Abstract

In this paper I reconsider Sweetser's proposal to include 'speech-act modality ' in the categories of modality expressed by natural language alongside the traditional cases of root and epistemic modality. I propose a reanalysis of her examples using the relevance-theoretic notion of metarepresentation. Rather than assuming that there is a separate speech-act domain for modal operators in natural language to range over, I suggest that the material embedded under modal operators is sometimes used metarepresentationally, a possibility which is independently motivated and well manifested in other logical operators. © 2000 Elsevier Science B.V. All rights reserved