would be (see e.g. Granath 1998). Thus, Römer (2004, 2005) demonstrates that the language of German EFL textbooks differs radically from authentic language use, as shown in her comparison of corpus data with textbook material. Although sometimes included in university courses for advanced learners of English, the direct exploration of corpora by students appears to be primarily found in elective project courses, where the aim is to teach students research methods in corpus linguistics (see e.g. Davies 2004; Kirk 2002; Mair 2002), rather than in basic, mandatory courses in second and foreign languages. Consequently, the number of students who are given the opportunity to try out this tool on their own makes up only a small minority of all language students at our universities. What could be the reason for this? Is it because analysing concordances is considered to be beyond what students on the intermediate level can be expected to do, or is it because those who teach these courses themselves have never learnt to use corpora? Based on my own experience teaching English syntax to Swedish university students, I will, in what follows, suggest that different types of corpus exercises can profitably be made part of regular foreign language courses. First, I will demonstrate how students can be introduced to corpus analysis in the classroom; next, I will describe how they can be taught to perform teacher-guided corpus searches in the language lab. I will also give a survey of student attitudes to these hands-on exercises, and finally sketch an answer to the question: Who benefits from learning how to use corpora?

2.

Looking back: What are the stumbling-blocks, and what are the advantages?

When I first started using corpora in my own teaching, in the spring of 1998, I was very optimistic about it. I foresaw a rapid development, with teachers and students alike being eager to put this new tool to use. Here was the perfect supplement to dictionaries and grammars, a tool to be used in conjunction with these, which would help raise students’ awareness of how language works by giving them access to a ‘native-speaker consultant’ who would be at their beck and call. At the time, I expected students to share my excitement about the possibility of

Who benefits from learning how to use corpora?

being able to study authentic language use to find out more about language patterns. But it was not long before I came to realize that students would not automatically catch on and begin to experience the thrills of using corpora. That a number of my colleagues have had the same experience came out quite clearly at the colloquium on Corpora and Language Teaching at the Corpus Linguistics conference in Birmingham in the summer of 2005 (CL 2005). Several speakers mentioned that “corpora ruin students’ regulated world”. Students want simple, straightforward answers, and are disappointed by the “blurry” responses they get from corpora. So why bother? As I propose to show, there are many benefits to be gained from introducing students to corpora, even if the aim does not necessarily have to be to make proficient corpus users of all students. Rather, integrating corpora into language courses can be done with several purposes in mind. The most important, perhaps, is to allow students to encounter ‘real language’ rather than made-up examples. Corpora are invaluable for teachers, in that they can employ them in a number of ways, such as, for example, to create exercises, demonstrate variation in grammar, show how syntactic structures are used to signal differences in meaning and level of style, discuss near-synonyms and collocations, and last (but not least) to give informed answers to student questions. Unfortunately, the time that can be set aside for integrating corpus exploration into regular language courses such as grammar, translation and written and spoken proficiency will usually be fairly restricted; in addition, the time for actual teacher-led training in a language laboratory will probably also have to be limited to a minimum. A good idea is therefore to make sure that corpus-based exercises are not only focused on corpus analysis, but also form an integral part of the course. In the following, I will outline the method I use to introduce students in a step-wise fashion to conducting corpus analysis using untagged text corpora, and show how these exercises simultaneously allow students to practise general grammar skills.

3.

Introducing students to corpus analysis: Teacher-prepared exercises

If first-term university-level EFL students are to be able to explore corpora on their own, there are certain basic requirements. First and foremost, students need to know how to identify different word classes, since in English, for many lexical items, it is the use of a word in context rather than inflectional morphology that signals the word class. Second, when working with corpora, students must be able to identify those hits that are relevant and weed out those that are not. Third, and this type of training applies above all to courses where syntactic variation is of

49

50

Solveig Granath

interest, students need to know how they can calculate variation based on a limited number of sample sentences. In the following, I will give examples of teacher-prepared exercises that I have used in EFL syntax courses in order to help students develop these skills. As already mentioned, being able to determine the category of a word is essential when working with text corpora; this is also an integral part of the grammar course as a whole. Consequently, the first corpus-based exercise that my students encounter consists of fifteen sentences all containing the word round and fifteen sentences containing that, where the students’ task is to determine the lexical categories of these words, working in groups of three or four. These first concordances that the students encounter have been carefully edited by me, to make sure that all five word classes of round (Noun, Adjective, Verb, Preposition, Adverbial particle) are represented, as well as the four possible word classes of that (demonstrative pronoun, relative pronoun, subordinator and adverbial intensifier). As can be seen from Concordance 1 (ten sentences containing round), this is not a KWIC concordance, but one that has been edited so that it contains full sentences. This has turned out to be a good way to start, since some students get rather distracted by seeing only sentence fragments. MicroConcord search SW: round (The Guardian/The Observer 2003) 1 2 3 4 5 6 7 8 9 10

Colin Montgomerie could find no consistency in his second round. I turned round and the turret was jus a well of fire behind me. There was fire everywhere. But the burglars have been round and stolen all the verbs and adjectives so we have to make do. I had night-vision goggles on and I looked down and I could see the round shape of landmines. And round every corner you could see Mount Teide itself – all 3,700m of it. In the 12th round he was trapped on the ropes, too exhausted to escape. David Starkey guides me round his impressive art collection in his own inimitable way. I can see him clearly – slim, pale, debonair, his hair smoothed back, a college scarf round his neck. Eventually I told him to give me back his front-door key, and phone first before coming round. ’… the infantry would come in between the tanks and the prisoners of war and round them up,’ Brisebois said.

Concordance 1. Determining the word class of round

The second corpus exercise, which aims to show students how variation can be studied by means of using corpora, is also an exercise in identifying the predicate verb and deciding whether the singular or the plural form is used – something


one would expect students to be able to do when they start university studies in English, but which inevitably causes some students considerable confusion. The starting point of the exercise is the grammar section on verb forms with collective nouns in British and American English that we study together in class. Briefly, the grammar says that whereas the singular form of the verb is used with collective nouns in American English, in British English they are often treated as plurals (Svartvik & Sager 1996: 150). These concordances have been edited to include only sentences where the collective noun functions as a subject and where the verb is either in the present tense or is a past tense form of be. The students examine ten sets of sentences from British and American newspapers in which five different nouns listed as ‘collective’ in the grammar appear, and then compare their results with the rule given in the grammar. Interestingly, one of the collective nouns listed is couple, and as the students will discover, this noun is normally treated as plural both in American English (see Concordance 2) and in British English. This leads to a discussion in class of how rules in the grammar should be interpreted – what does it mean when the grammar says that “there is a choice” between two alternatives? MicroConcord search SW: couple (The New York Times 1996) 1 for older people. There’s an elderly couple next door to us who haven’t used 2 Soraya Wiriadinata, an Indonesian couple who appear on seating charts at a 3 35 freelance writers busy, but the couple still run the operation from the 4 e labor would be cheaper. But the couple were committed to the mill and to 5 later, they were married. The couple now live in Brooline, Mr. Feuerst 6 at Oxford Brookes University. The couple were introduced by friends who two years 7 products manager. The couple were introduced at the wedding of 8 Rhyne, of the founding family. The couple have two children, the oldest, 129 stormed out of the apartment. The couple were not married and did not live 10 arpeted platform, upon which the couple has placed a chaise longue. A deep

Concordance 2. Investigating variation in the verb form used with collective nouns

Since each set of examples consists of ten instances of the structure, it is easy to calculate the percentage with which the singular and the plural verb forms are used with these nouns in the two varieties of English. Usually, there will be students who observe that ten sentences is too low a number for the results to be reliable, which is of course an important point for discussion. What I do when this point is brought up is extract a concordance from a much larger corpus of text (the newspapers on CD-ROM that I use contain roughly 45–50 million words per year), so that we are able to compare the percentage found in a small sample with that of a much larger corpus. In most cases, the results agree astonishingly well.

51

52

Solveig Granath

The third teacher-prepared exercise serves to caution students that when they do corpus searches, they will be likely to come across irrelevant hits. The exercise I use for this is one involving the split infinitive. The grammar used in the course includes a rule which says that although the split infinitive is not unusual in English, there are some adverbs that must never split the infinitive, namely not, merely and only (Svartvik & Sager 1996: 402). An important part of this exercise is to make students aware that to can be either a preposition or an infinitive marker, and to have them learn to distinguish between the two. The exercise is made up of six concordances of three short adverbs, one of which is not (the other two vary from term to term; some of the ones I have used are always, never, actually, really and sometimes). In the first ten sentences, the adverb occurs immediately before to, and in the next ten immediately after it; i.e. the latter may or may not constitute cases of split infinitives. The students’ task is to mark the sentences that contain infinitives, and cross out the ones that contain prepositions. Concordance 3 shows sentences where not follows to; as many as six out of the ten sentences actually do contain split infinitives, which gives rise to another discussion of how trustworthy the rules found in prescriptive grammars are. MicroConcord search SW: to not (The Guardian/The Observer 2003) 1 plains, is born of “a desperate need to not think, because thinking is so painful 2 brought and I looked round, trying to not meet anyone’s eye. At 7.30, on the 3 prehensive enough for newcomers to not feel ignored and succinct enough for 4 at. That was when I had the option to not go. ‘Grandpa ain’t going, I’m gonna 5 are people who are still committed to not accepting him. One blink at 6 lly,” he says at one point, referring to not only a personal alienation but a 7 r nearly 10 months before deciding to not make a bid. Mr Ahmed declined to 8 nation of things that all came down to not having much freedom. Record 9 right of way, and cars are restricted to not much more than walking pace. The 10 the Soviet leader Leonid Brezhnev to not let philosophical differences get in About 2,532,086 words were examined = one to not every 253,208 words

Concordance 3. Identifying relevant hits (split infinitives)

For each of the sets of 10 sentences, the number of words searched is given, which means that frequency of occurrence can be calculated for the structure. Thus, in addition to finding out that not can indeed split infinitives in English, the students also learn that this is rare, occurring as in this example only once per 400,000 words, whereas another concordance (not included here) shows that the ‘unsplit’ infinitive (not to V) is a common structure, occurring once per 12,000 words. Once this has been established, the question is raised why the split infinitive is used at all – when the grammar says it can never be used with not – which brings


up the question of differences between spoken and written registers, as well as stylistic and other pragmatic considerations. As can be seen from the above exercises, much can be achieved by relatively simple means, and even students who will never use corpora again can increase their awareness of the way the language works through the analysis of concordances and the follow-up discussions in class. A further use of corpora is to give informed answers to students’ questions. If students have some awareness of how corpus analysis is done, the teacher’s explanations will make more sense to them. This is the topic of the following section.

4.

Using corpora to answer student questions

When students ask questions about usage in class, I usually suggest an answer straight away, based on my own experience, and then as a follow-up do a corpus search which I present to the students the next time we meet. This has the advantage that the information is based on what the students are interested in finding out about, and it gives the teacher an additional opportunity to show students how corpora can provide information beyond what we find in reference books. The following three questions have come up in my own courses: Question 1: I have heard both in the spring and in spring. What is the correct way of saying it? Here, the British newspaper corpus actually does indicate that it is up to the speaker’s preference, as the two patterns are of similar frequency. Out of 53 examples, 24 had in the spring and 23 in spring. One pattern that emerged was that if spring is premodified by an adjective, the article is not used, and thus one says in early/late spring (there were altogether six sentences with this phrase). If the year is specified, the pattern in the spring of + year is somewhat more prevalent than in spring + year (with five instances of the former as compared to three of the latter). The different patterns that emerged from this simple query both gave the students the satisfaction that they in fact had a choice, as well as added information on how names of seasons collocate with adjectives and years. Question 2: Does it matter whether I say the Ganges River or the river Ganges? In order to answer this question, I consulted both American and British corpora, and the answer, for once, could be stated very clearly: American speakers will put ‘river’ after the name (and capitalize it in writing, i.e. the Ganges River), whereas in Britain, ‘river’ precedes the name and can be capitalized, but is more commonly written with a lower-case r: the river Ganges/the River Ganges. The most common

53

54

Solveig Granath

variant in both varieties, nevertheless, is to use just the Ganges, without the apposition river. Question 3: It says in the grammar that the United States and the United Nations are usually treated as singular entities. Is there a rule for which verb form is used with names in the plural? This turned out to be a very interesting question. Even large grammars that were consulted (by me) had no information on this, but the corpus search gave the following answers: Concord with plural names A. Family names – plural The Browns have left already. B.

Names of countries – usually singular The United States is still outperforming European markets. The Netherlands is to host the World Youth Championships in 2005. The Philippines was America’s first bona fide imperial possession. But: The Philippines have a terrorist problem.

C.

Groups of islands – usually plural The British Isles have had their own share of iconoclasts. The Canaries are geographically African rather than Spanish. The Balearic Islands are recruiting a new tourist police force. But: The Cayman Islands has implemented sophisticated plans for evacuation.

D.

Mountain ranges – usually plural The Alps are in the distance. The Himalayas are all pain and suffering. But: The Pyrenees was sunnier than the Alps yesterday.

E. Newspapers/magazines – singular The New York Times was, as usual, dryly factual.

The results indicate that with family names and names of newspapers, the speaker has no choice as to verb form. With the other three categories (names of countries, groups of islands, and mountain ranges), either the singular (names of countries) or the plural (groups of islands, mountain ranges) predominates, but the other form is also used occasionally. A survey such as this one can be quite helpful – not only to the students, but also to the teacher. Often, students ask questions about things that the teacher has not thought about (or, perhaps, has taken for granted).


Finding out the answer, and communicating it to the students with the help of authentic sentences, means that the language comes alive in the classroom in a way that is not possible by just relying on reference books and made-up examples. For students to be able to analyse concordances on their own, however, the most important part is, of course, hands-on training. Some examples of how this can be practised in the language lab are given in the following section.

5.

Corpus exercises in the language lab

The main problem with incorporating training in using corpora into regular courses is, as far as I can see, that only a limited amount of time can be set aside for actual hands-on exercises in the language lab. Still, for students to become proficient in using corpora, repeated teacher-led exercises are required. Unfortunately, it was only in the first year when I taught the course that I was able to have several sessions in the computer room; since then, I have had to limit the time I spend with the groups there. Initially, I met with small groups of students for three one-hour sessions; after that, the number of meetings was reduced to two one-hour sessions, and in the last four years, the students have had only one session in the computer lab, albeit two hours instead of one. The software I have been using for many years is MicroConcord, mainly because it is so simple that students can learn both simple and complex queries in a matter of minutes; with MicroConcord, it is also easy for the students to limit the number of hits to a small number that they can easily handle. I usually suggest that they get 20 examples when they start out – the number of lines that can be displayed simultaneously on the screen. In addition, I also try to give students a chance to familiarize themselves with the Collins Cobuild Concordance Sampler. Even though it takes a little longer to learn how to formulate queries with the Sampler than with MicroConcord, one of the greatest benefits with learning how to use it is that the students may access it from any computer. The corpora that I use with my students in the computer room consist primarily of British and American newspaper CD-ROMs (the British The Guardian/The Observer and the American The New York Times), which can be accessed from that room only, due to licensing restrictions.

. Free version at http://www.liv.ac.uk/~ms2928/software/ . http://collins.co.uk/Corpus/CorpusSearch.aspx

55

56

Solveig Granath

Regardless of the number of sessions the students have had in the lab, the tasks always proceed from simple searches to complex searches, ending with the students’ own queries. Below, I give a few examples of such tasks. One type of search that I often have students do in the computer lab is identifying the prepositions that collocate with certain nouns, verbs and adjectives, since prepositions generally cause learners problems. Moreover, it is rather difficult to glean this information from dictionaries. As an example, consider the preposition used after to be interested: interference from Swedish often makes Swedish learners select of rather than in after this adjective. A simple search of interested, however, clearly demonstrates that if interested is followed by a preposition at all, it is invariably in (see Concordance 4). This exercise also involves learning the simple command for ‘right sort’, which puts the words after the search word in alphabetical order. This makes it easy for the students to count the number of times in occurs after interested, as opposed to other alternatives. MicroConcord search SW: interested (The Guardian/The Observer 2004) 1 ur thoughts on renewable energy, we’re interested.”As to 2 and Simon in Monaco – but are always interested in a deal. 3 y in Kiss of Life is about love but I was interested in how people deal 4 n primary schools. “The Tories are only interested in more selection, not 5 money. This could mean that no one is interested in solar power, but it 6 een to secure new talent and Spurs are interested in the pair. 7 her father’s recantation. “I think what interested me,” says 8 nsultation and is not set in stone. Any interested parties can give their 9 the teenage pregnancy strategy will be interested to note that, in areas 10 sewage, then everybody’s going to be interested.”Whether the

Concordance 4. Investigating preposition use with interested

Interestingly, among our ten examples in Concordance 4, one sentence turns up where interested is followed by to. By now, students should know how to distinguish between the preposition to and the infinitive marker to, and be able to discern that in sentence 9, to is an infinitive marker. From experience, I know that interested to V is a pattern Swedish learners are not familiar with, so in a follow-up query, I have them type in interested to as search words. When the results have been displayed, students are asked to consider the meaning of the verbs following to (Concordance 5). Students usually have no problem seeing that these verbs all involve ‘receiving information’.


MicroConcord search SW: interested to (The Guardian/The Observer 2004) 1 2 3 4 5 6 7 8 9 10

who can’t drive but likes books, I’d be some of the senators may have been eterian shark, in Finding Nemo, will be the teenage pregnancy strategy will be ichard Webb As statisticians, we were are notoriously instransigent, so I was es: Guardian Online Mike Galvin I was Online Dr Desmond Turner MP I was was very hot, spat it out. Boswell was ess” to a modest buyer. I should be

interested to interested to interested to interested to interested to interested to interested to interested to interested to interested to

know whether it’s learn firstlearn that he was note that, in areas note that your scores read a recent read David Bolton’s letter see a Commons early see how the see Walker’s

Concordance 5. The structure interested to V

Making complex queries involves adding context words occurring at a certain distance to the right or to the left of the search word(s). In the lab sessions, this usually involves working with some aspect of grammar that we have discussed in class. As an example of a complex query, consider Concordances 6 and 7. The purpose of this exercise is to have students compare the use of the mandative subjunctive in British and American English. In short, the grammar says that whereas the mandative subjunctive is a regular part of American English, British English prefers the should-construction (Svartvik & Sager 1996: 76). Of all the ‘triggers’ of the mandative subjunctive, demand remains the most frequent, and the students are therefore instructed to use the search phrase demand* that (the asterisk will make sure all inflectional forms of both the verb and the noun are included) and as context words (specified to be 0,1, which means that the context word must occur immediately to the right of that) students type in he\she\it. By including only third person singular pronoun subjects, unclear cases (where the verb form is the same for the indicative and the subjunctive) are eliminated. Concordance 6 illustrates such a query.

57

58

Solveig Granath

MicroConcord search SW: demand* that (The Guardian/The Observer 2003) Collocation: he\she\it within one word to the right. 1 1 it politics after his local party demanded that he stand down following allega 2 1 s as intelligence officers and demanded that he accompany them for 3 1 ack for the coat-check man, demanding that he relinquish her 4 1 ng into the hairdressers and demanding that she be serviced instantly 5 1 D were occupied by activists demanding that it refuse to fund the pipe 6 1 it set out in a letter to Mr Blix demands that it be given advance notice of 7 2 haracter of his project rightly demands that he should be written back in 8 2 an Central Bank is bowing to demands that it should be ready to cut 9 4 write to the New York Times demanding that it sacked Muschamp. 10 5 noid, one professional man demanded, that it is impossible to love

Concordance 6. The mandative subjunctive in British newspaper text

In this task, students are also asked to sort their examples using numbers (the second figure on each line; 1 = subjunctive, 2 = should-construction, 3 = present indicative, 4 = past indicative, 5 = irrelevant hit). As Concordance 6 shows, out of nine relevant hits, six contain the mandative subjunctive and only two the shouldconstruction (in number 10, the that-clause is not subordinated to demanded but rather to the previous clause). Here we see that the claim made in the grammar cited above, that the should-construction is preferred in British usage, does not agree with the data. Concordance 7, on the other hand, turns out to corroborate what the grammar says, namely that in American English, the mandative subjunctive is preferred. This gives the instructor a chance to discuss on-going changes in English grammar. MicroConcord search SW: demand* that (The New York Times 1996) Collocation: he\she\it within one word to the right. 1 1 have soured on Mr. Colosio, ultimately demanding that he abandon his campaign. Mr. 2 1, whose work as a freelance journalist demands that he use his home phone all 3 1 extinction. Can he be blamed if his job demands that he venture to exotic places in 4 1 e scaffolding, he says, a police officer demanded that he move, saying, 'This is the big 5 1 been unwilling to accept prosecutors' demands that he first plead guilty to a felony. 6 1 ard remained watchful and that it had ‘demanded that he get help' running the 7 1 alled for boycotts of the company and demanded that it produce a plan to change its 8 1 has held and the American people are demanding that it continue to do so.' 9 1 to patronize him or his illness, lovingly demands that he behave and make music, and 10 1 bly, under President Plavsic's control, demanded that he and other top army officers be

Concordance 7. The mandative subjunctive in American newspaper text


Ideally, there should also be a third component in the session, namely time for students’ own queries. When I first started developing the course, the third session was devoted entirely to this. With only one 2-hour session in the computer room, which is what we have at present, many students do not have the time to get to the stage of making queries, and the ones who do get enough time to do their own searches rarely do very advanced queries. (As an example, consider a query posed in spring 2006: When do you use warm and when hot in English?) The ultimate goal of hands-on corpus exercises is of course for the students to be able to formulate queries and draw conclusions based on the concordances on their own, but in order for this to work well, students initially need a lot of guidance from the teacher, both in formulating queries and in analysing the results. Whether this can be made part of a course of course depends on the number of hours allotted for the course. In sum, using corpora as an integrated part of language classes with first-term university students will serve two main purposes: it will give students an insight into what we can learn from corpora and how it is done, and it will introduce new types of exercises for practising language skills, which will add variation to the language class. Nevertheless, for this to be successful, it is important that students find this way of working with language appealing. In the next section, I will present student reactions to the use of corpora, as demonstrated in the course evaluations.

6.

Students’ evaluation of corpus exercises

Except for the very first year when this course was taught, students have generally not found hands-on exercises in the language lab very difficult (see Table 1). The reason why one third of the students thought it was difficult in 1998 could be due to several things: corpus analysis was just being introduced into the course, and the students may have been given unclear instructions on how to do the tasks. It is also possible that many students found the third session, where they had to come up with their own queries, quite taxing. Since the spring of 1999, however, when the time in the computer room was reduced, close to half the students have reported that this part of the course is easy. Except for the very first year when corpora were used, the figures for the percentage of students who found this part of the course difficult have been quite low. What is obvious from my teacher perspective is that the students who face problems with the hands-on exercises are usually those whose language skills in general are quite poor at the beginning of the course.

59

60 Solveig Granath

Table 1. Students’ evaluation of the level of difficulty of hand-on corpus exercises Term

Easy

Intermediate Difficult

No answer

No. of students

Spring 1998 Spring 1999 Spring 2001 Spring 2002 Spring 2006

25% 47% 55% 51% 44%

35% 35% 42% 34% 51%

7% 10% – 13% –

57 60 44 57 43

33% 8% 2% 3% 5%

Table 2. Students’ evaluation of the usefulness of hands-on exercises Term

Useful

Not very useful

No. of students

Spring 1998 Spring 1999 Spring 2001 Spring 2006

13% 12% 23% 42%

87% 88% 76% 58%

51 54 44 44

Table 3. Do students believe they will use corpora in order to get information about grammar/vocabulary in the future? Term

Yes/probably

Maybe

No/probably not No. of students

Spring 1998 Spring 2000 Spring 2003 Spring 2006

20% 34% 31% 38%

24% 22% 36% 28%

57% 44% 33% 35%

51 41 39 40

Table 2 shows students’ evaluation of how useful they considered the computer room exercises to be. On the one hand, the results are rather depressing; the majority of students feel that they could do without this part of the grammar course. More encouraging, however, is the fact that the percentage of students who find this a useful part of the course has increased over the years. It was suggested at the Symposium on Corpora and Language Teaching in Göteborg, Sweden in December 2005 that this might be the result of students being increasingly computer-literate. Another reason might be that the tasks have been refined over the years and instructions for how to do them have become clearer. Another interesting question in connection with this is whether students have become aware of the usefulness of corpora as a language tool, and whether they believe they will use it in the future. Again, as Table 3 shows, there is a positive trend, so that the proportion of students who believe they may do so has increased from less than half to roughly two thirds of the group. It would of course be extremely interesting to know whether any students have in fact used corpora after the conclusion of the course. Since the instructor


does not meet with them after the grammar course is over, this has not been possible to determine. However, the course evaluation from spring 2006 contained the question Have you done any searches on your own after our session in the computer room? which three students answered in the affirmative, one even saying Yes, I actually do that quite often. In the evaluation, students are also encouraged add their own comments on the computer room sessions. Such comments help explain the underlying reasons for the figures presented in Tables 1–3. Here are some comments (with semester and year when the comment was made added in brackets). The fact that many students felt this was not too useful a part of the course is illustrated by the following: – As a first-semester English student one concentrates so hard on getting the basics of grammar straight that I believe many students could not see the “research training” for future purposes at this stage in their studies. (Spring 1998) – It was fun to work with grammar in another way but there were too few lessons in the computer room to get into it. (Spring 1998) – My impression is that you have to know a lot of grammar to be able to interpret the material. But at least now I know how to do it. (Spring 2000) – I was more worried about how to use the computer rather than finding out about grammar. (Autumn 2002) – The computer room session was just confusing and I felt that I could actually have skipped it because there was already too much to keep track of. (Autumn 2005) – I think it was a little too complicated. I didn’t really understand how to do it but I think it’s great that you’re able to find out about things like that. (Spring 2006) Among comments such as these, it is possible to discern two different groups of students: the ones who feel they have problems coping with the rest of the material in the course – to them, corpus analysis just adds to the general confusion – and those who actually find this part of the course quite interesting, but who feel they would need more time to acquire the skills needed to use corpora. It is also interesting to note that it is usually during the session in the computer room that many students realize the usefulness of knowing grammar (see the comment from spring 2000 above). A few comments were very negative; partly this has to do with the fact that not all students like working with computers:

61

62

Solveig Granath

– The computer room sessions were quite boring. I think many students felt that they could use the time better studying at home. Late in the afternoon in a room full of electricity… (Spring 1999) – I want to be able to communicate with someone or look in a book when I am looking for answers. (Autumn 2000) Other students expressed the opposite opinion; to them, this part of the course was not only useful but also added variation to the course: – Good idea with computer room sessions – useful – and something different from just having lectures. (Autumn 2002) – The computer room was useful. Forced you in another way to understand and answer questions. (Autumn 2005) – It was interesting, and it feels good to be able to use it as a tool for proving/ testing a point. (Spring 2006) So how useful is corpus work as part of an EFL syntax course? The results of a project conducted over a two-year period at Växjö University, Sweden, reported in Estling Vannestål & Lindquist (2007), corroborate many of the observations I have made above concerning the question. The purpose of the project was to determine in what way concordancing may influence students’ attitudes towards the study of grammar and whether the use of an inductive method such as this can lead to improvement in students’ proficiency. In their report, the authors concluded that it takes time to make skilled corpus users out of students at this level, especially when it comes to learning how to formulate their own queries. Furthermore, although many students appreciated finding out about this tool and realized its usefulness, particularly for their own writing, others, especially weak students, found it difficult or boring. Students’ lack of familiarity with inductive methods appears to explain why many students found this way of working difficult. Finally, it was obvious that working with corpora did not suit all types of learners, which led the authors to suggest that corpus activities might best be offered on a voluntary basis in EFL grammar courses on the university level. In the light of these results and my own experience, it is difficult to see how the use of corpora might be made appealing to all categories of students. So who, in the end, benefits from learning how to use corpora? This is the topic of the final section.

7.



As I hope I have shown in this paper, corpora can provide a useful and stimulating addition to second and foreign language learning. If the teacher has the skills required, corpus exercises can be used to vary classroom tasks, and examples of authentic language use can give students a much more vivid picture of the language than that which they get from their reference books. Even if students end up just trying out this tool once, the fact that they have used it at all (and seen it being used by the teacher) means that they may be ready at a later stage in their language training to become more proficient in using it on their own. It also introduces students to methods of linguistic research, which is often included as an objective in courses at university level. But who benefits from learning how to use corpora – and how can one become a skilled user? First of all, it is important that language teachers at university level know how to use corpora. Even though this is where corpus use is most widespread today, it is by far not the case that this is a skill that all language teachers have to date. Here, I am not speaking only about non-native speakers of the language, but also those teachers who teach their own native language. Unfortunately, ever since Chomsky’s criticism of corpora as a source of linguistic evidence, there has been a widespread belief that it is enough to have “native speaker intuition” and use introspection to determine whether a sentence is grammatical or not. However, computers can aid the user in discovering facts about the language that go beyond native speaker intuitions. Second, advanced students definitely benefit from working with corpora. This can actually provide them with the extra stimulus they need – something that tends to be neglected when we teach large groups of students. At our university, students in their second semester may opt to take a course where they do a small corpus study (morphology, phraseology or syntax), and in the third semester, they have the choice of doing a corpus investigation for their term paper. For both, it is an advantage if they have some familiarity with corpus analysis before they start. Many of the students who have used corpora for their term papers have done excellent studies and have shown both independence and creativity in their use of corpora. Intermediate students can also profit from hands-on exercises using the computer. For less advanced students, more teacher guidance is needed, but if the exercises are on the right level, corpus work can help raise their awareness of structures.

63

64 Solveig Granath

Finally, and most important – this is probably the best tool we can provide future language teachers with. Ideally, teacher training courses should include a course in corpus analysis in one of the last years of university studies. Students who have completed their language courses can be assumed to be equipped with the skills necessary for interpreting and evaluating corpus data. In such a course, it should also be possible to make the tasks directly relevant to classroom work, so that students would raise questions, based on problems they have come across when giving feedback to their own students. Such a course must clearly be designed to include exercises in using corpora to find out about lexicon, phraseology and grammar patterns, as well as background reading on corpora and corpus linguistics. The same kind of course can be offered to language teachers already in service – this will give them access to a “native speaker consultant” who does more than any native speaker could do. Or, in the words of Alan Durant (2005: 42): Asking a native speaker is not like consulting a reference book or corpora of language data. With a speaker’s intuition, it is not the language as a whole, or as represented by a wide-ranging sample, that is being consulted, but one particular speaker’s repertoire and awareness of her or his varied and changing language environment. Idiosyncrasies of personal history may complicate what appears to be the common wisdom of the language community. (…) Only with data from many native speakers can variant models combine into an overall, polyphonic image of a public language.

References Aston, G., Bernardini, S. & Stewart, D. (eds). 2004. Corpora and Language Learners. Amsterdam: John Benjamins. Davies, M. 2004. Student use of large, annotated corpora to analyze syntactic variation. In Aston et al., 257–269. Durant, A. 2005. There may be regular guys but there are no regular native speakers: Lexis and native-speaker-like competence. In The Power of Words: Studies in Honour of Moira Linnarud, S. Granath, J. Miliander & E. Wennö (eds), 31–44. Karlstad: Karlstad University Press. Estling Vannestål, M. & Lindquist, H. 2007. Learning English grammar with a corpus: Experimenting with concordancing in a university grammar course. ReCALL 19(3): 329–350. Granath, S. 1998. Using corpora in teaching English syntax to EFL students at the university level. In Proceedings of Teaching and Language Corpora (TALC 98), 87–92. Oxford: Keble College.

. See e.g. Tsui (2004), who demonstrates how corpus evidence can contribute in an important way to raising EFL teachers’ awareness of the use of lexical items and language structures.


Ketteman, B. & Marko, G. 2002. Teaching and Learning by Doing Corpus Analysis. Amsterdam: Rodopi. Kirk, J. 2002. Teaching critical skills in corpus linguistics using the BNC. In Ketteman & Marko, 155–164. Mair, C. 2002. Empowering non-native speakers: The hidden surplus value of corpora in continental English departments. In Ketteman & Marko, 119–130. Römer, U. 2004. Comparing real and ideal language learner input: The use of an EFL textbook corpus in corpus linguistics and language teaching. In Aston et al., 151–168. Römer, U. 2005. Progressives, Patterns, Pedagogy. A Corpus-driven Approach to English Progressive Forms, Functions, Contexts and Didactics. Amsterdam: John Benjamins. Stewart, D., Bernardini, S. & Aston, G. 2004. Introduction: Ten years of TaLC. In Aston et al., 1–18. Svartvik, J. & Sager, O. 1996. Engelsk universitetsgrammatik. Stockholm: Almqvist & Wiksell. Tsui, A. B. M. 2004. What teachers have always wanted to know – and how corpora can help. In How to Use Corpora in Language Teaching, John Sinclair (ed.), 39–61. Amsterdam: John Benjamins.

65

Oslo Interactive English Corpus-driven exercises on the Web* Signe Oksefjell Ebeling

This paper describes the interactive learning environment, Oslo Interactive English, developed at the University of Oslo in 2004/2005. Oslo Interactive English (OIE) offers a wide range of corpus-driven exercises suitable for undergraduate students of English. The exercises are divided into seven grammatical topics and there are four different exercise types, viz. multiple choice, gapfill, error correction and open. To do the exercises a corpus must be consulted. The main aims of OIE are to offer both on- and off-campus students an interactive learning environment and at the same time introduce them to the use of corpora as a language learning resource.

1.

Introduction

As part of the University of Oslo’s initiative in strengthening its flexible learning profile, Oslo Interactive English received one year’s funding in 2004. Flexible learning environments focus on learning activities that are characterised by their innovative use of various forms of Information and Communication Technology. An additional requirement made by the university was that, although ICT-supported, the projects should not only cater for distance learning but also encourage on-campus students to use such tools. Oslo Interactive English (OIE) is thus an interactive learning environment for on-campus undergraduate students of English and contributes to blended learning. Blended learning combines technology-based materials (OIE) with face-to-face sessions (on-campus teaching). As traditional classroom teaching does not always give students enough practical language training, OIE was created as stand-alone supplementary exercise material to on-campus teaching, aiming to encourage students to practise * I would like to thank Maria Leedham for useful comments on an earlier version of this article.

68 Signe Oksefjell Ebeling

grammar in their own time in a user-friendly web environment. In a setting where corpus-based research and corpus-based teaching have been nurtured over the past decades, viz. the English department at the University of Oslo, it was a natural step to merge exercises on the web with corpus linguistic methods. Since language corpora are in themselves an electronic resource, it seems only logical that they should form an integral part of e-learning. OIE is highly interactive, involving corpus searches and reflection on the part of the students to solve online grammar tasks. This paper describes the main features of OIE, including an overview of the exercise types used and a short presentation of the corpus and its search engine. It will also focus on the user’s path from language learner to language researcher. Finally, in order to log student activity on the web site, the paper looks at user statistics for OIE over a ten-month period.

2.

Aims and background

The main aim of Oslo Interactive English was to give students the opportunity to actively use a corpus in their language studies, combining their knowledge of grammar with practical language skills. At the same time as providing English language training, OIE also introduces the students to the use of corpora in order to improve their writing skills and general knowledge of English. In the past decade or so, we have seen a growing interest in the use of corpora in language teaching. As pointed out by Hunston (2002: 13), “[i]ncreasingly, language classroom teachers are encouraging students to explore corpora for themselves [...], allowing them to observe nuances of usage and to make comparisons between languages”. However, the main focus of what has been written on the subject has been on corpus use in the classroom under the supervision of a corpus-competent teacher. Discovery-learning and data-driven learning are terms used to describe the use of corpora in language teaching/learning. The idea is that, by observing language in context, students discover for themselves how language is used, be it a particular word, phrase, construction, etc. Although OIE does not differ significantly from models set up for corpus teaching in the classroom, there are two main differences that should be mentioned: – OIE can also be used as a resource outside the classroom – OIE provides “set” answers (i.e. suggested answers in order to supply automatic correction on the fly)

Oslo Interactive English

OIE can be accessed at any time and students can get exercises automatically corrected whenever they want. To answer the questions posed in the exercises they have to consult a corpus; they are immediately given a suggested answer after submitting their own answers. No discussion of alternative or wrong answers is offered, although the students are pointed to online grammars where they can read about the topic in question. It is relatively easy to point to the weaknesses of such an approach, particularly from a pedagogical point of view. The presence of a corpus-literate teacher is often stressed as a basic criterion in data-driven learning (cf. Mauranen 2004: 100). Moreover, experience shows that, for various reasons, the introduction of corpora to students may be challenging. Many students do not immediately take to such a method even in a classroom setting (cf. Kennedy & Miceli 2001). Some struggle with the search options and search interface, while others have difficulty with the analysis of the corpus output. Admittedly, the challenge is even greater in an online environment such as OIE than it is in the classroom. However, the OIE exercises give the students an opportunity to explore language learning with corpora in their own time and at their own pace. OIE is user-friendly, having been constructed with the user in mind, and students are given step-by-step instructions as they familiarise themselves with the corpus. The students are also carefully guided through the exercises; in some cases they are told which search strings to use and what to look for. As Leech (1997: 8) puts it: “[i]n the course of learning how to manipulate corpus searches, the student will need, at least initially, to be supplied with sample tasks of exercises”. The aims of OIE come together in this process, so to speak; guidance on how to use corpora is combined with interpretation of the output and as a result students may improve their language skills. It should also be stressed that OIE is a supplement to on-campus teaching, and not a replacement. This is very much in line with what Bernardini says: ... corpus-learner, and indeed corpus-teacher interaction are not replacements for learner-learner and teacher-learner interaction, but rather should be seen as an added value offered by corpus-aided discovery learning. (Bernardini 2004: 32)

All the exercises available through OIE are data-driven and have been created with the help of a corpus. As pointed out by Sinclair (2004: 7) “[c]orpus evidence is essentially indirect, which means that it cannot be taken at face value but must go through a process of interpretation”. It is important that students grasp this fact about corpus studies early on so that they can quickly learn how to interpret the results. As we shall see below, the structure of Oslo Interactive English is such that students have to familiarise themselves with the use of a corpus in order to do the exercises. This entails nothing more sophisticated than the ability to know how to search for words and phrases in a corpus, how to sort and interpret the output and

69


make sense of concordance lines. Once these challenges are overcome, students have every opportunity of becoming successful “discovery learners”.

3.

The structure of OIE

3.1

The exercises and exercise types

The exercises constitute the main ingredient of OIE. They focus on English grammar in use and have been divided into seven main categories. These categories coincide with topics that are covered in the undergraduate on-campus teaching in the form of either lectures or seminars. To easily distinguish the various topics from one another, each one is represented by a different colour: – Adjectives & adverbs (red) – Nouns & verbs (orange) – Function words (yellow)

– Vocabulary (green) – Text (turquoise) – English vs. Norwegian (blue)

In addition, there is a Miscellaneous category (purple), which includes mixed exercises and also exercises based on previous years’ exam questions. These seven categories appear as a menu on the web pages, where the students choose exercises according to which grammatical topic they wish to practice. The main categories include exercises on various subtopics, each of which contains eight exercises that all follow one of four different exercise types, viz. Multiple choice, Gapfill, Error correction and Open. In every set, exercises 1 & 2 are multiple choice, 3 & 4 are gapfill, 5 & 6 are error correction and 7 & 8 are open. This ordering of tasks ensures learning progress both as regards language skills and corpus skills. Before the students are comfortable with the corpus as a resource it is recommended that the exercises are done in successive order. However, once the students’ corpus skills have improved, the exercises can be answered in any order. The use of the corpus as a resource is most prominent in the multiple choice and open exercises, where the corpus has to be consulted in order to answer the questions. First, let us have a look at the multiple choice exercises.

3.1.1 Multiple choice In the multiple choice exercises the students are given specific corpus search tasks to perform. This exercise type is therefore first and foremost an introduction to the use of corpora. However, the students must also interpret the concordance lines and make use of their knowledge of English grammar, as illustrated in the example given in Figure 1.


Figure 1. Screenshot of a multiple choice exercise in OIE (the full exercise includes 8 questions)

Figure 2. Example of gapfill exercise in OIE

Figure 1 also shows that the students can search the corpus directly from the exercise page. The corpus search results, or concordances, will load in a new window so that the students can work with the exercises in one window and the concordances in another. In the corpus search window students can sort the concordance lines according to certain criteria which we will discuss more fully in Section 4.1. Further, the screenshot in Figure 1 tells us that the subtopic is “it and there” and that the exercise number is 2. The subtopic banner is blue, indicating that the exercise is from the category “English vs. Norwegian”.

3.1.2 Gapfill In the gapfill exercises, it is up to the students whether and how they want to use the corpus. Again the students have the corpus at their disposal and can quickly access it and search contexts similar to the ones found in the exercise sentences, e.g. search for contexts with it and there as shown in Figure 2. It could be argued that this kind of exercise provides a grammatical drill that in only a few cases necessitates students’ use of the corpus. However, this is also very much part of the OIE setup; students are meant to practice their language skills with or without the support of a corpus. In exercises where students can use

71

72

Signe Oksefjell Ebeling

Figure 3. Example of an error correction exercise on spelling

other means than the corpus to arrive at the correct answer, they are free to do so. Nevertheless, as all the exercises in Oslo Interactive English have been made on the basis of real data – they are genuine samples of English usage that can be found in the corpus – the students will benefit from making use of the corpus. This will also ensure that they become more confident corpus users.

3.1.3 Error correction In the error correction exercises, the students are asked to identify the errors and correct them directly on the web. Ideally, they should also be asked to give a reason for the changes they make, but the nature of the automatic correction system does not allow for individual answers. If the exercises are used in a classroom setting, however, this is the kind of task the tutors could set their students. Many of the erroneous sentences that the students are asked to correct have been taken from the Norwegian part of the International Corpus of Learner English (NICLE), which is to say they are genuine student errors. The choice of error correction as part of the OIE exercise material is to some degree based on the frequent use of error correction as a much-used exam question type. As was the case with the gapfill exercises, it is up to the students whether they want to use the corpus to do this type of exercise. Figure 3 illustrates part of an error correction task. 3.1.4 Open exercises The open exercises are “open” in the sense that the students are free to write full answers; they are not restricted to e.g. multiple choice or gapfill, and generally they do not get too many leads as to how to search the corpus. Also, the answer they give will not be corrected in the same way as for the other three exercise types. While the other exercises are corrected on the fly by a computer program, the open exercises cannot be dealt with automatically. The students will have to evaluate their answers on the basis of a suggested answer provided when they click the answer button. The open exercises also require that the students use the corpus. And in some cases more corpus skills are needed than was the case with the multiple choice


Figure 4. Example of an open exercise

Figure 5. Example of key to exercises with links to grammatical topic and corpus search hints

exercises. The reason for this is that guidance as to which searches to perform is not given to the same extent here. Furthermore, the open exercises are more theoretical in nature, i.e. in addition to corpus skills, the students need more theoretical knowledge of English grammar to solve the tasks given. Figure 4 gives an example of an open exercise on verbs ending in -ing vs. the to-infinitive.

3.2 The score and key to exercises I will now move on to the automatic correction of the exercises. When students submit their answers by clicking the submit button found at the bottom of every page, a small score window pops up, telling students how many correct answers they have achieved. From the score window there is a link to the key-to-exercises page. In addition to the actual answer, the key-to-exercises page gives a web link to the topic in question. Thus, in the example shown in Figure 5, they can read about it and there at a particular location on the web. There is also a link that takes you to the OIE corpus search hints page, which offers a short introduction to how to search the corpus. The fact that the exercises are corrected upon submission gives the students an immediate indication of how they are doing in a particular area; it also suggests to what extent they were successful in their interpretation of concordance lines.

73

74


4.

The corpus

Integrated within the structure of OIE is, as we have seen, a corpus. The exercises are corpus-driven and the idea is that, if the students use the corpus, they should be able to answer all questions correctly, or adequately in the case of the open type – provided that their knowledge of English grammar corresponds to their level of study. To ensure that all answers can be found in the corpus, a certain control is required in the sense that the developer decides what changes can be made to the corpus and the search tools. Thus, a corpus was put together for use with OIE instead of making use of existing concordancers / corpora freely available on the web. The OIE corpus is password protected and at the moment it is only available to students at the University of Oslo. The corpus comprises approximately 7 million words of 20th century fictional and non-fictional English. The texts include books, reports, speeches, and film scripts. The main bulk of the corpus is taken from an early version of the Longman/Lancaster English Language Corpus (Summers 1993), which was acquired by the English department in Oslo in the early 1990s. However, the original Longman corpus has been trimmed down by removing reduplicated texts and by excluding some text types, such as recipes and poetry, in addition to novels from the 19th century. Speeches and film scripts were not part of the original Longman corpus. The speeches are mainly political speeches by world leaders, downloaded from the internet. Finally, the film scripts were also taken from various sites on the internet, and include both British and American films. The film scripts (and to some extent the speeches) were added in order to provide the corpus with some spoken material, although it is a text category that only imitates speech, and has been more accurately termed “written to be spoken as if not written” (Gregory and Carroll 1978: 42). In order to include more recent data, some non-fictional texts, all of which are government reports from various English-speaking countries, have also been added for the purpose of OIE. The corpus is stored in a simple text file with minimal mark-up. Each text includes a header where meta information about the text, such as country of origin and year of publication, can be found. The relatively simple form of mark-up adopted in the OIE corpus does not allow users to perform advanced searches. In the following section we will take a look at the various search options available in OIE.


