Abstract
The purpose of this study was to define the different roles exercised by consonants and formant transition in recognition of speech sounds. Three sets of stimulus sounds were generated by computer from natural Japanese monosyllables. The stimuli of the first set had both natural consonants and natural formant transitions. The stimuli of the second set had natural consonants but no transition. The stimuli of the third set had natural formant transitions but no consonants. The stimuli were recorded on compact disc and presented binaurally. The subjects repeated the syllables. The subjects were 31 aphasics who had lesions in the left hemisphere and 18 normal geriatrics. It was revealed that a correct response rate at repetition could indicate performance of speech sound recognition, because the longer the duration of the consonants or transitions of the syllables were, the higher were the correct response rates of the aphasic patients. The consonant was the cue of recognition for some Japanese syllables (sa, za, ma, ba) ; however, both the consonant and transition were indispensable to recognize other syllables (pa, ta, ka, da, ga, na, la). Information on both consonant and transition was necessary to identify the point of articulation. Information concerning the consonant was required to identify the manner of articulation as well as nasality. Information on the consonant or transition was necessary for identification of vocality. The findings indicated that lesions of the left inferior supramarginal gyrus are responsible for disturbance of recognition of speech sounds. Correct and prompt recognition of human speech seemed to be realized when features are extraced initially from the consonant (first stage), then from the transition (second stage).
