Two experiments tested the extent to which surprise is indicated by overstatement, understatement, and irony. When events turn out unexpectedly and the expectations of characters were explicit, irony was more effective than overstatement and understatement. When expectations were not explicit and the situation had a negative outcome, overstatement was more effective than irony and understatement. The results show that the expressions do more than provide an inconsistency to trigger conversational implicature. The implications of these findings for theories of nonliteral language comprehension are discussed. (Herbert Colston)