You are here

Second Lecture. The Failure of Reductionism

Christopher Longuet-Higgins

It has fallen to me to open the bowling for the sciences, at the risk of being hit for six by Kenny, who is to be opening batsman for the humanities. Unfortunately, I shall not be able to do what might be expected of me by Lucas, which is to assert that everything except matter is immaterial—if he will excuse the pun. But I suppose I shall have to expound that view, if only to improve upon it, which is what I want to do in this lecture. In brief, I propose to examine the doctrine of scientific reductionism, to give reasons why I regard it as untenable, and to drop some hints as to what will actually have to be done by scientists, and is already being done, if we are to create a worthy science of the mind, rather than being condemned to mechanistic platitudes on the one hand, and philosophical exhortations on the other.

The reductionist position, as I see it, takes several forms, varying in sophistication. The confident young scientist described last time by Lucas, had obviously been impressed by the beauty and order of modern physics, and by its apparent ability to give a coherent account of all the phenomena—or nearly all—manifested by matter in motion. Objects as different as the sun and the moon, substances as different as chalk and cheese, influences as obscure as magnetism and X-rays, could all—or so it seemed—be brought within the compass of the physical sciences, and made comprehensible in terms of a small number of basic concepts and fundamental constants of nature. It was a heady experience, calculated to provoke a conversion to the view immortalised in Rutherford's famous dictum There is physics, and there is stamp collecting’. There were, of course, many matters of detail which had not been fully cleared up. One could state with great accuracy the laws describing the motion of a falling apple, but people hadn't yet explained exactly how a pip grew into a Cox's Pippin. Living organisms were certainly very different in their observable behaviour from crystals, atoms, and bar magnets. But as far as one could tell, the laws of conservation of mass and energy, and even the second law of thermodynamics, applied with equal force to living and non-living matter. The organic chemists had already synthesised many of the complex molecules which are formed in the tissues of plants and animals, and vitalism had been rudely expelled from polite scientific society. What better proof could there be of the overriding supremacy of the physical sciences and their ultimate ability to illuminate the whole universe of our experience?

So much for the Zeitgeist from which naive reductionism has sprung, at least in my own generation, though I suspect that the pendulum is now swinging rapidly in the opposite direction. But what is the intellectual content of this position, as opposed to its emotional symptoms?

The starting point of the reductionist argument is the supremacy of physics, the doctrine that any assertion which carries implications about the properties of matter must either conform with the laws of physics or be discarded. Let me illustrate this entirely reasonable view with one or two examples. First, suppose someone asserted that by taking thought he could determine his velocity of motion through space. Would we take him seriously? Of course not. Why? Because one thing which we know for certain now about space and time is that absolute motion is not only impossible to detect but impossible to define, if we take seriously, as we must, the experiments and the reasoning on which relativity is based. My second example is clairvoyance, the claim to be able to perceive future events. At the heart of physics is the causality restriction: if two events are in a relation of temporal precedence, then causal influences cannot be propagated from the later one to the earlier one. If anyone is genuinely clairvoyant, then modern physics is fundamentally in error. There are no two ways about it. Finally, to anticipate our discussions on free-will and determinism, does Heisenberg's principle of uncertainty represent a gap in physics, through which we can escape from the tyranny of physical determinism? No, says the physicist. The principle of indeterminacy is a principle of impotence. Unless quantum mechanics is all wrong, no-one can predict the way an excited electron will jump, or explain why it jumped the way it did, in terms of antecedent circumstances. So if an electron in my head jumps in a certain direction when I make a free choice, I cannot afterwards account for the way it jumped by referring to my choice, or in any other way either. It would, in any case, seem very odd to advance such an explanation for a physical event—a point to which we shall doubtless return in our later discussions.

