The event-related evoked potential (ERP) responses to sentence endings that either confirm or violate syntactic/semantic constraints have been extensively studied. Very little is known, however, about the corresponding situation with respect to music. The current study investigates the brain- wave (ERP) responses to perceived phrase closure. ERPs are a potentially valid measure of how language-like or uniquely musical the perception of phrase closure is. In our study, highly trained musicians (N= 16) judged whether or not novel musical phrases were closed (melodically or harmonically). Three stimulus series consisted of seven- note tunes with four possible endings: closed (tonic note or tonic chord), open/ diatonic (dominant chord or a member thereof), open/ chromatic (a chromatic note or chord outside the key of the melody), or open/white noise (a nonmusical control). One series included melodies alone, a second series included melodies harmonized, and a third series included melodies in which the melodic contexts were disrupted rather than the endings. In the recorded ERPs, a statistically significant negative drift in the waveforms occurred over the course of the context series, indicating anticipation of closure. The drift-corrected poststimulus waveforms for all series were subjected to a principal components analysis/analysis of variance. Two subject variables were also considered: sex and absolute pitch. All four stimulus types elicited identifiable responses. The waveform peaks for the four stimulus types are clearly differentiated by principal component analysis scores to two components: one with a maximum value at 273 ms and one with a maximum value at 471 ms. Taking the closed endings as the expected "standard," the waveforms for the two types of musical deviant endings were significantly below the standard at 273 ms and above the standard at 471 ms. The amount of negativity was proportional to the amount of deviance of the ending. The positive peak in the closed condition and the reduced peak in the open/diatonic condition are contrary to the normal inverse relationship between peak size and stimulus probability; the former agrees with peaks found in response to syntactic closure in language. Significant, though isolated, interactions involving both sex and absolute pitch also emerged.