The ability to perform accurate limb movements may require learning mechanisms that continually tune the motor system. In the current study, we isolate a form of pure limb motor plasticity. Participants reached to targets that were turned off just after the onset of an initial eye movement, reappearing at a new location at the end of the reaching movement. In contrast to classical prism or virtual reality paradigms, our task eliminated sensory adaptation by always maintaining a congruency between the seen and felt limb position. We also minimized awareness and potential adaptation processes on the basis of volitional strategies by progressively increasing the size of the target perturbations. In this manner, our adaptation procedure mimicked conditions used to study saccadic adaptation. The results indicated that adaptation under these conditions led to a robust after-effect that generalized to a large range of movements within the workspace. This fully natural, nonim-posed generalization of adaptation is not expressed in a spatial coordinate system, but more likely in a joint-centered coordinate space.