Now, perhaps, we are in a better position to state the doctrine of scientific reductionism, and to see what important questions, if any, it leaves unanswered about the nature of mind. It is, in a sense, a natural extension of scientific materialism, if that label is taken to signify a rejection of any concept which cannot be directly related to the world of things and stuff. Reductionism makes a less sweeping claim. The laws of physics, which carry the ultimate authority about the material world, are taken to be irreducible. The reductionist recognises, however, that there are other sciences, such as chemistry, which have almost as high standards of generality and precision, and that these must find a place in the scientific scheme. But he notes with satisfaction that, after many centuries of illusory independence, chemistry has been fitted into the framework of physics, at least in principle. As Paul Dirac remarked in the first chapter of his book on quantum mechanics, the whole of chemistry and a large part of physics could in principle be explained by the new theoretical discoveries. So chemistry, the reductionist claims, is really physics, and, furthermore, biology is really chemistry. Witness the resounding success of molecular biology, which unlocked the secrets of reproduction and inheritance under the noses of the classical biologists. So perhaps neurophysiology is really molecular biology? Some scientists studying the brain certainly think so. It is seriously, though in my view implausibly, suggested that DNA, which embodies the lessons of our evolution, is also the answer to the problems of memory, which nobody could deny is central to mental activity. What about psychology? There are, of course, old-fashioned psychologists who concern themselves with how human beings learn to speak or to adjust themselves to other people, but psychology is really neurophysiology; the most up-to-date psychologists need electrodes to stick into nerves and brains, knives for cutting bits out, and complicated electronic recording equipment for processing their measurements—though what the process of ‘processing’ is intended to achieve is not always made quite clear. And so on. Sociology is really psychology, economics is really sociology, history is really economics, and there the trail becomes indistinct.

I have been deliberately satirical in so describing the scientific reductionist, because his position only becomes clear when carried to its logical conclusion. But where is the fallacy? Because if we cannot find one, we must follow the argument where it seems to lead, into an intellectual wilderness populated by mad scientists trying to measure the positions and velocities of all the molecules within reach. Perhaps the best place to pick holes would be at a rather low level, with the assertion that chemistry is really physics, and see what this assertion amounts to. Roughly speaking, chemistry is concerned with the properties of matter under rather special conditions, such as those which prevail on earth. Under these conditions we can recognise distinct chemical substances, and the business of the chemist is to reveal the internal structure of these substances, and to account for their properties, including the ways in which they react with one another. The identity of a substance depends on the manner in which its constituent atoms are joined together. Usually, but not always, the atoms are bound together into identical clusters called molecules, so that the study of molecules is the major part of chemistry. But molecules come in all shapes and sizes, and must be brought to order before the chemist can state any significant generalisations about their behaviour. It is important to realise that no-one except a chemist is competent to do this; certainly not a physicist, if only because physicists are so contemptuous of chemical distinctions—of ‘stamp-collecting’ in Rutherford's words. Let me take an actual example. In recent years there has been much interest among chemists in what are now called electrocyclic reactions. I imagine that only a handful of people in this room will have the slightest idea what electrocyclic reactions are, and I have no intention of trespassing on the ignorance of those who do not. Along comes a physicist, expert in quantum mechanics, which, he maintains, explains all chemistry in principle, and we ask him for an explanation of electrocyclic reactions. Does he offer one? No. His first words are ‘What are electro-cyclic reactions?’. In order to answer his question we shall have to introduce him to chemical concepts which are not part of his intellectual armoury, and even then he may not understand why we asked the question, Actually to answer it he has to become, for the moment at least, a chemist. Only by so doing can he see what principles of physics may be relevant to its answer. Insofar as physics is what physicists do when they are left to get on with their own work, chemistry is not part of physics in any important sense. It has its own concepts and its own problems, the concepts being those which are relevant to the problems. This, of course, is not to deny that physical principles can be brought to bear on chemical phenomena, but the questions must be asked at the higher level before they can be examined at the lower.

But this is not the whole story; if it were, my objections to the reductionist position might seem tiresomely pedantic to, let us say, a molecular biologist engaged in the crystallographic study of virus particles. There is an even more cogent objection to the view that all higher level concepts must derive in the last resort from concepts in physics. Let me give a historical example. After Newton and Laplace it seemed that the secret of the universe was to be found in the laws of mechanics. These laws were thought to be deterministic in the sense that, given the states of motion of all material particles at one time, they would prescribe the states of motion of all the particles at any later time. But in the nineteenth century, before there had been any hint of indeterminacy in mechanics, a quite independent set of physical concepts emerged. Certain phenomena, such as the passage of heat from a hot body to a cold one, proved impossible to describe in purely mechanical terms. They, called for the new concepts of temperature and entropy which, like the idea of disorder, only make sense when applied to physical systems in which the states of motion of the constituent particles are largely unknown. Statistical Mechanics, as the new subject was called, could obviously not be founded on mechanical principles alone, because heat and temperature were not mechanical but statistical concepts. There was nothing in mechanics itself to stop hot bodies from getting hotter, or cold bodies from growing colder. This in spite of the unquestioned authority of mechanics, when applied to systems in which the initial state was specified in every detail. So if, even within physics, one set of concepts are not reducible to another, what grounds can there be for asserting that the concepts and laws of any other science must be reducible to those of physics?

