The Routledge Handbook of Corpus Linguistics

19
What can a corpus tell us about specialist genres?

Michael Handford

1. Corpora, specialised corpora and mega-corpora

Corpus and the study of genres: an introduction

While corpus linguistics is undoubtedly one of the modern success stories in applied linguistics, evidenced in part by the creation of this handbook, the value of corpora is not universally accepted (Flowerdew 2004). Many of the criticisms within applied linguistics are concerned with the following three areas:

• Corpus data are decontextualised data (Widdowson 1998, 2000).

• Corpora require a bottom-up approach (Swales 2002).

• Corpora are quantitative, number-crunching tools (see Baker 2006: 8).

A further issue is that of size, which in corpus linguistics impacts notions of representativeness. A widely held assumption in corpus linguistics, and a reason for some linguists, such as discourse analysts, being put off by corpora (Baker 2006), is that researchers need to collect considerable quantities of data because:

• The bigger the corpora the better (Sinclair 1991; Stubbs 1996).

Through exploring a selection of studies that combine corpus and genre approaches, this chapter will critically deal with all of these arguments. In so doing, we will be looking at how two of the most powerful approaches in discourse studies today (Bhatia et al. 2008: 3) which are ‘charting new courses’ in discourse analysis, can be fruitfully combined. This combination of genre analysis and corpus linguistics occurs through the development and analysis of what will be referred to as corpora of specialised genres (CSGs). Such corpora are created to research various aspects of language such as gender comparisons in spoken conversations, changes over time in region-based communication, or styles of writing in academic journal articles (see Hunston 2002: 14). In this chapter we will look at CSGs involving written academic, spoken business and service-encounter discourse.

This complementary combination of corpus and genre is a logical and desirable development in discourse analysis: corpora have much to say about language, but they can be lacking in contextual interpretability; genres are intrinsically contextual entities, but their linguistic features may be under-exposed. The combination is possible because, as Lee reminds us, ‘Most discourse analysts would probably concur with the corpus-based view of language: after all, what is discourse but actual, situationally embedded, used and re-used language?’ (Lee 2008: 88).

This chapter therefore has relevance to two of the most important and recurring problems in social science in general (Giddens 1984) and in the analysis of language in particular (Gee 2005): how can we relate the specific instance (such as the text, the discourse move or the lexicogrammatical item) to the wider social context within which it occurs and which it reflexively recreates, and how can quantitative and qualitative approaches be effectively combined?

Specialist corpora and mega-corpora

According to Flowerdew (2002: 96), two recent developments in the field of corpus linguistics have been:

1. the increase in the number of ‘mega-corpora’;

2. the creation of smaller, genre-based corpora.

‘Mega-corpora’ are large general corpora, comprising millions or hundred of millions of words, such as the BNC or the Bank of English. At the time of writing, the Cambridge International Corpus (CIC) is one of the world’s biggest corpora, weighing in at over one billion words.

Many of the criticisms levelled against corpus linguistics have been aimed at such large, general corpora (Flowerdew 2004): if you are dealing with such an enormous amount of text, then indeed you cannot say much about context, and probably the only way into the data is through a bottom-up, automated, quantitative approach. Some have even argued that the point of corpora is to provide decontextualised data (Hunston 2002: 110). While such very large corpora have borne paradigm-shaking fruit in the fields of lexis and grammar, most notable in the works of Sinclair both theoretically (e.g. 1991) and also practically (the revolutionary Cobuild dictionary), general corpora do not tend to be suitable for studying specialised language (Biber 1988; Flowerdew 2004; Lee 2008).

There are several reasons why smaller corpora are more suited for studying specialist genres. As Lee (ibid.) states, the sampling of most mega-corpora is not based on genre, and while classifying a mega-corpus post hoc is possible, it is hardly ideal (see Lee 2001 for an ad hoc classification of the BNC corpus). Furthermore, because of the way megacorpora are designed, accessing the individual sub-corpora is not always possible; access can also be blocked because of the confidentiality of the data itself (Flowerdew 2004). There is also a widespread tendency for mega-corpora to be comprised of far more written than spoken data, although this emphasis on writing is arguably more to do with practical and financial as opposed to ideological issues than would traditionally have been the case. Moreover, there are no multimodal mega-corpora to date (Lee 2001). A further problem with such general corpora is that they do not always contain complete texts, but may be limited to 2,000 word excerpts (see Nelson, this volume). Given that one of the possible defining features of a genre is that it specifies conditions for beginning and ending texts (Couture 1986: 86), not having the endings (or other sections) of longer texts would be something of a concern. It also means studying the extent to which items are primed to occur according to their position in the overall discourse (Hoey 2005) is not possible.

We will discuss the methodological advantages of smaller corpora over general corpora below; before that, we shall look at what is meant by ‘specialist genres’, the related term ‘specialised corpora’ and their rationale.

Specialist genres

In a way, the expression ‘specialist genres’ is somewhat tautological, as all genres could be seen as inherently specialised. Whether we are thinking about business letters, funeral speeches, PhD thesis abstracts or even conversations between friends, it is possible to pinpoint certain aspects of the genre which distinguish it from other communicative events and therefore make it a ‘specialist’ genre. Looking at one of the most invoked definitions of genre should help us understand this:

genre comprises a class of communicative events, the members of which share some set of communicative purposes. These purposes are recognised by the expert members of the parent discourse community, and thereby constitute the rationale for the genre. This rationale shapes the schematic structure of the discourse and influences and constrains choice of content and style.

