, 2001 and Takahashi et al., 2001), add to uncertainty about the environment. Moreover, in the real world outside the laboratory, we can be uncertain what our tasks are and which actions or tasks might lead to reward rather than punishment. Such uncertainty makes the control problem more difficult. The motor system is also nonstationary, in that its properties can change on multiple timescales. Throughout growth and development, the properties of our motor system change dramatically as our limbs lengthen and change in weight. Similarly, our muscles become stronger, so that similar activation

patterns give rise to larger forces. Nerve conduction delays initially decrease in the first 2 years after birth but then increase in proportion to the lengthening of the limbs (Eyre et al., 1991). click here As we age, other changes occur with delays becoming larger (Dorfman and Bosley, 1979) and muscle strength decreasing (Lindle et al., 1997) due to the decrease in cross-sectional area (Jubrias et al., 1997) and changes in muscle fiber properties (Brooks and Faulkner, 1994). Moreover, sensory acuity also decreases with age, for example, visual acuity is reduced as we become older (Owsley et al., 1983), adding uncertainty to the visual feedback. On a shorter timescale the way our motor system responds to our motor commands can change as we interact with objects or as our muscles become fatigued. The ever-changing nature of the motor system places

a premium on our ability to adapt control appropriately. Control is further AZD6244 cost complicated by the highly nonlinear nature of our motor system. In linear systems, once the response to two different time series of motor command is known, it is straightforward to predict the response to both applied together as simply the sum of the responses. This makes control of linear systems relatively simple, because by knowing the response of the system to a simple input such as

a pulse, one knows the response to any arbitrary input. For nonlinear systems this is no longer the case. The descending motor command undergoes a highly nonlinear transformation as it is converted into endpoint force or movement. Although the output from the nervous system sets the activation level of the motor neuron pool, the number, strength, and temporal properties of the motor units crotamiton that are activated exhibit nonlinearity. Although the measured activation level of muscle fibers exhibits a roughly linear relation with muscle force in an isometric situation, this simple relation disappears once the muscles and limbs move. The force of a muscle depends on activation level in a very nonlinear manner with respect to both the muscle velocity and muscle length and is further affected by tendon properties (for a review see Zajac, 1989). In addition, the moment arms of muscles can vary by a factor of three as the joint angles change during limb movements (Murray et al., 1995 and Murray et al., 2000).