It will, perhaps, be as well if I try and summarise this critique of scientific reductionism before launching into our main topic, the nature of mind. It's all very well to say that one science rests upon another, if all we mean is that the laws of the former do not actually conflict with those of the latter. But this demand does not entail that the concepts of the higher science can necessarily be explicated in terms of the concepts of the lower science. Nor does it even imply that the laws of the higher science follow from those of the lower: this is clear from the universally conceded fact that thermodynamics does not follow logically from dynamics. Dynamics is quite indifferent to the arrow of time, and thermodynamics emphatically is not. Most thoughtful biologists implicitly accept this thesis. In his recent admirable book, Le hazard et la necessité, Jacques Monod is at pains to lay bare the distinguishing features of living systems, or perhaps one should say, of life, in relation to the non-living world. He reinterprets the Darwinian thesis in modern terms, and shows how at every level, from the molecular to the visible, the evolution of life can be seen as the selection of what he calls ‘teleonomic’ variations, and their preservation by the fidelity of the hereditary process. This, if I may call it the First Law of Biology, is in style and conception totally unlike a law of physics, or even of chemistry, as I am sure Waddington would agree. But there are other scientists who remain unconvinced.

I recently attended a series of scientific meetings near Paris, at which a number of distinguished physicists and biologists were discussing the implications of theoretical physics for biology. It was a strange, indeed an unnerving, experience. In the politest possible way, some of the theoretical physicists seemed to be suggesting that if only the biologists would open their minds to certain theoretical possibilities which had been revealed by quantum mechanics, many of the more puzzling properties of living cells, and even of the brain, might become clear to them. The biologists, almost to a man, reacted to this advice with less than gratitude, and it was easy to see why. Indeed, at one point in the proceedings I could not help recalling the story of an Edwardian lady who found herself seated next to a stranger at dinner, and asked him about his work. ‘Madam’, he replied, ‘I am a student of physics’. ‘Oh really’, she said, ‘my husband always says that anyone with a classical education could get up physics in a fortnight’. But I must not give the impression that the scientific battle of the mind is being waged in the field of theoretical physics. To suggest that it was would be to incur the justified indignation of those scientists who actually study animals and their brains and their overt behaviour. An architect might well take offence if a town planner exchanged ideas over his head with a bricklayer, even though towns are ultimately built of bricks. So let us leave physics to the physicists, and see what the psychologists and neurophysiologists might be able to tell us about the nature of mind.

Psychology is still a young subject, struggling to be recognised as a science. I am not competent to review its history, but I suspect that it might make a rewarding Ph.D study in the sociology of science. The enormous success of the physical sciences in the nineteenth and early twentieth centuries called for some explanation, and generated a theory of science which laid great emphasis on quantitative measurement. Not surprisingly, this theory suited physics and chemistry very well, having been inspired by their example, but on the biological sciences its influence was perhaps a mixed blessing. It put pressure on the biologist to make his observations quantitative where possible, or at least to submit them to statistical analysis, but, as a consequence, it undervalued descriptive or taxonomic observations, and concepts which could not be worked into a quantitative mathematical theory. Little wonder, then, that psychology should suffer a severe attack of cold feet, and disown one of its great men, Sigmund Freud, in favour of white-coated experimenters on the learning abilities of rats and pigeons. Not that Freud was above criticism for his lack of statistical rigour; but how could anyone conduct an honest statistical analysis of material as complex and diverse as human dreams, for example? Nor would I suggest that nothing of value can be learned from carefully controlled experiments on animal and human behaviour; but what sort of information do such experiments yield, and can it possibly be used to illuminate our understanding of thought, as opposed to physical activity?

