Kinetic outcomes are an essential part of clinical gait analysis, and can be collected for many consecutive strides using instrumented treadmills. However, the validity of treadmill kinetic outcomes has not been demonstrated for children with cerebral palsy (CP). In this study we compared ground reaction forces (GRF), center of pressure, and hip, knee and ankle moments, powers and work, between overground (OG) and self-paced treadmill (TM) walking for 11 typically developing (TD) children and 9 children with spastic CP. Considerable differences were found in several outcome parameters. In TM, subjects demonstrated lower ankle power generation and more absorption, and increased hip moments and work. This shift from ankle to hip strategy was likely due to a more backward positioning of the hip and a slightly more forward trunk lean. In mediolateral direction, GRF and hip and knee joint moments were increased in TM due to wider step width. These findings indicate that kinetic data collected on a TM cannot be readily compared with OG data in TD children and children with CP, and that treadmill-specific normative data sets should be used when performing kinetic gait analysis on a treadmill.