Tone profiles following short chord progressions: Top-down or bottom-up?

Three experiments explored the relationship between chroma-salience profile s of individual chords and tone profiles obtained after short chord progres sions. Musicians' tone profiles for diatonic progressions of one, two, and three chords were compared with predictions of three models: a bottom-up st imulus model (number of times each chroma ol:curs in the progression), a to p-down or schema-driven key model (best-fitting key profile of C. L. Krumha nsl & E. J. Kessler, 1982), and an intermediate pitch model that includes b oth top-down and bottom-up components (cumulative pitch salience; R. Parncu tt, 1989, 1993). For single chords, all predictors significantly matched to ne profiles, except the key model applied to the diminished triad. For pair s of chords, the pitch and key models consistently outperformed the stimulu s model, consistent with the assumption that a (top-down) key had been esta blished; in the pitch model, the second chord influenced the tone profile m ore than the first (recency effect). Progressions of three chords comprised forward (e.g., F-G-C) and backward (C-G-F) cadences in major and minor key s. The pitch and key models were successful for all progressions, ,ut the k ey model predicted the tonic of backward cadences in C major and minor to b e E Predictions of the stimulus model were clearly worse than those of the other models, especially for backward cadences. Both primacy and recency ef fects were observed. In summary, the pitch model was the most consistently successful model over all experiments. To successfully predict tone profile s following chord progressions, it was necessary to account not only for re cency land primacy) but also for variations in pitch salience within chords . Results are consistent with a model of tonality induction in which bottom -up processes interact in real time with top-down processes of two kinds: r ecognition of harmonic pitch patterns and recognition of key profiles.