The role that measurement error plays in the regression effect is discussed with particular reference to the RMC models for evaluating the effects of Elementary Secondary Education Act Title I programs. The norm referenced evaluation model assumes a theory of growth where the relative ranking of students remains the same from pretest to posttest in the absence of any supplementary treatment (the equipercentile growth assumption). However, if the pretest is used to select low scoring candidates for Title I programs, there is certain to be a preponderance of negative errors among these scores, so that any unbiased posttest is certain to indicate growth even in the absence of any treatment. However, it is recommended that Title I programs use three tests: a selection test, an independent pretest for the selected group, and a posttest. A regression problem remains, however, if the correlation between the selection test and the pretest is not the same as the correlation between the selection test and the posttest. (Author/CTM)