4.1 Searching the OIE corpus Another important feature of OIE, in addition to the corpus, is the search interface. At the same time as it was decided that a controllable corpus was preferable to one already in existence, the advantage of creating an OIE search interface was also recognised. This makes it easier for the developer to make the interactive dialogue between student and corpus as simple and smooth as possible. Without controlling both the corpus content and search interface, there is always a risk that the search results may change and thus do not correspond to the expected answer to any given exercise. Students should also be made aware that a corpus containing authentic native language data may include samples of grammatically incorrect usage. This, of course, is the same for all corpora and it should be stressed that the students’ grammatical knowledge and interpretation of the concordances must always be what guides their conclusions based on corpus data. The students have the corpus at their disposal at all times, either directly from the exercise page, as we have seen, or from a separate corpus search window. When the corpus search facility was set up for this project, it was considered important to make the search options intuitive and easy for new users. As pointed out in Section 2, students need to be able to interpret the corpus results in order to arrive at the correct answer. This process also involves another important aspect of corpus use, namely that “[t]he [corpus] evidence needs to be interpreted with some awareness of the design of the software query package” (Sinclair 2004: 7). To help beginners to quickly become proficient users with guidance only in writing from the web, very few search options are available. In a simple search you can type in a single word (e.g. “pass”) or a string of words (e.g. “pass the ball”). Or, the slightly more advanced search option, where you can use a wildcard (e.g. “pass*”). You can also sort the concordance by keyword, by the word immediately preceding the keyword (i.e. sort by left word) or by the word immediately following it (i.e. sort by right word). These relatively restricted search options do of course not introduce students to more advanced uses of a corpus (e.g. lemma searches, contextual data searches including year of publication, genre, etc.), but it is believed that the simple search interface makes it easier for the student to get it right and not give up as a result of a more complex set of search options. One of the main aims of the project was to introduce students to this method so that they can become advanced and enthusiastic corpus users in the future. In the search facility for OIE, concordance lines are presented in an accessible way. The search results are given in a KWIC concordance and a link takes you to

75

76


Figure 6. OIE wildcard search, concordance sorted by the word to the right of the search string

a very limited set of contextual data, available from the header of each text of the corpus, for each line in the concordance. The example given in Figure 6 shows a KWIC concordance of a wildcard search where the concordances have been sorted by right word. The search string used in the example is “it’s * who” and returns 9 hits in the corpus.

5.

The users

Username and password are only given to students at the University of Oslo, either by their tutor or the OIE web administrator. The OIE environment is aimed at undergraduate students of English; this is reflected in the degree of difficulty presented in the exercises as well as the choice of grammatical topics to be practiced. In the following we will focus on the language learner and how he or she can benefit from an interactive learning environment such as OIE. Then, in Section 5.2 we will take a look at some usage statistics for OIE.

5.1

From language learners to language discoverers

Through OIE, students at undergraduate level are introduced to the use of corpora in language studies and are also given a snapshot of how linguistic research may be carried out with the assistance of a corpus. OIE illustrates that the integration of corpora in language teaching is appealing since the step from research method to language teaching is relatively short, paving the way for students to become researchers. However, as pointed out by Bernardini (2004: 23), far from all stu-


dents have research ambitions and may therefore benefit more from a “learning as discovery” approach since this “encourages learners to follow their own interests whilst providing them with opportunities to develop their capacities and competences” (ibid.). Such an exploratory way of learning is very much present in OIE, where student corpus searches as well as their interpretation of the concordances must be precise in order to give satisfactory answers to the exercises. Although students are guided in the sense that they are given particular problems to solve or questions to answer, they will have to use the corpus to discover for themselves what the concordances tell them about grammar and language use. With the help of a corpus, students acquire knowledge and skills in the language they are studying. Introducing undergraduate students to this model for language learning may serve as a stepping stone for language studies and research at more advanced levels. Corpus linguistics can thus be seen as a learning model where students take responsibility for their own learning. OIE offers a setting of problem-based learning where the students’ discoveries and interaction with a corpus give them a learning curve very much triggered by their own activity and motivation. Further, by being exposed to authentic data, students are given a more nuanced encounter with language than traditional grammars are able to offer. Working with authentic data will also make students aware that not all native language use is grammatically correct according to grammatical standards for English. This emphasises the importance of interpreting corpus data in the light of previously acquired knowledge of English grammar. The interactive learning model adopted in OIE gives the language learners the opportunity to be language discoverers / researchers. OIE may be seen as an introductory course in corpus use that can be of value later in life either within or outside a university setting in the form of linguistic research, language teaching, or in the use of corpora as a language resource on a par with other resources, e.g. dictionaries, phrase books, and TOEFL material. It is also thought that discovery learning and problem-based learning are models that have a motivating effect on students. Nevertheless, motivation is seen as a challenge, particularly since OIE was set up primarily as a supplement to oncampus lectures and seminars. It was therefore thought to be important to map the site activity; the following section shows how this was done and how successful OIE has been to recruit students using relatively few resources.

5.2 Usage statistics Two different approaches were chosen in order to map the OIE site activity: web statistics provided by the department and a separate, more detailed OIE log. First,

77

78


Table 1. Departmental Web statistics of OIE over two semesters1 Semester

Month

Rank in departmental Web statistics

OIE’s 1st semester

August 2005 September 2005 October 2005 November 2005 December 2005 January 2006 February 2006 March 2006 April 2006 May 2006

not listed 13th 9th 11th 9th 54th 61st 25th 24th 61st

OIE’s 2nd semester

the usage statistics of the department were tracked to get an impression of the overall use of OIE since its launch in August 2005. Table 1 follows OIE’s popularity over two semesters among all pages whose URL contains the address of the Department of Literature, Area Studies, and European Languages (ILOS) at the University of Oslo. The semester started mid-August 2005, so OIE does not figure on the usage statistics for that month. In September, though, the OIE exercise page had quite a few visitors and was listed as the 13th most popular web page in the department. In October OIE had moved up to 9th place. The popularity of OIE is fairly constant until the beginning of a new semester in January 2006, then it climbs throughout the semester before another dip by the end of the semester in May 2006. One factor to keep in mind when reading these statistics is the increase of total web pages on the department server with 1,012 in September 2005 and 6,744 in May 2006. Thus, behind the rank within the department it is revealed that the number of actual visitors/hits has been fairly stable, in fact we have seen an increase in OIE visitors, going from 259 hits in September 2005 to 364 in May 2006. OIE’s own log was set up in October 2005, returning statistics on exercises that have been displayed and exercises that have actually been submitted and corrected automatically. It does not give absolutely foolproof statistics, but at least robots accessing the site have been disregarded. It does not include statistics on the “open” exercises, since the answers are not submitted and can therefore not be logged. Although the figures shown in Table 1 were not too disappointing, they become far from satisfactory when entering into the details of the OIE log. Table 2 . Web statistics for ILOS were generated by Webalizer Version 2.01 (http://www.mrunix.net/ webalizer/).


Table 2. OIE log – User statistics October 2005 – August 2006 Unique users correcting exercises Users correcting more than 5 different exercises

112 10

shows how many unique users submitted and corrected exercises in the period from October 2005 to August 2006. First, on the positive side, we can note that, while in December 2005, only 12 unique users were recorded as having submitted exercises, 100 “new” users have been added since then. Of these 112 users only 10 had submitted and corrected more than 5 exercises; however, 5 of these users had corrected more than 10 exercises. From the log it can also be seen that there is quite a discrepancy between people only looking at (displaying) the exercises and people actually doing them. Of more than 900 unique users who had visited one of OIE’s exercise pages, only 112 chose to submit any of them for correction. Of course, this discrepancy can, at least partly, be explained by the fact that only students at the University of Oslo are entitled to a password, and therefore are able to search the corpus. The exercise pages are freely available and thus people might visit the pages, learn that they cannot get a password to the corpus, and choose not to attempt the exercises. Of the three exercise types that have been logged, multiple choice is by far the most popular one; while a multiple choice exercise was corrected 138 times, a gapfill exercise and an error correction exercise were corrected 71 and 51 times, respectively. The multiple choice exercises require the use of the corpus in order to get the answers right, so these statistics bode well for the students’ use of the corpus. Finally, the log also shows which grammatical topics were favoured by the users. For the exercise types multiple choice and gapfill, the categories Adjectives vs. Adverbs and Vocabulary scored highest. In the case of multiple choice, also Nouns & Verbs did relatively well. The most popular topic for error correction was the category English vs. Norwegian. This is useful information both in the sense that we see which topics the students find the time to practice and also in that it indicates which topics need to be made more visible. It also suggests that more exercises are needed on popular topics. One of the main conclusions to be drawn from the statistics is that the relatively nice figures provided by the general log service at the university belies the real success rate of OIE. The next step would therefore be to draw up an agenda for how to recruit more students to use OIE on a regular basis and convince them that it is for their own benefit.

79


5.2.1 How can student participation be increased? With the statistics information in mind, an important goal for OIE in the future must be to increase student participation. This will also contribute to another important issue, namely to get a full evaluation of OIE, both from the tutors’ and students’ point of view. However, in the following we will concentrate only on how to motivate students to become more frequent users of OIE. Potential solutions to boost OIE usage statistics include: 1. 2. 3. 4. 5.

Raise students’ awareness of OIE Give a 5–10-minute demonstration of OIE in a lecture Create a (weekly) discussion forum within OIE Integrate OIE into on-campus teaching Make OIE obligatory for on-campus students

The suggestions are listed according to the effort required in introducing them. While it may be too weak an initiative simply to try to raise the students’ awareness of OIE by e.g. sending out flyers or emails, it may be too strong to make it obligatory, as suggested in the last point. If possible, a discussion forum with a “live” tutor could be set up and maybe spark interest in the students. As students become more used to accessing the Internet, including search engines, discussion forums and brief demonstrations in a lecture may have a positive effect. However, at present, experience from other projects suggests that integration into on-campus teaching would be the way to go (cf. Granath this volume; Kennedy & Miceli 2001). Even when a tutor is present, many students are not immediately comfortable with a method that makes use of a corpus; problems may relate both to the search interface or the concordance lines and how to interpret them. To have a tutor present hence seems to be the best solution so that students will not give up if they cannot find an immediate answer to their problem on the web pages. This last suggestion seems to be too ambitious at the moment. However, as suggested in point 2 above, a demonstration of OIE will be given in a lecture early on in the first semester. It remains to be seen whether this will encourage more students to use OIE and indeed use a corpus both to solve OIE exercises and in other aspects of their studies. Another feature that may be seen as a motivating factor is the integration of a corpus-based mind game into OIE. Trivial Corpus Pursuit (TCP) was also developed with funding from the University of Oslo and is based on the idea behind the more famous board game Trivial Pursuit™. The aim of TCP is to answer grammar-related questions with the help of a corpus in order to fill a pie with six wedges. Figure 7 shows that a question in the red category Adjectives & Adverbs has been answered correctly, indicated by the wedge in the pie. Furthermore, a


Figure 7. Screenshot of the corpus-based mind game Trivial Corpus Pursuit

new question in the category Vocabulary has been loaded and the required corpus search has been performed. TCP makes use of the same corpus and search interface as OIE and, in terms of equipment, requires only a web browser to be played. TCP follows in the footsteps of OIE by offering similar corpus-driven exercises, but differs in that it is set up as a game and is therefore thought to bring an additional motivating factor to student participation.

6.

Concluding remarks

From a pedagogical point of view it is quite evident that the integration of corpora in language teaching / learning is beneficial to students. By exposing students to authentic data in the form of concordance lines, corpus-driven exercises allow for a more student-centred approach to learning, giving them the opportunity to evaluate data and draw their own conclusions. As a stand-alone online resource, OIE seeks to develop students’ skills and to make them reflect on the data in grammatical terms. In many ways, OIE can be characterised as an introductory course in the use of corpora, providing exercises to improve language skills. On the other hand, it may be argued that corpus analysis is such a demanding task that only the good students will benefit from it. However, the structure of . For a fuller account of Trivial Corpus Pursuit, see Ebeling (2006).

81

82


OIE seeks to capture students of all levels, by offering a range of exercises with a varying degree of difficulty. In some exercises, very little analysis is involved, and the answers are relatively obvious once you see the concordance lines. Other exercises, notably the open ones, are more demanding and may be better suited for more motivated students, showing a particular interest in English grammar. OIE has now been available to two cohorts of undergraduate students, one in the autumn semester of 2005 and one in the spring semester of 2006. It is hoped that a full evaluation of OIE will take place in the near future to analyse student benefits both as regards improved language / grammar skills and corpus skills. If the result of such an evaluation is positive, flexible learning environments of this kind could be even more integrated into on-campus undergraduate teaching in the future.

References Bernardini, S. 2004. Corpora in the classroom: An overview and some reflections on future developments. In J. M. Sinclair (ed.), 15–36. Ebeling, S. O. 2006. Trivial Corpus Pursuit: A corpus-driven mind game for language learners. In Ringer i vann. Fleksibel læring – Kvalitetsreformen i praksis, S. K. Koch, (ed.), 93–104. Oslo: University of Oslo. Gregory, M. & Carroll, S. 1978. Language and Situation. Language Varieties and Their Social Contexts. London: Routledge and Kegan Paul. Hunston, S. 2002. Corpora in Applied Linguistics. Cambridge: CUP. Kennedy, C. & Miceli, T. 2001. An evaluation of intermediate students’ approaches to corpus investigation. Language Learning and Technology 5(3): 77–90. Leech, G. 1997. Teaching and language corpora: A convergence. In Teaching and Language Corpora, A. Wichmann, S. Fligelstone, T. McEnery & G. Knowles (eds), 1–23. London: Longman. Mauranen, A. 2004. Spoken corpus for an ordinary learner. In J. McH Sinclair (ed.), 89–105. Sinclair, J. M. (ed.). 2004. How to Use Corpora in Language Teaching. Amsterdam: John Benjamins. Sinclair, J. M. 2004. Introduction. In J. M. Sinclair (ed.), 1–10. Summers, D. 1993. Longman/Lancaster English Language Corpus – Criteria and design. International Journal of Lexicography 6 (3): 181–208.

Websites Oslo Interactive English – http://www.hf.uio.no/ilos-dy/OIE Trivial Corpus Pursuit – http://www.hf.uio.no/ilos-dy/OIE/tcp.php International Corpus of Learner English – http://cecl.fltr.ucl.ac.be/Cecl-Projects/Icle/icle.htm

Corpus research and practice What help do teachers need and what can we offer? Ute Römer

Despite the progress that has recently been made in the field of corpus linguistics and language teaching, it is not clear what impact corpora have actually had so far on English language teaching practice. Corpus researchers often claim that corpus linguistics can make a difference for language teaching and that it has an immense potential to improve pedagogy, but perhaps do not focus enough on the interface of research and practice. They do not make sufficient efforts to reach practitioners, especially teachers, with the ‘corpus mission’, do not know enough about the needs of teachers, and do not show them where corpora can help them solve everyday problems. The present chapter aims to address these issues. It centres on a teachers’ needs analysis carried out to capture important aspects about the situation of teachers, their problems and wishes they might have. A questionnaire was devised that covered topics such as the quality of existing teaching materials, authenticity in language teaching, and the teachers’ language competence and exam marking. This chapter will present selected results from the teachers’ survey and discuss where corpus linguistics (or corpus linguists, rather) could, and should perhaps, offer help to teachers, hence showing them that corpus research can have an impact on pedagogical practice and that corpora can actually make a difference.

1.

Introduction: Research and practice

There has recently been a lot of research activity in the field of corpus linguistics and language teaching (see, for example, the contributions in Aston et al. (eds) 2004; Connor and Upton (eds) 2004; Hidalgo et al. (eds) 2007; Kettemann and Marko (eds) 2002; and Sinclair (ed.) 2004). Some considerable progress has unquestionably been made in providing corpus-based descriptions of the language for learners and of the language of learners. Researchers in the TaLC (Teaching and Language Corpora) tradition, including myself, keep claiming that, taken

84

Ute Römer

seriously, corpus linguistics can make a difference for language learning and teaching and that it has an immense potential to improve pedagogical practice. Johansson (2007), for example, notes that “corpora can be a good means of teaching and learning” and that “[t]hrough corpora we now have a new generation of dictionaries and grammars reflecting the way language is actually used.” An obvious question, however, that is rarely asked in this context (though see Mukherjee 2004 and Tribble 2000) is ‘What effect has corpus research actually had on the English language teaching practice so far?’ I would hence suggest that we ask ourselves what we (as corpus researchers) know about the situation of teachers, and that we focus more on the interface of corpus research and teaching practice. An important task in this context will be to find out what we could do for language teaching practitioners. In this paper I will report on a survey among qualified English language teachers at secondary schools (Gymnasien) in Germany. The aim of the survey was basically to elicit information on the teachers’ work situation and to collect comments on problems and wishes they might have. In the following sections I will first turn to a discussion of selected results from this survey and then sketch possible ways in which corpus linguists could react to these results.

2.

Helping teachers with their everyday problems

In Römer (2006) I provide an overview of pedagogical corpus applications and devise a wish list for future activities in the field. Among other things, this wish list addresses the issue of ‘missionary work’ by which I refer to the idea of ‘spreading the word’ about corpora and what they can do among language teaching practitioners. I argue, for instance, that “[a] next important step will be to reach teachers” and that “[t]he task of the corpus linguist would be to show the teacher where corpora can help her/him solve everyday problems [...]” (Römer 2006: 128, 129). To do this, however, we first have to determine what help teachers actually need and what their “everyday problems” are. As long as we are not in touch with practitioners and ask them about their situation, it will remain difficult to provide them with tailored tools and resources for corpus-inspired teaching and learning. . Since 1994, conferences on this topic (Teaching and Language Corpora) have been organised every other year. The first eight events in this series took place in Lancaster (1994 and 1996), Oxford (1998), Graz (2000), Bertinoro (2002), Granada (2004), Paris (2006), and Lisbon (2008). For details about the coverage of TaLC, readers are referred to the proceedings volumes of TaLC 1 to TaLC 6 (Wichmann et al. 1997; Botley et al. 1996; Burnard and McEnery 2000; Kettemann and Marko 2002; Aston, Bernardini and Stewart 2004; and Hidalgo, Quereda and Santana 2007).

Corpus research and practice

2.1 What help do teachers need? – Selected results from a survey In order to find out more about the needs of English language teachers, I carried out a survey. I teamed up with an experienced secondary school teacher (L1: German, subjects: English and Dutch) to devise a questionnaire that would cover central aspects of the challenges teachers face in their job and that would allow the informants to make comments on required or desired support in dealing with these challenges. This questionnaire covered topics such as the quality of existing teaching materials, the availability of appropriate materials and reference tools, and the teachers’ language competence and exam marking. The questionnaire consisted of 13 statements related to these topics, e.g. “Existing teaching materials provide sufficient support in teaching vocabulary.” The informants were asked to agree or disagree with the statements, by ticking one of the following four options: “I fully agree”, “I partly agree”, “I disagree”, “I don’t know”. Below each statement there was a “Comments?” box which enabled the teachers to state their opinion and to justify their decisions on agreement or disagreement with the statements. About 120 questionnaires were sent out, 78 of which were completed and returned. The informants were all non-native speakers of English (with German as their L1) and taught pupils from grades 5 to 13 (first-year English to English A-level pupils) at secondary schools in different parts of Germany, mainly in Lower-Saxony and North Rhine-Westphalia. All teachers responded to all statements in the questionnaire and about a third of them made extensive use of the comments option, which provided some interesting insights about their situation and about where they would like to get help. Let us now look at some results from the survey. As mentioned above, one of the topics addressed in the questionnaire was the grading of pupils’ in-class essays (or exams in general). The teachers were, for example, asked to agree or disagree with the statement “When marking exams, I feel insecure at times about right/wrong decisions and would need the advice of a native speaker.” The pie chart in Figure 1 shows that only 9 out of 78 teachers (11.54%) disagree with this statement. More than 88% of the informants either fully or partly agree that marking exams poses problems that relate to their language competence. The teachers were also asked to write down which resources they use . In retrospect, I think that it would have been better to use a grading scale from 1 to 6 (or at least a fifth option “I partly disagree”) in order to allow for more graded responses. . Many thanks are due to the following teachers who helped me distribute and collect the questionnaires: Simon Chlouba, Ulla Dvořák, Gabriela Fellmann, Stefan Hanke, Christiane Lütge, Hermann-Josef Müller, Thomas Müller, Walburga Müller, Sven Naujokat, Claudia Reichel, Ulrich Salden, Annika Seligmann, Susanne Wolff, and Birgit Ziegenmeyer.

85

86 Ute Römer

“disagree” 11.54%

“fully agree” 37.18%

“partly agree” 51.28%

Figure 1. Teachers’ agreement and disagreement with the statement “When marking exams, I feel insecure at times about right/wrong decisions and would need the advice of a native speaker” 25.00%

21.84% 21.84%

20.00% 10.14%

7.80%

s.

s

ct di

bo

a or

ca

ar

rp co

co llo

m m

3.90% 1.56%

ok

U ’s gr a

an Sw

te

rn

PE

et

s In

ue ag lle

p es tiv a n

rs ke a e

co

ct

io

na

rie

s

0.00%

di

3.90%

3.90%

5.00%

n

10.00%

tio

15.00%

Figure 2. Resources consulted by teachers in exam marking (multiple answers were possible)

in such situations to answer the question “Can you say that in English?” Figure 2 summarises the responses given on that issue. Most frequently mentioned as resources (each by 21.84% of the teachers) were (mainly monolingual) dictionaries and native speakers. Several teachers referred to “native speaker friends” who they can phone or email when they need advice on language problems. Other teachers appear to be in a less fortunate situation. One of them noted that s/he would often like to check with a native speaker whether a certain construction or collocation in a learner’s essay was acceptable or not, but that “unfortunately, native speakers are not always immediately available when you need them”, so all s/he is left with


in such cases is her/his non-native speaker intuition. On a more critical note, one teacher wrote that “even native speakers don’t always know the answer” and that “some kind of online service would probably be faster and better – and definitely more readily available (24 hours)”. Another informant questioned the reliability of native-speaker judgements, claiming that “a native speaker is by no means a guarantee for correct English”. Apart from dictionaries and native speakers, several teachers also ask colleagues for help (10.14%) or consult the Internet (7.8%). It did not become clear, however, whether this mainly refers to the use of Google or other commercial search engines, or whether “the Internet” stands for dictionaries that are available online. Both tools (Google and online dictionaries) were actually mentioned on the questionnaires, but only by two teachers. Collocation dictionaries, grammar books, and Michael Swan’s Practical English Usage (Swan 2005) were also mentioned as useful resources by a small number of teachers (each resource by 3.9%). Two informants said that they sometimes consult corpora (one of them mentioned the BNC but did not specify which version or interface s/he uses) but only if they “have enough time”. Interesting comments were also made on the question of the availability of reliable resources that teachers could consult to solve the “Can you say that in English?” problem. While the distribution of answers presented in Figure 3 gives the impression that the majority of teachers are reasonably happy with what is available (less than 22% of them expressed disagreement), the actual comments convey a different picture. Several informants noted that it often takes too long to find answers in dictionaries, and that dictionaries are usually not a very useful resource when they need information on the acceptability of word-combinations. One teacher commented that “simply looking up solutions for problems with complex items (e.g. phrasal verbs) is not always straightforward.” Another teacher thought that “it would be ideal to have an electronic device (perhaps a database or computer program?) where you could type in two words to check whether a preposition or collocation actually fits.” A further topic addressed in the questionnaire was the quality of the coursebooks currently used in the ELT classroom. The teachers were asked to state whether the teaching materials they had at their disposal provided sufficient support in teaching grammar and vocabulary, or whether they had to put a considerable amount of work into the creation of materials themselves. As the chart in Figure 4 shows, only 10.26% of the informants think that the coursebooks they use actually offer enough support in teaching grammar and teaching vocabulary. The rest of the teachers either fully (39.74%) or partly (50.00%) agree that they have to put a considerable amount of work into the creation of materials themselves. More revealing than the mere percentages, however, are again the teachers’

87

88

Ute Römer


“don’t know 5.31%


“partly agree” 37.18%

Figure 3. Teachers’ agreement and disagreement with the statement “There are enough reliable resources available that I can consult in the correction of class tests”


“partly agree” 50%


Figure 4. Teachers’ agreement and disagreement with the statement “Teachers have to put a considerable amount of work into the creation of teaching materials themselves”

comments on this issue. The comments hint at a number of shortcomings of the textbooks that are currently used, for example that “they offer too few exercises” “they do not include enough texts which capture real-life situations” “they focus mainly on British English and not enough on American English” “they marginalise language variation” “there is a lack of interesting, authentic, and up-to-date material” “vocabulary is not presented in authentic collocations” “the texts do not catch the pupils’ interest and are often inappropriate for the target group” – “textbook language cannot keep track of language change since real language develops fast and it takes too long to develop a new textbook series”. – – – – – – –

Teachers who want to present their pupils with better materials are hence forced to invest time in searching for more interesting and more authentic texts and in


creating additional exercises. Judging from the teachers’ comments, this is particularly true for classes on more advanced levels (seven teachers explicitly stated that a lot of work is left to them in the creation of materials for grades 11 to 13, i.e. the final three years before the A-levels). On the issue of authenticity, one teacher noted that “many of the textbook dialogues sound staged” and that they “may not be useful to train the learners’ communicative competence”. On a similar note, another informant observed that “the texts in the coursebooks have been constructed, so that certain words and structures can be introduced.” S/he went on stating that, as a result of this, s/he is “constantly searching for more natural and more appropriate texts that include examples of a particular structure and at the same time catch the pupils’ interest” (s/he mentioned song lyrics, such as Midge Ure’s “If I was” to introduce conditionals). By way of summarising the results of the survey, we can say that what teachers seem to want and need is – a wider range of better teaching materials with more interesting, longer and genuine texts, and with more exercises – textbooks that reflect actual language use and cover variation – textbooks that keep track of language development – tools which enable an ad-hoc creation of suitable materials, e.g. worksheets on particular lexical items or structures – help with exam marking – more reliable and more quickly accessible resources or reference tools that help them solve the question “Can you say that in English?”, perhaps “an online service or database”, as one teacher wrote – an always available native-speaker consultant, and, last but not least – more time. In the following sections I will respond to these needs and wishes from a corpus linguist’s perspective.

2.2 What help can corpus linguistics offer? Now that we know more about the problems teachers face in their job and about the needs they have, we can discuss where and how corpus resources and the corpus linguist could come in to offer help. The problems and wishes mentioned by the teachers centre around four key issues which will now be dealt with in turn: better teaching materials, support in creating materials, native speaker advice, and more reliable reference tools.

89

90 Ute Römer

2.2.1 Better teaching materials and support in creating materials The launch of the first COBUILD learners’ dictionary in 1987 (Sinclair et al. 1987) indicated an important turn in pedagogically-oriented lexicography and had a strong impact on dictionaries which have been compiled since then. Soon other publishers followed the COBUILD team in basing their dictionaries on large corpora, and nowadays it is hard to find an English language dictionary that is not corpus-based. Unfortunately, corpora and the results of corpus research have not yet exerted such a strong influence on EFL textbooks. The coursebook series that are used in schools across Germany, for instance, are to my knowledge not based on corpus evidence and still include a large share of invented texts and constructed dialogues (see the example from English G 2000 A1 in (1); cf. also Römer 2005: 277–280). (1)

JENNY Have you got a red felt-tip, Debbie? DEBBIE No, I haven’t. JENNY Of course you’ve got a red felt-tip. Look. DEBBIE Oh, you’re right. Sorry. Yes, I have, Jenny. Here. JENNY Thanks. BEN Have you got a garden, Nick? NICK Yes, we have. DEBBIE Jenny, have Sita and her family got a garden? JENNY No, they haven’t. BEN And have you got a swimming-pool? NICK No, of course we haven’t. (Schwarz 1997: 30)

The good news is, however, that the first entirely corpus-based EFL textbook series are currently being compiled, mainly for the American market. Worth mentioning here are McCarthy et al.’s Touchstone volumes and Barlow’s CorpusLAB books (cf. McCarthy et al. 2005; Barlow and Burdine 2006). It is to be hoped that materials of this kind, which include genuine instead of invented language and take corpus findings into account, will soon also be available to the teachers who participated in our survey. What is certainly needed in this context is a close collaboration between corpus researchers and textbook publishers. While the corpus linguist can provide access to a large number of authentic language samples, insights into language patterning, and information about the distribution of items across different language varieties and registers, the materials writer has the required knowledge about the pedagogical principles that have to be observed when it comes to making decisions about the sequencing and, in particular, the presentation of lexical items and grammatical structures.


What is the missing word in each of the following sentences – ‘speak’ or ‘talk’? I’m not here to _____ on behalf of the theatre at all. Are you able to _____ English fluently? I’d like to _____ about something with you. I managed to put her off that idea, managed to _____ her out of that. So you’re free to _____ your mind. Excuse me could you _____ up just a little bit? Yes yes er thank you. I will _____ to David about it as well. Mothers and fathers _____ differently to sons and daughters. Men tend to _____ like that, don’t they? You’re not allowed to _____ for the rest of the week.

Figure 5. An example of a data-driven learning exercise on the near-synonyms “speak” and “talk”

A possible response to the teachers’ call for “more exercises” could be the creation of suitable data-driven learning (DDL) materials for learners on different proficiency levels. DDL exercises present the learner with concordance samples directly taken from a corpus or with selected instances of lexical-grammatical items in context. Based on this “used” language material (in Brazil’s 1995 terms), the learner is encouraged to explore the use of words and phrases in an autonomous way, which means that s/he learns “by discovery” (cf. Bernardini 2000, 2002 and 2004) and becomes a better noticer and more language aware. Using a somewhat more controlled type of data-driven learning (cf. Johns 1991 and 2002), teachers could also present their pupils with sets of pre-selected concordance lines, for example from concordances of near-synonyms or, in general, expressions that are commonly confused but show differences in usage (e.g. “speak” and “talk”, “if ” and “whether”, or “big” and “large”). In the selected lines, the teacher could then blank the searchwords and ask the learner to fill the gaps. An example of such a DDL task, based on the items “speak” and “talk”, is displayed in Figure 5. Although exercises of this type can already be found in some publications by applied corpus linguists (cf. e.g. Sinclair 2003; Tribble and Jones 1997) and on the Web, it will be an important future task for the corpus researcher to create more DDL materials that address particular language items (especially items which cause constant problems for learners) and that could be used directly in the EFL classroom. A further step towards integrating corpus-based exercises and discovery learning in language teaching would be to train teachers in working with corpora . Links to a number of data-driven learning exercises can be found on Tim Johns’ and Passapong Sripicharn’s websites at http://www.eisu.bham.ac.uk/johnstf/timconc.htm and http:// www.geocities.com/tonypgnews/units_index_pilot.htm (accessed 1 November 2006).

91

92

Ute Römer

so that they could design required materials themselves whenever they needed them. I would argue that corpora and concordance packages present very useful resources for the creation of exercises that motivate the learner and promote her/ his language awareness. What is hence required are courses taught by corpus linguists and tailored to the teachers’ needs. Such courses could either be part of the general teacher training programme that every English language teacher has to do, or they could be offered to practicing teachers in the form of “advanced teacher training workshops”, as sketched by Mukherjee (2004: 240). However, since the schedules of teachers and teacher trainees tend to be rather full already, it might be more sensible to start the ‘corpus mission’ at the level of initial teacher training at universities and introduce future teachers to corpora and their pedagogical potential at this early stage. An obligatory university course in “Corpus linguistics and language teaching”, for instance, could show students where linguistic research findings can be relevant for the teaching practice and how they could profit from the use of corpora in language teaching and learning in their future job. My experience with teaching a course of this kind to third and final year students at the University of Hanover during the summer term 2006 was very positive, and the feedback I received from the course participants (all students intending to enter the teaching field), especially on the relevance of what we discussed for their careers as teachers, was overwhelming. As mentioned above, the creation of suitable materials should, however, not be left entirely to the teacher. What is undoubtedly required is more support from the research community in developing resources for the hand of the teacher. To give just one example, tools similar to Mike Scott’s Guardian Keywords Database (see Scott 2002), perhaps based on text collections of texts that are even more relevant for the learner than Guardian newspaper texts (e.g. texts from spoken domains), could be immensely valuable for the teacher and greatly facilitate the creation of suitable materials on particular word fields (e.g. sports, travelling, politics, etc.). Another very useful resource might be an online platform which combines a database with templates of DDL exercises and corpora with an easyto-use search interface. If the exercise templates (e.g. a gap-fill template) could then be filled with raw or filtered concordance output from simple queries of the teacher’s choice, an ad-hoc creation of DDL materials on a wide range of lexicalgrammatical phenomena would be possible.

2.2.2 Native speaker advice and more reliable reference tools One of the things our survey has highlighted is the teachers’ need for help in the correction of class tests and learners’ essays. It is hence not surprising, that a frequently mentioned item on the informants’ wish lists was the “always available native speaker”. I would argue that this is a wish that corpus linguistics can easily


fulfil – after all, a corpus usually contains the language output of thousands of native speakers (unless we are dealing with a learner corpus). In a similar vein, Barnbrook (1996: 140) has very appropriately described the computer corpus as “a tireless native-speaker informant, with rather greater potential knowledge of the language than the average native speaker”. What is more, unlike native speakers, corpora are available 24 hours a day, seven days a week and thus enable teachers to check language points and find information on common word-combinations or the typical usage of an item whenever they want (on this issue, see also Boulton and Wilhelm 2006). Of course, Web search engines can basically perform the same trick, which suggests that teachers could also google for answers to the question “Can you say that in English?” However, since the Web is anything but a principled and controlled (or controllable) collection of texts (which is true for linguistic corpora), the output of Google and other commercial search engines has to be treated with a lot of caution, in particular with respect to the sources of Web-attested examples. I would therefore refer teachers who desire native-speaker advice on seemingly odd word-combinations in their pupils’ exams or essays to one of the larger corpora that are searchable online. To give just one example, a teacher who stumbles over a phrase like “he was forced to make a diet” and is unsure about more idiomatic ways of expressing what the learner wanted to say here could simply look up “diet” in a corpus and check the phraseology of the word. The teacher would then find out that the patterns preferred by native speakers are “to go on a diet” and “to be on a diet” and that “make a diet” is an unusual collocation (see the concordance sample from a BNC search in Figure 6). The corpus can thus be seen as a valuable reference tool for the teacher – a tool which is perhaps more reliable than many of the available teaching materials or (non-corpus-based) usage handbooks. Corpus searches can provide answers

. Barnbrook’s statement links up nicely to one teacher’s criticism of the reliability of nativespeaker judgements (see Section 2.1). . Worth mentioning here are the COBUILD Concordance and Collocations Sampler that provides access to part of the Bank of English (see http://www.collins.co.uk/Corpus/CorpusSearch.aspx, accessed 1 November 2006) and the following search interfaces to the British National Corpus: BNC Simple Search (http://thetis.bl.uk/lookup.html, accessed 1 November 2006), Phrases in English, PIE (http://pie.usna.edu/simplesearch.html, accessed 1 November 2006), and COCA, the Corpus of Contemporary American English (http://www.americancorpus.org, accessed 17 October 2008). . This is an attested example taken from the German part of ICLE, the International Corpus of Learner English (cf. Granger, Dagneaux & Meunier 2002). This subcorpus of ICLE comprises argumentative essays written by 3rd and 4th year university students and has a size of about 234,000 words.

93

94 Ute Römer

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20

his is, and erm she went pression. I mean you may be ittle tiddy bit, I might be when you’re supposed to be colate muffins you won’t be it’s funny since I’ve been adache every, cos I’ve been I’ve been up when I’ve been I could say well I could go parts . Well no she did go ut the trouble is when I go e you went well you must go d I said I can’t wait to go for a we, we’d eat, we’d go it could be. But I’m going I’m just too fat, I’m going have Cos I’m But I’m going some chocolate though. I’m How do you know I’m

on on on on on on on on on on on on on on on on on on on on

a, a diet, I don’t know what, ho and see a wonderful cr a diet you try and get rid of a diet The best of cheese a diet. will you? You know. No a diet er, it is, I must be g a diet and she turned round a a diet I mean smelling, smok a diet, but, er and keep on t a diet, she used to go and se a diet, the first place I a diet I had been on diets a a diet! Oh alright then and a diet! and then we wouldn’t e a diet I’m on a diet this wee a diet and you have sufficie a diet, anyway. So I a diet I’m on a diet this week I a diet again. This year I wen a diet Put put three down fr a diet?

Figure 6. An extract from a BNC concordance of “a diet” (sorted to the left)

to many language-related questions, especially questions about the combinability of words and the appropriateness of collocations. An alternative to encouraging teachers to get their hands on online corpora would be to fulfil the teachers’ wish for an “online service or database” that provides answers to tricky language questions. Such a service already exists in Hong Kong. English language teachers at Hong Kong schools have access to a website called TeleNex, short for “Teachers of English Language Education Nexus” (Tsui 2005: 337), which serves as a forum for all sorts of language-related questions they may have. TeleNex is supported by a team of language specialists from the University of Hong Kong who use corpus evidence to respond to questions discussed in the TeleNex “Language Corner” (Tsui 2004 and 2005). An example query posted by a teacher and taken from Tsui (2005) is given in (2) below. To respond to the teacher’s problem, the language specialists would summarise corpus findings on the use of the items in question and provide selected concordance lines to highlight their central usage patterns. The TeleNex website presents a wonderful aid to language teachers in that it supports them in dealing with everyday language-related problems and helps raising their language awareness. It would be good to see a development of more websites of this kind so that similar services could be offered to a larger number of teachers in different countries around the world.


(2) Hi there, A student asked my colleague when to use “big” and when to use “large”. I couldn’t give her a definite answer. Is there some kind of rule that we should follow? (Tsui 2005: 340)

3.

Conclusion: Research for practice

The present paper has aimed to build a bridge between corpus research and pedagogical practice. By focussing on the situation of language teaching practitioners and their needs, it has explored the role of corpus linguistics (and corpus linguists) in supporting teachers in their work. The results of a survey among 78 practicing English language teachers indicate that many of the problems teachers have could be solved, at least partially, if they were introduced to some basic corpus resources and received more support from corpus researchers. I here agree with Kennedy (1992: 368) who notes that “[m]any teachers need persuading that corpus linguistics can make a contribution to their professional activity.” A major task for us (corpus researchers) will hence be to ‘spread the word’ about corpora and to inform teachers about what is already available (e.g. online corpora, corpus-based materials and reference works, collections of data-driven learning exercises). Like Mukherjee (2004: 243) who rightly stresses “the need for a large-scale popularization of corpus linguistics among English teachers”, I think that it is about time that we convince more practitioners that they (and their pupils) could greatly benefit from consulting corpora. In order to reach teachers, universities could for example organise “open days” and offer lectures and workshops on issues that directly relate to the teachers’ problems and needs. An existing and very successful event of this type is the annual “Teachers’ Day” hosted by the Universität des Saarlandes in Saarbrücken, Germany. In the past three years a number of well-known corpus linguists, including Michael Hoey, Joybrato Mukherjee, Anne O’Keeffe and John Sinclair, gave talks and workshops at this event and introduced teachers to what corpora could do for them. Another task for the linguist will be to carry out more research ‘at the interface’ and find out more about the needs of language teaching practitioners. If we know what teachers want and what help they need, we can react accordingly. The . Information on the Teachers’ Day can be found at http://fr46.uni-saarland.de/index. php?id=467 and http://www.uni-saarland.de/fak4/norrick/esf/2006/TeachersDay06.htm (accessed 1 November 2006).

95

96 Ute Römer

job of a teacher clearly is a difficult and stressful one, so if corpus linguists can do anything to offer support, they should do it.

Acknowledgements I would like to thank Karin Aijmer for inviting me to speak at the Gothenburg symposium on “Corpora and Language Teaching”, 2–3 December 2005. I am also grateful to the participants at this symposium, and to my audience at the 7th TaLC (Teaching and Language Corpora) conference in Paris, 1–4 July 2006, for stimulating questions and suggestions after my presentations, and to Michael Barlow for helpful comments on an earlier version of this paper.

References Aston, G., Bernardini, S. & Stewart, D. (eds). 2004. Corpora and Language Learners. Amsterdam: John Benjamins. Barlow, M. & Burdine, S. 2006. American Phrasal Verbs (CorpusLAB Series). Houston TX: Athelstan. http://www.corpuslab.com (accessed 1 November 2006). Barnbrook, G. 1996. Language and Computers. A Practical Introduction to the Computer Analysis of Language. Edinburgh: Edinburgh University Press. Bernardini, S. 2000. Systematising serendipity: Proposals for concordancing large corpora with language learners. In Rethinking Language Pedagogy from a Corpus Perspective, L. Burnard & T. McEnery (eds), 225–234. Frankfurt: Lang. Bernardini, S. 2002. Exploring new directions for discovery learning. In Teaching and Learning by Doing Corpus Analysis. Proceedings of the Fourth International Conference on Teaching and Language Corpora, Graz 19–24 July, 2000, B. Kettemann & G. Marko (eds), 165–182. Amsterdam: Rodopi. Bernardini, S. 2004. Corpora in the classroom: an overview and some reflections on future developments. In How to Use Corpora in Language Teaching, J. M. Sinclair (ed.), 15–36. Amsterdam: John Benjamins. Botley, S., Glass, J., McEnery, T. & Wilson, A. (eds). 1996. Proceedings of Teaching and Language Corpora 1996. Lancaster: University Centre for Computer Corpus Research on Language. Boulton, A. & Wilhelm, S. 2006. Habeant corpus – they should have the body: Tools learners have the right to use. Paper presented at 27e Congrès du GERAS Cours et Corpus, Université de Bretagne Sud, 23–25 March 2006. Brazil, D. 1995. A Grammar of Speech. Oxford: OUP. Burnard, L. & McEnery, T. (eds). 2000. Rethinking Language Pedagogy from a Corpus Perspective. Frankfurt: Lang. Connor, U. & Upton, T. A. (eds). 2004. Applied Corpus Linguistics. A Multi-dimensional Perspective. Amsterdam: Rodopi.


Granger, S., Dagneaux, E. & Meunier, F. (eds). 2002. International Corpus of Learner English. Handbook and CD-ROM. Louvain-la-Neuve: Presses Universitaires de Louvain. Hidalgo, E., Quereda, L. & Santana, J. (eds). 2007. Corpora in the Foreign Language Classroom. Amsterdam: Rodopi. Johansson, S. 2007. Using corpora: from learning to research. In Corpora in the Foreign Language Classroom, E. Hidalgo, L. Quereda & J. Santana (eds), 17–31. Amsterdam: Rodopi. Johns, T. F. 1991. Should you be persuaded – Two samples of data-driven learning materials. In Classroom Concordancing [ELR Journal 4], T. F. Johns & P. King (eds), 1–16. Johns, T. F. 2002. Data-driven learning: The perpetual challenge. In Teaching and Learning by Doing Corpus Analysis. Proceedings of the Fourth International Conference on Teaching and Language Corpora, Graz 19–24 July, 2000, B. Kettemann & G. Marko (eds), 107–117. Amsterdam: Rodopi. Kennedy, G. 1992. Preferred ways of putting things with implications for language teaching. In Directions in Corpus Linguistics. Proceedings of Nobel Symposium 82. Stockholm, 4–8 August 1991, J. Svartvik (ed.), 335–373. Berlin: Mouton de Gruyter. Kettemann, B. & Marko, G. (eds). 2002. Teaching and Learning by Doing Corpus Analysis. Proceedings of the Fourth International Conference on Teaching and Language Corpora, Graz 19–24 July, 2000. Amsterdam: Rodopi. McCarthy, M., McCarten, J. & Sandiford, H. 2005. Touchstone Student’s Book 1. Cambridge: CUP. Mukherjee, J. 2004. Bridging the gap between applied corpus linguistics and the reality of English language teaching in Germany. In Applied Corpus Linguistics. A Multi-dimensional Perspective, U. Connor & T. A. Upton (eds), 239–250. Amsterdam: Rodopi. Römer, U. 2005. Progressives, Patterns, Pedagogy. A Corpus-driven Approach to English Progressive Forms, Functions, Contexts and Didactics. Amsterdam: John Benjamins. Römer, U. 2006. Pedagogical applications of corpora: Some reflections on the current scope and a wish list for future developments. Zeitschrift für Anglistik und Amerikanistik 54(2): 121–134. Schwarz, H. (ed.). 1997. English G 2000 A1. Berlin: Cornelsen. Scott, M. 2002. Picturing the key words of a very large corpus and their lexical upshots or getting the Guardian’s view of the world. In Teaching and Learning by Doing Corpus Analysis. Proceedings of the Fourth International Conference on Teaching and Language Corpora, Graz 19–24 July, 2000, B. Kettemann & G. Marko (eds), 43–50. Amsterdam: Rodopi. Sinclair, J. M. 2003. Reading Concordances. London: Longman. Sinclair, J. M. (ed.). 2004. How to Use Corpora in Language Teaching. Amsterdam: John Benjamins. Sinclair, J. M. et al. (eds.). 1987. Collins COBUILD English Dictionary for Advanced Learners. London: HarperCollins. Swan, M. 2005. Practical English Usage, 3rd edn. Oxford: OUP. Tribble, C. 2000. Practical uses for language corpora in ELT. In A Special Interest in Computers: Learning and Teaching with Information and Communications Technologies, P. Brett & G. Motteram (eds), 31–41. Whitstable: IATEFL. Tribble, C. & Jones, G. 1997. Concordances in the Classroom: A Resource Book for Teachers. Houston TX: Athelstan.

97

98 Ute Römer

Tsui, A. B. M. 2004. What teachers have always wanted to know – and how corpora can help. In How to Use Corpora in Language Teaching, J. M. Sinclair (ed.), 39–61. Amsterdam: John Benjamins. Tsui, A. B. M. 2005. ESL teachers’ questions and corpus evidence. International Journal of Corpus Linguistics 10(3): 335–356. Wichmann, A., Fligelstone, S., McEnery, T. & Knowles, G. (eds). 1997. Teaching and Language Corpora. London: Longman.

part iii

The indirect corpus approach

Themes in Swedish advanced learners’ writing in English Jennifer Herriman and Mia Boström Aronsson

Advanced non-native learners’ writing may contain very few grammar mistakes but still appear somewhat deviant from native writing. This study investigates to what extent this deviance may be a result of differences in the non-native learners’ organization of information. Using a Systemic Functional Linguistic approach, it compares theme selection and thematic variation in argumentative writing by Swedish advanced learners in the International Corpus of Learner English and native speakers in the Louvain Corpus of Native English Essays. The results show that the Swedish learners’ themes tend to contain more interpersonal information. The Swedish learners also tend to overuse cleft constructions, especially to thematize new information and to express personal opinions. This results in a more emphatic style of persuasion and greater involvement and interaction with the reader.

1.

