The purpose of this study was to evaluate seven reading comprehension tests according to the construct of comprehension which they claimed to measure. The 367 items from these tests were classified by two judges according to the following four levels indicated on a reading comprehension taxonomy: recognition and recall, inference, evaluation, and appreciation. The percentage of items in each category was compared within and across tests. Inference items and recognition tasks together accounted for 81% to 100% of the total items, while evaluation and appreciation items appeared less frequently and were far less developed. The relative weight among items varied from test to test. The following conclusions emerged from the study: most tests contained few questions that related to higher level cognitive processes, judges encountered difficulties in categorization due to differences in definitions, and the low rate of inter-judge agreement in item categorization reflects a need to control the subjective aspects of further analysis. Nine pages of references and ten appendixes are included. (MAI)