This study examined the variance in teacher ratings attributable to teacher, occasion, and child on the original Behavior Assessment System for Children (BASC). Teachers rated approximately 1,100 girls (52%) and boys (48%) at four public elementary schools over 3 years. About half of the students were rated twice and about a seventh 3 times. Approximately 30% of students were White, 50% African American, 9% Hispanic, 2% multiracial, and 2% Asian. Children were in kindergarten to fifth grade. Cross-classified random effects models were used to account for multiple occasions of data nested within both children and teachers. Overall, the child accounted for the largest amount of variance in ratings (52.3%) followed by occasion (34.2%) and teacher (13.5%). However, large differences existed across composites and scales. The findings suggest that teachers' ratings of adaptive competencies, externalizing behaviors, and school problems are based primarily on child behavior/characteristics. Ratings of internalizing problems may be more situational.