Introduction

Advanced non-native learners may make few or no mistakes at all in their writing as far as their grammar and lexicon are concerned, and yet their writing may still appear to sound somewhat “nonnative” and deviant. What exactly causes this deviance may be difficult to pinpoint. It could, for instance, be the result of subtle differences in the ordering of the informational content or a difference in the frequency in which a certain linguistic structure is used. Now that large learner corpora, such as the International Corpus of Learner English (ICLE) (see further Granger 1993, 1994, 1998a, 1998b), are available, it is possible to investigate such differences in frequency and usage by comparing large quantities of . The International Corpus of Learner English consists of a number of different subcorpora of approximately 200,000 words each of mainly argumentative writing produced by learners of English at university level, each subcorpus representing one mother tongue.

102 Jennifer Herriman and Mia Boström Aronsson

non-native speakers’ (NNS) and native speakers’ (NS) writing. Mauranen (1993), for instance, has found that English texts written by Finnish NNS students tend to differ in their thematic progression from those written by NS. Similarly, Lorenz (1998) has found that texts by German NNS students tend to differ in their organization of information, with elements that are “new, relevant and noteworthy enough to be intensified” (1998: 62) placed earlier in the sentence than would normally be expected. This paper presents some of the chief findings from a comparison of how Swedish advanced learners of English and native English speakers organize information in their argumentative writing (Boström Aronsson 2005). The main focus of the comparison is on their selection of theme and thematic variation. Adopting a Systemic Functional Linguistic approach (SFL) (Halliday 2004; Thompson 1996), the analysis follows Halliday’s definition of theme as the element which is placed first and “which serves as the point of departure of the message; it is that which locates and orients the clause within its context” (Halliday 2004: 64). The remaining part of the clause is seen as the Rheme, the focus of the message. The selection of the theme of a clause serves an important function in the structuring of the text as a whole. It also contributes to the meaning of the clause itself, as Halliday illustrates in the contrasting selection of themes in (1a), in which a halfpenny is the theme and the clause says something about a halfpenny, and (1b), in which the smallest English coin is the theme, and the clause says something about the smallest English coin. (1) a. a halfpenny is the smallest English coin b. the smallest English coin is a halfpenny (Halliday 1994: 38, themes in italics)

A comparison of theme selection and thematic variation by Swedish and advanced learners of English and native speakers, then, will reveal whether there are differences in what the nonnative writers tend to select as theme and, if this is the case, whether these differences would explain some of the “nonnative soundedness” of advanced nonnative writing which cannot be attributed to observable mistakes in grammar or lexicon. The texts compared are argumentative essays in the Swedish subcorpus of the International Corpus of Learner English (SWICLE) and the Louvain Corpus of Native English Essays (LOCNESS). The former are written by native speakers of Swedish in their second year of university studies of English at the Universities of Lund, Göteborg, and Växjö and the latter by British and American students, mainly at university level. In the next section we will look further into the classification of themes and thematic variation.

2.

Themes in Swedish advanced learners’ writing in English 103

Themes and thematic variation

From a Systemic Functional Linguistic perspective, themes are classified according to which metafunction they serve in the language, i.e. whether they serve the experiential, interpersonal and textual metafunctions (Halliday 2004: 29–31 and 79–87). Experiential themes, as illustrated in (1) above, are thus concerned with semantic aspects of language and can be described within the system of transitivity in terms of processes including participants and circumstances. Interpersonal themes are concerned with interaction between people, such as expressions of the speaker’s/writer’s attitude and assessment of probabilities, and whether an utterance is a statement, question, or command. The former are, for instance, modal adjuncts, such as probably, perhaps, certainly, etc. which serve as modal interpersonal themes, and the latter are finite verbal operators and wh-interrogatives, which serve as Mood marking themes. Interpersonal themes may also be preposed clauses, such as I believe…, I think…, and It is possible…, It is desirable…, which carry out similar functions to modal adjuncts. These are described by Halliday (2004: 613–625) as metaphors of modality. The clauses with first person singular pronoun subjects attribute the attitude they express explicitly to the speaker/writer and are thus subjective interpersonal metaphors. The clauses with the subject pronoun it, on the other hand, do not overtly identify the source of the attitude they express and are thus objective interpersonal metaphors. Objective interpersonal metaphors may express attitudes such as epistemic modality, e.g. it is true, likely, clear; deontic modality, e.g. it is necessary, desirable; dynamic modality, e.g. it is possible, easy, difficult, and evaluation, eg it is nice, important, amazing (Herriman 2000: 296). Textual themes, finally, are concerned with the construction of text by building up sequences of discourse and creating cohesion and continuity, etc. (Thompson 1996: 38; Halliday 2004: 29–31 and 81). These are, for instance, conjunctions, which serve as structural textual themes, e.g. but, because, although, so that, and conjunctive adjuncts, which serve as conjunctive textual themes, and reflect how the message relates to the context e.g. moreover, instead, however (Halliday 2004: 79–87). Writers may select a simple theme representing the experiential metafunction only as in (1) above, for example, or they may combine several themes with one or more different metafunctions (multiple themes). The multiple theme in (2), for example, is a combination of all three metafunctions.

. Objective interpersonal metaphors are generally referred to as extraposition of subject clauses (Huddleston and Pullum 2002: 1403; Seppänen 1987: 337; Seppänen and Herriman 2002: 35).


(2) textual (structural) theme + interpersonal (modal) theme + experiential theme: But, surprisingly, the delivery was on time.

Writers may also select various constructions to vary the choice of information in the theme (Halliday 2004: 68–71). The types of thematic variation examined here are cleft constructions. These include it-clefts, as in (3), and pseudo-clefts, both basic pseudo-clefts and reversed pseudo-clefts, as in (4a) and (4b), respectively. Pseudo-clefts may also be th-clefts, i.e. clefts which are formed with “the in conjunction with the pro-form equivalents of the English interrogatives (thing, one, place, time, reason, way) and an optional relative” (Collins 1991: 27), as in (4c) and (4d), and all-clefts, as in (4e) and (4f). (3) It was my address book that George was looking for. (4)

a. b. c. d. e. f.

What George was looking for was my address book. My address book was what George was looking for. The one who was looking for the address book was George. George was the one who was looking for the address book. All George was looking for was my address book. That was all George was looking for.

Cleft constructions organize the information in the clause into a two-part theme and rheme structure. One part of the information is highlighted and the other is downranked in a subordinate clause, and thereby associated with an existential presupposition, in (3) and (4), for example, that “George was looking for something” (see e.g. Collins 1991: 71; M. Johansson 1996: 129). These two parts of the message are linked to each other by the identifying verb be. This identifying relation between the cleft clause and the highlighted element creates an exclusiveness implicature, which means that in (3), and (4), for example, it is implied that “George was looking for my address book” and nothing else (see Collins 1991: 69; M. Johansson 1996: 130 on the exclusiveness implicature of cleft constructions; see also Grice 1975: 45 on implicatures). The exclusiveness implicature of the cleft construction gives the proposition more argumentative force than it would have had as a regular declarative clause. Within Appraisal theory (White 2003), cleft constructions may therefore be regarded as a contractive heteroglossic rhetori. In Systemic Functional Linguistics, it-clefts are referred to as theme predications (Halliday 2004: 95–98) and psuedo-clefts as thematic equatives (Halliday 2004: 68–71). . Depending on whether the information in the theme and rheme is given or new, clefts can be contrastive or summative, or serve a topic linking or topic launching function (M. Johansson 2002: 185).


cal strategy, i.e. they close down negotiation by challenging and fending off alternative positions (Herriman 2003 & 2005). Earlier studies have shown, for instance, that it-clefts are used frequently in persuasive discourse, such as editorials (Collins 1991:179f.). This paper compares the following three aspects of information organization in NS and NNS writing: 1. Theme selection (simple or multiple; experiential, interpersonal, or textual themes) 2. The usage of objective and subjective interpersonal metaphors as themes 3. The usage of cleft constructions For the comparison of themes, a selection of approximately 22,400 words of argumentative essays from each corpus has been used and the themes have been studied on the sentence level manually. The comparison of interpersonal metaphors and cleft constructions, on the other hand, has been based on both of the whole corpora of argumentative texts (each approximately 175,000 words). Using the WordSmith Tools concordance program (Scott 2004), all the instances of the following subjective interpersonal metaphors were collected: I think, I believe, I am sure, I know, and I feel. The results of these three comparisons will be given in Sections 3.1–3.3, respectively. Section 4 discusses the effects that the differences found have on the NNS’ argumentative writing. Section 5 discusses the possible reasons for these differences, and Section 5, finally, concludes with some pedagogical recommendations.

. The data for the comparison are from Boström Aronsson’s doctoral dissertation (2005). . The unit of analysis is what Fries (1995: 45) refers to as an independent conjoinable clause complex. This consists of “an independent clause together with all hypotactically related clauses and words that are dependent on that independent clause”. . All the instances of objective interpersonal metaphors and it-clefts were collected by using it as the search word. Pseudo-clefts were collected by using the search words what, who, where, when, why, how, thing, one, place, time, reason, and way. Interpersonal metaphors and cleft constructions which are not thematic because they are embedded in other clauses have not been included in the investigation.


Table 1. Themes in NNS and NS writing (% denotes the proportion of independent clauses with this type of theme)* Themes N Simple Themes Multiple Themes with textual elements Multiple Themes with interpersonal elements Multiple Themes with textual and interpersonal elements Total Multiple Themes

865 286 105 89 480

NNS % 62.5 20.7 7.6 6.4 34.7

N 1,078 269 71 45 385

NS % 72.6 18.0 4.8 3.0 25.9

* 39 (2.8%) of the 1,384 independent clauses in the NNS writing and 22 (1.5%) of the 1,485 independent clauses in the NS writing consist only of a rheme and have no theme.

3.

Results

3.1

Theme selection

The comparison of theme selection by NS and NNS (Table 1) reveals a much higher frequency of multiple themes in NNS writing (about a third of the NNS’ independent clauses have a multiple theme compared to only a quarter of the NS’). Above all, the NNS use more multiple themes containing interpersonal themes. There are almost twice as many multiple themes containing interpersonal elements, both modal and Mood marking themes (194 vs. 116 instances; 14.0% vs. 7.8% of the total number of independent clauses). This indicates that the NNS tend to thematize their opinions to a greater extent than NS do in this type of writing, as illustrated in the following extracts (modal themes in italics): (5) a. b.

Maybe assimilation is not the right word after all. I mean it is a little bit harsh to say that someone should “become like the people in a group, or a country” – especially in ways of behaving or thinking”. Noone should be maimed to that degree. But it depends what you put into the words “behaving” or “thinking”. In a sense I think it is impossible not to alter your behaviour or thinking when you live in a different culture. But it should not mean that you alter your individuality. Maybe integration is not the right word after all, when it comes to how to deal with immigrants. (ICLE-SW-UG-008) Obviously, some immigrants will integrate easier in Swedish society than others. It is important though, that everybody is treated equally. Sadly, it is no secret that a kaukazian immigrant coming from a Western civilization will blend in more easily than, for instance, a muslim African. (ICLE-SW-UG-001)


The higher frequency of Mood marking interpersonal themes (61 vs. 42 instances) is chiefly because there are more interrogative clauses, as in (6a) and (6b): (6) a. How can we make people in all countries feel dependant on Nature? One solution is information. If you are not informed about the conse quences of your actions, how are you then supposed to act properly? The next step comes naturally. In what way shall the information be trans mitted? The answer also comes naturally: in all possible ways, because some people react to one form, others to another. (ICLE-SW-UG-010) b. Is there a hope for Man without Nature. The answer is naturally no since Man is a part of Nature. Still he is about to destroy major parts of it, both forests and seas. What are the reasons for destroying a part of one self? (ICLE-SW-UG-002)

A high frequency of interrogative clauses has been found in earlier studies of Swedish NNS writing (Ädel 2006: 133) and Finnish NNS writing (Virtanen 1998: 97), which suggests that the NNS tend to interact more with their readers by means of questions in argumentative writing. As well as more interpersonal themes, the NNS’ essays also contain more textual themes, in particular conjunctive themes, which express internal relations and metalanguage as in (7) (139 vs. 105 instances) (conjunctive themes in italics): (7) Therefor the traditional Swedish way of imposing new information on the generations on age, would be less effective than to via their children invoke a guilty conscience. Thus, instead of having an exhorted, disconcerted people, it would be seething of new positive energy, seemingly by its own will. What the Western World need is not new laws as much as a new way of thinking, and this “sap” can only coaxed from within. First of all we must restore the long-lost feeling of affinity between man and nature, and that by supporting contact between consuments and primary produce, to start with. (ICLE-SW-UL-011)

3.2 Subjective and objective interpersonal metaphors The comparison of subjective and objective interpersonal metaphors in NS and NNS writing (Table 2) reveals a much higher frequency of both subjective and objective interpersonal metaphors in the NNS writing.

. Spelling errors, etc. in the NNS writing have not been corrected.


Table 2. Subjective and objective interpersonal metaphors in NNS and NS writing (f = tokens per 10,000 words) Metaphors Subjective Objective

NNS

NS

n

f

n

f

314 363

18.0 20.8

129 160

7.4 9.2

Subjective interpersonal metaphors are more than twice as frequent in the NNS writing (314 vs. 129 instances). The large number of subjective interpersonal metaphors in the NNS writing is mainly due to an excessive use of I think, as in (8a) and (8b) (subjective interpersonal metaphors in italics). There are five times as many examples of this type in the NNS writing as in the NS writing (223 vs. 45 instances). (8) a. b.

I think the most essential thing is that immigrants learn Swedish fluently. (ICLE-SW-UL-007) I advocate totally free speech because I think it is a human right to be allowed to express your opinions. I think that advocating censorship on either the written or spoken word you are saying that the people are not clever enough to have their own opinions. I think that filmmakers are entitled to express their opinions, but I do not approve of unnecessary violence. Unnecessary violence is, according to me, unprovoked, abusing children or other defenseless people or animals, or not reflecting a true event e.g. the Holocaust. (ICLE-SW-ULW-008)

This overuse of I think by NNS has also been noted in earlier studies based on the ICLE corpus. Ringbom (1998: 44), for instance, reports that I think was far more frequent in all seven of the learner varieties that he studied (Swedish, Finland-Swedish, French, German, Finnish, Dutch and Spanish), and that it was most common in the writing produced by Swedish and Finland-Swedish NNS. The clause I think has been noted to be particularly common in spoken language. Altenberg, in his study of phraseology in the London-Lund Corpus of spoken language (1998: 113), found that among the structures used as “the springboard of utterances leading up to the communicatively most important – and lexically most variable – element”, I (don’t/didn’t) think/thought (that) was the most common type. Objective interpersonal metaphors are also more than twice as frequent in the NNS writing (363 vs. 160 instances). This frequency (20.8 per 10,000 words) is higher than that found in the Lancaster–Oslo Bergen Corpus (LOB) corpus of


Table 3. Semantic representations of NNS’ and NS’ objective interpersonal metaphors (f = tokens per 10,000 words)

Dynamic Modality Evaluation Epistemic Modality Deontic Modality Total

n

NNS %

f

n

NS %

f

113 155 74 23 365

31.0 42.5 20.3 6.3 100.1

6.5 8.9 4.2 1.3 20.9

40 61 33 26 160

25.0 38.1 20.6 16.3 100

2.3 3.5 1.9 1.5 9.2

written British English (16.5 per 10,000 words) (Herriman 2000: 296). It appears then that there is a tendency for NNS to foreground their opinions and evaluative comments by selecting interpersonal metaphors as themes. It was also noted that the NNS tend to combine a thematic objective interpersonal metaphor with another modal theme more frequently than NS (60 vs. 10 instances), as in (9a) and (9b): (9) a. b.

First of all I believe that it is important that our environmental problems get international media attention. (ICLE-SW-UL-022) I think that it is possible to use the development in order to make the environmental conditions on earth better. (ICLE-SW-UG-109)

This heavy foregrounding of personal opinions and attitudes appears to be even more common here than in spoken English. In Gómez-González’s (2001) study of themes in spoken English based on the Lancaster IBM Spoken English Corpus, only 10% of her examples of this type were combined with another modal theme (Gómez-González 2001: 274). A closer look at the types of attitudes expressed in NNS’ and NS’ objective interpersonal metaphors (Table 3) reveals further differences. The biggest difference in the types of attitudes expressed by NNS’ and NS’ objective interpersonal metaphors is that there are three times more interpersonal metaphors expressing dynamic modality (113 vs. 40 instances), especially potentiality, as in (10). (10) a. It is not easy to change people’s attitudes. (ICLE-SW-UL-021) b. It is difficult to keep all the changes in perspective. (ICLE-SW-ULX-006) . The difference between the frequency in the NS writing studied here and the LOB corpus is probably partly due to the fact that Herriman (2000a) included interrogative and imperative constructions and examples that were not placed thematically. It may also be due to register differences, as Herriman’s study comprises different types of written registers.


There are also more objective interpersonal metaphors expressing evaluations (155 vs. 61 instances), as in (11), and more objective interpersonal metaphors expressing epistemic modality (74 vs. 33 instances) as in (12). (11) a. but it’s tragic that it happens at all. (ICLE-SW-UL-039) b. It is important to know about one’s roots in our fast changing world. (ICLE-SW-UL-101) c. It is nice to go to a pizzeria and meet an Italian who speaks Swedish and makes Italian pizzas. (ICLE-SW-UL-007) (12) a. b.

It is clear that education is something that society simply cannot do without. (ICLE-SW-UG-103) It is very obvious that technology can be a danger if it is not controlled | by intelligent and sound minds. (ICLE-SW-UL-048)

These types of objective interpersonal metaphors all occur more frequently in the NNS writing than in the Lancaster–Oslo Bergen Corpus of written British English (Herriman 2000: 596). Their frequency per 10,000 words in NNS writing and LOB is as follows: dynamic modality 6.5 vs. 4.6, evaluations 8.9 vs. 6.1, and epistemic modality 4.2 vs. 3.3. Objective interpersonal metaphors expressing deontic modality, as in (13), are, on the other hand, not used more often in the NNS writing than in the NS writing or than in the Lancaster–Oslo Bergen Corpus (23 vs. 26 instances, cf. frequency 1.3 per 10,000 words in NNS writing and 1.7 in LOB). (13) a. It is necessary that they pay regard to our laws and rules… (ICLE-SW-UL-017) b. It is not society’s choice to decide when he or she has had enough of life. (ICLE-US-MRQ-0002.1)

3.3 Cleft constructions The comparison of cleft constructions in the NNS and NS writing (Table 4) shows that the NNS use cleft constructions more frequently than NS (266 vs. 141 instances). All three types of cleft constructions, it-clefts, basic and reversed pseudo-clefts, as illustrated in (14), are used more frequently. (14) a. b. c.

It is the people’s opinion, not the government itself, that worries me. (ICLE-SW-ULX-037) What we should try to understand is that we in fact can learn from change and renewal. (ICLE-SW-UL-108) That is what scares me. (ICLE-SW-ULX-029)

Themes in Swedish advanced learners’ writing in English

Table 4. Cleft constructions in NNS and NS writing (f = tokens per 10,000 words) Cleft construction It-clefts Basic pseudo-clefts Reversed pseudo-clefts Total

NNS

NS

n

f

n

f

73 100 93 266

4.2 5.7 5.3 15.2

31 48 62 141

1.8 2.7 3.5 8.1

The clefts with the most notable difference in frequency are it-clefts, which are 2.4 times more frequent in the NNS writing (73 vs. 31 instances). Basic pseudo-clefts are twice as frequent (100 vs. 48 instances). It is, however, only the type of basic pseudo-cleft in which the cleft clause begins with what that is over-represented in the NNS writing (80 vs. 25 wh-clefts), whereas th-clefts and all-clefts are almost equally common (14 vs. 15 th-clefts and 6 vs. 8 all-clefts). Reversed pseudo-clefts, finally, are 1.5 times more frequent in the NNS writing (93 vs. 62 instances). On the whole, this high frequency of cleft constructions in the NNS writing (it-clefts: 4.2, basic pseudo-clefts: 5.7, and reversed pseudo-clefts: 5.3 per 10,000 words) is closer to the frequency of clefts found by Collins (1991: 181) in the spoken language in the London-Lund corpus, LL (it-clefts: 4.3; basic pseudo-clefts: 6.3; reversed pseudo-clefts: 8.1 per 10,000 words) than that found in the written language of the Lancaster–Oslo Bergen Corpus, LOB (it-clefts: 5.7; basic pseudoclefts: 1.9; reversed pseudo-clefts: 2.2 per 10,000 words). If we compare how the NNS and NS distribute given and new information10 in their cleft constructions, we find that there is a tendency for the NNS to use all three cleft types to thematize new information more frequently than the NS do (cf. 25 vs. 5 it-clefts,11 37 vs. 10 basic pseudo-clefts and 12 vs. 4 reversed pseudoclefts), as illustrated in (15)–(17) below. (15) Integration or assimilation? Integration or assimilation is and has always been a tricky question. People are afraid of things and customs that they do not recognize. The less knowledge they have about other cultures the more frightening they seem. 10. Information that has been introduced into the discourse earlier is classified as given information, and information which has not been introduced into the discourse before or has not been active in the discourse within the past 20 sentences is classified as new information (see e.g. Halliday 1967, 1994; Halliday and Hasan 1976; Chafe 1976; Fries 1992, 1995; Brown and Yule 1983: 153; Prince 1981 on given and new information). 11. It-clefts thematizing new information tend to be particularly common in private and dialogic speech as well as in writing (Collins 1991: 191).

111


It is understandable that it usually is young, uneducated working-class men who sympathize with racist organisations once they feel a threat from the immigrants whit whom they have to compete about the few jobs there are. (ICLE-SW-UL-084)

In particular, the NNS tend to use basic pseudo-clefts to thematize new information in terms of different types of evaluative comments (33 vs. 7 instances), as in (16a) and (16b): (16) a. b.

It makes me angry that I have to overlook aspects I find important, concerning child care just because this “welfare state” forces me to do so. It is a shame that people who are educated in the field of child care will not be able to get employment because of economic (again!) cut-backs but what is even worse is the fact that the real loosers are our children. (ICLE-SW-ULX-041) Most Swedes living in the south of Sweden, see Sweden’s entry into the EU as an great opportunity, to get out and work in the rest of Europe, to have a greater exchange with other European countries. This is all good, but what I cannot understand is why the European mainland is called Europe and we up here in Scandinavia are called just that: Scandinavia. (ICLE-SW-UL-087)

It is also notable that the NNS sometimes use reversed pseudo clefts to thematize new information with cataphoric reference, as in (17). This is different from the expected pattern for information in reversed pseudo-clefts, as these typically highlight a demonstrative pronoun with anaphoric reference and are used to sum up the discussion (Collins 1991: 139). (17) I feel that today people are confused about this difference. They are afraid to say what they think of fear of being branded racists. There is certainly a rapid growing movement of racists and nazis in Sweden, but these terms should be kept for those deserving it. They are still only a small fraction of the population and it is important to keep them apart from the rest of society. I don’t mean that they are vile, foul-mouthed people to be kept under constant surveillance, but I think it is imperative not ot let them be spokes-men for the average person. This is where the danger lies; we are slightly wary of new, different habits and customs and this together with the constant threat we face today of losing our jobs and houses makes us look for a scape-goat. This is the foundation of the racist movement fo today: people’s innate suspicion of alien things and the bad times are perfect breeding-grounds for their warped device as they once used 50 years ago and we all know what terrible disaster it led to. (ICLE-SW-ULW-026)


In sum, the comparison of NNS and NS writing has revealed a number of significant differences between NNS’ and NS’ selection of themes and their usage of thematic variation. First, the NNS tend to overuse multiple themes, in particular multiple themes containing interpersonal themes but also those containing textual conjunctive themes. Second, the NNS overuse subjective interpersonal metaphors, such as I think, and objective interpersonal metaphors expressing certain types of modal expressions and evaluative comments, e.g. expressions of potentiality, emotive reactions, significance, and truth. Finally, the NNS use it-clefts and pseudo-clefts (both basic and reversed) more frequently than NS. In particular, they use these cleft constructions to thematize new information. They also use basic pseudo-clefts to express personal opinions more often than NS. The overuse of these types of themes and thematic variation means, then, that, even though the NNS language itself may be more or less grammatically correct, their style of writing deviates somewhat from NS’ argumentative writing. In the following we will show that it results both in a style of writing which is more appropriate for spoken language than for argumentative writing, and in a style of persuasion which is overly direct and emphatic.

4.

Features of NNS’ argumentative writing

In contrast to written language, which reflects the writer’s detachment from the audience, from himself or herself, and from concrete reality, spoken language tends to contain more features of involvement and interaction with the audience, such as interpersonal themes and modal expressions (McCabe 1999: 227; Biber et al. 1999: 486, 859), references to the writer and the reader in terms of first and second person pronouns, references to the speaker’s mental processes, e.g. think and feel, and devices for monitoring the flow of information, e.g. questions (Chafe 1982: 45; Chafe and Danielewicz 1987: 105; Biber 1988: 105). Many of the types of themes and thematic variation overused by the NNS are thus primarily features of spoken language. First, they overuse subjective interpersonal metaphors which contain first person references and references to the writer’s mental processes. According to Aijmer (2001: 252), the use of I think in NNS’ argumentative essays “reflects the writers’ concern to express a high degree of involvement with the issue discussed”. Second, the NNS overuse basic pseudo-clefts with a first person reference expressing a comment (what I think, what this means), which also contributes to their spoken and involved style. Jones and Jones (1985: 5f.) point out that although their main function is thematic, basic pseudo-clefts are often used to indicate the speaker’s/writer’s perspective or contrast. Biber et al. (1999: 963) note that the frequency of basic pseudo-clefts in conversation may be due to the


low information content of the cleft clause and that constructions such as what I think..., what I want to say..., and what we need... are often used in conversations “as a springboard in starting an utterance”. Third, the NNS overuse modal interpersonal themes and Mood marking interpersonal themes, in particular those marking questions, both of which express involvement and interaction with the reader and are typical of spoken language. According to Virtanen (1998: 105), an overuse of questions may reduce their argumentative value and increase the more informal style of the NNS writing. The examples in (18) illustrate how NNS may use questions to interact with their readers: (18) a. If this would have been a question of voting for or against the union as such, my choice would have been easier to make. I do not believe in the EC, I do not believe this is the birth of a new nation and I definitely do not like the thought of giving foreign men in black suits and ties the right to decide what is best for me. It scares me to think about where a union like this could lead us in the future. With the battle for freedom still going on in eastern Europe I wonder if we will end up like them, fighting for independence. Will the small countries in the EC be suppressed by the more powerful ones? Will Sweden’s voice be heard at all? Unfortunately, this is not a question of the EC’s being or not being. To be or not to be a member of the EC, that is the question. But what is the answer? The EC already exists, and that fact makes me look upon the union as a necessary evil. (ICLE-SW-UL-093) b. You cannot wipe out a person’s background, and why should we? Trying to suppress other people’s cultures only leads to trouble. (ICLE-SW-UL-016)

Finally, the NNS overuse conjunctive textual themes as in (19), which show how different parts of the text are connected. This suggests that they also tend to frequently express involvement by interfering in the text to show how different parts of the text are connected. (19) When attempting to integrate immigrants in Swedish society it is important that they are able to keep their own traditions and values to a certain extent. However, they need to understand and accept that certain laws and rules are different in Sweden… (ICLE-SW-UG-018)

As well as making their style of writing more appropriate for spoken than written language, some of the types of themes and thematic variation overused by the NNS contribute towards making their style of persuasion more emphatic. This can be observed in their overuse of certain types of conjunctive themes, their


overuse of subjective interpersonal metaphors such as I think, their overuse of objective interpersonal metaphors which express strong claims as to the truth and significance of the proposition, and their overuse of cleft constructions. Conjunctive themes have been found by Francis (1989: 209) to be used more frequently in persuasive expository writing (editorials and letters) than in news reports, which suggests that the NNS’ frequent use of conjunctive themes may reflect a persuasive style. According to Francis (1989: 209), they “function as labellers of chunks of previous text, and as such are an important feature of argument”. Subjective interpersonal metaphors such as I think have been described as a “powerful argumentative device/expression since it suggests that the speaker/writer is announcing/claiming something as an authority” (Aijmer 2001: 251). Aijmer notes that the Swedish NNS’ overuse of I think makes the NNS’ argumentation sound different from NS’: “[l]earners overuse I think in order to make their claims more persuasive and to provide more weight to the issues discussed while native speakers use a less rhetorical style” (2001: 256). Example (20) illustrates the direct and less subtle type of persuasion in the NNS’ use of the interpersonal metaphor I think: (20) Here in Sweden we often discuss the question of immigration. There are those who say that we should not allow any immigration, and there are those who claim that immigration is good for Sweden. What we seldom discuss, though, is how we should handle those immigrants who are already here to stay. I think that we all can agree that our immigrants should be a part of the Swedish society. But we have to find out whether we want them to be an integrated part, or totally assimilated. (ICLE-SW-UG-019)

Objective interpersonal metaphors can be seen as subjective opinions dressed up to appear more objective (Collins 1994: 19f.). Objective interpersonal metaphors, which express a strong claim about the accuracy of the proposition such as It is true and It is a fact and objective interpersonal metaphors which express significance such as (21) are commonly used in persuasive texts, such as editorials, because they function as a means of presenting the writer’s opinion explicitly without revealing its source (Herriman 2000: 214). (21) Some people would perhaps say, – But if others do not think of the environment why should I? This way of thinking is dangerous. It is important to think that each individul effort makes a lot of results in the end. It is important to think of Nature as irreplaceable. (ICLE-SW-UG-002)

Finally, the NNS’ emphatic style of persuasion is also reflected in their overuse of cleft constructions. Clefts function as contractive heteroglossic rhetorical strategies, which close down negotiation in argumentation and are therefore frequently used to strengthen the force of persuasive discourse (Herriman 2003: 3).


5.

Discussion

We have found, then, that the NNS’s overuse of certain type of themes and thematic variation contributes to a style of writing that is closer to spoken language and which makes the argumentation more emphatically persuasive. Why then do the NNS overuse these types of themes and thematic variation? We would like to suggest three main reasons for this: transfer from Swedish, learner strategies, and a lack of knowledge of certain aspects of argumentative writing. First, the NNS may overuse certain English structures because there are similar constructions in Swedish, which makes them easy to use. Subjective and objective interpersonal metaphors have counterparts which are very common in Swedish and which the NNS recognize and may transfer to their writing in English.12 It-clefts also have counterparts which are frequent in Swedish. The NNS may, however, sometimes use English it-clefts without being aware of the fact that there are some differences between the English and Swedish constructions. According to M. Johansson (2002: 145, 154, 169), Swedish it-clefts are less marked in terms of distributional freedom than English it-clefts, less associated with contrast than English it-clefts, and are also used as cohesive devices more often than English it-clefts. Thus, it is possible that Swedish NNS may use it-clefts in English where they would have used it-clefts in Swedish, whereas NS would have chosen another structure. As well as transferring familiar linguistic constructions, the NNS may also transfer cultural differences in how to write argumentative texts in Swedish to their writing in English. There may, for instance, be differences in their use of politeness strategies. Hyland and Milton (1997: 186) note that “students from different cultures may have preconceptions about the formal features of culturally and rhetorically appropriate writing which may differ from those which operate in English academic settings”. It is also possible that there is a higher tolerance in Swedish upper secondary schools towards informal language and features of spoken language in written English texts than there is within the American and British school systems. Second, the NNS’ overuse of certain types of themes and thematic variation may be influenced by different types of learner strategies. Learner strategies compensate for the NNS’ lack of sufficient knowledge of the foreign language (Ädel 12. A study of subjective and objective interpersonal metaphors in a sample of essays of 72,345 words in Swedish written by Swedish students at upper secondary school suggests that these constructions are also used very frequently in written Swedish and therefore the NNS may be influenced by their mother tongue in their choices of these structures as themes (Boström Aronsson 2005).


2006: 151). One such learner strategy is the use of formulaic expressions. These are multi-word or multi-form strings that can be used as fluency devices and discourse organizers (Weinert 1995: 182). The NNS may see interpersonal metaphors such as I think (that) and It is clear (that), and the structure It is X (that) of it-clefts as useful formulas that help them get started on an utterance. Similarly, the NNS may see certain types of basic and reversed pseudo-clefts, such as What I/we VP BE... and This/That is what..., as useful formulas. This may, then, partly explain the large number of subjective and objective interpersonal metaphors and certain types of cleft constructions in the NNS writing. Finally, the NNS’ overuse of certain types of themes and thematic variation may, quite simply, be caused by their lack of knowledge in certain areas. One area where they lack knowledge is how to write argumentative texts in English and they have little experience of arguing for their point of view and making their points clear to their readers in a foreign language. As a result they compensate for their lack of experience by overusing emphatic expressions, such as cleft constructions and objective interpersonal metaphors with emphatic meanings, as well as features of involvement, such as interpersonal themes and basic pseudoclefts expressing comments, whereas the NS, on the other hand, may use a different and more subtle way of putting forth their arguments. The NNS’ insecurity may also explain the larger number of basic pseudo-clefts conveying contextually derived information, such as What this means and What I want to say, which are used to comment on and/or clarify the preceding text. Another area where the NNS lack knowledge is that of which linguistic features are appropriate in writing and speech and therefore they are not able to master register differences. As a result they tend to use expressions in their writing which are more typical of spoken language than of written, for instance, subjective interpersonal metaphors, such as I think. The NNS’ problems distinguishing between different registers have also been noted by Wiktorsson (2003: 110, 121), who found that Swedish NNS tend to use expressions in their writing which are more likely to be found in NS’ speech than in their writing. The NNS also appear to lack sufficient knowledge of textual organization in English. This is suggested by their overuse of explicit expressions of cohesion, such as conjunctive themes which express relations between ideas and parts of the text, which, in turn, may result in a clumsier style in their essays. This overuse may be due to an insecurity among the NNS as regards the creation of cohesive ties when writing in a foreign language. Difficulties in the organization of information can also be seen in the NNS’ tendency to use more cleft constructions which thematize new information. Thematizing new information violates the principle of placing given information before new and this may place a greater burden on the reader and make the text more difficult to follow. This can be observed, for instance, in the NNS’ use of reversed pseudo-clefts in which the


highlighted element has cataphoric reference as in (17) above, and in examples in which the reference is both anaphoric and cataphoric, as in (22): (22) The best way in which to celebrate Christmas is to gather family and/or friends and celebrate together. I think this has always been, and still is what many people think is most important about Christmas, to be able to celebrate with the people you love, no matter how you are celebrating. (ICLE-SW-UG-044)

6.

Conclusion

Learner corpora can tell us a lot about learner writing. By studying a large amount of learner text it is possible to uncover various differences between learner language and native speaker language which are difficult or impossible to notice without large collections of learner writing and without the possibility to compare NNS writing to NS writing. A number of significant differences have been found here between NNS writing and NS writing. Above all, the NNS express more involvement in the text, interact more with their readers and use a more direct and emphatic type of persuasion than the NS do. As these features are more typical of NS speech than of NS writing, the style of the NNS’ argumentative writing is similar to NS’ spoken language. These differences between NNS writing and NS writing found in this study and the causes of these differences discussed here point to certain areas within which the teaching of argumentative writing may need to be improved for the NNS to be able to produce texts that appear more native-like. One such area is contrastive rhetoric (Connor 1996). Students need to be taught about differences in the cultural norms of argumentative writing in the foreign language and in their mother tongue. Moreover, the NNS also need to be made aware of register differences in general. If NNS are made aware of these differences, they may be more confident in their stylistic choices and rely less on formulaic expressions and emphatic constructions and more on strategies used by NS in argumentative writing.

References Ädel, A. 2006. Metadiscourse in L1 and L2 English. Amsterdam: John Benjamins. Aijmer, K, 2001. I think as a marker of discourse style in argumentative student writing. In A Wealth of English: Studies in Honour of Göran Kjellmer, K. Aijmer (ed.), 247–257. Göteborg: Acta Universitatis Gothoburgensis.


Altenberg, B. 1998. On the phraseology of spoken English: The evidence of recurrent wordcombinations. In Phraseology: Theory, Analysis, and Applications, A. P. Cowie (ed.), 101– 122. Oxford: Clarendon Press. Biber, D. 1988. Variation across Speech and Writing. Cambridge: CUP. Biber, D., Johansson, S., Leech, G., Conrad, S. & Finegan, E. 1999. Longman Grammar of Spoken and Written English. London: Longman. Boström Aronsson, M. 2005. Themes in Swedish Advanced Learners’ Written English. PhD dissertation, Göteborg University. Brown, G. & Yule, G. 1983. Discourse analysis. Cambridge: CUP. Chafe, W. L. 1976. Givenness, contrastiveness, definiteness, subjects, topics, and point of view. In Subject and Topic, C. N. Li (ed.), 27–55. New York NY: Academic Press. Chafe, W. L. 1982. Integration and involvement in speaking, writing, and oral literature. In Spoken and Written Language: Exploring Orality and Literacy, D. Tannen (ed.), 35–53. Norwood, New Jersey: Ablex Publishing Corporation. Chafe, W. L. & Danielewicz, J. 1987. Properties of spoken and written language. In Comprehending Oral and Written Language, R. Horowitz & S. J. Samuels (eds), 83–113. San Diego CA: Academic Press. Collins, P. C. 1991. Cleft and Pseudo-cleft Constructions in English. London: Routledge. Collins, P. C. 1994. Extraposition in English. Functions of Language 1(1): 7–24. Connor, U. 1996. Contrastive Rhetoric: Cross-cultural Aspects of Second Language Writing. Cambridge: CUP. Francis, G. 1989. Thematic selection and distribution in written discourse. Word 40(1–2): 201– 221. Fries, P. H. 1992. The structuring of information in written English text. Language Sciences 14(4): 461–488. Fries, P. H. 1995. Patterns of information in initial position in English. In Discourse in Society: Systemic Functional Perspectives. Meaning and Choice in Language: Studies for Michael Halliday, P. H. Fries & M. Gregory (eds), 47–66. Norwood NJ: Ablex. Gómez-González, M. A. 2001. The Theme-Topic Interface: Evidence from English. Amsterdam: John Benjamins. Granger, S. 1993. International Corpus of Learner English. In English Language Corpora: Design, Analysis and Exploitation. Papers from the Thirteenth International Conference on English Language Research on Computerized Corpora, Nijmegen. 1992, J. Aarts, P. de Haan & N. Oostdijk (eds), 57–69. Amsterdam: Rodopi. Granger, S. 1994. The learner corpus: A revolution in applied linguistics. English Today 39(10:3): 25–32. Granger, S. 1998a. The computer learner corpus: A versatile new source of data for SLA research. In Learner English on Computer, S. Granger (ed.), 3–18. London: Longman. Granger, S . 1998b. Prefabricated patterns in advanced EFL writing: Collocations and formulae. In Phraseology: Theory, Analysis, and Applications, A. P. Cowie (ed.), 145–160. Oxford: Clarendon Press. Grice, H. P. 1975. Logic and conversation. In Syntax and Semantics, Vol. 3: Speech Acts. P. Cole & J. L. Morgan (eds), 41–58. New York NY: Academic Press. Halliday. M. A. K. 1967. Notes on transitivity and theme in English. Part 2. Journal of Linguistics 3: 199–244. Halliday, M. A. K. 1994. An Introduction to Functional Grammar, 2nd edn. London: Arnold.


Halliday, M. A. K. 2004. An Introduction to Functional Grammar, 3rd edn, revised by C. M. I. M. Matthiessen. London: Arnold. Halliday, M. A. K & Hasan, R. 1976. Cohesion in English. Harlow: Longman. Herriman, J. 2000. The functions of extraposition in English texts. Functions of Language 7(2): 203–230. Herriman, J. 2003. Negotiating identity: The interpersonal functions of wh-clefts in English. Functions of Language 10 (1): 1–30. Herriman, J. 2005. Negotiating a position within heteroglossic diversity: Wh-clefts and it-clefts in written discourse. Word 56(2): 223–249. Huddleston, R. & Pullum, G. K. 2002. The Cambridge Grammar of the English Language. Cambridge: CUP. Hyland, K. & Milton, J. 1997. Qualification and certainty in L1 and L2 students’ writing. Journal of Second Language Writing 6(2): 183–205. Johansson, M. 1996. Contrastive data as a resource in the study of English clefts. In Languages in Contrast: Papers from a Symposium on Text-based Cross-linguistic Studies. Lund 4–5 March 1994, K. Aijmer, B. Altenberg & M. Johansson (eds), 127–150. Lund: Lund University Press. Johansson, M. 2002. Clefts in English and Swedish: A Contrastive Study of IT-clefts and WHclefts in Original Texts and Translations. PhD dissertation, Lund University. Jones, L. B. & Jones, L. K. 1985. Discourse functions of five English sentence types. Word 36 (1): 1–21. Lorenz, G. 1998. Overstatement in advanced learners’ writing: Stylistic aspects of adjective intensification. In Learner English on Computer, S. Granger (ed.), 53–66. London: Longman. Mauranen, A. 1993. Cultural Differences in Academic Rhetoric: A Textlinguistic Study. Frankfurt: Lang. McCabe, A. M. 1999. Theme and Thematic Patterns in Spanish and English History Texts. Unpublished PhD dissertation, Aston University. Prince, E. F. 1981. Toward a taxonomy of given-new information. In Radical Pragmatics, P. Cole (ed.), 223–255. New York NY: Academic Press. Ringbom, H. 1998. Vocabulary frequencies in advanced learner English: A cross-linguistic approach. In Learner English on Computer, S. Granger (ed.), 41–52. London: Longman. Scott, M. 2004. WordSmith Tools version 4. Oxford: OUP. Seppänen, A. 1987. On the syntax of seem and appear. Zeitschrift für Phonetik, Sprachwissenschaft and Kommunikationsforschung 40: 336–351. Seppänen, A. & Herriman, J. 2002. Extraposed subjects vs. postverbal complements: On the so-called obligatory extraposition. Studia Neophilologica 74: 30–59. Thompson, G. 1996. Introducing Functional Grammar. London: Arnold. Virtanen, T. 1998. Direct questions in argumentative student writing. In Learner English on Computer, S. Granger (ed.), 94–106. London: Longman. Weinert, R. 1995. The role of formulaic language in second language acquisition: A review. Applied Linguistics 16(2): 180–205. White, P. R. R. 2003. Beyond modality and hedging: A dialogic view of the language of intersubjective stance. Text 23(2): 259–284. Wiktorsson, M. 2003. Learning Idiomaticity: A Corpus-based Study of Idiomatic Expressions in Learners’ Written Production. Lund: Lund Studies in English.

Thematic choice and expressions of stance in English argumentative texts by Norwegian learners Hilde Hasselgård

English and Norwegian have been shown to differ somewhat as regards thematic choice, due to the V2 constraint in Norwegian and to different frequency distributions of some constituent types in clause-initial position. Using the Norwegian component of the International Corpus of Learner English, the present study explores the extent to which Norwegian students apply Norwegian patterns in their choice of thematic structure, and to what extent they have acquired the grammatical and stylistic norms in relevant genres of English. A further aim is to identify areas of difficulty for Norwegian advanced learners of English. The learners make rather few grammatical errors with themes, but certain features have a different distribution than in comparable texts by native writers. E.g. clause-initial adverbials are used by Norwegian learners with a frequency more similar to Norwegian than to English patterns. Formal subjects (it/ there) are also over-represented, which is almost entirely due to it-extraposition expressing stance. Furthermore, first-person pronouns are overused as theme in clauses which function as subjective stance markers. One conclusion is that the students seem to master the grammatical structures, but not the discourse conventions of argumentative and academic writing in English.

1.

Introduction and background

The Norwegian component of the International Corpus of Learner English (NICLE) has been completed relatively recently, and little research has so far been carried out. The present study is thus an exploratory one, aiming to identify discourse patterns chosen by Norwegian learners of English and to compare these to corresponding patterns in native-produced English and Norwegian. The focus is on the thematic section of the sentences, theme being defined according to Systemic-Functional Grammar as the initial part of a clause, extending up to and

122 Hilde Hasselgård

including the first participant, process or circumstance; i.e. the first experiential (referential) element in the clause (Halliday 2004: 175). Functionally, the theme is the element ‘which locates and orients the clause within its context’ (Halliday 2004: 62). Previous contrastive studies of word order in English and Norwegian have uncovered some differences in thematic structure (e.g. Hasselgård 1997 and 2004). The present study seeks to investigate whether Norwegian learners apply Norwegian patterns in their written English. Furthermore, previous studies of learner language indicate that learners overuse features from informal conversation in their written output (e.g. Aijmer 2002: 73; Boström Aronsson 2005: 188). In the present study I have looked for such features in the thematic section of the clause. Particular attention has been given to expressions of stance, which turned out to be thematized very frequently by the learners. The ultimate aim is to identify areas of difficulty for Norwegian learners of English as regards thematic choice and clause organization in relation to style and genre.

2.

Material and method

The study is based on the Norwegian component of the International Corpus of Learner English (ICLE). This corpus consists of essays written by Norwegian university students of English, most of them in their first year. The essay questions are designed to yield argumentative texts, although a close-reading of the texts shows that this has not always been achieved by the learners. None of the essays are timed or written under exam conditions. In order to identify thematic structures the material had to be analysed manually, and thus a selection of texts was made. A search for essay titles containing the word crime yielded 23 texts; a total of about 15,000 words. Most of them are entitled “Does crime pay?” while a few discuss the question “Is our prison system out-dated?”. The study starts from insights gained from studies of the English-Norwegian Parallel Corpus (ENPC), which contains English and Norwegian original texts with translations into the other language. The NICLE material is also compared to some genres of the British component of the International Corpus of English (ICE-GB): student writing, academic writing, press editorials, and conversation. The student writing consists of the untimed essays from ICE-GB. These are largely expository rather than argumentative, but still provide interesting material for comparison with NICLE. Academic writing has been included because it pro. For a description of the ICLE project and the corpus, see Granger (ed.) 1998 and http://cecl. fltr.ucl.ac.be/. For further information on the other corpora used, see URLs given in the reference list.

