NTT Basic Res. Labs., 3-9-11 Midori-cho, Musashino, Tokyo, 180 Japan
To investigate how acoustic information is integrated into phonemic restoration, reaction times to noise-replaced plosives in a vowel-consonant-vowel (VCV) sequence were measured. There were (a) a naturally spoken VCV, (b) a VCV made by concatenating the independently spoken V and CV, and (c) a VCV made by cross splicing the initial V and the following CV. In a naturally spoken VCV, the consonant information at the end of the initial V matches that at the beginning of the following CV. In an independently spoken VCV, however, there is no relationship between the consonant information at these locations, and in a cross-spliced VCV, they conflict. When the noise-replaced plosive was perceptually restored to the original plosive, reaction times to the plosive in the naturally spoken VCV was shortest. The next shortest reaction times were to the plosive in the independently spoken stimuli, and the longest times were to that in the cross-spliced VCV. This order was reversed, however, when the noise-replaced plosive was perceptually restored to the nonoriginal plosive. These results show that the integration of acoustic information at the end of the initial V is primary and faster than the integration of information at the beginning of the following CV.