The effects of the length of service were tested by randomly assigning families to 3- and 6-month service contracts and, in Portland, to a control condition with no specified time limit. The location of service provision (home or office), target populations, and staffing patterns (teams or individual workers) varied among the sites. Participating families were assigned to the study over a 2-year period from 1989 to 1991. The final sample consisted of 188 families in Portland, 79 families in Pendleton, and 193 families in Baltimore. Findings on outcomes were consistent with prior research on family preservation services. Results revealed low rates of placement; low rates of subsequent maltreatment; and positive changes in child, parent, and family functioning. Findings indicated that programs should consider both the importance of time-limited services and the need to match length of service to the service model and population. Careful consideration of these dimensions makes it possible to minimize costs by providing short-term, office-based services to families with older children and longer-term, in-home treatment for families with multiple needs. Some cases may require a more comprehensive treatment model. In any case, family preservation services need to be integrated with other child welfare and community resources to provide appropriate concurrent and follow-up services. Tables and 25 references