By George Johnson
George Johnson is an editor of The Week in Review of The New York Times and the author of "In the Palaces of Memory."

BRIGHT AIR, BRILLIANT FIRE
On the Matter of the Mind.
By Gerald M. Edelman.
Illustrated. 280 pp. New York:
Basic Books. $25.

ACCORDING to one of the weirder interpretations of quantum theory, electrons and the other subatomic particles that make up creation don't really come into existence -- taking on definite positions in time and space -- until they are beheld by a conscious observer. Extending this notion to a cosmic scale, the most radical proponents of what has come to be called the anthropic cosmological principle argue for a dizzying symbiosis in which the universe gives rise to conscious beings who in turn give rise to the universe through the act of observation.

It probably shouldn't be surprising that creatures who spend most of their time indoors thinking would convince themselves that mind is as fundamental as the Big Bang. But what then are we to make of the neuroscientists, who have worked for a good part of the century to show that mind -- consciousness included -- is very much an earthbound thing, arising from the chemistry inside our heads? While one part of science is trying to reduce matter to mind, another is trying to reduce mind to matter. Something obviously is going to have to give.

In "Bright Air, Brilliant Fire: On the Matter of the Mind," Gerald M. Edelman sides with the biologists who consider it a little parochial to suppose that the existence of the universe depends somehow on brain cells -- tiny sacs of seawater that developed the ability to strike sparks of electricity when predators approach. Like most neuroscientists, he sees the brain as a kind of Rube Goldberg contraption, a grab bag of evolutionary tricks that accumulated over eons to help insure survival on this spinning cinder.

But that is about all that Mr. Edelman and his colleagues seem to agree on. Whether they contemplate the behavior of brain cells or the behavior of women and men, most researchers in the cognitive sciences believe that what evolution has left in our heads is a very powerful if somewhat haphazard computer.

To Mr. Edelman that is just replacing one mystery with another: if mind is computation, then what is computation? Or more generally, what is mathematics? Mathematics arises from minds and minds arise from mathematics and we're stuck inside another of those tautological loops that drove Wittgenstein to take up carpentry.

The notion that the brain is a kind of computer is an error of such magnitude, Mr. Edelman believes, that cognitive science is on the brink of a crisis. "I claim," he writes, "that the entire structure on which the cognitivist enterprise is based is incoherent and not borne out by the facts."

Mr. Edelman's alternative to the computer model of the brain is something he calls neural Darwinism. But despite its catchy name, the theory has not had the influence he clearly thinks it should. The problem, it seems, is that hardly anyone but Gerald Edelman understands it.

Like the most treacherous peaks in the Himalayas, neural Darwinism is known primarily for those souls who have fallen trying to scale its heights. Opening one of Mr. Edelman's earlier books, "Neural Darwinism: The Theory of Neuronal Group Selection," a reader comes face to face with a sheer granite cliff with barely a handhold in sight. In 1988, Gunther Stent, one of the pioneers of modern biology, expressed his exasperation at trying to penetrate the theory. "I consider myself not too dumb," he said. "I am a professor of molecular biology and chairman of the neurobiology section of the National Academy of Sciences, so I should understand it. But I don't."

Francis Crick also found the parallels between the workings of the brain and the workings of Darwinian evolution rather elusive. He concluded that the theory should be called neural Edelmanism.

It hasn't helped matters that Mr. Edelman states flat-out that the brain does not work by processing information. Many neuroscientists find this assertion baffling. But it's by considering seemingly absurd notions (the bending of time and space, the acausal nature of the subatomic world) that we dig out of our conceptual caverns and see new light.

Mr. Edelman won a Nobel Prize in 1972 for establishing that the immune system works according to Darwinian principles. So why not the brain? After all, immunity is a kind of memory -- invaded by a virus the body remembers, so that next time it can more ferociously fight back. For that matter, a population of furry mammals can be said to carry in its genes a memory of the ice ages. How elegant it would be if, as Mr. Edelman believes, all three kinds of memory were related.

It's exciting then to be presented with "Bright Air, Brilliant Fire," which promises to explain not only the ideas in "Neural Darwinism," but also those in Mr. Edelman's books "Topobiology," which speculates on how embryos unfold to become people, and "The Remembered Present," which lays out a theory of consciousness. His previous books were written for scientists, who continue to argue the merits of his ideas. This book is for general readers, so it must be judged foremost by how well it manages to get these difficult ideas across.

In the first third of the book, Mr. Edelman guides us through his ideas about embryology and immunology in a leisurely, literate way. When a bacterium or virus enters the body, he explains, it is recognized as a foreigner because of the unfamiliar molecules, called antigens, on its surface. The immune system responds by churning out lymphocytes -- cells that are studded with antibodies, complementary molecules tailored to match the bumps and crevices of the antigens. Like Lego blocks, antibody and antigen snap together, setting off a cascade of cellular events that leads to the invader's destruction.

Before Mr. Edelman came along it was widely believed that there was a single kind of lymphocyte whose all-purpose antibody molecules had the chameleon-like ability to mold themselves to invading antigens. Mr. Edelman showed instead that the body randomly generates millions of different kinds of lymphocytes, each bearing antibodies as specialized as the various species that occupy the niches of an ecosystem. Confronted with an invader, the lymphocytes compete for the honor of finishing it off. Those whose antibodies happen to provide the closest fit latch onto the enemy's antigens. Thus selected, these winning lymphocytes become fruitful and multiply, proliferating into an arsenal against the infection. Some of the newly minted lymphocytes linger in the system. It is in this sense that there is a memory of the invasion: if the enemy is encountered again, the body is prepared for a swifter counterattack.

