*  ESSAY *

Of Clouds and Clocks


Karl Popper


Objective Knowledge: An Evolutionary Approach

For the notes, see the PDF Part 1 and Part 2.

MY predecessor who in this hall gave the first Arthur Holly Compton Memorial Lecture a year ago was more fortunate than I. He knew Arthur Compton personally; I never met him.

But I have known of Compton since my student days in the nineteen-twenties, and especially since 1925 when the famous experiment of Compton and Simone refuted the beautiful but short-lived quantum theory of Bohr, Kramers, and Slater.3 This refutation was one of the decisive events in the history of quantum theory, for from the crisis which it created there emerged the so-called `new quantum theory'-the theories of Born and Heisenberg, of Schrodinger, and of Dirac.

It was the second time that Compton's experimental tests had played a crucial role in the history of quantum theory. The first time had been, of course, the discovery of the Compton effect, the first independent test (as Compton himself pointed out) of Einstein's theory of light quanta or photons.

Years later, during the Second World War, I found to my surprise and pleasure that Compton was not only a great physicist but also a genuine and courageous philosopher; and further, that his philosophical interests and aims coincided with my own on some important points. I found this when, almost by accident, I got hold of Compton's fascinating Terry Lectures which he had published in 1935 in a book entitled The Freedom of Man.

You will have noticed that I have incorporated the title of Compton's book, The Freedom of Man, into my own title today. I have done so in order to stress the fact that my lecture will be closely connected with this book of Compton's. More precisely, I intend to discuss the same problems which Compton. discussed in the first two chapters of this book, and again in the second chapter of another of his books, The Human Meaning of Science.

In order to avoid misunderstandings I must stress, however, that my lecture today is not mainly about Compton's books. It is rather an attempt to look afresh at the same ancient philosophical problems with which he grappled in these two books, and an attempt to find a new solution to these ancient problems. The sketchy and, very tentative solution I am going to outline here seems to me to fit in well with Compton's main aims, and I hope-indeed I believe-that he would have approved of it.


THE central purpose of my lecture is to try to put these ancient problems simply and forcefully before you. But first I must say something about the clouds and clocks which appear in the title of my lecture.

My clouds are intended to represent physical. systems which, like gases, are highly irregular, disorderly, and more or less unpredictable. I shall assume that we have before us a schema or arrangement in which a very disturbed or disorderly cloud is placed on the left. On the other extreme of our arrangement, on its right, we may place a very reliable pendulum clock, a precision clock, intended to represent physical systems which are regular, orderly, and highly predictable in their behaviour. 

According to what I may call the commonsense view of things, some natural phenomena, such as the weather, or the coming and going of clouds, are hard to predict: we speak of the `vagaries of the weather'. On the other hand, we speak of `clockwork precision' if we wish to describe a highly regular and predictable phenomenon.

There are lots of things, natural processes and natural phenomena, which we may place between these two extremes-the clouds on the left, and the clocks on the right. The changing seasons are somewhat unreliable clocks, and may therefore be put somewhere towards the right, though not too far. I suppose we shall easily agree to put animals not too far from the clouds on the left, and plants somewhat nearer to the clocks. Among the animals, a young puppy will have to be placed further to the left than an old dog. Motor cars, too, will find their place somewhere in our arrangement, according to their reliability: a Cadillac, I suppose, is pretty far over to the right, and even more so a Rolls-Royce, which will be quite close to the best of the clocks. Perhaps furthest to the right should be placed the solar system?

As a typical and interesting example of a cloud I shall make some use here of a cloud or cluster of small flies or gnats. Like the individual molecules in a gas, the individual gnats which together form a cluster of gnats move in an. astonishingly irregular way. It is almost impossible to follow the flight of any one individual gnat, even though each of them may be quite big enough to be clearly visible.

Apart from the fact that the velocities of the gnats do not show a very wide spread, the gnats present us with an excellent picture of the irregular movement of molecules in a gas cloud, or of the minute drops of water in a storm cloud. There are, of course, differences. The cluster does not dissolve or diffuse, but it keeps together fairly well. This is surprising, considering the disorderly character of the movement of the various gnats; but it has its analogue in a sufficiently big gas cloud (such as our atmosphere, or the sun) which is kept together by gravitational forces. In the case of the gnats, their keeping together can be easily explained if we assume that, although they fly quite irregularly in all directions, those that find that they are getting away from the crowd turn back towards that part which is densest.

This assumption explains how the cluster keeps together even though it has no leader, and no structure-only a random statistical distribution resulting from the fact that each gnat does exactly what he likes, in a lawless or, random manner, together with the fact that he does not like to stray too far from his comrades.

I think that a philosophical gnat might claim that the gnat society is a great society or at least a good society, since it is the most egalitarian, free, and democratic society imaginable.

However, as the author of a book on The Open Society, I would deny that the gnat society is an open society. For I take it to be one of the characteristics of an open society that it cherishes, apart from a democratic form of government, the freedom of association, and that it protects and even encourages the formation of free sub-societies, each holding different opinions and beliefs. But every reasonable gnat would have to admit that in his society this kind of pluralism is lacking.

