The relationship between cloze and multiple-choice tests as measures of reading comprehension at two grade levels, fifth and eighth, was investigated. The reading passages used in the testing were at difficulty levels either 2 years below, 2 years above, or at the subjects' grade level. The final subject sampling included 708 students from each grade. Each of the 35 fifth-grade classes and 32 eighth-grade classes were randomly assigned to either cloze or multiple-choice tests. The California Achievement Test was given to all subjects before the study began. Independent variables considered were the type of test, the grade level, and the passage difficulty. An overall analysis of results showed the cloze test to be sensitive to grade levels but less sensitive to change in passage difficulty. The two tests overlapped in measuring some parts of comprehension, but the multiple-choice tests measured more. It was concluded that caution should be used when applying cloze test scores as measures of comprehension until detailed analyses show what the cloze procedure is measuring and the grade and difficulty ranges for which it is appropriate. References are included. (NH)