(Swales 1990: 58)

There are several words and phrases here that indicate the inherently ‘specialist’ nature of genre: class, expert members, parent discourse community, rationale, the schematic structure and perhaps most importantly constrains. As any parent of young children would testify, even ‘having a chat’ within a specific culture takes years to master. The child needs to develop tacit knowledge of various constraints and possibilities, such as knowing suitable topics and their relevant vocabulary, when they can/cannot be discussed (for example at the dinner table) and with whom (the neighbours, teachers, friends), appropriate turn-taking, listening skills and behaviour, and various paralinguistic features like touching and volume.

Nevertheless, some genres are certainly more specialist than others in that they are employed by highly trained, expert members of particular communities (Swales 1990; Bazerman 1994; Wenger 1998; Scollon and Scollon 2001) with very specific goals in mind and are achieved through highly constrained and specialised communicative behaviours, linguistic and otherwise.

The rationale for a genre approach

Genre itself is a highly contested notion, sometimes contrasted and sometimes used synonymously with other related terms such as register, style and text (Biber 1988; Martin 1992; Stubbs 1996; see Lee 2001 for an extended discussion). Under the umbrella term ‘genre analysis’, there are three broad branches (Hyon 1996): genre as goal-driven communicative event associated with discourse communities (Swales 1990, 2004); genre as rhetorical and social action (Bazerman 1994), associated with the New Rhetoric School; and genre as a staged, social process (Martin 1992), associated with Australian systemic functional linguistics (SFL).

These approaches differ in terms of the emphasis given to predictability and dynamism of the genre’s formal features, or the importance given to the wider social context. Despite the differences, the three branches all prioritise conventions, and in so doing provide a justification for conducting genre analysis. As Bhatia (2004: 23) states:

Genre essentially refers to language use in a conventionalised communicative setting in order to give expression to a specific set of communicative goals of a disciplinary or social institution, which give rise to stable structural forms by imposing constrains (sic) on the use of lexicogrammatical as well as discoursal resources.

This quote highlights the related concepts of constraint and convention and, crucially for corpus linguists, how such concepts find regular, recurrent realisation in lexicogrammatical and discoursal form. Studying genre can show us how to link the words speakers and writers use to the social practices they follow, recreate and sometimes challenge, such as maintaining manager/subordinate relationships in the workplace (Handford, forthcoming).

2. What are the methodological advantages of specialised corpora in analysing genres?

In the beginning of this chapter I listed four issues that have been seen as problematic in corpus linguistics: size, context, quantitative analysis and a bottom-up approach. This section shows how a specialised genre approach, as opposed to one utilising general corpora, can practically address these concerns.

One key methodological issue for anyone either compiling or analysing a corpus is that of size: how big should the corpus be? The issue of size is important for obvious practical and financial reasons, but also because it affects the representativeness of the corpus in question (see chapters by Adolphs and Knight, Nelson and Reppen, this volume). One clear advantage that CSGs have over general corpora is that they can be markedly smaller and still validly claim to be representative to some degree (Tognini Bonelli 2001; Tribble 2002). Although guaranteeing or even evaluating representativeness may not be objective (Tognini Bonelli 2001: 57), it seems plausible that the more specialised the genre, the smaller the corpus can be (Lee 2008).

In terms of actual size, a specialised corpus can be defined as large (O’Keeffe et al. 2007) if it contains a million words such as CANBEC, the Cambridge and Nottingham Business English Corpus, a one-million-word corpus of authentic spoken business English, or CANCODE, the Cambridge and Nottingham Corpus of Discourse English, which contains five million words of authentic everyday speech; however, some such specialised spoken corpora that have produced powerful and wide-ranging findings are made up of fewer than 60,000 words (for example Koester’s (2006) corpus of American and British Office Talk, ABOT). As noted above, spoken corpora are usually much smaller than written corpora, but it should also be remembered that transcripts of spoken corpora usually contain far more than just words, for example paralinguistic and even functional information (e.g. interrupting codes). Perhaps the comparison in terms of word count is erroneous: if the comparison is made on contextual information, many spoken corpora are far more detailed.

A further point concerning the issue of accurately describing the genre is made by Hyland (2006: 58). He argues that language corpora have enabled genre analysts, for example those involved in language education, to demonstrate that the observed principles and regularities they highlight are representative of the genre in question. The corpus can provide ‘an evidence-based approach to language teaching’ (ibid.). Such an approach is needed because human intuition is not a reliable guide to the linguistic patterns a genre involves.

