Three experiments investigated the cues used by musicians and nonmusicians asked to rate their confidence that an implied harmony change had or had not occurred at a specified point in pseudo- randomly generated, isochronous, triadic melodies that conformed to the rules of elementary classical harmony. In the first experiment, a single melody note followed the putative point of harmony change, and subjects' responses were predictable on the basis of whether this note was contained in the previous triad and the size of the melodic interval between this note and the one preceding it. Musicians were relatively more influenced by the former factor, and nonmusicians, by the latter one. In the second experiment, five to seven melody notes followed the putative point of harmony change, which greatly reduced control by the harmonic relationship of the first of these to the previous triad, but the effect of melodic interval size persisted. The third experiment verified the findings of the first two by using the cues that had been identified in them as design variables and evaluating their effect with conventional analytic statistics. This experiment also parametrically manipulated the number of notes after the putative point of harmony change and showed that maximum confidence concerning the occurrence of an implied harmony change was reached with the first note that did not belong to the previous triad. Possible mechanisms for use of these cues to implied harmony changes are discussed, and directions for future research are indicated.