Principled Reasoning In Conversational Implicature: A Formalist Program

Conversational implicature is one of the essential pragmatic phenomena manifested in common talks. Despite its numerous superficial varieties, its interpretation is highly regularized by some principled mechanism of reasoning. The ultimate goal of this paper is to represent the principled reasoning in conversational implicature in a formal fashion.Much emphasis is laid on the two species of conversational implicature—the generalized conversational implicature (GCI) and the particularized conversational implicature (PCI), which are modeled by the Default Logic for GCI Inference (DL_G) and a Hybrid Model of PCI Inference (HMPI).As a participant-independent utterance-type-meaning that can be motivated from factors other than conformity with the maxims, GCI is standardly predicted from utterance-level cues regardless of specific contexts on the accounts of the Q-, M-, and I- principles. Its inference is characterized by default reasoning and defeasibility, based on which a nonmonotonic model is developed.The model is called the Default Logic for GCI Inference, or DLg, an adapted version of Reiter's (1980) default logic and its refinements. The three components of it are DLg-1 for scalar and I-induced implicatures, DLg-2 for M-induced implicatures, and DL_G-3 for clausal implicatures and implicature cancellations, which are implemented in that order.On the other hand, PCI is an utterance-token-meaning associated with contextual interferences. Like GCI, it is derived by nonmonotonic reasoning from the participant-independent utterance and is observed as meaning other than arising from the maxim manipulations. Unlike GCI, it takes more elusive forms in the specific context saliently represented by the prior utterance, and its inference is a multiplex combination of (agenda) relevance, abductive and default reasoning, utterance context and rhetorical relations.The model to characterize the principled reasoning patterns in PCI is called a Hybrid Model of PCI Inference, or HMPI. It has absorbed the contributions by various theories and approaches, none of which alone can provide an adequate account of PCI inference. The end product is a 6-tuple model based on a modified version of agenda relevance, with the notion of agenda substantiated by the logics of conversation. In this light, PCI represents an optimized solution among all possible...