Perception of the English /s/–/∫/ distinction relies on fricative noises and transitions, not on brief spectral slices

A series of experiments compared two approaches to fricative identification, spectral template matching and articulatory dynamics. Natural-speech /s/ and /integral of/ noises from fricative-vowel or vowel-fricative syllables were cross spliced so that "hybrid" noises started out as either /s/ or /integral of/ and ended up with the other fricative noise in varying proportions. With both initial and final fricatives, listener judgments most often agreed with the longer part of the noise even when spectral templates would predict the other category. Also, the vocalic formant transitions contributed to the judgment. In another experiment, open transcriptions by four expert listeners similarly showed that all the cues were used; there were also some instances of nonspeech percepts that would be predicted by gestural models. One further experiment had subjects identify two fricatives from hybrid noises between two vocalic segments. When the order of the noises differed from the order of the transitions, the perceived ordering of the fricatives was often the reverse of the order of the noise segments. Taken together with previous results, these experiments indicate that listeners take the whole fricative noise, as well as the transitions, into account in fricative identification.