Our three pound engine

Paul Churchland argues that as the secrets of the brain are uncovered by neural science, its many ills will become as easy to treat as the ills of the body.

The past five years have seen an unprecedented flowering of popular and semi-popular books about the brain - not just about the mind, a perennial subject of attention, but about the brain in particular. Simon LeVay's The Sexual Brain explores the roots of human sexual orientation as grounded in a tiny hypothalamic nucleus whose cell population is large in most males, small in most females, and - it turns out - similarly small in homosexual males. Francis Crick's The Astonishing Hypothesis provides a Cook's tour of the brain's anatomy with the aim of approaching the central mystery of the nature of human consciousness. Antonio Damasio's Descartes' Error details how the brain's rich connections with its primitive bodily viscera are essential to normal social, emotional, and deliberative consciousness; and he explores the striking social pathologies that result from the disease-induced loss of normal brain-viscera communication.

While these are some of the best of the recent "brain" books, they are but three of many. This new trend represents a turning point, I think, in the nature of human self-consciousness.

The brain has led an invisible and largely anonymous existence for most of human history. Despite its pivotal role in every aspect of human life, it might as well have existed on the dark side of the moon, instead of inside the head, for all that its possessors knew or cared about its internal dynamics, its computational capacities, its unfolding development, or its illnesses and occasional dysfunctions. Hidden behind a quarter-inch of tough bone, and opaque to our understanding even in those rare instances where it finally was made visible, the brain has been a mystery of a depth matched only by the public's indifference to it. Our emotional, intellectual, and social commerce may be as vigorous as you please, but it is typically conducted in a vocabulary that makes no reference to the underlying neural reality. At a theoretical level, of course, people know that they have brains. And they know that brains have much to do with thinking. But at the level of their day-to-day practical consciousness, the brain might as well be, as Aristotle thought it was, a device for cooling the blood.

All this is about to change. Slowly at first, but with compounding momentum, an accessible conception of brain function will spread through both the language and the practice of human social commerce. We will all become more explicitly conscious of the chemical, neuronal, and computational details of each other's ongoing brain activities, and we will respond to and manipulate those activities accordingly. They will eventually become a seamless part of our Lebenswelt, as familiar as our fingers.

This expansion of our neural consciousness will be driven by two distinct forces. The first is a push from within the several sciences of the brain. Here there is genuine cause to marvel. The last decade has brought dramatic improvements in experimental technology, especially in the form of non-invasive brain-scanning machines that allow us to discern - in live, awake human subjects - brain anatomy and physiological activity to a spatial resolution of a millimetre and to a temporal resolution of fractions of a second. What was hidden before is hidden no longer.

Or, more accurately, not entirely hidden: the significance of specific brain areas and their fleeting neural activities is certainly not written on their sleeves, even when they are rendered visible in the laboratory. But once opened to the extended gaze of science, localised brain damage can be correlated with evident deficits in the patient's behaviour, and focal neural activity can be correlated with familiar cognitive and emotional activity. The result is a map of localised cognitive function and computational interaction, a map slightly different for each individual, a map unprecedented in its informed resolution.

Experimental results at the chemical level are no less important, and they reveal further correlations with the cognitive phenomena picked up in brain scans. Communication between neurones is always modulated by a proper mix of site-specific neurochemicals. Disruption of that mix, by disease or by drugs, can corrupt the emotional, behavioural, or cognitive profile of the affected individual. Equally, restoration of that neurochemical mix, by compensatory drugs, can often reconstitute the victim's original profile. This is quite real. Our best handles on schizophrenia, mania, major depression, and obsessive-compulsive disorder, for example, are all purely pharmacological in nature: chlorpromazine for the first, lithium salts for the second, and fluoxetine for the last two. These are all imperfect, especially the first two, but their benign effects are dramatic, and no other form of therapy seems to touch these crippling disorders.

Experimental and clinical advances aside, cognitive neuroscience is flushed with a new wave of theoretical activity, most of it centred on the computational properties of artificial neural networks. Living neural networks are difficult to access and observe. They have millions of elements, are too microscopic for complete knowledge, and are too fragile and inaccessible for manipulation of all the relevant factors. An artificial neural network, by contrast, can be made manageably modest in its numbers, can be specified in exhaustive detail, and can be manipulated down to the least hiccup of its smallest synapse. Most important, these artificial networks have given us a managable conception of how their microbiological analogues really work.

