Our primary goal has been to elucidate a model of pitch memory by examining the brain activity of musicians with and without absolute pitch during listening tasks. Subjects, screened for both absolute and relative pitch abilities, were presented with two auditory tasks and one visual task that served as a control. In the first auditory task (pitch memory task), subjects were asked to differentiate between diatonic and nondiatonic tones within a tonal framework. In the second auditory task (contour task), subjects were presented with the same pitch sequences but instead asked to differentiate between tones moving upward or downward. For the visual control task, subjects were presented again with the same pitch sequences and asked to determine whether each pitch was diatonic or nondiatonic, only this time the note names appeared visually on the computer screen. Our findings strongly suggest that there are various levels of absolute pitch ability. Some absolute pitch subjects have, in addition to this skill, strong relative pitch abilities, and these differences are reflected quite consistently by the behavior of the P300 component of the event-related potential. Our research also strengthens the idea that the memory system for pitch and interval distances is distinct from the memory system for contour (W. J. Dowling, 1978). Our results are discussed within the context of the current absolute pitch literature.

In this study, event-related brain potentials (ERPs) are used to measure subjects' responses to violations of musical expectancies elicited in seventone sequences. We manipulated the structure of the sequence (strongly vs. weakly constraining), the congruity of the terminal tone (congruent vs. incongruent), the form of sequence presentation (both types intermixed vs. blocked), and subjects' musical training. The degree of expectancy violation was measured by the P300 component of the ERP known to be sensitive to subject's expectations. As predicted, musically trained subjects generate specific expectancies in strongly constraining sequences. Expectancies may also be generated in weakly constraining sequences, however, only when these are intermixed among strongly constraining sequences. Otherwise, musically trained subjects refrain from expectancy generation and rely on their knowledge of the structure of musical scales. Musically naive subjects respond, overall, slower and less accurately than trained subjects but show similar overt behavioral effects, suggesting that these differences reflect only varying degrees of processing efficacy. ERP data, on the other hand, suggest that processing differs significantly between musically trained and untrained subjects. One such difference is that musically naive subjects generate expectancies regardless of the level of constraints in the sequence.