Melody recognition revisited: influence of melodic Gestalt on the encoding of relational pitch information

Melody recognition entails the encoding of pitch intervals between successive notes. While it has been shown that a whole melodic sequence is better encoded than the sum of its constituent intervals, the underlying reasons have remained opaque. Here, we compared listeners' accuracy in encoding the relative pitch distance between two notes (for example, C, E) of an interval to listeners accuracy under the following three modifications: (1) doubling the duration of each note (C - E -), (2) repetition of each note (C, C, E, E), and (3) adding a preceding note (G, C, E). Repeating (2) or adding an extra note (3) improved encoding of relative pitch distance when the melodic sequences were transposed to other keys, but lengthening the duration (1) did not improve encoding relative to the standard two-note interval sequences. Crucially, encoding accuracy was higher with the four-note sequences than with long two-note sequences despite the fact that sensory (pitch) information was held constant. We interpret the results to show that re-forming the Gestalts of two-note intervals into two-note "melodies" results in more accurate encoding of relational pitch information due to a richer structural context in which to embed the interval.