This study was carried out to make clear the mechanism of comprehension and memory of indirect requests. Undergraduates were required to process a short story, which contained either a highly conventional or a nonconventional indirect request. Subject's main task was to decide if the illocutionary force of the indirect request was the same as that of the direct one. Right away and 1 week later, subjects were every time given a recall test for the request they had read. The decision time was significantly faster when subjects had read conventional uses of requests than when they had read nonconventional ones. The result supported the conventional meaning model (Gibbs, 1981a) in that the subjects were automatically biased toward the conventional interpretation of nonliteral expressions. Moreover, the results on the two recall tests mostly showed that the surface form of the sentence was forgotten faster than its content, but the illocutionary force was accurately recalled even 1 week later. The results were discussed in terms of the conventionality and processing time of discourse.