The behaviourist approach to psychology, as I interpret it, is founded on the determination to do away with human testimony as essentially unreliable and incapable of quantification. People's reaction times and pulse rates can be objectively measured, but their opinions and their interpretations of their experiences, though they may be noted by the experimenter, do not count as scientific evidence. In effect, the behaviourist treats the subject as a black box, which he subjects to measurable stimuli of various kinds, and which emits measurable responses. He attempts to establish a functional relationship between the two, in whatever terms suggest themselves to him. Needless to say, this is very difficult, and there is a quite irresistible temptation to cheat. One way of cheating is to ask the subject why he responded as he did, but this would really give the game away. To prevent this kind of cheating, it is therefore recommended that all behavioural experiments should be conducted on dumb animals, but this does not prevent the experimenter from trying to interpret their behaviour by analogy with his own. The other way of cheating is, of course, to prise open the box and look inside. There must be some causal connection, surely, between the input stimulus and the output response; let us see if we can't trace the nervous pathways which lead from one to the other. At this point, of course, the behaviourist has turned into a neurophysiologist, so let us see what the professional neurophysiologist has to tell us.

To the neurophysiologist, the word ‘mind’ is even more suspect than to the experimental psychologist. His intellectual orientation is set by the word ‘neuron’, and anything which isn't neuronal isn't neurophysiology. The mind, whatever it is, certainly isn't composed of neurons, but the brain is. So let us put behind us the outworn concept of mind, and try to understand how the brain works. It may take many years, or even decades, but patience and experimental skill will eventually be rewarded. We already know quite a bit about axons, dendrites, synapses, and what have you; ultimately we may hope to have a more or less complete map of the central nervous sytem and we shall be home. This little caricature of the neurophysiologist is, I admit, quite unfair to those devoted scientists who have told us what we know about the structure of the most complex system in the universe, and it does very much less than justice to those who feel that we have a long way to go before we shall be able to understand even the main principles on which the brain works. But, loyalty to one's colleagues apart, I suspect that neurophysiology alone can never lead to a full understanding of the brain. The real difficulty, as I see it, is in defining the problem. It is all very well to ask for a neuro-physiological interpretation of the physical activity of an animal, but how can the neurophysiologist provide us with an interpretation of its mental activity, unless we can find an independent way of describing mental activity? It's no good defining mental activity in neurophysiological terms, because this would preclude explaining it in those terms. In short, if we want the neurophysiologist to help us to understand how the brain works, we must tell him, in non-physiological terms, what we mean by the word ‘works’. And at this point we find ourselves on the frontiers of thought.

Again it will be as well if I summarise before continuing. In the last few decades—though this trend is less evident now—psychologists have tended to play down the significance of subjective evidence and to concentrate on those things that can be measured with, clocks, electrodes, and chemical tests. But the baby is in danger of being thrown away with the bath water. We might, it is true, be able to control rats, pigeons, and other pests, by feeding them with brain hormones, or otherwise manipulating their stimulus-response patterns; but when it comes to human beings surely the only ultimate justification for a scientific study of the mind is to enable us to understand ourselves better, and if possible to improve our mental capacities. How can we possibly do this, if we turn our backs on thought itself, and consider only its physical manifestations? Surely it is time that we admitted mental concepts into the scientific study of the mind; but what is a mental concept?

At this point I leave the conventional sciences for a while to consider what other attempts have been made to elucidate the nature of thought. As usual, we find that many of the best ideas have a very long history. One of the oldest intellectual disciplines is Logic—as Kenny said yesterday—the study of the validity of inferences. If there are any concepts which must find a place in a theory of the mind, then inference is surely among them, and logic used to be regarded as embodying the laws of thought. The phrase ‘the laws of thought’ sounds strange to the modern ear at a time when scientific laws are taken to be business-like statements about down-to-earth matters like magnetism, or crystals, or cell membranes; and logic has come a long way since Aristotle. In a modern textbook of logic one is rather unlikely to discover any reference to the way our minds work. Certainly none of the arguments will appeal in any way to facts about human intelligence. The only symptom of concern with human thought is an occasional appeal to the reader's intuition as to how the symbolism may be informally interpreted, and perhaps a few examples for him to work out. If the author offers any apology for his subject this is much more likely to refer to the foundations of mathematics than to human reasoning per se. All this is entirely healthy, so long as the ultimate goal of the enterprise is not overlooked. If logic is to be justified solely as a critique of mathematical reasoning, then what is the justification for mathematics itself, or rather, what is mathematics? To ask this question is to invite the sarcastic reply that mathematics is what mathematicians do, either for their own benefit, if they are pure, or for the benefit of others, if they are applied. But even this riposte leaves open the question: how are we to decide whether a particular piece of mathematics is right or wrong? And here an appeal to some independent court is unavoidable. In the last resort, a piece of mathematics must stand or fall by whether it meets the demands of human reason; and we are back in the realm of the mind. In any but the most formalistic age, it would seem entirely natural to suggest that our mathematics is inspired by the attempt to capture the essence of our own processes of reasoning.

