Getting your inferences in order: Limiting variability in pragmatic inferences

Abstract

This paper offers a theoretical survey of the kind of constraints posited by Relevance Theory to tackle variability in inferential processes. Focusing on inferences about speaker knowledge, it is argued that this type of inference has an impact on a great variety of other inferential process. As a consequence, any model that does not give priority to these inferences runs the risk of generating unintended interpretations and of forcing the system to restart in order to correct mistakes which appear once speaker-related assumptions have been derived. Such a risk is faced in relevance-theoretic accounts through what is known as parallel mutual adjustment of inferences.While Epistemic Vigilance has been envisaged as a possible solution (Mazzarella, 2013), it is argued that such a solution still faces some difficulties. More specifically, it is argued that pervasive higher-order explicatures also appear to require that speaker-related inferences are derived first. It is shown that a similar requirement exists for strength inheritance on inferential chains. These phenomena point in the direction of a theoretical necessity to assume priority of speaker-knowledge inferences in order to constrain inferential variability. An alternative solution is presented to prioritise speaker-knowledge inferences during mutual adjustment processes.