ABSTRACT
It is often said that the human motor system possesses surplus degrees of freedom or redundancy. This redundancy bestows upon the system a great deal of flexibility, enabling the same goal outcome to be achieved in a wide variety of different conditions – if one method for achieving the goal will not work, another can be used instead. This kind of flexibility is essential for an animal that interacts with a changing, dynamic environment, but it necessarily involves the problem of having to determine, in each instance, the particular method (of the potentially infinitely many possible) by which an outcome is to be achieved. Most theorists have supposed that this problem is solved by the nervous system using an hierarchical control architecture: the intention to achieve a particular goal outcome is progressively transformed into an appropriate pattern of efference to the contributing muscles in a sequence of stages (Saltzman, 1979). Saltzman (1979), for example, suggests a seven-stage (or level) architecture in which the first (highest) level represents the goal outcome to be achieved, such as a change in the location of an object (say from on the floor to on top of the table). Subsequent levels then progressively determine how this will be done, starting with a plan for how the object should be moved (its trajectory specified in an extrinsic set of coordinates) and culminating in the efferent commands sent to the contributing muscles.
