This paper analyzes data from three administrations (spring and fall 1977 and spring 1978) of reading and mathematics achievement tests across grades three through eight. The study asks two major questions: (1) how accurately are students identified for Elementary Secondary Education Act Title I services based upon the first test administration?; and (2) how accurately are students not identified based upon that test? Through proportional analysis, and comparisons of actual test performances to regression-predicted test performances for several subgroups within the tested population, it was found that students scoring low on the selection test were much more likely to score low on subsequent tests than were high scoring pupils likely to continue to score high. Regression-predicted performance was much more accurate for low scoring pupils than for high scoring pupils. Coaching is strongly suggested by these data. Implications include that the school system in question needs to change its project identification procedures and needs to monitor its testing program. (Author/CTM)