The recognition of text and tune in songs was examined in a music-agnosic patient and five matched controls. Listeners had to focus on one component of the song at a time (text or music) and had to decide whether the component was familiar or unfamiliar. Songs were either matched (i.e., an original familiar or an original unfamiliar song) or mismatched (a combination of a familiar component with an unfamiliar one). Normal listeners displayed response patterns that are congruent with those obtained previously in different experimental settings and which showed that text and tune are difficult to separate. Data collected in the patient, however, suggest some independence between text and music in songs. Moreover, the usual asymmetry in favor of text was much reduced when later verses were used. Overall, the results are interpreted as revealing strong association, not integration, between the musical and the verbal component of familiar songs.