ATR Human Information Process. Labs., 2-2 Hikaridai, Seika-cho, Soraku-gun, Kyoto, 619-02 Japan
To investigate how the auditory mechanism restores the noise-replaced speech segments, the apparent durations for the perceptually restored speech were measured. The results were obtained that supports the hypothesis that the subjects restore the ``missed'' part of speech based on the internal template whose temporal structure corresponds to the ``original'' sample. The threshold of duration discrimination was measured using an up-and-down two-interval task. In one interval, a ``standard'' stimulus was presented that had the temporal structure of the natural sample. In the other interval, a ``comparison'' was presented in which the third mora was either lengthened or shortened. There were four conditions in a two-factorial design. The first factor was the noise replacement in the comparison, depending on whether the part of the modified segment in the comparison was replaced by a noise. The second factor was the noise replacement in the standard. In the lengthening direction, only the noise replacement in the comparison resulted in the increment of the discrimination threshold irrespective of the noise replacement in the standard. This suggests that only the noise replacement in the modified part is effective in the perceptual judgment of the durations.