WHY NEUROECONOMICS?

In an uncertain world, people and other animals make their living by predicting which of alternative courses of action is likely to yield the best return. For humans the return might take many forms, such as material, financial, social, or esthetic, but the underlying currency involved for any species is “inclusive fitness,” the rate at which an animal's genes are propagated. Professor Glimcher demonstrates that Economics methods are applicable to decision-making under conditions of uncertainty, both at the behavioral and the neuronal level. This approach has been called neuro-economics, although “econometrics” characterizes it more precisely. Econometrics is the application of statistical and mathematical methods in the field of economics to test and quantify economic theories and the solution to economic problems. Specifically, individuals' decision-making benefits from knowing how likely a response is to be reinforced, and knowing the reinforcement's value. Even single neurons are sensitive to these variables. Glimcher reaches beyond the heavily studied neural substrate for sensation and response to predictive neural circuitry that factors in the prior probability of reward, and its expected value. Indeed, he and his colleagues have identified neurons in monkey's inferior parietal lobule whose firing rates reflect both probability and value.

Glimcher argues that all behavior is indeterminate, the outcome of decisions based on a reckoning of probabilities, whether conscious or not. Determinate machine-like reflex responses such as are elicited from the isolated spinal cord are not in the repertoire of intact individuals. Formally, a system is considered determinate if, when one isolates it and repeatedly imposes identical starting conditions, its behavior is the same every time. Machinery is the obvious instance, but even ostensibly random outcomes such as occur during coin-flipping have been found to be determinate if the starting conditions are precisely identical. In intact biological systems starting conditions can probably never be held completely constant. So whether any behaviors are strictly speaking determinate is moot anyway. But this is not critical to Glimcher's argument that what animals do is not machine-like, but based on probabilities. Adapting the words of Stephen J. Gould, if one could rewind the tape of a life and let events play out again, the results would almost certainly differ dramatically.

SECOND GUESSING THE PAST

Blame the usual suspect, René Descartes, for the reductive approach. He grasped the machine-like nature of human behavior, but avoided attributing all of human existence to gross matter by regarding the will as a product of the soul, the brain conveying its instructions to the body. Did he really believe in this dualistic dichotomy, or did he present it to safeguard himself from charges of heresy? Either way, he has led neuroscientists up a garden path ever since. Cartesian dualism became a favorite means for kicking currently insoluble problems upstairs.

Such impoverished systems as the isolated spinal cord are used precisely because they react more predictably, being free of less controllable top-down influences. The result\.suggests a functional anatomy, but does not purport to represent functioning that occurs autonomously in the intact individual. In so far as findings from the isolated spinal cord have been extrapolated to parts of the brain not yet understood, this strategy succeeds if the system as a whole uses the same building blocks as the part studied. It fails if, as Glimcher believes, the very act of reduction has changed the manner in which the system works, throwing out the baby with the bathwater. Whereas Sherrington hedged his bets about extending his reflex model of behavior, no less an investigator than Pavlov believed that his findings on cortical reflexes are applicable to the more complex behaviors. Others treat behaviors that appear to be indeterminate as distinct from behaviors that characterize “man as machine,” attributing them to a qualitatively different non-biological source, such as the soul, or instruction direct from God. Glimcher believes that he has a better theory, and one that can be empirically validated. Having demolished reflexology at length in 10 entertaining as well as informative chapters on the history of neuroscience, he propagandizes theoretical advances in mathematics and evolutionary theory. In the final three chapters he presents illustrative experiments from his own laboratory both on the whole individual and the single neuron.

Critics as gifted and articulate as Hughlings Jackson, Henry Head, Paul Schilder, Kurt Goldstein, and Heinz Werner, protested that the intact brain functions along quite different lines from truncated parts of the nervous system. But they could not convincingly support their global theories empirically, because the technologies required to test their claims experimentally were not yet in place. Scientists are conservative. Kuhn remarked that a theory will not be rejected simply because it is inadequate, but only if it is confronted by a better one. So, although few neuroscientists would deny that the bottom-up reflex approach was ultimately inadequate, they continued to adhere to it for want of a testable better model. The necessary technologies are now available, but the theorizing has lagged behind. The new wine of technological advance has been poured into antiquated bottles of theory. Even Glimcher's neuroeconomic approach falls far short of what the holistic theorists would have considered an adequate account of behavior.

