1As Aijmer (2009: 2) points out:
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.
2The ways that (pre-existing) corpora can be exploited in the classroom is addressed elsewhere (Aijmer, 2009; Frankenberg-Garcia, this volume). The current article reverses the assumption that the corpus comes first – and that it can be exploited in the classroom – in order to put primacy on language teaching and learning, and the corpora which emerge if the learner takes centre-stage. The article suggests that the relationship between language teaching and corpora can take at least three forms, and these are illustrated in its three main sections: from syllabus to corpus (for general language courses), from corpus to syllabus (for specialist language learners) and learner corpora (in this case a role-play corpus). First of all, however, in order to answer the question « What is a corpus in language teaching and learning? », we have to define what we mean by a corpus and what we mean by language teaching. It is to these definitions that I turn my attention in the following sub-sections. The first question is perhaps more straightforward than the second.
3McEnery et al. (2006: 4) provide the following definition of what is normally meant by the word corpus:
The term corpus as used in modern linguistics can best be defined as a collection of sampled texts, written or spoken, in machine-readable form which may be annotated with various forms of linguistic information.
4Corpora provide a body of data which is representative of the language at a particular point in time. The British National Corpus is perhaps the best-known example of a corpus with representative texts gathered from the 1980s-1993. Arising out of Quirk’s Survey of English Usage, the BNC contains a 100-million-word text corpus of samples of written and spoken English from a wide range of sources. The project to create the BNC involved the collaboration of three publishers (with Oxford University Press as the lead collaborator, along with Longman and Chambers), two universities (Oxford and Lancaster) and the British Library. The creation of the BNC started in 1991 under the management of the BNC consortium and the project was finished by 1994. There have been no additions of new samples after 1994, but the BNC underwent slight revisions before the release of the second edition BNC World (2001) and the third edition BNC XML (2007). A new project has been launched, funded by the ESRC, in a collaboration between Lancaster University and Cambridge University Press, to create a further spoken corpus (http://cass.lancs.ac.uk/), the BNC Spoken Corpus 2014.
590% of the BNC is made up of samples of written language use. These samples were extracted from regional and national newspapers, published research journals or periodicals from various academic fields, both fiction and non-fiction books, leaflets, brochures, letters, essays written by students of differing academic levels, speeches, scripts and many other types of texts. The remaining 10% of the BNC is composed of samples of spoken language. The spoken corpus consists of two parts: one part is demographic, containing the transcriptions of spontaneous natural conversations produced by volunteers of various age groups, social classes and originating from different regions. The second part comprises « context-governed » samples such as transcriptions of recordings made at specific types of meetings and events. These are sub-divided into Business, Leisure, Education and Institutional, and the latter contain extracts from courts of law, amongst other institutional contexts. The original recordings transcribed for inclusion in the BNC have been deposited at the British Library Sound Archive and the sound-files are now being linked to the electronic transcriptions by researchers at the University of Lancaster and may be accessed via http://bncweb.lancs.ac.uk.
6As the BNC is a large mixed corpus which set out to be representative of British English as a whole, it is unsuitable for the study of highly specific text-types or genres, as any one of them is likely to be inadequately represented. Those wishing to explore their specific conventions of particular genres would do better to compile a small corpus including only texts of those types.
7To sum up this section, we can say that a corpus is generally understood to be a collection of:
-
authentic texts (including transcriptions of spoken data) which have been
-
sampled so that they are
-
representative of a particular language or variety of a language, and which are
-
machine-readable.
8Before turning to the more general question of what might be covered by the broad term didactique des langues (language teaching and learning), let us consider the two general ways in which corpus material can be used in language teaching. Firstly, publishers and researchers can use corpus samples to create language-learning syllabuses and materials. The learners themselves do not have access to the corpus but the corpus informs the way that language is presented to students in learning materials. Secondly, the analysis of corpus data can be incorporated directly into the language teaching and learning environment. With this method, language learners are given the opportunity to categorize language data from the corpus and subsequently form conclusions about the patterns and features of the target language from their categorizations. This method involves a greater amount of work on the part of the language learner and is referred to as « data-driven learning » or as « hands-on » corpus use (see Frankenberg-Garcia, this volume). Thirdly, in a « hands-off » approach, a tutor can use corpus examples to illustrate particular language points. This demands considerable insight and work on the part of the tutor. A large representative corpus like the BNC is particularly useful as a reference source when studying the use of individual words in different contexts, so that learners become familiar with the different ways to use particular words in context. As Hunston (2009) points out, however, explanations of this sort only accentuate our perceptions of the complexity of language rather than providing the type of straightforward « rule » that learners crave. Arguably, a representative corpus can show what company a word keeps (its collocations) and also its frequency, so that translators, for example, could select a word which is equally frequent in the target language as in the source language.
