Musical melody and speech intonation: singing a different tune

Abstract

Music and speech are often cited as characteristically human forms of communication. Both share the features of hierarchical structure, complex sound systems, and sensorimotor sequencing demands, and both are used to convey and influence emotions, among other functions [1]. Both music and speech also prominently use acoustical frequency modulations, perceived as variations in pitch, as part of their communicative repertoire. Given these similarities, and the fact that pitch perception and production involve the same peripheral transduction system (cochlea) and the same production mechanism (vocal tract), it might be natural to assume that pitch processing in speech and music would also depend on the same underlying cognitive and neural mechanisms. In this essay we argue that the processing of pitch information differs significantly for speech and music; specifically, we suggest that there are two pitch-related processing systems, one for more coarse-grained, approximate analysis and one for more fine-grained accurate representation, and that the latter is unique to music. More broadly, this dissociation offers clues about the interface between sensory and motor systems, and highlights the idea that multiple processing streams are a ubiquitous feature of neuro-cognitive architectures.

Figure 1. Acoustical representations of speech and song.

The top panels show the waveforms (amplitude as a function of time) of 2-s excerpts of samples of spoken and sung speech, respectively. The bottom panels show spectrograms (frequency as a function of time) of the same sound samples; intensity is coded by a color scale in this representation. Note the prominent fundamental frequency and harmonics (horizontal lines) present in the sung speech.

Figure 2. Each panel represents the fundamental frequency (F0) contour of a spoken utterance (left side of figure) or of sung speech (right side of figure).

Note the more continuous F0 contours for speech compared to the more discrete contours for song. The blue traces are the original contours, while the red ones represent distortions in which the F0 was either compressed by 50% (top panels) or exaggerated by 50% (bottom panels). The associated sound files illustrate that the manipulation of F0 on the speech sample (Sounds S1, S2, S3) has little perceptual effect, since it continues to sound natural (in fact, the change is hardly detectable). In contrast, the same degree of F0 distortion on the music (Sounds S4, S5, S6) is readily noticeable, as the familiar melody sounds obviously out of tune.

Figure 3. Three melodies in musical notation (left) and their corresponding fundamental frequency contours (right).

Melodies B and C are identical to Melody A, except for one changed tone (indicated by red arrows in both the musical notation and the pitch traces). Melodies A and B have the same contour (up, down, up, down, down, down), whereas Melody C has a different contour (up, down, up, up, down, down). The associated sound files illustrate that Melody C (Sound S9) is generally more easily distinguished from Melody A (Sound S7) because of this contour change, whereas Melody B (Sound S8) sounds more similar to Melody A because it has the same contour.