Toward a Motor Theory of Melodic Perception

Narmour's (1990) implication-realization theory postulates a set of bottom-up principles for melodic expectancy operating independently of tonality. Narmour describes two successive notes as an "implicative interval" that generates a set of melodic expectancies for a third note. The third note forms with the second note a "realized interval" that may fulfil or violate the expectancy. The set of principles has been quantified for empirical testing by Krumhansl (1995). Krumhansl describes predictions from each principle as follows: 1) Registral Direction predicts that small intervals imply a continuation in the same direction while large intervals imply a continuation in the reverse direction; 2) Intervallic Difference predicts that small intervals imply a realized interval of similar size while large intervals imply a realized interval of smaller size; 3) Registral Return predicts a general implication to return to the pitch region of the first note of the implicative interval; 4) Proximity predicts a general implication for small intervals; 5) Closure is realized when (a) pitch contour changes, and/or (b) when a large interval is followed by a smaller interval. Several experimental studies have tested the predictions. Cuddy &;Lunney (1995) presented listeners two-note contexts, followed by a third note and collected judgments of how well the third note continued the two-note context. Results supported all principles with the exception of closure. Krumhansl (1995) presented melodic fragments and collected judgments of continuity for a note following the fragment. Results supported all principles. Thompson, Cuddy, &;Plaus (1997) asked participants to use a piano keyboard to create a melodic continuation for a two-note context. Results of analysis of the first continuation note following the two-note context supported all principles. These and other related studies lead to questions about the source(s) of these expectancies. A new hypothesis is proposed: the principles are rooted in vocal production constraints. More specifically, the perception and composition of both vocal and instrumental melodies may be influenced by vocal constraints. Preliminary evidence is presented supporting the relation between vocal constraints and melodic expectancy. Twelve participants (4 with no music training, 4 with non vocal music training, and 4 with vocal music training) were asked to sing three-note melodic sequences. All sequences were transposed into each singer's comfortable range. Three methods of analysis were used to assess singing accuracy: 1) expert judges' assessment of singing errors; 2) singers' assessment of difficulty; 3) fundamental-frequency analysis of singing errors. Results from all analyses and across training levels converged to suggest that singing accuracy was related both to 1) degree of fulfilment of the principles as quantified by Krumhansl, and 2) perceptual expectancy as quantified by listeners' judgements in previous empirical studies. This evidence supports the notion that vocal constraints are a plausible origin of the principles. It urges further experimental scrutiny of concerns and issues arising from the motor theory.