Tired of stale philosophical debates, and vociferously opposed to "folk psychology," the author takes her readers on a tour of the landscape in both philosophy of mind and contemporary neuroscience. The central theme of the book, if it can be said to have but one, is: How can mental states be expressed in terms of neural processes? Churchland weighs the arguments against reducing mental states to physical states, and then to particular neural states, and finds them all wanting. She finds no difficulty in the notion that patterns of neural activity can represent, and that computations can be conceived as transitions in these neural patterns.
The book is divided into three sections. The first presents the basics of neuroscience for the philosopher, the second presents the basics of philosophy of mind for the neuroscientists, and the third explores some of those new possibilities. Even as recent a book as this must inevitably fail to fully capture the current scene, but it lays a fine foundation, and should be read by anyone interested in the emergence of a powerful new domain.
Cognitive scientists, the artificial intelligentsia, physicists and mathematicians in search of a good problem, and selected denizens of a host of other scientific domains have recently descended upon the brain. Though she does not focus on it in Neurophilosophy, Churchland documents how most of these folks previously paid little attention to the actual wetware of the brain in their theories of how the mind worked. Instead, the symbol-crunching serial computer served as a model, and all that mattered about the brain was that its elements could instantiate the fundamental logical operations.
We now know that the brain processes vast amounts of information in parallel, and that it is only by doing so that it can out perform the very best computer at such things as face recognition. We also know that most of this processing takes place in dense networks of highly, yet precisely, interconnected nerve cells, networks which could easily serve to represent features of experience in terms of particular patterns of activation.
Churchland discusses some of these advances in the final section of the book, but events have out-stripped even her optimistic assessments. Parallel distributed processing (PDP) networks can be analyzed in the language of nonlinear dynamics, and this has opened up the investigation of neural representation and computation to the use of the tools of modern physics and mathematics. There is so much activity in this new domain—variously called "neural networks" or "computational neurobiology" or the like—that one could be excused for feeling, as Churchland obviously does, that a utopian mind/brain science is on the horizon. Neurocomputers are already all the rage, and wafers with as many switching elements as the brain, and capable of running as fast as, or faster, than real time are within sight.
However, some cautionary notes really are in order. Exciting as connectionism, PDP models, and neural networks are, there remain some serious problems in fleshing them out. One that looms particularly large is that of development— how do these networks get built up, and how much of the "program" that is realized in their internal wiring results from such development?
Another problem, central to Churchland's concerns, is the level of detail at which neuroscientific structure will gracefully enfold cognitive function. PDP models, connectionist networks, and neural nets all utilize rather primitive notions of the nervous system at present, being neurally inspired rather than neurally informed. In fact, the details about the nervous system presented in the first section of Churchland's book go far beyond what is in evidence in most current models generated outside of the neuroscientific community. A question for the future, then, is the extent to which networks of real "neurons," hooked up in anatomically correct ways, and fitted out with physiologically appropriate parameters, can be endowed with cognitive functions. Will there turn out to be unique neuronal "signatures" whose presence will tell the investigator that certain computations are being performed? Will the very nature of human mentality be inextricably linked to the biological stuff of which it is made?
The point of this timely, readable, and most enjoyable book is that we cannot prejudge the issue. I agree.