Feature detection process in speech perception.

One current view of speech perception is that speech is initially processed by the extraction of independent features which are then integrated over time by a synthesis process into a phonetic or syllabic percept. This feature synthesis model was contrasted with the alternative view that speech is initially processed as a continuous multidimensional whole. Two selective adaptation experiments were carried out to provide a test between these views. In the first experiment, the feature synthesis model predicted an increase in adaptation with an increase in adaptor processing time. The alternative view can also explain this result based on the perceptual simiiarity of the adaptation sequence and the test items. In the second experiment, the alternative view predicted an increase in adaptation with a decrease in adaptor processing time in a case in which the short adaptation sequence was more perceptually similar to the test items than the long sequence. The feature synthesis model cannot readily handle this reversal of the effect found in the first experiment. Some problems with the current view of the feature detection process in speech perception are discussed.