The rhetorical tactic of 'posing questions that expect no answer' is historically well known to be an effective persuasive device, and one which influential speakers continue to use today. However, while skilled communicators acknowledge the strategic effectiveness of rhetorical questions in affecting interactive outcomes, and most English speakers can supply examples on demand, a review of the literature suggests that this interrogative form is as difficult to define as it is communicatively successful. This paper explores some of the limitations of traditional Speech Act Theory, and contemporary cextensions of this theoretical framework, when dealing with the purposes served by rhetorical questions. Using examples excerpted from three different sets of naturalistic data, I show how the pragmatic complexity inherent in rhetorical questions, given their diverse and imprecise definition, contributes to the difficulty in their identification and interpretation as indirect speech acts. I argue that these problems may be illuminated, but incompletely resolved, by using concepts and methods drawn from Discourse Analysis. Based on an analysis of the data, I tentatively conclude that these approaches, even if used in combination and augmented by the insights of cognitive psychology, are insufficient to identify and explain the multiple communicative functions of rhetorical questions in spontaneous conversation.