Prosodic elements such as stress and intonation are generally seen as providing both ‘natural’ and properly linguistic input to utterance comprehension. They contribute not only to overt communication but to more covert or accidental forms of information transmission. They typically create impressions, convey information about emotions or attitudes, or alter the salience of linguistically-possible interpretations rather than conveying distinct concepts or propositions in their own right. These aspects of communication present a challenge to pragmatic theory: how should they be described and explained? This paper is an attempt to explore how the wealth of insights provided by the literature on the interpretation of prosody might be integrated into the relevance-theoretic framework (Sperber and Wilson 1986/95; Blakemore 2002; Carston 2002). We will focus on four main issues. First, how should the communication of emotions, attitudes and impressions be analysed? Second, how might prosodic elements function as ‘natural’ communicative devices? Third, what (if anything) do prosodic elements encode? Fourth, what light can the study of prosody shed on the place of pragmatics in the architecture of the mind? In each case, we hope to show that the study of prosody and the study of pragmatics can interact in ways that benefit both disciplines.