Some Inter-Relationships of Readability, Cloze and Multiple Choice Scores on a Reading Comprehension Test

Abstract

Abstract. Seven passages of a published reading comprehension test were administered to college students to compare the regular multiple chioce test scores with scores on two forms of a cloze test, standard blank and dash (dashed-line) blank, and with readability formula values. The mean on the dash form was significantly higher than on the standard form and subjects' scores on both were significantly related to their scores on the multiple choice test. Regression equations were derived to predict multiple choice scores from cloze scores. At the passage level, mean scores on (both) cloze forms correlated with readability values, but multiple choice means were correlated with neither. In a supplementary study college students answered the multiple choice questions without reading the passages, in order to determine prior knowledge on each passage. When the multiple choice scores were corrected for this, they correlated positively with scores on both forms of the cloze test and with readability values.The cloze procedure (Taylor, 1953) provides a convenient method of testing reading comprehension. The "test items" consist of blanks replacing every nth (usually every 5th) word in a passage, with subjects being asked to try to replace the missing words. Item generation is simple and objective compared to the same process for multiple choice tests. Consequently, cloze scores on sets of passages provide a useful criterion measure for the development of readability formulas, and have been used for that purpose by a number of recent investigators (

Testing a readable writing approach to text revision.

“…The findings from the study are not consistent with a similar analysis carried out by Entin and Klare (1978), using the passage from an alternative form of the Nelson-Denny test. They found a correlation coefficient of only .10 between multiplechoice performance and passage readability.…”

The Psychology of Reading for Pleasure: Needs and Gratifications

“…The Fog Index shows significant positive relations with all three variables, whereas the cloze has low to verv low and nonsignificant correlations in all cases except one. On the cloze, such a failure to predict difficulty rankings seems to support the "most surprising results" reported by Entin and Klare (1978), who found near-zero rho values for comparisons of dash-line cloze difficulty rankings with comprehension test difficulty.…”

Traditional and alternative methods of measuring the understandability of accounting narratives

“…Third, the benchmark criteria points by which the Cloze procedure is measured vary widely in tests against multiple comprehension tests. Jones (1997, p. 120) looked at five tests (Bormuth, 1967(Bormuth, , 1968Rankin and Culhane, 1969;Rankin, 1971;Entin and Klare, 1978). Jones (1997, p. 119) concludes: "If their explanations are accepted, the appropriate range is between 36 and 44 per cent for the instructional levels and between 50 and 61 per cent for the independent levels.…”

“…In particular, much effort has been expended to find a criterion reference score with which Cloze scores can be compared (see, for example, Bormuth, 1967Bormuth, , 1968Entin and Klare, 1978;Rankin and Culhane, 1969). In general, the Cloze scores have been benchmarked against results from multiple-choice scores: 75% (instructional level) and 90% (independent).…”

“…Second, Shanahan, Kamil and Tobin, (1982); Kintsch and Miller, (1984) suggest that the Cloze procedure measures the ability to guess correctly missing words rather than the comprehension per se of the passages Third, the benchmark criteria points by which the Cloze procedure is measured vary widely in tests against multiple comprehension tests. Jones (1997, p.120) looked at five tests (Bormuth, 1967(Bormuth, , 1968Rankin and Culhane, 1969;Rankin 1971 andEntin andKlare 1978). Jones (1997, p. 119) concludes:…”