8
Conversation and preference structure
The previous chapter focused on aspects of social awareness which can have an impact on what gets communicated by what is said during an interaction. The term 'interaction' could actually apply to a very large number of quite different social encounters. For example, a teacher talking to students in a classroom is one kind of interaction; others include a doctor talking to a patient in a clinic, or individuals taking part in courtroom proceedings, attending a committee meeting, buying stamps at the post office, and dozens of other different experiences people have in which there is interpersonal exchange of talk. The kind of talk is likely to differ according to the different contexts of interaction. However, the structure of the talk, the basic pattern of 'I speak—you speak—I speak—you speak', will derive from that fundamental kind of interaction we acquire first and use most often. This is the structure of conversation. Conversation structure is what we have been assuming as familiar throughout much of the preceding discussion. It is time to look more closely at that structure as a crucial aspect of pragmatics.
Conversation analysis
There are many metaphors used to describe conversation structure. For some, conversation is like a dance, with the conversational partners coordinating their movements smoothly. For others it's like traffic crossing an intersection, involving lots of alternating movement without any crashes. However, the most widely used analytic approach is based, not on dancing (there's no music) nor on traffic flow (there are no traffic signals), but on an analogy with the workings of a market economy.
CONVERSATION AND PREFERENCE STRUCTURE 71
In this market, there is a scarce commodity called the floor which can be defined as the right to speak. Having control of this scarce commodity at any time is called a turn. In any situation where control is not fixed in advance, anyone can attempt to get control. This is called turn-taking. Because it is a form of social action, turn-taking operates in accordance with a local management system that is conventionally known by members of a social group. The local management system is essentially a set of conventions for getting turns, keeping them, or giving them away. This system is needed most at those points where there is a possible change in who has the turn. Any possible change-of-turn point is called a Transition Relevance Place, or TRP. Within any social group, there will be features of talk (or absence of talk) typically associated with a TRP.
This type of analytic metaphor provides us with a basic perspective in which speakers having a conversation are viewed as taking turns at holding the floor. They accomplish change of turn smoothly because they are aware of the local management system for taking those turns at an appropriate TRP. The metaphor can be applied to those conversations where speakers cooperate and share the floor equally. It can also be used to describe those conversations where speakers seem to be in competition, fighting to keep the floor and preventing others from getting it. These patterns of conversational interaction differ substantially from one social group to another. In order to illustrate the system at work, we will focus on the conventions of one social group—middle class English speakers in public—while remaining aware that other social groups will have substantially different assumptions about the meaning of various features.
Pauses, overlaps, and backchannels
Most of the time, conversation consists of two, or more, participants taking turns, and only one participant speaking at any time. Smooth transitions from one speaker to the next seem to be valued. Transitions with a long silence between turns or with substantial overlap (i.e. both speakers trying to speak at the same time) are felt to be awkward. When two people attempt to have a conversation and discover that there is no 'flow', or smooth
rhythm to their transitions, much more is being communicated than is said. There is a sense of distance, an absence of familiarity or ease, as in the interaction shown in [i] between a student and his friend's father during their first meeting.
[i] Mr. Strait: What's your major Dave?
Dave:       English—well I haven't really decided yet.
(3 seconds) Mr. Strait: So—you want to be a teacher? Dave:       No—not really—well not if I can help it.
(z.5 seconds)
Mr. Strait: Wha—// Where do you— go ahead
Dave: I mean it's a—oh sorry   //1 em—
As shown in [1], very short pauses (marked with a dash) are simply hesitations, but longer pauses become silences. The silences in [1] are not attributable to either speaker because each I has completed a turn. If one speaker actually turns over the floor to another and the other does not speak, then the silence is attributed to the second speaker and becomes significant. It's an attributable silence. As shown in [2], the non-response of Dave is treated, by his girlfriend, as possibly communicating something.
[z] Jan:    Dave I'm going to the store. (2 seconds) Jan: Dave?
