The predictive power of first-grade peer and teacher ratings of behavior: sex differences in antisocial behavior and personality at adolescence.

Peer and teacher assessments of aggression, social withdrawal, and likability using the Pupil Evaluation Inventory were recorded for 104 French Canadian girls and boys in grade 1. Self-reported delinquency and personality measures were administered to these children when they were in junior high school 7 years later. Linear regression analyses revealed significant predictive differences between the grade 1 assessment of girls and boys: For antisocial behavior, teacher and peer assessments of boys were equally good predictors, and the combination of the two assessments did not improve prediction; for girls, peer and teacher assessments taken separately were weak predictors, but, taken together, they were better predictors for girls than for boys; for personality, peers were better predictors than teachers both for girls and for boys. When categorical analyses were used to predict extreme antisocial behavior, peer and teacher assessments were equally good predictors for girls and boys. The use of peer and teacher ratings together, however, decreased the number of false positives. The implications of these findings for research and clinical work are discussed.