ECOLOGICAL VALIDITY OF PROBABILITY THEORY

Having introduced Von Neumann's game theory, Glimcher explains why it behooves animals to behave unpredictably. Animals make a living out of predicting what will happen next. What happens depends not only on physical circumstances, but also on the motives of other animals, predators and prey. Unpredictable behavior by prey frustrates the predator's anticipations. For humans, so intensely social, the biggest challenge is to predict what other humans will do. This entails resort to probability theory. Glimcher claims that probability theory can clarify behavior at all levels. Questions remain, however. It is not clear how probability theory clarifies simple or automatic behavior. Moreover, Glimcher's discussion of strategic behavior omits emotion as an intervening variable and smacks of the behaviorist. Glimcher does concede that mind is possible, but doubts its relevance: “mind, though it may very much exist, simply does not figure in that equation” (p. 343), he writes, referring to the neuroeconomic brain–behavior architecture. Other neuroeconomists do include emotions in their theoretical framework.

We can extend the analysis to the world outside the laboratory, where multiple opportunities exist for responses that are potentially rewarded. What happens when two responses offer equal probability of reward, of equal value, as they did to Buridan's hypothetical Ass? Jean Buridan, fourteenth century philosopher of mind and Rector of the University of Paris, devised this thought experiment. Or perhaps an adversary devised it, in order to make an ass out of Rector Buridan. I contributed the predator.

GAMES DONKEYS PLAY

A hungry ass finds himself equidistant to two piles of hay, one right and one left, which are equal in size as well as quality. Having no reason to prefer one pile to the other, the ass cannot decide which pile to eat first. Relentlessly rational, he stays in place and starves to death.

A predator lurks nearby. He wants to intercept the ass's path to dinner, and make dinner out of the ass. However, he does not know which path that will be. He seeks a clue on which to base a prediction.

What would an actual ass do, faced with an approach–approach conflict between heading right and heading left, while keeping his intention inscrutable? The response conflict is within an opponent processing system that consists of a right-turn vector represented in the left brain, and an equally activated left-turn vector in the right brain. However, the equilibrium between right and left turning tendencies is unstable.

The cerebrum houses 10¹¹ neurons, making 10¹⁴ synaptic connections, all firing all the time. Most of this activity is autonomous, and not in response to stimuli whether external or internal in origin. Such intense ongoing traffic in a totally connected network guarantees moment-to-moment fluctuations of the activation levels. If one representation gains a momentary edge, it will increase its inhibitory hold on the other, which consequently will diminish the other's inhibitory hold on it. The momentary disparity will suffice to swing intention to one of the two trajectories, in a rapidly enlarging activation imbalance, like a seesaw tipping to one side. Two adaptively advantageous outcomes result from this non-rational randomly generated difference-amplifying feedback. The animal gets to eat. Since the outcome is randomly determined, the predator has no basis for predicting it. I intend this illustration to suggest that animals do not need a random sequence generator in order to behave randomly. Randomness is inherent in the “noisy” neural network.\.

WHAT DOES THIS BOOK OFFER THE NEUROPSYCHOLOGIST?

It offers the company of an educated mind on a short trot through seldom encountered domains of knowledge. The reader is unlikely to find elsewhere as clear and readable an exposition of topical issues such as Nash Equilibrium or Prey Theory. The book is not a compendium of neuroeconomics, however, but intensely personal. Glimcher's references to neuropsychology are rudimentary. But the decision mechanisms that Glimcher studies at the behavioral and the neuronal level could be identified in patients with brain lesions, especially of prefrontal cortex. A significant literature on decision making under conditions of uncertainty, not reviewed in Neuroeconomics, implicates ventromedial prefrontal and left inferior parietal cortex, as well as lateral cerebellum. Whatever an animal can do must derive from the activity of specific brain circuitry, and therefore must be reflected in the activity of individual neurons within the circuit. Hence probabilities are reflected in the firing of the single neuron. Finding such neurons is fascinating, but not revolutionary.