9The theory and practice of language teaching and language learning is a vast field which it would be inappropriate to attempt to encapsulate here. In general, we can say that, since the 1960s, the field has moved from a focus on grammar-translation (the aim of learning a language was to read its literature) towards an interest in communicative competence (the aim is to be able to function practically in daily interactions of different types). This communicative revolution took hold in the 1970s and 1980s and there was a strong focus on « authentic » language which I will come back to in Section 1.3. Most language teaching materials take an eclectic approach which covers the acquisition of grammar and vocabulary in everyday and thematic situations which are relevant to the student body in question. Most syllabuses also highlight the four skills (listening, speaking, reading and writing) and are adapted to the level of the student (beginner, intermediate or advanced, to give but the broadest categories). Over the last 40 years, the focus in language syllabuses has shifted from grammar to situations, themes, functions and notions, to task-based learning, the lexical syllabus and more learner-centred approaches built around needs analyses. It was, however, back in the 1980s that these developments were beginning to shape language teaching and learning. As Nunan (2007: 10) suggests:
The 1980s was the decade in which the principles of communicative language teaching, which had evolved in the preceding decade, began to gain traction in the classroom. We began to see curricula and materials that took as their point of departure an analysis of learners » communicative needs, rather than inventories of language systems. Needs analysis procedures and needs based programming emerged to support the development of differentiated curricula to meet different learner needs.
10Theories of language learning and corpora came together particularly forcefully in the early 1980s when I began my French text-book writing career: the focus was on « authenticity », the provision of samples of language which were produced in the target culture in real communicative situations. This was due in part to a reaction against the unnatural model sentences favoured by the grammar-translation approach which were fabricated to illustrate particular aspects of structure. As far as I was aware, there was no material of an authentic sort of this type which was easily available for adaptation for the teaching of French in schools, though researchers/teachers from the University of Reading had been trail-blazers in this area in creating the Enquête Sociolinguistique sur Orléans Corpus in 1968.
11Through the 1970s and on into the 1990s and the first decade of the current century, the team of researchers at the Crapel (Centre de Recherches et d »Applications Pédagogiques en Langues) in Nancy were at the forefront of discussions of the interaction between « authenticity » and what might be « authentic » for a language learner (Duda et al., 1972; Holec, 1974, 1990; Boulton, 2009; Duda & Tyne, 2010). Authentic texts are generally described as spontaneous, spoken, non-scripted texts produced in a real communicative situation. For a learner, however, in order to be « authentic », a text must be both relevant and accessible. As Widdowson (1998: 714-715) pointed out, language learning tasks "must take account of the interests, attitudes, and dispositions of the learners… the appropriate language for learning is language that can be appropriated for learning". Rühlemann (2008: 685) remarks that:
authenticity in Widdowson »s sense does not depend on the text being invented by a materials designer or captured in a spoken corpus, but on the successful mediation through careful selection and motivating teaching.
12The notion of authenticity, then, is not something which is inherent to the text but is a negotiation between the teacher and the learners.
13The remaining sections of this article illustrate three very different types of spoken corpora which I have created and used for language teaching and learning: firstly, in Section 2, a corpus which resulted from the development of a series of text-books published by Oxford University Press and Cambridge University Press between 1981 and 1993 and aimed at the different needs of British students studying for their GCSE and A-level examinations in French; secondly, in Section 3, a small corpus of French for Specific Purposes (to meet the needs of student gardeners and golf greenkeepers) created in 1995; and, thirdly, in Section 4, a corpus composed of a collection of role-plays in English, designed to help researchers and learners better understand the role of pragmatic markers in spoken discourse.
Figure 1 – Covers of text-books of French (1980-1993).
- 1 School exams taken at the age of 16 and 18 respectively, in England, Wales and Northern Ireland.
