Four approaches to the classification of 110 students as mentally retarded were investigated. The frequency of classification of educable mentally retarded (EMR) and trainable mentally retarded (TMR) and nonretarded (NMR) Ss from three racial-ethnic groups (Anglos, Mexican Americans, and Blacks) and two socioeconomic groups (middle and low) was compared. Classification approaches included the full scale IQ approach; the psychometric approach which involves multiple criteria and multiple tests; the pluralistic approach which uses J. Mercer's Estimated Learning Potential (ELP) to adjust Wechsler Intelligence Scale for Children - Revised IQ scores; and the adaptive approach which incorporates the traditional psychometric tests and the Adaptive Behavior Inventory for Children. Results indicated that the adaptive behavior approach classified the fewest students as mentally retarded at both levels (EMR, TMR), while also declassifying the most students as not mentally retarded. The pluralistic approach classified fewer students as EMR or TMR than the modified psychometric or traditional IQ approaches, but more students than the adaptive behavior approach. The adaptive behavior approach had the greatest impact on Mexican American and Black Ss, followed by the ELP approach. The adaptive behavior approach was most conservative in the classification of middle and low socioeconomic status Ss. (Legal and practical considerations involved in nondiscriminatory evaluation and declassification of students as retarded are addressed.) (CL)