The same principles hold in neural Darwinism. But here, instead of a population of different lymphocytes, there is a population of what Mr. Edelman calls neuronal groups -- brain cell clusters randomly wired together, each in a different way. Like the immune system, Mr. Edelman believes, the brain uses a kind of Darwinian selection to mold itself to the contours of the outside world.

In this case the invader is an unfamiliar stimulus entering through the senses. Many groups of neurons respond, but some happen to be configured in a way that makes them respond more strongly than the others; they fit the stimulus better, just as some antibodies fit the antigens more closely. This resonance between signal and circuit sets off a biochemical reaction. But instead of causing the selected group of brain cells to multiply (as with the lymphocytes) the links between the neurons -- the synapses -- are strengthened. In the future, this group of neurons will react to the stimulus more strongly.

In the cerebral jungles, a group of neurons can also recruit neurons from neighboring tribes. Alliances of neurons are constantly forming and shifting boundaries, providing an ever-changing population of neuronal groups available for selection.

Not all of this is new. It is generally accepted that memories are made when information from the senses sets off biochemical reactions that strengthen the connections between neurons, fine-tuning the circuitry we use to interpret the world. The difference with Mr. Edelman's theory is that, as he sees it, no information actually passes from the environment to the brain, any more than information from the environment can be said to pass directly into an animal's genome; the environment simply selects from a random population those creatures best suited to survive. That, I think, is what he means when he says the brain is not an information processor.

This fairly subtle distinction becomes even subtler when Mr. Edelman describes how he has simulated his theory on a supercomputer. A program he calls Darwin can learn to categorize, sorting striped objects from bumpy objects, for example. But though neural Darwinism can be simulated with a computer, Mr. Edelman insists that what is being simulated is not computation. I suppose that's a defensible position. Computers can simulate the traffic patterns of the Los Angeles freeways, but that doesn't mean that Los Angeles is a computer. Still, for all Mr. Edelman's assertions, I was struck again and again by how much neural Darwinism sounds like computing.

For example, he speculates that the brain is divided into an array of ever-changing "maps," each dedicated to a different task -- detecting motion, sorting out shapes and colors. By sending signals back and forth, the maps give rise to thinking and consciousness. But it's not clear from the book how Mr. Edelman's maps differ from the "agents" in Marvin Minsky's "society of mind" or the "stupid homunculi" in Daniel C. Dennett's "Joycean machine" -- two theories that unabashedly describe the brain as a kind of computer.

Mr. Edelman insists that it is wrong to believe that "individual neurons carry information, just as some electronic devices carry information." Yet when he describes the brain he uses words that imply that the brain is something like a computer. He speaks of "the brain's output," parts of the brain that "receive input" or "spatial signals." Early on he writes, "A piece of brain tissue is an intricate network that responds to electrical and chemical signals in three-dimensional space and in time." He concedes that "What I have described so far may sound like the organization of a vast telephone exchange or perhaps even that of a digital computer."

WHY then does he think his poor colleagues are so benighted? A clue comes when he speaks of his theory as "a crisis for those who believe that the nervous system is precise and 'hard-wired' like a computer." There's the problem. Nobody really believes that.

Though one would hardly know it from reading "Bright Air, Brilliant Fire," dozens of neuroscientists, psychologists and computer scientists have developed simulations in which randomly wired networks of artificial neurons learn to generalize, categorize and perform other computations. These designers consider it axiomatic that much of the brain is randomly connected. Proceeding from there, they use statistical thermodynamics and other mathematical tools to show how orderly behavior can arise from randomness.

Even in a digital computer, randomness on one level gives rise to order on another level. Under a microscope, the circuitry of a computer chip looks precise and geometrical, like a city viewed from the air. But the electrons that bumble their way through the streets and avenues each behave unpredictably, according to the vagaries of quantum mechanics. It is only by harnessing electrons en masse that we can use them as juice for computation.

Mr. Edelman argues that the brain is not a computer because it is analog, not digital. But that is like saying that an LP record is inherently more musical than a compact disk because it records sound in the form of squiggles instead of binary code. Information processing is information processing, whether you use the 1's and 0's of digital logic or fluctuating voltages in neurons.

He also contends that there is no such thing as software in the brain, that it is the morphology -- what some would call the hardware -- that counts. But the distinction between software and hardware is not nearly so clear-cut. Some of the programs that animate the computer I'm using to write this review consist of 1's and 0's permanently burned into little slivers of silicon called ROM chips -- the software is indistinguishable from hardware. Or think of the old vacuum tube computers that were programmed for each new task by plugging and unplugging cables in an old-fashioned switchboard. The software consisted of patterns of wires. When cognitive scientists talk about software in the brain, they are referring to patterns of neurons connected to perform various tasks.

Again, Mr. Edelman himself falls back on the obvious comparisons. In describing how a brain cruising along on automatic pilot might interrupt its normal behavior to focus on a potentially dangerous event, he writes about " 'alarm' signals" dispatched to systems in the midbrain that "send back signals" that "block the execution of a motor program."

Sometimes it seems that Mr. Edelman is trying to use the book to goad his rivals, to draw attention to his theory by making it sound stranger than it is. I think the ink could have been better spent on exposition. In many ways, neural Darwinism is a tantalizing notion. "Bright Air, Brilliant Fire" makes it clearer than it was before, but not a whole lot clearer.

GRAPHIC: Drawing: Exposed surface of the cerebral cortex, drawn by the 16th-century anatomist Andreas Vesalius, often called the founder of modern anatomy. (From "Bright Air, Brilliant Fire")

LANGUAGE: ENGLISH