Thematic choice and expressions of stance 123

vides a ‘target genre’ that the students are expected to learn. Press editorials are interesting in that they represent an argumentative genre. Conversation has been included because previous studies of learner language report that learners overuse features from informal conversation in their written output. Finally, some comparisons are made with the Swedish ICLE component (SWICLE) and LOCNESS (the NS reference corpus compiled by the ICLE team at Louvain-la-Neuve), but only on the basis of findings by other linguists (Aijmer 2002; Ädel 2006; Boström Aronsson 2005). The study uses the methodology of the ‘Integrated Contrastive Model’ (Granger 1996; Gilquin 2003), which couples the study of learner language with contrastive studies. The approach has a number of advantages: on the basis of insights from contrastive analysis one can predict areas of difficulty for learners and identify potentially interesting topics of research. Conversely, peculiarities of learner (inter)language can be linked to features from the learner’s native language or to other features of the learning situation. The present investigation starts from insights from previous contrastive studies and compares thematic patterns in English and Norwegian with those found in NICLE. The analysis is largely based on T-units; i.e. a main clause with any associated dependent clauses (Thompson 2004: 156). However, the division into orthographic sentences made by the learners has been respected, so a dependent clause or a fragment occurring on its own has not been considered part of another T-unit.

3.

Previous research on “Scandinavian English” word order

Previous studies of Scandinavian-produced English point in the same direction as regards word order patterns. The tendency seems to be towards an overuse of patterns that are acceptable both in English and in the closely related Scandinavian languages, but which are more frequent in Scandinavian. Drew (1998) investigates the writing skills of Norwegian students in a teacher training college. Some findings that are relevant to the present study are that the Norwegian students use sentence-initial adverbial clauses much more frequently than their British peers. They also use conjunctions and conjuncts more often (Drew 1998: 126 and 129). Shaw (2004) compares a collection of academic papers written in English by Danes to papers written in Danish by the same authors as well as academic papers written by native speakers of English (NS writers). He finds that sentence-initial adverbials are used far more often in articles written by Danes than in articles by NS writers, while sentence-initial subjects dominate more in the NS texts


(2004: 75). To a great extent the texts written in English by Danes follow the same syntactic patterns as the texts in Danish by the same writers. Boström Aronsson (2005) investigates themes in Swedish advanced learners’ written English as evidenced in SWICLE. Like Drew (1998) she finds that the learners use more interpersonal and textual themes than native speakers. Further she finds a clear overuse of extraposition, all types of cleft sentences, and ‘subjective interpersonal metaphors’ (2005: 84ff.), particularly I think. On the whole, Swedish learners overuse themes which express the writer’s involvement with the text and interaction with the writer, a finding also reported by Ädel (2006).

4.

Some features of thematic structure in NICLE material

4.1

Syntactic choices: A contrastive perspective

Previous contrastive studies of theme in English and Norwegian (e.g. Hasselgård 1997, 2004 and 2005) have uncovered a number of differences in thematic choice between the languages. The following features are found more frequently in Norwegian than in English: initial adverbials, formal subjects and clause-coordinating conjunctions. English favours agentive subjects as theme to a greater extent than Norwegian, and has a greater tolerance of indefinite noun phrases in subject position (Hasselgård 2005: 36ff.). On the basis of these findings and those of the studies referred to in Section 3, I hypothesized that the NICLE essays would display many of the features found in Norwegian. Table 1, building on Hasselgård (1997: 14), shows the choice of sentence-initial constituent in the original texts of the ENPC as well as in NICLE. The survey includes only declarative main clauses. Table 1 should be interpreted with some caution; the ENPC and NICLE texts represent different genres, viz. fiction and argumentative prose. It is thus possible that some of the differences are due to genre. A further complication is the developmental factor; all the ENPC texts have been written by professional authors. But with these reservations in mind, it is clear that the Norwegian learners use patterns that are closer to those found in the Norwegian ENPC texts than in the English ones. This pertains particularly to the distribution of clause-initial adverbials and subjects. It is, however, noteworthy that both initial adverbials and anticipatory subjects are used more frequently in the learner texts than in the Norwegian ENPC texts. A feature that is not transferred from Norwegian is the fronting of objects and predicatives; in fact, the learners avoid such fronting almost entirely, thus achieving a lower percentage of fronted objects than that of original English in the


Table 1. Choice of first constituent in NICLE and in original texts of the ENPC2 First constituent* Subject Anticipatory subject Direct object Subject predicative Adverbial Conjunction Other Total T-units

N

NICLE %

E orig (ENPC) N %

N

536 46 4 3 274 59 16 938

57.1 4.9 0.4 0.3 29.2 6.3 1.8 100.0

380 16 58 0 105 38 3 600

315 17 73 2 149 32 12 600

63.3 2.7 9.7 0 17.5 6.3 0.5 100.0

N orig (ENPC) % 52.5 2.8 12.2 0.3 24.8 5.3 2.0 99.9

* In coordinated main clauses starting with a coordinating conjunction (not-sentence initial), the first constituent after the conjunction has been considered.

ENPC. As regards fronted objects, however, the apparent difference between the NICLE texts and the English ENPC texts may have a simple explanation: almost all the fronted objects in the ENPC texts are instances of direct speech preceding their matrix clause. As such constructions are less likely to occur in argumentative prose, the learners’ avoidance of fronted objects seems appropriate.

4.2 Types of experiential theme An analysis of experiential themes shows what kind of content is verbalized first in a sentence; typically what type of participant is used as a starting point. The material was thus analysed for experiential structure according to Systemic-Functional Grammar (Halliday 2004: 260) with participant roles depending on the process type of the verb. The results are shown in Table 2, building on a similar investigation of original texts in English and Norwegian in the ENPC (cf. Hasselgård 2004: 192). Note that the category of ‘empty’ does not occur in Halliday (2004) and has instead been inspired by Berry’s (1995: 66) ‘pass option’ as regards

. For the studies reported in Hasselgård (1997) and (2004) only a small part of the ENPC was used; 150 T-units were selected randomly from each of four texts in each language. . The process types are associated with participants are as follows: Material processes (of ‘doing’ and ‘happening’): actor and goal; relational processes (of ‘being’ and ‘having’): carrier and attribute (predicative) or token and value (equative); mental processes (of ‘sensing’ and ‘thinking’): senser and phenomenon; verbal processes (of ‘saying’) sayer and verbiage; behavioural processes (physiological, non-directed activity): behaver; existential processes; existent; cf. Halliday (2004: 260).


Table 2. The ten most frequent types of experiential theme in English and Norwegian (from the ENPC) and in NICLE English original N % 1. 2. 3. 4. 5. 6. 7. 8. 9. 10.

actor carrier circumstance verbiage senser sayer empty behaver token goal

164 117 98 56 51 27 20 18 17 11

27.3 19.5 16.3 9.3 8.5 4.5 3.3 3.0 2.7 1.8

Norwegian original N % 1. 2. 3. 4. 5. 6. 7. 8. 9. 10.

actor circumstance carrier senser verbiage empty process sayer behaver token

122 115 83 72 66 29 30 26 22 10

20.3 19.2 13.8 12.0 11.0 4.8 4.7 4.3 3.7 1.7

NICLE 1. 2. 3. 4. 5. 6. 7. 8. 9. 10.

actor circumstance carrier senser empty token goal sayer value phenomenon

N

%

145 136 104 84 77 49 41 33 33 18

20.9 20.0 19.5 11.3 7.4 5.2 4.4 3.6 3.5 1.3

thematic choice; i.e. the option of filling the thematic slot with a non-referential element such as the anticipatory it or the existential there. Like Table 1, Table 2 suggests that the learners are using Norwegian patterns in their written English. The role of circumstance corresponds to the syntactic function of adverbial adjunct, and is used more frequently both in Norwegian originals and in NICLE than in English originals. The roles of actor and carrier are typical subject roles (in material and relational clauses, respectively), and are frequent as themes in all three subcorpora. It is perhaps noteworthy that the roles of senser (‘experiencer’) and sayer are so frequently thematized in NICLE as I did not expect to find as much reported speech and thought in argumentative essays as in fiction. The discourse functions of these roles are further explored below.

4.3 Initial adverbials As shown in Table 1, initial adverbials are overused in NICLE as compared to English original texts. All the major types of adverbials are frequent; 162 adjuncts, 56 disjuncts, and 56 conjuncts. Example (1) seems typical of ‘NICLE English’. The fronted adverbial is not ungrammatical, but represents a slightly more marked choice in English than in Norwegian; it seems to announce a comparison between Norway and other countries which in fact does not come in the text. The sentence

. According to Halliday (2004: 257) the anticipatory it will usually be a carrier in a relational process, while the existential there is not assigned an experiential role.


Table 3. Use of textual themes (co-ordinating conjunctions, conjuncts and continuatives) Textual themes

sentence-initial co-ordinating conjunction non-sentence-initial conjunction conjunct continuative

N

NICLE SWICLE* LOCNESS* % of total no of % of total no of % of total no of T-units (N = 938) T-units T-units

59

6.3

108

11.5

63 14 244

6.7 1.5 26.0

27.1

21.2

* Figures from Boström Aronsson (2005: 75).

could be made less marked by means of the paraphrase shown in (1a), incorporating the information of the adverbial in the subject NP. (1) In Norway the prisons are quite nice comparing to prisons in other countries. a. Norwegian prisons are quite nice compared to prisons in other countries.

As Norwegian is a verb-second language, it was expected that the V2 rule of Norwegian might be erroneously transferred into the English of the weaker students, thus causing S-V inversion after an initial adverbial. However, there was only one such error in the essays studied, namely (2) below. (2) Perhaps was the successful snatch what triggered a more serious crime, changed their lives and made them suffer.

The NICLE texts also seem to have an overuse of conjuncts and other textual themes similar to that reported by Boström Aronsson (2005: 75) and Drew (1998: 126); cf. Table 3. This is another feature that does not constitute a grammatical error, but which marks learner English off from NS English. Furthermore,

. Drew (1998: 130) reports a higher frequency of premodified NPs in native speaker writing than in learner writing. (1) is an example where NP premodification would be a good substitute for an initial adverbial. . The tendency was confirmed by Eia (2006), who found an underuse of linking adverbials in NICLE compared to LOCNESS, but a great overuse of sentence-initial and and but, which more than compensated for the underuse of linking adverbials.


40 35 30 25 20 15 10 5 0

NICLE NS student essays conversation NS academic writing ex is

ex i tcl e tra te ft p os nti al i tio n

wh -c

lef

t

re ve rs ed

NS press editorials wh -c le ft

Figure 1. Special theme constructions in NICLE and four genres of the ICE-GB (frequencies per 10,000 words)

the use of continuatives such as well, yes and no, may not be appropriate to the genre. This topic will, however, have to await further study.

4.4 Special theme constructions A notable feature of Table 1 was the clear overrepresentation of anticipatory subject in the NICLE texts as against ENPC texts. In Figure 1, the two types of anticipatory subject constructions have been distinguished (existential there and it-extraposition). These are viewed by Gómez-González (2001: 245ff.) as ‘special theme constructions’ along with clefts and dislocation constructions. Because of the overuse of clefts in SWICLE texts reported by Boström Aronsson (2005: 97) these constructions were also counted for Figure 1. The NICLE texts have been compared to four genres of NS English from the ICE-GB: untimed student essays, academic writing (by professionals), press editorials, and conversation. Figure 1 shows that the main reason for the high number of anticipatory subjects in NICLE is the massive overuse of extraposition in relation to all the other genres, including NS press editorials, which are said by Herriman (2000: 224) to

. It-extraposition has been broadly defined, similarly to Kaltenböck (2005: 126), as not requiring a predicative in the matrix predicate. The extraposed constituents are of the following types: that-clause, wh-clause (including indirect questions with if), to-infinitive, ing-clause. Unlike Kaltenböck, I have not included NP + relative clause.


be the genre where extraposition was most frequent in her material. A possible explanation for this overuse may be found in the discourse functions of extraposition in the NICLE essays. These will be briefly explored below. In contrast to the Swedish learners (cf. Boström Aronsson 2005: 97) the Norwegians do not seem to overuse cleft constructions, with the possible exception of reversed wh-clefts. Contrary to expectation there were few errors among the anticipatory subjects; only three instances of it for there in existential constructions, (3)–(4), and one example of there with extraposed clause (5). (3) It is always a risk that it can happen again.

(4) Maybe it is more to it than just to make them to quit criminal activities. (5) Therefore I think there will take more than a decade or a couple of governments to figure out what works.

5.

The use of extraposition

According to Herriman (2000: 223), extraposition is a means of expressing attitudinal meaning, and as such it has the following main functional properties: – to make the attitudinal meaning into an explicit and negotiable proposition, by expressing it in a separate (matrix) clause – to make it possible to elaborate the description of the attitudinal meaning, since the matrix predicate can be modified (cf. example 6) – to make it possible to conceal the source of the attitudinal meaning – to make the attitudinal meaning into the perspective from which the content of the extraposed clause is interpreted; i.e. thematization of the attitudinal meaning (cf. Thompson (2004: 152), who uses the term ‘thematized comment’ for this construction). Examples with a modified predicative, such as (6), are, however, rare in the material. Slightly more commonly, the predicate is modalized, either by means of a modal verb or by means of a disjunct adverbial, both of which are shown in (7).

. Kaltenböck (2005: 145) similarly reports that extraposition of ‘new complements’ is particularly frequent in persuasive writing. . The use of anticipatory it vs. existential there is a classic topic in English grammar teaching in Norway, as Norwegian uses the pronoun det in both constructions (e.g. Dypedahl et al. 2006: 65). Note that the context of (3) shows that the construction is intended to be existential.


(6) It was almost impossible to keep it clean, and it was always sticky and dusty. (7) Of course, it could be argued that crime does occasionally pay.

When the predicate of the matrix clause expresses the modal meanings of usuality and probability, as in (6), Halliday (2004: 615) regards the construction as an ‘objective’ and ‘explicit’ expression of modality.10 Although the typical extraposition structure does not include the source of the attitudinal meaning, the source may be added, as shown in (8). (8) In the second line to the top/bottom I presume it is possible one day to find bad genes and select or manipulate them.

The source of the attitude is, however, not normally given. The reader is left to assume that any opinion expressed is that of the writer, or indeed that the assessment is more objective. It is thus Herriman’s first and last functions of extraposition that are most evident in the NICLE essays; the learners use the construction most often for the thematization of evaluation or opinion. What they evaluate are typically degrees of difficulty, importance, expectedness and appropriateness. Almost as frequent are expressions of modality; possibility/necessity and truth/ fact. Extraposition is also used to thematize evidentiality, sometimes implying the writer as senser. Conversely, it can place the source of the attitudinal meaning with somebody other than the writer. A survey of extraposition meanings found in NICLE is given in Table 4. Boström Aronsson (2005: 88ff.) uses the term ‘objective interpersonal metaphor’ for the extraposition construction, thereby emphasizing that its main discourse function is to convey an attitudinal meaning while making it seem objective. The high frequency of objective interpersonal metaphors might lead one to assume that the learners are using the construction in order to avoid mentioning themselves – more or less as a substitute for I think. However, if this is the case, it does not preclude the frequent use of self-reference in the NICLE essays, and it is to such ‘subjective interpersonal metaphors’ (Boström Aronsson 2005: 84) that we now turn.

10. Usuality, roughly corresponding to indefinite frequency, is a modal category in Halliday (2004: 147).


Table 4. Patterns and meanings in clauses with extraposition in NICLE (raw frequencies) Meaning

Expression

evaluation / opinion

it is hard/difficult to (7); it is easy to (3); it is important to / that (3); it is strange that/how (2); it does not matter if (2); it is better that

18

possibility / necessity

it is possible to (3); it was impossible to; it might be that; it is likely* to; it is certain that; it was necessary to (2)

9

truth / fact

it is true that; it is the case that; it is a truth that; it is a fact that

4

subjective evidence

it looks like* (3); it appears that; it seems (to me) that; it seems like*; it didn’t strike me that

7

external source / evidence it could be argued that; it is said that; it is generally agreed that; it is prohibited by law to (3); it’s up to the governments if; it was her idea that

8

other

it takes more work to; it is not usual to; it is needless to say that; it is a question if; it depends on X how

5

* It looks/seems like and it is likely to are used in extraposition constructions by the learners: It looks like it can be hard to separate between what is reality and what is TV-entertainment. ; … it is likely to get the impression that from an economic point of view crime does pay after all.

6.

Self-reference and subjective stance

A striking feature of the NICLE essays is the frequent use of self-reference; i.e. direct reference to the writer as writer or as participant in a story. This is shown mainly in the use of first-person singular pronouns, but there are also first person plural pronouns where the reference is not general, but specifically includes the speaker and a limited number of other people. Seventeen per cent of the T-units in the material have clear cases of self-reference, almost half of which occur in subject position.11 This agrees with the observation of Hyland: In addition to announcing the writer in the text, pronouns typically occur in the thematic position in the clause. While the important focus of academic writing tends to be the events or concepts under discussion in the rheme, the choice of

11. Incidentally, the pronouns I, we and you all come out as keywords by comparison with the LOB corpus. A keyword is a word that ‘is found to be more frequent in one individual text than its frequency in a reference corpus would suggest’ (Scott 2001: 48). Keywords are calculated using a function of WordSmith Tools.


first position is very significant. The way a writer begins a clause not only foregrounds important information, firmly identifying the writer as the source of the associated statement, but also helps the writer control the social interaction in the text. (2002: 1093)

Table 5 shows the frequency of the first-person singular pronoun as subject in NICLE and the four genres of ICE-GB. I is clearly overused as subject compared to all the NS written genres, though it is used far less frequently than in conversation. One possible explanation is that the essay questions may seem to invite personal views, thus triggering the thematization of the writer as the source of an opinion. However, in press editorials, a clearly opinionated genre, I (referring to the writer) does not occur at all, and we occurs as subject only 15.4 times per 10,000 words. This suggests that the learners apply interactive features from conversational English to their written output as regards the use of first-person subjects. Incidentally, Petch-Tyson (1998: 112) found a greater frequency of interpersonal involvement among Finnish and Swedish learners than in the Dutch and French ICLE components and the American NS control corpus. It is, furthermore, an interesting sign of cultural differences that Hyland (2002: 1098), found a clear underuse of first-person subjects in the English of Hong Kong learner English (expository texts) as compared to NS texts. For an interpretation of the avoidance of first-person subjects, see Hyland (2002: 1107ff.). The majority of self-reference subjects in NICLE were found with mental and verbal processes (typically ‘think’ and ‘say’, respectively), which explains the frequent occurrence of senser and sayer as theme pointed out in connection with Table 2 above. Like extraposition, the constructions with I as senser or sayer typically present the writer’s evaluation or opinion of the state of affairs presented in a following subordinate clause. In choosing these constructions, writers thematize themselves as well as their subjective perspective on a state of affairs. Examples are given in (9)–(11). (9) I believe many also regret it when they get caught. (10) I think many answers will be required at the same time. Table 5. The use of I as subject (frequency per 10,000 words) NICLE ICE-GB conversation ICE-GB student essays ICE-GB academic writing ICE-GB press editorials

53.2 285.4 13.6 5.3 0


Table 6. Patterns of subjective stance (raw frequencies) Meaning

Expression

belief / assumption (probablilty)

I believe (14); I think (11); I don’t think (3); I guess (3); I suppose

32

prefacing own opinion (evaluation)

I would say (6); I would like to say; I’m sorry to say (2); I would argue; I must emphasize

11

truth / fact

I know (2); I remember

3

external source

I’m not saying (3)

3 49

(11) I would argue that many people take this calculated risk, people with low incomes all the way up to the incredibly wealthy.

Petch-Tyson reports similar findings for French, Dutch and Swedish learners: ‘…a main function of I seems to be to talk about the writer within the context of the piece of discourse, either saying something about the writer functioning within the text or what the writer thinks’ (1998: 111–114). In the present study, constructions such as those shown in (9)–(11) have been subsumed under the term ‘subjective stance’. The patterns and the meanings they express are summarized in Table 6. As shown in Table 6, the subjective stance markers are most frequently used to express the modal meaning of probability, thus functioning as a hedge as regards the content of the following that-clause while at the same time thematizing the writer as source. The thematization of writer as source is the main function of the second most frequent subjective stance meaning, viz. the prefacing of the writer’s own opinion or evaluation. Like extraposition, subjective stance markers can be used to mark something as fact, and – curiously – as coming from a different source. Subjective stance markers have been found to be a common feature of learner English, cf. Biber and Reppen (1998: 155), Ringbom (1998: 44), and Boström Aronsson (2005: 85f.). To get an impression of the frequency of these subjective stance expressions, the most frequent ones were investigated in ICE-GB for comparison (I think, I believe, I would say, I don’t think). As Figure 2 shows, the frequency of subjective stance expressions in NICLE is far higher than in the academic (written) genres of ICE-GB, and approaching the frequency for conversation. It is also slightly higher than the corresponding figures for SWICLE. None of the subjective stance markers considered for Figure 2 were found in press editorials, which is why this genre is not included in the figure.


40 35 30 25 20 15 10 5 0

subjective st ance NI C S W NS NS N ICL s S LE E tu de na ca de co n ve t wr m ic r sa ti i tin g wr i tin on g

Figure 2. Frequencies per 10,000 words of some subjective stance markers in NICLE, SWICLE and three genres of ICE-GB12

A note may be made of the expression I would say. Some readers may have objected to its being said to signal opinion/evaluation in Table 6 above, as it is not typically used with this meaning by native speakers. Example (12), from ICE-GB, shows the more characteristic use in NS English, namely to mark the content of the complement clause as approximate. Alternatively, the expression can be used quite literally, as in (13). (12) He was in his late forties I would say

(13) And I would say to the honourable lady uh as I’ve said to my honourable friend that … (14) I would say that in any crime for, profit or pleasure, where you actually get away with it, it does pay.

In NICLE, on the other hand, the expression typically occurs towards the end of the essay, often prefacing (part of) the conclusion, as in (14). The expression is also clearly overused with 6 occurrences in the 15,000-word sample of NICLE. In the whole corpus (approximately 210,000 words), it occurs 27 times as against only 18 times in the whole ICE-GB (1 million words).

12. SWICLE figures have been taken from Boström Aronsson (2005: 84). Her table summarizes all types of interpersonal subjective metaphor and thus includes more expressions.


NICLE

14

SWICLE LOCNESS NS editorials

12 10 8

NS conversation

6 4 2 0 certainly

surely

of course

indeed

probab ly

perhaps

mayb e

possib ly

Figure 3. Adverbials of stance and modality in NICLE, SWICLE, LOCNESS and conversation and editorials from the ICE-GB (frequencies per 10,000 words)

7.

Other markers of stance

The overrepresentation of both extraposition and subjective stance markers in the NICLE material might lead one to assume that students should be trained to use other stance markers instead, such as disjuncts. However, in a study by Aijmer (2002) it emerges that Scandinavian students overuse this type of stance marker too. On the whole, Norwegian and Swedish learners seem extremely keen to express modality and evaluation explicitly in their texts. Figure 3 shows the frequencies of some adverbials of stance and modality in NICLE, SWICLE, LOCNESS and two registers of ICE-GB.13 Figures for SWICLE and LOCNESS have been taken from Aijmer (2002: 67). Compared to LOCNESS, Swedish and Norwegian learners overuse all these adverbials except surely, indeed and possibly. Compared to editorials, however, the overuse is less pronounced. The clearest cases of overuse in NICLE are found with of course, probably and maybe. Figures for NICLE and SWICLE follow each other rather closely, although the tendency to overuse disjuncts is even more marked in NICLE than in SWICLE, with 43.9 and 33.5 occurrences per 10,000 words, respectively. In conversation, the corresponding frequency is 25.9. Notably, the adverbials that are most frequent in conversation are also most clearly overused in NICLE in relation to the other written genres. However, the learners cannot really be said to copy the patterns of conversation, but rather to overuse them, presumably unaware that they are not appropriate to the genre of writing.

13. All occurrences of the adverbials have been counted, not just those in thematic position.


Table 7. Patterns that introduce other voices projected counterargument

you could/may say that (2); some may/might say that (2); one can say that (2); many think; let’s say that

8

pseudo-interaction

you think; you could have guessed; you may ask; I hear you saying

4

8.

Other voices

With their apparently massive urge to express their own opinions, one may wonder if the students represented in NICLE leave any space at all for other voices. However, some patterns are found that explicitly introduce a different voice in the thematic section of the T-unit, although they are not particularly frequent. As shown in Table 7, the ‘other voices’ have two main functions: to introduce a counterargument, as shown in (15)–(16), and to represent a pseudo-interaction with the reader, as illustrated in (17)–(18). (15) I do not think so, even though some may say that it is too easy to get away with criminal behaviour in Norway. (16) Many think the time to serve is worth the risk. But altogether the punishments both from state and society are not worth a possible gain. (17) And yes you could have guessed it, he got caught. (18) Would you push the button if it meant crime would be abolished?

As shown by both the frequent use of self-reference and the attempts at interaction with the reader, the ICLE material shows a rather interactive style of writing. Further indicators of this are a relatively frequent use of questions (5.4 per cent of the main clauses) and imperatives (2 per cent of the main clauses). By comparison, only 1.7% of the main clauses in the ICE-GB student essays are interrogative and 0.1% are imperative.

9.

Concluding remarks

Two main questions were asked at the outset of this paper: (i) Do Norwegian students apply Norwegian patterns in their choice of thematic structure? and (ii) To what extent do Norwegian learners (over-)use features from informal conversation in their written language? The answer to the first question seems to be ‘yes’,


though with some modifications. The use of initial adverbials seems to follow Norwegian patterns; i.e. adverbials are placed clause-initially more frequently than in NS English. This does not constitute a grammatical error, but contributes to giving the learner English a ‘foreign accent’ (Shaw 2004: 81). Another trace of a Norwegian accent in the NICLE essays was the frequency of empty themes (anticipatory subjects), which was also more like Norwegian than native English. However, the overuse concerned mostly it extraposition, which was found to be overused also in comparison with native Norwegian. Like previous studies of Scandinavianproduced English (e.g. Boström Aronsson 2005), the present study found that the NICLE material contains more textual themes (conjuncts and conjunctions) than comparable NS material. There were hardly any errors with S-V vs. V2 order or fronted predicatives and objects; the learners thus seem to master the syntactic differences between English and Norwegian sentence openings. As to the question of features from informal conversation, the NICLE essays exhibit an interactive writing style with a high degree of writer and reader visibility. Subjective stance markers are greatly overused in NICLE compared to written genres of English, and are in fact only slightly less frequent than in NS conversation. As regards other expressions of evaluation and modality it is less clear whether they are due to influence from informal English conversation or from the mother tongue. However, the comparison with Norwegian suffers from a mismatch of genres. There is a particular need for investigating more genres of spoken and written Norwegian, including argumentative prose by professionals as well as students. A striking feature of the NICLE essays included in this investigation was the high frequency of expressions of modality, opinion and evaluation. Two constructions were selected for closer scrutiny: extraposition and subjective stance markers. They are similar to the extent that they thematize modal and attitudinal meaning and place the ideational content in a dependent clause at the end of a sentence. However, subjective stance markers foreground the source of the modal/attitudinal meaning, while the extraposition construction typically leaves the source of attitudinal meaning implicit. Still, the two constructions do not seem to be in complementary distribution in learner English. On the contrary – they are both overused in relation to NS writing. A similar overuse is found for other stance markers, such as modal adverbials. This could perhaps be linked to the overall interactive flavour of the texts, in which the writers are very much present, conveying their views to the readers. Any conclusions based on the findings of the present study must be very tentative, due to the small size of the material studied (15,000 words) and to the exploratory nature of the investigation. However, the following seems to be the case: The students by and large master the grammatical structures of sentence


openings, but not the discourse conventions of argumentative / academic writing. Thus it is possible that language instruction at this advanced level could focus less on e.g. the grammar of it/there and the use of (projected) complement clauses. The students may need to learn how to vary their repertoire of stance markers, but above all they need to focus on the association between linguistic expressions and genre / style level / medium and on widening their repertoire of styles (e.g. formal / informal, impersonal / interactive). They need to learn not only how to use extraposition and other stance markers correctly, but also when it is appropriate to tone down the writer’s explicit presence in the text. Finally, the present study is a good example of how the exploration of learner corpora can lead to insights that were not even sought for at the outset of the investigation. The initial focus was on syntactic structure and thematic roles. It was in examining the discourse meanings underlying the quantitative differences between NICLE and the other corpora as to choice of theme that the predominance of stance markers appeared, thus moving the focus slightly from the ideational and syntactic features of sentence openings to the interpersonal functions of thematization. It seems clear to me that both the discourse-analytical and the contrastive perspective are indispensable in analysing advanced learner language with a view to identifying areas in which the learners need to adjust or improve their performance.

References Ädel, A. 2006. Metadiscourse in L1 and L2 English. Amsterdam: John Benjamins. Aijmer, K. 2002. Modality in advanced learners’ written interlanguage. In Computer Learner Corpora, Second Language Acquisition and Foreign Language Teaching, S. Granger, J. Hung & S. Petch-Tyson (eds), 55–76. Amsterdam: John Benjamins. Berry, M. 1995. Thematic options and success in writing. In Thematic Development in English Texts, M. Ghadessy (ed.), 55–84. London: Pinter. Biber, D. & Reppen, R. 1998. Comparing native and learner perspectives on English grammar: A study of complement clauses. In Granger (ed.), 145–158. Boström Aronsson, M. 2005. Themes in Swedish Advanced Learners’ Written English. PhD dissertation, Göteborg University. Drew, I. 1998. Future Teachers of English: A Study of Competence in the Teaching of Writing. Kristiansand: Høyskoleforlaget. Dypedahl, M., Hasselgård, H. & Løken, B. 2006. Introducing English Grammar. Bergen: Fagbokforlaget. Eia, A.-B. 2006. The use of linking adverbials in Norwegian advanced learners’ written English. Unpublished MA thesis, University of Oslo. Gilquin, G. 2003. The Integrated Contrastive Model: Spicing up your data. Languages in Contrast 3(1): 95–124.


Gómez Gonzáles, M. A. 2001. The Theme-Topic Interface: Evidence from English. Amsterdam: John Benjamins. Granger, S. 1996. From CA to CIA and back: An integrated approach to computerized bilingual and learner corpora. In Languages in Contrast. Papers from a Symposium on Text-based Cross-linguistic Studies, K. Aijmer, B. Altenberg & M. Johansson (eds), 37–52. Lund: Lund University Press. Granger, S. (ed.). 1998. Learner English on Computer. London: Longman. Halliday, M. A. K. 2004. An Introduction to Functional Grammar, 3rd edn, revised by C. Matthiessen. London: Arnold. Hasselgård, H. 1997. Sentence openings in English and Norwegian. In Corpus-based Studies in English. Papers from the Seventeenth International Conference on English Language Research on Computerized Corpora, M. Ljung (ed.), 3–20. Amsterdam: Rodopi. Hasselgård, H. 2004. Thematic choice in English and Norwegian. Functions of Language 11(2): 187–212. Hasselgård, H. 2005. Theme in Norwegian. In Semiotics from the North: Nordic Approaches to Systemic Functional Linguistics, K. L. Berge & E. Maagerø (eds), 35–48. Oslo: Novus. Herriman, J. 2000. The functions of extraposition in English texts. Functions of Language 7(2): 203–230. Hyland, K. 2002. Authority and invisibility: Authorial identity in academic writing. Journal of Pragmatics 34: 1091–1112. Kaltenböck, G. 2005. It-extraposition in English: A functional view. International Journal of Corpus Linguistics 10(2): 119–159. Petch-Tyson, S. 1998. Writer/reader visibility in EFL written discourse. In Learner English on Computer, S. Granger (ed.), 107–118. London: Longman. Ringbom, H. 1998. Vocabulary frequencies in advanced learner English: A cross-linguistic approach. In Granger (ed.), 41–52. Scott, M. 2001. Comparing corpora and identifying key words, collocations, frequency distributions through the WordSmith Tools suite of computer programs. In Small Corpus Studies and ELT, M. Ghadessy, A. Henry & R.-L. Roseberry (eds), 47–65. Amsterdam: John Benjamins. Shaw, P. 2004. Sentence openings in academic economics articles in English and Danish. Nordic Journal of English Studies 2(3): 67–84. Thompson, G. 2004. Introducing Functional Grammar. 2nd edn. London: Arnold.

Corpora used The English-Norwegian Parallel Corpus (ENPC): www.hf.uio.no/ilos/OMC/English/Subcorpora.html The International Corpus of English, British component (ICE-GB): www.ucl.ac.uk/english-usage/ice/icegb.htm The International Corpus of Learner English (ICLE) – Norwegian and Swedish components (NICLE / SWICLE): cecl.fltr.ucl.ac.be/Cecl+++Projects/Icle/icle.htm

The usefulness of corpus-based descriptions of English for learners The case of relative frequency Susan Hunston

The paper argues that some of the most interesting information about word frequency is difficult to present in a way that is useful to learners. This is particularly true when the issue is not one of absolute frequency but of frequency that is relative and dependent on complex variables. The first instance of this is phrases which are not frequent in themselves but which can be counted as phrases because of the significance of the unit to each of its components. In cases such as these the unity of a perceived multi-word unit depends on the strength of collocation between progressively longer sequences. The second instance explores the factors influencing the relative frequency of wordforms within a lemma. These include complementation pattern and modality. It is hypothesised that a relatively high frequency of the base wordform is likely to co-occur with a high frequency of modal-like expressions. The third and final instance is ‘semantic sequences’, which describe what is often said, though not what should be said. These emerge as a consequence of a combination of discoursal constraints, including the phraseology of a discipline. The paper raises questions concerning the relevance of such frequency information to language teaching and the difficulty of translating an approach to describing a corpus into a useful tool for language learners and teachers.

1.

Introduction

Discussions of the relevance of corpus studies to language learning or teaching tend to focus on either the use of corpora by learners, as an awareness-raising tool, or on the presentation to learners (and teachers) of information derived from corpora. This paper focuses on the second of these. It links information about corpus frequency to the concept of phraseology, something which is increasingly believed to be important to learners. The paper argues, however, that

142 Susan Hunston

some of the most interesting information on this topic is difficult to present in a way that is useful to learners. The tradition of applied corpus study that this paper is based on is that which provides information about words, phraseology and grammar to learners, typically in reference books such as dictionaries and grammars, based on notions such as the idiom principle and units of meaning (Sinclair 1991, 2004). Behind this tradition is the assumption that phraseology is key to language description, that the behaviour of individual words drives rather than is driven by grammatical description, and that the investigation of individual words can form the basis of a language learning syllabus, with grammar being taught as an area of lexis rather than as separate from it. A further assumption is that corpus research should as far as possible avoid preconceptions about linguistic categories (the ‘corpusdriven’ approach, see Tognini-Bonelli 2001). The idiom principle proposes that most naturally-occurring language consists of a series of chunks rather than a series of independent words. This facilitates language processing for the hearer (whereas a single word may have multiple interpretations, a group of words typically has only one) and for the speaker (only one choice is made instead of a series of choices) (Sinclair 1991: 110–115). The object of investigation when looking at the idiom principle is phraseology. This can be conceived of in terms of ‘fixed phrases’ or ‘lexical bundles’, that is, as strings of letters and spaces (in written language or in transcribed speech) that are treated as if they were single words (e.g. Biber et al. 1999: 990). An alternative view is that of ‘units of meaning’, that is, sequences that are fixed in essence but not in actual wording. An example might be Sinclair’s observations of the word brink, which very frequently occurs in the sequence on the brink of, which in turn is preceded by a number of verbs including teetering, tottering and hovering and is followed by a noun phrase indicating a momentous personal event (Sinclair 1999b). Although the middle part of the unit of meaning is relatively fixed in terms of wording, the outer parts are fixed only in terms of general meaning. For language teachers, both these approaches have obvious applications. Students can be taught lexical bundles in the same way as they are taught any vocabulary item. They can also be made aware of the typical uses of words such as brink, such that learning the phraseology associated with a word becomes as much part of learning a word as its pronunciation or part of speech. Central to both approaches in the concept of the frequency of words and sequences of words in a corpus. Lexical bundles are identified as strings of two, three or four words that occur together more frequently than other strings. Units of meaning are frequently-occurring sequences of kinds of words, rather than recurring strings, but frequency is equally crucial in identifying typicality. Related

The usefulness of corpus-based descriptions 143

concepts such as ‘collocation’ depend on relative frequency, where one item is more frequent in a given context than in a corpus overall. In this paper I shall discuss instances not of absolute frequency but of frequency that is relative and dependent on specified but often complex variables. Unlike Sinclair (1991: 110), who deals with instances of ‘simultaneous choice’, where it can be argued that selection of the core of the unit of meaning makes selection of the rest of the unit almost inevitable, I suggest instances where selection of one item only alters the probability of selection of other items. This skewing of probabilities, I suggest, is important to the description of language in a corpus, but is difficult to present to learners. I shall use two examples that illustrate the role of relative frequency: phraseology and relative frequency, and the interaction between individual wordforms, their associated patterns, and modal meaning.

2.

Phraseology and relative frequency

Multi-word units are often ranked in the same way as individual words, in terms of their raw frequency. Each sequence of letters and spaces is treated as unique and is given a measurement of frequency, either raw or normalised. Typically, the longer the sequence the less frequent it is. For example, Biber et al (1999: 992–993) note that in both conversation and academic prose the ratio of five-word, four-word and three-word ‘lexical bundles’ is about 1:10:100. As a result, in deciding which bundles to include in their study, they take a cut-off point of ten occurrences per million words for up to four-word bundles but of only five occurrences per million for five-word bundles. An alternative view is to recognise not the absolute frequency of the whole unit, but the significance of the unit to each of its components. In other words, it is possible that the unity of a perceived multi-word unit depends on the strength of collocation between progressively longer sequences (see Danielsson 2007 for a proposed method of automating searches for this). For example, the phrase after a few moments of occurs only 11 times in the 450 million-word Bank of English (BoE). No conceivable study of lexical bundles would consider items that occur only so rarely. A study of the typical phraseology of moments would hardly identify this either. I suggest, however, that it has a unity which depends on conceptualising it as a progressively lengthening sequence, where each additional item collocates with the preceding items taken together, thus:

. The Bank of English is a general corpus of 450 million tokens jointly owned by HarperCollins publishers and the University of Birmingham.

144 Susan Hunston

– The most frequent collocate of moments is of. The sequence moments of accounts for 20% of the total occurrences of moments. – The second most frequent left collocate of moments of is few. The sequence few moments of accounts for 5% of the total occurrences of moments of. – The most frequent collocate of few moments of is a. The sequence a few moments of accounts for 56% of the total occurrences of few moments of. – A frequent collocate of a few moments of is after. The sequences after a few moments of accounts for 13% of the total occurrences of a few moments of. As each longer sequence is considered, the probability of co-occurring words alters. If moments is chosen, moments of is likely. Moments of has a wide variety of co-occurrences, but few is one of the most frequent. Once few moments of is selected, a few moments of is very likely. Given a few moments of, after is the most likely choice, followed closely by for and within. A further example can be given using the very frequent word time (Nakamura 2003). Below I shall consider the sequence it at the time, which occurs 562 times in 450 million, again below Biber et al’s threshold, and which does not constitute an identifiable grammatical unit. The most frequent left collocates of time are:

the time this time first time a time of time

155 per thousand words 66 ptw 62 ptw 45 ptw 45 ptw

In passing, it is striking to note that the time and a time occur with markedly different frequencies, adding corroboration to Sinclair’s point (Sinclair 1999a) that in a given context there may be no free choice between apparently paradigmatic items. The time is more frequent than a time, either because time more frequently occurs in a specific than in a non-specific context (e.g. by the time I had finished is more typical than at a time of my own choosing), or because idiomatic phrases containing the time (e.g. have the time of one’s life) occur more frequently than those containing a time (e.g. Once upon a time). To continue the investigation using the same methodology as above, we then consider the left collocates of the time, again treating this sequence as though it was a single word. The most frequent left collocates in descending order are:

at the time by the time + clause all the time (most/some) of the time for the time (being)


Thus, the phrase at the time can be treated as a unit, not only because of its overall frequency (though it is of course very frequent) but also because of the strong collocation between the and time and between the time and at. The investigation can continue treating at the time as a single word. It most often occurs at the beginning of a clause, very frequently at the beginning of a paragraph. However, the third most frequent left collocate is it. As noted above, this occurs 562 times in the Bank of English (accounting for 0.07% of the 705,993 instances of time and 1.7% of the 32,862 instances of at the time). The most frequent left collocates of it at the time are words indicating thought or knowledge, such as know, aware of, recognise. Furthermore, the clauses in which these items appear are usually negative, and there is often an indication of contrast, such as but or although in the context. The it is invariably cataphoric. In other words, the following represent one relatively frequent use of the phrase at the time: I didn’t know it at the time but… Although he wasn’t aware of it at the time… Though we didn’t recognise it at the time…

It is apparent that these perform a single discourse function: creating suspense before a key point in a narrative is introduced. There is a unity to the sequence, not because it is particularly frequent, certainly not in relation to time, but because it is built up of items that are statistically important to one another: at to the time, it to at the time, and know / aware / recognise to it at the time. There is an argument, then, that the sequence ‘negative + be aware + it + at the time + contrast’ is a viable unit, though it is neither a lexical bundle nor a unit of meaning. What is not clear is how this information is to be made available to learners. A sequence of this kind is unlikely to appear in a vocabulary list, even one that includes phrases as well as individual words. It can hardly be said to be an essential part of a learner’s knowledge of how the word time is used, as its frequency is low relative to other sequences involving that very common word. It is conceivable that it would appear as part of a conscious-raising exercise based on a given text, though it is not easy to recognise it as a relatively frequent item except in the context of a corpus search for the phrase at the time. There is even a question as to whether information about sequences such as these is of use to learners, even very advanced ones. After all, it does not represent information about ‘what is correct’ in English, only information about ‘what is (relatively) often said in a particular context’. This is a point we will return to below, but at the moment we must leave these questions unanswered.

146 Susan Hunston

3.

Wordform, pattern and modality

Although the concept of ‘lemma’ has been challenged as a theoretical concept (because the wordforms comprising a lemma often do not behave in ways that suggest they are equivalent items, see Sinclair 1991: 44–51), it is often useful as a practical one. Whilst it may not come as a surprise that the wordforms belonging to a single lemma do not occur with equal frequency (typically, and unsurprisingly, the base and -ed forms occur more frequently than the -ing and -s forms), it is perhaps more remarkable that lemmas differ substantially in the relative frequency of their wordforms. Numerous examples of such apparent discrepancies can be given: below are six examples with information taken from Leech et al. (2001) and based on the British National Corpus (BNC). – – – – – –

fill 35%; filled 47%; filling 14%; fills 5% explain 41%; explained 37%; explaining 9%; explains 13% agree 35%; agreed 58%; agreeing 3%; agrees 4% disagree 54%; disagreed 23%; disagreeing 8%; disagrees 8% appear 36%; appeared 35%; appearing 5%; appears 25% disappear 25%; disappeared 59%; disappearing 9%; disappears 7%

This very basic information raises a number of questions: why is explains so frequent compared to fills? why is agreed so frequent compared to disagreed and disagree so frequent compared to agree? why is appears so frequent compared to disappears and disappeared so frequent compared to appeared? Such differences are known to have a number of explanations, including the type of process represented (Halliday and Matthiessen 2004), the registers in which each lemma is most likely to occur (Matthiessen 2006), and the phraseology of the individual lexical item (Sinclair 1991). It is this final phenomenon that we will explore here, in the context of wordforms and what comes after them and before them. Hunston (2003), for example, considers a number of lemmas including DECIDE. Overall, the wordform decide accounts for 28% of the total lemma, while decided accounts for 61%. These figures are greatly altered, however, if only those instances of DECIDE that occur with a particular complementation pattern (or ‘grammar pattern’ (Hunston and Francis 1999)) are considered. Where DECIDE is followed by a that-clause, for example, 13% of instances occur with decide and 80% with decided, but where DECIDE is followed by a wh-clause, 70% of instances occur with decide and only 9% with decided. It must be remembered at this point that we are discussing wordforms, not verb phrases. Decide does not necessarily represent the present tense and decided is not necessarily past tense. On the other hand it is true that decided is more often associated with a finite verb


phrase incorporating tense (e.g. he decided or it has been decided), and this time reference does correlate logically with the representation in the that-clause of a completed decision, as in: Martin decided that he would move to the Midlands.

However, although decide sometimes occurs as present tense, it more often occurs as a base form following a modal auxiliary. This correlates with the function of a wh-clause indicating a potential rather than actual decision, as in: The court must decide what is best for the children .

Hunston (2003: 38) points out that the base form decide also frequently follows other phrases which express modal meaning although not with a modal auxiliary. Some examples are given here:

Expressing necessity or obligation: …have to decide what …have until time to decide whether …up to them to decide where …forces the United Kingdom to decide whether …is expected to decide what

Expressing possibility or a degree of difficulty: …it took several days to decide whether …attempting to decide what …gives students more freedom to decide what …been able to decide whether …free to decide what …have the right to decide whether …has yet to decide whether

Using the information from DECIDE we might hypothesise that a relatively high frequency of the base wordform is likely to co-occur with a high frequency of modal-like expressions. This might be tested by taking two verbs at random: ACCOMMODATE and ACCOMPANY. All these verbs have in common is their closeness in the alphabet. According to Leech et al.’s (2001: 27) figures, the proportion of each lemma accounted for by the base form is very different: 64% in the case of ACCOMMODATE and only 19% in the case of ACCOMPANY. Perhaps surprisingly, relatively few instances of the base form are accounted for by occurrence of present tense, in either verb (information from the Bank of English). Accompany is a finite present tense in about 20% of its occurrences, accommodate in only 3%. If we calculate how many instances are preceded by to (indicating

148 Susan Hunston

the likely presence of a modal-like expression), the positions are reversed: 70% of the instances of accommodate are preceded by to compared to 48% of the instances of accompany. In other words, the base form accommodate accounts for a higher proportion of its lemma, is relatively infrequently the present tense, and is relatively frequently preceded by to, all compared with accompany. Instances of to accommodate show a number of modal-like expressions such as: Expressing obligation: have to, need to, is to, be asked to, be up to someone to, force someone to, be vital to Expressing difficulty: has yet to, try to, take [time] to, be difficult to Expressing ability or right or willingness: have the chance to, have the right to, be able to, agree to

From the above we might conclude that the frequencies of wordform, complementation pattern and modal meaning are interrelated, with the probability of occurrence of each linguistic feature being affected by the others. In addition, the probability of occurrence of each feature is dependent of the phraseology of the lexical items involved. Again, however, the relevance of this for learners is problematic. It is possible to give learners examples which are ‘typical’ in the sense of representing what combination of features occurs most frequently (as dictionary examples would hope to do), but it would be misleading to suggest to learners that they should restrict their output only to that which occurs most frequently. After all, although 70% of the instances of DECIDE + that-clause in the Bank of English occur with decided, a full 30% do not. Information about relative frequency of this kind is not information about right and wrong but about ‘what is often said’, and the learner might be expected to make a rather different use of this kind of information than he or she would of grammatical rules.