(2 seconds) Jan:    Dave—is something wrong? Dave: What? What's wrong? Jan:    Never mind. Silence at a TRP is not as problematic for the local management system as overlap. If the expectation is that only one person speaks at a time, then overlap can be a serious problem. Returning to example [1], the final two lines illustrate overlaps, conventionally marked by a double slash (//) at the beginning of the overlapping talk. Typically, the first overlap occurs as both speakers attempt to initiate talk. In accordance with the local management system, one speaker will stop to allow the other to have the floor. However, for two speakers who are having difficulty getting into a shared conversational rhythm, the stop-start-overlap-stop pattern may be repeated.
72 SURVEY
CONVERSATION AND PREFERENCE STRUCTURE
73
The type of overlap shown in [i] is simply part of a difficult first conversation with an unfamiliar person. There are other kinds of overlap and they are interpreted differently. For many (often younger) speakers, overlapped talk appears to function like an expression of solidarity or closeness in expressing similar opinions or values. As shown in [3], the effect of the overlapping talk creates a feeling of two voices collaborating as one, in harmony.
[3] Min: Did you see him in the video?
Wendy: Yeah—the part on the beach
Min: Oh my god //he was so sexy Wendy: he was just being so cool
Min: And all the waves // crashing around him! Wendy: yeah that was really wild!
In example [3], overlap communicates closeness. In example [4], overlap communicates competition.
[4] Joe: when they were in
// power las— wait CAN I FINISH?
Jerry:     that's my point I said—
In example [4], the speakers may appear to be having a discussion, but they are, in fact, competing for the floor. The point at which overlap occurs is treated as an interruption and the first speaker actually has to make a comment about procedure (with a louder voice, shown by the capital letters in 'CAN I FINISH?') rather than about the topic of conversation.
By drawing attention to an expectation that he should be allowed to finish, the first speaker in [4] is appealing to some of the unstated 'rules' of conversation structure. Each potential speaker is expected to wait until the current speaker reaches a TRP. The most obvious markers of a TRP are the end of a structural unit (a phrase or clause) and a pause. Notice that, in [4], the first speaker has uttered 'when they were in—' at the point where the second speaker begins to talk. There is no pause and it is not the end of a phrase or clause. This is a clear interruption and breaks the 'rules'.
Normally, those who wish to get the floor will wait for a possible TRP before jumping in. Of course, those holding the floor in a competitive environment will avoid providing TRPs. To do so,
they must avoid an open pause at the end of a syntactic unit. As illustrated in [5], the speaker fills each of his pauses ('um' or 'uh'), which are placed inside, not at the end of, syntactic units. (Just prior to this turn, another speaker had attempted to take the floor, so the speaker in [5] seems concerned to protect his turn.)
[5] I wasn't talking about—um his first book that was—uh really just like a start and so—uh isn't—doesn't count really.
Another type of floor-holding device is to indicate that there is a larger structure to your turn by beginning with expressions of the type shown in [6].
[6] a. There are three points I'd like to make—first...
b. There's more than one way to do this—one example would be ...
c. Didn't you know about Melvin?—oh it was last October...
d. Did you hear about Cindy's new car?—she got it in ...
The expressions in [6a.] and [6b.] are associated with discussions of facts or opinions whereas those in [6c] and [6d.] are preludes to storytelling. In all cases, they are used to get the regular exchange of turn process suspended and allow one speaker to have an extended turn. Within an extended turn, however, speakers still expect their conversational partners to indicate that they are listening. There are many different ways of doing this, including head nods, smiles, and other facial expressions and gestures, but the most common vocal indications are called backchannel signals, or simply backchannels. Some of these are present in Mary's contributions to [7].
