Experiment I: A story recorded in Japanese and the one in English were partitioned by pauses whose length was one or two seconds at the end of each clause (version A), phrase (B) and word (C). No pause (D) and pause-at-every-sentence versions (E) were also made up. These versions were sent to 8 groups of Japanese high school students homogeneous in their academic levels of language. The subjects were requested to write the contents of the stories in their own language freely soon after having listened to the stories.
Results: Scores increased in the order of versions (D), (E), (A), and (B) both in Japanese and English materials (p < 0.001), but ‘the more pauses, the higher score' principle did not go on in version (C), which remarkably reduces the scores (p < 0,001). These facts are quite the same as the results of Kohno (1981). A phrase therefore may be thought of as a candidate of the perceptual unit.
Experiment II: The same procedure was taken up as Experiment I except having used college teachers of English as subjects and only English versions (A) and (B) as materials. The results were (A): x = 25.0, (B): x = 27.3, both of which were nearly the full marks. This means that the P-unit may become longer when the subject's listening ability increases, but the original P-unit as long as a phrase unit is kept up without being forgotten by the subjects.
Experiment III: Three kinds of rhythm whose inter-beat intervals (IBI) were 250, 500, 1000ms were shown to one patient with infarction involving the forebrain commissural fibers (Subject 1), and children younger than about 4 years of age, S-2) and children older than 5, S-3) and one normal adult (S-4). They were all requested to tap the table simultaneously in time with the above rhythm.
Results: S-1's left hand could not follow the slow rhythm of 500 and 1000ms IBIs, but he could follow the rapid stimuli of 250ms IBI. This phenomenon was also observed by S-2's both hands. Interestingly, the right hand of S-1 and S-3 could synchronize their tapping with all rhythms as well as S-4. Negative autocorrelations were detected among the adjacent IBIs in the slow response beats by S-3, S-4 and by the right hand of S-1, but those correlations were never found in the rapid response movements (250ms) of all the subjects nor in any responses of S-2 and of S-1's left hands.
This means that normal adults use ongoing, analytic processing to fit the slow rhythm, but holistic processing to the rapid rhythm. Evidence was found however,that S-2 and the left hand of S-1 can use only the holistic approach not only to fit the rapid rhythm but also to the slow rhythm, and that it is the very reason why they cannot follow the slow tempos. A phrase, on the other hand, consists of one to several syllables combined one to another and the length of one syllable is about 100ms (in the case of Japanese) or 100 to 300ms (in English), either of which is so rapid that it must be processed holistically. Kohno (1981) shows that extremely slow pronunciation in which eadl syllable is drawled more than 500ms decreases listenability (p < 0.01), but if the time per one syllable is shortened into 250ms, it recovers completely. Long intervals among the phrase units such as 5000ms never bring about negative effects.