Reflection: Defeasibility of inference versus cancellability of pragmatic meaning

Pragmatic inference differs from logical inference in that it is defeasible, which means such inferences allow for the possibility of error. In practice, this has been taken to mean that implicatures and other pragmatic meaning representations which derive from such inferences are cancellable (as we briefly discussed in the case of presuppositions). It is important to note that defeasibility is a characteristic of a cognitive process (e.g. pragmatic inference), while cancellability is a characteristic of the product of that process (e.g. an implicature). In practice, this means that while inferences can be blocked or suspended, implicatures can only be corrected/repaired (Haugh 2013b).

Blocking inferences involves cases where a potential inference (which could conceivably follow from what is said) is not allowed through by the speaker (an anticipatory orientation). For example, we can see in the utterance Some, in fact all, the people said they liked the food, that by adding in fact all after some, the putative “not all” scalar implicature is blocked from arising (Jaszczolt 2009: 261). Suspending inferences, on the other hand, involves cases where the speaker removes his/her commitment to an inference that has more than likely already been drawn (a retroactive orientation). For instance, in the following interaction on the phone between friends, what starts out as looking like Cameron will invite Steve to go out somewhere, ends up with Cameron hinting that he would like Steve to invite him out.

We can observe here how a (potential) default inference, which arises from Cameron’s utterance in line 18, namely that Cameron is very likely going to invite Steve somewhere if given a “go-ahead” response, is subsequently suspended by a nonce inference arising from Cameron’s subsequent utterances in lines 20–21, where it becomes apparent that he is hinting that he would like Steve to invite him somewhere. Steve obliges with an invitation in line 22. In this sense, then, we can see that pragmatic inferences are always contingent on what precedes and follows the utterance in question, and this is what makes them defeasible. We will return to consider this issue further.

To correct or repair an implicature, in contrast, involves denying, retracting or clarifying what was meant (normally by the speaker), or at least attempting to do so. We saw an example of this at the beginning (see example [5.1]). Defeasibility thus arises from cognitive operations on the part of individual users, while cancellability arises through social actions that are jointly achieved by participants.