The purpose of this study was to examine the accuracy of teachers' abilities to identify problem children. Ninety-five third grade children served as subjects. Teachers designated each child as a conduct problem, withdrawal problem, or normal, resulting in 16, 14, and 65 children being assigned to the three groups. Behavioral observations, peer sociometric ratings, and academic achievement scores were collected. The results indicated that conduct problem children differed from normal children on behavioral, sociometric, and academic measures whereas withdrawn children differed from normals on the latter two measures. Conduct problem and withdrawal problem children did not differ on any of the outcome measures. The results suggest that teachers can identify children who are having difficulties in class but may be less accurate in differentiating between different types of problem children. (Author/MP)