The present study investigated how young children encode a song in memory, specifically, whether they integrate or store independently in memory the text and the tune of a song. Preschool children and, for purposes of comparison, adults, each participated in two test sessions: in one they were presented three novel tunes each with rhyming text and in the other they heard the same three tunes but with nonrhyming text. Following this familiarization phase they received a recognition test. On each trial, they heard one of five types of songs; the original song, a completely new song (new words, new tune), a mismatch of the tune of one song with the words of another, old words with a new tune, and new words with an old tune. The subject's task was to decide if each test song was "exactly the same," "somewhat the same," or "not at all the same" as any one of the three original songs. At both ages the proportion of "same" responses was significantly greater for the original than mismatch songs, indicating that listeners were more likely to remember the exact pairing of text and tune than to retain the components independently. However, integration was significantly greater for adults than children. At both ages, for songs judged as similar to the original songs, the words more often than the tune proved to be the most salient feature, particularly for children. There were no significant effects of rhyme on performance at either age. The results are discussed in terms of Serafine, Crowder, and Repp's (1984) hypothesis that one integrates the tune and text of a song in memory, rather than storing components separately.