Churchland presents an interpretation of some recent advances in the neurosciences which permits a reductionistic view of sensori-motor coordination, sensory representation, and potentially other mental phenomena as well.

The centerpiece of Churchland's presentation is what he calls the state space sandwich. One side of the sandwich is a map of say, a visual field in which the two dimensions of the field represent the respective angles of two eyes; the other side of the sandwich in this example could represent a motor field in which the two dimensions represent the angles of an upper arm and a forearm. Churchland shows that in this case the motor field is actually distorted compared to real space and compared to the visual field. But this hardly matters, for as long as various gridpoints on each side of the sandwich are tied together by neurons, the visual and the motor fields can be coordinated. Thus, the brain need not perform the sort of mathematical calculations we would do to specify how two functions (visual and motor in this case) identify the same point in space.

This model is both functionally and physically realistic according to Churchland. Functionally, we would expect three consequences:

1. such a system would remain partly operating despite localized damage;
2. the system will be very fast; and
3. the quality of, in this case sensorimotor coordination, will not be equal in all parts of the field if one or both of the fields is distorted along one or more axis.

All three of these conditions are biologically accurate, which suggests that the model may be valid. The model makes physical sense as well, for the layered structure we seem to be specifying is just what we find in the brain. Although the mapping of actual brain layers is only just begun, there is some evidence that the sandwich model fits at least one area of the brain (the superior colliculus).

Churchland also considers the possibility of state space sandwiches with more than two dimensions. He discusses a mechanism proposed by A. Pellionisz and R. Llinas that can handle maps of three dimensions or more -- indefinitely more. This turns out to be important because it could help us explain more complex functions than eye-hand coordination.

For state spaces can do more than emulate computations. They can also be used in representation, even of qualitative phenomena.

In color vision, for example, it is clear that to some extent colors arrange themselves on a continuum. Certain colors are more similar to one another than other colors are, and a color such as orange can be seen as being between red and yellow. Using the color cube suggested by Edwin Land, Churchland argues that color perception could be seen as occurring in a three-dimensional 'color state space'. This would again correspond to functional reality in that we know that simple genetic errors can create color blindness over broad ranges of color -- just as we would expect from such a model. Similarly taste sensation could occur within a four-dimensional 'gustatory state space', since there are four basic tastes. An olfactory state space would have six dimensions, or more.

Such an analysis could even account for facial recognition, if we posit, say, a 20-dimensional state space. Body image and even, conceivably, language could fit into this scheme as well. Considerably more theoretical and neurophysiological work needs to be done before we can validate such applications, as Churchland acknowledges. But he believes that he has presented a plausible mechanism by which a number of mental events could be reduced to simple brain structures.

Contents

Index