A multilevel modeling approach was used to investigate the influence of age composition in 70 urban preschool classrooms. A series of hierarchical linear models demonstrated that greater variance in classroom age composition was negatively related to development on the Child Observation Record (COR) Cognitive, Motor, and Social subscales. This was true when controlling for class size, general classroom quality, and socioeconomic status at the classroom level and for age, gender, and baseline ability at the child level. Additionally, to address possible concerns related to nonrandom assignment to classrooms, a series of models were run including variance in developmental age (i.e., baseline ability) at the classroom level and at the child level. The results were consistent for chronological age composition and developmental age composition at the classroom level; greater variance in classroom developmental age composition was negatively related to Time 2 scores on the COR Cognitive, Motor, and Social subscales. Furthermore, a cross-level interaction indicated that negative influence of greater variance in classroom developmental age composition was stronger for children older in developmental age. Implications for early childhood education policy are discussed.