The Relative Merits of Item Transformations and the Cloze Procedure for the Measurement of Reading Comprehension

Abstract

The relative merits and potential of item transformations and the cloze procedure are evaluated. The two strategies for assessing reading comprehension were reviewed in terms of three criteria: (1) the extent to which literal comprehension skills are validly measured; [2] the extent to which the issue of passage dependency is addressed; (3) utility for classroom teachers, reading specialists, and evaluators of reading programs. The analysis revealed that the cloze procedure offers several distinct advantages over six sentence-derived transformations. They include the assessment of both semantic and syntactic relationships, dependency on contextual clues, and ease of development, administration, and scoring. Both item generation methods were applicable to only content domains that could be decomposed into passages and sentences. Unfortunately, this limitation encompasses many of the functional skills and minimum competencies that are in the process of being measured in 38 states.

Evaluating a history text: A comparison of cloze and maze procedures

“…DISCUSSION The widely divergent scores achieved by students on the cloze and maze tests undoubtedly reflect differences in the research base and construct validity of the two instruments. Cloze, unlike maze, has an extensive research history supporting its application as one estimate of text comprehension (Berk, 1979). It is easy to construct and provides a reasonably accurate, albeit conservative match of students and text.…”

A CONSUMERs' GUIDE TO CRITERION‐REFERENCED TEST RELIABILITY

“…The progress that has been made over the past decade is reflected in the methodologies of amplified objectives, fox test specifications, mapping sentences (facet theory), item transformations, item forms, and algorithms (See MiUman, 1980;Popham, 1980;Roid & Haladyna, 1980). Proponents of the latter three strategies, in particular, claim that generating an item domain is an objective, mechanical process (Berk, 1979(Berk, , 1980a. While the precision of the different strategies varies markedly and much more research lies ahead, the strategies do provide some evidence to refute the skepticism expressed by Thorndike in 1967, "As soon as we try to conceptualize a test score as a sample from some universe, we are brought face to face with the very knotty problem of defining the universe from which we are sampling (p. 285)... the universe is considerably restricted, is hard to define, and the sampling from it is hardly to be considered random" (p. 288).…”

The developmental nature of meta‐cognition and the relationship between knowledge and control over time

“…CLOZE tasks require readers to infer the meanings of the deleted words using the semantic and syntactic constraints within the passage. CLOZE tasks have long been advocated as logical measures of literal and inferential comprehension, understanding the main idea and structural awareness (Berk, 1979). Paris and Jacobs (1984) argued that CLOZE tasks are sensitive to the types of activities used by strategic readers because they must invoke strategies to decipher the missing words and monitor the comprehensibility of the text.…”