14From 1980 to 1993, on the back of my MA dissertation on the exploitation of authentic resource materials (ARMs) and an article on this topic (Beeching, 1982), I was commissioned by Oxford University Press and Cambridge University Press to write a series of French text-books which addressed the need for samples of spontaneous spoken (and written) French for students preparing for GCSE and A-level1 examinations in the UK. The first, Vrai de Vrai! Authentic French for Listening and Reading Comprehension (Beeching, 1985), focused on receptive skills and was tailored to the GCSE syllabus with chapters devoted to transactions in shops and at the station, hotel and post-office, but also discussions about clothes and fashion, music, feminism and nuclear power. Authentic materials were expected to be not only relevant but also motivating for students. As I say in the introduction:
The exploitation of authentic material is being increasingly recommended both because of the obvious relevance of such things as menus and tourist information brochures and for the effect they have on students » motivation. It has been shown that motivation is one of the factors – if not the factor – which is crucial in learning a foreign language. The national criteria for GCSE French stress that examination tasks should be of value outside the classroom and that the material used should be carefully selected authentic materials. (Beeching, 1985: 3).
15Not only were the recordings collected for this particular purpose but the exploitation of them is carefully graded, and teachers are enjoined to adapt the way that the texts are presented to suit the level of proficiency of the students. The notion of « gist » listening was at that time quite a new one and the caveat is issued that the student is not expected to understand every word. Various tactics are deployed to ensure that students are supported in their understanding, however: the provision of vocabulary lists to be presented before listening, followed by « sign-post » questions, a range of multiple-choice, gap-fill, matching or other types of activity in English or in French. « Authenticity » is thus constructed as being both naturally-occurring French speech (and writing) and a negotiation which involves a careful consideration of learner proficiency and learner needs.
16A similar approach was taken in the ensuing publications. All of these were based on recordings of spontaneous speech. A Vrai Dire… Authentic French for Role-Play (Beeching, 1986), focuses on productive skills at GCSE. Contrastes (Beeching & Page, 1988) was aimed at the development of a range of skills in A-level students. Ça se Dit et ça s’écrit (Beeching & Le Guilloux, 1990) was designed to develop advanced spoken and written skills, and to differentiate between the two modes, while La Passerelle (Beeching & Le Guilloux, 1993) provided a much-needed bridge between GCSE and A-level, focusing on developing basic grammar in students whose syllabus up till then had not done so.
17After more than 10 years of recording and transcribing spoken French, I had collected a fairly large corpus of spoken language, 17½ hours and 155,000 words of which are available as an online corpus: www2.uwe.ac.uk/faculties/CAHE/ELC/Documents/iclru/corpus.pdf. The corpus features speakers from Paris, Brittany, the Lot, the Minervois and from Belgium and includes school-children, bakers, butchers, nurses, doctors, salespersons and even a toiletteur de chiens/dog stylist.
18In 2014, there is still no fully-fledged reference corpus of French to equal the British National Corpus though a number of projects are afoot to create one. In addition to my small Beeching Corpus (1980-1993) mentioned above, the following are accessible online:
-
ELILAP (Étude LInguistique de la LAngue Parlée) and LANCOM (LANgue et COMmunication) (http://bach.arts.kuleuven.be/elicop/ProjetLANCOM.htm) ;
-
ESLO and ESLO 2 Enquête Sociolinguistique sur Orléans (http://eslo.tge-adonis.fr/), which will contain the ESLO Corpus from 1968, replicated in 2014, as ESLO2 ;
-
CRFP (Corpus de Référence du Français Parlé) ;
-
CFPP (Corpus du Français Parlé Parisien) (http://cfpp2000.univ-paris3.fr/) ;
-
CLAPI (Corpus de LAngue Parlée en Interaction) (http://clapi.univ-lyon2.fr/).
19There is a fuller list of French spoken corpora here:
-
Inventaire des corpus oraux de la DGLFLF (Délégation Générale à la Langue Française et aux Langues de France). Disponible en ligne. http://www.culture.gouv.fr/culture/dglf/recherche/corpus_parole/Inventaire.pdf
-
CNRTL (Centre National de Ressources Textuelles et Lexicales). Disponible en ligne. http://www.cnrtl.fr/
20These corpora provide data from different genres and regions of the French-speaking world, some with aligned audio, video and transcriptions. All of them can be exploited by the language teacher (or language researcher). However, until they are grouped together, and selections are made, none can be said to have been sampled so as to be representative of the French language as a whole (see Pierrel, this volume, for an initiative to make French corpora more accessible).
