S. Eggins & D. Slade Analysing Casual Conversation.London: Cassell, 1997. Chapter 2. RELEVANT APPROACHES TO ANALYSING CASUAL 2.

1 INTRODUCTION Although, as we pointed out in Chapter 1, casual conversation has generally received limited analytical attention, conversation (as a general label for spoken interactive discourse) has been more fortunate. In fact, conversation has been analysed from a variety of perspectives, with sociological, philosophical, linguistic and critical semiotic approaches all making important contributions towards understanding the nature of spoken discourse. In this chapter we briefly review key ideas from the approaches we consider most relevant to the analysis of casual conversation, suggesting that an eclectic approach to analysing casual conversation is not only richer but also essential in dealing with the complexities of casual talk. 2.2 A TYPOLOGY OF RELEVANT APPROACHES TO ANALYSING CASUAL CONVERSA TION The pervasiveness of spoken interaction in daily life has made it an interesting domain of study for researchers with backgrounds in ethnomethodology, sociolinguistics, philosophy, structural-functional linguistics and social semiotics. At various times and in various ways, analysts from all these perspectives have sought to describe aspects of how talk works. Within ethnomethodology, new ways of thinking about conversation emerged in the 197S from Conversation Analysis, notably the work of Sacks, Schegloff, Jefferson and their followers. Sociolinguistic approaches arise from the interdisciplinary connections between sociology / anthropology and linguistics. From sociolinguistics we have contributions emerging from the work of Hymes in the Ethnography of Speaking, and Cumperz in Interactional Sociolinguistics, including the work of Labov and associates in Variation Theory. From a more logical-philosophical perspective, both Speech Act Theory and Pragmatics have added important insights to our understanding of how people interpret conversation.' And within linguistics, the study of conversation has been pursued most actively by approaches interested in both the structure and the function of authentic discourse, notably the Birmingham School and Systemic Functional Linguistics. More recent perspectives have emerged from social semiotic orientations which arise from interdisciplinary connections between linguistics and critical and cultural theory, including Critical Linguistics and Critical Discourse Analysis (CDA). Figure 2.1 provides a brief typology of the approaches we consider most relevant to analysing casual conversation. However, although these perspectives have all

contributed to ideas about spoken interaction, relatively few of them have specifically addressed the challenge of analysing casual conversation. We have therefore found it most useful to adopt a rather eclectic theoretical base, drawing on insights from all these different approaches, but with particular reference to Conversation Analysis (1), Systemic Functional Linguistics (4b), and Critical Discourse Analysis (5b). From some approaches we take perspectives on the microstructuring of casual conversation, including the analysis of the Iocalistic organization of turn-taking from Conversation Analysis, itemization of linguistic features relevant to variation in conversational style from Interactional Sociolinguistics, the production and interpretation of speech acts from Speech Act theory and Pragmatics, and the grammatical, semantic and discourse characteristics of casual talk from Systemic Functional Linguistics. CONVERSATION ethnomethodological Conversation Analysis (CA) Ethnography of Speaking sociolinguistic Interactional Sociolinguistics Variation Theory logico-philosophic Speech Act Theory Pragmatics structural-functional Birmingham School Systemic Functional Linguistics (SFL) social semiotic Critical Discourse Analysis (CDA) Critical Linguistics Figure 2.1 Relevant approaches to analysing casual conversation From other approaches we take perspectives on the more macro patterns of text types or genres. The concept of genre stems from work in traditional literary studies, but for our purposes we are concerned with definitions of genre which have relevance to casual conversation. The term genre is used mainly in Ethnography of Speaking, Systemic Functional Linguistics and Critical Discourse Analysis. In Ethnography of Speaking, genre refers to one component in the complex communicative context of interactions. In Systemic Functional Linguistics, the term has been used to describe how people use language to achieve culturally recognized goals. In Critical Discourse Analysis genre is defined as 'a socially ratified way of using language in connection with a particular type of social activity (e.g. interview, narrative exposition)' (Fairclough 1995a: 14). In Variation Theory, although the term genre is not used, the notion of 'overall' text structure, corresponding to the notion of generic structure, has been central to much of Labov's (1972a) work on

discourse." In the following sections we review the contributions we take from each approach listed in the typology. 2.3 SOCIOLOGICAL PERSPECTIVES ON CASUAL CONVERSATION: ETHNOMETHODOLOGY AND CONVERSATION ANALYSIS (CA)3 It is one of the sharper ironies of modern linguistics that the founding work in conversation analysis was done outside the field of linguistics by a group of American sociologists working during the 1960s and 197s. Their approach to conversation was strongly influenced by the sociologist Harold Garfinkel's concern to understand how social members themselves make sense of everyday life. The "ethnomethodology" he developed centred on: 'paying to the most commonplace activities of daily life the attention usually accorded extraordinary events' (1967: 1). The early conversation analysts, sociologists such as Sacks, Schegloff, Jefferson and their successors, combined a concern with following a rigorously empirical methodology with the ethnomethodological aim of finding methods for making the commonsense world visible.4 In the study of talk, this meant: 'an insistence on the use of materials collected from naturally occurring occasions of everyday interaction' (Atkinson and Heritage 1984: 2). Conversation Analysis (CA), as a branch of ethnomethodology, focuses on

conversation because it offers a particularly appropriate and accessible resource for ethnomethodological enquiry: Seeing the sense of ordinary activities means being able to see what people are doing and saying, and therefore one place in which one might begin to see how making sense is done in terms of the understanding of everyday talk. (Sharrock and Anderson 1987: 299) In trying to explain how it is that conversation can happen at all, Sacks et al. (1974: 700) found it necessary to account for two "grossly apparent facts" that they observed about spoken interactive data. These facts are that:

a)only one person speaks at a time. b) speaker change recurs. These two "facts" lie behind our commonsense observation of conversation, that it is, fundamentat1y, a turn-taking activity. In trying to explain how it is that speakers keep taking turns, the Conversation Analysts modelled conversation as a generative mechanism, designed to fulfil two distinct functions. First, speakers have to be able to work out when it is appropriate to transfer the role of speaker. Second, there has to be a way of determining who the next speaker is to be. For example, in text 1.1, analysis has to account for how a change of speaker occurs once S I (David) has said Dead space in the conversation, and how it is that there is a largely orderly transfer of the speaker role, in this case to S2 (Liz). Sacks et al. (1974) suggested that speakers recognize points of potential speaker change because speakers talk in units which they called Turn Constructional Units (TCUs). They defined a TCU as a grammatically complete unit ofIanguage, such as a sentence, clause or phrase, the end of which represents to the interactants a point at which it is possible for speaker transfer to occur. Thus, in text 1.1, when David has completed his sentence, Liz knows it is appropriate to take the turn, as a range of cooccurring factors such as falling intonation, grammatical structure of a completed sentence, posture and gaze, suggest that in this instance the sentence is a TCU. Had Liz begun speaking just as David had said This has been, then her talk would have been interpreted as an interruption, rather than orderly turn-taking, as she would have been beginning her talk where there was no TCU boundary. The second problem of who should be the next speaker also depends on TCUs. Sacks et al. (1974) note that at the end of a TCU there are two possibilities for determining the allocation of turns. The first possibility is that the current speaker selects the person who is to be the next speaker. Typical ways by which the current speaker can select the next speaker include the use of names or vocatives, gaze, posture, and the targeting of moves such as directing questions to particular interactants. Thus, for example, in text 1. 1, when Nick asks Where's the cigarettes, David? he has selected David as the next speaker both by the use of the vocative and by formulating a question. While it would be possible for anyone else present to take the role of speaker at this point, their contributions would be considered to interrupt the orderly transition of speaker role to David. The second possibility is that the next speaker may self-select. Thus, for example, when Liz comes in with In France they say an angel is passing, she does so without having been nominated or invited to speak by David. This system of turn allocation is summarized in Figure 2.2.

Sacks et at. (1974) point out that this system operates at the end of every turn or what they call "locally", rather than on an overall or "global" basis. In other words, turn allocation cannot be agreed in advance at the beginning of the conversation, but must be continually renegotiated at each TCD boundary. The system has one aim: to ensure that when the current speaker finishes her turn at talk, some other speaker will start talking. The Conversation Analysts thus modelled conversation as an infinitely generative turn- taking machine, whose design suggests that the major concern of interactants is to avoid lapse: the possibility that no one is speaking point of possible turn transfer current speaker selects next speaker next speaker self-selects selects a different speaker selects self The turn-taking system Source: derived from Sacks et al. 1974 Given that conversation is driven by a turn-taking machinery expressly designed to keep going, it becomes quite problematic to determine how a conversation could ever stop. It was in exploring an answer to this issue of conversational closure that CAs made what many see as their most significant contribution to the analysis of interaction: the identification of the adjacency pair. Consider the following pair of turns from text 1.1 in Chapter 1: 11b Nick [turns to David] Where's the cigarettes, David? 12b David Sorry, Nick. I've cut you off. Sacks et at. (1974) noticed that the occurrence of the second turn can be explained by the first: a question in some ways implies that the next turn will be an answer. Conversational sequences where there was some kind of "special relatedness" operating between a~jacent utterances as in this example are called adjacency pairs. They were identified as typically having three characteristics:
. . .

a) two utterance length b) adjacent positioning of component utterances c) different speakers producing each utterance. (Schegloff and Sacks 1974: 238)

The classic adjacency pair is the question/answer sequence, although other adjacency pairs include complaint/denial; compliment/rejection; challenge/ rejection; request/grant; offer/accept; offer / H;ject and instruct/receipt (Sacks et al. 1974: 717). Adjacency pairs were identified as functioning as a turn-transfer technique, i.e. they

function both to allocate the next turn, and to exit from the current turn. It is important to understand that the system is not one of determination, but of expectation. Given that a first pair part (the first utterance of the adjacency pair) has been produced, there is a very strong likelihood that the addressed participant will be the next speaker, and will produce a relevant second pair part. The Conversation Analysts recognized that in fact there are two types of second pair parts. There is first a preferred second pair part. For example, in text 1.1, had David responded to Nick with Here, helP yourself, he would have been providing the preferred second pair part: the compliance predicted by Nick's initiating demand. However, as David's actual response indicates, it is always possible for the addressee to produce some kind of discretionary alternative - what in CA terms is referred to as a dispreferred second pair part. Preferred responses tend to be briefer, linguistically simpler, supportive or compliant and oriented towards closure. Dispreferred responses tend to be longer as respondents may seek to apologize, explain or otherwise justify their dispreferred response. For example, in text 1.1 David produces four utterances in his refusal of Nick's request, providing an apology and several explanations. Dispreferred responses are therefore linguistically more complex, and involve noncompliance or conflictual action. The identification of the adjacency pair was the basis for two further developments in CA. The first was the recognition of sequences longer than two units, and the second was the formulation of the theoretical concept of sequential implicativeness. Consider the following sequence of turns from text 1 .4 in Chapter 1:
. . . . . . .

6 C and have you got ... the... first day covers of ... 7 S yes 8 C (Anzac) [2 secs] 9 S 10 C 11 S 12 C

13 S

how many would you like four please two of each? what have you got theres two different designs on the- [5 secs S shows C the covers]
. .

