Four studies of the reliability and validity of needs assessment instruments developed by the Taylor Public Schools, Michigan, are described. The studies focused on the stability of student scores, classification stability, content validity, and concurrent validity. Consisting of separate tests for kindergarten, first and second grade, the instruments were designed to assist the school system in student selection for its compensatory education programs. Test scores were only modestly stable over a period of two to three weeks. The instruments did not reliably classify students for eligibility for compensatory education. Classroom teachers believed the tests measured important skills well. Correlations between teacher nomination for compensatory education and test scores were low. The studies indicate districts should place greater emphasis on teacher judgment rather than test scores to identify participants. Researchers expect that using cut off scores further from the median will result in improved classification stability. The reliability of selection test scores should be examined even if staff believes tests are adequately measuring important skills. The effectiveness of locally developed needs assessment instruments for student selection into compensatory education programs must be judged in comparison to alternative procedures. (DWH)