Signals that drive motor learning can arise in different modalities, such as through vision or proprioception, and have differential importance in driving learning. For example visual feedback of hand trajectories is not required for adaptation

to novel stable (DiZio and Lackner, 2000, Scheidt et al., 2005 and Tong et al., 2002) or unstable dynamics (Franklin et al., 2007a). This result may not be unexpected because congenitally blind individuals are able to walk Bortezomib purchase and use tools (two examples of adaptation to unstable dynamics), and can adapt to the perturbing effects of a Coriolis force field (DiZio and Lackner, 2000). This demonstrates that visual feedback is not critical for adaptation to dynamics. Interestingly, when subjects were presented with no visual information regarding the errors

perpendicular to the movement direction, they could straighten their movements (adapting to the dynamics) but were unable to modify their movement direction and, therefore, unable to reach the original targets (Scheidt et al., 2005). This suggests that visual information appears to be responsible for learning the direction of the movement and path planning. Indeed, subjects without proprioception are able to adapt to visuomotor rotations (Bernier et al., 2006), suggesting that Selleck SCR7 the visual signal is enough for the remapping of movement direction planning. However, subjects without proprioception are unable to learn the correct muscle activation patterns to adapt to their self-produced joint-interaction torques during reaching (Ghez et al., 1995 and Gordon et al., 1995). Visual feedback does provide useful information for dynamical control, in particular to select different internal models of objects (Gordon et al., 1993). However, whereas visual feedback may predominately affect the learning and remapping of path planning, it appears that proprioceptive feedback predominately drives the learning and generalization of dynamics. Models of trial-by-trial adaptation have been developed to relate errors experienced on one trial to the update of internal representation of the forces or joint torques that will be produced on the subsequent trial (Kawato

et al., 1987, Scheidt et al., 2001 and Thoroughman and Shadmehr, Tryptophan synthase 2000). However, this approach is limited in several respects. First, it has been shown not to function in unstable environments, where the control of the limb impedance is required (Burdet et al., 2006 and Osu et al., 2003). Second, within the optimal control framework, motor learning should not be viewed as a process that only acts to reduce error. Indeed, other factors such as energy consumption (Emken et al., 2007), risk (Nagengast et al., 2010), and reward play a role in the determination of the manner in which adaptation occurs and may explain why subjects change to curved movements under certain circumstances (Chib et al., 2006 and Uno et al., 1989). Optimal control can predict the trajectories learned after force field adaptation (Izawa et al.