4.

Semantic sequences and the learner

In this paper so far I have spoken in fairly general terms about sequences that express ‘what is often said’. I would now like to give a name to these and refer to them as ‘semantic sequences’. A number of semantic sequences have been identified above, including: – ‘negative + be aware + it + at the time + contrast’ e.g. I didn’t realise it at the time but…


– ‘modal meaning + decide + wh + potential decision’ e.g. It’s up to you to decide what to do Another example of a word-pattern combination that co-occurs with modal meaning more often than might be expected is DISTINGUISH followed either by ‘between plural-noun phrase’ or by ‘noun from noun’. In this case, modal meaning is associated with all three wordforms that have the capacity to be non-finite, i.e. distinguish, distinguished, distinguishing. In addition, the non-finite is surprisingly frequent. Of the 279 instances of distinguishing between in the Bank of English, for example, only 7 (fewer than 3%) are preceded by a form of the verb BE. Examples of modal meaning preceding DISTINGUISH include: – Obligation and necessity: ought to distinguish; must be able to distinguish; the importance of distinguishing; it has been useful to distinguish; there is no need to distinguish – Possibility and difficulty: can be distinguished; ways of distinguishing; could not distinguish; fails to distinguish; it became impossible to distinguish; enable humans to distinguish; capable of distinguishing; had no trouble distinguishing We have therefore information both about what follows and what comes before the verb. Some of this is formal (the verb is often followed by a preposition between or from; distinguish often follows to while distinguishing often follows a preposition such as of) and some is functional (the verb often follows an expression of modal meaning). As a result there are two observable sequences: – ‘necessity’ + ‘distinguish’ + ‘two related things’ – ‘possibility’ + ‘distinguish’ + ‘two related things’ A different kind of example involves nouns frequently followed by appositive that-clauses, that is, nouns with the grammar pattern N that. Two such nouns appear in this text: The history of AIDS took a new twist last week with the disclosure that a seaman from Manchester appears to have had the disease as far back as 1959. In the earliest case of AIDS on record, a team of researchers has shown that stored tissues from the man’s body contained genetic material from HIV. Other scientists said the work could shed new light on the evolution of the virus and the rate at which it mutates. The data, published in a letter in the current issue of The Lancet, have generated intense media interest and speculation that the man became infected in Africa. However, Trevor Stretton, one of the physicians who cared for the man, stressed that there was no firm evidence that he had visited the continent. [New Scientist]

150 Susan Hunston

The sequence the disclosure that occurs 241 times in the BoE. Taking 100 lines as a sample, the following recurring phraseologies can be noted: – The disclosure that is clause-initial in 22 lines. – The disclosure that follows a word or phrase indicating that the disclosure is the cause of something in 48 lines. The most frequently-occurring examples are after, FOLLOW, with. – The disclosure that follows a word or phrase indicating that the disclosure is caused by something in 8 lines. In other words, in over 60% of those lines where the disclosure that is not clauseinitial, the disclosure is expressed as causing some other phenomenon. The text example, where a disclosure causes a ‘new twist’ and ‘speculation’, turns out to be unexpectedly typical. Repeating the exercise with speculation that, we note that: – – – – –

In 29 lines, speculation that follows a variant of there is/was. In a further 5 lines, speculation that is clause-initial. In 29 lines, the speculation is expressed as the outcome of something. Frequent preceding items include led to, prompted, fuelled. 4 lines contain the phrase despite speculation that. In 10 lines the speculation is described as being ended or dismissed.

In other words, most of the lines that indicate causality show the speculation as arising from an event or situation. Again the text example reflects something that is relatively frequent. The commonly-occurring sequences described here can be expressed as – ‘something follows or is caused by + disclosure + that-clause’ – ‘something precedes or causes + speculation + that-clause’ These sequences incorporate a well-known grammar pattern: the abstract nouns speculation and disclosure are frequently followed by appositive that-clauses. They also capture the fact that ‘a speculation that something is the case’ is mostly likely to be the outcome of some process whereas ‘a disclosure that something is the case’ is most likely to produce an outcome of some kind. These observations give us a great deal of detailed information about how the words in question often occur in natural discourse, but the information goes beyond what might legitimately be put in a dictionary, for example. The sequences describe what is often said, but not what should be said. They emerge as the consequences of a combination of discoursal constraints, or even of how scientific discovery is commonly represented, rather than as a consequence of the vagaries of collocation. As description


they provide a possible link between lexis, grammar and discourse. As prescription they are not useful. A more practical aspect of semantic sequences is their occurrence in discourses of specific kinds, where their identification is useful not in accounting for how a given lexical word is used but in recording what is often said (and how) by a given discourse community. One example comes from work by Yoshihito Kamakura on a corpus of texts in English written by Japanese learners (Kamakura 2007). In investigating the learners’ use of prepositions such as in, Kamakura has noticed a frequently occurring sequence which might be expressed as: – ‘modal meaning + representation of discourse + in English’ e.g. For me it is very difficult to write in English. I like to chat in English. It is important to talk more often in English. Unlike most studies of learner corpora, identification of a sequence such as this tells us very little, except incidentally, about the learners’ level of expertise in English. It is neither good nor bad that they produce a number of sentences along these lines, and proving that such sentences are less likely to occur in native speaker English tells us only that language learners talk about language learning, either because they are preoccupied with it or because their teachers set up opportunities for them to do so. This is hardly surprising. Where the identification of such sequences might be useful, though, is in identifying what learners do talk or write about and in devising teaching materials that reflect these common concerns at a micro-level (what individual sentences or utterances are doing) rather than at a macro-level (what whole discourses are about). Teaching the next generation of language learners to produce language such as that exemplified above involves exposing them to: the function of expressing that something is difficult or desirable; the vocabulary associated with language acts such as write, chat, talk and the names of languages; and the fairly trivial grammar point that in is the correct preposition to link a language act with a language name. Knowing which semantic sequences occur frequently allows the materials writer to prioritise those functions, vocabulary, and grammar items that learners most need. The process is circular – what these learners talk about becomes in effect the basis of the syllabus for the next group of learners – but it does have a number of advantages. It offers learners’ own discourse as a learning resource, in a tradition developed by Seidlhofer (2005) for example. It provides an approach to analysing a corpus that is both functional and formal. It is simple for teachers and learners to carry out, as it does not require a native speaker corpus as a reference point.

152 Susan Hunston

Further instances come from work by Nicholas Groom on the phraseology of disciplinary discourses (Groom 2007). To give a very brief example from this work, Groom notes that the phrase it is not surprising that, which is of course fairly common in all academic discourse, tends to occur in published English Literature papers in a particular sequence: – ‘logical cause + it is not surprising + that-clause’ e.g. Given the absolutist nature … it is not surprising that its paths must be restricted… With spatial freedom as a good indicator of power, it is not surprising that Spenser turned elsewhere to cartographic metaphors. Most of Groom’s work involves a study of prepositions in English Literature and in History. To give again only a brief example, he looks at beyond and to and notes a sequence that might be expressed as: – ‘entity’ + ‘existence etc’ + beyond + ‘conventional’ + to + ‘new domain’ Examples include: The “villainy” of those at the lower end of late medieval English society extended beyond vulgarity and repulsiveness to rapacity and violence. (History) De Gaulle proposed moving beyond periodic foreign ministers’ meetings to the creation of a formal political organisation for the Six. (History) Few have gone beyond recounting dramatic episodes taken almost exclusively from the first wave of plague to compare levels of violence before and after the Black Death. (History) [an analysis that may help] us go beyond the contingent events of one revolution and explore the revolutionary process in a broader perspective. (History) Shelley’s interest in drama extended beyond the poetic to the melodramatic. (English Literature)

Examples such as this raise interesting possibilities for the description of academic discourse because they offer an opportunity to describe ‘what is often said’ in functional as well as in formal or phraseological terms. The first example goes beyond the observation that a phrase such as it is not surprising that is common and notes the role of that phrase in argument construction in a very specific disciplinary context. The second example points out that an obvious and unremarkable use of two prepositions can have particular significance in disciplinary contexts.

5.


Discussion and conclusion

The discussion in this paper has consequences for methods of investigating a corpus and for language description. It suggests that raw frequency figures, and particularly comparative figures, reveal anomalies that warrant further investigation. The explanations for such anomalies are very often phraseological. More specifically, investigations of frequency anomalies often reveal the existence of sequences of meaning elements, or semantic sequences. The discussion also suggests that it is the interaction between frequencies that is important, that is, the interaction between the probabilities of occurrence of given language features. The notion of ‘typicality’ is very sensitive to immediate co-text. Questions as to the relevance of this to language teaching have been raised above. These should perhaps be reformulated in terms of the difficulty of translating an approach to describing a corpus into a useful tool for language learners and teachers. It has been proposed that describing phraseology is a matter of identifying units that are longer than a word, but that these units are not (or not only) specified strings of words, neither are they (only) units of meaning. Rather they represent a complex interplay of relative frequency and probability. Recognition of such an interplay is of little use to learners. What has been suggested in this paper is that statistical clues can be used to identify semantic sequences that in turn encode ‘what is often said’ in a given discourse. In themselves, the sequences are not what learners expect, in that they do not distinguish between correct and incorrect sequences. On the other hand, they provide a link between the generalities of lexis and grammar and the specificity of discourse, and they can be expected to interact with what learners wish to talk about most frequently.

References Biber, D., Johansson, S., Leech, G., Conrad, S. & Finegan, E. 1999. Longman Grammar of Spoken and Written English. London: Longman. Danielsson, P. 2007. What constitutes a unit of analysis in language? Linguistik Online 31. http:// www.linguistik-online.de/index.html. Groom, N. 2007. Phraseology and Epistemology in Humanities Writing. PhD dissertation, University of Birmingham. Halliday, M. A. K. & Matthiessen, C. 2004. An Introduction to Functional Grammar, 3rd edn. London: Arnold. Hunston, S. 2003. Lexis, wordform and complementation pattern: A corpus study. Functions of Language 10: 31–60.

154 Susan Hunston

Hunston, S. & Francis, G. 1999. Pattern Grammar: A Corpus-driven Approach to the Lexical Grammar of English. Amsterdam: John Benjamins. Kamakura, Y. 2007. Prepositions in a Corpus of Learner English. MPhil thesis, University of Birmingham. Leech, G., Rayson, P. & Wilson, A. 2001. Word Frequencies in Written and Spoken English: Based on the British National Corpus. London: Longman. Matthiessen, C. 2006. Frequency profiles of some basic grammatical systems: An interim report. In System and Corpus: Exploring Connections, G. Thompson & S. Hunston (eds), 103–142. London: Equinox. Nakamura, M. 2003. A Corpus Study of Four Frequent Nouns in English. MA thesis, University of Birmingham. Seidlhofer, B. 2005. Engaging the language learner in corpus-based learning. Paper read at the 37th Annual Meeting of the British Association for Applied Linguistics, King’s College London. Sinclair, J. M. 1991. Corpus Concordance Collocation. Oxford: OUP. Sinclair, J. M. 1999a. A way with common words. In Out of Corpora: Studies in Honour of Stig Johansson, H. Hasselgård & S. Oksefjell (eds), 157–180. Amsterdam: Rodopi. Sinclair, J. M. 1999b. The computer, the corpus and the theory of language. In Transiti Linguistici e Culturati, Vol. 2, G. Azzaro & M. Ulrych (eds), 1–15. Trieste: EUT. Sinclair, J. M. 2004. Trust the Text: Language, Corpus and Discourse. London: Routledge. Tognini-Bonelli, E. 2001. Corpus Linguistics at Work. Amsterdam: John Benjamins.

part iv

New types of corpora

Income/interest/net Using internal criteria to determine the aboutness of a text Winnie Cheng

This paper describes the application of a corpus-driven approach to the learning and teaching of phraseology in English for Specific Purposes (ESP). It describes two specialised corpora which have been compiled, using texts from the fields of economics and financial services, with the aim of providing resources for the study of the patterns and uses of English language specific to these fields. This paper demonstrates an innovative method to provide an initial indication of the aboutness of two texts, each drawn from a different specialised corpus. First, the search engine ConcGram© is used to identify the phraseological patterns specific to each text. The phraseological patterns found in each of the texts are then compared with those of each of the specialised corpora, and finally compared with those of a reference corpus of general English usage. This iterative process provides an indication of the aboutness of the texts. The process itself and the findings have important implications for ESP, especially if the students are the ones who generate the results.

1.

Introduction

Since the beginning of corpus linguistics proper in the late 1950s and early 1960s, the study of corpora has revolutionised the study of language (Tognini-Bonelli 2001; Hunston 2002). Applications of corpus investigation are found in an array of areas such as lexicography, language teaching, translation, stylistics, grammar, gender studies, forensic linguistics, and computational linguistics (TogniniBonelli 2001). Wilson, Archer and Rayson (2006) note that corpus-linguistic research has developed from the more traditional areas, including vocabulary, spoken language, synchronic and diachronic variation, Languages for Specific Purposes, tagging, and corpus development, to exciting new areas, such as crosscultural rhetoric and social psychology. In language teaching, despite the fact that

158 Winnie Cheng

corpora and corpus evidence have been used for the last two or three decades (see, for example, Sinclair 1987, 1991, 2004a, 2004b; Johns 1986, 1991), as observed by Römer (2006: 121), corpora have yet to become part of the “pedagogical landscape”. Teubert (2005: 203) outlines a type of diachronic corpus of “texts that relate to each other in a pre-defined way”, which differs from “the traditional corpus as a random selection of texts representing a certain sub-discourse”. In LSP teaching, learning and research, the use of specialised corpora, as described by Teubert (2005), for linguistic evidence, input and insights has been gaining importance (Hewings 2005). Bowker and Pearson (2002), for example, discuss how small specialized corpora that contain texts of a particular genre can be extremely useful for language teachers and learners and translators, including identifying specialized terms and detecting collocations in the specialised target language for glossary compilation, term extraction, and writing. Sinclair (2005b) remarks that compared to a general language corpus, a specialised corpus will have a greater concentration of vocabulary. He outlines the possibility of an automated process to compare the lexical profiles or aboutness of texts (Sinclair 2005a). Building on Sinclair (2005a), Cheng (2006) examines a small specialised spoken public corpus collected in Hong Kong during and in the immediate aftermath of the SARS crisis in 2003. She first determines the overlapping patterns of co-selection of the most frequently occurring lexical words in the specialised SARS corpus, and then describes the cumulative effects of the habitual co-selection in the lexical items that contribute to both textual meanings and intertextual coherence within and across the texts. In LSP research, few cross-cultural studies have investigated business and financial services English in corpora. Two recent studies investigate the use of metaphors in corpora of newspaper financial reports, comparing English and Spanish newspapers (Charteris-Black and Ennis 2001), and British and German newspapers (Charteris-Black and Musloff 2003). The study reported in this paper makes a case for using the evidence from two specialised economics and financial services corpora, developed at the University of Siena, Italy (2005–2007) and the Research Centre for Professional Communication in English, The Hong Kong Polytechnic University (2006–2007), in an attempt to describe aboutness (Phillips 1989; Sinclair 2005a, 2006), i.e. what a text is about, through a process of describing the phraseological patterns of a text from each corpus, comparing the text-specific patterns with those of the two specialised corpora, and finally comparing the text-specific patterns with the pattern of a reference corpus. Keywords

. Severe acute respiratory syndrome (SARS).

Using internal criteria to determine the aboutness of a text 159

and phraseology are not usually studied together. It is rather common to compile glossaries of key words, but the wider phraseological and textual context in which key words appear is often not accounted for. The key words supplied by authors of books and academic papers are used to access potentially relevant topics or content in texts, but the reliability of the key words in yielding the desired results is often found wanting. The importance of English language as an international lingua franca in business and professional communication is beyond dispute. In order to become successful intercultural communicators in English, it is important for professionals and professionals-in-training to have linguistic competence, among other competencies, namely sociolinguistic, discourse, strategic, socio-cultural, and social competence (Byram 1997; van Ek 1986). The premise of this study is that a high level of linguistic competence can be developed through identifying and understanding the phraseology used by writers in specific contexts. ‘Phraseology’ has been defined as “the recurrent co-occurrence of words” (Clear 1993: 277), “the more-or-less fixed co-occurrence of linguistic elements” (Hunston 1995), and “the study of the structure, meaning, and use of word-combinations” (Cowie 2004: 673). This study aims to describe and demonstrate a corpus-driven, innovative method to describe the phraseological patterns of text and corpora, which would be indicative of the aboutness of text and corpora (Sinclair 2005a). It argues that the method itself and the findings have important implications for ESP and LSP, especially if the students are the ones generating the results.

2.

The notion of aboutness

The key words of a text are often viewed as informative of the aboutness of the text (see, for example, Scott 1997, 2000; Tribble 1997; Scott and Tribble 2006). Scott and Tribble (2006), in their studies of the general BNC and corpora differentiated by domain, aim to find out whether key key words (KKWs) reflect aboutness. They conclude that “Aboutness is much more clearly attained as the genre becomes domain-specific” (Scott and Tribble 2006: 82). The focus of these studies on aboutness has been on words, which happen to be key “in a particular text” or “in lots of texts, where ‘lots’ is defined (subjectively) by the number of texts in the database” (Scott and Tribble 2006: 78). As Sinclair (2004b: 148) observes, “the word is not the best starting-point for a description of meaning, because meaning arises from words in particular combinations”. He argues that “a word on its own is usually not distinctive enough to deliver a stable and precise meaning (outside the protected words which are recognised as technical terms – and even they are always at risk)” (Sinclair 2005a).

160 Winnie Cheng

The over-reliance on keywords means that most of the important information regarding the content of individual texts has not been utilised. Sinclair (2005a) also observes that even when word clusters, rather than single words, are entered, the search does not treat the words as a linguistic structure, but rather as a set of co-ordinates, within which might lie the texts required, if the user is successful. Sinclair (2005a) states that an accurate description of a text’s content can be achieved by examining “intercollocability”. Central to this notion are the patterns of co-occurrence of words (collocation) found within texts, which are central to meaning-making and contribute, by means of particular phrasings, to genrespecific usage (see, for example, Sinclair 1991; Scott 2001; Williams 1998, 2002; Tognini-Bonelli 2002). Intercollocability is determined by generating, in a fully automated form, a text’s collocational profile. This is done by first establishing keywords and the collocates of the keywords, and then the collocates of collocates to determine groups of words, which in turn determine what Sinclair (2005a) terms the aboutness of the text. Once the criteria for aboutness have been established through collocational profiling, texts can be compared and their “aboutness distance” (Sinclair 2005a) can be calculated, and hence their membership of the same genre determined. Importantly, as Sinclair (2005a) maintains, establishing these collocational patterns for a particular text provides a set of criteria that is internal to the text for the description of its phraseology, and hence the aboutness of the text can be objectively established. Nevertheless, up until now, models for topic identification in texts have assumed that all texts can be labelled based on discrete external criteria, such as demographic and contextual criteria, which are subjectively arrived at and so unreliable (Sinclair 2005a). By ignoring criteria internal to texts, these models are considered to have major limitations.

3.

Texts from specialised corpora

This study posits that an accurate description of the aboutness of economics and financial services texts is important for professionals and professionals-in-training as well as ESP learners and teachers in the domain of economics and financial services English. The study compares the aboutness of two individual texts, respectively taken from the Corpus of Academic Economics Texts (CÆT) (TogniniBonelli 2006) and the Hong Kong Financial Services Corpus (HKFSC) (Cheng 2007). Tables 1 and 2 summarise the contents of the two specialised corpora. A text was chosen from each of the two specialised corpora: an article from an economics journal (14,610 words) from the CÆT, and an annual report (30,581 words) of a major bank in Hong Kong from the HKFSC.


Table 1. Composition of the CÆT, Siena, Italy (22 October 2006) Text-type

Number of words

Economist Survey European Central Bank Bulletin Textbook Journal Article European Union Report Total number of words

913,978 428,796 295,363 312,904 81,457 2,032,498

Table 2. Composition of the Hong Kong Financial Services Corpus (31 January 2007) Text-type

Number of words

Annual Report Prospectus Speech Standards Code of Practice Rules Interim Report Total number of words

639,182 1,991,140 843,963 102,533 69,981 8,369 17,015 3,830,000

4.

From n-gram to skipgram to concgram

Since the 1960s, researchers in the fields of natural language processing, computational linguistics and corpus linguistics have described patterns of phraseology in text and corpora (Sinclair, Jones and Daley 1970). ‘N-grams’, sometimes termed ‘word clusters’, ‘lexical clusters’ or ‘bundles’ (see, for example, Scott 2004; Biber et al. 1999; Carter and McCarthy 2006), e.g. a lot of, one of the, you know, and I don’t think, are contiguous words that constitute a phrase, or a pattern of use. This means that many instances of “word-combinations” (Cowie 2004: 673) or word co-occurrrences or multi-word units that mostly occur in non-contiguous sequences and with positional variation (i.e. AB, BA) (Cheng, Greaves and Warren 2006) may not be found. Skipgrams or ‘phrase-frames’ (Fletcher 2006) such as the past three years, the past few years uncover patterns of phraseology with non-contiguous sequences, but not those with positional variation. The limitations of the existing search engines that generate n-grams and skipgrams have led to the design of a search

162 Winnie Cheng

engine, ConcGram© (for details, see Cheng, Greaves and Warren 2006; Greaves 2006), which, on top of its capability to handle both constituency variation (i.e. AB, A*B) and positional variation (i.e. AB, BA), conducts fully automated searches and searches for word-combinations comprising up to five words. The products of the searches are termed ‘concgrams’. A concgram is all of the permutations of constituency variation and positional variation generated by the co-occurrence of two or more words. Below is a sample of the concordance lines of a three-word concgram ‘allocated/economic/capital’ from a journal article from the CÆT: Example 1: allocated/economic/capital 1 2 3 4 5 6 7 8 9 10 11

cost of allocated economic capital. The cost of economic capital is defined as the activity’s allocated more) of large banks’ total risk and allocated economic capital. 31 Thus, assessing capital for credit and Relationship between PDF and allocated economic capital. Note: The shaded area under the PDF to credit losses), less the cost of allocated economic capital. The cost of economic capital is defined that unexpected losses will exceed the allocated economic capital. Exhibit 2. Overview of risk measurement capital is defined as the activity’s allocated economic capital times the bank’s ROE target or hurdle which economic capital is allocated. The total economic capital allocation for the bank as a whole is where the amount of supporting equity is the economic capital allocated to that activity. Specifically, framework, at many large banks the amount of economic capital allocated against operating risk is of risk that are measured, and against which economic capital is allocated. Besides credit and market are made for each type of risk, against which economic capital is allocated. The total economic capital

An examination of the eleven concordance lines illustrates some of the kinds of variation to be found in many concgrams. First, line 1 contains two instances of the concgram ‘allocated/economic/capital’. The first instance is repeated in line 4, and the second instance (The cost of economic capital is defined as the activity’s allocated) does not constitute a linguistic unit in text, called “chunks” in Sinclair and Mauranen’s (2006: 134) Linear Unit Grammar, and so is excluded and not analysed. Lines 2–6 contain the positional variation ‘allocated economic capital’; and lines 7–11 contain the positional variation ‘economic/capital/allocated’ (note that the first of the two instances of this positional variation in line 7 is repeated in line 11). There is constituency variation in lines 10 and 11, with one intervening word ‘is’. Altogether three concgram configurations are found: ‘allocated economic capital’ (5 instances), ‘economic capital allocated’ (2 instances) and ‘economic capital * allocated’ (2 instances). ESP researchers, teachers and learners can study the concgram configurations in the co-text to work out whether they constitute meaningful phrasings, and then go on to identify the phraseological patterns, and finally the aboutness, specific to the text or corpus. How this is done is described below. . ConcGram© was developed by Chris Greaves, Senior Project Fellow, English Department, The Hong Kong Polytechnic University.

5.


Methodology for using concgrams to determine aboutness

Initial findings in Cheng, Greaves and Warren (2006) suggest that concgrams are a means to uncover aboutness. The process adopted in this study involves searching for the concgrams in a text taken from each of the two specialised corpora. When each text was concgrammed automatically to generate concgram lists, an exclusion list based on Ahmad’s (2005) list of the fifty most frequently occurring words in the British National Corpus, all of which are function words, was used to focus the searches on the co-occurrences of lexical words. A cautionary note is needed here. The idea that aboutness is primarily located in the co-occurrences of lexical words will need to be monitored closely in subsequent studies. The role of function words in the co-selection of lexical items (Sinclair 1996) is currently underestimated, and certainly not fully understood. However, research work on collocational frameworks (Sinclair and Renouf 1988; Renouf and Sinclair 1991) and the phraseological patterns and characteristics of language (Biber, Conrad and Reppen 1998; Biber et al. 1999; Cowie 1998; Hunston and Francis 2000; Simpson and Swales 2001; Bartsch 2004; Nesselhauf 2005; Sinclair 1987, 1996, 2004a, 2004b) suggests that the use of an exclusion list is unwise, if we are to fully describe the phraseological patterns, and hence aboutness, of texts and corpora. Lists of three-word concgrams were first generated (Tables 3 and 4), followed by listing out a rank-ordered series of variable expressions which characterise the phraseological patterns found in the text. The iterative processes of then determining whether or not each list of concgrams captures the aboutness of each text requires that a search for each concgram is conducted in the specialised corpus from which it is taken (i.e. CÆT and HKFSC), and then searched for again in a general corpus. The general corpus used in this study is a five-million-word crosssection (three million written and two million spoken) of the British National Corpus (henceforth BNC (5 m)). Those concgrams which are not matched in terms of relative frequencies in CÆT and HKFSC or BNC (5 m) are downgraded in importance, or dropped, and those which remain can be said to represent the language of the particular text based on its distinctiveness. This methodology, once it is further refined, should help to uncover the “aboutgrams” (Sinclair personal communication 2006) of not only the text under investigation, but also those of CÆT and HKFSC.

6.

Findings and discussion

This section discusses the findings resulting from the iterative process of identifying the aboutness of texts and corpora, and then the phraseological patterns of

164 Winnie Cheng

Table 3. Ten most frequent three-word lexical concgrams in Journal Article (CÆT) Three-word concgrams

Journal Article

CÆT

BNC (5m)

credit/models/risk internal/models/risk credit/internal/risk credit/risk/spreads allocation/capital/economic correlations/factors/risk credit/market/risks credit/internal/models credit/modelling/risk allocated/economic/capital

46 20 16 15 14 14 13 13 12 9

46 20 16 15 14 14 13 13 12 9

0 0 0 0 0 0 0 0 0 0

prevalent concgrams of the two corpora. Tables 3 and 4 show the top ten threeword lexical concgrams in each text, based on frequencies, and the frequencies of each concgram in CÆT and HKFSC respectively and in BNC (5m). Since the lists of concgrams cannot be automatically classified into those which contain meaningful word co-occurrences and those which do not, human intervention is required to examine the concgrams in their concordance lines. This means that the raw frequencies may need to be adjusted in cases of repeats in the same concordance line (as discussed in example 1). It is then necessary to remove instances where the words in a concgram do not form a linguistic unit in text, or do not conform to the canonical form of the concgram, with ‘canonical form’ being defined as the “base form with the prototypical meaning” (Cheng and Warren 2006). Table 3 lists the ten most frequent three-word concgrams of the Journal Article. None of the ten concgrams are found in the rest of CÆT, and none are found in BNC (5 m). This study, therefore, argues that these concgrams embody the aboutness of the Journal Article. Now that the criteria for aboutness have been established, in future studies this Journal Article can be compared with other journal articles, and their “aboutness distance” (Sinclair 2005a) can be calculated and hence their membership of the same genre determined. Another finding is that the concgrams in the list seem to provide evidence of the process by which the software builds up a profile of the intercollocation of collocates (Sinclair 2005a). After establishing the unique words (i.e. ‘types’) in the text, the search engine ConcGram© then searches for words that co-occur with each one, and lists all of the two-word concgrams found. The two-word concgrams then become the origin for the next search which finds all the words that co-occur with the two-word concgrams. This iterative process, by which concgrams are built up automatically by the search engine, is the same process as that described by Sinclair (2005a) for determining the aboutness of documents. Sinclair (2005a)


Table 4. Ten most frequent three-word lexical concgrams in Annual Report (HKFSC) Concgrams

Annual Report

HKFSC

BNC (5m)

income/interest/net designated/fair/value income/recognised/statement assets/financial/liabilities instruments/fair/value designated/financial/value financial/instruments/value designated/fair/instruments designated/value/instruments

37 36 36 28 26 24 23 19 19

164 102 176 161 131 67 89 44 43

0 0 0 0 0 0 0 0 0

states that the “process (called the intercollocation of collocates) also disambiguates words, and a group gives a strong sense of the content, scope and argument of the document”. For example, the three most frequent two-word concgrams in the Journal Article are ‘credit/risk’ (130 instances), ‘models/risk’ (61 instances) and ‘credit/models’ (50 instances). The most frequent three-word concgram, ‘credit/ models/risk’, can be seen to be the product of the intercollocation of the collocates of these three most frequent two-word concgrams in the text. Table 4 shows the ten most frequent three-word lexical concgrams in the Annual Report 2005 in HKFSC. An examination of Table 4 shows that the aboutness of the Annual Report is not so quickly arrived at, because of additional instances of all of the concgrams in the rest of HKFSC. The reason for this can be explained by the nature of the two texts that were chosen. The academic Journal Article examined earlier (Table 3) is likely to be a more unique text with regard to its aboutness, compared to the rest of CÆT, whereas Annual Report 2005 is one of the ten annual reports contained in HKFSC, and so it is to be expected that there will be patterns of language use that occur in the other annual reports. To see if this is the case, a sub-corpus of all of the ten annual reports in HKFSC was created and searched (Table 5). There are two main findings. First, as all the ten three-word concgrams listed are also found in the other annual reports, identifying the aboutness of one constituent text in the Annual Report sub-corpus can be a good indicator of the aboutness of the entire sub-corpus. Second, five out of ten three-word lexical concgrams, highlighted in Table 5, namely ‘designated/fair/value’, ‘designated/financial/value’, ‘designated/fair/instruments’, ‘designated/value/instruments’ and ‘income/net/trading’, are almost exclusively confined to the annual reports. These five concgrams are, therefore, much more useful for describing the distinctive

166 Winnie Cheng

Table 5. Ten most frequent three-word lexical concgrams in Annual Report (HKFSC) Concgrams

Annual Report

All Annual Reports3

HKFSC

BNC (5m)

income/interest/net designated/fair/value income/recognised/statement assets/financial/liabilities instruments/fair/value designated/financial/value financial/instruments/value designated/fair/instruments designated/value/instruments income/net/trading

37 36 36 28 26 24 23 19 19 19

81 95 102 71 90 60 71 41 41 67

164 102 176 161 131 67 89 44 43 73

0 0 0 0 0 0 0 0 0 0

aboutness of the annual reports as a genre. These findings suggest that there are a number of phrases which are commonly used in annual reports, and which could be used to identify other members of this genre with a high degree of probability in an automated fashion. For the remaining five concgrams, namely ‘income/interest/net’, ‘income/ recognised/statement’, ‘assets/financial/liabilities’, ‘instruments/fair/value’ and ‘financial/instruments/value’, while they are more likely to occur in an annual report, compared to other text types in HKFSC, they are also present in quite large numbers in other financial services texts (see Table 2). Given that none of the concgrams are found in BNC (5m), this group of concgrams is indicative of financial services language use, rather than general language use. Similar to the Journal Article findings, the concgrams in Tables 4 and 5 also demonstrate the notion of intercollocability. For example, one of the three most frequent concgrams, ‘designated/fair/value’, is the product of the intercollocation of the collocates of three of the most frequent two-word concgrams in the text: ‘fair/value’ (173 instances), ‘designated/fair’ (41 instances) and ‘designated/value’ (40 instances). This process can also be seen at work when looking at larger word co-occurrences. For example, in this Annual Report text, the five-word concgram ‘designated/fair/financial/instruments/value’ has ‘financial instruments designated at fair value’ as the canonical form. Below, the concordance lines of the three most frequent three-word concgrams in each text are examined to show how the phraseological patterns are identified. Examples 2–4 come from the Journal Article in CÆT, and Examples 5–7 . There are ten annual reports of major Hong Kong listed companies in the HKFSC, totalling 395,589 words, i.e. 11.63% of the HKFSC.


the Annual Report 2005 in HKFSC. The discussion focuses on valid instances, positional and constituency variation, concgram configurations, the canonical form of a concgram, and meaningful word associations. In Example 2, six of the 52 instances of the concgram ‘credit/models/risk’ (lines 1, 3, 8–10 and 52) found are not included in the total. This is because the three lexical words do not constitute a linguistic unit in the text. This particular concgram ‘credit/models/risk’ is typically contiguous (44 instances), and the configuration in these cases is ‘credit risk models’. The two exceptions to this are found in line 7 ‘credit and other risk models’ and line 51 ‘risk-based pricing models for credit products’, both of which share a meaning similar to the most frequent form. They are represented as ‘credit ** risk models’ and ‘risk-**models*credit*’ respectively. Example 2: credit/models/risk (Journal Article, CÆT) 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32

models, in principle, might be used wherever credit risk is incurred. In regulatory practice, VAR models, appropriate stress testing of credit risk models for the banking book might partially using these models. Similarly, when estimating credit risk, most practitioners assume that all the potential use of models and, in particular, credit risk models within the banking book, which to trading account VAR models, internal credit risk models are used in estimating the economic risk models. In lieu of formal backtesting, credit risk models tend to be validated indirectly including risk-based pricing models for credit products, the setting of portfolio concentration Within most credit risk models, each customer’s credit rating at the end of the planning horizon is risk is comparable to that allocated against credit risk. Models for estimating operating risk, procedure.20 Risk Factors Affecting Changes in Credit Ratings. Within most credit risk models, each among the risk factors. In particular, credit risk models nearly always assume zero of a particular bank. Although banks’ internal credit risk models could, in theory, be incorporated financial statements and Call Reports). Top-down credit risk models generally are vulnerable to the same a context for later discussions of internal credit risk models, Section II presents a general a special case. Within both DM- and MTM-type credit risk models, the model-builder is required to effects are considered explicitly. IV. Credit Risk Models: Building Blocks This section reviews the core analytical components of credit risk models. For each component, we first discuss Affecting Changes in Credit Ratings. Within most credit risk models, each customer’s credit rating at the capital policies. Specifically, internal credit risk models may be useful in: 1) the setting of Given Default. Within the current generation of credit risk models, LGDs are usually assumed to be Soundness Within the current generation of credit risk models there are a number of important To illustrate this process, consider that credit risk models often invoke simplifying assumptions VI. Possible Near-Term Applications of Credit Risk Models Although the reliability concerns We argue that the current generation of internal credit risk models raises important concerns in each of of the main building blocks of (bottom-up) credit risk models used within banks’ large corporate degree of subjectivity in the specification of credit risk models, the need for effective model formal backtesting procedures for their credit risk models. In lieu of formal backtesting, tail of a credit portfolio’s PDF (the focus of credit risk models) is likely to be highly sensitive to used extensively to gauge the reasonableness of credit risk models and internal capital allocation and integrity of the current generation of credit risk models at large banks. To anticipate this in backtesting. Most of the uncertainty within credit risk models (and the infeasibility of stress tests might be developed for internal credit risk models used within the banking book, we

168 Winnie Cheng

33 any efforts in this direction. Integration of Credit Risk Models into Decision-Making Processes The 34 model. In practice, the extent of reliance on credit risk models differs greatly among banks. Much of 35 caution when contemplating the use of internal credit risk models for setting regulatory capital 36 management. To give one example, some banks use credit risk models to estimate an efficient portfolio 37 in connection with the current generation of credit risk models are substantial, they are not 38 there are notable instances where the output of credit risk models was not considered in situations which the inputs or outputs of banks’ internal credit risk models might usefully be incorporated into 39 40 These include: 1) the selective use of internal credit risk models in setting formal RBC requirements 41 internal credit ratings and other components of credit risk models for purposes of developing specific 42 might involve the direct use of internal credit risk models in setting formal RBC requirements 43 securitized assets. The application of internal credit risk models, if deemed reliable by supervisors, 44 of examination guidance on the use of internal credit risk models would provide useful insights into 45 in terms of how they might eventually use credit risk models for regulatory purposes, such an 46 examples by which information from internal credit risk models might be usefully incorporated into 47 an accounting distinction. Large banks’ credit risk models, in principle, might be used wherever 48 pricing on individual loans, banks also use credit risk models in active portfolio management. To 49 these requirements below because, with regard to credit risk models, formal compliance with these 50 situations where it might have been. In general, credit risk models are not used in determining loan loss 51 necessarily insurmountable. It may well be that credit and other risk models are evolving so rapidly 52 noted that estimation of the extreme tail of a credit portfolio’s PDF (the focus of credit risk models)

Example 3 shows all of the concordance lines (22) for ‘internal/models/risk’ in the Journal Article, CÆT. The instances in lines 2 and 3 are excluded because the former is repeated in line 9, and the latter does not constitute a linguistic unit in the text. Two configurations are found, namely ‘internal risk models’ (7 instances) and ‘internal * risk models’ (13 instances). When there is constituency variation, the intervening word is always ‘credit’. Example 3: internal/models/risk (Journal Article, CÆT) 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18

Analogous to trading account VAR models, internal credit risk models are used in estimating the from banks’ internal risk models, these internal grades could provide a basis for developing the reasonableness of credit risk models and internal capital allocation processes. Another near-term regulatory and supervisory uses of internal risk models. II. Overview of Economic Capital possible near-term roles for incorporating internal risk models into prudential capital policies. of course, are predicated on the reliability of internal risk models. This paper evaluates the near-term ignored altogether. The selective application of internal risk models in this area could fill an important to encompass additional elements of banks’ internal risk models, including analytical tools based on combination with other information from banks’ internal risk models, these internal grades could provide problem is much more acute in the context of internal risk models for the banking book. At most large adequacy of a particular bank. Although banks’ internal credit risk models could, in theory, be We argue that the current generation of internal credit risk models raises important concerns in areas in which the inputs or outputs of banks’ internal credit risk models might usefully be incorporated policies. These include: 1) the selective use of internal credit risk models in setting formal RBC comparable stress tests might be developed for internal credit risk models used within the banking book, 1note of caution when contemplating the use of internal credit risk models for setting regulatory capital pools of securitized assets. The application of internal credit risk models, if deemed reliable by into prudential capital policies. Specifically, internal credit risk models may be useful in: 1) the

19 20 21 22


To provide a context for later discussions of internal credit risk models, Section II presents a general proposal might involve the direct use of internal credit risk models in setting formal RBC testing of examination guidance on the use of internal credit risk models would provide useful insights provides examples by which information from internal credit risk models might be usefully incorporated

In Example 4, the instances in lines 1–2 and 19–22 are excluded because they do not constitute a linguistic unit in the text. All of the other instances are the invariant contiguous canonical form ‘internal credit risk’. Example 4: credit/internal/risk (Journal Article, CÆT) 1 internal credit ratings and other components of credit risk models for purposes of developing specific 2 many banks use internal capital allocations for credit risk within a variety of decision-making 3 These include: 1) the selective use of internal credit risk models in setting formal RBC requirements 4 to trading account VAR models, internal credit risk models are used in estimating the economic 5 capital policies. Specifically, internal credit risk models may be useful in: 1) the setting of 6 a context for later discussions of internal credit risk models, Section II presents a general of examination guidance on the use of internal credit risk models would provide useful insights into 7 8 described in Section V, incorporating internal credit risk measurement and capital allocation systems 9 could objectively validate a bank’s internal credit risk model. Although similar concerns were raised 10 stress tests might be developed for internal credit risk models used within the banking book, we are 11 We argue that the current generation of internal credit risk models raises important concerns in each of caution when contemplating the use of internal credit risk models for setting regulatory capital 12 which the inputs or outputs of banks’ internal credit risk models might usefully be incorporated into 13 securitized assets. The application of internal credit risk models, if deemed reliable by supervisors, 14 15 of a particular bank. Although banks’ internal credit risk models could, in theory, be incorporated 16 might involve the direct use of internal credit risk models in setting formal RBC requirements 17 Task Force which has been reviewing the internal credit risk modeling and capital allocation processes of 18 examples by which information from internal credit risk models might be usefully incorporated into 19 Basle Accord; and 2) the use of internal credit ratings and other components of credit risk 20 risk measurement purposes, a customer’s internal credit rating is generally used as a summary statistic 21 used extensively to gauge the reasonableness of credit risk models and internal capital allocation 22 Credit Risk Modeling and Internal Capital Allocation

Examples 5–7 show the most frequent concgrams in the Annual Report 2005, HKFSC. In Example 5, the instances in lines 1–5 and 7–11 are repeated elsewhere, and are therefore excluded. The instances in lines 46–50, 52 and lines 54–56 are also excluded because they do not constitute a linguistic in the text. The remaining 38 instances are all the invariant and contiguous canonical form ‘net interest income’. Example 5: income/interest/net (Annual Report, HKFSC) 1 2 3 4 5 6 7

Net interest income in 2005 also includes income of HK$989 million from held-to-maturity 2005 Net trading income comprises interest income and expense and dividend income attributable to from ‘Net interest income’ and ‘Dividend income’, to ‘Net trading income’. This added HK$817 spreads. Net interest income also includes income of HK$694 million from held-tomaturity in ‘Net interest income’ and ‘Dividend income’ respectively. Financial instruments designated groups. Net Interest Income Net interest income of HK$43,491 million was HK$6,521 million, or recognised in ‘Net interest income’ or ‘Dividend income’ as appropriate. Advances to customers and

170 Winnie Cheng

8 Included in net interest income in 2005 is income earned on held-to-maturity investments in the 9 is included in ‘Net interest income’ and ‘Net income from financial instruments designated at fair 10 million from ‘Net interest income’ to ‘Net income from financial instruments at fair value’. 11 HK$505 million of net interest income to ‘Net income from financial instruments designated at fair 12 well as their reclassification to ‘Net interest income’. Derivatives and hedging (Note 3j) HKAS 39 13 of HK$505 million of net interest income to ‘Net income from financial instruments 14 of HK$199 million from ‘Net interest income’ to ‘Net income from financial instruments at 15 deposit spreads in Hong Kong. Net interest income from Corporate, Investment Banking and Markets 16 2004, as a result of a decline in net interest income in Global Markets which more than offset a to a widening of deposit spreads. Net interest income in 2005 also includes income of HK$989 million 17 18 trading assets and liabilities and net interest income from financial instruments designated at fair 19 Hong Kong, Taiwan and Indonesia. Net interest income from the Commercial Banking business was 20 This added HK$817 million to net interest income but was more than offset by the maturity of 21 and improved deposit spreads. Net interest income also includes income of HK$694 million from 22 includes a full 12 months’ Net interest income was lower, reflecting an share of profits from 23 In the rest of Asia-Pacific, net interest income increased, notably in Singapore, mainland China 24 was coupled with strong growth in net interest income in the rest of Asia-Pacific, driven by increased 25 to other customer groups. Net Interest Income Net interest income of HK$43,491 million was 26 assests and liabilities from ‘Net interest income’ and ‘Dividend income’, to ‘Net trading income’. 27 a strong trading performance. Net interest income fell by HK$2,533 million, or 22.5 per cent, 28 compared with 2004. In Hong Kong, net interest income improved by HK$4,656 million, or 29.6 per cent. 29 movements which could reduce future net interest income while balancing the cost of such hedging 30 the residual period to maturity in net interest income. Where the adjustment relates to the carrying throughout the region. Included in net interest income in 2005 is income earned on held-to-maturity 31 basis points, and the inclusion of net interest income on held-to-maturity investments in the insurance 32 of Asia-Pacific respectively. Net interest income increased by HK$4,024 million, or 57.0 per cent, 33 34 contributed to an increase in net interest income of 41.4 per cent. The payments and cash 35 and dividends were recognised in ‘Net interest income’ and ‘Dividend income’ respectively. Financial 36 per cent, higher than in 2004. Net interest income from the Personal Financial Services business 37 earnings. This has reduced ‘Net interest income’ and ‘Operating profit’ in 2005 by HK$3,010 38 the sensitivity of projected net interest income under varying interest rate scenarios. The 39 In the rest of Asia-Pacific, net interest income rose by HK$1,362 million, or 26.8 per cent, 40 related assets is included in ‘Net interest income’ and ‘Net income from financial instruments 41 and improved deposit spreads. Net interest income increased by HK$6,016 million, or 28.9 per cent, 42 portfolios is to optimise net interest income. Market risk in non-trading portfolios arises 43 qualifying hedges is included in ‘Net interest income’. Fair value hedge Changes in the fair value 44 income were recognised in ‘Net interest income’ or ‘Dividend income’ as appropriate. Advances 45 in 2005, including accrual book net interest income and funding related to dealing positions, was income’ and ‘Dividend income’, to ‘Net trading income’. This added HK$817 million to net interest 46 47 income From 1 January 2005 Net trading income comprises interest income and expense and 48 in the income statement within ‘Net trading income’ along with accrued interest. Amounts 49 in the income statement as ‘Net trading income’ as they arose. Related interest income and 50 5statement captions according to their nature. Income from related assets is included in ‘Net interest 51 interest rate and amortised through net interest income, rather than through net fee income as in prior 52 interest income, rather than through net fee income as in prior years. Net Trading Income Net 53 interest was classified under ‘Net interest income’ in prior years. Gains Less Losses from 54 Interest income and expense, and dividend income were recognised in ‘Net interest income’ or 55 the reclassification of interest and dividend income on trading assets and liabilities from ‘Net 56 as they arise, together with related interest income and expense and dividends, within ‘Net income