Hyland’s position presupposes that the corpus we are using has sufficient relevant contextual information to allow plausible interpretation of the item in question. Yet it is the role of context that has given rise to perhaps the most serious criticisms against corpus linguistics (Hunston 2002). Widdowson argues that corpus data is not authentic language, given that the text is separated from its original context (Widdowson 1998, 2000): it is at best a decontextualised sample and should therefore be used with great care in, for example, language-learning settings. While Widdowson’s point may carry some weight when looking at the larger general corpora discussed earlier, CSGs provide a powerful riposte to this challenge because they can, in principle, allow for not only description, but also interpretation and explanation of the data (see Candlin and Hyland 1999: 1–17). Methodologically, genre corpora are of particular value because the analyst is ‘probably also the compiler and does have familiarity with the wider socio-cultural dimension in which the discourse was created’ (Flowerdew 2004: 16). The possibility to look at a particular linguistic feature across a collection of instances of the relevant genre, and use text-external understanding to interpret the data which contains that feature, is revolutionising not only corpus linguistics and genre analysis, but also the interdisciplinary field of discourse analysis (Lee 2008).

The perception that corpora are number-crunching quantitative tools that necessitate a bottom-up approach has, nevertheless, led to a methodological and indeed theoretical aversion on the part of many social scientists (see Baker 2006: 1–10). The size and composition of many CSGs, however, means that they do lend themselves to qualitativebased analyses (Flowerdew 2004). In one such approach, the user inserts tags which demonstrate the specific move structures across a collection of texts, a necessarily manual procedure which can therefore only be conducted with smaller corpora (Flowerdew 1998). Other studies which employ qualitative analyses will be discussed below. Also, recent research by Swales (Swales and Lee 2006), on the pedagogical applications and implications of using comparative corpora with language learners, shows that Swales now agrees corpora can, in fact, be interpreted with top-down approaches more typically associated with genre analysis (Lee 2008).

After all, a CSG is merely a collection of contextually linked, machine-readable texts which allows for a multi-methods analysis: whether we approach the texts individually or as a whole, manually or automatically, is not dictated by the data itself, but by the research question we want to answer. In the following sections we will explore various studies that exemplify such an approach.

3. What can a corpus tell us about academic genres?

There have been numerous corpus studies of academic language, and although these tend largely to be of written texts (Flowerdew 2002), there have also been some outstanding studies of spoken discourse. The Michigan University MICASE corpus should be noted, both because of the research emanating from it (e.g. Simpson 2004; Swales 2004) and also because it is a free, publicly accessible corpus of over 1,800,000 words which permits straightforward analysis of various spoken texts recorded on the university campus. These include lectures, dissertation defences and meetings, but also service encounters (a web search of MICASE will bring you to the home page). The British companion corpus is the BASE corpus is also freely accessible online. This corpus has video as opposed to transcribed data. The written equivalent, the British Academic Written English (BAWE), is also freely available online. Another excellent site is the Purdue University ‘Owl’ writing laboratory website. Further work on spoken academic genres, as well as other speech genres, has been conducted on Nottingham’s CANCODE corpus (O’Keeffe et al. 2007: Ch. 10) and on an academic subset of it (Evison 2008), and Csomay (2005) has explored variation in differing genres using corpora.

A comprehensive survey of the extensive literature on written and written/spoken genres is beyond the scope of this chapter, but some of the most important studies are Biber 1988; Swales 1990, 2004; Biber et al. 1998; Flowerdew 2002; Hyland 2003; Connor and Upton 2004. Research has been conducted on comparing different academic disciplines (Flowerdew 1998; Hyland 2000; Ghadessy et al. 2001). Another approach is the analysis of particular features such as vague language (Cutting 2007) and hedges (Hyland and Milton 1997) in academic genres.

The study we will look at, by Tribble (2002), is of an invited short academic journal report written by a respected researcher who discusses recent developments in his field. The text is taken from the RAT corpus (Reading Academic Text Corpus, of Reading University). There are three basic steps in Tribble’s study: choose a text which is an ‘expert performance’ (Bazerman 1994: 131) of the relevant genre; compile contextual information about the creation of the text; and conduct linguistic analyses which can be integrated with the context.

Tribble’s study has been chosen because he explicitly integrates many of the themes we have discussed so far, most importantly specialised (written academic) corpus data and genre analysis. He also combines the three branches of genre analysis (Flowerdew 2008) outlined above, allowing for a thorough contextual integration with the corpus data (thereby addressing one of the most recurrent criticisms of corpus linguistics). The study is also interesting because it develops useful pedagogical practice based on the findings, which is not always the case with corpus studies (Swales and Lee 2006: 57).

In terms of traditional corpus linguistic approaches (e.g. Sinclair 1991), Tribble’s analysis is somewhat unusual in the extent to which the text-external context is emphasised (see Table 19.1); furthermore, the analysis is primarily concerned with a single text. Nevertheless, he does employ recognisable corpus techniques such as frequency and concordance searches, and accesses a larger reference corpora (the one-million-word BNC sampler) to conduct keyword searches (Scott 1996). Keywords (ibid.) are statistically significant items in a corpus or text, and can provide an empirically valid impression of the typifying language of the genre (see Scott, this volume).

The first seven headings and associated questions provide an explanation of the social/ cultural dimensions of the text in question. For example, under the second point, social context, Tribble (2002: 134) states: ‘The article was written for publication in a specialist academic journal. In writing such a short piece, the author faces special constraints in terms of content and extent, but also has to meet normal academic standards of warrant and referencing.’ Such insights are important because they help us, and Tribble’s students, understand the context in which the text is produced, and the constraints under

Table 19.1 Tribble's analytical framework

which the writer is working. On a practical pedagogical level, such understanding can then be transferred to academic writing requirements the students may have.