21In the early 1990s, I was asked to teach French to a group of trainee gardeners and golf greenkeepers from the nearby Agriculture College in Bridgewater. These students had GCSE French and a very remote interest in French grammar. They undertook a field-trip each year to a partner-college in Coutances in Normandy and some of them went on to do internships at gardening establishments in France such as a bambouseraie (bamboo plantation) in Lyon. In order to meet their needs and to harness their motivation, I developed the course to include references to their specific subject-area: Horticulture. Drawing on horticultural magazines in French, the students prepared talks – in French – on subjects as diverse as maintaining the lawn-courts at Wimbledon, dealing with moles and badgers, using vegetation shredders, and so on. The level of language achieved was astonishing despite the students » lack of knowledge of basic grammar rules. The students invited me to join them on their annual field trip to Coutances. This was my opportunity to collect the data I needed to tailor next year »s course to their specific needs. The trip was organised around a series of visits to experimental stations where a range of plants were grown. One of the students video-recorded the presentations in French while I did some improvised interpreting where required. There were frequently occasions when I filled in the gaps in the students » French while the students filled in the gaps in my knowledge of horticulture. So, when periodicité was mentioned, I could translate it as « periodicity » without having a clear knowledge of what was meant. The students were able to explain to me the process that was being referred to, and were proud to do so. This type of collaboration between language and subject specialists is particularly fruitful in specific purposes contexts leading to a greater sense of equality between teacher and learner and to collaborative and negotiated learning (for more on Specific Purposes, see Kübler, this volume).
22On our return to the UK, I transcribed the video-recordings and analysed the data; the results are published in Beeching (1997). The main findings which informed the syllabus for the following year in terms of grammatical items, important structures and lexis were as follows:
23Grammatical items:
-
the synthetic future was not used at all, so priority was placed on the analytic future (aller + infinitive) ;
-
the qui/que distinction which poses a problem for English learners of French was not crucial – most uses were of qui in presentatives (see below).
24Important structures:
25Structures which were very frequent in the data and which are not traditionally covered in intermediate courses included characteristically spoken structures such as presentatives and also structures used to express cause and effect. These are illustrated below:
Il y a / On a un/e ……. qui …….
il y a une usine qui s »est construite
on a deux fois 4500 plants qui arrivent en même temps.
C »est un… qui/que/qu »…
c »est des variétées qui sont assez porteuses
c »est un chauffage qu »on met au coeur de la végétation
on met le chauffage 60 degrés > effect > ça permet de chauffer le cœur
si on plante au mois d »avril un cutting dans le sol > effect > il ne va pas fleurir ici où vous avez une forte luminosité > effect > vous avez une chute importante de l »hydrométrie
26Lexical items:
-
Names of plants and parts of plants ;
-
the processes applied to plants, equipment, including computer-assisted administration ;
-
growth conditions ;
-
commercial terminology.
27Most of the plants have cognate forms in English, e.g. aster, astroméria, carotte, chrysanthèmes, escarole, frisée, frésia, géranium, gerbera, orchidée, pelargonium, rununcule. What is more, a high proportion of semi-technical terms can be inferred from English, e.g. bulbe, botritis, cellule, climat, compartement, CO2, concentration, cutting, évaporation, humidité, injecter, irrigation, monopole, modem, périodisme, PPM, programmer, qualité, quantité, température (ambiante), tester, variété, végétation. In terms of pedagogy, grouping terms together can help to focus the learner »s attention on families of forms (fleur, fleur coupée, double fleur, simple fleur, mise à fleur, floraison, date de floraison, fleuron, fleurir) or on forms which look similar but which have different translations in English: plant (« seedling »), plante (« plant »), plante à massifs (« bedding plant »), plantation, planter.
28By corpus standards, this corpus was extremely small. It was, however, precisely targeted at this group of learners » specific needs and thus forms an authentic representative sample of « French for Amenity Horticulture ». Collecting a spoken corpus on the ground allows the teacher of Language for Specific Purposes to discover the communicative acts, the lexis and the grammar which are vital in the contexts in which the learners can envisage operating. The transcription and detailed linguistic analyses bring important insights which can inform the construction of the syllabus. In this case, the data collection and transcription also brought to light certain characteristics of spontaneous spoken French and of multimodality which are interesting from a theoretical point of view and form part of a lively and ongoing field of research (Bruxelles et al., 2006; Mondada, 2009).