Another idea of great antiquity, referred to yesterday by Kenny, is that the secret of human thought is to be found in the study of human language. Language may not be the only medium in which thoughts can be expressed, but it is the faculty which most obviously distinguishes us from other animals, and permits the philosopher to discuss, the scientist to describe, not only the world but the nature of man. As most members of this audience are undoubtedly aware, the study of language has suddenly entered a new age. The leader of the new linguistics was, of course, the American linguist, Noam Chomsky. Chomsky's special contribution to linguistic thought was his comparison between the grammatical sentences of a natural language and the theorems of a logical system. Modern logic is conducted in symbols, as language is conducted in words, and there are strict rules in logic for determining whether a string of symbols represents a theorem; that is, whether it can be derived from a particular set of strings called the axioms. To say this might give the impression that Chomsky's primary concern was with form rather than with content, and this impression would not be entirely misleading. But to Chomsky and his school the concept of form gives place to that of structure, and the structure of a sentence in a natural language is identified with the manner in which that sentence is derived from the vocabulary of the language by the application of grammatical rules. I shall not go into detail now, as we shall be discussing language very fully in later sessions. I merely want to emphasise that language is an exceedingly rich mine of information about our mental processes, that its description is a highly non-trivial undertaking, which has already led to some general insights which are far from obvious, and that the study of language is no less scientific an enterprise than the study of aggressive behaviour or of courtship patterns, as its primary data is equally open to observation and to theoretical interpretation. As in other branches of science, a particular observation, or its interpretation, must always be open to revision, but the categories within which ideas about language must be framed seem to be much more nearly in keeping with a science of the mind than those which must be used for interpreting, let us say, the blink reflex, or the effects of narcotics upon attention.

So logic and linguistics seem to be directly concerned with the nature of mental processes in a way in which neurophysiology is not, and physics could not possibly ever be. But how are they to be integrated into psychology, or at least into that part of psychology which has survived the ravages of behaviourism? Because the picture is not as black as I painted it; there are many psychologists who pay their subjects the compliment of attending to their reports, and are prepared to incorporate into their theories concepts such as recognition, or interpretation, or decision, which do not lend themselves readily to statistical analysis. I want to suggest that the problem of describing the mind becomes very much clearer if we recognise that in speaking of ‘the mind’ we are not speaking of a static or passive entity, but of an enormously complex pattern of processes, far too rapid for us to reflect upon as we carry them out. One of the most interesting psychological case-studies on record is that of the late Professor A. C. Aitken, of this university. Aitken was probably the most prodigious mental calculator in all history, as well as being a first class mathematician. He was able to recount, in moderate detail, what happened in his mind when he was asked, for example, to work out the cube root of a nine-figure number; and his reports, which it would be perverse to disregard, reveal an ability to run through an incredibly complex set of mental processes in far less time than it took him to report them. He lived long enough to witness the marvels of modern computing, and it is said that he regarded the computer as an unfair competitor. And this brings me to my last point.

The computer is, without doubt, the most interesting of modem inventions. Many people see it as a threat, of course, but the threat arises from its ability to do what no machine has ever done before: to carry out logical operations. It is interesting to think what a neurophysiologist or a physicist might make of a computer, if they had never set eyes on one before. The physicist would discover inside it a large number of magnetic memory elements, and there his interest would probably stop. The neuro-physiologist would trace the wires leading in and out of the memory, and make an anatomical map, including the various pieces of peripheral equipment; he might also note that sharp pulses travelled hither and thither, and activated the peripherals in an irregular way. But neither of them would really understand what was happening unless someone came along and explained about programs, and about computing languages, in which programs are written. The analogy with human thinking begins to fail at this point, because human beings can of course think very well without being programmed; perhaps I shouldn't say without ever having been programmed, but without being programmed at the time they do the thinking. But it does bring out one point of substance for our discussions, namely that the whole enterprise of understanding our minds is doomed to failure unless someone—and it had better be the psychologist—is prepared to undertake the description of mental processes in terms at least as abstract as those which are needed for describing computing ‘software’, as it is called. It is a commonplace to say that computing is the implementation of logical algorithms, that is, precisely ordered sets of logical instructions. Isn't it about time that psychology embraced the idea of an algorithm, and began to formulate a theory of thought in algorithmic terms? The only risk I can see in adopting this strategy is that it might fail to throw any light on the nature of thought—which I very much doubt—or that it might fail to illuminate all our problems—which seems very likely but can hardly count as an argument against making the attempt.