The second part of Tribble’s analytical framework deals with three aspects of linguistic analysis: lexicogrammatical features, text relations/textual patterning, and text structure. The first step is to conduct a keyword search of the article, showing that words such as membrane, enzyme and ethylene are statistically significant in comparison to the reference corpus. According to Tribble, this type of information is essential for those who wish to participate in the target discourse community, because keywords pinpoint the requisite content knowledge of the text (2002: 137), as well as the collocational and colligational patterns which feature these genre-specific terms. Keywords can also clarify which interpersonal features are typical in a genre (Handford, forthcoming).

The next step in the analysis is to do basic frequency counts across the text and the reference corpora, and run left and right concordances of the most frequent words. Interestingly, Tribble argues that this allows for deeper insights into the ‘stylistically salient’ lexicogrammatical features of the text in question than either keyword searches or a single frequency list can produce.

The keyword and frequency analyses have also thrown up aspects of the next level of linguistic analysis, textual patterning. One such pattern was that that clauses are employed in the reporting of claims, as in this sentence. Textual patterns are important because they contribute to the text’s particular identity and are evidence of ‘the constraints which the writer has had to respond to in order to ensure that the text is an allowable contribution to a specialist genre’ (Tribble 2002: 142). In this instance, the constraints include the requirement for claims to be made in academic research, and for them to be communicated in a particular stylistic fashion, often through referencing others, e.g. ‘Y found that …’

The final stage of Tribble’s linguistic analysis looks at text structure, specifically the moves that organise the text’s information. He finds that the text follows the Situation– Problem–Response–Evaluation discourse model (Hoey 1983) in terms of the overall relational patterns, but also in terms of the paragraphing of the text. Such understanding of a text’s moves is of considerable value to novice learners, and according to Swales and Lee (2006) is the first step in acquiring genre-awareness. For a fuller exploration of how texts can be linguistically explored using concordances see Tribble, this volume.

Flowerdew (2008) has argued that Tribble’s (2002) approach is suitable for analysts who are ‘unfamiliar with the institutionalized practices of the genre’, perhaps because it first involves a thoroughgoing analysis of the contextual features. She states that Tribble’s position ‘is to see the role of context as very much informing corpus-based analyses’ (2008: 115), and by implication the importance of collecting and digesting sufficient relevant information about the communicative context and the author’s purpose before we analyse the language.

A possible criticism of this approach, given that it is ‘corpus-based’ (Flowerdew 2008) and not ‘corpus-driven’ (Tognini Bonelli 2001: 65) is that the relationship between the corpus data and the context might be ‘vague’ and ‘difficult to validate’ (ibid.). In a corpus-driven approach, the methodology is ‘observation leads to hypothesis leads to generalisation leads to unification in theoretical statement’ (ibid.: 85). But we should remember that the corpus-driven approach has been developed in alliance with large corpora. CSGs, in contrast, are contextually driven corpora.

4. What does a corpus tell us about professional genres?

As with academic discourse, most of the research on professional genres has explored written texts, for example Bhatia (1993, 2004), Berkenkotter and Huckin (1995), BargielaChiappini and Nickerson (1999), as well as collections which include academic as well as professional studies such as Christie and Martin (1997), Candlin and Hyland (1999) and Connor and Upton (2004). Furthermore, it is only the Connor and Upton collection that explicitly applies corpus linguistics. Research into spoken professional genres is a challenging proposition, not least because of the difficulty in obtaining permission to record authentic data. Nevertheless, work on CSGs of professional discourse includes McCarthy and Handford (2004), Koester (2004, 2006), Adolphs et al. (2007) and Handford (forthcoming), Atkins and Harvey (this volume).

A relevant question is, ‘What counts as a corpus, professional or otherwise?’ Some very important studies of spoken professional discourse have been referred to as ‘corpusbased’, such as Bargiela-Chiappini and Harris’s (1997) study of business meetings, or the outstanding, multidimensional Language in the Workplace (LWP) project headed by Janet Holmes in New Zealand, accessible through the Victoria University website. Studies such as these have been highly influential, not least in the area of spoken genres, and yet the data does not seem to be fully transcribed, as there are no quantitative findings from either study (although see Vine 2004 for work on modal verbs). It is debatable whether such studies are indeed corpus-based, because the data itself does not appear to be in machine-readable form. Therefore, in this chapter, only machine-readable datasets will be considered as corpora and analysed as such.

Another important question is ‘What do we mean by professional?’ The term has received various definitions (e.g. McCarthy 1998; Sarangi and Roberts 1999), but here it is defined as intra-organisational or interorganisational communication between people whose professional identities are salient. In answer to the question of why genres are important in professional communication, it is because professionals achieve their goals and develop professional solidarity through genres (Bhatia 2004: 21).

