Background and Purpose: Virtual reality (VR) creates an exercise environment in which the intensity of practice and positive feedback can be systematically manipulated in various contexts. The purpose of this study was to investigate the training effects of a VR intervention on reaching behaviors in children with cerebral palsy (CP). Participants: Four children with spastic CP were recruited. Method: A single-subject design (A-B with follow-up) was used. All children were evaluated with 3 baseline, 4 intervention, and 2 follow-up measures. A 4-week individualized VR training program (2 hours per week) with 2 VR systems was applied to all children. The outcome measures included 4 kinematic parameters (movement time, path length, peak velocity, and number of movement units) for mail-delivery activities in 3 directions (neutral, outward, and inward) and the Fine Motor Domain of the Peabody Developmental Motor Scales-Second Edition (PDMS-2). Visual inspection and the 2-standard-deviation- band method were used to compare the outcome measures. Results: Three children who had normal cognition showed improvements in some aspects of reaching kinematics, and 2 children's change scores on the PDMS-2 reached the minimal detectable change during the intervention. The improvements in kinematics were partially maintained during follow-up. Discussion and Conclusion: A 4-week individualized VR training program appeared to improve the quality of reaching in children with CP, especially in children with normal cognition and good cooperation. The training effects were retained in some children after the intervention.