29The final example illustrates the benefits of creating a learner corpus (for more on this topic, see De Cock, this volume; Granger, 2009). A comparison of native speaker and learner productions in specific contexts can highlight the differences between L1 and L2 usages, and can help to inform language teaching. In this case, I wanted to investigate the use of discourse markers (DMs) (see Beeching, forthcoming, for fuller details). Previous studies of DMs conducted by applied linguists and language teachers have suggested that, far from being irritating « tics » or « parasites », markers have an important role to play in spoken interaction. Back in the 1980s, Svartvik noted that:
If a foreign language learner says five sheeps or he goed, he can be corrected by practically every native speaker. If, on the other hand, he omits a « well », the likely reaction will be that he is dogmatic, impolite, boring, awkward to talk to, etc., but a native speaker cannot pinpoint an « error ». (Svartvik, 1980: 171)
30Fung and Carter (2007), too, recommend the inclusion of markers in the curriculum:
Incorporation of DMs into the language curriculum is necessary to enhance fluent and naturalistic conversational skills, to help avoid misunderstanding in communication, and, essentially, to provide learners with a sense of security in L2. (Fung & Carter, 2007: 433)
31Further research is, however, required into the functions, polysemies and sociolinguistic salience of individual markers before language teachers can feel confident about presenting them as part of the language curriculum. As Kasper (1997: n.p.) has pointed out, pragmatic competence is not easily amenable to intuition:
Because native speaker intuition is a notoriously unreliable source of information about the communicative practices of their own community, it is vital that teaching materials on L2 pragmatics are research-based (Myers-Scotton & Bernstein, 1988; Wolfson, 1989; Olshtain & Cohen, 1991; Bardovi-Harlig et al., 1991).
32Research of this sort can most reliably be conducted using corpus data.
33Typically, and understandably, learners are expected to simply « pick up » DM-usage as they become more integrated into a particular speech community (see Hellerman & Vergun, 2007). Given that the meanings and functions of DMs can be complex and opaque and that the sociolinguistic salience of particular markers may vary across different parts of the English-speaking world, it seems premature to rush into the teaching of them; we may not know, for example whether a learner will end up in the UK, the US, Australia, New Zealand or Canada, not to mention countries where English is a widely spoken colonial language, like India, or others where it is a Lingua Franca. Similarly, teachers typically do not know whether their learners are destined for either further study or employment in one of the countries in which English is spoken or will for example require a purely reading knowledge of the language to facilitate study in the host country. A grasp of more colloquial spoken forms like DMs will more likely prove to be an asset for the former group of students than for the latter.
34Table 1 (from Beeching, forthcoming) summarises the range of studies which have previously been conducted on the acquisition of different DMs across a range of L1s, levels of competence and contexts, and employing different research methods.
Table 1: Previous studies of native and non-native usages of discourse markers.
Study
|
L1
|
Level of competence in English
|
Context
|
Research method and markers studied
|
Müller, 2005
|
German
|
9-12 + years of formal instruction university students (95,555 words)
|
Silent movie stimulus > oral narrative in pairs, followed by exchange of opinions – conducted in experimental conditions
|
- Information gap oral narrative - Exchange of opinions - « Use of English » questionnaire So, well, you know, like
|
Fung & Carter, 2007
|
Cantonese
|
49 Intermediate-advanced school pupils (17-19 year-olds) (14,157 words)
|
Tape-recorded group role-play recorded in Hong Kong in a pedagogic context
|
« Naturally occurring » interaction (a very wide range of markers are included)
|
Denke, 2009
|
Swedish
|
Advanced (37,152 words)
|
Video-recorded natural science seminars – mainly monologal
|
Naturalistic spoken data Y »know, well, I mean
|
Liao, 2009
|
Chinese
|
6 Advanced graduate students in a study-abroad context 6 x 40-50 min. lectures + 6 x 40 min. interviews > 1,422 tokens of 9 markers
|
Teaching-Assistant led discussions in the US; sociolinguistic interviews
|
Lecture with quantitative analysis + qualitative ethnographic interviews Yeah, oh, you know, like, well, I mean, ok, right, actually
|
Baumgarten & House, 2010
|
US English and mixed non-natives: Indonesian, German, Chinese, Korean, French, Nepali, Gujarati
|
University students in Germany; 3 x 30-min. conversations on a discursive topic:
English L1
ELF1
ELF2
|
