The authors suggest that efforts to comply with the protection in evaluation procedures provisions of P.L. 94-142 (the Education for All Handicapped Children Act) have been characterized by attempts to identify a single fair test for use with specific groups of minority children, and that such efforts will not result in improved efforts to eliminate bias in assessment. Defining assessment as the process of collecting data to make decisions about pupils, the authors conceptualize a model in which instructional decisions are made using data on the history of intervention effectiveness with individual students. Operationalization of an instructional cascade and systematic documentation of pupil progress throughout levels of the cascade are proposed as alternatives to current norm referenced decision making practices. (Author)