Webbed Reprint Collection|
William H. Calvin
University of Washington
Seattle WA 98195-1800 USA
Email || Home Page
Book review, |
by William H. Calvin, of
Gerald M. Edelman's book,
as it appeared in
copyright ©1988 by William H. Calvin and the American Association for the Advancement of Science
While not the first to use Darwinian analogies and biology's powerful population thinking to approach the problem of how to wire up a brain, Gerald Edelman came to it from having tackled the analogous self-organization problems in the immune system; in the last dozen years, he has studied the counterpart selectivity theories for the nervous system in greater depth than anyone else, and so this book has been eagerly awaited.
Neural Darwinism is a fine example of Edelman's broad, bottom-up approach to how nervous systems get themselves organized, store information, and create new behavioral patterns. It is in the tradition of Donald Hebb's 1949 Organization of Behavior, asking "What is the nature of categorization, generalization, and memory, and how does their interaction mediate the continually changing relationships between experience and novelty?" (p. 241). It should be read by neurobiologists, developmental biologists, the cognitive cognoscenti--including the artificial intelligentsia--and by and those hopeful technologists who are flocking to the banner of neural-like networks as an alternative way of shaping up smart machines.
But became Neural Darwinism is so ambitious (a "global brain theory") in its attempt to model neural ontogeny, it is an easy book to misjudge--and an even easier book to lay aside and never finish. It is marred by mindnumbing sentences such as "As a result, combinations of those particular groups whose activities are correlated with various signals arising from adaptive behavior are selected" (p. 5) and by nonbiological terminology such as "re-entrant" and "degeneracy." Edelman seldom unbends enough to use the tutorial approximations "feedback" and "different ways of doing the same thing." Incredibly, there is no glossary.
And the introduction omits the very items that could motivate readers to endure the theoretical presentation. Several decades ago, biologists began to realize that there was a lot of cell death going on in developing nervous systems, and theorists began to suggest that carving away cells might create functional patterns corresponding to long-term memory storage. Richard Dawkins in a 1971 Nature paper (which Edelman omits) made this explicit, though J. Z. Young's 1965 model for octopus memory is closer to the modern mainstream in selectively eliminating some synapses rather than eliminating entire cells.
Whatever the synapse turnover rate is (and no one even has estimates), there is an imbalance in the rates of making and breaking synapses during childhood. It causes us to reach adolescence with little more than half the number of cortical synapses that we had eight months after birth. What principles control the editing? Here, surely, is "neural Darwinism" in action. Since Edelman's models seem particularly relevant to the postnatal tuning-up process, it is even more surprising that this conceptually important background, from research on both humans and monkeys (for example, Science 232, 232 ), is omitted. Edelman treats "wiring up" as preceding "tuning up," but such data suggest overlap throughout childhood.
Because biologists are often impatient with even lucid theoretical discussions, some will unfortunately skim until reaching the unusually attractive specific examples-- which, in order to achieve their clarity, lack the richness of Edelman's more general theory. Most attractive of all is the single foldout color plate: this computer display reminds me of the back side of a colorful tapestry, little threads running here and there, as if they were axons in a tangential section of brain; their colors denote synaptic strengths between "cortical neurons." In the first frame, thanks to the randomized initial conditions, the picture is so haphazard as to suggest that Jackson Pollack himself had finally designed a true tabula rasa.
As the neural-like network gains experience (the sensory surface is stimulated, one point at a time and each point connects to many "cortical neurons"), one starts to see (in the second time frame of the color plate) red patches of strongly connected cells emerging from an increasingly blue boundary area where cells are weakly interconnected. Groups emerge, the physiological boundaries becoming far sharper than the underlying smear of anatomical connections -- and all without instruction. Once you comprehend it, you may feel that this one color plate is worth the price of the book.
Eventually, in this map of a model hand, each patch will correspond to a top or bottom surface of a finger, looking not unlike the detailed maps of somatosensory cortex in monkeys, the plasticity of which has been studied by Michael Merzenich and colleagues. More impressively, Edelman and co-workers Leif Finkel and John Pearson can mimic the cortical rearrangements that occur when a finger is amputated (or overstimulated), though I note a revealing exception. Real cortical maps globally rearrange themselves, including boundaries between more distant digits--but the model's map shows only a local effect on the boundary between the affected digit and its immediate neighbor. It is as if, were an enlarging California to expand north into mid-Oregon, the Oregon-Washington and Canadian borders remained fixed (rather than also distorting, as real cortical boundaries tend to do).
Not everything that involves random initial conditions and selective survival deserves to be called Darwinism. The dance evolutionary biologists call the "Darwinian two-step," randomness-then-selection continuing back and forth for many rounds to increasingly shape up nonrandom-looking results, usually cannot be seen in Edelman's examples of neural Darwinism; these repeated injections of randomness lie at the heart of what some would consider as delimiting 'Darwinism from simpler forms of self-organization such as clumping and zero-sum "Matthewism."
And while the group selection of the subside may involve both groups and selection, Edelman's examples tend not to demonstrate what usually comes to mind when the two words are used together following an invocation of Charles Darwin--who invented group selection as a striking exception (much debated by contemporary sociobiologists) to the usual rule that natural selection acts only on individuals and their progeny: It must not be forgotten that . . . a high standard of morality gives but a slight or no advantage to each individual man and his children over the other men of the same tribe.... [But a tribe whose members] were always ready to aid one another, and to sacrifice themselves for the common good, would be victorious over most other tribes, and this would be natural selection. At all times throughout the world tribes have supplanted other tribes; and as morality is one important element in their success, the standard of morality [will rise by natural selection] [The Descent of Man].If one looks carefully, Edelman's more general theory usually encompasses both the Darwinian two-step and this type of group-qua-group selection--but both tend to be missing from the examples on which biologists will likely focus their attention.
It is very easy to pick out an unphysiological assumption or unfulfilled prediction (my grumble is about assuming shunting rather than subtractive inhibition in cerebral cortex) in a work with the breadth and depth of Neural Darwinism. On such a pretext, many readers will rationalize laying this admittedly difficult book aside, unfinished. Yet those who persevere may come away feeling, "The brain really could work that way," not only because Edelman's assumptions are usually close-to-physiological but because he frames the issues in ways that have been repeatedly successful, in stochastic and Darwinian contexts, in revealing emergent properties. If you are concerned with the questions Edelman addresses, this book may well be worth your time.