Audio-taped elicited conversations on a topic of general interest « the role of English in the world » or « men and women in the contemporary arts scene »
|
Quantitative and qualitative analysis of the uses of I think and I don »t know
|
Polat, 2011
|
Turkish
|
5 years formal EFL instruction in Turkey; emigrated to US aged 25
|
One individual who has emigrated (married to a US citizen). Audio-taped conversations every 2 weeks for 1 year
|
Developmental learner corpus – one speaker »s development over a year
You know, like and well
|
Beeching, forthcoming
|
Chinese L1
Mixed L1 recorded at UWE
Native speakers recorded by Qun Zheng at UWE
|
Intermediate / advanced university students, in 2 groups:
1. 24 Beijing Foreign Studies University: SWELC Corpus (6,326 words)
2. 28 L1 Albanian, Czech, French, German, Spanish and Vietnamese students at UWE (7,419 words)
3. 44 native speakers of English, students at UWE (14,308 words)
|
3-min. argumentative role-play in pairs; exam situation; « practice » exam situation; experimental situation
|
Quantitative (role-play data) + qualitative: some brief ethnographic interviews of mixed L1 subjects
Well, you know, I don »t know, like, sort of, I think, I mean
|
35The scope of the current article does not permit detailed elaboration on each of these studies. Some studies (Denke, 2009; Liao, 2009) look purely at DM usage in non-natives while others compare native and non-native usages. Fung and Carter (2007) compare native and non-native usage in corpora which were collected independently of each other (they compared their own data with occurrences in the Cancode corpus). Polat (2011) compares the data from her one individual subject that she studies longitudinally with native-speaker data drawn from the Louvain Corpus of Native English Conversation (Locnec: De Cock, 2004). Müller (2005), Baumgarten and House (2010) and my own forthcoming study draw on comparable corpora: native and non-native data were collected following the same protocol with both natives and non-natives in order to ensure that like was being compared with like.
36Of the studies which did not include detailed comparison with a native-speaker corpus, Denke »s (2009) investigation looks specifically at repair and repetition in the context of monologal seminar presentations. The more interpersonal functions of DMs are less frequent in such a context. The interest of Denke »s study lies in the way that genre influences both the frequency of occurrence and the functions which DMs can have. Liao »s (2009) study of Chinese Teaching Assistants in the US highlights the extent to which the speaker »s desire to integrate in the target culture influences their rate of usage of DMs. One of her participants lived with her Chinese husband, had little contact with natives of English and intended to return to China – her attitudes to DMs as informality markers were largely negative and she avoided using them. Another participant was, however, intending to settle permanently in the US, had a number of American friends and used DMs plentifully as a means of creating friendly rapport. Ethnographic studies of this sort, which couple attitudinal interviews with transcribed recordings of spoken data which can be drawn on for detailed frequency counts, allow the researcher to correlate language usage with extralinguistic information in a particularly interesting way. Polat (2011) extends this type of approach to her longitudinal study of a single individual, a native speaker of Turkish who emigrated to the US aged 25 and married an American native speaker of English. Polat used the developmental learner corpus she created to trace patterns of use of three markers: you know, like and well. She found that you know was heavily used by the participant though rates of usage dropped by half over the year of recordings. Like increased from almost no usages at the beginning to 230 per 10,000 words by the mid-point, then reduced by half by the end of the study. Rates of both you know and like far exceeded those of the native speakers in the Locnec corpus. Finally, well was not used at all as a discourse marker.
37Turning now to the studies which explicitly focus on eliciting native and non-native performance, Müller (2005) compares the use of markers by American and German native speakers performing the same tasks in experimental conditions. She has a large corpus to draw on (the Giessen – Long Beach Corpus) and her study compares two communicative functions, narrating and opinion-giving. She provides an exhaustive analysis of the uses of four markers: so, well, you know and like. She found that the Americans used so, you know and like to a statistically significantly greater extent than the Germans. Well, on the other hand, was used more by the Germans than the Americans, though not to a statistically significant extent. The native speakers of German used well for different functions – when searching for the right phrase, giving an indirect answer, continuing an opinion and concluding (to a statistically significant degree).