14 C I ll take 15 S uhum

two of each Here we see that turns 6-7 constitute a question/answer adjacency pair. However, this pair is followed by a sequence of seven turns which in some sense seem to "go together" or have a particular relatedness to each other. It takes these seven utterances for the salesperson's question (How many would you like?) to be resolved so that the customer's initial request can be complied with. Instances such as these prompted CA to recognize the existence of sequences, of which the adjacency pair was merely the minimal version. In their description of sequences, CA has frequently concentrated on those which are particularly "visible" in that they interrupt, suspend, or prepare for the ongoing interaction. The types of sequences referred to most frequently in the CA literature are: insertion sequences (Schegloff 1972), pre-sequences (Schegloff 1980), side sequences (Jefferson 1972), closing sequences (Schegloff and Sacks 1974), and repair or clarification sequences (Schegloff et at. 1977). From the analysis of adjacency pairs and turn sequences came a recognition of the more general principle underlying conversational organization: that of the sequential relevance or sequential implicativeness of talk (Schegloff and Sacks 1974: 296). This is the notion that conversational turns make sense because they are interpreted in sequence. As Atkinson and Heritage point out: 'no empirically occurring utterance ever occurs outside, or external to, some specific sequence. Whatever is said will be said in some sequential context' (Atkinson and Heritage 1984: 6). CA suggests that the most significant placement consideration in conversation is that of adjacency. Thus, wherever possible the speakers current turn will be interpreted as implicating some action by the responder in the immediate next turn. Similarly, the respondents subsequent talk will, where possible, be interpreted as related, immediately prior turn. Thus, adjacency pairs can be seen as merely the prototypical variety of the general conversational principle of sequential relevance.

This principle explains why we will work extremely hard to interpret any two adjacent turns as related, despite the absence of any other indications of cohesion. Consider, for example, the two turns at talk below: Turn Speaker 1 A What's that floating in the wine? 2 B There aren't any other solutions.5 You will try very hard to find a way of interpreting B's turn as somehow an answer to A's question, even though there is no obvious link between them, apart from their appearance in sequence. Perhaps you will have decided that B took a common solution to a resistant winecork and poked it through into the bottle, and it was floating in the wine. Whatever explanation you came up with, it is unlikely that you looked at the example and simply said "it doesn't make sense", so strong is the implication that adjacent turns relate to each other. Similarly, sequential implicativeness explains why, if a speaker does not wish an utterance to be interpreted as related to immediately preceding talk, s/he needs to state that explicitly, using expressions such as To change the subject.. ., or By the way... . This orientational perspective of talk is what gives conversation its essential nature as a dynamic process of recipient design, another important concept developed by CA and well summarized by Taylor and Cameron: My behaviour is designed in light of what I expect your reaction to it will be: i.e. you will react to it as conforming to the relevant rule or as in violation of it, thereby leading you to draw certain conclusions as to why I violated the rule... Thus, by the inexorable fact that interactions progress, any component action inevitably is temporally situated in a sequential context, a context to which it is an addition and within which it will be interpreted, held accountable and responded to in turn. (1987: 103). At the same time as it focuses close up on the turn-taking organization of talk or who gets to be speaker, CA has always recognized that topic management (what people talk about when they do get to be speaker) is a distinct, though interrelated, aspect of conversational organization. In asking how people manage to get their topics made "mentionable" in a conversation, Schegloff and Sacks (1974), and later others (e.g. Maynard 19Ho) have developed an account of topic placement and topic fitting which shows the interaction of local and overall conversational structure with topic. Through notions such as step-wise topic progression, topic shift, and topic change, CA has tried to categorize the apparently "natural" or smooth procedures speakers use to progress from one topic to another (Sacks et al.

1974, Schegloff and Sacks 1974, Maynard 1980, Jefferson 1984). Although the focus in CA has not been on contrasting different types of spoken interactions, there are references in Sacks et al. (1974) to the relationship between casual conversation and other types of turn-taking interactions. Casual conversation can be positioned at one end of a continuum which reflects the degree to which turns are pre- allocated in an interaction (Sacks et al. 1974: 729). The post office interaction of text 1.4 would presumably occur towards the other end of this pole, since turn allocation is largely predetermined by the nature of the social task. In general, CA work has focused very much on micro structural issues, rather than on the larger, macro-structure of conversation. Thus although Sacks (1992 b: 17-32) discusses some of the characteristics of extended talk, i.e. talk that "goes on over more than a single turn" (ibid: IH), his account deals with the micro-features of the talk, rather than with the overall structure of such segments. Jefferson and Lee (1992), in their article on the analysis of "conversations in which 'troubles' are expressed", argue in relation to their conversational corpus that: Although many of the conversations were long and multifaceted, they were not amorphous. There seemed to be a shape to them; a shape which recurred across the range of conversations; a shape which could be sensed to be rather well formed in some of the conversations and distorted or incomplete in others. (521-3) Although they do not use the term 'genre', they discuss a similar phenomenon when they describe the global text structure of talk about troubles: we had a strong, if vague, sense of troubles talk as a sequentially formed phenomenon, a seed collection of elements which might constitute the components out of which a troubles-telling "sequence" could be constructed, and a set of categories which might distribute the components across appropriate speakers. In short, we had the basis for a troubles-telling sequence. (Jefferson and Lee 1992: 522) They argue, however, that although this underlying abstract structure exists, participants negotiate their way through the structure and regularly disrupt it. Thus there are ideal types or 'templates' (ibid: 524) which can be described, but in reality interactants regularly depart from them. Influenced by this and the Critical Discourse Analysis perspective (discussed below), our description of the storytelling (Chapter 6) and gossip genres (Chapter 7) is an account of the underlying abstract structure that speakers, in particular cultural

contexts, orient to. The generic descriptions we present, however, are not to be interpreted as a fixed or rigid schema. Our data support the findings of Jefferson et al. that in casual conversation there are often disruptions to the generic flow, but we will be suggesting that the reason that we can recognize these deviations is precisely because there is an underlying abstract structure to each generic type. A further major contribution of CA has been to make everyday interaction a worthy subject of academic research. Not only have their "discoveries" about conversation drawn attention to the many insights to be gained from its detailed analysis, but they have also offered a powerful way of thinking about casual talk, by emphasizing that it is a dynamic creation of interacting and co-operating participants: The discourse should be treated as an achievement; that involves treating the discourse as something 'produced' over time, incrementally accomplished, rather than born naturally whole out of the speaker's forehead.... The accomplishment or achievement is an interactional one... it is an ongoing accomplishment, rather than a pact signed at the beginning. (Schegloff 1981: 73) The strength of the CA observations of conversation comes in part from the fact they are always based on actual recorded data of naturally occurring interactions, transcribed in meticulous detail. Believing that intuition is an extremely unreliable guide for work in conversation, CA has always rejected experimental methods of collecting conversational data such as simulating dialogues or setting up artificial interactive contexts, and has challenged discourse analysts to access the data offered by everyday social life. The debts owed to early CA by all subsequent approaches to conversation analysis (and to discourse analysis more generally) cannot be overstated, and CA is still a thriving perspective on interaction a(. collections such as Roger and Bull (1989), Boden and Zimmerman (1991) and Drew and Heritage (1992) attest. However, despite its many contributions, it has three major drawbacks in the analysis of the casual conversational data which we are concerned with. These drawbacks are its lack of systematic analytical categories, its "fragmentary" focus, and its mechanistic interpretation of conversation. Lack of systematicity is a problem for all aspects of the CA account of conversation. For example, CA has not provided us with an exhaustive list of all adjacency pairs in English, nor, more seriously, has it made clear exactly how adjacency pairs might be recognized. In addition, while CA identified TCUs as the critical units of conversation, it has not specified exactly how a TCU boundary can

be recognized in anyone situation. One of the major problems with this lack of systematicity is that comprehensive quantitative analysis becomes impossible. Yet quantitative analysis is just what is necessary to give empirical validity to claims that conversations are typically organized in particular ways, and to provide evidence of any statistically significant variations in conversational behaviours. To develop such systematic analyses we believe it is necessary to draw very specifically on linguistic expertise, employing linguistic methodologies to relate aspects of conversational organization to aspects of the organization of language as a whole. We will suggest in later chapters that specitying criteria for the identification of adjacency pairs requires the use of linguistic categories, terms and analyses. The systematic relationship between categories of first pair parts and their preferred and dispreferred second pair parts needs to be specified in grammatical terms. Specifying TCU boundaries depends on being able to describe the co-occurrence of linguistic patterns, involving rhythm, intonation, grammatical structure and semantics. The second drawback is that while the close up focus on small excerpts of talk has been responsible for CA's major discoveries about conversation, CA is limited in its ability to deal comprehensively with complete, sustained interactions. While Sacks used fragments to uncover the social meanings being achieved through talk (e.g. how affiliation is achieved, how category membership is determined), much subsequent work in CA has focused on more mechanistic concerns (e.g. the "precision timing" of turn- taking, etc.). This has meant that the reality of conversations (that many are very long and indefinitely sustainable) has not been addressed. A further explanation for this concerns the data CA often works on. While all the data are recorded in authentic situations, relatively little of it constitutes casual conversation as we have defined it in this chapter. Thus, some of the very issues which are central for us do not arise for CA. The final drawback is that while on the one hand CA offers a powerful interpretation of conversation as dynamic interactive achievement, on the other hand it is unable to say just what kind of achievement it is. Modelling conversation as a machine does not explain adequately just what interactants use the machine for, nor how the machine functions in relation to macro social structures. To address these issues requires a shift of orientation away from conversation as a form of social interaction that is incidentally verbal, and towards conven,ation as a linguistic interaction that is fundamentally social. Rather than seeing conversation

merely as good data for studying social life, analysis needs to view conversation as good data for studying language as it is used to enact social life. We now review some of the most relevant linguistic approaches to analysing casual conversation. 2.4 SOCIOlINGUISTIC APPROACHES TO CONVERSATION We group together as "sociolinguistic" approaches to discourse analysis those approaches which may have disparate multi-disciplinary origins, but which share in their practice an orientation to the use of language in the social contexts of everyday life. 2.4.1 Ethnography of Speaking: conversation and cultural context Ethnographic approaches to conversation have been led by Dell Hymes (1972b, 1974) and are concerned with understanding the social context of linguistic interactions. In seeking to account for "who says what to whom, when, where, why, and how" (Hymes 1972b), Hymes developed a schema for analysing context that has as its prime unit of analysis the speech event in which language occurs: The speech event is to the analysis of verbal interaction what the sentence is to grammar... It represents an extension in the size of the basic analytical unit from the single utterance to stretches of utterances, as well as a shift in focus from... text to ... interaction. (I7). The term "speech event" refers to "activities... that are directly governed by rules or norms for the use of speech" (Hymes 1972b: 56). Speech events include interactions such as a conversation at a party, ordering a meal, etc. Any speech event comprises several components and these are listed in the grid in Table 2.1. With each letter acting as an abbreviation for a different component of communication, Hymes's grid has become known as the "SPEAKING grid". Table 2.1 Hymes's Speaking Grid (Hymes 1972b) S setting scene P participant E ends A act sequence K key I instrumentalities N norms of interaction and interpretation G genre

