The convergent and discriminant validities of tests designed to measure the subskills of literal and inferential reading comprehension in different content areas were assessed. A multitrait-multimethod matrix was generated, using three measuring methods: (1) a standardized reading test; (2) individually devised tests; and (3) teacher ratings. Subjects were 136 sixth-grade students. Results indicated moderate convergent validity and no discriminant validity for the subskills. Evidence from this study suggests that reading tests do not measure discrete skills in reading as well as many educators would believe, and that the reading process remains relatively stable across material in different content areas. (Author/PN)