An Educational Testing Service (ETS) procedure was evaluated, which is based on item response theory and estimates true scores on tests not taken. The reading, vocabulary, and mathematics tests of high school seniors from the National Longitudinal Study (NLS) of 1972 and the High School and Beyond (HSB) seniors of 1980 and 1982 were found to share common blocks of items but differ in terms of total test length and overall difficulty. As the common blocks were too short to permit reliable comparisons, LOGIST was used to simultaneously estimate the item and person parameters for all three cohorts. The estimate of a student's ability (theta) based on the achievement test taken was combined with the item parameters for a test not taken. Thus, a number-right true score or "estimated number-right score" was obtained for a 1982 HSB senior on a 1972 NLS mathematics test. Since the group's expected true score equaled its expected number-right score, the desired cohort comparisons across time could be made. In an evaluation of this method, which generates hypothetical true scores based on tests not actually taken, comparable test forms (X and Y) of varying common block size were simulated for three conditions. Two hundred simulated examinees each for the three conditions were used. For each simulated examinee, the probability of a correct response was computed for each item. Comparisons of simulated true scores with true scores based on LOGIST estimates were made for both the tests taken and the hypothetical tests not taken. Results indicate that the ETS method should remain experimental. (TJH)