The analysis of speech creates some different challenges from that of writing. Whereas written texts are much more clear-cut and are usually the finished product when we see them, speech is far fuzzier – it unfolds as a process, and it may not be clear where it’s going, where it begins and where it ends, and whether the audience differs from the participants. But for all of this, there is the argument that speech is more important systemically, more layered semantically, and more interesting analytically, than writing (Halliday 1985/1994). This is because spoken language ‘responds continually to the small but subtle changes in its environment, both verbal and non-verbal –… The context of spoken language is in a constant state of flux, and the language has to be equally mobile and alert’ (Halliday 1985/1994: xxiii–xxiv). It is partly for these reasons that two of the three studies explored in this chapter look at spoken as opposed to written discourse.

In this section we will be looking at Koester’s work on spoken workplace communication (Koester 2004, 2006). Koester’s ABOT (American and British Office Talk) corpus is taken from thirty hours of recordings from various workplaces in the US and UK, although the actual amount of transcribed and contextualised data is 34,000 words. Her work is a demonstration of the depth of analysis possible with a small but carefully selected and painstakingly studied dataset.

Koester is concerned with accounting for the stability and variation evident in spoken genres, and argues that a Swalesian approach, which prioritises communicative purpose, is well suited to do this. She also draws insights and applies tools from systemic function linguistics (SFL) in particular, but argues communicative purpose should be the defining feature of a genre. She distinguishes between transactional and non-transactional (or ‘relational’) workplace discourse, and then categorises such discourse into separate genres, such as decision-making and office gossip, respectively (see Figure 19.1).

In contrast with Tribble (2002), the first methodological step for Koester is to identify the discourse goals by looking at the evidence of the data, and as such is more in line with corpus-driven (Tognini Bonelli 2001) approaches. By applying ethnographic methods from communicative goals studies and interactional sociolinguistics, the genre classification is gleaned from the discourse frames of the participants, as opposed to those of the analyst.

TRANSACTIONAL DISCOURSE
unidirectional	Discursive roles of the participants	collaborative
Genres briefing service encounters procedural and directive discourse requesting action/permission/goods reporting		Genres making arrangements decision-making discussing and evaluating
NON-TRANSACTIONAL DISCOURSE
	Genres Office gossip Small talk

Figure 19.1 Summary of Koester’s (2006) generic framework.

In accordance with a Hallidayan perspective, Koester quantitatively and qualitatively explores interpersonal meanings in both transactional and relational goal-oriented discourse; the interpersonal language features she analyses include modal verbs, idioms, vague language, hedges and intensifiers. The point is to understand the relationship between such features and the workplace genres. For example, in decision-making (a transactional genre), speakers tend to use particular modal verbs (an interpersonal language feature), such as have to much more frequently than other modals (e.g. should), and such differences can be explained by considering interpersonal issues like the perceived degree of obligation of the action and the relationship between the speakers. Also, modal verbs (like idioms) in general occur more frequently in collaborative genres like decision-making than in unidirectional or non-transactional discourse (see Figure 19.1).

In addition to specific language items, Koester also discussed the structural patterns in genres. For instance, she pinpoints the same ‘problem–solution’ pattern (Hoey 1983) that was found in Tribble’s analysis above, in decision-making episodes, arguing that such episodes usually begin with the speakers identifying some problem. She gives the following example from an American food cooperative, where the bookkeeper needs to read a handwritten item from a staff list:

Extract 1

Eventually, after over thirty turns, they decide the word is ‘sherry’. The first turn signals a problem through the word decipher, i.e. a word Ann cannot read. It is also worth noting that this turn contains the interpersonal modal verb wanna, and the request is worded grammatically as a question, a choice which can also be interpreted interpersonally. Ann could instead have worded the request, ‘Somebody decipher this now please,’ as the transactional goal is the same, but the interpersonal force would be very different and highly inappropriate in this context. By thus pinpointing typical language features in genres and speculating on other ways of representing the message, such features can cast light on how speakers are communicatively constrained by their communities, roles and goals.

5. What can a corpus tell us about non-institutional genres?

Some discourse is clearly non-institutional, as demonstrated by research into CSGs of everyday communication by Carter and McCarthy on the CANCODE corpus (McCarthy 1998; O’Keeffe et al. 2007), as well as that of Eggins and Slade (1997) who employ a systemic functional linguistics (SFL) approach to the analysis of gossip and other relational genres. Nevertheless, the boundary between institutional and non-institutional discourse is necessarily fuzzy. Given the participation of interlocutors whose professional identity is not salient, we could say there are also transactional genres that are not strictly institutional (i.e. neither wholly professional nor academic). CANCODE data, for example, has five contextual categories: intimate, socialising, pedagogical, professional and transactional (see McCarthy 1998). Examples include communication in the media between professionals and members of the public, such as O’Keeffe (2006), and studies of service encounters.

This section is based around McCarthy’s (2000) chapter on the genre of close contact service encounters, in this case a hairdressing encounter, which appears in Coupland’s edited collection Small Talk (Coupland 2000). The data is from the transactional section of the CANCODE corpus, and although McCarthy does not conduct keyword or concordance searches, he does provide quantitative comparisons of the amount of transactional and interpersonal language, which are discussed below. Nevertheless, the analysis of the text is largely qualitative, demonstrating how a corpus of whole texts and discourse analytical techniques can be integrated.