temporal and physical circumstances subjective definition of an occasion speaker / sender/addressor /hearer / receiver / audience/ addressee purposes and goals outcomes message form and content tone, manner channel (verbal, non-verbal, physical forms of speech drawn from community repertoire) specific properties attached to speaking interpretation of norms within cultural belief system textual categories Hymes argues that the values of the factors identified in the SPEAKING grid on any specific occasion determine our use of language and our interpretation of what people say. It is the analysis of these components of a speech event that is central to what became known as the "ethnography of communication" or the "ethnography of speaking", with the ethnographer's aim being to discover rules of appropriateness in speech events. As the SPEAKING grid shows, Hymes used the term "genre" to refer to just one component of the speech event. Defining genre as including such categories as joke, story, lecture, greeting, and conversation, Hymes argued that: Genres often coincide with speech events, but must be treated as analytically independent of them. They may occur in (or as) different events. The sermon as a genre is typically identified with a certain place in a church service, but its properties may be invoked, for serious or humorous effect, in other situations. (Hymes 1974: 61) The SPEAKING grid provides a necessary reminder of the contextual dimensions operating in any casual conversation. These contextual dimensions are similar to the systemic analysis of register, discussed later in this chapter. Hymes's analysis of genre has not been applied to conversation, perhaps because of the lack of explicitness in his account of the relationship between genre and the other components of the SPEAKING grid, and their expression in language. However, the ethnographic framework he initiated led not only to broader notions of the "communicative competence" language users display, but also to a recognition of the close relationship between speech events and their social! cultural contexts.

2.4.2 Interactional Sociolinguistics: the contextualization of discourse Interactional Sociolinguistics is the label associated with approaches to discourse which grow out of the work of the anthropologist John Gumperz (1 982a, b), strongly influenced by the sociologist Erving Goffman (19.59, 1967, 1974, 1 9H 1). Like Hymes, GUlJ1perz was centrally concerned with the importance of context in t6"e production and interpretation of discourse. Through detailed analyses of grammatical and prosodic features in interactions involving interracial and interethnic groups (e.g. interactions between British and Indian speakers of English in England, see Gumperz 1 982a), Gumperz demonstrated that interactants from different socio-cultural backgrounds may "hear" and understand discourse differently according to their interpretation of contextualization cues in discourse. For example, intonation contours may be interpreted by some interactants as indicating rudeness and aggression, while for others such intonation patterns denote deference and consideration. As Gumperz explains, our interactions take place against the background of, and are critically affected by, our socio-cultural context: What we perceive and retain in our mind is a function of our culturally determined predisposition to perceive and assimilate. (1982a: 4) In other words, in our participation in discourse events we are always bound by our cultural context. Because we interact with orientations only to those contextualization cues that our cultural conditioning prepares us for, miscommunication can occur when we come into contact with interactants who do not share our cultural context. Gumperz's work is strongly focused on exploring the contextualization cues operating in different sociocultural "styles", in order to identity and explain sources of what are referred to as "communication difficulties". While Gumperz did not focus on the analysis of casual conversation, the most relevant empirical applications of his work to casual conversation include studies by Deborah Tannen (19H4, 19H9, 199o) and Deborah Schiffrin (198sa,b, 1987, 199O), each of whose work we consider briefly below. 2.4.2. 1 Analysing style in casual conversation Asking questions about overall characteristics of conversational discourse, rather than about its sequential organization, Tannen's approach to conversation draws not only on Hymes's ethnography, but also on Gumperz's work on cross-cultural perspectives on discourse (cf. Gumperz and Hymes 1964, 1972; Gumperz 1982a, b). Her work results in a description in terms of cultural preferences or conventions. Despite the fact that Tanncn uses data which fall clearly within the scope of our

definition of casual conversation presented in Chapter 1 (Tannen analysed dinner party talk among friends), her work differs from ours in both its aims and approach. Tannen (l984) describes her underlying investigation as asking: How do people communicate and interpret meaning in conversation? (7) Her study involves the identification and characterization of different conversational stylcs in terms of the kind of "rapport" they convey. She suggests that this description of conversational styles is: a step toward the goal of understanding conversational interaction: what accounts for the impressions made when speakers use specific linguistic devices? What accounts for the mutual understanding or lack of it in conversation? (7) Tannen identifies a number of stylistic devices as an indication of alternative strategies in creating rapport, and relates the stylistic variation to Lakoff's "Rules of Rapport": I. Don't impose (Distance) 2. Give options (Deference) 3. Be friendly (Camaraderie). (Lakoff in Tannen 1984: 11) The features Tannen identifies include: topic choice, pacing, narrative strategies, and "expressive paralinguistics" such as pitch and voice quality. Tannen's analysis results in statements of preferences and conventions. For example, preferences under the "narrative strategies" category include: a. Tell more stories b. Tell stories in rounds. c. Prefer internal evaluation (i.e. point of a story is dramatized rather than lexicalized). (30-31). d. Tannen suggests that the use of these features characterizes a "high involvement style" (1984: 31), and she compares two groups of interactants at her dinner party in terms of whether they used and expected this high involvement style or, instead, a high considerateness style. Tannen's focus on different interactive styles is extremely useful in highlighting the importance of variation in conversational behaviour. Analysis of conversational style reveals that we are not talking about individual speakers so much as groups of speakers sharing common styles. Interactive styles provide recognition criteria for subcultural groups, and indicate dimensions of difference that are significant for cultural members. These dimensions reflect the powerful stratifying dimensions in

that culture - typically gender, race, ethnicity and social class. However, the identification of conversational styles which can be empirically correlated with particular social groups demands careful explanation. If, for example, we identify women as showing a preference for certain conversational strategies, and if those preferences differ from men's conversational strategies, we need to explain just what conversational styles mean. While we share Tannen's interest in uncovering and understanding conversational variation, we reject her suggestion that style differences in conversational behaviour are no more than indications of equal but different modes of behaving. Tannen develops this thesis most clearly in her popular work on male and female conversational behaviour (Tannen 1990), and it relates to what is referred to as the "dominance/difference" debate in studies in language and gender (Cameron 1992). While "difference" advocates (for example Tannen) suggest that subcultural groups, such as women, have different preferred ways of interacting, they are uncritical of the implications of those differences. "Dominance" theorists, on the other hand, point out that some groups are heavily disadvantaged by their conversational style: i.e. that the conversational styles of some groups have unfavourable material consequences for their members (see, for example, Fishman 1980). Thus, speakers whose style is low in conversationally assertive strategies are less likely to get floor time, less likely to be heard seriously, and less likely to control the topic. It is not just that their style may be different from the more assertive strategies used by their co-interlocutors, but that their style of interacting may serve the interests of their co-interlocutors better than it serves their own interests. The approach to conversational style thatwill be taken in this book is informed by social semiotics (see critical linguistics/critical discourse analysis, discussed below). By interpreting style in more critical terms, as socially acquired orientations towards particular semantic choices, we are able to explore the material consequences and political implications of particular stylistic features and generic patterns. 2.4.2.2 Analysing conversational continuity A more micro-focused application of quantitative interactive sociolinguistic analysis is found in the work of Schiffrin (1 98sa, b), especially Schiffrin' s (1987) study of discourse markers. Unlike Tannen's study, which provides an overall characterization of features of conversational talk, Schiffrin's work is localistic in focus, centred on the turn as a basic unit. Her perspective also involves her in issues of sequential organization, since she states her basic concern to be with "the accomplishment of conversational

coherence" (Schiffrin 1985a: 640). Schiffrin asks: How can what one speaker says be heard as following sensibly from what anothcr has said? She then draws on CA insights, to focus on the recipient design of utterances: i.e. on how speakers design their utterances so that they are accessible to their listeners. As an explicit example of this recipient design, she looks at a number of "discourse markers", which she defines as: 'sequentially dependent elements which bracket units of talk' (31). Into this category Schiffrin groups a range of items, including: oh, well, and, but, or, so, because, now, then, y'know, and I mean . As texts 1.1 to 1.3 in Chapter 1 show, these markers occur extremely frequently in casual conversations. Using a corpus collected through peer-group interviews (thus, not drawn from a naturally occurring situation), Schiffrin argues for the importance of both qualitative and quantitative/distributional analysis in order to determine the function of the different discourse mar~rs in conversation. Schiffrin's analysis of the function and distribution of a limited number of linguistic devices in conversation offers useful insights into questions of structure, or which turns at talk are signalled as structurally related to others, and of continuity. For example, her discussion of oh finds a common function in its various distributional contexts, as a marker of information management, which "initiates an information state transition" (99). The use of oh indicates shifts in speaker orientation to information. These shifts occur as speakers and hearers manage the flow of information produced and received during discourse (100-l01). From Schiffrin's work we take specific insights into the function of particular items to signal special sequential relatedness in talk, information which is very relevant in determining the boundaries of conversational exchanges (see below). We also take seriously her concern to explore conversational patterns quantitatively as well as qualitatively. One criteria we demand of the analyses we develop in Chapters 3 to 6 is that results can be quantified, so that sections of one casual conversation can be compared with sections of others. This comparison gives substance to any claims we make about patterns in casual conversation and its varieties. 2.4.3 Variation Theory: texts-in-talk A different sociolinguistic perspective which we find relevant to conversational analysis is Variation Theory. This was initially developed by Labov (see, for example, 197n). Although the major part of Labov's contribution was in the analysis

of phonological variation, he has also been involved in important work on the structure of texts within conversations. Of most relevance to our work is the detailed study of narratives, first presen ted in Labov and Waletzky (1967). Labov and Waletzky argue that fundamental narrative structures are evident in spoken narratives of personal experience. Although their analysis is based on narratives collected in interviews (and thus not examples of spontaneous conversation), their findings can be usefully extrapolated to an analysis of narratives occurring in spontaneous conversation. They were concerned with relating the formal linguistic properties of narratives to their function. They argue that the 'overall structure' of a fully formed narrative of personal experience involves six stages (1967: 32-41): 1. Abstract 2. Orientation 3. Complication 4. Evaluation 5. Resolution 6. Coda In this structure the Abstract and Coda stages are optional; and both the Orientation and Evaluation may be realized either before or as part of the Complication and Resolution respectively. Apart from these variations, the stages must occur in this sequence for a text to be a successful narrative. These structural labels are demonstrated in one ofLabov's texts collected in a sociolinguistic interview (Labov 1972a: 3SR-9). Speaker A is the interviewer and speaker B responds with the narrative. Text 2.1 The Fight
.

A: What was the most important fight that you remember, one that sticks in your mind? B: Well, one 1 think was with a girl.

