How does an animal know the appropriate context in which to use its different vocalizations? I studied this problem with a songbird, the blue-winged warbler (Vermivora pinus). Mated males in nature tend to sing one song form (II) at dawn, under low light levels, and at a median rate of 15 songs/min. Their other song (I, the more familiar bee-buzz) is used later in the day, under higher light levels, and at a median rate of 6 songs/min. In the laboratory I tutored different groups of hand-reared males under normal circumstances (group 1: song II early in the morning, under low light, at a rapid rate; song I later, under high light, at a slow rate), under reversed circumstances (group 2: song I early, low light, rapid rate; etc.), and with no contextual cues (group 3: songs I and II both early and late, under both light levels, and at an intermediate rate). Few males developed entirely normal blue-wing song, but males of group 2 clearly reversed the use of song forms and components. The relationship between signal form and function in the songs of this warbler appears flexible and to some extent learned by the males.