Abstract

A scalar implicature is a Quantity-based conversational implicature that is derived from a set of salient contrastive alternatives usually linearly ordered in semantic or informational strength such as a Horn scale. It is dependent on the non-use of the semantically or informationally stronger alternatives that could have been used in the same set. A typical example is that the use of the semantically weak existential quantifier some implicates ‘not the semantically stronger universal quantifier all’. But this is not always the case. In this article, I discuss three ‘marked’ cases of scalar implicature in a number of different languages: the use of ‘I like you’ to implicate ‘I love you’ in Chinese, the use of a general noun such as olona ‘person’ to refer to the speaker’s husband or her boy’s father in Malagasy, and the use of a semantically weak scalar expression to imply the meaning of its stronger alternatives in a face-threatening context in Chinese, Japanese, and English. Following Huang : 25–39, 2020), I call this type of marked scalar implicature, which has received very little attention in the literature ‘non-canonical scalar implicature’. I point out that the use of ‘non-canonical’ scalar implicature is implicated in a classical way, with maximum theoretical parsimony, from Grice’s co-operative principle and its component maxims of conversation. Finally, I provide a novel neo-Gricean pragmatic analysis of non-canonical scalar implicatures, combining both Horn’s and Levinson’s Q- and R/I-principles. In my account, two aspects of non-canonical scalar implicature are distinguished: epistemic and non-epistemic. For generating the non-epistemic aspects of non-canonical scalar implicature ‘from weak to not stronger’, Horn scales and the Q-principle are invoked. On the other hand, for engendering the epistemic aspects of non-canonical scalar implicatures ‘from weak to stronger‘, a scale such as is treated as forming an Atlas-Levinson rather than a Horn scale and the computation of it is subject to the R/I-principle. Therefore, a more accurate term for a non-canonical scalar implicature may be ‘Q-scalar/I-implicature’.