Both on-level and out-of-level tests were administered to third and fifth grade children in order to compare the scaled scores of different level tests of the same testing program and to discover whether the relationship between levels might be distorted by ceiling or floor effects. Only reading tests were used in this study. The Stanford Achievement Tests were used for the third grade, and the Metropolitan achievement Tests were used for the fifth grade. The results are presented in tables and graphs, and the students' attitudes regarding difficulty of the tests are discussed. Overall, the comparative standard scale scores across the levels examined did not appear to interlock perfectly. Whether such measurement error was offset by reduction in the floor and ceiling effects for the out-of-level tests apparently depended heavily on two factors: the relative difficulty of the adjacent out-of-level test, and how much scores were initially depressed or inflated for the given group tested on-level. (Author/CTM)