In response to the tradition of examining a new test's validity by comparing it with a well-established multiple-choice test, a study investigated whether multiple-choice tests with one right answer can measure the reading comprehension process as defined by constructivists. The study compared the information obtained from two different multiple-choice tests (one traditional with one good answer per question and one novel with multiple good answers per question) and from two free-response measures (retellings and semi-structured interviews). Subjects were 28 third graders and 25 sixth graders in Illinois--two schools were in the inner city of Chicago, one was suburban, and two were in central Illinois. Of particular interest were readers' text understandings and readers' rankings on the four measures. Qualitative results (from case studies) indicated that subjects' performance varied according to the assessments used. Results also pointed to discrepancies in the use of text characteristics and salient text information by subjects and by test questions. Different assessment formats, then, seem to tap different processes or different components of one process. The study raises questions about task demands, text characteristics, reader factors, and sources of responses in reading assessment. (One figure is included and 30 references are attached.) (SR)