The purpose of the study was to establish diagnostic validity for a modified version of the Classroom Adjustment Rating Scale (CARS). To do this, two a priori multivariate contrasts were specified in order to detect differences on the three CARS scales between three groups of pupils. Subjects were 170 normal (NORM), 125 Learning Disabled (LD), and 48 Emotionally Handicapped (EH) elementary age boys who were rated by their classroom teachers. The original CARS was modified by maintaining and adding items that had high correlations with the original three scales of Acting-Out, Shy-Anxious, and Learning Problems. Results with both the CARS and modified CARS (MCARS) indicated a significant multivariate contrast between the NORM group and a weighted combination of the LD and EH groups. Also, a significant multivariate contrast between the LD and EH group resulted for the CARS and MCARS. Follow-up with discriminant analysis found 71.4% of all boys correctly classified by the original CARS scales. However, the MCARS scales resulted in only a slight improvement in overall correct classification, with 72.3% of boys correctly classified. Teacher-judged intelligence, social class, and prosocial classroom behavior were included as possible biasing variables. It was found that the set of prosocial items had considerable impact on the ratings of EH boys, thus demonstrating the importance of including positive behaviors on a rating scale.