This study investigated relationships among teacher, parent, and observer behavioral ratings of 3- and 4-year-old children using intra-class correlations and analysis of variance. Comparisons within and across children from middle-income (MI; N = 166; mean age = 54.25 months, standard deviation [SD] = 8.74) and low-income (LI; N = 199; mean age = 51.21 months, SD = 7.22) backgrounds revealed significant agreement between the raters but also considerable differences in both ranking and absolute scores between raters. Teachers and parents consistently rated children from LI classrooms as having more behavioral problems and fewer prosocial behaviors. Results are conceptualized with respect to how differential expectations, comparison groups, and types of interaction with children can affect the evaluation of child behavior. Results point to the need for multiple sources of evaluation when assessing children for behavioral difficulties, particularly children from lower income backgrounds. (Contains 3 tables.)