This study grew out of the observation of a remarkable sparing of emotional responses to music in the context of severe deficits in music processing after brain damage in a nonmusician. Six experiments were designed to explore the perceptual basis of emotional judgments in music. In each experiment, the same set of 32 excerpts taken from the classical repertoire and intended to convey a happy or sad tone were presented under various transformations and with different task demands. In Expts. 1 to 3, subjects were required to judge on a 10-point scale whether the excerpts were happy or sad. Altogether the results show that emotional judgments are (a) highly consistent across subjects and resistant to brain damage; (b) determined by musical structure (mode and tempo); and (c) immediate. Experiments 4 to 6 were designed to assess whether emotional and non-emotional judgments reflect the operations of a single perceptual analysis system. To this aim, we searched for evidence of dissociation in our brain-damaged patient, I.R., by using tasks that do not require emotional interpretation. These non-emotional tasks were a 'same-different' classification task (Expt. 4), error detection tasks (Expt. 5A,B) and a change monitoring task (Expt. 6). I.R. was impaired in these non-emotional tasks except when the change affected the mode and the tempo of the excerpt, in which case I.R. performed close to normal. The results are discussed in relation to the possibility that emotional and non-emotional judgments are the products of distinct pathways.