38Fung and Carter’s (2007) study compares rates of DM usage in the interactive classroom discourse of secondary school pupils in Hong Kong with usage in the Cancode corpus; native speakers were found to use DMs for a wider range of pragmatic functions. Baumgarten and House (2009), on the other hand, found that I think was used more frequently and in a wider range of contexts by ELF than by L1 English speakers: it was the most common way of « expressing subjective meanings and taking stances ». This finding was replicated in my own forthcoming study with respect to the Chinese learners of English. In this study, native speakers, a mixed non-native group on a stay in the UK and Chinese non-natives who had not been to an English-speaking country were recorded doing a 3-minute role-play in pairs. Their conversations were transcribed and raw distributional frequencies were charted for a range of lexemes which have DM functions. Figure 1 charts rates of occurrence per 10,000 words of the lexemes well, you know, I don »t know, like, sort of, I think and I mean. It is important to point out that these are raw rates, in other words, counts for like include all occurrences of like (as in I like bananas and He looks just like his father, not just discourse-marking like).
Figure 2 : Rates of occurrence per 10,000 words of a range of discourse markers in role-plays conducted by Chinese and mixed non-native and native speakers of English.
39Figure 2 shows that well is most often used by the natives at a rate of just over 40 occurrences per 10,000 words, with the mixed non-natives coming in a rate of 20 and the Chinese non-natives using it about 5 times per 10,000 words. In general, the Chinese beginners have low rates of DM usage apart from, and quite markedly so, you know and I think. It seems that these learners rely quite heavily on these two markers to elicit agreement or approval from their interlocutors and to express hesitation – or, indeed, their opinion. The mixed non-natives who have spent some time in the UK have rather higher rates than the Chinese non-natives and equal, or exceed, the native speakers in their rates of usage of you know, I think and I mean. That is to say that these learners are very good at using interactional markers which are addressee-oriented (you know) and speaker-oriented (I think and I mean). They also use I don’t know to a greater extent than native speakers do. These learners use DMs as compensation strategies which help them to fill their pauses with appropriate fillers. Polat (2011: 3752) suggests that her participant uses you know as a « catch-all gap-filler » and finds that native speakers have far higher rates of other hesitation markers such as uh/eh, um and er which would support such a hypothesis. In my own data, the mixed non-natives have lower rates of both like and sort of than the native speakers but they have clearly picked up the usage of like from their native-speaking peers and are beginning to use it appropriately from both a functional and sociolinguistic point of view –for their age-group, this is the mitigating marker of choice. Corpora of this sort allow not only raw frequency counts, but also access to the contexts in which the markers were used. In the vast majority of cases, like was used as an approximative and not in its focaliser or quotative functions (see Dailey O’Cain, 2000; Macaulay, 2001; Buchstaller & D »Arcy, 2009). The example below shows the way in which like was used alongside other hesitation gap-filling markers, such as I mean, you know and I don »t know to “act as a stock-phrase mainly used to help speakers process and plan their output, and link spans of discourse” (House, 2009: 187):
L: yeah sure but I mean like you can we can probably work for money during the year and you know it »s much more rewarding when you know you get to work for free and you know like for me for example I I I »m going to do volunteering this summer you know helping people er like I don »t know feel more yeah rewarding in the end you should do that.
40Brief conversations were held with the participants after their role-plays were recorded, and subject « L » in particular had not been in the UK for long (under 3 months). She declared herself to be very interested in staying on in the UK – indeed, she claimed that she wanted to marry an Englishman! This desire in principle to be accepted and integrated into UK society appears to play a role in the adoption of markers such as like.
41In conclusion, there is considerable variability in the use of DMs across native and non-native speakers, in terms of both their distributional frequency and the functions for which they are pressed in to service. Natives tend to use more, but this is not invariably the case: non-natives can also depend on certain markers like « lexical teddy bears » (Hasselgren, 2002) to bridge the gaps between sequences of talk, to hesitate and fill pauses. These appear to vary according to their native language: there is some evidence that natives of German use well, Dutch-speaking Belgians use so (Buysse, 2012) and Chinese learners use I think. Most learners depend on you know to gain confirmation from their interlocutor that they have said the right thing. As Polat »s participant Alex puts it:
I wanted to make sure you understood me. I don »t trust my English. I am saying right or wrong, and I want to get approval from you. (Polat, 2011: 3754).
