Didier Maillat: Getting Your Inferences in Order: The Limits of Mutual Adjustment


This paper take a theoretical look at mutual adjustment (Wilson and Sperber 2002, 2012; Carston 2002). This notion constitutes a challenge for pragmatic models because not only does it introduce variability in inferential processes but it also rules out the possibility of a meaning calculation process which is sequential.

This form of parallel processing of ex/implicatures is extremely flexible, and “retroactive” (Wilson and Sperber 2012), as it can involve “backwards inferences”. However, the notion of parallel processing argued for here is not completely straightforward either, since mutual adjustment is not devoid of sequentiality. Indeed, since comprehension processes conform to the comprehension heuristic, they will have to follow “a path of least effort” determined by the order of accessibility of contextual premises.

Building up on the proposal made by Mazzarella (2013), we explore how inferences about the speaker’s knowledge constitute a challenge for mutual adjustment and parallel processing of ex/implicature. Mazzarella (2013) shows how accessibility can mislead the comprehension heuristic and lead to accidental relevance (Wilson 2000, Carston 2007). Specifically, Mazzarella envisages a corrective course to prevent accidental relevance interpretations from firing based on Sperber et al.’s (2010) Epistemic Vigilance. This paper will address some potential remaining issues in solving the problems raised by inferences about speaker knowledge using EV. First, a corrective epistemic filter will not suffice to explain how, once the accidental interpretation has been ruled out, a new interpretation can be arrived at, following a path of least effort. Second, it is not clear how the same EV module posited in Sperber et al. (2010) could be responsible for the corrective mechanisms needed to rule out accidental interpretations. Third, and more fundamentally, a post-hoc corrective mechanism that rejects an accidental interpretation faces a considerable challenge in cancelling the cognitive effects triggered by the accidental interpretation. This, in turn, calls for a re-appraisal of parallel processing.