A test pattern consisting of 1 to 15 dots was presented for 50 ms on a computer screen. Each of fifteen college students was asked to report orally the number of dots displayed as quickly as possible. In the experimental condition, the subject was presented short sentences successively via headset during the dot-counting task. Half of the experimental subjects judged the gender of the voice reading each sentence, and the other half judged truth of the sentence. In the control condition, the subject conducted only the dot-counting task. The results showed that the span of attention (the upper limit for 50％ correct numerosity judgment) was not influenced by the concurrent task, but the reaction time to dot-counting under the concurrent task was longer than that under the control condition.
We analyzed the relationships between impression ratings and phonetic properties of monologues using the Corpus of Spontaneous Japanese (CSJ). The impression ratings were of subjective attributes such as speaking rate and speech style. Each attribute was rated on a seven-point scale. The phonetic properties were objective attributes such as speaking rate and pause ratio, as measured directly from the CSJ speech data. Using six phonetic properties and five impression ratings as variables, structural equation modeling was conducted to explain not only the relations between the impression and phonetic properties but also the relations among the impressions. The model showed that the impressions “Skillfulness” and “Liking” could be predicted from the impressions “Speech rate”, “Activity”, and “Formality”, which in turn could be related with the objective phonetic properties. Moreover, it was shown that the higher the “Skillfulness” evaluation, the higher the “Liking” evaluation.