Robotic rehabilitation of upper limb in post-stroke subjects (ST) has demonstrated to improve their activities of daily living (ADL). The transfer of training effects to ADL is usually assessed through clinical scales, that are often not sensitive enough to detect mild improvements or differences in how a functional task is performed. Based on these considerations, in this study we applied instrumented motion analysis to a group of ST patients to quantitatively describe the transfer of the effects of planar robot-therapy to a non-trained 3D task typical of ADL.