In Example 6, in 32 out of 36 instances, the non-contiguous ‘designated * fair value’ configuration is found with ‘at’ as the intervening word. In the last four lines, other words intervene between ‘designated’ and ‘fair value’ but they share a meaning similar to that of the canonical form ‘designated * fair value’ (‘designated under the ‘fair value option”, line 33; ‘designated and qualify as fair value hedging instruments’, line 34; ‘designated as hedging instruments in a fair value’, line 35; and ‘designated are recognised initially at fair value’, line 36). The constituency variations are represented as ‘designated ** fair value *’ (line 33), ‘designated *** fair value **’ (line 34), ‘designated ***** fair value’ (line 35), and ‘designated **** fair value’ (line 36). Example 6: designated/fair/value (Annual Report, HKFSC) 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33

at fair value’. Where issued debt has been designated at fair value and there is a related derivative, and advances classified as held for trading or designated at fair value are reported as trading of certain assets to ‘Financial assets designated at fair value’. Holdings of debt securities assets and liabilities, including those designated at fair value, are recognised and measured on to ‘Net income from financial instruments designated at fair value’. Both of these reclassifications under certain restricted circumstances, to be designated at fair value with changes in fair value being and ‘Net income from financial instruments designated at fair value’. The related assets are included in the balance sheet under ‘Financial assets designated at fair value’ and ‘Financial investments’. The 2005 Net income from financial instruments designated at fair value comprises all gains and losses from and financial assets and liabilities designated at fair value, the estimation of fair value at except those classified as held for trading or designated at fair value are recognised in 'Interest income' Net income from financial instruments designated at fair value From 1 January 2005 Net income income’ respectively. Financial instruments designated at fair value From 1 January 2005 A financial have not been classified as held for trading or designated at fair value. Loans and advances are recognised within ‘Net income from financial instruments designated at fair value’. Gains and losses arising from trading instruments, or financial instruments designated at fair value, respectively (Note 3e and 3f). in ‘Net income from financial instruments designated at fair value’. Where issued debt has been are included in ‘Financial liabilities designated at fair value’. Previously all income and 2004 The category, Financial instruments designated at fair value’ was introduced on 1 January 2005 with financial assets or financial liabilities designated at fair value are included in 'Net income from securities and financial instruments designated at fair value that are quoted in active markets, derivative, and the combined contract is not designated at fair value through profit and loss. These in conjunction with financial instruments designated at fair value. These activities are described securities unless they have been designated at fair value (see Note 3f) or they are contracts are used with financial instruments designated at fair value, in which case gains and losses in ‘Net income from financial instruments designated at fair value’. From 1 January 2004 to 31 in ‘Net income from financial instruments designated at fair value’. Premiums receivable and amounts Debt securities issued for trading purposes or designated at fair value are reported under the appropriate in conjunction with financial instruments designated at fair value. Ineffective hedging derivatives contracts and the linked financial assets are designated at fair value, and the movements in fair value net interest income from financial instruments designated at fair value was 24 basis points, and the or ‘Net income from financial instruments designated at fair value’ unless they qualify as hedges for in the balance sheet. Unless the securities are designated under the ‘fair value option’ (see below), then

172 Winnie Cheng

34 35 36

derivatives (net of interest accrual) that are designated and qualify as fair value hedging instruments Hedge accounting is applied to derivatives designated as hedging instruments in a fair value, cash flow Financial assets and financial liabilities so designated are recognised initially at fair value, with

In Example 7 showing ‘income/recognised/statement’, lines 2, 6 and 40–41 are excluded because they are repeated elsewhere in the lines. Line 1 is also excluded because the words do not constitute a linguistic unit in text. In the remaining lines, the same positional variation is found, ‘recognised/income/statement’, but with different kinds of constituency variation, ‘recognised ** income statement’ with usually two intervening words, ‘in the’ (26 instances) or ‘through the’ (1 instance), and sometimes in combination with an adverb, ‘immediately’ (5 instances, lines 7–11) and ‘directly’ (1 instance, line 12), and therefore with the constituency variation ‘recognised *** income statement’. The remaining instances show greater deviation from the canonical form, but nonetheless share a similar meaning (‘recognised as a gain in ‘Other operating income’ in the income statement’, line 5; ‘recognised as rental income in the income statement’ line 14; and ‘recognised in ‘Interest income’ and ‘Interest expense’ in the income statement’, line 39). Example 7: income/recognised/statement (Annual Report, HKFSC) 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26

impairment losses which are charged to the income statement. Negative goodwill is recognised are recognised in the income statement in ‘Net income from financial instruments designated at fair recognised in equity are recognised through the income statement and classified as ‘Gains less losses recognised in reserves are recognised in the income statement. Provisions Provisions for is recognised as a gain in ‘Other operating income’ in the income statement. Debt securities or million) was recognised as rental income in the income statement in respect of operating leases. inputs, is not recognised immediately in the income statement but is recognised over the life of the goodwill is recognised immediately in the income statement as it arises. At the date of disposal portion is recognised immediately in the income statement within ‘Net trading income’ along with portion is recognised immediately in the income statement. Gains and losses accumulated in HKAS 39 are recognised immediately in the income statement and reported in ‘Net trading income’, properties to be recognised directly in the income statement as they arise. Prior to the at fair value are recognised in ‘Interest income’ and ‘Interest expense’ in the income statement 2004: HK$211 million) was recognised as rental income in the income statement in respect of operating fair value gain or loss is recognised in the income statement. r Pension and other post-retirement under these circumstances is recognised in the income statement in the period in which it occurs. on impaired loans was not recognised in the income statement, but was credited to an interest financial asset previously recognised in the income statement) is removed from equity and recognised transaction is ultimately recognised in the income statement. When a forecast transaction is no of these securities was recognised in the income statement as it arose and included in ‘Gains less held for trading are recognised in the income statement. Where derivatives are designated and 2004 Interest income was recognised in the income statement as it accrued, except in the case of on an appropriate basis, or recognised in the income statement when the inputs become observable, or equity instruments previously recognised in the income statement that are no longer required are with changes in fair value recognised in the income statement. Derivative assets and liabilities on any gain or loss arising was recognised in the income statement as ‘Net trading income’, after


27 movements in fair value are recognised in the income statement in ‘Net income from financial 28 after the impairment loss was recognised in the income statement, the impairment loss is reversed 29 assets and liabilities are recognised in the income statement as they arise, together with related 30 in a foreign operation are recognised in the income statement of the separate subsidiary financial 31 deferred tax. Income tax is recognised in the income statement except to the extent that it relates 32 to equity and is subsequently recognised in the income statement when the deferred fair value gain or changes in fair value being recognised in the income statement as they arise. This is described in 33 fees receivable are recognised in the income statement over the period of the provision of 34 35 (2004: HK$nil). Expenses recognised in the income statement in respect of defined benefit schemes 36 is removed from equity and recognised in the income statement. If, in a subsequent period, the fair changes in fair value being recognised in the income statement (in ‘Other operating income’) with 37 38 expense and dividends, are recognised in the income statement within ‘Net trading income’ as they 39 assets and liabilities were recognised in the income statement as ‘Net trading income’ as they arose. 1 January 2004 to 31 December 2004 Interest income was recognised in the income statement as it 40 41 for the year comprises current and deferred tax. Income tax is recognised in the income statement except

Analysis of the six examples of three-word concgrams taken from the Journal Article, CÆT and Annual Report, HKFSC has demonstrated how phraseological patterns are identified from concordances generated by ConcGram©. The analysis has also illustrated the canonical form of a concgram, as well as positional and constituency variations that exist in different configurations of a concgram. Most importantly, the concgram examples show how the phraseologies of specific texts, and specialised corpora, can be identified and retrieved.

6.

Conclusions

This paper has described the most frequently-occurring concgrams of texts taken from specialised ESP economics and financial services corpora, compiled for different groups of users in their respective domains. The paper argues that the canonical form of a concgram, as well as the non-canonical forms of a concgram which share a similar meaning with the canonical form, constitute phraseology. The study has important implications for ESP teaching, learning and research in raising language awareness and increasing knowledge about the aboutness associated with the topics in the discipline- and professional-specific discourses. Searching a specialised corpus, such as the economics and financial services corpora discussed in this study, not only gives the user pertinent information about the frequencies and patterns of key phraseologies, but also examples of all the contextualised uses of particular phraseologies. The findings can be useful to describe the institutionalised small culture nature (Holliday 1999) specific to the discourses and genres, to identify the communicative event, setting, and writer and reader roles (see, for example, Partington 1998; Dudley-Evans and St John

174 Winnie Cheng

1998; Castello 2002; Tribble 1997, 2000; Scott and Tribble 2006; Gunnarsson 2005; Palmer-Silveira, Ruiz-Garrido and Fortanet-Gómez 2006; Warren 2007). The implications of the corpus evidence discussed in this paper for ESP teaching methods and the design and writing of learning and teaching materials are worth exploring in greater detail. ESP writers could draw on the findings of corpus researchers, in the form of research papers, dictionaries, grammar books, and other resources, when they write and revise teaching materials, tasks and activities. The main pedagogical implication of this study is that the approaches of corpusdriven study (Tognini-Bonelli 1996) and data-driven learning (DDL) (Johns 1991) could be combined. In explaining DDL, Johns (2002: 108) suggests that one should “confront the learner as directly as possible with the data, and to make the learner a linguistic researcher”. In this context, the learners can be ESP teachers, materials writers, learners, or professionals. With the necessary knowledge and skills, they could become ESP language researchers, involving themselves in collecting their specialised texts, or even compiling their own specialised corpora, and generating the results for achieving an understanding of the aboutness of texts and corpora. When they are engaged in the process of describing the aboutness of a text, and then the aboutness of the specialised corpus of which the text is part, they will acquire a much more insightful and generalised understanding of the language used in their specific disciplines or professions.

Acknowledgements The work described in this paper was the result of the collaborative research efforts in the direction of aboutness with Professor Elena Tognini-Bonelli, University of Siena, Italy, and Professor John McH. Sinclair, the Tuscan Word Centre, Italy. Professor Tognini-Bonelli kindly provided access to the CÆT. The work described in this paper was substantially supported by a grant from the Research Grants Council of the Hong Kong Special Administrative Region (Project No. B-Q02J).

References Ahmad, K. 2005. Terminology in Text. Paper presented at the Tuscan Word Centre International Workshop: Dial a Corpus. Certosa di Pontignano, Italy. June, 2005. Bartsch, S. 2004. Structural and Functional Properties of Collocations in English: A Corpus Study of Lexical and Pragmatic Constraints on Lexical Co-occurrence. Tübingen: Narr.


Biber, D., Conrad, S. & Reppen, R. 1998. Corpus Linguistics: Investigating Language Structure and Use. Cambridge: CUP. Biber, D., Johansson, S., Leech, G., Conrad, S. & Finegan, E.. 1999. Longman Grammar of Spoken and Written English. London: Longman. Bowker, L. & Pearson, J. 2002. Working with Specialized Language: A Practical Guide to Using Corpora. London: Routledge. Byram, M. 1997. Teaching and Assessing Intercultural Communicative Competence. Clevedon: Multilingual Matters. Carter, R. & McCarthy, M. 2006. Cambridge Grammar of English. Cambridge: CUP. Castello, E. 2002. Tourist-information Texts: A Corpus-based Study of Four Related Genres. Padova: Unipress. Charteris-Black, J. & Ennis, T. 2001. A comparative study of metaphor in Spanish and English financial reporting. English for Specific Purposes 20: 249–266. Charteris-Black, J. & Musloff, A. 2003. ‘Battered hero’ or ‘innocent victim’? A comparative study of metaphors for euro trading in British and German financial trading. English for Specific Purposes 22: 153–176. Cheng, W. 2006. Describing the extended meanings of lexical cohesion in a corpus of SARS spoken discourse. International Journal of Corpus Linguistics 11(3): 325–344. (Special Issue Corpus Linguistics and Lexical Cohesion, J. Flowerdew & M. Mahlberg (eds)). Cheng, W. 2007. Hong Kong Financial Services Corpus: Empowering professionals to learn the language of their profession. Association for Business Communication 7th Asia Pacific Convention, 28–30 March 2007. Cheng, W. & Warren, M. 2006. Corpus-driven analysis of phraseology: The concgram. Special and Varied Corpora, The Tuscan Word Centre, 27–31 October 2006, Certosa di Pontignano, Tuscany. Cheng, W., Greaves, C. & Warren, M. 2006. From n-gram to skipgram to concgram. International Journal of Corpus Linguistics 11(4): 411–433. Clear, J. 1993. From Firth principles: Computational tools for the study of collocation. In Text and Technology: In Honour of John Sinclair, M. Baker, G. Francis & E. Tognini-Bonelli (eds), 271–292. Amsterdam: John Benjamins. Cowie, A. (ed.). 1998. Phraseology: Theory, Analysis, and Applications. Oxford: Clarendon Press. Cowie, A. 2004. Phraseology. In Concise Encyclopedia of Pragmatics, J. Mey (ed.), 673–677. Amsterdam: Elsevier. Dudley-Evans, T. & St John, M.-J. 1998. Developments in ESP: A Multi-disciplinary Approach. Cambridge: CUP. Fletcher, W. 2006. Phrases in English. Retrieved 15 February 2006, from /http://pie.usna.edu/. Greaves, C. 2006. Concgrams and their implications in language learning and teaching. Third Inter-Varietal Applied Corpus Studies (IVACS) International Conference: Language at the Interface. University of Nottingham, UK, 23–24 June, 2006. Gunnarsson, B.-L. (ed.). 2005. Communication in the Workplace. Uppsala: Uppsala University. Hewings, M. 2005. Editorial. English for Specific Purposes 24(1): 1–3. Holliday, A. 1999. Small culture. Applied Linguistics 20(2): 237–264. Hunston, S. 1995. A corpus study of some English verbs of attribution. Functions of Language 2(2): 133–158. Hunston, S. 2002. Corpora in Applied Linguistics. Cambridge: CUP.

176 Winnie Cheng

Hunston, S. & Francis, G. 2000. Pattern Grammar: A Corpus-driven Approach to the Lexical Grammar of English. Amsterdam: John Benjamins. Johns, T. 1986. Microconcord: A language-learner’s research tool. System 14(2): 151–162. Johns, T. 1991. Should you be persuaded: Two samples of data-driven learning materials. In Classroom Concordancing (EnglishLanguage Research Journal 4), T. Johns & P. King (eds), 1–16. Birmingham: ELR. Johns, T. 2002. Data-driven learning: The perpetual challenge. In Teaching and Learning by Doing Corpus Analysis. Proceedings of the Fourth International Conference on Teaching and Learning Corpora, Graz, 19–24 July 2000, B. Kettemann & G. Marko (eds), 107–117. Amsterdam: Rodopi. Nesselhauf, N. 2005. Collocations in a Learner Corpus. Amsterdam: John Benjamins. Palmer-Silveira, J. C., Ruiz-Garrido, M. F. & Fortanet-Gómez, I. (eds). 2006. International and Intercultural Business Communication. Theory, Research and Teaching. Frankurt: Lang. Partington, A. 1998. Patterns and Meanings: Using Corpora for English Language Research and Teaching. Amsterdam: John Benjamins. Phillips, M. 1989. Lexical Structure of Text [Discourse analysis monographs 12]. Birmingham: University of Birmingham. Renouf, A. & Sinclair, J. M. 1991. Collocational frameworks in English. In English Corpus Linguistics: Studies in Honour of Jan Svartvik, K. Aijmer & B. Altenberg (eds), 128–143. London: Longman. Römer, U. 2006. Pedagogical applications of corpora: Some reflections on the current scope and a wish list for future development. Zeitschrift für Anglistik und Amerikanistik (ZAA) 54(2): 121–134. Scott, M. 1997. PC Analysis of key words – and key key words. System 25(1): 1–13. Scott, M. 2000. Focusing on the text and its key words. In Rethinking Language Pedagogy from a Corpus Perspective, Vol. 2, L. Burnard & T. McEnery (eds), 103–122. Frankfurt: Lang. Scott, M. 2001. Comparing corpora and identifying key words, collocations, and frequency distributions through the WordSmith Tools suite of computer programs. In Small Corpus Studies and ELT: Theory and Practice, M. Ghadessy, A. Henry & R. L. Roseberry (eds), 47–67. Amsterdam: John Benjamins. Scott, M. 2004. WordSmith Tools version 4. Oxford: OUP. Scott, M. & Tribble, C. 2006. Textual Patterns. Key Words and Corpus Analysis in Language Education. Amsterdam: John Benjamins. Simpson, R. & Swales, J. (eds). 2001. Corpus Linguistics in North America. Ann Arbor MI: University of Michigan Press. Sinclair, J. M. 1987. The nature of the evidence. In Looking Up: An Account of the COBUILD Project in Lexical Computing, J. M. Sinclair (ed.), 150–159. London: Collins. Sinclair, J. M. 1991. Corpus Concordance Collocation. Oxford: OUP. Sinclair, J. M. 1996. The search for units of meaning. Textus 9(1): 75–106. Sinclair, J. M. 2004a. English Collocation Studies. London: Continuum. Sinclair, J. M. 2004b. Trust the Text. London: Routledge. Sinclair, J. M. 2005a. Document Relativity. Ms, Tuscan Word Centre, Italy. Sinclair, J. M. 2005b. Corpus and text. Basic principles. In Developing Linguistic Corpora: A Guide to Good Practice, M. Wynne (ed.), 1–16. Oxford: Oxbow Books. Sinclair, J. M. 2006. Aboutness 2. (manuscript), Tuscan Word Centre, Italy.


Sinclair, J. M., Jones, S. & Daley, R. 1970. English Lexical studies, Report to the Office of Scientific and Technical Information (OSTI) on Project C/LP/08. Department of English. University of Birmingham, January 1970. Sinclair, J. M. & Mauranen, A. 2006. Linear Unit Grammar. Amsterdam: John Benjamins. Sinclair, J. M. & Renouf, A. 1988. A lexical syllabus for language learning. In Vocabulary and Language Teaching, R. Carter & M. McCarthy (eds), 140–158. London: Longman. Teubert, W. 2005. Evaluation and its discontents. In Strategies in Academic Discourse, E. Tognini-Bonelli & G. Del Lungo Camiciotti (eds), 185–204. Amsterdam: John Benjamins. Tognini-Bonelli, E. 1996. Corpus Theory and Practice. Birmingham: TWC. Tognini-Bonelli, E. 2001. Corpus Linguistics at Work. Amsterdam: John Benjamins. Tognini-Bonelli, E. 2002. Functionally complete units of meaning across English and Italian: Towards a corpus-driven approach. In Lexis in Contrast: Corpus-based Approaches, B. Altenberg & S. Granger (eds), 73–96. Amsterdam: John Benjamins. Tognini-Bonelli, E. 2006. The corpus as an onion. Special and varied corpora, The Tuscan Word Centre, 27–31 October 2006, Certosa di Pontignano, Italy. Tribble, C. 1997. Improvising corpora for ELT: Quick-and-dirty ways of developing corpora for language teaching. In Practical Applications in Language Corpora, B. LewandowskaTomaszczyk & P. J. Melia (eds), 106–117. Lodz: Lodz University Press. Tribble, C. 2000. Genres, keywords, teaching: towards a pedagogic account of the language of project proposals. In Rethinking Language Pedagogy from a Corpus, L. Burnard & T. McEnery (eds), 75–90. Frankfurt: Lang. van Ek, J. A. 1986. Objectives for Foreign Language Learning, Vol. 1: Scope. Strasbourg: Council of Europe. Warren, M. In press. 2007. An initial corpus-driven analysis of the language of call-centre operators and customers. ESP Across Cultures, Vol. 4. Williams, G. 1998. Collocational networks: Interlocking patterns of lexis in a corpus of Plant Biology research articles. International Journal of Corpus Linguistics 3(1): 151–171. Williams, G. 2002. In search of representativity in specialised corpora: Categorisation through collocation. International Journal of Corpus Linguistics 7(1): 43–64. Wilson, A., Archer, D. & Rayson, P. 2006. Corpus Linguistics around the World. Amsterdam: Rodopi.

New types of corpora for new educational challenges Collecting, annotating and exploiting a corpus of textbook material Fanny Meunier and Céline Gouverneur

The present paper addresses the relationship between corpora and one commonly used type of pedagogical material in instructed English as a Foreign Language (EFL) settings, namely the textbook. Major English Language Teaching (ELT) publishers increasingly use native and learner corpora as input material on which to build, at least in part, new series of reference and pedagogical material such as dictionaries, grammar or vocabulary books. Surprisingly, however, English for General Purposes (EGP) textbooks are exceptions to the rule and still shy away from corpora. After a first section on the corpus tradition in ELT, we provide a survey of textbook research in Section 2. The third section presents a new type of pedagogically annotated textbook corpus: the TeMa corpus (corpus of Textbook Material). From a methodological viewpoint, we will show that the collection and annotation of pedagogical corpora allow for new automated search options which are not available in the manual ‘page by page’ approach often adopted in traditional textbook analyses. Concrete examples of search options and their preliminary results will be presented in Section 4, together with suggestions on how pedagogically annotated corpora can help learners, teachers, and material designers meet new educational challenges.

1.

ELT material and the corpus tradition

Corpora are no longer absent from the learning and teaching scene. Their legitimacy as useful pedagogical aids has now been established, and publications addressing the use of corpora in second/foreign language teaching abound: Burnard and McEnery (2000), Sinclair (2004) or Connor and Upton (2004) provide comprehensive edited volumes on the use of corpora in language teaching and

180 Fanny Meunier and Céline Gouverneur

learning; Botley, McEnery and Wilson (2000) focus on the use of multilingual corpora in teaching and research; Granger, Hung and Tyson (2002) address the links between computer learner corpora, second language acquisition and foreign language teaching; Mukherjee and Rohrbach (2006) and O’Keeffe, McCarthy, and Carter (2007) deal with the use of native and learner corpora in the classroom and the necessary mediation between research findings in corpus linguistics and classroom pedagogy. Corpora have not only been valued in applied linguistics circles, they have found their way to the offices of major ELT publishers. The latter increasingly use native and learner corpora as a source of authentic data on the basis of which they build, at least in part, new series of reference and pedagogical material such as dictionaries, grammar or vocabulary books. The use of corpora has even become a selling point. Cambridge University Press (CUP) uses a special logo to advertise corpus-based publications and presents the various types of publications in tables classified according to levels (beginner, elementary, pre-intermediate, intermediate, upper-intermediate and advanced) and type of publication (including a mixture of target audience and contents specification: i.e. ‘adult’, ‘exam’, ‘professional English’, ‘Cambridge copy collection’, ‘grammar’, ‘vocabulary’, ‘dictionaries’, ‘methodology and linguistics’). Other publishers focus on the link between their in-house corpus and their dictionaries (see for instance Longman and the Longman Corpus Network). As for the type of corpus used, ELT publishers have a marked preference for native corpora as a reference resource to inform their material. This preference for native corpora is justified by the need to present real, authentic English. In that respect, whilst CUP offers a ‘real English guarantee’ to the buyers and users of their material (http://www.cambridge.org/elt/corpus/corpus_based_books.htm), Longman assures its readership that [they] ‘only see real English, as it is really used’. http://www.longman.com/dictionaries/corpus/index.html. As for Macmillan, the use of their World English Corpus is described as ‘a unique modern database of over 200 million words revealing fresh information on how words are used and natural examples of English as it is written and spoken now!’ (http://www. macmillandictionary.com/aboutcorpus.htm). Learner corpora are also being used . See http://www.cambridge.org/elt/corpus/corpus_based_books.htm for a presentation of the tables. . The six levels usually correspond to the A1, A2, B1, B2, C1 and C2 levels of the Common European Framework for languages (CEF, see http://www.coe.int/t/dg4/linguistic/Source/ Framework_EN.pdf ) . Visit http://www.longman.com/dictionaries/corpus/index.html for more details on the Longman Corpus Network.

New types of corpora for new educational challenges 181

by ELT publishers, but to a much smaller extent. Here again, when publishers refer to learner corpora, they seem to privilege in-house learner corpora (see for instance CUP and the Cambridge Learner Corpus at http://www.cambridge.org/ elt/corpus/learner_corpus2.htm or Longman and the Longman Learner Corpus at http://www.longman.com/dictionaries/corpus/learners.html). One of the first points that will be made in the present paper is that one extensively used type of ELT material which has, to date, not yet benefited from the corpus revolution is the English for General Purposes (EGP) textbook. Indeed, whilst publishers tend to acknowledge some sort of connections between EGP textbooks and corpora in terms of vocabulary selection or grammar syllabus, they admit that current textbooks are not corpus-based. To a query on the possible corpus-based nature of various EGP textbooks (Merlevede 2006) Longman stated that “although [name of a textbook] uses the corpus-based Longman Grammar of Spoken and Written English for both the grammar and vocabulary syllabuses, the course itself is not corpus-based” (idem, 2006: 94), Macmillan stated that “the choice of vocabulary in [name of a textbook] is heavily influenced by the Macmillan English Dictionary which is based on the World English Corpus” (idem, 2006: 94), and CUP explained that they “have tried to tie [name of a textbook] into the Common European Framework as far as possible” […], that they “have [their] own Cambridge Corpus of over 700 million words” […] but that “[h]owever, [they] would not say that [name of a textbook] is corpus based as it began life before the Cambridge Corpus was really in full swing across the Press”. CUP however states that they “hope to include [their] Corpus in all Secondary ELT titles in the future” (idem, 2006: 95). Publishers seem to acknowledge the importance of corpora in ELT but fail to give information on how exactly the corpus is used (or could be used) to flesh out the linguistic contents of their textbooks.

2.

A survey of textbook studies

2.1

General overview

Increased interest in textbook analysis goes back to the early 1980s and several lines of research can be distinguished: development of general criteria for textbook analysis (Williams 1983; Cunningsworth 1984 and 1995; Chambers 1997; Sheldon 1988), assessment of textbooks as a useful/useless type of pedagogical material (Swales 1995; O’Neill 1993; Ranalli 2003; Harwood 2005), focus on specific classroom activities (Jacobs and Ball 1996), and focus on specific lexical or grammatical contents (Gabrielatos 1994; Biber et al. 2004; Römer 2004a and b;


Koprowski 2005; Meunier and Gouverneur 2007). Table 1 provides a non-exhaustive survey of textbook studies carried out over the last two decades in terms of learning context (e.g. EFL, ESL, EGP, etc.), type of textbook analysed (e.g. international, national etc.), method of textbook analysis adopted (i.e. manual pageby-page analysis or automated corpus-based approach), level of proficiency addressed and number of volumes examined. As can be seen from Table 1, studies have addressed a wide range of linguistic aspects with grammar and vocabulary taking centre stage. The focus on the authenticity of input is equally important and reflects the heated debate over authentic versus non-authentic (or adapted/simplified) language, i.e. whether pedagogical materials should be authentic or not, and to what extent they should be adapted to specific learners or classroom contexts (for an in-depth discussion on the issue, see Widdowson’s (2003) review on the topic). The table also shows that several studies deal with specific registers such as English for Academic Purposes (Swales 2002; Paltridge 2002; Biber et al. 2002) and that phraseology constitutes a recent research interest (Biber et al. 2004; Koprowski 2005; Meunier and Gouverneur 2007; Gouverneur 2008).

2.2 Analysing textbooks: Page-by-page vs. corpus approach? The methodological approach adopted in the studies listed in Table 1 deserves further attention. Most studies were carried out using a manual, page-by-page approach. Only six recent studies have been conducted using more automatic methods: Biber et al. (2004) Römer (2004b, 2006); Chujo (2004), Anping (2005), Meunier & Gouverneur (2007) and Gouverneur (2008 and forthcoming). To our knowledge, the corpora exploited in the aforementioned studies are the only published cases of textbook corpora. The first textbook corpus is one component of the TOEFL 2000 Spoken and Written Academic Language Corpus (T2K-SWAL Corpus), designed by Biber et al. (2002). The corpus is a collection of academic language which students are exposed to in American universities. It contains 2,7 million words of spoken and written American English, among which 760,619 are written texts taken from academic textbooks. The corpus, originally compiled with the aim of suggesting design principles for the new TOEFL, also serves other purposes. Biber et al. (2004) investigated what they call lexical bundles in

. Most of these studies deal with more than one linguistic or language teaching aspect. For clarity’s sake, the main focus was retained as the criterion for classification. . For a detailed description of the T2K-SWAL Corpus, see Biber et al. (2002: 19).

progressives (spoken data)

discourse features

vocabulary; grammar

modals

possessives; demonstrative

grammatical tasks

Römer (2006)

Gilmore (2004)

Anping (2005)

Hyland (1994)

Gabrielatos (1994)

Nitta & Gardner (2005) Boxer & Pickering (1995) Vellenga (2004)

Cane (1998)

Miura (1997)

if clauses – spoken language

Römer (2004b)

EFL EGP EFL EGP ELT

EAP

EFL EGP EFL

EFL

EFL

EFL

conversation skills

EFL

EFL: integrated skills ESL: grammar books “government-authorized” conversation textbooks

?

international

international

writing textbooks

international + local (China)

international

local: German

local: German EFL textbook & grammar local: German

Learning Textbook type context

metalang.; explicit treatment of ESL & speech acts; metapragm. information EFL oral communication ELT

speech acts: complaints

modal auxiliaries

Focus

Römer (2004a)

Research Author area

Table 1. Overview of textbook research over the last two decades

Authenticity

Grammar

Pragmatics

page by page

unspecified

page by page

page by page

unspecified

page by page

page by page

corpus-based

corpus-based (GEFL TC) corpus-based (GEFL TC) page by page

page by page

Method adopted

senior high school ?

?

?

intermediate

beginner

5 levels: beginner to university university

secondary school secondary school secondary shool ?

Level

3

16

8

7

9

1

22

7+ 3 50

12

12

No of vol. 6


Other

Vocabulary and phraseology

Speaking

learning strategies: repetition, resource use, recording vocabulary norms

collocations

verb form clustering

lexical bundles

lexical phrases

phraseology

high-frequency verbs

pronunciation

argumentative writing

group activities writing

register variation

Ranalli (2003)

Gabrielatos (1994)

Hill (1996)

Biber et al. (2004)

Koprowski (2005)

Meunier & Gouverneur (2007) Gouverneur (2008)

Gabrielatos (1994)

Swales (1995)

Jacobs & Ball (1996)

Biber et al. (2002)

dissertation writing

language of cause and effect

Paltridge (2002)

Moreno (2003)

Reda (2003)

vocabulary levels

Chujo (2004)

Table 1. (continued)

EAP

EAP

EFL EGP EFL EGP EFL EGP EAP ESP; ESL EFL EGP EAP

EFL EGP

EAP

EFL

EGP ESP EFL EGP EFL EGP voc books EFL

task-based texbooks

international

random (published since 1990) American

academic writing

international

international

international

international

American

international

international

international

international

local: Japanese

intermediate and upper-intermediate advanced

university

?

?

page by page

corpus-based (T2K-SWAL) page by page

unspecified

unspecified

upper-inter. and advanced

university

university

all

advanced

11

8

?

10

3

1

3

5

3

?

?

3

intermediate 7 and advanced upper-interme- 3 diate beginner to 6 advanced

corpus-based (TeMa corpus) corpus-based intermediate (TeMa corpus) and advanced page by page beginner

corpus-based (T2K-SWAL) manual (list)

unspecified

unspecified

unspecified

corpus-based (wordlists) page by page



EAP classroom use and textbooks. The study reveals that classroom academic discourse and EAP textbook discourse display specific language features. The second example of a textbook corpus is the German English as a Foreign Language Textbook Corpus (GEFL TC). It was compiled by Römer (2004a) and consists of texts taken from two series of general textbooks intended for German learners of English. The texts included in the corpus are supposed to represent spoken language (e.g. dialogues). GEFL TC consists of about 100,000 words. In her two corpus-driven studies, Römer (2004a, 2006) compared what she calls ‘school English’ with authentic English by examining two aspects of grammar (modal auxiliaries and progressives). The results she obtained revealed striking discrepancies between the spoken language included in the textbooks and real spoken data. As for Chujo’s work (2004), it aimed to measure the gradations of vocabulary across levels in EGP and ESP textbooks The texts included in the textbooks were scanned, proofread and part-of-speech tagged. Lemmatised wordlists were then computed and compared to a lemmatised frequency wordlist from the British National Corpus. Chujo also compiled specialised vocabulary wordlists from the coursebooks and from tests he had selected. Although he does not mention the term corpus explicitly, Chujo (2004) uses a corpus of texts taken from textbooks and tests. A fourth example of textbook corpus has been compiled by Anping (2005) and consists of over one million words of text taken from international EFL textbooks and from textbooks made in China for Chinese EFL learners. Most parts of the corpus are lexically and semantically tagged. Anping’s (2005) corpus-driven study consisted in finding out whether the design of EFL textbooks used in China reflected recent learning theories and teaching approaches. Meunier & Gouverneur (2007) and Gouverneur (2008 and forthcoming) are corpus-driven studies focusing on the treatment of phraseology in ELT textbooks. The data analysed come from a corpus of textbook material, the TeMa corpus (see Section 3 for a detailed description) which contains over seven hundred thousand words. Before describing the features of the TeMa corpus, some terminological issues will be addressed. Two adjectives are usually used to refer to textbook corpora or ‘corpora of coursebooks’ as Gabrielatos calls them (2005: 5): ‘pedagogic’ and ‘pedagogical’. According to dictionaries, they both mean ‘relating to teaching methods or to the practice of teaching’. In the literature however, the adjective ‘pedagogic’ is the preferred label and the expression ‘pedagogic corpus’ . For a detailed description of the GEFL TC, see Römer (2004a). . For a detailed description of the corpus, see He Anping (2005).


was coined by Willis (n.d.) and defined by Hunston (2002: 16) as “a corpus consisting of all the language a learner has been exposed to. []. It can consist of all the coursebooks, readers etc a learner has used, plus any tapes etc they have heard.” If one sticks to Hunston’s definition, collecting a pedagogic corpus seems rather utopian. No learner, let alone his/her teachers, is in a position to provide an exhaustive list of all the language input he/she has been submitted to, be it inside or outside the classroom. Given what precedes, we suggest a more realistic definition of a pedagogic corpus as being a large enough and representative sample of the language, spoken and written, a learner has been or is likely to be exposed to via teaching material, either in the classroom or during self-study activities. Typical teaching material includes texts, tapes and exercises. Taking into account the new definition provided, the examples of textbook corpora described above can be referred to as pedagogic corpora as they all consist of representative samples of textbook data intended for the teaching of EFL. Hunston (2002: 16) also suggests a number of possible exploitations of pedagogic corpora. First, a pedagogic corpus can be used for awareness-raising purposes by providing the learner with all the instances of a word or phrase he/she has encountered in various contexts (see for instance Biber et al. (2004) and their study of EAP vocabulary in textbooks). Secondly, the data included in a pedagogic corpus can be compared with a corpus of authentic English to check the authenticity of the language presented to the learners (as is the case for the two textbook analyses carried out by Römer in 2004a and 2006). As will be shown in Sections 3 and 4, the TeMa corpus allows for a number of additional exploitations thanks to the annotation that has been inserted.

3.

A new type of pedagogically annotated corpus for textbook research

The TeMa corpus has been collected in the framework of a research project on phraseology in language learning and teaching. A review of textbook studies convinced us that one way of facilitating an in-depth analysis of textbook material was to have a computer-readable version of the material. Once collected, the newly created pedagogic corpus could then be analysed with the help of typical corpus linguistics tools such as text retrieval software.


Table 2. Textbooks included in the TeMa corpus Title

Level

Accelerate

Intermediate Advanced Advance Intermediate your English Advanced Clockwise Intermediate Advanced Cutting Intermediate Edge Advanced English Intermediate Panorama Advanced Initiative Intermediate Advanced Inside Out Intermediate Advanced Matters Intermediate Advanced New Intermediate Cambridge Advanced New Intermediate Headway Advanced

3.1

Authors

Year

SB

WB Publisher

Lodge, P. & B. Wright-Watson Scott-Malden, S. & J. Wilson / Broadhead, A. Jeffries, A. Forsyth, W. Cunningham, S. & P. Moor Cunningham, S. & P. Moor / O’Dell, F. / Walton, R. & M. Bartram Kay, S. & V. Jones Jones, C. & T. Bastow Bell, J & R. Gower Bell, J & R. Gower Swan, M. & C. Walter Jones, L. Soars, L. & J. Soars, L. & J.

1995 1997 / 2003 2001 2003 2005 2003 / 2003 / 2000 2000 2001 2003 2001 2003 2002 2003 2003

X X / X X X X X / X / X X X X X X X X X

/ / / X / / X X / / / X X X X X X / X X

MacMillan Heinemann CUP OUP Longman CUP CUP MacMillan Longman CUP OUP

The textbooks in TeMa

The textbooks used for the compilation of the TeMa corpus were selected among recent best-sellers on the international ELT market, in similar proportion among the most renowned publishers. Thirty-two volumes of English for General Purposes (EGP) coursebooks were chosen for inclusion in TeMa. They include the student’s book and workbook of the textbook series at the advanced and/or intermediate levels. Table 2 presents the title, level, authors, date of publication and types of volumes available (i.e. Student’s Book ‘SB’ and/or Workbook ‘WB’).

3.2 Corpus markup The TeMa corpus went first through a markup stage in order to identify each section of the corpus. Figure 1 illustrates the incremental mark-up stage of the . Only one textbook series was published in the 90s. The nine other series were published after 2000.

Exercises 6123

Guidelines 6124

Exercises 6113

Guidelines 6114

Figure 1. Markup stage of the TeMa corpus

Guidelines 6214

Exercises 6213

Tapescripts 6212

Tapescripts 6122

Tapescripts 6112

Texts 6111

Guidelines 6224

Exercises 6223

Tapescripts 6222

Texts 6221

Workbook 622

Intermediate 62 Student’s book 621 Texts 6211

Workbook 612 Texts 6121

Student’s book 611

Advanced 61

New Headway 6



corpus. A first subdivision is based on the textbook series, the levels, and the type of book (i.e. SB or WB). Each textbook series is first given a code number. New Headway, for instance, has been assigned number 6. An extra digit is then added to represent the levels (1 for advanced and 2 for intermediate). New Headway advanced is thus 61 and New Headway intermediate 62. Each level is then further divided into student’s book (1) and workbook (2). The last subdivision represents the types of input provided: the texts (1), the transcription of the tapescripts (2), the vocabulary exercises (3) and the guidelines to these exercises (4). Each coursebook series is divided in sixteen potential subcorpora, each identified by a 4-digit markup. The TeMa corpus is innovative in a number of ways. First, it is rather large. With over 700,000 words of textbook material, it is one of the largest pedagogic corpora available. A second aspect is the richness of the pedagogic input, i.e. not only texts (as was the case for T2K-SWAL and for Anping’s corpus), not only spoken data (as was the case for GEFL TC) but both, plus vocabulary exercises and guidelines to the exercises. Thirdly, the language collected in the TeMa corpus comes from international EGP textbooks intended for any learners of English as a Foreign Language, with no mother tongue background restriction. The type of target audience of the textbooks collected in the four other pedagogic corpora was more specific: T2K-SWAL Corpus only contains EAP books for American students, GEFL TC is based on textbooks designed specially for German EFL learners, and a large part of Anping’s corpus consists of texts taken from EFL textbooks made in China for Chinese learners.

3.3 Pedagogical tagging Another aspect that singles out TeMa from other types of textbook corpora is the pedagogical tagging that has been applied to the vocabulary subcorpus. Corpus annotation is very common in corpus linguistics and many corpora can be partof-speech tagged, syntactically parsed, or even error-tagged (as is the case for many learner corpora). The type of tagging that has been used for TeMa belongs to what is commonly called ‘problem-oriented tagging’ and is defined by de Haan (1984: 125) as the phenomenon whereby users will take a corpus and add to it their own form of annotation, oriented particularly towards their own research goal. The subcorpus of vocabulary exercises in TeMa has been tagged according to the pedagogical tasks the learners have to perform when doing the exercises. This pedagogical tagging has been applied to all the vocabulary exercises on the basis of both the learning activities learners have to engage in (e.g. match words and their definitions, choose an item from a multiple-choice option, etc.) and on


Figure 2. Vocabulary exercise as it appears in the textbook

the pedagogical status of the lexical items in the exercises (e.g. a pre-selected list of words in a box). The following excerpt from the student’s book of Clockwise Intermediate (Forsyth 2003) illustrates how vocabulary exercises have been coded. Figure 2 is a scan of the actual page of the book, whilst Figure 3 presents its annotated corpus version. As can be seen from Figure 3, each exercise is given a unique reference. The example presented in Figure 3, i.e. , is taken from CLockwise Intermediate Student’s Book – Unit 6 – Page 24 – Exercise 1. The four-digit tag before each word or sentence (1213 in this case) refers to the origin of the exercise (see Figure 2 for an illustration). The two-letter tags between brackets (BC) indicate the pedagogical status of the lexical items presented. BC is used when words are presented in a box (hence B) from which items are to be selected to complete (hence C) the sentences in the exercise. As for the introductory tag in front of each exercise line, it gives information on the pedagogical task that has to be performed. CB in Figure 3 means “complete the sentence with words from a box”. The tags within the exercise sentences (e.g. AB) refer to the status of the lexical items within the exercise sentences: “from”, “to”, “too” are all preceded by AB as they are all answers taken from a box, hence (AB). Each sentence ends with a dollar sign ($) and within the sentences, the answers are followed by a hash (#). These additional signs make it easier to spot the beginning and the end of sentences as well as to extract and contextualise the exact lexical items being practised in the exercises. To tag our corpus pedagogically, a list of over 80 tags has been drawn. Seven main types of pedagogical tasks were identified during the compilation of


1213(BC)–#$ 1213(BC)as#$ 1213(BC)between#$ 1213(BC)from#$ 1213(BC)in#$ 1213(BC)that #$ 1213(BC)to#$ 1213(BC)too #$ 1213(BC)very#$ 1213(CB)He’s completely different 1213(AB)from# her$ 1213(CB)They’re quite similar 1213(AB)to# each other in age$ 1213(CB)I think she’s 1213(AB)too# young for him. She’ll get bored with him$ 1213(CB)They’ve got a lot 1213(AB)in# common$ 1213(CB)I think they’re quite a good couple : they look 1213(AB)very# similar$ 1213(CB)the single woman looks quite like 1213(AB)no word# the older man-except 1213(AB)that# she’s a woman of course !$ 1213(CB)there are so many differences 1213(AB)between# them ; they’ll split up before long !$ 1213(CB)She looks about the same height 1213(AB)as# him$

Figure 3. Pedagogically annotated vocabulary exercise as it appears in the corpus

the vocabulary exercises subcorpus: complete, define, match, replace, understand, correct, (re)write. Each main type was in turn divided into subcategories. Two tasks will be illustrated and explained hereafter: complete and match.

3.3.1 The COMPLETE tag The “complete” category, represented by the generic capital letter C*, can be subdivided in seven more specific tasks. (C) alone is used when learners have to complete a sentence, fill in a blank in a sentence without any prompt (such as a pre-selected list of words, multiple choice options, first letter of the words, etc). Learners have to retrieve the words or expressions from their mental lexicon on the sole basis of the context provided. 2113 (C) Can I get past please? oh I’m sorry, are my bags 2113 (A) in the way#? I’ll put them up in the locker.$

The answer, which has to be provided by the learner, is preceded by the status tag (A). . The complete list of tags will be available in Gouverneur (forthcoming).


(CB) stands for “complete from a box” and refers to tasks where learners have to complete sentences choosing from a pre-selected set of words provided in a box. The lexical items which are presented in the box, and which should be used to complete the sentence, are given the tag (BC) for “box to complete”. The answers get the AB tag (“answer from a box”). Here follows an example of (CB): 5113(BC)dubbed#$ 5113(BC)subtitles#$ 5113(CB)Foreign-language films can be shown with 5113(AB)subtitles# or they may be 5113(AB)dubbed#.$

In (CE), “complete-exercise”, learners have to complete sentences using lexical items that have to be chosen from an exercise done previously. In such cases, the answers are referred to as (AE) (answer from exercise), such as in: 3213 (CE) All I did was ask him to smile for the camera and he 3213 (AE) gave me a black eye#.$ 3213 (CE) His aggressive behaviour is unacceptable. He should be 3213 (AE)charged with assault#.$

The expressions give a black eye and to be charged with assault had been practised in an exercise done previously. The (CT) tag, “complete-text”, is used in front of sentences which should be completed with lexical items taken from a text seen previously in the unit, as in: 0213(CT)Products with too much 0213(AT)packaging# use a lot of energy to produce and distribute.$ 0213(CT)Find out if there is anything harmful in a product by writing to the 0213(AT)manufacturer#.$

The words packaging and manufacturer are words included in a text read previously in the unit. The answers are preceded by (AT), which stands for “answer from a text”. The (CZ) tag was chosen to refer to sentences which learners have to complete with a lexical item they have to pick from a multiple choice included in the sentence. The options provided are referred to with the tag (BZ) and the correct answer is given the tag (AZ), as illustrated below:


2113(CZ)A person who resembles a famous person can be called 2113(BZ)a lookalike/ a lookout/ an onlooker# 2113(AZ)a lookalike#.$ 2113(CZ)The proverb “Look before you 2113(BZ)jump/ leap/ strike# 2113(AZ)leap# means you should think about the possible dangers before you do something.$

(CZX) tasks are similar to the previous ones but instead of choosing the correct answer from a list of possible options, learners have to cross out the wrong answer. 2213(CZX)remind 2213(BZX)someone to do something/ someone about an appointment/ someone of another person/ to phone someone# 2213(AZX)to phone someone#$ 2213(CZX)forget 2213(BZX)to do something/ someone’s birthday/ of something/ about something# 2213(AZX)of something#$

The tag (CW) is used when learners have to provide a morphologically derived form for a given word, for instance the adjective derived from a particular noun, as in: 2213(CW)industry# 2213(AW)industrial#$ 2213(CW)history# 2213(AW)historical#$ 2213(CW)crowd# 2213(AW)crowded#$

The tag to refer to the answer in those cases is (AW).