Abstract Orientation Like I was a kid, you know, and she was the baddest girl, the baddest girl in the neighbourhood. If you didn't bring her candy to school, she would punch you in the mouth; and you had to kiss her when she'd tell you. This girl was only about 12 years old, man, but she was a killer. She didn't take no junk; she whupped all her brothers. And I came to school one day and 1 didn't have no money. My ma wouldn't give me no money... So I go to school and this girl says "Where's the candy?" I said "I don't

have it." She says, powww! So 1 says to myself, "There's gonna be times my mother won't give me money because (we're) a poor family. And I can't take this all, you know, every time she don't give me any money." So 1 say, "Well, I just gotta fight this girl. She's gonna hafta whup me. 1 hope she don't whup me." Complication Evaluation Resolution Coda (Data and analysis from Labov 1972a: 35H-9)' And I hit the girl: powwww! And I put somcthing on it. I win the fight. That was one of the most important. The description of each of the stages that occur in the storytelling genres of spontaneous casual conversation will be explored in Chapter 6. Labov and Waletzky's analysis was the first attempt to offer a functional description of narratives of personal experience and the strength of their analysis lies in its clarity and applicability. However, there are two main problems in applying their narrative structure to the stories people tell in spontaneous casual conversation. The first problem is that the accurate identification of the Resolution is problematic, given that it is defined solely by its position within the text, as coming after the Evaluation stage. It is even more problematic when Labov and Waletzky (1967) argue that Evaluation is not only realized as a discrete stage, but is also spread throughout the text. Thus they suggest that there may be more than one Evaluation and 'not all the evaluation sections have the structural feature of suspending the complicating action' (1967: 37). A~ the criteria for describing the evaluations are semantic and not grammatical, it makes it difficult to locate where an Evaluation is realized discretely, and this in turn makes it difficult to determine exactly where the Resolution begins. The second problem is that Labov and Waletzky's definition of narratives fails to distinguish between narrative texts which have the pivotal stages of Complication and Resolution, and other narrative-like texts, such as recounts that do not have a crisis of any sort but merely involve the temporal retelling of events." As we will see in Chapter 6, participants in casual conversations produce a range of texts which have many features of a narrative, but which do not display the structural changes of a Complication followed by a Resolution. Because Labov and Waletzky's data

were elicited in interviews,7 they did not encounter these different kinds of storytelling texts, leaving later analysts with the task of extending their description of narrative structure to describe the variety of agnate story genres that occur in casual conversation. This is a task we take up in Chapter 6. 2.5 LOGICO-PHILOSOPHIC APPROACHES TO CONVERSATION Further complementary insights can be drawn from both Speech Act Theory and Pragmatics, which offer a logico-philosophic perspective on conversational organization by focusing on the interpretation rather than the production of utterances in discourse. 2.5.1Speech Act theory: conversation as a sequence of speech acts All linguistic descriptions of conversational structure owe much to Austin's (1962) and Searle's (1969, 1976) notion of the illocutionary force of speech acts. This means that every utterance can be analysed as the realization of the speaker's intent to achieve a particular purpose. The iIlocutionary force of many utterances is directly derivable from the linguistic form of the utterance. For example, a statement has the illocutionary force to inform. However, Searle and Austin also alerted researchers to the indirectness of many speech acts. Thus, Nick's utterance Where's the cigarettes, David? (text 1.1 in Chapter 1) has the same illocutionary force (to request) as its more direct alternative Can I have a cigarette, David? Neither Austin nor Searle were concerned with the analysis of continuous discourse, casual or otherwise. However, other analysts from a range of backgrounds have drawn on the implication of their work, that the speech act, or the illocutionary act, is the basic unit of discourse analysis. Identification of this unit allowed researchers (such as Labov and Fanshel 1977) to address one of the principal problems of discourse analysis: the lack of a one-to-one match up between discourse function (illocutionary force) and grammatical form (type of clause). For example, Nick (text 1.1) could have asked for a cigarette in any of the following ways: Can I have a cigarette, David? (modulated interrogative) Where's the cigarettes, David? (wh-interrogative) Give me a cigarette, David (imperative) I want a cigarette (declarative) What I'd do for a cigarette! (exclamative).

Given that there is extensive variety in the syntactic forms that can realize a particular illocutionary force, speech act accounts explore how the listener can interpret which illocutionary force is meant on any particular occasion. In one development useful for analysing casual conversation, Labov (1970, 1972b) suggested a distinction in interaction between the world as one interactant sees it (A-events) and the world as another interactant sees it (B-events). This enabled him to come up with rules for interpreting utterances as the intended speech act. For example: If A makes a statement about a B-event, it is heard as a request for confirmation. (Labov 1970: 80) For example, in the following sequence of turns from text 1.3 in Chapter 1, we see Alex volunteering a completion to Bill's narration in turn H, when she says And showed her. However, Alex cannot be assumed to know what actually happened next as it is a Bill- event. Her declarative clause is in fact heard as needing some kind of confirmation or correction, which is what Bill provides in his next turn, I said "THAT":
. . . . .

6 Bill 7 Mavis and Alex 8 Alex 9 Bill 10 Mavis and Alex

Labov formulated a more complex set of procedures for explaining how our post office customer could have initiated her encounter in text 1.4 in Chapter 1. She could have said I'd like to post these letters please and had her declarative recognized as a directive (demand for service) by the salesperson because the following interpretive procedures apply: There is a general rule for interpreting any utterance as a request for action (or command) which reads as follows: If A requests B to perform an action X at a time T, A's utterance will be heard as a valid command only if the following preconditions hold: B believes that A believes (=it isan AB-event that) 1. X would be done for a purpose Y 2. B has the ability to do X 3. B has the obligation to do

X 4. A has the right to tell B to do X. (Labov 1970: 82) From Speech Act Theory we take the insight that the basic unit of conversational analysis must be a functionally motivated, rather than formally defined, one. The speech act, under its systemic name of speech function, is central to our account of discourse structure in casual conversation (see Chapter 5). Recognition of the tension between discourse function and grammatical expression is also taken up, though reinterpreted through Halliday's (1994) notion of grammatical metaphor (explained in Chapter 5)' Labov suggests that interaction involves in part the distribution of knowledge, and consequently that roles in discourse are in part the enactment of differential access to knowledge. This suggestion is relevant to the development of models of conversational exchange structure (see the Birmingham School below). 2.5.2 Pragmatics: formulating "maxims" of conversational behaviour Another perspective we acknowledge as relevant is that of Gricean Pragmatics (e.g. Grice 1975, Leech 1983, Levinson 1983), which formulates conversational behaviour in terms of general "principles", rather than rules. At the base of the pragmatic approach to And I said "Could you direct me in the direction where the men's singlets are? I'm after... a Dr Flannel. She said "DR FlANNEL!" She said "What's that?" I said "WAIT A MINUTE!" [Makes motion of tearing shirt open]. [laughter] And showed her [laughs] I said 'THAT!" [pointing at singlet] [laughter] conversation analysis is Grice's co-operative principle (CP). This principle seeks to account for not only how participants decide what to DO next in conversation, but also how interlocutors go about interpreting what the previous speaker has just done: We might then formulate a rough general principle which participants will be expected (ceteris paribus) to observe, namely: Make your conversational contributions such as is required, at the stage at which it occurs, by the accepted purpose or direction ofthe talk exchange in which you are engaged. (Grice 1975): 45) This basic principle is then broken down into specific maxims and submaxims which are implicated in the CP, including maxims of Quantity ("say only as much as is necessary"), Quality ("try to make your contribution one that is true"), Relation

("be relevant") and Manner ("be brief and avoid ambiguity") (46). Although the pragmatics approach does not lead to a comprehensive description of conversational interactions, the concepts of maxims and principles provide a useful heuristic technique. For example, the description of casual conversation could lead to statements. in the form of maxims such as "in casual conversations in specific contexts interactants try to be provocative" and "in other casual conversations they try to be consensual", etc. Such maxims could then provide a useful means of characterizing different varieties of conversation. However, it is important to stress that we regard maxims as merely a shorthand summary of the results of research. The goal of conversation analysis pursued in this book is to offer explicit and detailed linguistic descriptions which can explain any conversational maxims proposed. Fairclough (1 99sa: 46) points to a further, more significant problem with the pragmatics approach: that it implies that conversations occur cooperatively, between equals. That is, that power is equally distributed between the consciously co-operating, autonomous conversational contributors. Such an account fails to recognize two facts about casual conversations. First, many conversations involve levels of disagreement and resistance. In fact, we will be offering support later for Kress's claim that: most or many conversations are marked by disagreement, and by absence of support. Conversations, like all texts, are motivated by difference. (1958a:21) Thus, we will be arguing that disagreement is essential to the motivation and the maintenance of casual talk. Second, in most conversations power is not equally distributed but is in fact constantly under contestation. Gricean pragmatics implies a noncritical idealizing of conversations as homogeneous, co-operative and equal. This view of conversation is hard to sustain if one works, as we do, with authentic examples of spontaneous interactions involving groups of participants from a range of different social backgrounds, rather than with hypothesized examples of decontextualized dialogic utterances. 2.6 STRUCTURAL-FUNCTIONAL APPROACHES TO CONVERSATION We use the label structural-functional to refer to two major approaches to discourse analysis which have relevance to the analysis of casual conversation: the Birmingham School and Systemic Functional. Linguistics. These two approaches

share a common orientation to discourse in that they both seek: to describe conversation as a distinctive, highly organized level of language. (Taylor and Cameron I987: 5). Structural-functional approaches ask just what is conversational structure, and attempt to relate the description of conversationa1 structure to that of other units, levels, and structures oflanguage. In addition, both approaches draw on the semantic theory of British linguistJR. Firth (1957) and Palmer (1968) and seek to offer functional interpretations of discourse structure as the expression of dimensions of the social and cultural context. 2.6.1 The Birmingham School: specifying the structure of the conversational exchange Logico-philosophic approaches to conversation offer ways of explaining how casual conversation is interpreted. However, we turn to the work of the Birmingham School, characterized as a structuralist functionalist approach to conversation (cf. Taylor and Cameron 1987),8 for insights into the linguistic structure of conversational exchanges: i.e. how interactants can keep taking turns. The Birmingham School was established through the work of Sinclair and Coulthard (1975), whose approach to discourse analysis has been extended beyond classroom discourse to conversation by Burton (1978, 1980, 1981), Coulthard and Brazil (1979) and Berry (1981 a,b). It shares a common origin with the systemic functional approach discussed later. Both approaches derive from the sociosemantic linguistic theory of J. R. Firth (see Firth 1957, and Palmer 1968), particularly as developed by Halliday in the early description of scale-and- category grammar (Halliday 1961, Halliday and McIntosh 1966). However, the evolution of the two approaches has differed. The Birmingham School has maintained its focus on discourse structure, while Halliday's development of the systemic perspective has led to the semiotic orientation in his work (e.g. Halliday 1978). The pioneering contribution of the Birmingham School approach involved recognizing discourse as a level of language organization quite distinct from the levels of grammar and phonology. Distinct discourse units, as opposed to grammatical units, were identified for the analysis of interactive talk. These units were seen to be related in terms of ranks or levels, i.e. each discourse unit being made up of one or more of the units immediately "below" it. Thus, in Sinclair and Coulthard's (1975) study of classroom discourse, acts combined to make up moves,