In this talk I have probably done very much less than was expected of me. I have not attempted to analyse the essential nature of mind, or to list all those mental faculties which we hold most precious. Others are better qualified than I to do that. All I have tried to do is to show that neither neurophysiology nor behaviourist psychology will suffice for the construction of a science of the mind, because their concepts are not mental but physical. The initiative must come from a more abstract level of description, such as logic or linguistics, and a psychological theory worthy of the name must accommodate the concepts of these subjects. It goes without saying that the psychology of the future must harmonise with the findings of neurophysiology, just as chemistry must harmonise with physics; but neither physics nor physiology can possibly dictate the laws which describe how our minds work.



I've been asked to start the discussion, but perhaps I'm not a very good person to start, because I am in sympathy with ninety-five percent of what Longuet-Higgins said. In the first part of his paper he argued in a general way against the programme of reductionism in science, and with that I'm in entire agreement. In the second part of his paper, he argued more in particular that he thought that the science of the mind, which is, as yet, something rather in the future, could not be reduced to either of the existing sciences of behavioural psychology or of neuro-physiology. And here there were two points with which I disagree. Professor Longuet-Higgins admitted that his account of the neuro-physiologist was something of a caricature. Though he didn't say so, I think that perhaps his account of the behaviourist was also something of a caricature, and, having myself last night been critical of behaviourism, I'd like to defend the behaviourist in one respect.

I think that behaviourism was characterised by Longuet-Higgins as an approach determined to do away with human testimony in investigating its subject matter. I think that he is wrong and that behaviourists are right, in saying that one shouldn't pay much attention to people's testimony about their thoughts, if by testimony we mean the kind of thing which Longuet-Higgins seems to mean—the remarks made, say, by the late Professor Aitken, about what went on in his mind when he did his prodigious calculation, or for that matter, the kind of thing which Waddington quoted last night from Einstein. What makes the thoughts of Einstein great—indeed, what makes them thoughts at all—is not what imagery or what visceral thrills he said occurred while he was thinking out the answer to his problems; it's rather that the answers, when—he comes to express them in symbols, whether in mathematical symbols or in language, can be understood by others, can be criticised by others, can be used by others to guide and inform their own researches and their own experiments. If we are to study human intelligent thought, we mustn't take the expression of those thoughts as being mere causal results of the true hidden thought of which these are just the visible effects. The verbal expression of the thought is itself an instance, and indeed the paradigm instance, of the phenomenon which is to be studied. It's the preeminent instance of that phenomenon and it's the first thing that we have to study. To this extent the behaviourist, it seems to me, is perfectly right, though I'm not sure how far Longuet-Higgins would in the long run disagree, since later in his paper he said that language is a primary datum which is open to observation, and that I certainly agree with.

In the second place I want to take up a remark which is perhaps unfair to take up because it was meant at least partly as a joke. Longuet-Higgins said that the late Professor Aitken regarded the computer as an unfair competitor of the mathematician. Now, though this was a joke, I think that Longuet-Higgins does really believe that computers can do arithmetic. In one sense, of course, it is obviously true that computers can do arithmetic, and can do it better than we can. But computers can do arithmetic better than we can in precisely the same sense as clocks can tell the time better than we can; in the sense that, if you want to know what the time is, you do better to look at a clock than to introspect or to ask a neighbour who hasn't got a watch. But of course, in another sense, clocks can't tell the time at all: it is we who tell the time, using clocks as our instruments. Clocks can't tell the time, because clocks can't know the time, and if clocks could tell the time they wouldn't know what to do with it when they told it.


Well, I should be ashamed to spoil that last point; let me take Kenny's earlier point. I think that Einstein's thoughts and Aitken's thoughts are interesting, and I think they are instructive and they tell us something about the way people think, or the way some people can think—and after all it's what the mind can do which is of more interest than what it always does. The highest manifestations obviously are exceedingly interesting, but if we didn't believe that it was of significance to tell other people what went through our minds when we solved problems, then what the hell is a university for? At a university we try to teach people how to think, and we do this by introspecting as best we can. There isn't very much theory about this yet, but we do describe to other people our own thought processes, hoping that the descriptions which we give them may appeal to them and may help them to direct their minds in a similar way. No, I just cannot allow that it's ultimately of no concern as to how Einstein solves his problems. If we could teach ourselves to think in the way that Einstein does, no doubt we'd think a great deal better.

