A longitudinal matched-groups design was used to examine the quality and costs of community-based residential supports to people with mental retardation and challenging behavior. Two forms of provision were investigated: noncongregate settings, where the minority of residents had challenging behavior, and congregate settings, where the majority of residents had challenging behavior. Data were collected for 25 people in each setting. We collected information through interviewing service personnel in each type of setting on the costs of service provision, the nature of support provided, and the quality of life of residents. We also conducted observations in each setting. Results suggest that noncongregate residential supports may be more cost effective than congregate residential supports.