Cloze item types and second language comprehension

Abstract

This paper describes the further analysis of one portion of the data reported in Jonz (1989). Four 50-item fixed-ratio cloze tests were administered to 238 nonnative university freshman. Half of the subjects were randomly assigned to take the cloze tests in their naturally occurring order; the other half of the subjects were randomly assigned to take the tests after the original sequence of the sentences in the passage had been thoroughly scrambled. Additionally, half of the subjects were allowed to read the entire correctly sequenced unaltered passage prior to completing the cloze tests. After each item on the two normal-order tests had been assigned to one of five item types according to a scheme introduced in Bachman (1985) and further developed in Jonz (1990b), the patterns of response of each of the item types on all four cloze tests were then studied. The data analysis revealed that under the prior-knowledge condition, textually cohesive cloze-test items showed a significant sensitivity to sentence scrambling. The paper concludes that (1) intersentential ties are particularly salient in the comprehension processes of nonnative speakers (cf. Jonz, 1987), and that (2) in specific respect of nonnative-speaker responses, fixed-ratio cloze tests are significantly sensitive to textual variations and continuities at levels well beyond local phrase structure. That is to say, attained scores on standard fixed-ratio cloze procedure reflect to a significant extent a nonnative-speaker examinee's utilization of text-level linguistic structures and discourse comprehension processes.