The paper reports two studies which investigate whether banked cloze procedures reflect test-takers’ discourse level comprehension processes. In Study One, two banked cloze tests were administrated to 165 second-year students in a Chinese university. They were randomly divided into two groups. Half of them took the normal test where the sentences were in their original order and the other half completed the scrambled test where the sentences were rearranged. Study Two involved 141 second-year students in the same Chinese university and the same research procedures were applied. One of the main differences between the two studies lied in the cloze texts: the passage for the first study was a loosely ordered encyclopedic description while the text for the second was a chronological narration. T-test analysis between subjects’ scores in normal and scrambled conditions revealed that the normal cloze in Study One were not sensitive to sentence scrambling while that in Study Two were. The paper concludes that there may be considerable variation in the effectiveness of banked cloze in assessing discourse level abilities. Tester intervention is necessary for developing sound banked cloze tests.
II. STUDY ONE
III. STUDY TWO