6.3.3 Preference

The purpose of this short section is to introduce and clarify the notion of ‘preference’, which will be a significant factor in the analysis to come in Section 8.

Levinson notes that because of the wide variety of possible responses to, for example, questions (there are not only insertion sequences of variable length but also items like re-routes – ‘Better ask John’ – or challenges to the sincerity of the question etc), this might seem to undermine the situational significance of adjacency pairs. However, preference organisation ensures that not all potential second parts are of equal standing: there is a ranking at work in which there is at least one preferred or dispreferred category of response (1983:307). Atkinson and Heritage point out that such choices ‘arise at the level of lexical selection, utterance design, and action or sequence choice’(1984:53). In spite of the associations which the word ‘preference’ may have (most attribute the notion to Sacks) commentators are keen to stress that it does not refer to the individual preferences of interlocutors, a person’s private desires, or participants’ subjective and psychological leanings (Levinson 1983:307, Heritage 1984a:207, Atkinson and Heritage 1984:53). Rather, it deals with ‘highly specialised and …institutionalised methods of speaking’ (Heritage 1984a:207) and is ‘a structural notion that corresponds closely to the linguistic concept of markedness’ (Levinson p.307 original emphasis). Let us look at some concrete examples for further clarification.


Child: Could you…could you put on the light for my .hh room
Father: Yep.


1. C: Um I wonder if there’s any chance of seeing you tomorrow
2. sometime (0.5) morning or before the seminar
3. (1.0)
4. R: Ah um (.) I doubt it
5. C: Uhm huh
6. R: The reason is I’m seeing Elizabeth.
(Levinson p.309)

In A we have a request and its acceptance (preferred); in B we have a request and its refusal (dispreferred). While this may still not take us away from ideas of personal preference (we would all like our requests to be accepted) the above commentators’ remarks that preference is situated and institutional is supported if we look at how the preferred and dispreferred responses in A and B respectively are organised. The former is immediate, minimal and not accounted for, whereas the latter, the dispreferred, is delayed, strung out over two turns, and is accountable. One of the earliest comments on this (1972) comes from Sacks in a public lecture. Whereas a preferred response ‘pretty damn well occurs contiguously’ a dispreferred response ‘may well be pushed rather deep into the turn that it occupies’ (1987:57).

Heritage, too, remarks that preferred responses have the features of:

1. simple acceptance
2. no delay

as in:

B: Why don’t you come up and see me some



I would like to


However, dispreferred turns are noticeably different in their organisational features. Sacks remarks that ‘[c]omponents like “well” and/or “I don’t know”, for instance, at the beginning of an answer turn, characteristically precede something less than agreement’ (1987:59) Levinson shows more fully that they are marked by

1. delays – a pause before delivery, displacement by use of such items as insertion sequences
2. prefaces – markers such as ‘uh’, ‘well’, apologies, token agreements
3. accounts – explanation of dispreferred act
4. declination component – often indirect and mitigated

If we apply this to exchange B above we can see there is delay in lines 3 and 4, a preface in line 4 (‘Ah um’) followed by an indirect declination (‘I doubt it’ rather than ‘no’), and finally in line 6 an account (‘The reason is I’m seeing Elizabeth’).
As further evidence of the structural rather than the psychological nature of preference, Heritage (1984a:268-9) points to Pomerantz (1984), who considered self-deprecations, where the preferred response is disagreement and not agreement (Pomerantz pp.83-90), and also to a study by Heritage and Drew (1979) where denial not admission is the preferred response to blaming as the latter could well signal a conflict between the accused and the accuser. (It must be said, however, that both these preferred responses could well lead to accounting – in the first instance, an explanation from the hearer of why the speaker’s self-deprecation is wrong, in the second, the hearer’s presentation of an alibi/excuse. Accounting, we have just seen, is generally presented as a feature of dispreferreds.) This underlines, Heritage continues, that preference organisation is strongly associated with politeness and the concept of ‘face’, and preferred and dispreferred responses can be seen to be ‘affiliative’ and ‘disaffiliative’, respectively (p.268). We will see later (8.6, 8.7) the strong connections between preference and politeness.

However, though these views are now largely accepted without too much questioning, it is possible to highlight some problems with this notion of preference, problems related to subjectivity, frequency, and context-sensitivity. If we go back to one of the early mentions of the concept by Sacks (in a public lecture in 1972) we find that it is but a tentative notion and one that does seem to involve people’s desires to some extent. Talking of question-answer adjacency pairs he says,

if a question is built in such a way as to exhibit a preference between ‘yes’ and ‘no’, or ‘yes-’ or ‘no-’ like responses, then the answerers will tend to pick that choice, or a choice of that sort will be preferred by answerers, or should be preferred by answerers.