which in turn combined to make up exchanges. Exchanges combined to make up transactions, which finally made up lessons, which were the largest identifiable discourse unit in the pedagogic context. Their analysis sought to describe systematically the relationships between these discourse units and grammatical units such as the clause. The major contribution of the Birmingham School which has relevance for analysing casual conversation is its work on specifying the structure of the conversational exchange. The exchange is the discourse unit which captures the sequencing of turns at talk in terms of a set of functional "slots". While CA recognized the adjacency pair (a two-turn structure) plus a range of different, more extended sequences, it did not propose a general theory of discourse structure. In contrast, Birmingham School analysts have tried to develop a general description, in functional-structural terms, of the exchange as the basic unit of conversational structure. Defining the exchange simply as "two or more utterances" (Sinclair and Coulthard 1975: 21), they suggested that the pedagogic exchange involved the following three moves, or exchange slots: Initiation /\ (Response) /\ (Feedback) An example of this three-move pedagogic exchange is: Exchange Speaker Initiation Teacher Response Student Canberra Feedback Teacher Right What is the capital of Australia? However, attempts to apply this exchange formula to non-pedagogic exchanges revealed its limitations - and the peculiarities of the pedagogic context. Conversational exchanges differ from pedagogic exchanges in two mall1 ways: (i) At the exchange level pedagogic exchanges typically consist of three "slots", in a sequence motivated by movement towards completion, while casual conversational contexts reveal far more open-ended exchange types. In conversation it is quite typical to find the prolonged multiple-slot exchange identified as sequences by CA. For example, in text 1.1 in Chapter 1, turns 1 to 11 a can be read as constituting a single exchange. This example also illustrates the need for more specific criteria to identity the boundaries of conversational exchanges. (ii) In casual conversation interactants rarely ask questions to which they already know the answers. Therefore the types of moves which occur in Initiating slots of conversation are very different from those in pedagogic exchanges. They include "real" questions, statements of opinions, commands, offers, etc. The slots which

occur after the Responding slot do not generally consist of evaluating moves but are either recycling types of moves (queries, challenges) or additional "afterthoughts" of various kinds. In an attempt to come up with a formula which could describe conversational exchanges, Coulthard and Brazil offered a more explicit definition of the exchange as "basically concerned with the transmission of information" (1979: 41). They replaced the label "feedback" with the label "follow-up" for the third element, and suggested that the conversational exchange could contain slots which were responses to previous initiations and at the same time initiations which elicited responses. The following sequence was necessary to account for conversational exchanges: Initiation /\ (Re-Initiation) /\ Response /\ (Feedback) Conversational exchanges thus consist of minimally two elements, and maximally four. Extending on this, they noted that:
. .

(i) feedback could itself be reacted to with further feedback; (ii) sometimes moves occurred which marked the beginning of an exchange without necessarily constraining the next element; (iii) sometimes moves occurred which indicated the end of an exchange (but which were not necessarily feedback).

Consequently, additional elements were added to their formula, resulting in the following exchange formula (Coulthard and Brazil 1979: 40): (Open) /\ Initiation /\ (Re-Initiation) /\ Response /\ (Feedback) /\ (Feedback) /\ (Close) This formula allows for a minimum of two, and a maximum of seven elements of structure in a single exchange. For example, we can use this formula in one possible analysis of the first 11 turns in text 1.1: Turn Speaker
. . .

1 S1 2 S2 3 S3

Exchange slot/move Open Initiation Response This eh has been a long conversation. [Dead space in the conversation] In France they say "An angel is passing". In English too. 4 S2 5 S3 6 S2 7 S4 8 S3 9 S2 10 S3 11 S3 However, this conversation: Reinitiation Response Reinitiation Reinitiation Response Reinitiation Response Close model still has Really? Umm == Oh I've never heard that before. == == I've never heard of that. Well I think so. I think I've heard it first in English but maybe they were just translating. I don't know. I thought in English it was "Someone's walked over Oh "over your grave". Y ou're probably absolutely right. have heard it only in French. Maybe I a number of problems for those interested in the analysis of The criteria for determining the boundaries of exchanges remain unclear. For example, are there perhaps more exchanges in turns 1 to lla above than we have identified? The criteria for allocating different moves to each slot are also unclear. The model remains rigidly sequenced and does not recognize that some conversational moves can appear at any point in an exchange. For example, it is possible to query. or challenge at any point, thereby achieving a Re-Initiation. Major retheorizing of Coulthard and Brazil's exchange structure formula was achieved by Margaret Berry (1981 a,b), whose four-slot exchange formula drawing on Labov's (1970) distinction between A-events and B-events has proved useful in

describing the structure of pragmatic exchanges (see Ventola 19H7). However, as it has limited applications to conversational exchanges, we will not review it in detail (see Martin 1992: ch. 2 for a discussion). One Birmingham analyst who did specifically work on casual conversation, though some of it fictional, was Burton. Building on Coulthard and Brazil, Burton (lg78) suggested that exchanges in casual conversation were in fact far more open-ended than the earlier formulae recognized. The typical structure, she argued was that following an Initiation and an obligatory Response, there could be a Re-Initiation (moves which extend on the previous Initiation /\ Response pair). This could then be followed by any number of Responses. More important than the details of Burton's exchange formula was her observation that the "polite consensus-collaborative model" (1978: 140) which underlies the exchange structure of both Sinclair and Coulthard, and Coulthard and Brazil (and subsequently Berry) did not fit casual conversational data well at all. While interactants in pragmatic encounters negotiate in order to achieve exchange closure, casual conversationalists are frequently motivated to do just the opposite: to keep exchanges going as long as possible. One of the ways in which this is achieved, as Burton noted, is for casual conversationalists to choose challenging rather than supporting moves, since challenges, such as moves which withhold the preferred response to an initiation, demand their own responses and thus compel further talk. In Chapter 5 we will take up these issues in 'developing an analysis of exchanges in casual conversation and we will integrate Birmingham insights about the structure of the discourse stratum with systemic interpretations of the links between discourse, grammar and semantics in casual talk. 2.6.2 Systemic functional linguistics: a functional-semantic interpretation of conversation An important influence on our approach to analysing casual conversation is that of systemic functional linguistics, based on the model of "language as social semiotic" outlined in the work of Halliday (e.g. 1973, 1975, 1978, 1994, Halliday and Hasan 1985). Eggins (1994) provides an introduction to the basic principles of the systemic approach, Eggins and Martin (1995) outline the general approach to discourse/text analysis, and Ventola (1987, 1995) illustrates some applications of the systemic approach to the analysis of pragmatic interactions. The systemic approach offers two major benefits for conversational analysis: 1. It offers an integrated, comprehensive and systematic model oflanguage which

enables conversational patterns to be described and quantified at different levels and in different degrees of detail. 2. It theorizes the links between language and social life so that conversation can be approached as a way of doing social life. More specifically, casual conversation can be analysed as involving different linguistic patterns which both enact and construct dimensions of social identity and interpersonal relations.

It is these two advantages of systemic linguistics which are responsible for its applications in a range of domains: e.g. in critical discourse analysis (discussed below), educational fields (Christie 1991a,b, Cope and Kalantzis 1993, Martin 1993), and computational linguistics (e.g. Bateman and Paris 1991, Matthiessen and Bateman 1991). In the following sections we briefly summarize those aspects of the systemic approach which we draw on in our analysis of casual conversation. We are particularly concerned with the realizational relationship between language and context, the analysis of interpersonal meaning in conversation, the levels of analysis for micropatterns, and the functional interpretation of genre. 2.6.2.1 Strands of meanings: ideational, textual, interpersonal The systemic functional model of language can be glossed as a functional-semantic theory of language. It is a functional theory in that it models conversation as purposeful behaviour. It is semantic in that it interprets conversation as a process of making meanings. One of the most powerful aspects of the systemic approach is that language is viewed as a resource for making not just one meaning at a time, but several strands of meaning simultaneously. These simultaneous layers of meaning can be identified in linguistic units of all sizes: in the word, phrase, clause, sentence, and text. This means that a casual conversation, itself an extended semantic unit or text, is modelled as the simultaneous exchange of three types of meaning. These three types of meaning, or metafunctions, can be glossed as follows:
. . .

(i) ideational meanings: meanings about the world; (ii) interpersonal meanings: meanings about roles and relationships; (iii) textual meanings: meanings about the message.

The three strands of meaning are summarized and exemplified from text 1.1 in

Table 2.2. As there are different strands of meaning being enacted in talk, so the analyst needs to analyse the talk from different perspectives. Thus different analytical techniques are used to uncover each strand of meaning. For example, to explore the ideational meanings in a text, the analyst focuses on patterns which encode the who, when, where, why, and how of a text. These patterns are seen in the analysis of lexical cohesion (chains of words from similar semantic domains) and the analysis of transitivity (for details see Halliday 1994, Eggins 1994)' The systemic model is rich in analytical techniques, allowing the analyst to focus on those patterns which are most relevant to specific data and research interests. This multi-semantic and multi-analytical perspective on language has proved a useful tool for studies of other semiotic systems (d. Kress and van Leeuwen 1 ggo, 1996, O'Toole 1 994). The richness of the model, however, necessitates a careful delimiting of analytical focus. Table 2.2 Types of meanings in the systemic model Types of meaning ideational Gloss / definition

Example:

meanings about the world, conversat representation of reality (e.g. topics, language cigarettes su~ject matter)

interpersonal

SI and S "banterin S2 and meanings about roles (e.g. status, relationsh intimacy, contact, sharedness between provokin interactants) challengi direct a assertive meanings about the messageextual

textual

Rapid tu ellipsis a of expres

2.6.2.2 Focus on interpersonal meaning in analysing casual conversation The pluri-semantic model outlined above offers us three main approaches to

analysing casual conversation:

(i) We can focus on the ideational meanings: this involves looking at what topics get talked about, when, by whom, and how topic transition and closure is achieved, etc. (ii) We can focus on the interpersonal meanings: this involves looking at what kinds of role relations are established' through talk, what attitudes interactants express to and about each other, what kinds of things they find funny, and how they negotiate to take turns, etc. (iii) We can focus on the textual meanings: this involves looking at different types of cohesion used to tie chunks of the talk together, different patterns of salience and foregrounding, etc.

Analysis of casual conversation could explore all of these three dimensions of casual talk. However, for theoretical reasons this book focuses on the analysis of interpersonal meanings in casual conversation. We offer the following two reasons for this focus. First, as outlined in Chapter 1, we believe that the primary task of casual conversation is the negotiation of social identity and social relations. Thus casual conversation is "driven" by interpersonal, rather than ideational or textual meanings. To support this claim, we point to: the absence of any pragmatic motivations or outcomes to casual talk; the observation that anything can be a topic of talk in casual conversation (no topics are a priori barred) which suggests that casual conversation is not focused on ideational meanings; the apparent triviality of much of the ideational content of casual talk, which suggests that the important work of casual conversation is not in the exploration of ideational meanings. Rather, any ideational domain (or Field) serves as the environment for the exploration of social similarities and differences. Second, it is the open-ended, turn-taking organization of conversation that differentiates it from other linguistic activities (Sacks et al. 1974)' As we will see in later chapters, this turn-by-turn structuring of conversation is realized through interpersonal patterns of mood and conversational structure.