[7] Caller: if you use your long distance service a lot then you'll Mary: uh-uh
Caller: be interested in the discount I'm talking about because Mary: yeah
Caller: it can only save you money to switch to a cheaper service Mary: mmm
These types of signals ('uh-uh', 'yeah', 'mmm') provide feedback to the current speaker that the message is being received. They normally indicate that the listener is following, and not objecting to,
74 SURVEY
CONVERSATION AND PREFERENCE STRUCTURE 75
what the speaker is saying. Given this normal expectation, the absence of backchannels is typically interpreted as significant. During telephone conversations, the absence of backchannels may prompt the speaker to ask if the listener is still there. During face-to-face interaction, the absence of backchannels may be interpreted as a way of withholding agreement, leading to an inference of disagreement. In conversation, silence is significant and will be interpreted as meaningful.
Conversational style
Many of the features which characterize the turn-taking system of conversation are invested with meaning by their users. Even within a broadly defined community of speakers, there is often sufficient variation to cause potential misunderstanding. For example, some individuals expect that participation in a conversation will be very active, that speaking rate will be relatively fast, with almost no pausing between turns, and with some overlap or even completion of the other's turn. This is one conversational style. It has been called a high involvement style. It differs substantially from another style in which speakers use a slower rate, expect longer pauses between turns, do not overlap, and avoid interruption or completion of the other's turn. This non-interrupting, non-imposing style has been called a high considerateness style.
When a speaker who typically uses the first style gets into a conversation with a speaker who normally uses the second style, the talk tends to become one-sided. The active participation style will tend to overwhelm the other style. Neither speaker will necessarily recognize that it is the conversational styles that are slightly different. Instead, the more rapid-fire speaker may think the slower-paced speaker just doesn't have much to say, is shy, and perhaps boring or even stupid. In return, he or she is likely to be viewed as noisy, pushy, domineering, selfish, and even tiresome. Features of conversational style will often be interpreted as personality traits.
Adjacency pairs
Despite differences in style, most speakers seem to find a way to cope with the everyday business of social interaction. They are
certainly helped in this process by the fact that there are many almost automatic patterns in the structure of conversation. Some clear examples are the greetings and goodbyes shown in [8] to [10].
[8] Anna: Hello. Bill: Hi.
[9] Anna: How are you?     Bill: Fine.
[10] Anna: Seeya! Bill: Bye.
These automatic sequences are called adjacency pairs. They always consist of a first part and a second part, produced by different speakers. The utterance of a first part immediately creates an expectation of the utterance of a second part of the same pair. Failure to produce the second part in response will be treated as a significant absence and hence meaningful. There is substantial variation in the forms which are used to fill the slots in adjacency pairs, as shown in [11], but there must always be two parts.
[11] First Part
A: What's up? A: How's it goin'? A: How are things? A: Howyadoin'?
Second Part
B: Nothin' much.
B: Jus' hangin' in there.
B: The usual.
B: Can't complain.
The examples in [11] are typically found in the opening sequences of a conversation. Other types of adjacency pairs are-illustrated in [iz], including a question-answer sequence [iza.], a thanking-response [izb.], and a request-accept [izc.].
[1 z] First Part
a. A: What time is it?
b. A: Thanks.
c. A: Could you help
me with this?
Second Part
B: About eight-thirty.
B: You're welcome.
B: Sure.
Not all first parts immediately receive their second parts, however. It often happens that a question-answer sequence will be delayed while another question-answer sequence intervenes. The sequence will then take the form of Qi-Qz-Az-Ai, with the middle pair (Qz-Az) being called an insertion sequence. Although there appears to be a question (Qz) in response to a question (Qi), the assumption is that once the second part (Az) of the
76 SURVEY
CONVERSATION AND PREFERENCE STRUCTURE 77
insertion sequence is provided, the second part (Ai) of the initial question (Qi) will follow. This pattern is illustrated in [13].
[13] Agent: Do you want the early flight? (=Qi)
Client: What time does it arrive? (=Q2.)
Agent: Nine forty-five. (=A2)
Client: Yeah—that's great. (=Ai)
An insertion sequence is one adjacency pair within another. Although the expressions used may be question-answer sequences, other forms of social action are also accomplished within this pattern. As shown in [14], there is a pair which consists of making a request—accepting the request (Q1-A1), with an insertion sequence of a question-answer pair (Q2-A2) which seems to function as a condition on the acceptance (Ai) being provided.
