Performance on Cloze Tests with Fixed-Ratio and Rational Deletions

Abstract

Although research on the cloze test has offered differing evidence regarding what language abilities it measures, there is a general consensus among researchers that not all the deletions in a given cloze passage measure exactly the same abilities. An important issue for test developers, therefore, is the extent to which it is possible to design cloze tests that measure specific abilities. Two cloze tests were prepared from the same text. In one, different types of deletions were made according to the range of context required for closure, while in the other a fixed-ratio deletion procedure was followed. These tests were administered to 910 university and pre-university students, including both native and non-native speakers of English, with approximately half assigned at random to take the fixed-ratio test and the other half taking the rationally deleted test. While both tests were equally reliable and had equal criterion validity, the fixed-ratio test was significantly more difficult. Analyses of responses to different types of deletions suggest that the difficulty of cloze items is a function of the range of syntactic and discourse context required for closure. The study also provides practical and empirically supported criteria for making rational deletions and suggests that cloze tests can be designed to measure a range of abilities.