Given the priority of interpersonal meanings in motivating and structuring casual conversation, any comprehensive analysis of casual conversation must be able to offer a framework for describing interpersonal patterns in talk. Our focus, then, is on the interpersonal meanings expressed in casual conversation. In taking the interpersonal as primary, we do not exclude from discussion ideational or textual patterns, but we discuss these as resources mobilized by interpersonal meanings. Ideational dimensions have been explored in detail, both outside systemics and within it, to the comparative neglect of the interpersonal meanings in talk. For example, CA approaches examine topic management (e.g. Schegloff and Sacks 1974, Maynard 1980), and topic development (e.g. Button and Casey 19H4) in conversation. Within systemics, although there has been little work on ideational patterns in casual conversation, there is a substantial literature available detailing the kinds of analyses that could be applied (e.g. Halliday and Hasan 1976, 1985, Martin 1992). Our focus on the interpersonal strand in casual conversation directly implicates the link between interpersonal meanings and the context of casual conversation, which we consider below. 2.6.2.3 Context: register and genre in the systemic model The systemic model is also relevant to conversation analysis because it seeks to explain language as a social resource. The model is based on Halliday's claim that language relates "naturally" to the semiotic environment, that 'language is as it is because of what it has to do' (1973: 34). Halliday claims that the three types of meaning we find represented in language are not there accidentally. They are there because those are the three types of meanings we need to make with each other. This implies that social life requires the negotiation of a shared ideational world. Simultaneously, it requires the continual renegotiation of our places within that world: who we are, how we relate to the other people in it, and how we feel about it all. In the process of negotiating those ideational and interpersonal actualities, we must also negotiate ways of talking about that world: what kinds of texts we can construct to represent ideational and interpersonal meanings. The tripartite structure of language is therefore an encoding of the tripartite structure of the contexts of situation in which we use language. The systemic interest in the analysis of context through the concept of register is the logical extension of Halliday's "natural" language thesis. The three register variables of field (activity or topic focus), mode (extent and type of feedback possible) and tenor (roles and role

relationships) are proposed to describe the major dimensions of any situations which have systematically predictable linguistic consequences. Each of the three register variables is realized through patterns in the different metafunctions. The field of a situation is realized through the ideational metafunction, the mode is realized through the textual metafunction, and tenor is realized through the interpersonal metafunction. These links between context of situation and language are summarized in Figure 2.3.
mode

field ideational tenor

Figure 2.3 Context and language in the systemic function model Source: Eggins and Martin (in press) A further level of context important in analysing conversational data is that of genre (cf. Eggins and Martin in press), which is discussed in more detail below. 2.6.2.4 Dimensions of social identity: the register variable of tenor By focusing on interpersonal meanings in casual conversation we are choosing to explore the register variable of tenor, of which those interpersonal meanings are the expression. By the term 'tenor' Halliday (1978: 143) is referring to "the role structure: the cluster of socially meaningful participant relationships" operating in a situation. Extending on the work of Poynton (1985) and Martin (in press), we will subclassify role relations into four main dimensions: status relations; frequency of contact or level of familiarity; degree of affective involvement; and orientation to affiliation. Each is briefly explained below: First, status relations: the construction of a social self through interaction involves taking on a recognized social role, and attributing to fellow interactants relevant social roles. These may involve an inequality of status, as in functionally differentiated roles such as customer/salesperson, or they may imply the (temporary) assumption of equal status, as in the functionally reciprocal roles friend/friend. As Poynton points out, the legitimation for unequal status relations may come from a number of different sources:
. .

(i) force; (ii) authority, where the culture attributes unequal status to particular role

 relations, .

ego parent-child, employer-cmployee;

(iii) expertise, which will, for example, unequally position the expert in relation to

textual

Context organized by metafunction

Language organized by metafunction the novice; or (iv) status symbols, which Poynton defines as 'a matter of relative ranking with respect to some unevenly distributed but socially-desirablc object or standing or achievement, e.g. wealth, profession/occupation, level of education, hereditary status, location of residence, overseas travel. (1985: 77)

Second, affective involvement: our interpersonal relations also vary according to the degree of affective involvement, the degree to which-we "matter" to those with whom we are interacting. Mfective involvement can range from nil (distant, unattached) through some (ey;. school friends or particular work colleagues) to high (e.g. lovers, close friends, family). In relationships which are largely low in emotional investment, interpersonal relations are towards the impersonal end of the continuum. Mfective involvement may be positive or negative, and may be a permanent feature of a particular relationship, or may be transient (as in a sudden argument with the bus driver about bus fares, where a typically neutral relationship becomes affectively charged for a short time). Third, contact: interpersonal relations also vary in terms of the level of familiarity constructed to operate between interactants. Familiarity is developed through high frequency of contact. Contact can be characterized as regular (e.g. our relations with our immediate families) or intermittent, with the most intermittent being simply a one-off encounter. At the same time, the contact we have with other people may be voluntary, when we choose to spend our time with them, or involuntary, as in the workplace where we are compelled to spend time with people we might otherwise prefer to avoid. A final dimension relevant to the kind of contact between interactants is the orientation that is operative in their interactions: whether they come into contact largely to achieve pragmatic tasks, or to relate to each other "as people". Fourth, orientation to affiliation: an important part of our social identity derives from our inclination or disinclination to affiliate with various formal and informal social groups, such as family, workmates, fellow students in our tutor group, residents in our area, or other passengers on a bus. Orientation to affiliation refers to the extent to which we seek to identifY with the values and beliefs of those we interact with in these different social contexts. We might already be well oriented towards identification with some social groups (e.g. close family), or we may be seeking to be accepted by others as an "insider" (e.g. a new employee in the workplace). On the other hand, we might be happily or unhappily positioned as "other" in a social group (e.g. the marginalized, unaccepted member of the office), or be contesting an affiliation from an earlier time in our social lives (e.g. a rebellious adolescent in the family). 2.6.2.5 Tenor and patterns of interpersonal meaning in casual conversation The systemic model goes beyond simply suggesting a list of tenor dimensions implicated in social identity. It also seeks to offer an account of the ways in which

language is used both to reflect and construct these dimensions. As we will demonstrate, there are four main types of linguistic patterns which represent and enact the social identities of participant,> in casual conversation. These patterns, which operate at different levels or within different linguistic units, are: (i) grammatical; (ii) discourse; (iii) semantic; (iv) generic patterns. i) Grammatical patterns - these are patterns which operate within turns, and have to do with the mood of the clauses interactants use. For example, whether interactants use declarative, interrogative or imperative clauses. We will be suggesting in Chapter 3 that grammatical patterns of mood choice are a key resource for enacting and constructing status differences. Reciprocal mood choice indicates functional equality of roles, while nonreciprocal mood choice indicates functional differentiation, or the linguistic acting out of status differences. ii) Semantic patterns - this group of patterns is revealed by studying attitudinal and expressive meanings in talk. Semantic patterns often concern the choice of lexical items, and so are revealed by examining the words used by conversationalists. They have to do with the types of evaluative and attitudinal lexis which interactants use, and the directness with which they can speak to or laugh at (or with) each other. We will be suggesting in Chapter 4 that these semantic resources are often used to develop group cohesion and to censure actual or potential deviant group behaviours. iii) Discourse structure patterns - these patterns operates across turns and are thus overtly interactional and sequential. They show us how participants choose to act on each other through their choice of speech functions (i.e. speech acts), such as "demanding", "challenging", "contradicting" or "supporting", and how participants' choices function to sustain or terminate conversational exchanges. We will be suggesting in Chapter 5 that choice of speech function is a key resource for negotiating degrees of familiarity. If interactants wish to explore their interpersonal relations, they must choose speech functions which keep the conversation going, and this frequently means that intimate relations involve interactants reacting to each other in confronting, rather than supporting, moves. iv) Generic structure patterns - these staging patterns operate to build "chunks" of talk, such as storytelling or gossiping segments within the flow of chat. We will be suggesting in Chapter 6 that the construction of text types or genres within casual talk is one way in which affiliation can be explored and developed. We will now review the work on genre within systemic linguistics. 2.6.2.6 Genre in systemic functional linguistics

Systemic accounts of genre owe much to the Firthian linguist, Mitchell (1957), who developed a detailed account of buying and selling in Cyrenaica. His analysis involved setting up text structures of the following kind for market auction and market transaction contexts. In the formulae below, /\ stands for the typical sequence of realization, although Mitchell notes that some variability and overlap is found. Market Auction: Auctioneer's Opening /\ Investigation of Object of Sale /\ Bidding /\ Conclusion Market Transaction: Salutation /\ Enquiry as to Object of Sale /\ Investigation /\ Bargaining /\ Conclusion In the 1970s, Hasan, working in the tradition begun by Mitchell, laid the foundations for a theoretical conception of genre within systemics with her accounts of generic structure. In her early work looking at the structure of nursery tales and later the structure of service encounters (1984, 1985), Hasan introduces the notion of generic structure potential to generalize the range of staging possibilities associated with a particular genre. Her analysis of staging in service encounters (19H5: 64) is outlined below. [ (Greeting) (Sale initiation) ] /\ [ (Sale Enquiry") {Sale Request /\ Sale Compliance}" /\ Sale] /\ Purchase /\ Purchase Closure /\ (Finis) This notation is one means by which linguists express what appear to be the obligatory and optional stages of genres. The caret sign /\ means that the stage to the left precedes the one to the right. The stages within the brackets ( ) are optional features of the genre, the brace { }" indicates that the stages within are recursive, and the square brackets [ ] enclose the elements which are recursive. Each stage in the genre (e.g. Sale Enquiry) is given a functional label which attempts to represent its meaning-what it does in semantic terms, in this case making an enquiry about a sale. Of course, people innovate on the basic sequence of obligatory stages and often play, in imaginative ways, with the optional stages of all genres. However, as Bakhtin suggested "genres must be fully mastered to be used creatively" (1986: Ho). Hasan's analysis of service encounters is exemplified in text 2.2 below (from Halliday and Hasan 1985: 59)' Text 2.2 Service Encounter Sale Request Customer Can I have ten oranges and a kilo of bananas please? Sale Compliance Vendor Yes, anything else? Customer No thanks