[14] Jean: Could you mail this letter for me?
Fred: Does it have a stamp on it? Jean: Yeah. Fred: Okay.
The delay in acceptance in example [14], created by the insertion sequence, is one type of indication that not all first parts necessarily receive the kind of second parts the speaker might anticipate. Delay in response symbolically marks potential unavailability of the immediate (i.e. normally automatic) expected answer. Delay represents distance between what is expected and what is provided. Delay is always interpreted as meaningful. In order to see how delay is locally interpreted, we need some analytic terms for what is expected within certain types of adjacency pairs.
Preference structure
Adjacency pairs are not simply contentless noises in sequence. They represent social actions, and not all social actions are equal when they occur as second parts of some pairs. Basically, a first part that contains a request or an offer is typically made in the expectation that the second part will be an acceptance. An accept-
(Qi = Request)
(Q»)
(A2)
(Ai = Acceptance)
ance is structurally more likely than a refusal. This structural likelihood is called preference. The term is used to indicate a socially determined structural pattern and does not refer to any individual's mental or emotional desires. In this technical use of the word, preference is an observed pattern in talk and not a personal wish.
Preference structure divides second parts into preferred and dis-
preferred social acts. The preferred is the structurally expected next act and the dispreferred is the structurally unexpected next act. (The general patterns are presented in Table 8.1.)
First part	Second part	
	Preferred	Dispreferred
Assessment	agree	disagree
Invitation	accept	refuse
Offer	accept	decline
Proposal	agree	disagree
Request	accept	refuse
TABLE 8.1	The general patterns	of preferred and dispreferred
structures (following Levinson 1983)
In considering requests or offers as first parts, acceptance is the preferred and refusal is the dispreferred second part. In examples [i5a.-d.], the responses in each second part all represent pre-ferreds. Thus, acceptance or agreement is the preferred second part response to a request [15a.], an offer [15b.], an assessment [15c], or a proposal [13d.].
[15]
a.
b.
c.
d.
Second Part Sure.
Yes, please. Yes, it is. That'd be great.
First Part Can you help me? Want some coffee? Isn't that really great? Maybe we could go for a walk.
To get a sense of how expected these preferred second parts are in the examples in [15], imagine each of the first parts being met with silence. We might say that in any adjacency pair, silence in the second part is always an indication of a dispreferred response.
SURVEY
CONVERSATION AND PREFERENCE STRUCTURE
Indeed, silence often leads the first speaker to revise the first part in order to get a second part that is not silence from the other speaker. This may be clearer via an example, such as [16], where Jack's silence in response to Sandy's comment prompts Sandy to restate her assessment. Jack then agrees (a preferred) with Sandy's assessment.
[i 6] Sandy: But I'm sure they'll have good food there. (1.6 seconds) Sandy: Hmm—I guess the food isn't great. Jack:    Nah—people mostly go for the music.
Notice that Jack's silence occurs where he would have had to produce a disagreement (i.e. a dispreferred response) regarding Sandy's assessment. Non-response communicates that the speaker is not in a position to provide the preferred response.
However, silence as a response is an extreme case, almost risking the impression of non-participation in the conversational structure. Generally speaking, when participants have to produce second part responses that are dispreferred, they indicate that they are doing something very marked.
In example [17], the first speaker has made a statement that the second speaker appears to disagree with. Agreement would be the preferred second part, eliciting a response such as 'Yeah' or even 'I think so'. The second speaker (Julie) finds herself in the position of producing a dispreferred.
[17] Cindy: So chiropodists do hands I guess.
Julie:   Em—well—out there—they they mostly work on people's feet.
Julie's dispreferred second part is marked with initial hesitations, as if it is difficult to perform this action (essentially correcting the other). There is a delay ('em', plus pause) in getting started and the actual statement which indicates disagreement only comes after a preface ('well'), an appeal to the views of others ('out there'), and a stumbling repetition ('they they'). Even the statement contains an expression ('mostly') which makes the information less challenging to the claim in the first part. The overall effect is that this speaker is presenting herself as having difficulty and is unwilling to have to say what is being stated.
