Another Turn in the Conversation: What Does Cloze Measure?

Abstract

This study addresses a controversy in cloze testing. At issue is whether the cloze procedure measures comprehension that ranges beyond the context immediately surrounding a cloze deletion. Eight cloze passages published over the past 15 years were analyzed, using a system that (a) estimates the quantity of text required to cue closure of any one blank (Bachman, 1985) and (b) considers the linguistic category of the deleted word. The research reported here demonstrates that across the cloze tests considered, the standard fixed-ratio cloze procedure has a high level of sensitivity to intersentential ties and lexical selections, and that the kinds of language knowledge required to complete cloze tests is virtually the same from one test to the next. The implication of these findings is that the fixed-ratio cloze procedure is far from erratic in its selection of item types. This study suggests that, for deriving tests of language comprehension, the cloze procedure produces tests that are generally consistent in the ways they measure the language knowledge of examinees.