Sale Vendor That'll be dollar forty Purchase Customer Two dollars Purchase Closure Vendor Sixty, eighty, two dollars. Thank you. (Data and analysis from Halliday and Hasan 1985: 59) Text 2.2 contains the obligatory elements of the structure of a service encounter, i.e. those elements that are a defining feature of the genre. The text begins with a request for goods: Can I have ten oranges and a kilo of bananas please? This is then followed by the Sale Compliance, which in this case is the granting of the Sale Request. A positive Sale Compliance is also likely to include an invitation for more purchases, as in text 2.2 when the vendor says Yes, anything else? (see Halliday and Hasan 19H5: 60). This stage is followed by the exchange of the goods through the three stages of Sale, Purchase and Purchase Closure. that: Hasanarguesthatitistheobligatoryelementsofstructurethatdefinethegenreand

the appearance of all these elements in a specific order corresponds to our perception of whether the text is complete or incomplete. (Halliday and Hasan, 1985: 61) Other elements which are optional may occur in some instances of the genre and may also occur in other genres. However, these do not occur randomly, and the conditions under which they are likely to occur can be described. For example, the optional element Sales Initiation, expressed by "Who s next?, is likely to occur in a crowded shop where it is necessary to get a customer's attention. It could also occur in other genres, as it is not a defining feature of a service encounter. Hasan also claims that it is the internal or generic structure of a text that is defining of a genre, so that texts of the same genre will realize the same obligatory elements of structure. Such a model gives us a way of accounting for both the similarities and differences between text 2.2 above and text 1.4, the post office transaction. Text 1.4 contains all the obligatory elements of a service encounter (Sale Request, Sale Compliance, Purchase, and Purchase Closure), but differs from text 2.2 in having both a Sales Initiation and more than one Sale Request /\ Compliance sequence. Drawing on this work, Martin (1992) developed an alternative but ultimately complementary theory of genre. He defines genre as a:

staged, goal oriented, purposeful activity in which speakers engage as members of our culture. (l984a: 25) The concepts within his definition can be elaborated as follows: Staged: a genre is staged as the meanings are made in steps; it usually takes more than one step for participants to achieve their goals. Goal oriented: a genre is goal oriented in that texts typically move through stages to a point of closure, and are considered incomplete if the culmination is not reached. Social process: genres are negotiated interactively and are a realization of a social purpose. More recently Eggins and Martin (in press: g) have defined genre as 'a theory of the unfolding structure texts work through to achieve their social purposes'. Genres are enacted in texts, and as texts have different purposes in the culture, texts of different genres will unfold in different ways, or work through different stages or steps. Martin refers to the overall staging patterning of texts as the schematic structures. These patternings are a realization of the overall purpose of the text. The stages, as with Hasan's work, are identified in functional terms. The fact that we can recognize the stages of a text as having particular functions is because of the patterning of language in each stage. For both Hasan (lg85) and Martin (1993), each stage is defined by its distinctive semantic and lexicogrammatical realizations. Each genre is seen as having different semantic and lexico-grammatical characteristics, and each stage of the generic structure is realized through specific discourse-semantic and lexico-grammatical patterns. Considerable work has been carried out on the analysis of written genres using Martin's model. In the 1980s Martin and Rothery analysed in detail some of the key written genres of schooling including reports, narratives, explanations, procedures and argumentative genres such as exposition and discussion (see Martin and Rothery 1986). This work has had a wide impact in education in Australia, and has formed the theoretical underpinning of what is now referred to as the genre-based approach to the teaching of writing in schools.

The work on spoken genres in this paradigm has included the analysis of service encounters (e.g. Ventola I987), the analysis of spoken, pedagogic discourse (e.g. Christie 1989, Hammond I995), and narratives elicited from a sociolinguistic interview (Plum I988). However, as yet there has been very little work in this paradigm on the analysis of the genres that occur in spontaneous, informal conversation. For early applications of the model to conversation, see Eggins (1990), Horvath and Eggins (1995) and Slade (1995). In Chapters 6 and 7 we will be exploring the structure and function of some of the more common genres that occur in casual talk, building on the techniques developed to describe the more micro-patterns of conversation which are introduced in Chapters 3, 4 and 5. By drawing on analytical techniques offered by systemic linguistics, and also on its theoretical interpretation ofthe relationship between language and social context, our analysis of casual conversation is simultaneously an analysis of linguistic patterns at several levels, and an account of the work that different linguistic patterns are doing in the construction of social identity. In the typology of approaches outlined in Figure 2.1 above we have positioned systemic functional linguistics as related to both structural-functional approaches to conversation and to social semiotic approaches. Its focus on structure and function in language (e.g. in the analysis of genre) is complemented by an interpretation of language as a social semiotic resource: a system for making meanings through which language users both reflect and constitute themselves as social agents. This semiotic approach is expressed by Halliday in his recognition of the relationship between the micro- and the macro-social worlds: By their everyday acts of meaning, people act out the social structure, affirming thcir own statuses and roles, and establishing and transmitting the shared systems of value and of knowledge. (1978: 2) This enactment relationship between language and social structure has provided a foundation for the systemic model. However, since Halliday's development of the more "semiotic-contextual" perspective, systemic analysts have devoted mIlch of their time to developing detailed tools for analysing language patterns. This has often left too little time for interpreting critically the results of their descriptions, and too little time for explaining in accessible terms the social effects of linguistic choices at both micro and macro levels. It is in this area of critical interpretive text analysis that systemics has recently been influenced by, and has also influenced, the

last approach we will consider: that of the critical linguists and critical discourse analysts. 2.7 CRITICAL LINGUISTICS AND CRITICAL DISCOURSE ANALYSIS (CDA) In this final section we group together two approaches to thinking about language which have a "critical" perspective in common. This critical perspective insists that the adoption of critical goals means, first and foremost, investigating vcrbal interactions with an eye to their determination by, and their efkcts on, social structures. (Fairclough, 1995a: 36) In other words, the micro-interactions of everyday life are viewed by critical analysts as the realization of macro-social structures, so that: the question of how discourse cumulatively contributes to the reproduction of macro structures is at the heart of the explanatory endeavour. (1995a: 43) By the label Critical Linguistics we are referring to social semioticians such as Gunther Kress and his associates (e.g. Kress 1985a, b, 1987, Hodge and Kress 1988, 1993, Kress and van Leeuwen 1996), whi1e the term Critical Discourse Analysis refers to work principally by Norman Fairclough (e.g. Fairclough 1989, 1992, 1995a, b). Both approaches are concerned to 'develop ways of analysing language which address its involvement in the workings of contemporary capitalist society' (Fairclough 1995a: 1). They share a focus on the relationship between language, ideology and power (see, for example, Kress and Hodge 1979, Fairclough 1989) and the relationship between discourse and sociocultural change (see, for example, Fairclough 1992). In these respects the critical strands are similar in focus to much of the work in contemporary critical theory. Where they differ from this work is in their emphasis on relatively detailed textual and lexicogrammatical analysis. Although researchers in the "critical" tradition have not focused on casual conversation, they have made various comments as to the interest of casual talk, and have provided guidance as to what a critical conversation analysis might look like. We will briefly review the contributions we take from three main areas of their work: (i) notions of text and difference; (ii) methods and techniques of critical discourse/conversation analysis; and (iii) critical accounts of genre. i) Notions of text and difference One of the major contributions of the critical

perspectives has been quite simply exploring why texts happen. Kress (1985a) is one of the few analysts to ask what we consider to be an obvious but unvoiced question for conversation analysts: What is the motivation for speech? (11) Kress's answer, that it is differences which motivate speech, is equally succinct: Most speech genres are ostensibly about difference: argument (differences of an ideological kind), interview (differences around powcr and knowledge), 'gossip' (differences around informal knowledge), lecture (difference around formal knowledge), conversation. (15) Kress's argument involves recognizing that individuals who come to interactions share membership of particular social groupings, and learn modes of speaking or discourses associated with those institutions. Drawing on Foucault, he defines these discourses as: systematically-organised sets of statements which give expression to the meanings and values of an institution. ... A discourse provides a set of possible statements about a given area, and organises and gives structure to the manner in which a particular topic, object, process is to be talked about. In that it provides descriptions, rules, permissions and prohibitions of social and individual actions. (7) People come to interactions with their own discursive histories, determined by their social history and social position. Interaction with others who do not necessarily share the same discourses (indeed, they cannot, for no two interactants have identical social! discursive histories) creates difficulties that need resolution. As Kress explains: There are likely to be problems at any time, arising out of unresolved differences in the individual's discursive history, the individual's present discursive location and the context of discourses in interactions. That difference is the motor that produces texts. Every text arises out of a particular problematic. Texts arc therefore manifestations of discourses and the meanings of discourses, and the sites of attempts to resolve particular problems. ( 12) He goes on to suggest that dialogues (whether casual conversation or more pragmatic, formal interactions) provide the clearest examples of this "discursive difference" or textual problematic: Successful dialogues come about in the tension between (discursive) differcnce and

the attempt to rcsolve that difference in some way. (13) This notion of discursive difference explains why couples in longstanding relationships have little to say to each other. They have simply exhausted their differences. To avoid silence they may try to construct difference, for: Where there is no differencc, no text comes into being. (12) Kress suggests that the structure of dialogue, of two (or more) interactants taking up positions in discourse and attributing positions to others, makes difference apparent in a way that written text does not. Texts are about (social/discursive) differences, and so texts are fundamentally about (social/discursive) power, for power is nothing more than "relations of difference" (Kress 1985a: 52). The interests of critical linguists and critical discourse analyst" in language is not accidental. It arises from their conviction that language and power stand in a particular relationship: Because of the constant unity of language and other social matters, language is entwined in social power in a number of ways: it indexes power, expresses power, and language is involved wherever there is contention over and challenge to power. Power docs not derive from language, but language may be used to challenge power, to subvert it, and to alter distributions of power in thc short or in the longer term. (52) To understand the social distribution of power, the analyst should look at language, because: Language provides the most finely articulated means for a nuanced registration of differences in power in social hierarchical structures, both as a static system and in process. All linguistic forms which can be used 10 indicate relations of distance, and those which can indicate 'state' or 'process' serve the expression of power. In fact there are few linguistic forms which are not pressed into the service of the expression of power, by a process of syntactic/textual metaphor. (53) Our analyses in Chapters 3 through 7 are concerned with identitying discursive differences. We will be arguing that these differences invariably express power relationships which have implications beyond the microsituation of the specific conversation in which they occur. ii) Methods and techniques of "critical" conversation analysis As well as suggesting why conversation occurs, critical analysts have also suggested how it

could or should be studied. For Fairclough (1995a), as for all critical analysts, the micro-event and the macro-social structures are inextricably linked: 'micro' actions or events, including verbal interaction, can in no sense be regarded as of merely 'local' significance to the situations in which they occur, for any and every action contributes to the reproduction of 'macro' structures. (34) Given this "reproductive" or "reconstituting" relationship between everyday interaction and more global social structures, Fairclough points out that: it makes little sense to study verbal interactions as if they were unconnected with social structures. (35) (emphasis in original) Yet this, he claims, is just what many of the approaches to conversation analysis have tried to do. He launches a strong critique against approaches as diverse as the Conversation Analysts, the Birmingham School, and Gricean Pragmatics for only describing the local organization of speech events, and not the social work such events achieve in maintaining social structures. Part of the problem he identifies with such approaches is that they perpetuate views of conversation as the speech events of conscious, independent, social actors, co-operatively achieving selfevident "goals" through homogeneous interactions. Such a view denies the fact that most interactants participate in micro-encounters quite unaware of their ideological! macro-structural implications (cf. Fairclough 1995a: 46ff). In fact, successful ideologies naturalize themselves, effectively making awareness of the social determinations of our actions virtually impossible. Fairclough also points out that: The descriptive approach has virtually elevated cooperative conversation between equals into an archetype of verbal interaction in general. As a result, even where attention has been given to 'unequal encounters'... the asymmetrical distribution of discoursal and pragmatic rights and obligations according to status...has not been the focal concern. (46-7) . Thus, the social effects and consequences of power inequalities have not received critical attention. In contrast to the purely "descriptive" goals pursued by many of these approaches, Fairclough argues for the pursuit of "critical" goals in discourse analysis. Basic to this critical approach is the question we raised earlier when considering the paradox of casual conversation: how can it be that people are standardly unaware of how their ways of speaking are