Hesitations and prefaces are also found in dispreferred second parts to invitations, as shown in [18].
[18] Becky: Come over for some coffee later.
Wally: Oh—eh—I'd love to—but you see—I—I'm supposed to get this finished—you know.
As is often the case, the expression of a refusal (a dispreferred second) can be accomplished without actually saying 'no'. Something that isn't said nevertheless gets communicated in [18]. After a preface ('Oh') and a hesitation ('eh'), the second speaker in [18] produces a kind of token acceptance (T'd love to') to show appreciation of the invitation. Then, the other's understanding is invoked ('you see') and an account is presented ('I'm supposed to get this finished') to explain what prevents the speaker from accepting the invitation. There is also a meaning conveyed here that the speaker's circumstances are beyond his control because of an obligation ('I'm supposed to') and, once again, the inviter's understanding ('you know') is invoked.
The patterns associated with a dispreferred second in English are presented as a series of optional elements in [19].
How to do a dispreferred		Examples
a.	delay/hesitate	pause; er; em; ah
b.	preface	well; oh
c.	express doubt	I'm not sure; I don't
		know
d.	token Yes	that's great; I'd love to
e.	apology	I'm sorry; what a pity
f.	mention obligation	I must do X; I'm
		expected in Y
g-	appeal for understanding	you see; you know
h.	make it non-personal	everybody else; out
		there
i.	give an account	too much work; no
		time left
)•	use mitigators	really; mostly; sort of;
		kinda
k.	hedge the negative	I guess not; not possible
80 SURVEY
CONVERSATION AND PREFERENCE STRUCTURE 8l
The overwhelming effect of a dispreferred is that more time and more language are used than in a preferred. More language essentially represents more distance between the end of the first part and the end of the second part. From a pragmatic perspective, the expression of a preferred (in response to an offer or invitation, for example) clearly represents closeness and quick connection. The expression of a dispreferred, as mapped out in [19], would represent distance and lack of connection. From a social perspective, it is easy to see why participants in a conversation might try to avoid creating contexts for dispreferreds. One obvious device for accomplishing this is to use those pre-sequences described at the end of Chapter 7. The best way to avoid a dispreferred second is not to get to the point where a first part of the pair is uttered. It must follow, then, that conversations between those who are close familiars will tend to have fewer elaborate dispreferreds than conversations between those who are still working out their social relationship. The amount of talk employed to accomplish a particular social action in conversation is a pragmatic indicator of the relative distance between the participants.
5
■
I
9
Discourse and culture
The emphasis in the preceding chapter was on the sequential structure of conversation, particularly on aspects, of the turn-taking procedures for control of the floor, with less attention paid to what speakers had to say once they got the floor. Having gained the floor, speakers have to organize the structure and content of what they want to say. They have to package their messages in accordance with what they think their listeners do and do not know, as well as sequence everything in a coherent way. If those speakers decide to write out their messages, creating written text, they no longer have listeners providing immediate interactive feedback. Consequently, they have to rely on more explicit structural mechanisms for the organization of their texts. In this expanded perspective, speakers and writers are viewed as using language not only in its interpersonal function (i.e. taking part in social interaction), but also in its textual function (i.e. creating well-formed and appropriate text), and also in its ideational function (i.e. representing thought and experience in a coherent way). Investigating this much broader area of the form and function of what is said and written is called discourse analysis.
Discourse analysis
Discourse analysis covers an extremely wide range of activities, from the narrowly focused investigation of how words such as 'oh' or 'well' are used in casual talk, to the study of the dominant ideology in a culture as represented, for example, in its educational or political practices. When it is restricted to linguistic issues, discourse analysis focuses on the record (spoken or written) of the
SURVEY
DISCOURSE AND CULTURE