Explicit cancelability, semantic content, and metalinguistic coding

In both philosophical and linguistic research, the explicit cancelability test is widely used to distinguish semantic contents from conversational implicatures. Assuming a straightforward relation between semantic content and explicit cancelability, a researcher might think that: if the proposition p is expressed semantically by an utterance, then p is not explicitly cancelable. In this paper, however, I argue for two amendments to this assumption. First, following Jerrold Sadock, I argue that the semantic content of an ambiguous utterance may be explicitly cancelable. Against this widely accepted view, a recent paper by Arthur Sullivan argues that the ambiguity problem should be rejected. I respond to Sullivan’s arguments and argue that the ambiguity problem remains despite Sullivan’s objections. Second, I argue that there is also another type of case complicating the relation between semantic content and explicit cancelability. In some cases where an utterance of a sentence s semantically expresses the proposition that p, certain resources that speakers have available for metalinguistic communication may be employed to make p satisfy a sufficient condition for explicit cancelability, although s is not ambiguous. These two weaknesses of the explicit cancelability test have important implications for its diagnostic use.