An experimental evaluation of three theories of auditory stream segregation.

Three theories of auditory stream segregation were evaluated. In two-part trials, subjects heard an induction sequence, whose effects upon an immediately subsequent test sequence were measured. The rhythm and total duration of Induction Sequence tones were varied in two experiments. The similarity between induction and test sequences aided segregation, but rhythmic predictability and longer tone durations did not. Frequency alternation during the induction sequence was not necessary to induce segregation in the test sequence. Furthermore, peripheral processes inadequately account for the segregation effects found. The data suggest that, once a distinct percept emerges from an auditory scene, properties derived from the percept (particularly changes) are fed back to control the ongoing analysis of that auditory scene. A neural adaptation to stimuli with constant properties may form part of this analysis.