Well now, the second point, arising out of that: I wasn't suggesting for a moment that one should disregard overt behaviour. I was merely suggesting that there are people who put blinkers on their scientific research. I wasn't criticising behavioural psychology, let me say; I was talking about a certain philosophical school in psychology, called the behaviourists. I don't suppose there are any behaviourists in this room, although doubtless there are many behavioural psychologists. But a behaviourist psychologist is a person who holds a certain dogma about psychology, and I couldn't find any other way in which to represent this dogma clearly, but to say that it excludes from serious consideration human testimony, so that we are only allowed to pay serious attention to the way that this thing responds when we do this and that to it. Of course, there are dangers in believing what people say, because people can be terribly misled about their own mental processes; but one doesn't want to disregard any clues which one might be able to use.


I am in an awkward position, because I agree with Longuet-Higgins’ conclusions, which are admirable and true and worthy to be believed, but my logical conscience doesn't allow me to accept the arguments by which he reaches them. That is to say, it seems to me quite evident, if one looks at the whole field of human knowledge, that reductionism doesn't work and is untrue. What is difficult, though, is to see how the reductionist arguments may be met, and I don't think as yet that their case has been properly answered. Some points have been made, for instance the stamp-collecting argument is a very important one, and I go entirely along with that. Aristotle was the first to put this forward—he called them ‘formal causes’—what sort of thing it is. And it's clearly of the greatest importance in chemistry, also in biology. Biologists have realised that it is so important that they have a special name for it: taxonomy, and if this isn't enough, we can consider the case of doctors where much the rarest skill and much the most important skill is that of diagnosis—‘what is it that is wrong with you?’ Now this far I go, with Longuet-Higgins, in seeing that there is more than one question that we have to face, and it's because there is more than one question that not all the sciences are the same science. Biology asks different questions, and therefore can't be reduced to chemistry or physics. But the difficulty is to be sure that the different questions don't get in the way of each other. And now I'm just going to raise two difficulties, and try and give one answer which may not do.

One point which I think is worth making, and that we are, and rightly, worried about is this: when the biologist is asking a question we feel that what the chemist or the physicist has to say is going to be of very great importance; but then how come that it doesn't completely answer the question? For instance, one can't manage without air. It seems that if you know all about the oxygen supply surely you must know all about respiration. And in answer to this I want to put forward a slightly sophisticated logician's move, which is to talk about bound variables. The crucial point is that, whereas the Laplacean physicist thought that he would be able to know where every molecule of oxygen was, and every other molecule, and work out from an initial state description the whole of the subsequent course of the universe, the biologist doesn't need this information, and would find it entirely irrelevant, because from his point of view one molecule of oxygen is as good as another. And this indefinite replaceability is one of the ways by which we can distinguish what is true, in the reductionist's case, from what is false.

A second point, which I'm not so sure about, is to take the difficulty, which seems to arise as we think about the reductionist case, that if I know all about something very simple then I must be able to answer more complicated questions, because they must be definable in terms of necessary and sufficient conditions of the very simple. This is an old, old thesis, a thesis of logical atomism, and how this is to be answered I think can be partly seen by an analogy in the human disciplines, where we often have some rather general, vague concept—motive was suggested earlier—which clearly is connected with behaviour. I can't be really and truly generous if I don't ever give anybody anything, yet it's not to be defined in terms of behaviour; because I can be generous in this way or in that, or the other; my behaviour is evidence for my generosity or not, but it's always only prima facie evidence, which can be rebutted by further evidence of my failure to be generous on some other occasion. And what I want just simply to air for the moment is the possibility that we shall find part of the answer to the reductionist's case in moving from the logic of necessary and sufficient conditions to the logic which we are much more familiar with, in history, in the law court, in philosophy, of stating a case, facing objections to it, rebutting those objections, and then having those rebuttals again turned against ourselves.


Could I make a remark about reductionism, because I think I'm somewhat more sympathetic to it than Longuet-Higgins. I've always in the past really considered myself rather strongly anti-reductionist. It's only after listening to him that I am beginning to have some doubts as to whether I'm as anti as I once thought.

