Reflection: Types of inference in utterance processing

One of the key measures of processing effort in relevance theory is the amount of inferential work that is required in order to derive explicatures and implicatures. These inferences are generally characterized as “pragmatic” as opposed to “logical”. Logical inference involves a chain of reasoning in which the truth of the premises guarantees the truth of the conclusions (leading to entailments). Pragmatic inference, on the other hand, results in conclusions that may seem necessary, permissible or reasonable to draw, but their truth is not guaranteed (leading to implicatures, for instance) (Wood 2010). However, the pragmatic inferences that are said to underlie utterance processing by Relevance theorists (as well as neo-Griceans such as Levinson) are nevertheless pseudo-logical, in the sense that they still involve moving from premises (or assumptions) to conclusions. This kind of “logical” inferencing contrasts with the “associative” inferencing we briefly introduced.

Recently, Relevance theorists have proposed a rather useful distinction between intuitive and reflective inference (Mercier and Sperber 2009). An inference is intuitive when a user accepts conclusions without attending to reasons, and so it is a representational process. An inference is reflective when a user derives conclusions from premises through reasoning. The latter thus requires reflection, that is, thinking about one’s thoughts, which is a metarepresentational process, albeit not necessarily a consciously experienced one. This distinction is useful as it sidesteps the problem of calling the kinds of inferential work we normally assume to underlie pragmatic meaning as “pseudo-logical” or even “pragmatic”. One question that remains somewhat open, however, is whether associative inference falls under the broader category of intuitive inference.

It is important to note that in a relevance theory account of utterance processing, it is not assumed that an explicature is always derived first, and then followed by inferences leading to implicatures. It is claimed instead that “explicatures and implicatures (i.e. implicit premises and conclusions) are arrived at by a process of mutual parallel adjustment, with hypotheses about both being considered in order of accessibility” (Wilson and Sperber 2004: 617). In this account of utterance processing, then, “Grice’s circle”, where “what is said seems both to determine and to be determined by implicature” (Levinson 2000: 186), is no longer a problem, because multiple representations of pragmatic meaning can be derived in a parallel rather than in a serial manner. In other words, users can process more than one pragmatic meaning representation at the same time. This is important because pragmatic meaning inevitably involves multiple layers of representations, as we discussed. There has thus been a shift away from the “traditional” Gricean view in pragmatics that what is said is always processed before what is implicated, to the view that in some cases the latter can be “directly” accessed in utterance processing (Gibbs 2002; Holtgraves 1999).