socially determined, and of what social effects they may cumulatively lead to? (36) While Garfinkel and Berger and Luckman sought explanations in the "taken-forgrantedness" of everyday interaction, Fairclough extends this to consider why participants have the impression of "orderliness" in interactions, and feel that "things are as they should be, i.e. as one would normally expect them to be" (28). Orderliness, he suggests, arises from participants' conformity with their background knowledge about the norms, rights and obligations appropriate in interactions in particular contexts. For example, a lecture at university is "orderly" to the extent that students and lecturer conform to dominant discourse practices whereby the lecturer talks (without invitation), and the students listen (without complaint). Orderliness will be experienced as having been disrupted if, for instance, a student begins interjecting critical comments during the lecturer's discourse, or the lecturer stands up and says Well, unfortunately I reallv don't know anything about this topic, so you guys better do the talking today. Fairclough suggests that the "knowledge base" which constitutes participants' background knowledge about how orderly interactions proceed involves four components: knowledge of language codes, knowledge of principles and norms of language use, knowledge of situation. knowledge of the world. (33) Fairclough claims that all four of these types of knowledge involve ideology, by which he means: each is a particular representation of some aspect of the world (natural or social; what is, what can be, what ought to be) which might be (and may be) alternatively represented and where any given representation can be associated with some particular 'social base'. (31) This means knowledge is not neutral, and always implies ways of doing which serve the interests of some social group, generally at the expense of others. For example, knowing about the principles and norms of language used in lectures involves knowing a way of using language in that context which serves the interests of one social group (academics, whose position as an educated authority is maintained) at the expense of another (students, who would certainly learn more through other modes and under other conditions). However, participants generally notice neither the social determinations of their speech (e.g. that I behave as a student in lectures in conformity to norms established

by an interested social group, academics), nor the social effects (e.g. surrendering of control over my own acquisition of knowledge). It is in the nature of ideology to "naturalize" itself. And thus critical discourse analysis involves "denaturalizing" everyday discourse, to expose the ideologies expressed and therefore the interests served. To achieve this denaturalization, Fairclough argues that: Textual analysis demands diversity of focus not only wirh respect to functions but also with respect to levels of analysis. (7) Fairclough finds the most useful linguistic model for achieving this diversity is systemic functional linguistics, which he recognizes to be: a functional theory of language oriented to the question of how language is structured to tackle its primary social functions. ... The view of language as social semiotic (Halliday, 1978) incorporates an orientation to mapping relations between language (texts) and social structures and relations. (10) Like Fairclough, we find the theory of systemic linguistics the most useful one for denaturalizing the ideological processes of casual conversation. iii) Critical accounts of genre The third aspect of CDA which has influenced our work is its perspective on genre, where it has contributed to our thinking about analysing both genres within conversation and the genre of conversation itself. Views of genre within CDA draw less on functional linguistic accounts of structure and more on current reconceptions of genre by critical literary theorist~. Drawing on Bakhtin (1986), these accounts make reference to the sociocultural and historical contexts embodied by types of texts and explore the ways in which genres function as discursive practices. Bakhtin's (1986) identification of speech genres, and his argument that the more complex "secondary" genres of writing are derived from these primary speech genres, has broadened the field of enquiry to include everyday as well as literary genres. Bakhtin was interested in the analysis of actual language use and not in the analysis of decontextualized sentences, and consequently his basic unit of analysis is the utterance rather than the sentence. He defines speech genres as: the typical f'orm of the utterance associated with a particular sphere of communication (e.g. the workplace, the sewing circle, the military) which have therefore developed into 'relatively stable types' in terms of thematic content, style and compositional structure.' (1986: 52) Examples of speech genres include 'short rejoinders of daily dialogue', 'everyday

narration", 'business documents' and 'the diverse world of commentary' (60). Their structure derives from the situations from which they arise such that: even in the most free, the most unconstrained conversation, we cast our speech in definite generic forms, sometimes rigid and trite ones, sometimes more flexible, plastic and creative ones. . . . We learn to cast our speech in generic forms and when hearing others' speech, we guess its genre from the very first words; we predict a certain length and a certain compositional structure; we foresee the end; that is from the very beginning we have a sense of the speech whole. If speech genres did not exist and we had not mastered them, if we had to originate them during the speech process and construct each utterance at will for the first time, speech communication would be almost impossible. (78) Bakhtin's perspective on genre is similar to that taken in systemic-functional linguistics. Bakhtin claims that the lexical, grammatical and compositional structures of particular genres are a reflection of the specific context of . communication and he has identified genres as 'relatively stable types' of interactive utterances with definite and typical forms of construction. These particular features of Bakhtin's analysis are echoed in Martin's approach to genre analysis reviewed above (see Eggins and Martin 1995). From a CDA perspective, Fairclough defines genre as 'a socially ratified way of using language in connection with a particular type of social activity' (1995a: 14). Genre is 'the abstract constituent of text types' (13). There are the ideal text types which people orient to but do not necessarily conform to, but there are also: texts which closely match ideal types (as well as others which do not), so that people learn them from concrete textual experience. (13) Genres in CDA are seen as social actions occurring within particular social and historical contexts. As Millar argues (l984), the similarities in form and discursive function are seen as deriving from the similarity in the social action undertaken. Thus texts are looked at not only for the textual regularities they display, and therefore also the generic conventions they flout, but also for the class, gender, and ethnic biases they incorporate. Thus, the ways in which the texts position readers or other participants and the ways in which the texts function as discursive practices are explored. Hodge and Kress's (l988: 8-10) analysis of a billboard advertisement helps to exemplify this approach. Their analysis focuses not on the advertisement format, text structure and lexico-grammatical features, but rather it explores the way that

contingent social structures and practices are realized in the advertisement. It considers how the text is institutionally legitimized, how the readers are positioned (e.g. are they positioned as consumers, and in what ways are they able to respond?). The analysis also considers what particular gender, class and ethnic bias is evident in the advertisement (e.g. do the writers of the text want to appeal to a female audience, and if so what 'feminine' stereotypes i~;, the advertisement appealing to?). What is not in the text is also considered (e.g. what ideological messages are carried by the very absence of certain features?). This new conception of genre in CDA sees genres as both social and textual categories, no longer fixed and immutable but dynamic and changing. While recognizing that there are generic conventions in text, both Fairclough and Kress stress the need to see genres not as fixed and rigid schema but as abstract, ideal categories open to negotiation and change. As Kress argues: Genres are dynamic, responding to the dynamics of other parts of social systems. Hence genres change historically; hence new genres emerge over time and hence, too, what appears as 'the same' generic form at one level has recognisable distinct forms in differing social groups. (Kress 1987: 42) Within this general retheorizing of genre, CDA makes passing references to the nature of conversation itself as a generic form. It recognizes the essential similarity which conversation has with other genres, in that it is motivated by difference. It also recognizes conversation as varying from other genres in significant ways. Kress, for example, suggests that of all genres, conversation is that with the "least or no power difference", since: in a conversation the participants all speak 'on their own behalf and take turns on their own initiative, without being directed by anyone member of the group. That is, the distribution of power in the interaction is such that the genre of conversation does not provide for anyone participant to assume a differentiated directing role. (Kress 1985: 25) One result of this power equality is, he suggests, that the "mechanisms of interaction", by which he seems to mean the turn-taking procedures, are less foregrounded, allowing the content to be most salient. In contrast, he points out that in genres in which power is unequally distributed (his example is educational genres), the reverse applies. The unequal power foregrounds the interactive conventions, thereby rendering least salient the content or substance of the interaction (Kress 1985: 25).

We agree with Kress that overt differences of power in hierarchically structured interactions generate or are reflected in more "closed" interactions (Kress 1985: 26). However, we believe that the data we will be presenting in the next six chapters challenge the claim that casual conversation is "the genre... which is formally least about power" (Kress 1985: 26). In fact, what we will show, through detailed, systematic analysis, is that conversation is always a struggle over power - but that the struggle goes "underground", being disguised by the apparent equality of the casual context. One of the tasks of our analysis is to denaturalize some of the means (e.g. humour) by which the power differences in conversation are rendered opaque for both interactants and "casual" observers. One final insight we take from CDA is to recognize our own positions as "readers" of the texts we will be presenting: The interpretation of texts is a dialectal process resulting from the interface of the variable interpretative resources people bring to bear on the text, and properties of the text itself. (Fairclough 1995a: 9) We recognize that the interpretations of the conversations that we will be presenting throughout this book implicate our own sociohistorical discursive practices. 2.8 CONCLUSION In this chapter we have reviewed ideas from a range of approaches to discourse analysis that we consider useful in the analysis of casual conversation. In the following chapter we begin our analytical exploration of casual talk by developing a description of the grammar of casual conversation as a resource for realizing interpersonal meanings. In relating grammatical patterns to relations of status, familiarity, affective involvement and affiliation, we will be making a first connection between the "differences" which we negotiate in the micro-semiotic encounters of everyday talk and the "differences" which permeate the macro-social structures of the cultural context within which casual conversations take place. NOTES 1. All the approaches mentioned so far are cogently reviewed in Schiffin (1994) . 2. It is important also to mention the applied work on genre by Swales (1990) who is concerned to show how a 'genre centred' approach to second language teaching 'offers a workable way of making sense of the myriad communicative events that

occur in the contemporary English-speaking academy' (199: I ). His use of the term "genre" is derived from several different traditions (see 1990: 13) and as such is not dealt with separately in this chapter. 3. Note that the term "conversation" is used in the following section as it was used by the Conversation Analysts themselves: that is, it refers to all linguistic interactions, and not only to casual conversational interactions as we have defined them in Chapter I. 4. CA work can be surveyed in, for example, Sudnow (1972), Schenkein (1978), Psathas (1979), Button and Lee (1987), Atkinson and Heritage (1984), Drew and Heritage (1992). 5. This is an authentic excerpt from data collected by Eggins. 6. See Rothery (1990) for an expansion of this argument in relation to written narratives in primary school. 7. Plum (1988) argues that Labov did not encounter different storytelling texts because during the interviews if a certain stage, such as Resolution, was not mentioned he would elicit it by asking questions such as "What happened then?". This resulted in the storytelling texts having the structure of Complication /\ Resolution. 8. See also Stubbs (1983), McTear (1985) and Wells et at. (1979) for extensions of the Birmingham School approach to other discourse varieties.