The allusional pretense theory claims that ironic remarks have their effects by alluding to a failed expectation. In normal conversation, this is accomplished by violating pragmatic rules of discourse, usually the maxim of sincerity. Such violations simultaneously draw a listener's attention to the failed expectation and express the speaker's attitude (normally but not necessarily negative) toward the failed expectation. Using a variety of utterance types, 3 experiments tested the theory. The first experiment, using 4 speech act types, showed that both insincerity and allusion were perceived far more frequently in ironically intended utterances than in literally intended ones. The second experiment demonstrated that the negative attitudes frequently expressed with ironic utterances are a function of the relative frequency of positive versus negative expectations and not an intrinsic characteristic of discourse irony per se. The third experiment found that over-polite requests are more likely to be used ironically than under-polite ones, presumably because the former can serve a speaker's politeness considerations while simultaneously conveying both an intended request and the speaker's attitude. It was concluded that irony is used primarily to express a speaker's attitude toward the referent of the ironic utterance, while simultaneously fulfilling other goals as well, such as to be humorous, to make a situation less face threatening, and to serve politeness considerations.