This study compares the validity of three cloze procedures (traditional, limited, and least-major-constituent limited) when reading comprehension is defined as cognitive change. Over 400 seventh and eighth graders from public schools in North Carolina and Arizona were randomly assigned to one of the three cloze formats for either fiction or nonfiction passages, and to either an experimental or a control group. The experimental groups took the cloze test and one week later read the complete passage and took the cloze test again. Control group subjects followed the same procedures, but read complete passages that were unrelated to the cloze test. Results of the comparative analysis indicate that all of the cloze procedures were reliable measures, but that not all the cloze formats were consistently valid measures of reading comprehension defined as cognitive change. Two of the cloze formats (traditional and limited) were sensitive to the differential effects of reading, but only for the fiction passages. The study concludes that alternatives to cloze procedures should be sought when investigating reading comprehension as a cognitive process. (RL)