3.3.2 The MATCH tag Matching exercises are extremely common in textbooks. To annotate such exercises, the tags are placed in front of the lexical items to be matched. Given the fact that, very often, the two parts of a matching exercise have the same “weight”, the left- and right-hand parts of the exercises were arbitrarily assigned the (MQ) for “match-question” and (MA) for “match-answer” tags respectively. 6213(MQ)strong# 6213(MA)coffee#$ 6213(MQ)full-time# 6213(MA)job#$ 6213(MQ)film# 6213(MA)star#$ 9113(MQ)I don’t believe a word of it!# 9113(MA)I don’t believe it at all#$ 9113(MQ)To eat your words# 9113(MA)To admit being wrong#$ 9113(MQ)By word of mouth# 9113(MA)By speaking and not by writing#$


As illustrated in the examples above, the two elements to be matched can be the two parts of multi-word units (collocations or compound nouns for instance) or paraphrases and synonyms. It must be noted that in the case of synonyms, paraphrases and meanings, the exercises have not been classified in the ‘definition’ category as the primary pedagogical technique underlying the task is to make the connections (matching) between the two elements and not to define a given item. Here again, both parts of the matching exercise can come from various types of input: a text (MQT/MAT), a previous exercise (MQE/MAE) or a box (MQB/ MAB). Some of the possible options are presented in the following box. 9113(MQT)date back# 9113(MA)be invented in#$ 9113(MQT)turn into# 9113(MA)change in form or nature#$ 9113(MQE)ground plan# 9113(MA)a drawn plan of a building at ground level 1#$ 9113(MQE)main artery# 9113(MA)big or principal road#$ 6123(BMA)for#$ 6123(BMA)from#$ 6123(MQ)The company isn’t liable# 6123(MAB)for# 6123(MA)any damages caused to vehicles parked on the premises#.$ 6123(MQ)Bill is emotionally detached# 6123(MAB)from# 6123(MA)his parents. He hardly ever speaks to them#.$

Answers may also be chosen from a multiple choice included in the sentence. The assigned tags are then (MQZ) for the question part, (BZM) for the box to choose from, and (MAZ) for the answer, as illustrated below: 1213(MQZ)my brother is married# 1213(BZM)a woman called Jenny/ to a woman called Jenny# 1213(MAZ)to a woman called Jenny#$ 1213(MQZ)my brother married1213(BZM)a woman called Jenny/ to a woman called Jenny# 1213(MAZ)a woman called Jenny#$ 1213(MQZ)he met# 1213(BZM)her at a party/ to her at a party# 1213(MAZ)her at a party#$ 1213(MQZ)he was introduced# 1213(BZM)her at a party/ to her at a party# 1213(MAZ)to her at a party#$ 1213(MQZ)he was fascinated# 1213(BZM)by her/ in her# 1213(MAZ)by her#$ 1213(MQZ)he was very interested# 1213(BZM)by her/ in her# 1213(MAZ) in her#$

Given the size of the pedagogical tagset created for the annotation of the corpus (about 80 in total), it must be acknowledged that the tagging stage was extremely time-consuming. The compilation of the vocabulary exercises required careful


selection and analysis. Suggestions of tags were inserted in the paper copy of the textbook for each exercise selected and, only then, was the exercise ready for compilation. Our progression in analysing the exercises often forced us to come back to previously annotated exercises whenever we discovered subtleties of tasks that had not been encoded, which inevitably led to numerous checks, revisions and adaptations. However, once the annotation stage is completed, the corpus offers numerous paths for exploitations, as will be shown in Section 4.

4.

Meeting new pedagogical challenges

We will now examine in what way the collection and annotation of a textbook corpus can help meet the new pedagogical challenges mentioned in the title of the paper. A pedagogically annotated corpus makes it possible to explore the data from a variety of perspectives never addressed before or addressed on a much smaller scale given the manual analysis involved. On a descriptive level, using a pedagogically tagged textbook corpus makes it possible to provide a solid empirical description of the material under analysis. A comparison of vocabulary selection across levels can be carried out and it becomes possible to determine what a specific level actually means in terms of vocabulary selection (e.g. by providing a list of all the words/expressions practised in the exercises of a level and by comparing it with a list of words form a lower/higher level). The relationship between the input provided in the texts and audio files and the words/expressions practised in the exercises can also be addressed. The various types of cognitive tasks that learners have to perform when doing the exercises can be analysed. The results of such studies can help raise the publishers’ or textbook writers’ awareness of the types of exercises they propose, in what proportion and to what target audience. Gouverneur (2008), in a pilot study on the aforementioned questions, reports on some preliminary results. A comparison of the collocations presented in the exercises reveals a total lack of consistency between the textbooks examined, i.e. very few were common to all textbooks. As to the weight of pedagogical tasks, the study shows that some tasks are common to all levels of proficiency, such as ‘complete’ tasks, for instance, whilst some others are more specific to one level, such as the ‘replace’ tasks which are very frequent in the advanced textbooks but can hardly be found at the intermediate level. The study also states that not enough tasks promote cognitive processes such as noticing or receptive and productive retrieval. Cognitively oriented SLA research has demonstrated the importance of noticing, extrapolation and rehearsal (see among others De Bot et al. 2005). A detailed analysis of a pedagogically tagged textbook corpus like TeMa helps researchers specify where and when exactly noticing, extrapolation and rehearsal


of lexical items are practised in the textbook and, subsequently, propose possible improvements to current practice. One such improvement could be the addition of extra electronic input to the textbook. As is the case with learners’ dictionaries, which now almost invariably include a CD-ROM version containing extra material such as concordance lines, thesaurus, extra examples, or exercises, it would be reasonable to expect a similar evolution for textbooks. The accompanying CDROM could not only include the transcript of the audio files and texts included in the paper version of the textbook but also more texts, more exercises and more authentic native corpus input. Such add-ons, combined with user-friendly search options, would make it possible, for teachers and learners alike, to access more contextualised instances of words and expressions and to practise independent or teacher-led data-driven learning activities. Another issue worth addressing with the help of a textbook corpus such as TeMa is the metalanguage used in the textbooks to refer to vocabulary and phraseology. Analysing the TeMa subcorpora containing the guidelines to the exercises helps us identify (1) the type of metalanguage used by material designers (general terms such as words or expressions or specific terms such as collocations or idioms?); (2) the consistency in their terminology (does the term idiom refer to fixed idioms, to conversational routines, to pragmatic phrasemes or to other types of multiword units?). A pilot study carried out on a small proportion of the corpus (Meunier and Gouverneur 2007) has shown that the metalanguage used in textbooks, and more particularly in the guidelines to the exercises, is still far too general and indirect. Textbook designers tend for instance to make use of terms such as ‘words’ or ‘expressions’ instead of using more specific terms which, as some argue (Lewis 2001), are not more difficult to remember and understand. The use of a (limited) set of specific and pedagogically-oriented terms might even facilitate the understanding of important concepts. Although the results reported here only deal with a limited number of research questions, the richness of the TeMa corpus allows for far more exploitations; the focus could be on one type of linguistic features (e.g. high frequency verbs); spoken and written input could also be compared in order to investigate language mode variations; and along the lines of what has already been done by Römer (2004a, 2004b) an analysis of the authenticity of textbook material could be carried out on larger corpora of textbook material. It should also be added that the pedagogical tagging of the TeMa corpus which was heavily oriented towards lexis could easily be extended to other aspects such as grammar exercises and metalanguage, or speech-act analysis.

5.


Conclusion

The previous sections have introduced what we believe to be a rather innovative type of corpus, namely a pedagogically annotated corpus of textbook material. Its description, collection and annotation procedures have been outlined, and preliminary results of exploitations have also been provided. We are aware that the methodology adopted in our research project has its limitations: the corpus collection and annotation is limited in size (focus on texts, audio transcripts, lexically-oriented exercises and guidelines to those exercises); no possible feedback on open questions types of exercises can be accessed (be it teacher-learner or peer interaction); no feedback on teacher input is available; and no general format-like information (pictures, maps, etc.) has been encoded. Despite these limitations however, we believe that the collection, annotation and exploitation of this new type of corpus makes it possible to access empirical evidence otherwise inaccessible. This type of empirical evidence includes access to frequency patterns of use not only in the input provided to the learners but also in the types of exercises suggested; access to the actual connections (made or not) between the input and its exploitation in the exercises; access to the metalanguage used to introduce formal aspects of language. Access to such type of information helps foster a reflexive approach to textbook editing and provide evidence-based guidelines to improving textbooks. Analyses such as those presented in Section 4 also helped us reveal what is good about textbooks. We have for instance demonstrated the growing awareness of the phraseological nature of the language and the presence of recycling and rehearsal exercises. As for potential areas of improvement, it should be made clear that they are not restricted to promoting the authenticity of textbook material. Although we believe it essential to offer learners input which is as authentic as possible, adapted or simplified input also has its relevance, especially perhaps at lower levels of proficiency. Other domains that could benefit from the analysis of pedagogically tagged corpora include: a reconsideration of the links between important issues revealed in the SLA literature and their possible inclusion and exploitation in textbooks; an improved awareness (on the part of the teachers) of what is contained in the material they use; and a possible revision of the grammatical and lexical metalanguage present in textbooks. We also believe that, thanks to a corpus approach,10 the inevitable initial limitations of the paper and ink format could disappear. 10. The ‘corpus approach’ mentioned is seen as: including much larger quantities of similar input for learners, providing opportunities for data-driven learning, providing access to corpus searches and words in context, etc.


We can only but hope that other types of pedagogically annotated corpora addressing similar or different issues as the ones presented here will soon be created and exploited.

References Anping, H. 2005. Corpus-based evaluation of ELT textbooks. Paper presented at the joint conference of the American Association of Applied Corpus Linguistics and the International Computer Archive of Modern and Medieval English, 12–15 May 2005, University of Michigan. Biber, D., Conrad, S., Reppen, R., Byrd, P. & Helt, M. 2002. Speaking and writing in the university: A multidimensional comparison. TESOL Quarterly 36(1): 9–48. Biber, D., Conrad, S. & Cortes, V. 2004. If you look at …: Lexical bundles in university teaching and textbooks. Applied Linguistics 25(3): 371–405. Botley, S., McEnery, T. & Wilson, A. (eds). 2000. Multilingual Corpora in Teaching and Research. Amsterdam: Rodopi. Boxer, D. & Pickering, L. 1995. Problems in the presentation of speech acts in ELT materials: The case of complaints. ELT Journal 49: 44–58. Burnard, L. & McEnery, T. (eds). 2000. Rethinking Language Pedagogy from a Corpus Perspective. Papers from the Third International Conference on Teaching and Language Corpora. Frankfurt: Lang. Cane, G. 1998. Teaching conversation skills more effectively. The Korea TESOL Journal 1: 31– 37. Chambers, F. 1997. Seeking consensus in coursebook evaluation. ELT Journal 51 (1): 29–35. Chujo, K. 2004. Measuring vocabulary levels of English textbooks and tests. Using a BNC lemmatised high frequency word list. In English Corpora under Japanese Eyes, J. Nakamura, N. Inoue, N. & T. Tabata (eds), 231–249. Amsterdam: Rodopi. Connor, U. & Upton, T. (eds). 2004. Applied Corpus Linguistics: A Multidimensional Perspective. Amsterdam: Rodopi. Cunningsworth, A. 1984. Evaluating and Selecting EFL Teaching Materials. Oxford: Heinemann. Cunningsworth, A. 1995. Choosing Your Coursebook. Oxford: Heinemann. De Bot, K., Lowie, W. & Verspoor, M. 2005. Second Language Acquisition. An Advanced Resource Book. London: Routledge. De Haan, P. 1984. Problem-oriented tagging of English corpus data. In Corpus Linguistics, J. Aarts & W. Meijs (eds), 123–139. Amsterdam. Rodopi. Gabrielatos, C. 1994. Collocations, pedagogical implications and their treatment in pedagogical materials. Ms, Research Centre for English and Applied Linguistics, University of Cambridge. Available at http://www.gabrielatos.com/Collocation.htm Gabrielatos, C. 2005. Corpora and language teaching: Just a fling, or wedding bells? TESL-EJ 8(4): A1, 1–37. Gilmore, A. 2004. A comparison of textbook and authentic interactions. ELT Journal 58 (4): 363–374.


Gouverneur, C. 2008. The phraseological patterns of high-frequency verbs in advanced English for general purposes: A corpus-driven approach to EFL textbook analysis. In Phraseology in Foreign Language Learning and Teaching, F. Meunier & S. Granger (eds). Amsterdam: Benjamins. Gouverneur, C. Forthcoming. Phraseology in Foreign Language Learning and Teaching: A Corpus-based Study of EFL Textbooks. PhD dissertation, Université Catholique de Louvain. Granger, S., Hung, J. & Petch-Tyson, S. (eds). 2002. Computer Learner Corpora, Second Language Acquisition and Foreign Language Teaching. Amsterdam: John Benjamins. Harwood, N. 2005. What do we want EAP teaching materials for? In Journal of English for Academic Purposes 4 (2): 149–161. Hill, V. J. 1996. Verb-form clustering and syllabus design. System 24 (4): 529–536. Hunston, S. 2002. Corpora in Applied Linguistics. Cambridge: CUP. Hyland, K. 1994. Hedging in Academic Writing and EAP textbooks. English for Specific Purposes 13: 251–281. Jacobs, G. M. & Ball, J. 1996. An investigation of the structure of group activities in ELT coursebooks. ELT Journal 50: 99–107. Koprowski, M. 2005. Investigating the usefulness of lexical phrases in contemporary coursebooks. ELT Journal 59(4): 322–332. Merlevede, J. 2006. Corpus Linguistics Implications in English Language Teaching. An Analysis of Corpus-based Pedagogical Materials. MA dissertation, Université catholique de Louvain, Louvain-la-Neuve. Meunier, F. & Gouverneur, C. 2007. The treatment of phraseology in ELT Textbooks. In Corpora in the Foreign Language Classroom, E. Hidalgo, L. Querada & J. Santana (eds), 119–139. Selected papers from the Sixth International Conference on Teaching and Language Corpora (TaLC), University of Granada, Spain, 4–7 July, 2004. Amsterdam: Rodopi. Miura, T. 1997. An Analysis of Aural/Oral Communications A. English Textbooks in Japanese Upper Secondary School. MA thesis, The University of Birmingham. Moreno, A. I. 2003. Matching theoretical descriptions of discourse and practical applications to teaching: The case of causal metatext. English for Specific Purposes 22: 265–295. Mukherjee, J. & Rohrbach, J. M. 2006. Rethinking applied corpus linguistics from a languagepedagogical perspective: New departures in learner corpus research. In Planing, Gluing and Painting Corpora: Inside the Applied Corpus Linguist’s Workshop, B. Kettemann & G. Marko (eds), 205–232. Frankfurt: Lang. Nitta, R. & Gardner, S. 2005. Consciousness-raising and practice in ELT coursebooks. ELT Journal 59(1): 3–13. O’Keeffe, A., McCarthy, M. & Carter, R. 2007. From Corpus to Classroom: Language Use and Language Teaching. Cambridge: CUP. O’Neill, R. 1993. Are textbooks symptoms of disease? Practical English Teaching 14(1): 11–13. Paltridge, B. 2002. Thesis and dissertation writing: An examination of published advice and actual practice. English for Specific Purposes 21: 125–143. Ranalli, J. M. 2003. The Treatment of Key Vocabulary Strategies in Current ELT Coursebooks: Repetition, Resource Use, Recording. MA thesis. Centre for English Language Studies, University of Birmingham. Reda, G. 2003. English coursebooks: Prototype texts and basic vocabulary norms. ELT Journal 57(3): 260–268.


Römer, U. 2004a. Textbooks: A corpus-driven approach to modal auxiliaries and their didactics. In How to Use Corpora in Language Teaching, J. Sinclair (ed.), 185–199. Amsterdam: John Benjamins. Römer, U. 2004b. Comparing real and ideal language learner input: The use of an EFL textbook corpus in corpus linguistics and language teaching. In Corpora and Language Learners, G. Aston, S. Bernardini & D. Stewart (eds), 151–168. Amsterdam: John Benjamins. Römer, U. 2006. Looking at looking: Functions and contexts of progressives in spoken English and ‘School’ English. In The Changing Face of Corpus Linguistics. Papers from the 24th International Conference on English Language Research on Computerized Corpora (ICAME 24), A. Renouf & A. Kehoe (eds), 231–242. Amsterdam: Rodopi. Sheldon, L. 1988. Evaluating ELT textbooks and materials. ELT Journal 42(4): 237–246. Sinclair, J. (ed.). 2004. How to Use Corpora in Language Teaching [Studies in Corpus Linguistics 12]. Amsterdam: John Benjamins. Swales, J. M. 1995. The role of the textbook in EAP writing research. English for Specific Purposes 14(1): 3–18. Swales, J. M. 2002. Integrated and fragmented worlds: EAP materials and corpus linguistics. In Academic Discourse, J. Flowerdew (ed.), 150–164. Harlow: Longman. Vellenga, H. 2004. Learning pragmatics from ESL & EFL textbooks: How likely? TESL-EJ, Vol. 8(2). Available from http://www.kyoto-su.ac.jp/information/tesl-ej/ej30/a3.html Widdowson, H. 2003. Defining Issues in English Language Teaching. Oxford: OUP. Williams, D. 1983. Developing criteria for textbook evaluation. ELT Journal 37 (3): 251–255. Willis, D. n.d. Syllabus as pedagogic corpus. Ms.

Textbooks Bell, J. and Gower, R. 2001a. Matters. Student’s Book Advanced. London: Longman. Bell, J. and Gower, R. 2001b. Matters. Workbook Advanced. London: Longman. Bell, J. and Gower, R. 2003a. Matters. Student’s Book Intermediate. London: Longman. Bell, J. and Gower, R. 2003b. Matters. Workbook Intermediate. London: Longman. Broadhead, A. 2003a. Advance your English. Coursebook. Cambridge: Cambridge University Press. Broadhead, A. 2003b. Advance your English. Workbook. Cambridge: Cambridge University Press. Cunningham, S. and Moor, P. 2003a. Cutting Edge. Advanced Student’s book. London: Longman. Cunningham, S. and Moor, P. 2003b. Cutting Edge. Advanced Workbook. London: Longman. Cunningham, S. and Moor, P. 2005a. Cutting Edge. Intermediate Student’s book. London: Longman. Cunningham, S. and Moor, P. 2005b. Cutting Edge. Intermediate Workbook. London: Longman. Forsyth, W. 2003. Clockwise. Advanced Classbook. Oxford: Oxford Uninversity Press. Jeffries, A. 2001. Clockwise. Intermediate Classbook. Oxford: Oxford Uninversity Press. Jones, C. and Bastow, T. 2001a. Inside Out. Student’s book Advanced. London: MacMillan. Jones, C. and Bastow, T. 2001b. Inside Out. Wordbook Advanced. London: MacMillan. Jones, L. 2002. New Cambridge Advanced English. Student’s Book. Cambridge: Cambridge University Press.


Kay, S. and Jones, V. 2000a. Inside Out. Student’s Book Intermediate. London: MacMillan. Kay, S. and Jones, V. 2000b. Inside Out. Workbook Intermediate. London: MacMillan. Lodge, P. and Wright-Watson, B. 1995. Accelerate Intermediate Student’s Book. MacMillan Heinemann. O’Dell, F. 2003a. English Panorama. Advanced Student’s Book 1. Cambridge: Cambridge University Press. O’Dell, F. 2003b. English Panorama. Advanced Student’s Book 2. Cambridge: Cambridge University Press. Scott-Malden, S. and Wilson, J. 1997. Accelerate. Advanced Student’s Book. MacMillan Heinemann. Soars, L. and Soars, J. 2003a. New Headway. Intermediate Student’s Book. Oxford: Oxford University Press. Soars, L. and Soars, J. 2003b. New Headway. Intermediate Workbook. Oxford: Oxford University Press. Soars, L. and Soars, J. 2003c. New Headway. Advanced Student’s Book. Oxford: Oxford University Press. Soars, L. and Soars, J. 2003d. New Headway. Advanced Workbook. Oxford: Oxford University Press. Swan, M. and Walter, C. 2003a. The New Cambridge English Course. Student. Intermediate Cambridge: Cambridge University Press. Swan, M. and Walter, C. 2003b. The New Cambridge English Course. Practice. Intermediate. Cambridge: Cambridge University Press. Walton, R. and Bartram, M. 2000. Initiative. Student’s book. Cambridge: Cambridge University Press.

The grammar of conversation in advanced spoken learner English Learner corpus data and language-pedagogical implications Joybrato Mukherjee

The present paper focuses on aspects of what has been labelled the ‘grammar of conversation’ in the Longman Grammar of Spoken and Written English (1999), i.e. syntactic features that are typical of spontaneous spoken language. Specifically, this paper will present some findings of three case studies of lexicogrammatical forms and structures in advanced German learners’ use of spoken English, including verb-noun collocations, the discourse marker you know and performance phenomena such as repetitions and contractions. The case studies are all based on the German component of the Louvain International Database of Spoken English Interlanguage (LINDSEI). The linguistic findings are taken as a starting-point for a discussion of language-pedagogical implications and applications, especially with regard to possible ways of teaching and learning the grammar of conversation in the EFL classroom.

1.

Introduction: Spoken grammar in corpus linguistics and language pedagogy

Spoken and written language differ considerably from each other with regard to lexis and grammar. It is now widely accepted that the lexicogrammatical differences between the two media are not primarily due to deficiencies of speech compared to writing. Rather, language should be viewed as structure-guided and systematic in both media: . I would like to thank Rosemary Bock and Christiane Brand for their feedback on an earlier version of the manuscript and Sandra Götz and Annedore Vorherr for their assistance with some of the corpus analyses on which this paper reports.

204 Joybrato Mukherjee

Spoken and written language do not differ in their systematicity: each is equally highly organized, regular, and productive of coherent discourse. [...] Spoken and written language do differ, however, in their preferred patterns of lexicogram(Halliday 1987: 69) matical organization.

It is hard to exaggerate the extent to which corpus-based research has paved the way towards an empirically sound description and theoretical modelling of this very ‘systematicity’ of the grammar of spoken English. Leech (2000: 676) speaks of a “new thinking on spoken grammar [that] has been sparked off by corpus studies” and provides the following list of characeristic features of spontaneous spoken language: (a) loose, relatively unintegrated structure with a very wide-ranging use of independent non-clausal (“fragmentary”) units; (b) the inappropriateness of the sentence to the analysis of spoken grammar; (c) simplicity of phrase structure (particularly of noun phrases); (d) repetitive use of a restricted lexicogrammatical repertoire; (e) grammatical features reflecting interactiveness and on-line (Leech 2000: 676) processing constraints.

This list captures perhaps the most significant overall result of corpus-based research into spoken language, namely the acceptance of spoken conversation – the most basic and prototypical type of speech – as a language variety in its own right. Accordingly, its lexicogrammatical organisation is best described by a “grammar of conversation” as sketched out in the corpus-based Longman Grammar of Spoken and Written English (Biber et al. 1999: 1037), including, for example, discourse markers, repetitions, hesitation phenomena and other disfluencies. The grammar of conversation is not only characterised by such speech-specific forms and structures, but also by quantitative differences with regard to many areas of the core grammar shared by speech and writing. For example, whereas both long passives (i.e. with an explicit by-agent) and left dislocations (e.g. The car, where is it?) are in principle available to language users in speech and writing alike, corpus data show that while the former is typically used in (academic) writing the latter is almost exclusively used in conversation (cf. Biber et al. 1999: 938, 957). Another case in point is the wide-ranging area of phraseology, including collocations and lexical bundles (i.e. recurrent multi-word sequences). Here, too, there are clear quantitative differences between speech and writing. For example, the collocation little baby and the lexical bundle I don’ know are much more frequently used in conversation than, say, in academic writing (cf. Biber et al. 1999: 988f.). In general, the wealth of corpus-based research into speech and writing over the past decades provides ample testimony of the fact that the lexicogrammar of spontaneous spoken English should not be squeezed – at least not in its entirety – into a traditional form of a “monolithic grammar” (cf. Conrad 2000, quoted in Hunston

The grammar of conversation in advanced spoken learner English 205

2002: 167), which is primarily designed for the description of written language, but should rather be captured in terms of a spoken grammar. While the concept of a spoken grammar is now generally accepted in the corpus-linguistic community and beyond, this concept has been widely neglected in language pedagogy until very recently. This is somewhat surprising because applied linguists and language-teaching professionals have often complained that the kind of spoken English used even by advanced learners of English is very often a far cry from native-like language use in spontaneous speech (cf. e.g. Kieweg 2000: 8; Mukherjee 2002: 142). As a consequence, derogatory terms such as ‘classroomese’ – and, in the German context, ‘Abiturspeak’ – have been used for the kind of orally performed written English that foreign language learners of English who have not benefited from studies abroad tend to use in spontaneous spoken conversations. Without any doubt, it will take much effort to establish the teaching of spoken grammar on the curricular agenda in the EFL classroom in countries like Germany, not only because empirically sound and corpus-based descriptions of authentic spoken English have not been available very long, but also because many language teachers are reluctant to accept that learners of English should actually be taught forms and structures of spoken English such as left dislocations and discourse markers (other than, say, well). On a more positive note, however, one should acknowledge the Common European Framework of Reference for Languages (Council of Europe 2001) as a highly significant step towards the systematic teaching of spoken grammar in the EFL classroom. In the Reference Framework, the competencies that foreign language learners need to acquire are described in terms of can-do statements at six proficiency levels (ranging from A1 ‘breakthrough’ to C2 ‘mastery’). For example, the ‘functional competence’ of ‘spoken fluency’ is described at the C2 level as follows: Can express him/herself at length with a natural, effortless, unhesitating flow. Pauses only to reflect on precisely the right words to express his/her thoughts or to find an appropriate example or explanation. (Council of Europe 2001: 129)

This is just one example of the all-pervading focus on communicative skills needed for real-life interactions with native speakers that is at the heart of the Reference Framework. It thus places special emphasis on spoken grammar. From a language-pedagogical perspective, not only is it relevant to describe the target norm for teaching English as a foreign language (on the basis of native corpora and/or in the form of can-do statements as in the Reference Framework), but it is also necessary to compile and analyse corpora of learner language in order . Abitur refers to the German A-levels. Note in this context also Cutting (2006: 172) who refers to spoken learner language as often being “bookish and pedantic”.


to find out what the “archetypal learner” (cf. Granger 1998: 7) utters at a given stage in the language learning process. It is only by carefully comparing native-like language use and actual learner language that it is possible to identify areas in which there is still a discrepancy between the target norm and learners’ speech and areas in which learners have already approximated to the target norm. In a wider setting, learner corpus linguistics has provided applied linguists, language-teaching professionals and materials designers with an entirely new kind of resource which makes it possible to abstract away from individual learners a supraindividual description of a specific stage in the learning process. A major milestone in the evolution of learner corpus linguistics has been the compilation of the International Corpus of Learner English (ICLE, cf. Granger et al. 2002), which also includes a German component with more than 200,000 words of argumentative essays written by advanced learners of English. The German component of ICLE has already been used in a number of studies, e.g. on adjective intensification (cf. Lorenz 1999), collocations (cf. Nesselhauf 2004a) and the progressive (cf. Römer 2005). As it is restricted to writing, a project of the same scale has been launched by the Centre for English Corpus Linguistics at the Université de Louvain: the Louvain International Database of Spoken English Interlanguage (LINDSEI, cf. De Cock et al. 2006). The German component of LINDSEI (LINDSEI-Ger) has already been compiled at the University of Giessen: it comprises 50 interviews, including some 90,000 words spoken by learners of English in their third and fourth years of university studies (cf. Brand and Kämmerer 2006). The corpus provides an ideal testbed for the exploration of the spoken grammar of advanced German learners of English – both for linguistic analyses of spoken learner English as a noteworthy variety in its own right and for language-pedagogical considerations of how to deal with spoken grammar in the EFL classroom. Accordingly, in the following sections of this paper I will address both linguistic and language-pedagogical issues related to spoken grammar. In Section 2, the main part of the present paper, I will provide an overview of three case studies which are all based on LINDSEI-Ger. These will include comparisons of non-native and native data along with spoken and written learner language. At the end of each case study, I will address some language-pedagogical issues and sketch out

. Note in this context that it is a hotly disputed issue whether native usage should serve as a target norm in the EFL classroom of the 21st century at all. While Mukherjee (2005) and Granger (2006) argue the case for retaining the native speaker norm in EFL teaching contexts, Jenkins (2004), for example, opts for a different model, namely an English-as-a-lingua-franca norm.


some suggestions for teaching and learning spoken grammar in the EFL classroom. In Section 3, I will offer some concluding remarks.

2.

The grammar of conversation in advanced German learners’ speech: Three case studies and their language-pedagogical implications

The LINDSEI-Ger-based case studies are concerned with three different aspects of spoken grammar. In the first case study (cf. Section 2.1), I will present some of the findings from the comparison of verb-noun collocations in spoken and written learner language. Generally speaking, collocations can be viewed as part of the core grammar shared by speech and writing; however, there are differences between spoken and written learner language with regard to the number, kind and range of collocations. In the second case study (cf. Section 2.2), I will focus on the discourse marker you know in LINDSEI-Ger. Discourse markers are items that are largely unique to speech and are typical of spontaneous spoken language – using discourse markers appropriately helps to render learner speech more natural and native-like. In the last case study (cf. Section 2.3), I will briefly address classic performance phenomena such as repetitions that occur only in speech and compare LINDSEI-Ger data with native data.

2.1 Verb-noun collocations in speech: spoken vs. written learner data There are various definitions of collocations on offer, and they can be plotted on a gradient from a more frequency-based understanding of collocations to a more phraseological definition. In much corpus-based work, a statistical orientation is clearly visible, e.g. in Hoey’s (1991: 6f.) definition: “[C]ollocation has long been the name given to the relationship a lexical item has with items that appear with greater than random possibility in its (textual) context”. By contrast, Cowie (1998) and Howarth (1998) focus more on the functional integrity of collocations as syntactic and/or pragmatic units. Howarth (1998: 24), for example, defines collocations as “combinations of words with a syntactic function as constitutents of sentences (such as noun or prepositional phrases or verb and object constructions)”. Nesselhauf (2004a) has provided a systematic and detailed study of verb-noun collocations (e.g. solve a problem, raise a question) in advanced German learners' argumentative essays in English. Her study is based on a 150,000-word pilot version of the German component of ICLE (GeCLE). She adopts a predominantly phraseological approach in that she considers verb-noun combinations as collocations in which the noun is used without an arbitrary restriction in its meaning,


Table 1. Top ten verb-noun collocations in GeCLE (cf. Nesselhauf 2004a) Rank

Collocation

Number of occurrences

Normalised frequency

1 2 3 4 5 6 7 8 9 10

solve a problem have a reason commit a crime have time have a chance do work get an impression have a look at raise a question take care of

25 19 18 17 17 15 14 14 12 11

0.16 ptw 0.12 ptw 0.12 ptw 0.11 ptw 0.11 ptw 0.10 ptw 0.09 ptw 0.09 ptw 0.08 ptw 0.07 ptw

but in which the verb is to some degree restricted to specific nouns (cf. Nesselhauf 2004a: 27ff.). In make a decision and shrug one’s shoulders, for example, the verbs make and shrug are considered to be restricted to a specific range of noun collocates, including decision and shoulders respectively. Table 1 provides a list of the most frequent verb-noun collocates in GeCLE. Vorherr (2006) compares some of the findings on verb-noun collocations in GeCLE with LINDSEI-Ger data. She observes, for example, that in advanced spoken English, learners tend to use other verb-noun collocations much more frequently. Table 2 provides a list of the most frequent verb-noun collocates in LINDSEI-Ger. A comparison of Tables 1 and 2 yields some interesting results. Firstly, the three verb-noun collocations that occur most frequently in GeCLE (i.e. solve a problem, have a reason, commit a crime) do not appear at all in LINDSEI-Ger. Secondly, the three most frequent verb-noun collocations in LINDSEI-Ger (i.e. have time, have a problem, have a chance) occur considerably less frequently in GeCLE – in fact, have a problem does not belong to the top ten verb-noun collocations in GeCLE. These two findings show that with regard to collocations advanced learners do distinguish between spontaneous speech and essay-writing, which could be viewed as a welcome contradiction to Kieweg’s (2000) description of spoken learner language as orally performed writing (cf. Section 1). As there are substantial differences between spoken and written learner language, it is thus . Note that collocations in Nesselhauf ’s (2004a) study thus form a category in between ‘free combinations’ (e.g. want a car) and fixed ‘idioms’ (e.g. kick the bucket). . Note that Vorherr (2006) takes as her starting point the list of verb-noun collocations in Nesselhauf 's (2004a) study – she thus restricts herself to a subset of potentially many more verb-noun collocations in LINDSEI-Ger.


Table 2. Top ten verb-noun collocations in LINDSEI-Ger (cf. Vorherr 2006) Rank

Collocation

Number of occurrences

Normalised frequency

1 2 3 4 5 6 7 8 9 10

have time have a problem have a chance do work have an experience do a job find a job have a look at have a feeling make an experience

53 40 22 21 21 21 14 14 14 12

0.58 ptw 0.44 ptw 0.24 ptw 0.23 ptw 0.23 ptw 0.23 ptw 0.16 ptw 0.16 ptw 0.16 ptw 0.13 ptw

also useful to base the teaching of spoken and written collocations on speech-specific and writing-specific target norms. Thirdly, it is interesting that the topmost verb-noun collocations in LINDSEI-Ger are used much more frequently than the topmost verb-noun collocations in GeCLE. For example, while have time and have a problem occur with relative frequencies of 0.58 ptw and 0.44 ptw respectively in LINDSEI-Ger, the two most frequent verb-noun collocations in GeCLE, i.e. solve a problem and have a reason, only occur with frequencies of 0.16 ptw and 0.12 ptw respectively. Note, however, that Vorherr (2006) has also found that the overall number of all verb-noun collocations is considerably higher in GeCLE than in LINDSEI-Ger. The last two findings do not contradict each other, but complement each other and seem to tie in with what might be seen as a variation of Hasselgren’s (1994) ‘teddy-bear principle’ in learner language: the overall range (or variety) of collocations is lower in speech than in writing, but the smaller set of collocations in speech is used much more frequently. The fact that learners are more restricted in speech with regard to the range of collocations they have at their disposal is also corroborated by the limited number of verbs in Table 2: in six collocations have is used, in two more it is do. In writing, on the other hand, the range of verbs is much wider (cf. Table 1).

. This also applies to other, less frequent collocations that are not included in Tables 1 and 2. For example, Vorherr (2006) finds that not only high-frequency verb-noun collocations (e.g. commit a crime, raise a question) but also less frequently used collocations in GeCLE do not occur in LINDSEI-Ger at all (e.g. set an example, find a solution). Note in this context that De Cock (2005) in her study of lexical bundles (i.e. combinations of more than two words) in learner language also finds that the most frequent word combinations are used considerably more frequently in speech than in writing.


It is quite clear that quantitative findings obtained from the comparison of verb-noun collocations in speech and writing have to be taken with a measure of caution. The most significant caveat is the different nature of speech and writing per se. Due to the constraints of on-line speech production, spoken language tends to be more formulaic and less varied than writing, which is usually edited and proof-read. It might well be the case, therefore, that the range of verb-noun collocations in speech is not as wide as in writing because of this formulaic and routinised nature of speech. A second caveat refers to the different degrees of ‘nouniness’ of speech and writing. De Haan (2001), for example, shows that while written language is more nominal (that is, characterised not only by many nouns, but also adjectives and prepositions), spoken language tends to be more clausal. If we compare verb-noun collocations in spoken and written language, we may thus find fewer verb-noun collocations in speech. Thirdly, it should also be noted that individual verb-noun collocations may be more frequent in GeCLE than in LINDSEI because they are more appropriate for the typical topics of argumentative essays, while others may be more suitable for interview situations. For example, it is not surprising that solve a problem is the most frequent verb-noun collocation in GeCLE (cf. Table 1) as it is particularly useful for argumentative essays. While the aformentioned caveats have to be kept in mind when dealing with overall quantitative comparisons of spoken and written learner language, such comparisons are nevertheless valuable because they may point to areas in learner language which are worth studying in more detail from a language-pedagogical perspective. A very relevant area is deviant collocations in learner language. Nesselhauf (2004a) finds that almost a third of all verb-noun collocations in GeCLE are unacceptable or at least questionable if compared with comparable native data and native speakers' intuitions. In LINDSEI-Ger, various weak spots in learners’ use of verb-noun collocations can be identified. For example, many verb-noun collocations involving the verb have seem to pose problems for German learners of English. The deviant form have difficulties occurs five times in LINDSEI-Ger, see Figure 1. The idiomatic form have difficulty is not attested at all in LINDSEI-Ger. The deviant form is presumably triggered by interference as in German the idiomatic phrase involves the plural form (cf. Germ. Schwierigkeiten haben). The concordance in Figure 1 shows that the deviant verb-noun collocation is either the only deviance in an otherwise correct flow of speech (e.g. in line 4, in which the noun is correctly complemented with in + ing-clause) or is combined with other inaccuracies (e.g. in line 3, in which difficulties is incorrectly premodified by a little bit). The same holds true for the deviant form have chances which ocurs three times in LINDSEI-Ger, see Figure 2.


1 2 3 4 5

th this Columbian girl . erm . I had most difficulties yes yes we’re ema m that was not too bad .. I had sometimes difficulties to understand .. ink we Germans today we have a little bit difficulties with our history epared dishes . it’s .. y= you might have difficulties in finding a real en th= pupils that d= that had real great difficulties erm . learning s=

Figure 1. Concordance for have difficulties in LINDSEI-Ger

1 2 3

found out that being a vet . had even less . chances for getting a job . xam as a Latin teacher I will have excellent chances there will be no pr ause: erm . well . I think I would have more chances . of getting a: mm

Figure 2. Concordance for have chances in LINDSEI-Ger

Again, interference may be at the heart of the deviant form have chances. What is more, the concordance in Figure 2 also corroborates Nesselhauf ’s (2004a: 103) observation that “many learners seem unaware of the different uses of chance and chances”. The LINDSEI-Ger data show not only deviations in the noun position but also in the verb position. Examples (1) to (3) illustrate wrong choices of the verb, with target hypotheses given in brackets. (1) I really seek to get a job where I really .. know the stuff I’m doing has some sense [→ makes sense] for er at least some people (2) like a little chil= er child in a school who should talk a story [→ tell a story] about some kind of er pictures (3) . yeah and erm . I made the experience [→ had the experience] erm to . be on my own for the first time

In the light of the various examples of verb-noun collocations in LINDSEI-Ger, it is obvious that this structure is interrelated with other areas of English lexicogrammar, for example noun premodification and noun complementation. Note that learners often tend to premodify the noun in verb-noun collocations with intensified adjectives. Examples (4) to (6) illustrate an overuse and/or a misuse of adjective intensification (given in italics) in the context of verb-noun collocations, which is in line with Lorenz’s (1998: 64) observation that “learners not only use more intensification, they also use it in places where it is semantically incompatible, communicatively unnecessary or syntactically undesirable”. (4) because the =a there you have a . quite . big choice afterwards . what to do . and


(5) you can . kind of collect points . during: the semester . by doing some little homework .. (6) pupils that d= that had real great difficulties erm . learning s= th= things

To the right of the noun of a verb-noun collocation, learners often use the to-infinitive (which is the structure that is usually used in equivalent contexts in German) instead of an ing-clause (with or without preposition). Examples (7) to (9) illustrate the overuse of the to-infinitive after verb-noun collocations. (7) that was not too bad .. I had sometimes difficulties to understand valenciano which (8) er .. yeah .. erm .. yeah . erm I . had . few problems to understand the Irish . they’ve . quit (9) clean in er Iceland er you don’t have er any problems to drink er . er eh eh the water di= direct

The results of Vorherr’s (2006) small-scale study of verb-noun collocations in LINDSEI-Ger on the basis of Nesselhauf ’s (2004a) observations in GeCLE make it clear that learner corpora provide a comprehensive and detailed picture of learners’ actual usage both quantitatively and qualitatively. It is also possible to systematically compare learners’ speech and writing. Learner corpora thus have a lot to offer, but so far it seems that it is mainly applied linguists and corpus linguists who have been profiting from the advent of learner corpora and who have added this new kind of resource to their linguistic toolkit. As Seidlhofer (2002), Granger (2004), Mukherjee and Rohrbach (2006) and others have noted, more interaction between corpus linguists, second language acquisition researchers and language teaching professionals including language teachers – with their language-pedagogical feet on the ground, so to speak – is needed. Learner corpus analyses always, at least implicitly, raise the question of what the language-pedagogical implications and applications might be. In the remainder of this section, I would like to address this question with regard to verb-noun collocations. An obvious way of improving learners’ collocational competence is by systematically including high-frequency verb-noun collocations in classroom materials, e.g. conversations in ELT textbooks. While, as Mauranen (2004: 90) notes, prefabricated patterns have received much attention in communicative language teaching, “the actual expressions taught as typical and useful rarely have attested origins”. A comprehensive description of verb-noun collocations in native corpora and in advanced learners’ output makes it possible to improve textbook materials by paying more attention to forms and structures that seem to pose particular


problems for learners (e.g. the construction have difficulty (in) V-ing). Various suggestions have been made as to how concordance output can also be used for designing fill-in-the-gap exercises (cf. e.g. Granger and Tribble 1998): native corpus data (e.g. selected concordance lines taken from the British National Corpus) lend themselves to such data-driven learning (DDL) exercises in which learners work with concordance output themselves. In the light of the availability of learner corpora, the question arises whether learners should also be confronted with learner corpus data in DDL scenarios, as suggested, for example, by Granger and Tribble (1998), Flowerdew (2001), Mukherjee and Rohrbach (2006). Nesselhauf (2004b) makes the following suggestion: Data-driven learning with learner data is probably particularly useful for points which have already been covered in the classroom, possibly even repeatedly, but which the learners nevertheless still get wrong. In this way, instead of being told once again that what they are doing wrong, learners have the opportunity to get something right, namely to identify and explain the mistake in question. (Nesselhauf 2004b: 140)

In the learner corpus context, DDL activities would make use of corpus data as negative evidence. Nesselhauf (2004b: 140) makes it clear that negative evidence needs to be complemented by positive evidence and that follow-up exercises are needed to stabilise the correct usage. In such a scenario, data from LINDSEI-Ger could be used, for example, for the teaching of verb-noun collocations, e.g. the use of have + experience instead of make + experience. While positive evidence is usually taken from native corpora, a variant of this procedure is the use of positive evidence from the learner corpus itself (if there is positive evidence in the learner corpus, that is). Psychologically, learners react much more positively to the use of positive evidence obtained from learner data because they do not get the impression of learner output being treated exclusively as a hotch-potch of mistakes and errors – it is neither desirable nor useful to establish a rigid dichotomy between good and correct usage in native data on the one hand and bad and incorrect usage in learner output on the other. Figure 3 provides negative evidence of make +

. Picking up on Seidlhofer (2002), Mukherjee and Rohrbach (2006) also develop ideas as to how teachers may compile and analyse ‘local’ learner corpora, including data produced by their own learners (also in the form of longitudinal monitor learner corpora). However, while this is a feasible and fruitful option for written material (e.g. student essays that have been produced electronically), the compilation of spoken local learner corpora seems unrealistic, given the resources needed for the recording and the transcription and the limitations of the classrom context.


1 2 3 4 5

. wouldn’t want to miss the . experiences I’ve made and . you want me making lot of lot of lots of experiences . yeah and erm . I made the experience erm to . be on my own for the ust f= er .. er made the same experience . erm .. yeah it’s quite hard it’s: I’m gonna make my own experiences but . then . I found out well

Figure 3. Make + experience in LINDSEI-Ger: negative evidence 1 2 3 4 5

eah erm I want to talk about an experience I had in a foreign country so . it was a really nice erm . experience . I had and . what I found don’t know if you ever had that experience but sometimes . when you e cares about time . I had this experience . erm .. mm she wo he experience I didn’t have the experience and I was not a carpenter

Figure 4. Have + experience in LINDSEI-Ger: positive evidence

evidence, which could be fruitfully combined with the positive evidence of have + evidence included in Figure 4. All the data are obtained from LINDSEI-Ger. The DDL scenario envisaged here thus consists of two phases: (a) Raising of language awareness by analysing and discussing negative and positive evidence (e.g. Figures 3 and 4); (b) stabilising the correct usage through follow-up exercises. Such DDL exercises could take the form of concordance lines (for example, from the British National Corpus) with gaps in which the verb-noun collocation has to be completed by the learner – ideally, such an exercise would include various verb-noun collocations. It is quite clear that such a DDL scenario can only work if the teacher carefully prepares the two phases and chooses appropriate material for classroom discussion and DDL exercises. From a language-pedagogical point of view, verb-noun collocations should be seen as a paradigmatic example of collocations in general. Not only the teaching of spoken grammar, but also the teaching of collocations has so far been widely neglected in the ELT classroom. The suggestions in this section show that learner corpora may help to open up new perspectives in this context. But as in many other areas of applied corpus linguistics, what is needed is much more classroom action research into whether and to what extent learners really profit from the use of learner corpora.