42As Polat remarks, this indicates that DMs may play a particular role in non-native discourse, one which is similar but more crucial than that played in native discourse, as a means of negotiating meaning.
43There are, however, other factors, too, which come into play in the acquisition of DMs. Some learners acquire markers in a naturalistic way in conversation with native speakers once they are in the host country, but identity factors play a large part in their desire to do so. DMs are complex, multifunctional and may be stylistically stigmatised. Teaching them may rely on developing students » observation using appropriate naturalistic data and allowing time for conversation of an interactional sort in the classroom.
44In this section, we have surveyed studies of the non-native acquisition of DMs which draw on different types of corpora, recorded in different communicative situations. The advantage of the experimental design of the role-play corpus is that it generates data which allow the researcher to compare like with like. It is relatively easy to collect and transcribe such a corpus: each role-play is only 3 minutes in length and a representative corpus of different speakers can therefore be collected fairly quickly. The role-play scenario was sufficiently stimulating to produce lively discussion – and, most importantly, a good number of DMs is generated. The data are thus robust and it is possible to make both statistical and qualitative analyses.
45On the downside, the role-play situation could be considered to be unnatural, set in a particular communicative situation and thus ungeneralisable. Also, as Müller (2005) points out, there may be significant differences depending on whether speakers took Role A or Role B – eliciting or giving information, or in the case of the UWE Role-play, defending the volunteering or the working-for-money position.
46Table 2 attempts to summarise the criteria that need to be borne in mind when considering a corpus which might be suitable for language teaching. In this article, we have considered three different scenarios: firstly, in « From syllabus to corpus », a corpus is collected and transcribed by a materials developer specifically with a large general population of students in mind, who are studying for exams based on a national curriculum. These data are authentic, non-scripted, representative, machine-readable and pedagogically relevant, with support in the form of vocabulary help, sign-post questions, and a range of gap-fill, matching and multiple choice comprehension and production exercises. Secondly, in « From corpus to syllabus », authentic, non-scripted and relevant recordings are made in situ in a Language for Specific Purposes situation. These recordings and their transcriptions are analysed in order to inform the syllabus in terms of the most common lexical and structural features which are pertinent in this specific communicative situation. This corpus may be small, and is not generalizable to students following a more global language course (though aspects of it, such as features of the spoken language which have not traditionally been taken on board in language syllabuses, may be of more general interest). Finally, in the section on « The role-play corpus », the place of corpora for research on the acquisition of particular features was highlighted, in this case the growing study of the functions of DMs in native and non-native discourse.
Table 2 : Criteria for corpora for teaching and learning
|
CORPORA
|
|
|
FOR TEACHING AND LEARNING
|
|
authentic? (non-scripted)
|
Representative?
|
machine- readable?
|
pedagogic?
|
|
|
|
|
|
relevant
|
support
|
Syllabus > corpus
|
✔
|
✔
|
✔
|
✔
|
✔
|
Corpus > syllabus
|
✔
|
✔ (of a particular genre)
|
☐
|
✔
|
no
|
Role-play
|
✔ (guided, improvised)
|
✔ (of a particular population)
|
✔
|
✔
|
no yet
|
47Looking to the future, developments will no doubt include video as well as audio as a necessary part of what is deemed authentic for the 21st century, with students » increasing access to YouTube clips and internet media. Multimedia corpora coupled with the necessary pedagogic support online in the form of sub-titles, transcriptions and language-learning activities can be tailored to students » specific needs. Multimedia applications are not only motivating and appropriate to the internet age, but provide greater contextualisation, including paralinguistic features and sociolinguistic background information, to better situate extracts in their authentic communicative situation.
48At more advanced levels and in university language learning situations, students themselves can create their own « do-it-yourself » corpora, which can be either « throwaway » or archived to be used by their peers (see Tyne, 2009; Boulton & Tyne, 2013: 108). The active engagement of students in the creation of corpora raises a more general and crucial point with respect to the relationship between corpora and pedagogy. Hands-on corpus activities involving concordance lines do not suit all learners: some prefer to see language embedded in social situations in which real speakers can be seen engaging with each other in specific sociolinguistic contexts. If students select extracts for transcription themselves, we can be more certain that not only the material itself but also the engagement that the student has with it is « authentic ».