Portal de Periódicos da CAPES

Octave generalization and melody identification

1978; Springer Science+Business Media; Volume: 23; Issue: 1 Linguagem: Inglês

10.3758/bf03214301

1532-5962

Diana Deutsch,

Music and Audio Processing

In a recent article, House (1977) describes an ex­periment which he interprets as evidence against thetwo-channeltheory for the abstraction of pitch rela­tionships proposed by Deutsch (1969). The method­ology used in this experiment was, however, in­appropriate for a test of this theory.According to the model of Deutsch (1969), theabstraction of pitch relationships takes place alongtwo separate and parallel channels. Along one ofthese channels, there is convergence of informationfrom neural units underlying tones which are separ­ated by octaves. This gives rise to the strong percep­tual similarity between single tones which stand inoctave relationship. Such tones are said to have thesame tone chroma. It is further proposed thatthe harmonic equivalence of chords whose com­ponents are placed in different octaves are mediatedby higher order convergence of information alongthis channel. Along the second channel, there is nosuch octave convergence, but instead abstractionof musical intervals takes place-thatis, abstractionof the ratios between tonal frequencies. This secondchannel thus mediates transposition of tonal se­quences or melodies.The two-channel hypothesis therefore predictsthat where judgments of single tones are concerned,octave generalization effects should occur. However,where judgments of melodic relationship are con­cerned, provided that there is no artifact due tooctave generalization for single tones, octave general­ization effects should not operate.This theory was supported in an experiment byDeutsch (1972), in which subjects were played thefirst half of the tune Yankee Doodle with itscomponents placed randomly in three octaves (withthe restriction that no two successive tones occurredas in the untransformed melody). The subjects wereasked to name the tune, but were provided with noclues to its identity, apart from being assured thatit was well known. Itwas found that the percentagecorrect recognition under these conditions was nobetter than when the sequence was played as a seriesof clicks with the pitch information removed entirely.Itwas concluded that octave generalization does notoccur along the channel that processes melodicrelationships.However, the discussion proceeds with the follow­ing caution: Another finding of interest is thatwhen the experimental group was played Condition D(randomized octaves) after they had heard Condi­tion A (and so knew the identity of the tune), theSubjects reported that they were now able to followthe tune to a large extent. Thus the Subjects wereable to use octave generalization to