Standardized rating scales are the most often used part of the multidimensional assessment of children's behavior. However, low cross-informant correlations raise concerns about the reliability and validity of the resulting scores. This study examined whether the manner in which behavioral constructs are measured differs across informants, across time, and across child gender for the Child Behavior Checklist for Ages 4-18 (CBCL/4-8) and the Teacher Response Form (TRF). Measurement models using multigroup confirmatory factor analysis were examined for child behavior ratings obtained, for the same child, at 54 months and first grade from mothers, fathers, and teachers. Model contracts were also evaluated to determine whether the measurement of behavioral problems was different for boys and girls. This study involved approximately 700 children from the prospective, longitudinal database of the NICHD Study of Early Child Care and Youth Development. The general form and factor loadings of the CBCL 4-18 and TRF were invariant across mothers, fathers, and teachers and held regardless of the target child's gender. These results suggest that the low correlations between parent and teacher ratings of behavior problems appear more likely to results from differences in situation-specific behavior than from the manner in which they are measured. (Contains 1 figure and 2 tables.)