Service encounters are essentially transactional genres, but they often contain a lot of interpersonal discourse. For example, 92.5 per cent of the communication that takes place in the above encounter is interpersonal (McCarthy 2000) as it does not directly relate to the task at hand (in this case, the transaction of cutting hair) but instead is relational in nature. This is an important finding because it shows that in an apparently transactional genre like a hairdressing encounter, the vast majority of the language can be non-transactional. Such a finding has implications for the degree of rigidity we can reasonably apply in describing the discourse of a genre, and means that goals-driven descriptions of genre need to account for the interpersonal as well as transactional (Coupland 2000). It also raises the issue of constraint: such interpersonal communication may be expected by those concerned within this context, and an encounter that comprised purely transactional language could be perceived as inappropriate and uncomfortable.

McCarthy breaks the encounter down into eight stages, from the client being invited to ‘come through’, to her paying the bill and leaving. In terms of whether the stages are primarily transactional or interpersonal, McCarthy argues that they are all transactional, but contain varying degrees of interpersonal talk. This is particularly evident in Extract 2 from the second stage of the encounter, the initial discussion of how the client’s hair will be cut.

Extract 2

1.	<Server> Are you alright?
2.	<Client> I’m fine thanks and you?
3.	<Server> I’m fine thanks you yes {<Client> {Laughs}} are we cutting it as normal or 4 anything different or?
5.	<Client> Erm any suggestions or {laughs}?
6.	<Server> I always ask you that {laughs}.

(McCarthy 2000: 93)

In this extract, the cutter refers to the fact that the client has had her hair cut on numerous occasions by this cutter (lines 3–6). The encounter also begins with phatic greetings. Other stages, while clearly fulfilling a transactional goal, also contain a high degree of interpersonal, positively evaluative language, for example Oh that’s lovely … better … smashing. Such language is statistically more significant in CANCODE than in the business-discourse corpus CANBEC (Handford, forthcoming).

In terms of how speakers achieve their interpersonal goals through their use of language, it is evident in the above extract and throughout the encounter that both the client and the servers make considerable efforts to create a positive and friendly atmosphere. Apart from the aspects noted in the previous paragraph, in this extract we also see an interesting choice of pronouns (are we cutting it as normal) which creates a communal tone, even though, strictly speaking, the client is not involved in the act of cutting. There is a high degree of laughter, and when the cutter does make a suggestion it is hedged (I mean you could … sort of have a … like a sort of … ). Such hedging ensures that the suggestions do not appear too forceful, thereby addressing the negative face needs of the client and allowing her the option to disagree with the suggestions without threatening the positive face needs of the cutter (Brown and Levinson 1987).

Furthermore, it should also be remembered that the stages are all primarily transactional: the other 90 per cent of the encounter is made up of language that is wholly interpersonal. Such findings show how important the maintenance and development of the relationship is in such regular service encounters. For the client, we can infer that the hair-cutting ability and price are not the only factors in choosing a hairdresser: having a comfortable relationship with the cutter is important. Conversely, for the cutter, it seems that the efforts spent on developing the relationship will help ensure the customer returns. Such findings add weight to the argument that interpersonal language can be employed to fulfil transactional goals (Iacobucci 1990; Handford, forthcoming).

On a more abstract level, this linguistic behaviour arguably shows how the participants in this interaction are very much ‘conforming to type’: the client and servers, through drawing on their previous relevant experiences, both assume and recreate the social roles and relationships associated with the constraints of this social context. In so doing, they ratify the social practices that characterise this genre (such as ‘having your hair done’ in a provincial British women’s hairdressers). As such, we could say that such a genre approach highlights the dependence of texts (the language that is produced) on social context, and allows the analyst to make powerful inferences about what is allowed and what is controlled (that is, permitted and not permitted) in terms of the discourse features. For example, while it is possible that the speakers might baldly interrupt each other, doing so would probably be seen as highly inappropriate in this context (but not in other contexts, such as a cross-examination in a British court). Therefore we can say that the text is dependent very much on the social context in which it is embedded and to which it contributes.

This chapter has provided a flavour of the combination of genre and corpus approaches, and shown how CSGs can answer many of the criticisms levelled at corpus linguistics. Areas including the role and importance of context, fruitful combinations of quantitative and qualitative techniques, and the types of top-down and bottom-up approaches to the data have been discussed. In so doing, we have seen that corpora are not only large, decontextualised collections of words necessitating bottom-up, statistical approaches, but can be employed to explore nebulous issues in applied linguistics and discourse analysis. For instance, we have also seen how primarily transactional genres feature interpersonal language, and how goals and practices are evidenced in the data. The underlying theme of this chapter is that it is the social actions the participants (both expert and apprentice) perform in relation to the particular conventions and constraints which best explains what happens in genres, that is why speakers and writers do what they do. Well-designed and compiled CSGs can thus allow the analyst to uncover empirical evidence of the goals and practices that manifest themselves in the typical and therefore normative patterns in the discourse. By applying the compatible range of methods outlined here that corpus and genre approaches provide, we can attempt to see ‘the whole of the elephant’ (Bhatia 2004: xv).