Computer models of these artificial neural networks have displayed some arresting behaviours. Various networks have learned to discriminate male from female faces, to re-identify individual faces across different photographs, to read printed text aloud, to show genuine stereoptic or 3-D vision, to recognise voiced phonemes and to distinguish grammatical from ungrammatical sentences. These achievements are certainly not beyond the reach of classical digital computing techniques, but researchers are impressed by the extraordinary speed with which the new neural-network computing architectures execute their acquired skills, the relative modesty of the neuronal resources they require, their functional perseverance in the face of scattered neuronal damage or degraded data, and the verisimilitude of the perceptual, cognitive, and behavioural skills they display. Their cognitive performance is seldom perfect, but even their failures resemble the failures of biological creatures.

All of this activity is based on a style of information coding and a style of computation quite different from the styles deployed in conventional digital computers. The brain represents a given aspect of the world with a signature pattern of neuronal excitations across a large population of proprietary neurones. (Your TV screen uses the same trick: any picture is just a pattern of brightness levels across its 250,000 pixels). And the brain "computes'' by transforming such patterns of excitation into new patterns as they are filtered through the trillions of synaptic connections by which one population of neurones communicates with another. It would appear that we finally have a revealing theoretical grasp of how the brain's microstructure sustains cognition.

So there is much in the way of news for the neural sciences to tell us. The current push, however, encounters an even more powerful pull. For assorted reasons, the general public has acquired a fascination for things neural, robotic, computational, or mind altering. It is evidently weary of a century of mostly impotent "talk'' therapies, yet needful still of a wisdom equal to the scourge of psycho and sociopathology, and to the intricate demands of even a normal social existence. The public is eager to explore a brain-based approach to social, moral, and practical cognition, if only for a change of pace.

In prospect, at least, the payoffs appear far-reaching and welcome. We feel them already in the medical and psychiatric domains. Using the new noninvasive scanning and computer-imaging techniques, a neurologist can now take a visual tour of a patient's brain throughout its entire volume without ever lifting a scalpel. He can locate and identify the offending tumors or lesions, and plan some carefully focused surgery or other treatment accordingly.

Equally, psychiatrists can now diagnose a profile of major depression, prescribe a "serotonin agonist'' that will renormalise the patient's neurochemical balance, and thereby lift the pall of darkness that had fallen over his every thought and deed. The redemptive drug is not an "upper"; it produces no "high'' in normal people. But, like insulin treatments for diabetics, it does redress a broadly metabolic deficit. It gives the afflicted individual the same chance at happiness had by the rest of us.

These examples portend a multitude. Once the brain becomes as familiar to us as the rest of the body, and once its many ills become as easy to diagnose and treat as the ills of the body, we can be sure that the public will insist on these benefits, and that the institutions of modern civilisation will attempt to provide them. The policies and techniques of public education, for example, are sure to exploit the new insights, most obviously in the remedial and corrective domains, but also for the baseline instruction of normals. In the past, educational philosophies have proved very sensitive - occasionally, too sensitive - to changing fashions in cognitive theory. We can expect greater and, let us hope, more effective shifts in the future.

Social and moral development in children will become more transparent under this same illumination, and we can expect to identify and ameliorate subclinical sociopathologies earlier and more effectively than at present. Such insight will also serve us better in the law. At the very least, it will permit better informed application of our existing rules concerning responsibility, sentencing, and corrective incarceration; and it may motivate more insightful and more effective practices altogether.

Industrial uses of neural technology are certain to proliferate; indeed, the process has already begun. The capacity of trained neural networks to analyse a complex situation, almost instantaneously, so as to recognise subtle global features buried in transient noise, makes them invaluable for such things as production quality control, mineral prospecting, weather forecasting, general medical diagnosis, and financial analysis. These examples are real, and they already include some social analysis.

The Chicago police department recently trained an artificial neural network (on the information contained in the thousands of administrative files of Chicago's finest) to pick out, from the files of young officers, those most likely to encounter chronic disciplinary hearings later in their careers. In a test run on the randomly chosen files of some long-term officers, more than half of the small group identified by the network as "potential bad cop'' turned out to have been independently identified, as already problematic, by their human administrators. The city's police union, it is said, was not amused by this new form of scrutiny. The rest of us may feel differently.

As the function of the human brain, and of artificial brain-like networks, come finally to be understood by unfolding research, that knowledge will inexorably make its way into the public consciousness. This is a good thing, and for two reasons beyond those already mentioned. The first is that the technologies referred to, like any others, contain the potential for occasional abuse, both accidental and deliberate. The best defence against such threats is, as always, a well-informed and perceptive public.

The second reason for applauding such publication concerns individual development within the personal and spiritual domains. The development of character, the realisation of our dreams, the acquisition of wisdom, and a fruitful commerce with the other brains near and dear to one - all of these goals will be better served by a deeper understanding of the three-pound engine that pursues them.

Paul M. Churchland is professor of philosophy at the University of California, San Diego. His book The Engine of Reason, the Seat of the Soul, will be published by the MIT Press in June.