Annotation for Serafine, Mary Louise, Glassman, Noah, and Overbeeke, Cornell

The Cognitive Reality of Hierarchic Structure in Music

Annotation (by Bill Tilghman):

This article reports a series of six experiments designed to investigate the degree to which hierarchic structure "exists" or plays a role in the perception of music. All six experiments made use of short excerpts involving "compound" melodies, in which a single voice might be represented by two simpler underlying voices in a late-foreground reduction. For each excerpt in the first experiment, subjects heard the original melody and two reductions, one of which was correct and the other of which was a slightly erroneous "foil," and were asked to identify which of the two reductions better matched the original. Subjects were only slightly more likely to choose the "correct" reductions, and a subsequent quantitative analysis of the reductions themselves showed that a good predictor of success at this task was the degree to which notes that were metrically accented in the original were retained in the reduction (as compared to the degree to which they were retained in the foil). In the second experiment, subjects listened only to the pairs of reductions (the correct one and the foil), not the original melodies, and were asked to choose which one of each pair they found more aesthetically pleasing. Since they showed no particular preference for the "correct" reduction, the authors concluded that subjects' correct matchings in the first experiment were unlikely to be based on mere aesthetic preference. The third experiment attempted to enhance the results of the first experiment by allowing subject-controlled, unlimited hearings of the stimuli, by providing as motivation a small monetary reward for correct answers, and by increasing the dissimilarities between the correct reductions and the foils. These experimental modifications did not, however, result in better performance at choosing correct reductions over foils.
In the remaining three experiments, the authors abandoned the apparently unsuccessful strategy of having subjects listen to reductions themselves; instead, subjects were asked to compare melodies that may or may not share a common underlying structure and to make similarity judgments about these melodies. For these three experiments, the authors composed a number of "model" melodies (designated M1) and, for each model, three "comparison" melodies: a slight variation (M2) of the original model that shares with it an identical harmonic and contrapuntal structure, a "harmonic foil" (HF) whose contrapuntal structure is identical to those of M1 and M2 but with different surface harmonies, and a "contrapuntal foil" (CF) whose harmonies are identical to those of M1 and M2 but whose underlying contrapuntal structure is quite different. In the fourth experiment, subjects rated various pairs of melodies for similarity and found that M1/M2 pairs were most similar, and that pairs involving one of the models (M1 or M2) and HF were more similar than were pairs involving one of the models and CF, indicating that subjects could perceive contrapuntal-structural similarities even when musical surfaces were different. In a matching task, however, these same subjects, in cases where they chose one of the foils over M2 as a match for M1, were more likely to choose CF than to choose HF, indicating that they were less able to discriminate the structurally similar CFs from their models than the structurally different HFs. The fifth experiment replicated the rating task from the fourth experiment, except that subjects gave a similarity rating after each of five hearings for each pair. Again, HFs were rated as more similar to M1s than were CFs (and in fact as slightly more similar to M1s than were M2s!), and furthermore, these perceptions were strengthened after repeated hearings. In the sixth experiment, yet another modification of the similarity-rating task, subjects recieved a "rehearsal" or "learning" period between the hearings of the two melodies in each pair. Here there was no significant difference between their responses for M1/HF pairs and M1/CF pairs, indicating no particular ability to recognize structural similarity in the face of surface differences.
The equivocal results of these experiments with regard to the cognitive reality of heirarchic structure are further confounded by the authors' debatable operant definitions of "surface" and "structure." They take the harmonic implications of a melody to be an element of its musical surface, independent of the voice-leading implications of the melody, which represent its true "structure." That is, they assume that a melody can be altered such that its underlying harmonies are truly and significantly modified, yet its contrapuntal structure, even in the late foreground, is left untouched.
Keywords: