It is widely thought that a good criterion for determining whether a certain content is (merely) conversationally implicated is to check whether that content is cancelable. It is also commonly assumed that one can test for successful cancelation by adding a cancelation clause and see if the result is felicitous. If it is, the cancelation has succeeded; if it is not, it has failed. My aim in this paper is to show that this latter assumption is mistaken. Moreover, I argue that its widespread acceptance is problematic, as it may lead to misdiagnosis of philosophically significant cases.