Deviant Case Analysis

Deviant case analysis is a procedure used in conversation analysis (CA) for checking the validity and generality of proposed phenomena of conversational interaction. It bears comparison to hypothesis testing and falsification as general methodological principles. However, it has special features that make it an essential ingredient in CA.

The main aim of CA is to discover robust rules of conversational sequencing that are not merely statistically observed regularities, but ones that are demonstrably used and oriented to, as normatively expectable, by conversational participants. Through inductive analysis of a corpus of conversational data, potential rules of conversational order may be proposed, such as in greetings exchanges, or in the ways in which invitations may be either accepted or rejected. Deviant cases are those that do not appear to conform to the proposed rule or pattern.

The importance of these cases, beyond their status as disconfirmations, is that they are often the examples that turn out to best demonstrate participants' normative orientations, somewhat echoing the “breaching experiments” that were characteristic of early ethnomethodology. Unlike those experiments, however, CA's deviant cases are ones found naturally occurring in a corpus of recorded talk.

Stephen Clayman and Douglas Maynard (1995) define three uses of deviant cases in CA: (a) Deviant cases may turn out, upon analysis, to instantiate the same normative orientations as are proposed for the standard cases; (b) a new analysis may be proposed that now takes full account of both the (apparently) deviant cases, and also the main set; and (c) a deviant case may turn out to exhibit a different kind of order in its own right, such that it properly belongs in a different collection.

The classic example of the second type, where a revised analysis is proposed, is an early CA study by Emanuel Schegloff (1968) on the ways in which telephone conversations routinely begin. Using a corpus of 500 examples, Schegloff had proposed a “distribution rule for first utterances,” to the effect that the person answering the call is, expectably, first to speak. One case was found to deviate from the rule. This led to a questioning and radical revision of what the rule proposed. Rather than proposing a rule in which call answerers make first turns, which might be considered first parts of a greeting exchange (e.g., “hello,” answered by “hello”), a revised rule could be proposed that accounted for all 500 cases. This was the summons-answer sequence, in which the first utterance of “hello” (or whatever) was now understood as the second part of a different kind of pair, the answer to the summons, which was the phone ringing.