Technical problems with norm-referenced achievement testing that can lead to the erroneous evaluation of schools for Chapter 1 Program improvement is discussed, and an alternative testing model is presented. The history of Chapter 1 testing and evaluation policies is briefly reviewed, and problems with the norm-referenced model are explored. Data from a large urban district with an extensive Chapter 1 system for the Iowa Test of Basic Skills are used to demonstrate the way in which regression bias can lead to unintended consequences that impair the validity of an identification scheme. An alternative testing model, based on item response theory, is proposed that would allow schools to develop their own assessment devices. To avoid inequities, the ideal testing system would require schools receiving Chapter 1 funds to administer two tests to their students, a locally developed assessment and the state test. The Federal government could establish criteria for program identification that the school would meet, such as 75% of students attaining the school's standards, and the state test would ensure that the school's standards were appropriate. (Contains 19 references.) (SLD)