Terms such as ‘tend to’ and ‘should’ do involve people’s desires to a more significant degree than later commentators acknowledge. Further, acknowledging preference’s intimate association with politeness, which Heritage does strongly, is an inadvertent way of connecting it with people’s subjective states, as politeness strategies are based on what the literature would call people’s ‘face wants’, that is, how they see their own public image. (A detailed discussion of politeness is to come in 8.7.) This does not negate the fact that preference does have a structural basis, but it is worth noting that is not entirely the case.

The second point too is non-structural. Bilmes (1988), in his discussion of the concept of preference, also notes the psychological point just made, and also observes that in CA literature there are references with a bias towards the significance of the frequency of preferred actions. He quotes, amongst others, Heritage & Watson as saying that confirmations are ‘massively preferred’, and Schegloff, Jefferson, and Sacks finding that self-correction is ‘vastly more common than other-correction’ (pp.172-3). We can also add here that Heritage talks of third-turn receipt objects in question-answer sequences in courtroom and news interview interaction as being ‘massively absent’ (1985:98). Bilmes is firm in his criticism of (what he sees as) such methodological lapses and reminds practitioners that ‘CA is a structural and not a statistical undertaking’ (p.173). While elsewhere I have criticised CA for being merely structural, the point here is not intended to insist on it being more structural, but simply to point out that it is incumbent on CA analysts to be consistent in the use of their own methods.

Further, context, as always, plays a significant part in preference considerations. Thus, in a therapy session, a self-deprecation would not, contra Pomerantz, necessarily get a disagreement, but might be used as the starting point for a serious discussion (Kotthoff 1993:196). We can add here that among friends a self-deprecation could well get an amused response of laughter. Moreover, there are cultural factors to take into account. Tannen (1981) and Schiffrin (1984) both point out that argument and disagreement are a normal part of (American) Jewish conversational style (Schiffrin talks of ‘Jewish argument as sociability’) and this clearly gives a different expectation concerning the preference for agreement.

Both Levinson and Heritage provide very similar tables of preferred and dispreferred turns for a variety of actions. This is Heritage’s.


Preferred Format

Dispreferred Format Response
















Table 3. Preference Format For Some Selection Types (1984a:269)

Levinson’s table of the ‘correlation of content and format in adjacency pair seconds’ has the following in addition to the above.






Expected answer


Unexpected answer or non-answer

Table 4. Correlation of Content And Format In Adjacency Pair Seconds (p.336)

Here we would like to extend these tables further and add the following pair:


Preferred response

Dispreferred response




Table 5. Preference Organisation Of The Action Of Humour

This is a not unreasonable addition. Sacks (1974) remarks that the expected response at the end of the joke he analysed is (amused) laughter. Norrick (1993), too, sees these features in a similar way: ‘we can say that joking and laughter are linked as two parts of an adjacency pair as well’ (p.23). In such formulations it seems clear that some type of amusement is the preferred response. It is worth repeating here that we are talking about amusement and not simply laughter. Hay (2001) criticises Norrick for citing joking and laughter as an adjacency pair when, she says, there are other forms of humour support. She lists the following (passim): contributing more humour, playing along with the gag, using echo or overlap, offering sympathy, contradicting self-deprecating humour, and heightened involvement in the conversation. She further notes that appreciation of humour can be withheld or there can be a complete lack of reaction. All of these examples can be subsumed in the adjacency pair formulated here. That is, the humour support items can be seen as types of ‘amusement’ (preferred response), and withholding appreciation or not reacting can be classed as ‘non-amusement’ (dispreferred response).

In this section we have seen how speech act theory shows, amongst other things, that to speak is to act, and also that we can speak directly or indirectly. Even indirect speech is interpretable as there are certain conversational maxims we follow (the chief one of which is the maxim of relevance) which, allied to shared background knowledge, allows us to assign meanings to such talk. Such features as these are also essential components of humour, which by design fully exploits them in its creation of multiple meanings. However, speech act theory is found wanting when it comes to the sequential complexity of talk in interaction, but great assistance can be had from CA with its findings on turn taking, adjacency pairs, and preference, which provide great insights into how we cooperatively organise talk in order to make sense to one another. But CA itself is of limited use when the analyst wishes to look at what lies beneath the structures of talk, so further help from ethnographic sources is called for, as will be particularly evident in the final analysis in Section 8. Before that a key feature of the disputed utterance – gender – is now examined and its relevant aspects for our purposes noted.


A Pragmatic Approach to Humour> 6.3.1 Speech Acts

6.2 Talk In Interaction>

6.3 Some Features Of Talk>

6.3.1 Turn Taking>

6.3.2 Adjacency Pairs>

6.3.3 Preference