The equivalency of achievement results obtained under Title I evaluation models A1 and C1 was examined. Data were reading comprehension and arithmetic scores on the Iowa Tests of Basic Skills, administered to fourth, sixth, and eighth grade students in St. Louis, Missouri. The St. Louis Public Schools had traditionally used Model A1, a pretest-post test design in which the pretest is not used for selection. Model C1 is the regression-projection model in which only data on students identified as eligible and served by a program are used to assess the impact of Title I programs against a statistical expectation. The expectation is derived from a no-treatment group containing pupils who were eligible, but scored above the cutting score for selection, and were not served by the program. Results indicated, among other findings, that Model C1 produced higher estimates at grade 4, while Model A1 produced higher estimates at grades 6 and 8. The results also raised questions about the selection criteria used to identify program participants. Although the models were not found to be different, definite conclusions could not be drawn about their equivalence. (Two brief post hoc studies, using non-Title I pupils and a mixed model, are appended). (GDC)