In terms of what the future holds, it seems plausible that research may increasingly attempt to combine such qualitative and quantitative, genre and corpus approaches, but with increased emphasis on the insights that close linguistic analysis can offer on wider social practices and discourses (Gee 2005). One possible observation of much CSG-based research is that it could still go further in interpreting and explaining the contexts in which the texts occur. Fairclough’s (2000) keyword-based study of political language is one example of how such insights can be produced. Furthermore, according to Bhatia, a tendency in genre analysis can be towards oversimplification (2004), but basing descriptions on appropriately sampled, carefully analysed CSGs should enable the researcher to avoid questionable idealisations. One relatively unexplored area is that of corpora of professional discourse.

Through the processes of intertextuality and interdiscursivity (Bhatia 2004) there are new genres forming all the time, which have yet to be adequately described and interpreted. These include teleconferencing, videoconferencing, as well as internet-based genres like blogs and other cybergenres relating to eParticipation. Also, while each of the three texts examined here have been analysed in isolation, the reality is that they each form part of an ongoing genre network (Swales 2004). CSGs can therefore provide fecund datasets for these research fields.

References

Adolphs, S., Atkins, S. and Harvey, K. (2007) ‘Caught between Professional Requirements and Interpersonal Needs: Vague Language in Health Care Contexts’, in J. Cutting (ed.) Vague Language Explored. Basingstoke: Palgrave Macmillan, pp. 62–78.

Baker, P. (2006) Using Corpora in Discourse Analysis. London: Continuum.

Bargiela-Chiappini, F. and Harris, S. (1997) Managing Language: The Discourse of Corporate Meetings. Amsterdam: John Benjamins.

Bargiela-Chiappini, F. and Nickerson, C. (eds) (1999) Writing Business: Genres, Media and Discourses. Harlow: Longman.

Bazerman, C. (1994) ‘Systems of Genres and the Enhancement of Social Intentions’, in A. Freedman and P. Medway (eds) Genre and New Rhetoric. London: Taylor & Francis, pp. 79–101.

Berkenkotter, C. and Huckin, T. (1995) ‘Rethinking Genre from a Sociocognitive Perspective’, Written Communication 10: 475–509.

Bhatia, V. (1993) Analysing Genre: Language Use in Professional Settings. Harlow: Longman.

——(2004) Worlds of Written Discourse. London: Continuum.

Bhatia, V., Flowerdew, J. and Jones, R. (eds) (2008) Advances in Discourse Studies. London: Routledge.

Biber, D. (1988) Variation across Speech and Writing. Cambridge: Cambridge University Press.

Biber, D., Conrad, S. and Reppen, R. (1998) Corpus Linguistics: Investigating Language Structure and Use. Cambridge: Cambridge University Press.

Brown, P. and Levinson, S. (1987) Politeness: Some Universals in Language Usage. London: Cambridge University Press.

Candlin, C. and Hyland, K. (eds) (1999) Writing: Texts, Processes and Practices. London: Longman.

Christie, F. and Martin, J. (eds) (1997) Genres and Institutions. London: Continuum. Collins Cobuild Learners’ Dictionary (1987) Glasgow: Collins.

Connor, U. and Upton, T. (eds) (2004) Discourse in the Professions: Perspectives from Corpus Linguistics. Amsterdam: John Benjamins.

Coupland, J. (ed.) (2000) Small Talk. Harlow: Pearson Education.

Couture, B. (1986) Functional Approaches to Writing: Research Perspectives. Norwood: Ablex.

Csomay, E. (2005) ‘Linguistic Variation within Classroom Talk: A Corpus-based Perspective’, Linguistics and Education 15(3): 243–74.

Cutting, J. (ed.) (2007) Vague Language Explored. Basingstoke: Palgrave Macmillan.

Eggins, S. and Slade, D. (1997) Analysing Casual Conversation. London: Cassell.

Evison, J. (2008) ‘Turn-openers in Academic Talk: An Exploration of Discourse Responsibility’, unpublished PhD thesis, University of Nottingham.

Fairclough, N. (2000) New Labour, New Language? London: Routledge.

Flowerdew, J. (ed.) (2002) Academic Discourse. Harlow: Longman.

Flowerdew, L. (1998) ‘Concordancing on an Expert and Learner Corpus for ESP’, CALL Journal 8(3): 3–7.

——(2002) ‘Corpus-based Analyses in EAP’, in J. Flowerdew (ed.) Academic Discourse. Harlow: Longman, pp. 95–114.

——(2004) ‘The Argument for Using English Specialized Corpora to Understand Academic and Professional Settings’, in U. Connor and T. Upton (eds) Discourse in the Professions: Perspectives from Corpus Linguistics. Amsterdam: John Benjamins, pp. 11–36.

——(2008) ‘Corpora and Context in Professional Writing’, in V. Bhatia, J. Flowerdew and H. Jones (eds) Advances in Discourse Studies. London: Routledge, pp. 115–27.

Gee, J. P. (2005) An Introduction to Discourse Analysis. London: Routledge.

Ghadessy, M., Henry, A. and Roseberry, R. (eds) (2001) Small Corpus Studies and ELT: Theory and Practice. Amsterdam: John Benjamins.

Giddens, A. (1984) The Construction of Society: Outline of a Theory of Structuration. Cambridge: Cambridge University Press.

Halliday, M. (1985/1994) An Introduction to Functional Grammar. London: Edward Arnold. Handford, M. (forthcoming) The Language of Business Meetings. Cambridge: Cambridge University Press.

