クローズ法と文理解と文章理解

Cloze procedure and the measurement of reading comprehension in disconnected sentences and continuous passages.

山田 純

読書科学

The question of the extent to which cloze tests can measure intersentential reading ability has not been investigated very extensively. If the range of contextual constraints affecting cloze performance is relatively small, viz. limited to phrases and clauses, then cloze test results should be essentially the same as those achieved on an ordinary sentence completion test.
This study attempts to determine if cloze scores are sensitive to contextual constraints across sentences. For this purpose, a scrambled-type cloze test was used along with a sequenced-type. Furthermore, the difference between continuous passages and collections of disconnected sentences is discussed in light of the cohesion theory developed by Halliday and Hasan (1976). It was postulated that a collection of disconnected sentences differs from a continuous passage in lacking items presupposed by cohesive items which connect sentences and thereby construct a well-formed text. It was hypothesized that any difference between scrambled and sequenced cloze test scores will depend upon cohesive cloze items, i. e., the mean cloze score on cohesive items of the sequenced cloze test should be significantly higher than that for cohesive items in the scrambled cloze test, whereas the mean scores of noncohesive cloze items should be the same in both tests. It was also hypothesized that adults' abilities to utilize presupposed items will enable them to score much higher than children in the sequenced cloze test; but less so in the scrambled cloze test.
Twelve third-year elementary school students and twelve university students served as subjects. All of the subjects were native speakers of Japanese. The passages used in the tests were taken from two different sets of textbooks of Japanese for third-year elementary school instruction approved by the Japanese Ministry of Education. The subjects had not previously read the passages.
A 2 x 2x2 ANOVA, adults vs. children, scrambled vs. sequenced, and cohesive vs. non-cohesive, with repeated measures on the last two factors, was computed. All of the main and interaction effects were significant, and the hypotheses were confirmed. An analysis of the responses indicated the extent to which the cloze test can measure intra- and inter-sentential reading ability.
A comparison of cohesive cloze scores between the scrambled and sequenced cloze test showed that these vary according to individual cohesive cloze items, e.g., gains of some cohesive items were less than +2 students making correct responses for a given item, while gains for some were more than +6. This result suggests that cohesive cloze scores in the sequenced cloze test are affected by intra-sentential semantico-syntactic constraints, and that the degree of cohesiveness varies, i.e., some cohesive items are strongly associated, and others are weakly associated with their presupposed items. It was suggested that cloze scores are determined by the reader’s ability to utilize both intra-sentential semantico-syntactic constraints and inter-sentential cohesive constraints.

日本語

日本読書学会

[ISSN] 0387-284X

[NCID] AN00172568