The silent reading test has become a ubiquitous part of the current American educational scene. It makes its first appearance near the end of first grade and continues in some form at least through senior high school-and usually appears in college and graduate placement examinations. The results of each administration become part of the student's permanent school record, available upon demand to teachers, parents, and counselors, who rely heavily upon them to separate the blue birds from the red birds, the terminal from the academic, the Ivy Leaguer from the junior college candidate. Certainly if they to play such a major role, they deserve the most careful criticism and explanation. Yet in fact they rarely get it. Teachers instead treat standardized test scores, imposing in all the trappings of a computerized printout, as an authoritative measure of the ability and accomplishment of each student. Only when a score deviates wildly from the teacher's expectation (already conditioned by dozens of previous scores amassed in the cum folders) will the teacher think to question the results, talking of a bad day, a noisy room, or the teacher with the short skirt who should have known better. In most cases it is up to the individual student to prove, if he can, that the machine's assessment of his ability has been misleading. Silent reading tests, which presumably test the ability to understand what is read, the point at issue here. These make the strong assumption, as Marks and Noll2 have pointed out, that responses are independent of specific previous knowledge, either partial or complete, or nonknowledge response biases. We have been using one series of these tests at the Child Development Laboratory, Massachusetts General Hospital, where we deal almost exclusively with children who have severe learning problems in school. The behavior of our subjects during testing, however, suggested that the tests-