再探閱讀中文文句時的斷詞是否為一種基於統計學習的結果

The present study applied the Fisher and Aslin paradigm to investigating statistical learning in reading a string of linguistic symbols (Korean letter, Hebrew letters, and Chinese characters). Statistical learning was observed regardless of the type of orthography and orthographic similarity between the participants’ native language and the experimental stimuli. The results indicate that statistical learning can be induced and observed with visual linguistic stimuli. More interestingly, the present study also succeeded in inducing and observing statistical learning with the RSVP paradigm, which is more similar to normal reading than the Fiser and Aslin paradigm. The statistical learning effect (.57) observed in the RSVP of Chinese characters was comparable to the effect (.57) observed with the auditory version of the Saffran et al. paradigm using Chinese characters in our previous study. It was also similar to the effect (.58) originally reported by Saffran et al. with auditory English syllables. Two conclusions can be drawn from the current and our previous investigations of statistical learning in reading Chinese. First, in reading unspaced Chinese sentences, Chinese readers compute statistical information contained in the sentence and use that information to find word boundaries and identify words. Second, statistical learning requires sequential processing of input, and can be observed only if a task can effectively induce sequential processing. The results of the present study bear interesting implications for the teaching of reading in Chinese.