OBJECTIVE: To evaluate the efficacy of 9 weeks of twice-weekly partial body weight-supported treadmill training (PBWSTT) for children with cerebral palsy (CP) and moderate to severe walking difficulty compared with overground walking. DESIGN: Randomized controlled trial. SETTING: Metropolitan Specialist School for children with moderate to severe physical and/or intellectual disabilities. PARTICIPANTS: Thirty-four children classified level III or IV by the Gross Motor Function Classification System were recruited and randomly allocated to experimental or control groups. Of these, 26 (15 girls, 11 boys; mean age 10 y, 10 mo +/- 3 y, 11 mo [range, 5-18 y]) completed training and testing. INTERVENTIONS: Both groups completed 9 weeks of twice-weekly walking training. The experimental group completed PBWSTT, and the control group completed overground walking practice. MAIN OUTCOME MEASURES: Ten-meter walk test (self-selected walking speed), 10-minute walk (walking endurance), School Function Assessment. RESULTS: The overground walking group showed a trend for an increase in the distance walked over 10 minutes (F=3.004, P=.097). There was no statistically significant difference in self-selected walking speed over 10 meters or in walking function in the school environment as measured by the School Function Assessment. CONCLUSIONS: PBWSTT is safe and feasible to implement in a special school setting; however, it may be no more effective than overground walking for improving walking speed and endurance for children with CP. Continued emphasis on progressive reduction of body weight support along with adding concurrent overground walking practice to a treadmill training protocol may increase the intensity of training and assist with carryover of improvements to overground walking. Treadmill training programs that include concurrent overground walking as an additional key feature of the training protocol need to be rigorously evaluated for children with CP.