A chord-priming paradigm was employed to test predictions of a neural net model (MUSACT). Through self-organization, the network encodes the grouping of tones and chords that are pervasive in the Western musical environment and then predicts chordal expectations. The model makes a non-intuitive prediction: Following a prime chord, expectations for the target chord are based on psychoacoustic similarity at short stimulus onset asynchronies (SOAs) but on implicit knowledge of conventional relationshipxsat longer SOAs. In a critical test, two targets were selected for each prime such that one was more psychoacoustically similar to the prime and the other was more closely related based on convention. With an SOA of 50 ms priming favored the psychoacoustically similar target; with SOAs of 500 ms and longer the effect reversed and priming favored conventional relatedness. The results underscore the limitations of models of harmony based on psychoacoustics factors alone. The predicted time course of priming was tested further in the second experiment. These studies demonstrate how neural net learning models that are apropriately constrained can be subject to strong empirical verification.