The real snag of reductionism, it seems to me, arises if you suppose that we really know all there is to be known about the physical entities and laws. Now I've lived long enough to have been taught chemistry at a time when what I was taught is now totally changed. I was taught that molecules were made up of groups of atoms which stick together with valancy bonds, like little hooks sticking out of them in certain directions, with which they could join together. This left absolutely no possibility of the very large scale protein molecules with their tertiary structure and allosteric behaviour and such things, which are now explained as depending on incomplete saturation of the bonds between the atoms that are primarily joined together in the basic molecule. That concept didn't occur forty years ago, at any rate in the chemistry I was taught; it's been added on since. You must, I think, always realise that we don't know all about the basic physical entities, we don't know all about the electron, we don't know all about the quantum. If we discover new phenomena, which can't be squeezed into what we already knew, we just add a bit to what we thought we knew about the original, elementary structures. At least, we can try to do that, but I don't think we always can do it. I am quite in agreement with Longuet-Higgins that mental phenomena have to be described in terms of mental systems, and the questions to ask about intentions and mental operations cannot be phrased into physical terms—or at least, they can't as yet be phrased into physical terms, in such a way that you could get an answer by any modification of basic physical concepts. Maybe we never will be able to. But I'm sympathetic enough to the reductionist position to say we should at least try to; that if we have complex mental phenomena, and have to describe them in a non-physical way, we will want to ask non-physical questions about them, but we can try to invent new physical explanations.

In biology, for instance, it might be said that to all the questions we want to ask about evolution we give an answer in terms of phenomena such as natural selection, which you can hardly translate into physical terms, or if you did it would be so fantastically clumsy as to be unusable. You may have to apply different types of explanation to answer the questions you want to ask. But unless you try to get your answers in terms congruent with the other sciences as you know them already, you are liable to go off and invent specific explanations which you will never be able to incorporate into the rest of science. In the specific case of neurophysiology, have the physical factors told us anything relevant to mental events? The sort of questions one wants to ask about linguistics, or about logic, are questions that have to be framed in terms of things like algorithmic programmes, or in some sort of mentalistic terms, and cannot be framed easily in terms of the passage of currents among neurons. On the other hand, the fact that we know that when events are going on in the brain they take the form of electric currents passing through many cells in many different regions of the brain, does give us some indication of the kind of animal we are dealing with; the kind of thing we are talking about. For instance, there are plenty of cells in the brain; many more single cells than are needed to provide one cell per word in the dictionary. I'm not certain whether there are enough to provide one cell for each sentence anybody says in his lifetime, but I shouldn't be terribly surprised. So you could have imagined that mental events involve single specific individual cells, or a smallish number of individual cells for each mental event. This would have meant that a mental event had quite a different basic character from that which it has when it involves a few millions of cells, reverberating and interacting, with currents passing between them. It does seem to me that although the neuro-physiological understanding is not yet refined enough to shed any bright light on mental concepts, it nevertheless does give us an indication of the kind of thing we are talking about when we speak about a logical algorithm.


I don't think I can really deal with John Lucas’ point in a moment, but I think I might be able to say something in reply to Waddington. I think there is quite a difference between the reductionist position as it might have been before 1927 and the position as it is now, because—though it sounds terribly dogmatic to say so—we really do know that chemical phenomena are determined by the Schrodinger equation. The equation has survived a fantastic amount of testing, and it really seems to be right. So it's no good trying to fiddle with the foundations. As for the other point which you made—if you want to understand the tertiary structure of proteins, well, if you've got a relevant chemical concept that might need revision, indeed it might; which just goes to show how chemical concepts are not the same thing as physical concepts.

Just to make a remark about the brain, and all those neurons and so forth, I think the computer is particularly helpful here, helping one to think clearly about the relation between different kinds of problem. If I have a computing language, and a program, and a computer, I can ask different kinds of question about a computation. We can ask what's the logic of the computation, or we can ask, how does the computer interpret the language. Finally we can ask how does the computer actually work—has it got solid state circuits and so forth. If I advance a theory about the way this computation is implemented, which simply doesn't fit with the actual physical structure of the computer, of course I'm talking rubbish. And it's that sense in which we can say that the higher level—more abstract—concepts or assertions have got to square with the assertions at the lower level. It's very difficult to make a general statement making it clear quite how the higher level generalisations relate to the lower level restrictions, but I think this is a case in which we can see very clearly in what manner that relation is to be considered.

From the book: