Contribution of linguistic memory to recognition of speech sounds with low intelligibility: investigation with noise-vocoded speech sound

摘要

Effects of the linguistic knowledge on speech perception have been investigated by using noise-vocoded speech sound (NVSS). NVSS was synthesized from natural speech sound by replacing speech signal with noise bands while preserving the time-varying amplitude envelopes. The amplitude envelope represents relatively slower modulation, which has information of syllabification and rhythm. However, these acoustic features were not enough to explain reasons that NVSS could be recognize. It was necessary for considering effect of linguistic knowledge instead of the acoustic features. In this study, effects of familiarity and recency on word intelligibilities were examined using a word list that familiarity was controlled and direct priming technique. This study consisted of three experiments. Intelligibility of NVSS was proportional to familiarity of word. It was shown that words in easy accessibility to mental lexical knowledge were recalled readily. In addition, a percent correct of a group trained with familiar words was higher than a group with unfamiliar words. Findings suggested that the training led subjects to obtain the listening strategy spontaneously. In the experiment using the direct priming technique, it was suggested that acoustical information was used to access the mental lexicon as well as semantic information.