Hoey, M. (1983) On the Surface of Discourse. London: George Allen & Unwin.

——(2005) Lexical Priming. London: Routledge.

Holmes, J. and Stubbe, M. (2003) Power and Politeness in the Workplace. London: Longman.

Hunston, S. (2002) Corpora in Applied Linguistics. Cambridge: Cambridge University Press.

Hyland, K. (2000) Disciplinary Discourses: Social Interactions in Academic Writing. Harlow: Longman.

——(2003) Second Language Writing. Cambridge: Cambridge University Press.

——(2006) English for Academic Purposes. London: Routledge.

Hyland, K. and Milton, J. (1997) ‘Qualification and Certainty in L1 and L2 Students’ Writing’, Journal of Second Language Writing 6(2): 183–206.

Hyon, S. (1996) ‘Genre in Three Traditions: Implications for ESL’, TESOL Quarterly 30: 693–722.

Iacobucci, C. (1990) ‘Accounts, Formulations and Goal Attainment Strategies in Service Encounters’,in K. Tracy and N. Coupland (eds) Multiple Goals in Discourse. Clevedon, Avon: Multilingual Matters, pp. 85–99.

Koester, A. (2004) ‘Relational Sequences in Workplace Genres’, Journal of Pragmatics 36: 1405–28.

——(2006) Investigating Workplace Discourse. Routledge: London.

Lee, D. (2001) ‘Genres, Registers, Text Types, Domains, and Styles: Clarifying the Concepts and Navigating a Path through the BNC Jungle’, Language Learning and Technology 5: 37–72.

——(2008) ‘Corpora and Discourse Analysis: New Ways of Doing Old Things’, in V. Bhatia, J. Flowerdew and H. Jones (eds) Advances in Discourse Studies. London: Routledge, pp. 86–99.

McCarthy, M. J. (1998) Spoken Language and Applied Linguistics. Cambridge: Cambridge University Press.

——(2000) ‘Mutually Captive Audiences: Small Talk and the Genre of the Close-contact Service Encounters’, in J. Coupland (ed.) Small Talk. Harlow: Pearson Education, pp. 84–109.

McCarthy, M. J. and Handford, M. (2004) ‘“Invisible to Us”: A Preliminary Corpus-based Study of Spoken Business English’, in U. Connor and T. Upton (eds) Discourse in the Professions: Perspectives from Corpus Linguistics. Amsterdam: John Benjamins, pp. 167–201.

Martin, J. (1992) English Text: System and Structure. Amsterdam: John Benjamins.

——(1997) ‘Analysing Genre: Functional Parameters’, in F. Christie and J. Martin (eds) Genre and Institutions: Social Processes in the Workplace and School. London: Continuum, pp. 3–39.

O’Keeffe, A. (2006) Investigating Media Discourse. London: Routledge.

O’Keeffe, A., McCarthy, M. J. and Carter, R. A. (2007) From Corpus to Classroom: Language Use and Language Teaching. Cambridge: Cambridge University Press.

Sarangi, S. and Roberts, C. (eds) (1999) Talk, Work and Institutional Order. Berlin: Mouton de Gruyter.

Scollon, R. and Scollon, S. W. (2001) Intercultural Communication: A Discourse Approach. 2nd edn. Oxford: Blackwell.

Scott, M. (1996) WordSmith Tools. Oxford: Oxford University Press.

Simpson, R. (2004) ‘Stylistic Features of Spoken Academic Discourse: The Role of Formulaic Expressions’, in U. Connor and T. Upton (eds) Discourse in the Professions: Perspectives from Corpus Linguistics. Amsterdam: John Benjamins, pp. 37–64.

Sinclair, J. (1991) Corpus, Concordance, Collocation. Oxford: Oxford University Press.

Stubbs, M. (1996) Text and Corpus Analysis. London: Blackwell.

Swales, J. (1990) Genre Analysis. Cambridge: Cambridge University Press.

——(2002) ‘Integrated and Fragmented Worlds: EAP Materials and Corpus Linguistics’, in J. Flowerdew (ed.) Academic Discourse. Harlow: Longman, pp. 150–64.

——(2004) Research Genres: Explorations and Applications. New York: Cambridge University Press.

Swales, J. and Lee, D. (2006) ‘A Corpus-based EAP Course for NNS Doctoral Students: Moving from Available Specialized Corpora to Self-compiled Corpora’, English for Specific Purposes 25: 56–75.

Tognini Bonelli, E. (2001) Corpus Linguistics at Work. Amsterdam: John Benjamins.

Tribble, C. (2002) ‘Corpora and Corpus Analysis: New Windows on Academic Writing’, in J. Flowerdew (ed.) Academic Discourse. Harlow: Longman, pp. 131–49.

Vine, B. (2004) ‘Modal Verbs in New Zealand English Directives’, Nordic Journal of English Studies 3(3): 205–20.

Wenger, E. (1998) Communities of Practice: Learning, Meaning and Identity. Cambridge: Cambridge University Press.

Widdowson, H. (1998) ‘Context, Community and Authentic Language’, TESOL Quarterly 32(4): 705–16.

——(2000) ‘On the Limitations of Linguistics Applied’, Applied Linguistics 21: 3–25.

19What can a corpus tell us about specialist genres?