Infelicitous Cancellation: The Explicit Cancellability Test for Conversational Implicature Revisited

This paper questions the adequacy of the explicit cancellability test for conversational implicature as it is commonly understood. The standard way of understanding this test relies on two assumptions: first, that that one can test whether a certain content is (merely) conversationally implicated, by checking whether that content is cancellable, and second, that a cancellation is successful only if it results in a felicitous utterance. While I accept the first of these assumptions, I reject the second one. I argue that a cancellation can succeed even if it results in an infelicitous utterance, and that unless we take this possibility into account we run the risk of misdiagnosing philosophically significant cases.