According to relevance theory (Sperber and Wilson 1986; Blakemore 1991) some cases of communication depend on the hearer recognising that a particular speech act, for example admitting, betting or promising, is being performed. These are ‘communicated’ acts. Other cases of communication do not depend on the hearer recognising that a particular speech act, for example predicting, warning or permitting, is being performed. These are ‘non-communicated’ acts. In the case of non-communicated acts communication is successful so long as the hearer recovers adequate contextual effects without having to recognise the speaker’s intentions. Against this view, I will argue that each of the speech acts considered to be non-communicated in the relevance theory literature fall into one of two categories. The speech acts in one category contribute to the strength of associated assumptions, and those in the other convey socially relevant information. I will argue that according to relevance theory both types of speech act must be recognised and that they are in fact communicated. If relevance theory is to be internally consistent, therefore, the distinction between communicated and non-communicated speech acts must be abandoned.