This study describes the reliability of a protocol for upper limb three-dimensional movement analysis (UL-3DMA) in children with hemiplegic cerebral palsy (HCP). The UL-3DMA is based on the ISB-recommendations, and contains a set of functional and clinically relevant tasks. Tasks were selected to reflect the characteristic movement deficits seen in children with HCP. The protocol consists of three reach tasks (forwards, upwards, sideways); two reach-to-grasp tasks (with objects requiring different hand orientations); and three gross motor tasks. Within and between session reliability was tested in a group of 12 children with HCP, aged 6-15 years. Reliability of movement duration/speed and joint angles at endpoint was assessed with the intraclass correlation coefficient; similarity of the waveforms with the coefficient of multiple correlation. Measurement errors were calculated for all parameters. Results indicated good within and between session reliability for movement duration/speed. Trunk, scapula, shoulder, elbow and wrist angles at endpoint generally showed moderately high to very high reliability. High levels of reliability were also found for scapula, shoulder and elbow waveforms and lower levels for the wrist and trunk. Within and between session measurement errors were below 5° and 7°, respectively, for most kinematic parameters. Joint angles in the transverse plane, as well as wrist flexion generally showed higher between session errors (7-10°).This study indicates that the proposed protocol is a reliable tool to quantify upper limb movements in children with HCP, providing a sound base for its clinical application. Further research is needed to establish the discriminative ability of the UL-3DMA. © 2011 Elsevier B.V.