2.2 The discourse marker you know: The corpus as a whole vs. individual learners Even before LINDSEI-Ger was compiled, advanced spoken learner language had been analysed from a pragmatic perspective at the University of Giessen. In the


1990s, the Giessen Long Beach Chaplin Corpus (GLBCC) was compiled, which includes – apart from a native-speaker component – more than 90,000 words of spoken English produced by 77 German learners of English as a foreign language (cf. Jucker et al. 2003). The design of the corpus is more limited in scope than LINDSEI-Ger because GLBCC is restricted to oral narratives and conversations between two students at a time, all of which are about the Chaplin silent movie The Immigrant. On the basis of GLBCC, Müller (2005) has analysed in detail the use of the discourse markers well, so, like and you know in advanced spoken learner English. Discourse markers are a particularly relevant area for the teaching of spoken grammar because, as Müller (2005) shows, they occur very frequently in authentic native-like usage and fulfil a wide range of communicative functions. What is more, the GLBCC data reveal that there is a considerable gap between native-like usage and advanced learner language with regard to the forms, frequencies and functions of discourse markers in speech. As Hasselgren (2002) notes, there is a strong correlation between learners’ discourse-marker competence and their overall fluency in speech. It seems, therefore, that it is important for learners to acquire discourse-marker competence because it enables them to perform much more naturally and with a much higher level of ‘spoken fluency’ (cf. Section 1) in spontaneous conversations. Hasselgren (2002) and Müller (2005) analyse learner corpora in their entireties, thus abstracting away an overall supraindividual trend across individual learners, which is taken to be characteristic – if not to say, representative – of a specific stage of learners’ progression. On the basis of selected data from LINDSEI-Ger, Mukherjee and Rohrbach (2006) show that it is also very useful from a language-pedagogical perspective to zoom in on individual learners’ output because the corpus-as-a-whole average may at times mask an amazing spectrum of individual competencies across the learners in a learner corpus. For language teaching, it is also necessary to ‘individualise’, as it were, the corpus analysis: In this context, individualisation of the analysis is intended to mean that for the purpose of individual assessment and analysis of the variation between learners, it would be useful to complement the learner-language-as-whole perspective by also taking the differences between learners into account. (Mukherjee and Rohrbach 2006: 217)

Let us, by way of exemplification, consider the discourse marker you know. In LINDSEI-Ger, you know is used as a discoure marker 86 times. As Table 3 shows, you know thus occurs 0.95 times per thousand words in LINDSEI-Ger. Table 3 also gives some comparable data from two different sources, namely the native . Hasselgren (2002) uses the term smallword for discourse marker.


Table 3. You know as discourse marker: Frequencies in learner language and native data Advanced spoken learner English (LINDSEI-Ger)

American English American English component of GLBCC conversation (cf. Müller 2005) (cf. Biber et al. 1999)

British English conversation (cf. Biber et al. 1999)

c. 0.95 ptw

c. 3.70 ptw

c. 2.00 ptw

c. 4.50 ptw

Table 4. You know as discourse marker: Frequencies across learners in LINDSEI-Ger Learner (No.) in LINDSEI-Ger

Instances of you know as discourse marker

Row total

022 026 027 001; 013; 029; 032 002; 015; 049 006; 009; 010; 023; 031 007; 011; 014; 021; 033; 040; 042; 043; 047; 048 remaining 25 speakers Sum total

24 10 7 4 each 3 each 2 each 1 each 0

24 10 7 16 9 10 10 0 86

American corpus used as a yardstick for comparison by Müller (2005) and the rough estimates given for American English and British English conversations in the Longman Grammar (Biber et al. 1999). The data in Table 3 suggest that at the corpus-as-a-whole level, learners underuse you know as a discourse marker. While this is undoubtedly true, it is also insightful to look at the range of discourse-marker competencies across the individual learners in LINDSEI-Ger both quantitatively and qualitatively. Table 4 gives a comprehensive overview of the number of instances of you know in the 50 individual learners’ output. Table 4 shows that the 86 instances of you know as a discourse marker are extremely unevenly distributed across LINDSEI-Ger. While a few learners (especially 022, 026, 027) use it as frequently as native speakers, half of all the learners (i.e. 25 learners) do not use you know as a discourse marker at all. The figures illustrate that the fiction of homogeneity that is often associated with the compilation of a learner corpus according to well-defined standards and design criteria may run counter to the wide range of differing individual levels of competence in the corpus. From a qualitative perspective, it is also very useful to individualise learner corpus analyses because the different extents to which you know is used as a discourse marker can be viewed as a signpost of the overall spoken fluency of learn-


ers. Two extreme examples should suffice to illustrate the link between discourse marker competence and overall spoken fluency. Example (10) is taken from the LINDSEI-Ger interview with learner 022 (speaker B). (10) you know you can . I mean kind of there’s this running gag that you know give a medic er the telephone [directory [mhm and he will . learn it [by heart [ by heart <end laughter> [yeah [ no now I know i=why I didn’t do medicine <end laughter> [and [ yes <end laughter> . and and this is this kind of you know . erm . these are the people . they want to have mhm you know the= they they have to be young mhm they have to be willing . you know and then they give them six months of training in whatever [direction [mm mm . yeah and: and they are bright anyway . [you know [yeah otherwise they wouldn’t . be medics

While it is true that this passage is marked by a very frequent use – perhaps even an overuse – of you know, the overall impression one gains also when listening to the recording is a very positive one. The speech produced by this learner is very fluent, and discourse markers like you know are used effectively, for example when introducing a turn, planning what to say next or thinking about appropriate words or phrases. To some extent, the spoken fluency of speaker 022 depends on – and is created by – a natural use of discourse markers like you know. Example (11), on the other hand, is taken from the LINDSEI-Ger interview with learner 017 (speaker B), who belongs to the 25 learners that do not use you know as a discourse marker at all. (11) it’s not that bad <end laughter> would you say that erm . okay .. er would you say Britain’s intrinsically xenophobic no . [no no no [mm


.. I think Britain proves that in well in some parts of Britain erm . it can be proved that a multicultural society is . possible . mhm for ex= . for example in London . mhm I mean it all depends on the[i:] economy I think . erm because i= if you go up North and . see: the problems they are facing there with unemployment and everything . mhm of course there are: . pff riots an= an= and there . are fights between some . <slaps thighs> gangs but .. I think in in well in most of the parts . they proved that it’s possible an= [and [mhm they are not . xenophobic at all

The overall impression of the output of learner 017 is clearly less natural and less fluent than in the case of learner 022. The learner’s penultimate turn in example (11) nicely illustrates that discourse markers like you know are a very powerful element in learners’ spoken repertoire: for example, you know would be a natural choice at positions in the stream of speech where the learner uses pff as a filler, repeats an=/and and where he slaps his thighs. These are all positions at which speech-planning processes take place, and in an authentic conversation in which the turnholder’s position is at constant risk it would be useful to use natural turnholding markers such as you know. From a language-pedagogical perspective it is necessary, firstly, to make learners aware of the natural use of discourse markers in spontaneous spoken language, and, secondly, to automatise, as it were, the use of discourse markers. DDL scenarios lend themselves to raising learners’ awareness of the natural contexts in which discourse markers like you know are used. Again, while native data seem to be the obvious choice, it might also be worth opting for the analysis of learner output, not only for psychological reasons (cf. Section 2.2), but also because the forms produced by advanced learners themselves appear to be cognitively more available to – and typical of – other learners at a certain stage in the learning process: [A]dvanced learners are not defective native speakers cleaning up a smattering of random errors, but rather learners working through identifiable acquisition (Cobb 2003: 419) sequences.

Figure 5 presents an excerpt from the concordance for you know in LINDSEI-Ger which learners could produce themselves and use to find out more about some of the typical contexts and functions of you know.


Figure 5. Excerpt from a concordance for you know in LINDSEI-Ger (WordSmith Tools, Version 4.0, Scott 2005)

From the concordance for you know, learners should be able to draw some conclusions as to the use of you know (which would have to be confirmed by the teacher), for example: – you know can be used at virtually all positions when you hesitate in the flow of speech; – you know also occurs at the beginning of a clause; – you know is used when you are thinking about the next words to utter (e.g. line 24); – you know can be used if you want to indicate that the next words are perhaps not very precise (e.g. line 25); – you know can be used when what you said before is not entirely precise (e.g. line 35); – you know is used when you want to explain something that you said in more detail (e.g. line 31). Once learners have found out about natural contexts and typical functions of you know (on grounds of learner data and/or native data) and described them in straightforwardly clear and easily accessible terms, it is for classroom interaction to provide a framework for actually using and practising the use of you know. This can be done at two different levels: the strategic level and the formal level.


At the strategic level, a very effective way would be to make use of the dialogue techniques that Kieweg and Kieweg (2000) have suggested for the ELT classroom. These dialogue techniques are designed in such a way that they increase the spontaneity and length of students’ verbal reactions and force them to produce speech under real-time on-line production constraints. One of the many techniques sketched out by Kieweg and Kieweg (2000) is the explication of new and unfamiliar games (including computer games), computer programs, technical equipment and any other tools and gadgets for which students lack the precise technical vocabulary. The idea here is that learners are forced to circumscribe the component parts or functions of such tools by using less precise vocabulary, which they should indicate by you know as an approximator. This could be combined with the regulation that longer unfilled pauses in speech are not allowed – they would have to be replaced, say, by fillers or discourse markers. Dialogue techniques of this kind provide a communicative platform for practising discourse markers in classroom contexts. In a wider setting, the task-based approach to language learning provides a very suitable framework for practising fluent and natural language use in general and the use of discourse markers in particular. Nunan (1989) defines a ‘task’ as a piece of classroom work which involves learners in comprehending, manipulating, producing or interacting in the target language while their attention is (Nunan 1989: 10) principally focused on meaning rather than form.

Kieweg and Kieweg’s (2000) suggestions should be viewed as some examples of how to design communicative tasks to make learners use the target language spontaneously and fluently. At the formal level, you know can be learned as part of larger chunks and formulaic sequences, e.g. ...you know because..., ...because you know..., the point is you know..., the thing is you know..., it’s you know..., this is a difficult question you know. Learning you know as part of such larger bundles that are frequently used in spontaneous speech ensures that learners use the discourse marker at least to some extent so that it is automatised and thus also more easily available in other contexts of hesitation, speech planning and approximation.

2.3 Performance phenomena in speech: Learner data vs. native data In this section I will put into perspective forms and structures that are typical of spontaneous spoken language but which have just begun to attract attention in learner corpus linguistics and language pedagogy, namely performance phenomena such as repetitions and other disfluencies (cf. Rühlemann 2006). Götz (2006)


Table 5. Repetitions of personal pronouns in LINDSEI-Ger and LOCNEC (cf. Götz 2006) Pronoun

LINDSEI-Ger Number of Normalised occurrences frequency

LOCNEC Number of Normalised occurrences frequency

I we you he she it they

168 16 30 15 9 30 42

187 23 46 12 8 73 25

1.9 ptw 0.2 ptw 0.3 ptw 0.2 ptw 0.1 ptw 0.3 ptw 0.5 ptw

1.6 ptw 0.2 ptw 0.4 ptw 0.1 ptw 0.1 ptw 0.6 ptw 0.2 ptw

has conducted a pilot study of some performance phenomena in LINDSEI-Ger and compared them to native-speaker data in the Louvain Corpus of Native English Conversation (LOCNEC). In this section, I will sketch out some of the findings in this study and their language-pedagogical implications. The focus of Götz’s (2006) study is on features that are responsible for the overall impression of fluency in speech. Among the wide range of phenomena that Götz (2006) analyses in LINDSEI-Ger are repetitions, including repetitions of pronouns. Table 5 gives the number of repetitions of personal pronouns in LINDSEI-Ger and LOCNEC. Table 5 displays no drastically different trends between the learner data and the native data, although some of the pronoun repetitions seem to be underused or overused in learner speech from a strictly quantitative perspective. However, when it comes to performance phenomena such as repetitions, the languagepedagogical aim cannot be the imitation of the frequencies that we find in native data. Rather, we need to abstract away from frequencies and address the question whether learners use stretches of speech that are typical of native speech in the same contexts and for the same communicative functions. From such a qualitative point of view, quantitative data such as those in Table 5 can only provide a starting point for a detailed analysis of repetitions in actual contexts and a functionally-oriented comparison of learner speech and native speech. The most frequent repetitions by far of the nominative pronoun I, occurring with similar frequencies in LINDSEI-Ger and LOCNEC, are a case in point. Biber et al. (1999: 1058) argue . The 118,000-word LOCNEC is not publicly available and only used internally by the Centre of English Corpus Linguistics at the University of Louvain and cooperating institutions, including the Department of English of the University of Giessen. A summary of the major findings on which Götz (2006) reports is provided by Götz (2007).


that in native speech repetitions of I “occur almost invariably at the beginning of a clause – often at the beginning of a turn – where the build-up of planning pressure on the speaker is likely to be great”, giving the following example: (12) Except of course, I er, I, I couldn’t read my road map there in, in Brussels. (BrE) (Biber et al. 1999: 1058)

In LOCNEC, repetitions of I fulfil this very function of “relieving that planning pressure” (Biber et al. 1999: 1058) in most of the cases. As examples (13) to (15) show, the repetitions are used at the beginning of a clause while the speaker is planning the exact wording. Note that in each example, the clause following the repeat is produced without any further hesitation phenomena. (13) but when I’m home I I do miss it now and then

(LOCNEC)

(14) after seeing her last week I I don't need to ask that question any more cos (LOCNEC) (15) I I saw him in an interview with a vampire and thought he was great in that (LOCNEC)

From the LINDSEI-Ger data, on the other hand, a slightly different picture emerges. It seems that the repeats at the beginning of a turn or a major syntactic unit are not sufficient for learners’ speech planning because in many cases the clause following the initial repeat is marked by additional disfluencies. Illustrative examples are given in (16) to (18). (16) the customer . or . the wife . I I guessed whoe= wherever she is erm . she (LINDSEI-Ger) (17) I I just find that really . erm . distracting .

(LINDSEI-Ger)

(18) I . I really fought the [i:] idea of having yeah . erm . . it but . (LINDSEI-Ger)

It seems that in learner speech planning pressure does not decrease after the beginning of a major syntactic constituent but remains high throughout. This is corroborated by the overall frequencies of co-occurrences of repetitions of I at the beginning and further hesitations to the right: in LOCNEC 26% of clauses with initial repetitions of I also include additional hesitation markers (e.g. more repetitions, filled/unfilled pauses, self-corrections), whereas the percentage is as high as 58% in LINDSEI-Ger. This is indicative of the fact that advanced learners, unlike native speakers, have to consider much more thoroughly not only what they want to say but also how they should say it even within syntactic constituents (cf. Fulcher 1996). It is unlikely that this problem can be solved entirely in foreign language teaching, so our attention should turn to ways of dealing with persistent


planning pressure. A very effective solution ensuring a high level of fluency in learner speech might be the use of discourse markers like you know or like within syntactic constituents (cf. Section 2.2). Function words such as pronouns may also be followed by filled or unfilled pauses if and when the speaker needs time for speech planning. Götz (2006) finds that after all personal pronouns – and many other function words such as conjunctions – unfilled pauses occur significantly more frequently in LINDSEI-Ger than in LOCNEC. She also shows that in many contexts learners tend to use filled pauses after pronouns whereas native speakers tend to prefer a repetition of the pronoun itself. Also, learners use filled pauses after conjunctions (such as and and because), after prepositions (such as with) and after the auxiliary verb forms is, was, were and can three to four times as frequently as native speakers. These observations not only corroborate the greater planning pressure of learners in speech throughout syntactic units, but the findings also explain why spontaneous non-native speech often appears to be less natural and less fluent than native speech. One of the major reasons for the overall difference in fluency between native speakers and learners appears to be the smaller set of (semi-)preconstructed and holistic phrases that even advanced learners have at their disposal. It is therefore necessary not only to teach the use of discourse markers more systematically to increase the naturalness and fluency of learners’ speech, but also to overcome the still prevailing “strict divide between grammar practice activities and vocabulary practice activities” (Müller-Hartmann and Schocker-von Ditfurth 2004: 66) in the EFL classroom and promote the use of formulaic language. As Levis (2006: 265) notes, a fluent speaker “is one who has access to a stock of verbal idioms, phrases that do not need to be consistently reconstructed” – if learners of English are to become genuinely fluent speakers they perhaps need even more formulaic sequences than native speakers because of the greater planning pressure on non-native speakers. As a third area in which advanced learners deviate clearly from the native target norm I would like to mention repetitions of subject-verb contractions, i.e. repetitions of forms such as I’m, we’re, she’s, it’ll, they’ve and I’d, as exemplified in (19) and (20). (19) er yeah I’m I’m in the Choral Society (LOCNEC) (20) sort of you know they’ve they’ve been sort of invaded from different directions (LOCNEC)

First of all, Table 6 gives the overall frequencies of non-repeated co-occurrences of pronouns and verb contractions in LINDSEI-Ger and LOCNEC for four groups: (1) pronoun + ’m/’re/’s, (2) pronoun + ’ll, (3) pronoun + ’ve, (4) pronoun


Table 6. Subject-verb contractions in LINDSEI-Ger and LOCNEC (cf. Götz 2006) Group

LINDSEI-Ger

LOCNEC

pronoun + ’m/’re/’s pronoun + ’ll pronoun + ’ve pronoun + ’d

17.5 ptw 0.5 ptw 2.4 ptw 0.8 ptw

23.8 ptw 1.4 ptw 5.8 ptw 3.2 ptw

Table 7. Repeated subject-verb contractions in LINDSEI-Ger and LOCNEC (cf. Götz 2006) Repeated subject-verb contraction

LINDSEI-Ger LOCNEC Percentage of all occurrences Percentage of all occurrences

I’m we’re you’re he’s she’s it’s they’re I’ll other pronouns + ’ll I’ve you’ve they’ve other pronouns + ’ve I’d we’d you’d they’d other pronouns + ’d

3.9% – 1.3% 4.2% 1.8% 6.1% – 3.3% – 1.9% – 8.3% – 1.7% – – – –

5.0% 1.7% 1.8% 6.7% 3.4% 6.6% 2.0% 7.3% – 5.4% 1.0% 1.7% – 5.7% 7.1% 4.2% 2.4% –

+ ’d. The figures show that subject-verb contractions are generally underused by learners of English. Table 7 zooms in on the four groups of subject-verb contractions listed in Table 6 and gives the frequencies of repetitions of several manifestations of the four groups (as percentages of the non-repeated instances) in LINDSEI-Ger and LOCNEC. Table 7 shows that learners in LINDSEI-Ger – compared to the native speakers in LOCNEC – use considerably fewer repetitions of subject-verb contractions. Tables 6 and 7 tie in with the general observation that genuinely spoken features such as contracted forms are notoriously underrepresented in


learners’ speech, which seems to be more oriented towards written grammar (cf. Section 1).10 The underuse of subject-verb contractions in general and the absence of repetitions of some contracted forms in particular clearly render learner speech less natural than native speech. The language-pedagogical conclusions to be drawn from corpus-based analyses of performance phenomena in learner speech are two-fold. Firstly, there are certainly forms and structures that are so typical of natural speech that it is necessary to pay much more attention to teaching and learning them in the EFL classroom. This includes not only discourse markers but also idiomatic and (semi-) preconstructed phrases that may help learners to be more fluent, even though planning pressure is higher than for native speakers. Psycholinguistic research into formulaic language (cf. e.g. Wray 2002; Schmitt 2004) and corpus-based work on phraseological chunks (cf. e.g. Granger and Meunier 2008; Meunier and Granger 2008) have already provided many significant results and insights into language-pedagogical applications and implications, e.g. with regard to materials design. Secondly, an area that seems to have been neglected so far is teacher education. While it is true that teaching materials may well profit from corpus-based descriptions of spoken language (e.g. by including more natural conversations in ELT textbooks), many performance phenomena such as repetitions and filled pauses cannot be expected to be systematically included in, say, written teaching material. For spontaneous spoken language, it is the teacher that provides the most important target model for learners of English in institutionalised classroom contexts: Teachers are models. They need to feel comfortable using the target language as a natural means of communication [...]. Teachers need to be able to present language as naturalistic examples of the target language, to expose learners to examples of language currently in use, with features which are characteristic of authentic discourse in the target language. (Müller-Hartmann and Schocker-von Ditfurth 2004: 28)

It is a truism that fluency in learner speech can only be achieved if the teacher model provides natural and fluent input. It is therefore vital for corpus-based insights into the nature of spoken language to play a much greater role in teacher education. In Germany, for example, teacher edcuation is still strongly influenced by grammar, writing and translation courses. Practising the grammar of conversation, by contrast, is very often not considered to be essential. Therefore, as in most other language-pedagogically relevant areas, it will be difficult to change 10. Note in this context that Götz (2006) also shows that the non-contracted equivalents of the forms listed in Tables 6 and 7 are overused by learners in LINDSEI-Ger.


classroom reality – and what the language students learn in EFL teaching – without a substantial shift in focus in teacher education.

3.

Concluding remarks

Firstly, I hope to have shown in the present paper that learner corpora in general and LINDSEI-Ger in particular are valuable resources for the description of the grammar of conversation in advanced learners’ speech from three different perspectives, as the three case studies in Section 2 have shown: a. With the advent of written and spoken learner corpora, it is possible to compare written and spoken grammar systematically, for example with regard to quantitative and qualitative differences in the use of verb-noun collocations. b. Learner corpora not only provide new insights into the abstract and supraindividual average of learner language at a given stage in the learning process (in terms of interlanguage uniformity), but they also shed light on the range of individual learner competencies within that very stage (in terms of interlanguage variation). For example, there is a wide range of discourse-marker competencies in LINDSEI-Ger. c. Last but not least, learner corpora can be compared with native data so that the gap between learner output and the native target norm can be described, e.g. with regard to the forms, frequencies and functions of performance phenomena such as repetitions. Secondly, the three case studies have illustrated the relevance of learner corpus data to applied (corpus) linguistics and also to the teaching and learning of English as a foreign language. The various language-pedagogical suggestions with regard to the grammar of conversation in the classroom context bear testimony to this. Thirdly, it has become obvious that much more research into the grammar of conversation in advanced learners’ speech – from all three perspectives mentioned above – is needed. The same holds true for syllabus design in teacher education and pedagogically oriented classroom action research into how to teach spoken grammar. For a more systematic and detailed study of the grammar of conversation to be feasible and successful, it is necessary for future research to keep corpus annotation on its agenda (cf. de Haan 2000; de Mönnink 2000). At present, a part-of-speech (PoS) tagged (let alone syntactically parsed) version of LINDSEIGer is not available. However, PoS tagging and syntactic parsing information is particularly useful for analyses of syntactic structures, e.g. for automatic searches


for verb-complementational patterns, left dislocations, repetitions and sentence fragments. A third relevant level of annotation is error tagging, i.e. the identification and categorisation of errors and the indication of corresponding target hypotheses. A very promising model seems to be the concept of a multi-level standoff annotation that Lüdeling et al. (2005) have developed for a learner corpus of German, FALKO (Germ. Fehlerannotiertes Lernerkorpus). The multi-level standoff architecture that they envisage is based on Extensible Markup Language for Discourse Annotation (EXMARaLDA, cf. Schmidt 2004). The plain learner texts and the annotations are stored separately, and the annotation is open to an infinite number of annotation levels, thus allowing, for example, for the coding of various target hypotheses for one and the same error in a learner corpus. It is to be hoped that learner corpora of English, including LINDSEI-Ger, will at some point be available in an enriched version with multi-level annotations of this kind.

References Biber, D., Johansson, S., Leech, G., Conrad, S. & Finegan, E. 1999. Longman Grammar of Spoken and Written English. Harlow: Pearson Education. Brand, C. & Kämmerer, S. 2006. The Louvain International Database of Spoken English Interlanguage (LINDSEI): Compiling the German component, In Corpus Technology and Language Pedagogy: New Resources, New Tools, New Methods, S. Braun, K. Kohn & J. Mukherjee (eds), 127–140. Frankfurt: Lang. Cobb, T. 2003. Analyzing late interlanguage with learner corpora: Québec replications of three European studies, The Canadian Modern Language Review 59(3): 393–423. Conrad, S. 2000. Will corpus linguistics revolutionize grammar teaching in the 21st century? Paper presented at the Second North American Symposium on Corpora and Language Teaching, Flagstaff AZ, 31 March – 2 April 2000. Council of Europe. 2001. Common European Framework of Reference for Languages: Learning, Teaching, Assessment. Cambridge: CUP. Cowie, A. P. (ed.). 1998. Phraseology: Theory, Analysis, and Applications. Oxford: OUP. Cutting, J. 2006. Spoken grammar: Vague language and EAP. In Spoken English, TESOL and Applied Linguistics, R. Hughes (ed.), 159–181. Houndmills: Palgrave Macmillan. De Cock, S. 2005. The routine aspects of speech and writing: A comparative corpus-driven study of native and non-native performance data. Paper presented at Phraseology 2005: The many Faces of Phraseology – an Interdisciplinary Conference, Louvain-la-Neuve, 13–15 October 2005. De Cock, S., Granger, S. & Petch-Tyson, S. 2006. The Louvain International Database of Spoken English Interlanguage – LINDSEI. Available at , accessed 10 June 2006. De Haan, P. 2000. Tagging non-native English with the TOSCA-ICLE tagger. In Corpus Linguistics and Linguistic Theory, C. Mair & M. Hundt (eds), 69–79. Amsterdam: Rodopi.


De Haan, P. 2001. Aspects of the syntax of spoken English. In A Wealth of English: Studies in Honour of Göran Kjellmer, K. Aijmer (ed.), 47–55. Göteborg: Acta Universitatis Gothoburgensis. De Mönnink, I. 2000. Parsing a learner corpus? In Corpus Linguistics and Linguistic Theory, C. Mair & M. Hundt (eds), 81–90. Amsterdam: Rodopi. Flowerdew, L. 2001. The exploitation of small learner corpora in EAP materials design. In Small Corpus Studies and ELT: Theory and Practice, M. Ghadessy, A. Henry & R. L. Roseberry (eds), 363–379. Amsterdam: John Benjamins. Fulcher, G. 1996. Does thick description lead to smart tests? A data-based approach to rating scale construction. Language Testing 13(2): 208–238. Götz, S. 2006. Fortgeschrittene englische Lernersprache: Eine korpusbasierte Studie zu ausgewählten Merkmalen gesprochener Morphosyntax. MA thesis, University of Giessen. Götz, S. 2007. Performanzphänomene in gesprochenem Lernerenglisch, Zeitschrift für Fremdsprachenforschung 18(1): 67–84. Granger, S. 1998. The computer learner corpus: A versatile new source of data for SLA research. In Learner English on Computer, S. Granger (ed.), 3–18. London: Longman. Granger, S. 2004. Computer learner corpus research: Current state and future prospects. In Applied Corpus Linguistics: A Multidimensional Perspective, U. Connor & T. Upton (eds), 123–145. Amsterdam: Rodopi. Granger, S. 2006. A plurilingual and multicultural perspective on learner corpus research. Paper presented at Multiligualism and Applied Comparative Linguistics (MACL), Brussels, 8–10 February 2006. Granger, S., Dagneaux, E. & Meunier, F. (eds). 2002. International Corpus of Learner English. Louvain-la-Neuve: Presses Universitaires de Louvain. Granger, S. & Tribble, C. 1998. Learner corpus data in the foreign language classroom: Formfocused instruction and data-driven learning. In Learner English on Computer, S. Granger (ed.), 199–209. London: Longman. Granger, S. & Meunier, F. (eds). 2008. Phraseology: An Interdisciplinary Perspective. Amsterdam: John Benjamins. Halliday, M. A. K. 1987. Spoken and written modes of meaning. In Comprehending Oral and Written Language, R. Horowitz & S. J. Samuels (eds), 55–82. San Diego CA: Academic Press. Hasselgren, A. 1994. Lexical teddy-bears and advanced learners: A study into the ways Norwegian students cope with English vocabulary. International Journal of Applied Linguistics 4(2): 237–260. Hasselgren, A. 2002. Learner corpora and language testing: smallwords as markers of learner fluency. In Computer Learner Corpora, Second Language Acquisition and Foreign Language Teaching, S. Granger, J. Hung & S. Petch-Tyson (eds), 143–173. Amsterdam: John Benjamins. Hoey, M. 1991. Patterns of Lexis in Text. Oxford: OUP. Howarth, P. 1998. Phraseology and second language proficiency. Applied Linguistics 19(1): 24–44. Hunston, S. 2002. Corpora in Applied Linguistics. Cambridge: Cambridge University Press. Jenkins, J. 2004 ELF at the gate: The position of English as a Lingua Franca. The European English Messenger 13(2): 63–69. Jucker, A. H., Smith, S. W. & Lüdge, T. 2003. Interactive aspects of vagueness in conversation. Journal of Pragmatics 35: 1737–1769.


Kieweg, W. 2000. Zur Mündlichkeit im Englischunterricht. Der Fremdsprachliche Unterricht Englisch 34(5): 4–9. Kieweg, M. & Kieweg, W. 2000. Praxiserprobte Dialogtechniken. Der Fremdsprachliche Unterricht Englisch 34(5): 17–24. Leech, G. 2000. Grammars of spoken English. Language Learning and Technology 50(4): 675– 724. Levis, J. M. 2006. Pronunciation and the assessment of spoken language. In Spoken English, TESOL and Applied Linguistics, R. Hughes (ed.), 245–270. Houndmills: Palgrave Macmillan. Lorenz, G. 1998. Overstatement in advanced learners’ writing: stylistic aspects of adjective intensification. In Learner English on Computer, S. Granger (ed.), 53–66. London: Longman. Lorenz, G. 1999. Adjective Intensification – Learners versus Native Speakers: A Corpus Study of Argumentative Writing. Amsterdam: Rodopi. Lüdeling, A., Walter, M., Kroymann, E. & Adolphs, P. 2005. Multi-level error annotation in learner corpora, Proceedings from the Corpus Linguistics Conference Series 1(1). Available at , accessed 12 June 2006. Mauranen, A. 2004. Spoken corpus for an ordinary learner, In How to Use Corpora in Language Teaching, J. M. Sinclair (ed.), 89–105. Amsterdam: John Benjamins. Meunier, F. & Granger, S. (eds). 2008. Phraseology in Language Learning and Teaching. Amsterdam: John Benjamins. Müller, S. 2005. Discourse Markers in Native and Non-native English Discourse. Amsterdam: John Benjamins. Müller-Hartmann, A. & Schocker-von Ditfurth, M. 2004. Introduction to English Language Teaching. Stuttgart: Klett. Mukherjee, J. 2002. Korpuslinguistik und Englischunterricht: Eine Einführung. Frankfurt: Lang. Mukherjee, J. 2005. The native speaker is alive and kicking: Linguistic and language-pedagogical perspectives. Anglistik 16(2): 7–23. Mukherjee, J. & Rohrbach, J. 2006. Rethinking applied corpus linguistics from a languagepedagogical perspective: New departures in learner corpus research. In Planing, Gluing and Painting Corpora: Inside the Applied Corpus Linguist’s Workshop, B. Kettemann & G. Marko (eds), 205–232. Frankfurt: Lang. Nesselhauf, N. 2004a. Collocations in a Learner Corpus. Amsterdam: John Benjamins. Nesselhauf, N. 2004b. Learner corpora and their potential for language teaching. In How to Use Corpora in Language Teaching, J. Sinclair (ed.), 125–152. Amsterdam: John Benjamins. Nunan, D. 1989. Designing Tasks for the Communicative Classroom. Cambridge: CUP. Römer, U. 2005. Progressives, Patterns, Pedagogy: A Corpus-driven Approach to English Progressive Forms, Functions, Contexts and Didactics. Amsterdam: John Benjamins. Rühlemann, C. 2006. Coming to terms with conversational grammar: ‘dyslocation’ and ‘dysfluency’. International Journal of Corpus Linguistics 11(4): 385–409. Schmidt, T. 2004. EXMARaLDA: Ein System zur computergestützten Diskurstranskription. In Automatische Textanalyse: Systeme und Methoden zur Annotation und Analyse natürlichsprachiger Texte, A. Mehler & H. Lobin (eds), 203–218. Wiesbaden: Verlag für Sozialwissenschaften. Schmitt, N. (ed.). 2004. Formulaic Sequences: Acquisition, Processing and Use. Amsterdam: John Benjamins. Scott, M. 2005. WordSmith Tools, Version 4.0. Oxford: OUP.


Seidlhofer, B. 2002. Pedagogy and local learner corpora: working with learning-driven data. In Computer Learner Corpora, Second Language Acquisition and Foreign Language Teaching, S. Granger, J. Hung & S. Petch-Tyson (eds), 213–234. Amsterdam: John Benjamins. Vorherr, A. 2006. Collocations in Spoken English Learner Language: A Corpus-based Study. PhD dissertation, University of Giessen. Wray, A. 2002. Formulaic Language and the Lexicon. Cambridge: CUP.

Index

A Aboutness 6, 159, 176 Accuracy 23, 32 Adverbial 50, 125 Annotation 30, 227, 229 Authentic data 38, 77, 81, 180 Authenticity 83, 89, 182, 183, 186, 196, 197 B British National Corpus (BNC) 146 C Cleft construction 104, 110, 111 It-clefts 111, 116 Cognitive 30, 32 Collocate(s) 6, 27, 53, 56, 142–145, 160, 164–166, 208 Collocation 32, 44, 58, 87, 154, 176, 198 Collocates of collocates 160 Collocational profile(s) 6, 160 Comparative fallacy 14, 18, 25 Complementation pattern 6, 141, 146, 148 Complexity 25, 26, 32 Concgram 161–167, 173, 175 Contrastive interlanguage analysis (CIA) 18 Corpus of Academic Economics Texts (CÆT) 160 Corpus-driven 10, 44, 65, 67, 97, 154, 175, 176, 229 D Data-driven learning (DDL) 16, 37, 41, 68, 69, 91, 95, 174, 196, 197, 213 Design criteria 5, 14, 17, 216 Disciplinary discourses 152

Discourse marker 8, 207, 215 Discovery-learning 68 E e-learning 21, 32, 68 English as a Lingua Franca (ELF) 24 English for Specific Purposes 157, 175, 199, 200 Error 23, 29, 30, 44, 70, 72 Error analysis 23 Error-tagging 24 Evaluation 109, 177 Extraposition 119, 130 F Fluency 27, 32 Foreign language teaching (FLT) 2, 3, 13, 15, 16, 19, 34, 179, 180, 222 Frequency 29, 154 Raw frequency 143, 153 Relative frequency 141, 143, 146, 148, 153 G Grammar of conversation 7, 203, 204, 207, 225, 226 Grammar pattern(s) 64, 146, 149, 150 H Hesitation phenomena 204, 222 I Idiom principle 27, 38, 142 Implicit teaching methods 3, 33 Intercollocability 160

International Corpus of Learner English (ICLE) 5, 17, 122, 139 Interpersonal metaphor 105 Objective interpersonal metaphor 103, 108, 110, 115 Subjective interpersonal metaphor 108, 115 Interrogative clauses 107 K Key key words 159 Key words 159 KWIC concordance 50, 75, 76 L Language competence 83, 85 Learner corpus 13, 16, 19, 27, 30, 203, 212, 228 Learner needs 22, 23 Lemma 6, 75, 142, 146–148 Lexical bundle 142, 198 Lexical profiles 158 Linear Unit Grammar 162, 177 Louvain Corpus of Native English Conversation (LOCNEC) 221 Louvain International Database of Spoken English Interlanguage (LINDSEI) 7, 203, 227 M Markup 188, 227 Metalanguage 7, 107, 196, 197 Multi-word unit 143 N n-gram 27, 161, 175 NICLE 72, 121–139

232 Corpora and Language Teaching

O Oslo Interactive English (OIE) 67 P Pattern 44, 154, 176 Pedagogical tagging 189 Performance phenomena 220 Phraseology 30, 31, 119, 143, 153, 159, 175, 199, 227–229 Phraseological pattern 157–159, 162, 163, 166, 173 Popularization 95 Problem-based learning 77 Pseudo-clefts 104, 105 Basic pseudo-clefts 111 Reversed pseudo-clefts 111 Q Question 53, 54 R Reference corpus 6, 47, 123, 131, 157, 158 Reference tools 85, 89, 92

S Search interface 69, 75, 80, 81, 92, 93 Second-language acquisition (SLA) 11, 33, 39, 41 Selective attention 42 Self-reference 131 Semantic sequence 148 Skipgram 161 Special theme construction 128 Specialised corpus 157, 158, 163, 173, 174 Spoken grammar 203, 227 see also Grammar of conversation Stance marker 121, 133–135, 137, 138 Student participation 80, 81 Student-centred approach 81 SWICLE 102, 123, 124, 128, 133–135, 139 Systemic Functional Linguistics 104, 139 T Teachability 22, 23

Teacher’s needs analysis 83 Teaching materials 3, 6, 22, 48, 83, 85, 87–90, 93, 151, 174, 225 Teaching objectives 22, 23 TeMa Corpus 7 Theme 105, 106, 119, 120, 139 Experiential theme 103 Interpersonal theme 103 Multiple theme(s) 103, 106, 113 Textual theme 103 Thematic structure 5, 6, 121, 122, 124, 136 Transitivity 103 Trivial Corpus Pursuit 80, 81 V Vocabulary 70, 79, 81, 120, 139, 177, 190, 199 W Word association 167 Wordform 146 Word order 5, 122, 123

In the series Studies in Corpus Linguistics (SCL) the following titles have been published thus far or are scheduled for publication: 36 Quaglio, Paulo: Television Dialogue. The sitcom Friends vs. Natural Conversation. xiv, 161 pp. + index. Expected February 2009 35 Römer, Ute and Rainer Schulze (eds.): Exploring the Lexis–Grammar Interface. vi, 315 pp. + index. Expected March 2009 34 Friginal, Eric: The Language of Outsourced Call Centers. A corpus-based study of cross-cultural interaction. xxii, 317 pp. + index. Expected February 2009 33 Aijmer, Karin (ed.): Corpora and Language Teaching. 2009. viii, 232 pp. 32 Cheng, Winnie, Chris Greaves and Martin Warren: A Corpus-driven Study of Discourse Intonation. The Hong Kong Corpus of Spoken English (Prosodic). 2008. xi, 325 pp. (incl. CD-Rom). 31 Ädel, Annelie and Randi Reppen (eds.): Corpora and Discourse. The challenges of different settings. 2008. vi, 295 pp. 30 Adolphs, Svenja: Corpus and Context. Investigating pragmatic functions in spoken discourse. 2008. xi, 151 pp. 29 Flowerdew, Lynne: Corpus-based Analyses of the Problem–Solution Pattern. A phraseological approach. 2008. xi, 179 pp. 28 Biber, Douglas, Ulla Connor and Thomas A. Upton: Discourse on the Move. Using corpus analysis to describe discourse structure. 2007. xii, 290 pp. 27 Schneider, Stefan: Reduced Parenthetical Clauses as Mitigators. A corpus study of spoken French, Italian and Spanish. 2007. xiv, 237 pp. 26 Johansson, Stig: Seeing through Multilingual Corpora. On the use of corpora in contrastive studies. 2007. xxii, 355 pp. 25 Sinclair, John McH. and Anna Mauranen: Linear Unit Grammar. Integrating speech and writing. 2006. xxii, 185 pp. 24 Ädel, Annelie: Metadiscourse in L1 and L2 English. 2006. x, 243 pp. 23 Biber, Douglas: University Language. A corpus-based study of spoken and written registers. 2006. viii, 261 pp. 22 Scott, Mike and Christopher Tribble: Textual Patterns. Key words and corpus analysis in language education. 2006. x, 203 pp. 21 Gavioli, Laura: Exploring Corpora for ESP Learning. 2005. xi, 176 pp. 20 Mahlberg, Michaela: English General Nouns. A corpus theoretical approach. 2005. x, 206 pp. 19 Tognini-Bonelli, Elena and Gabriella Del Lungo Camiciotti (eds.): Strategies in Academic Discourse. 2005. xii, 212 pp. 18 Römer, Ute: Progressives, Patterns, Pedagogy. A corpus-driven approach to English progressive forms, functions, contexts and didactics. 2005. xiv + 328 pp. 17 Aston, Guy, Silvia Bernardini and Dominic Stewart (eds.): Corpora and Language Learners. 2004. vi, 312 pp. 16 Connor, Ulla and Thomas A. Upton (eds.): Discourse in the Professions. Perspectives from corpus linguistics. 2004. vi, 334 pp. 15 Cresti, Emanuela and Massimo Moneglia (eds.): C-ORAL-ROM. Integrated Reference Corpora for Spoken Romance Languages. 2005. xviii, 304 pp. (incl. DVD). 14 Nesselhauf, Nadja: Collocations in a Learner Corpus. 2005. xii, 332 pp. 13 Lindquist, Hans and Christian Mair (eds.): Corpus Approaches to Grammaticalization in English. 2004. xiv, 265 pp. 12 Sinclair, John McH. (ed.): How to Use Corpora in Language Teaching. 2004. viii, 308 pp. 11 Barnbrook, Geoff: Defining Language. A local grammar of definition sentences. 2002. xvi, 281 pp. 10 Aijmer, Karin: English Discourse Particles. Evidence from a corpus. 2002. xvi, 299 pp. 9 Reppen, Randi, Susan M. Fitzmaurice and Douglas Biber (eds.): Using Corpora to Explore Linguistic Variation. 2002. xii, 275 pp. 8 Stenström, Anna-Brita, Gisle Andersen and Ingrid Kristine Hasund: Trends in Teenage Talk. Corpus compilation, analysis and findings. 2002. xii, 229 pp. 7 Altenberg, Bengt and Sylviane Granger (eds.): Lexis in Contrast. Corpus-based approaches. 2002. x, 339 pp. 6 Tognini-Bonelli, Elena: Corpus Linguistics at Work. 2001. xii, 224 pp.

5 4 3 2 1

Ghadessy, Mohsen, Alex Henry and Robert L. Roseberry (eds.): Small Corpus Studies and ELT. Theory and practice. 2001. xxiv, 420 pp. Hunston, Susan and Gill Francis: Pattern Grammar. A corpus-driven approach to the lexical grammar of English. 2000. xiv, 288 pp. Botley, Simon Philip and Tony McEnery (eds.): Corpus-based and Computational Approaches to Discourse Anaphora. 2000. vi, 258 pp. Partington, Alan: Patterns and Meanings. Using corpora for English language research and teaching. 1998. x, 158 pp. Pearson, Jennifer: Terms in Context. 1998. xii, 246 pp.

Corpora and Language Teaching (Studies in Corpus Linguistics)

Corpora and Language Learners (Studies in Corpus Linguistics)

How to Use Corpora in Language Teaching (Studies in Corpus Linguistics, 12)

An Introduction to Corpus Linguistics (Studies in Language and Linguistics)

Exploring Corpora for ESP Learning (Studies in Corpus Linguistics)

Using Corpora to Explore Linguistic Variation (Studies in Corpus Linguistics)

Studies In French Applied Linguistics (Language Learning and Language Teaching)

Perspectives on Corpus Linguistics (Studies in Corpus Linguistics)

New Trends in Corpora and Language Learning (Corpus And Discourse)

Cognitive Linguistics and Language Teaching

Corpus, Cognition and Causative Constructions (Studies in Corpus Linguistics)

Keyness in Texts (Studies in Corpus Linguistics)

Lexis in Contrast: Corpus-based Approaches (Studies in Corpus Linguistics)

English Corpus Linguistics: An Introduction (Studies in English Language)

Corpus Linguistics: Investigating Language Structure and Use

Lexis in Contrast: Corpus-based Approaches (Studies in Corpus Linguistics)

Collocations in a Learner Corpus (Studies in Corpus Linguistics)

Corpus Approaches to Grammaticalization in English (Studies in Corpus Linguistics)

Textual Patterns: Key Words And Corpus Analysis in Language Education (Studies in Corpus Linguistics)

Corpus Linguistics: Investigating Language Structure and Use

Corpora and Discourse: The Challenges of Different Settings (Studies in Corpus Linguistics, Volume 31)

Researching Specialized Languages (Studies in Corpus Linguistics)

Corpus-based Approaches to English Language Teaching (Corpus and Discourse)

Computer Learner Corpora, Second Language Acquisition and Foreign Language Teaching (Language Learning & Language Teaching)

Corpus Linguistics

Computer Learner Corpora, Second Language Acquisition and Foreign Language Teaching (Language Learning & Language Teaching)

University Language: A corpus-based study of spoken and written registers (Studies in Corpus Linguistics)

Metadiscourse in L1 And L2 English (Studies in Corpus Linguistics)

Small Corpus Studies and Elt: Theory and Practice (Studies in Corpus Linguistics)

Discourse In The Professions: Perspectives From Corpus Linguistics (Studies in Corpus Linguistics, SCL 16)

Sociolinguistics and Language Teaching (Cambridge Applied Linguistics)

Corpora and Language Teaching (Studies in Corpus Linguistics)

Corpora and Language Learners (Studies in Corpus Linguistics)

How to Use Corpora in Language Teaching (Studies in Corpus Linguistics, 12)

An Introduction to Corpus Linguistics (Studies in Language and Linguistics)

Exploring Corpora for ESP Learning (Studies in Corpus Linguistics)

Using Corpora to Explore Linguistic Variation (Studies in Corpus Linguistics)

Studies In French Applied Linguistics (Language Learning and Language Teaching)

Perspectives on Corpus Linguistics (Studies in Corpus Linguistics)

New Trends in Corpora and Language Learning (Corpus And Discourse)

Cognitive Linguistics and Language Teaching

Corpus, Cognition and Causative Constructions (Studies in Corpus Linguistics)

Keyness in Texts (Studies in Corpus Linguistics)

Lexis in Contrast: Corpus-based Approaches (Studies in Corpus Linguistics)

English Corpus Linguistics: An Introduction (Studies in English Language)

Corpus Linguistics: Investigating Language Structure and Use

Lexis in Contrast: Corpus-based Approaches (Studies in Corpus Linguistics)

Collocations in a Learner Corpus (Studies in Corpus Linguistics)

Corpus Approaches to Grammaticalization in English (Studies in Corpus Linguistics)

Textual Patterns: Key Words And Corpus Analysis in Language Education (Studies in Corpus Linguistics)

Corpus Linguistics: Investigating Language Structure and Use

Corpora and Discourse: The Challenges of Different Settings (Studies in Corpus Linguistics, Volume 31)

Researching Specialized Languages (Studies in Corpus Linguistics)

Corpus-based Approaches to English Language Teaching (Corpus and Discourse)

Computer Learner Corpora, Second Language Acquisition and Foreign Language Teaching (Language Learning & Language Teaching)

Corpus Linguistics

Computer Learner Corpora, Second Language Acquisition and Foreign Language Teaching (Language Learning & Language Teaching)

University Language: A corpus-based study of spoken and written registers (Studies in Corpus Linguistics)

Metadiscourse in L1 And L2 English (Studies in Corpus Linguistics)

Small Corpus Studies and Elt: Theory and Practice (Studies in Corpus Linguistics)

Discourse In The Professions: Perspectives From Corpus Linguistics (Studies in Corpus Linguistics, SCL 16)

Sociolinguistics and Language Teaching (Cambridge Applied Linguistics)

Recommend Documents