抄録
Subjects were presented a sentence of active or passive voice, followed by various cue words either with or without delay. The task of the subject was to follow the cue word (s) and complete such a sentence as was semantically similar to the original. (a) Mean reaction time (RT) was longer for the passive voice response than for the active, and was the longest for the soshubun. (Soshu-bun is such type of sentence as “N1-wa (topic) N2-ga (subject) V-shita.” For example, “Talking of the child, a dog bit (him).”) (b) RT was shorter for the output sentence the word order of which was more frequently used. (c) Under without-delay condition, the shortest RT was obtained for the output sentence equal to the input sentence in voice. Three different kinds of models were compared to explain the above data. Among them “the direct selection of appropriate sentence type” model seemed most appropriate. Moreover, it was found that the syntactic structure of sentence was often not preserved long in memory and decayed much faster than the content.
