Study of the Cloze procedure with native and non-native speakers of English

This study examined. various aspects of the methodology of the close procedure to determine their effect on the nature of close tests. It was hypothesised that changes in the frequency of word deletion, in the difficulty of the original text and in the procedure used to judge acceptable restorations of the deleted word would produce significantly different close tests and would result in varying correlations with measures of English proficiency. Three texts were selected and each was subjected to the deletion of every sixth, eighth, tenth and. twelfth word, to give twelve close tests. Five procedures were developed to score the responses to these tests for the degree of similarity they showed to the deleted. words. The tests were administered to 360 adolescent native speakers of English and. 360 adult non-native speakers of English wtho were pursuing further studies in Britain. It was found that significant differences existed among close tests when deletion frequency was changed, but that se scoring procedures reduced. this effect, The change in deletion frequency had. no effect on the measurement of text difficulty, but significant interactions were observed, among the three experimental variables. Different cloze tests gave unpredictably different measures of English proficiency. A study of identical deletions showed. that no increase in the predictability of deleted word was gained. by increasing context from five words to eleven words. Since the quantity of context had no effect on predictability, it was suggested that close is essentially sentence-bound. The nature of the correlations of cloze with measures of English proficiency and the results of factor analyses suggested that cloze is a better test of syntax and lexis than of higher-order reading abilities. Implications for future use of the doze procedure are presented and suggestions made for further research