An investigation of the construct validity of a taxonomy of reading comprehension

A test of reading comprehension containing 74 test items judged to represent levels I - IV of Bloom's taxonomy was administered to 415 Ss (83 high school juniors and seniors; 171 college freshmen and sophomores; and 161 college juniors, seniors, and graduate students). High levels of inter- and intra-judge agreement were obtained with a relatively brief training program. An examination of the item analysis data suggests that the mean item difficulty increased systematically across taxonomy levels (from I-IV) and across subsamples (high school juniors and seniors to college juniors, seniors, and graduate students). Item discrimination indices, although somewhat variable, showed no systematic trend across taxonomy levels or Ss. Kuder-Richardson reliability coefficients varied significantly between taxonomy levels but not between subsamples. In general, total score and subtest I showed relatively high K-R reliability while subtest III showed low K-R reliability. Results were subjected to item analyses, factor analysis, and analysis of variance. Four hypotheses derived from Bloom's Taxonomy of Educational Objectives were tested. Three hypotheses were not confirmed by the data and the fourth received only partial support. In general, the results contain some support for a hierarchical relationship among test items of different taxonomy levels. The analysis of variance between item difficulty resulted in a significant F value between taxonomy levels but not between grade levels. Thus, item difficulty seems to be a critical dimension in regard to the hierarchical relationship of taxonomy levels. An examination of inter-item correlations suggests a low relationship between items; in some cases the mean correlation coefficient between test items of different taxonomy levels was greater than the relationship between test items of the same taxonomy level. Factor analysis of inter-item correlation coefficients yielded 29 factors, none of which accounted for a significantly large proportion of the factor variance. No meaningful relationships between individual factors and taxonomy levels were observed. Examination of test-retest gain scores indicated that greatest gains were recorded in taxonomy level III, not level I or II. The general conclusion was that evidence favoring an inter-relationship of levels of the taxonomy as proposed by Bloom was minimal. In addition, evidence for a hierarchical relationship between taxonomy levels seems, at best, weak. Differences between behaviors associated with each taxonomy level seem quantitative rather than qualitative. Each of the levels of the taxonomy proposed by Bloom might well represent different aspects of knowing, something akin to cognitive schemas.