Expectancy violations of a highly constrained musical context were studied by presenting subjects with a chord sequence (I, IV, V) that generated a strong expectancy for a specific final chord, and then completing the sequence with either a “best-possible” (Tonic), harmonically plausible (Minor), or harmonically implausible (Dissonant) resolution. Subjects determined whether it was the best-possible resolution, and in one-half of the trials made their decision known with an overt response. Several ERP waveform components showed differences among resolution types, response conditions, and electrode locations. Among the affected components were two subclasses of the P300, the first of which (P3a) was largest in response to the Dissonant in both response conditions at all electrode sites. The area and peak amplitude of the P3b varied at several electrode sites as a function of the degree of expectancy violation represented by the resolutions. The peak latency of the P3b component reflected the behavioral response time measurement in which the Tonic and Dissonant were identified 300 msec more quickly than the Minor. A comparison between frequency and timedomain data analyses demonstrates several parallels, and it is concluded that both can serve to investigate the perception and processing of the probability structure underlying musical events and contexts.