Third International Conference on Spoken Language Processing (ICSLP 94)
Sternberg et al.  analyzed the influence of the number of words, such as ascending numbers or a sequence of weekdays, on the reaction time (RT) to produce the words. They found that RT increased linearly as the number of words increased. Since a primary stress was included in each word used in their experiments, they concluded that a stress group comprises a programming unit in English speech production. In this paper, we modify and extend their study, and apply the RT measurement paradigm to the investigation of the process of Japanese speech production. As a first step, we investigated the influence of the number of Japanese syllables on RT. We used four material sets consisting of non-words having one to four syllables, where each syllable was [ka] or [ta]. We measured the RT for the production of the nonwords and analyzed the effect of the number of syllables on RT among each material set. Here, RT was defined as the time difference between visual presentation of the onset signal and release of the initial consonant. Analysis of variance (ANOVA) revealed a significant positive effect of the number of syllables on RT for 10 material sets out of 20 in all the five subjects, although the RT did not increase linearly as the number of syllables increased. No subject showed a significant decrease of RT for any material set. These results suggest that a syllable might be the smallest segment for a programming unit in Japanese speech production, while there remains another possibility that the programming unit might be a larger segment than a syllable. These results are discussed in terms of the production process model and difference between English and Japanese.
Bibliographic reference. Masaki, Shinobu / Honda, Kiyoshi (1994): "Estimation of temporal processing unit of speech motor programming for Japanese words based on the measurement of reaction time", In ICSLP-1994, 663-666.