I do not intend, however, to discuss today any of the social or political issues connected with the problem of freedom; and I intend to use the cluster of gnats not as an example of a social system, but rather as my main illustration of a cloud-like physical system, as an example or paradigm of a highly irregular or disordered cloud.

Like many physical, biological, and social systems, the cluster of gnats may be described as a 'whole'. Our conjecture that it is kept together by a kind of attraction which its densest part exerts on individual gnats straying too far from the crowd shows that there is even a kind of action or control which this 'whole' exerts upon its elements or parts. Nevertheless, this `whole' can be used to dispel the widespread `holistic' belief that a `whole' is always more than the mere sum of its parts. I do not deny that it may sometimes be so. Yet the cluster of gnats is an example of a whole that is indeed nothing but the sum of its and in a very precise sense; for not only is it completely described by describing the movements of all the individual gnats, but the movement of the whole is, in this case, precisely the (vectorial) sum of the movements of its constituent members, divided by the number of the members.

An example (in many ways similar) of a biological system or 'whole' which exerts some control over the highly irregular movements of its parts would be a picnicking family-parents with a few children and a dog-roaming the woods for hours, but never straying far from the family car (which acts like a centre of attraction, as it were). This system may be said to be even more cloudy-that is, less regular in the movement of its parts-than our cloud of gnats.

I hope you will now have before you an idea of my two prototypes or paradigms, the clouds on the left and the clocks on the right, and of the way in which we can arrange many kinds of things, and many kinds of systems, between them. I am sure you have caught some vague, general idea of the arrangement, and you need not worry if your idea is still a bit foggy, or cloudy.


THE arrangement I have described is, it seems, quite acceptable to common sense; and more recently, in our own time, it has become acceptable even to physical science. It was not so, however, during the preceding 250 years: the Newtonian revolution, one of the greatest revolutions in history, led to the rejection of the commonsense arrangement which I have tried to present to you. For one of the things which almost everybody thought had been established by the Newtonian revolution was the following staggering proposition:

All clouds are clocks-even the most cloudy of clouds.

This proposition, `All clouds are clocks', may be taken as a brief formulation of the view which I shall call `physical determinism'.

The physical determinist who says that all clouds are clocks will also say that our commonsense arrangement, with the clouds on the left and the clocks on the right, is misleading, since everything ought to be placed on the extreme right. He will say that, with all our common sense, we arranged things not according to their nature, but merely according to our ignorance. Our arrangement, he will say, reflects merely the fact that we know in some detail how the parts of a clock work, or how the solar system works, while we do not have any knowledge about the detailed interaction of the particles that form a gas cloud, or an organism. And he will assert that, once we have obtained this knowledge, we shall find that gas clouds or organisms are as clock-like as our solar system.

Newton's theory did not, of course, tell the physicists that this was so. In fact, it did not treat at all of clouds. It treated especially of planets, whose movements it explained as due to some very simple laws of nature; also of cannon balls, and of the tides. But its immense success in these fields turned the physicists' heads; and surely not without reason.

Before the time of Newton and his predecessor, Kepler, the movements of the planets had escaped many attempts to explain or even to describe them fully. Clearly, they somehow participated in the unvarying general movement of the rigid system of the fixed stars; yet they deviated from the movement of that system almost like single gnats deviating from the general movement of a cluster of gnats. Thus the planets, not unlike living things, appeared to be in a position intermediate between clouds and clocks. Yet the success of Kepler's and even more of Newton's theory showed that those thinkers had been right who had suspected that the planets were in fact perfect clocks. For their movements turned out to be precisely predictable with the help of Newton's theory; predictable in all those details which had previously baffled the astronomers by their apparent irregularity.

Newton's theory was the first really successful scientific theory in human history; and it was tremendously successful. Here was real knowledge; knowledge beyond the wildest dreams of even the boldest minds. Here was a theory which explained precisely not only the movements of all the stars in their course, but also, just as precisely, the movements of bodies on earth, such as falling apples, or projectiles, or pendulum clocks. And it even explained the tides.

All open-minded men-all those who were eager to learn, and who took an interest in the growth of knowledge-were converted to the new theory. Most openminded men, and especially most scientists, thought that in the end it would explain everything, including not only electricity and magnetism, but also clouds, and even living organisms. Thus physical determinism-the doctrine that all clouds are clocks-became the ruling faith among enlightened men; and everybody who did not embrace this new faith was held to be an obscurantist or a reactionary.


AMONG the few dissenters was Charles Sanders Peirce, the great American mathematician and physicist and, I believe, one of the greatest philosophers of all time. He did not question Newton's theory; yet as early as1892 he showed that this theory, even if true, does not give us any valid reason to believe that clouds are perfect clocks. Though in common with all other physicists of his time he believed that the world was a clock that worked according to Newtonian laws, he rejected the belief that this clock, or any other, was perfect, down to the smallest detail. He pointed out that at any rate we could not possibly claim to know, from experience, of anything like a perfect clock, or of anything even faintly approaching that absolute perfection which physical determinism assumed. I may perhaps quote one of Peirce's brilliant comments: `. . . one who is behind the scenes' (Peirce speaks here as an experimentalist) `... knows that the most refined comparisons [even] ofmasses [and] lengths, ... far surpassing in precision all other. [physical] measurements, ... fall behind the accuracy of bank accounts, and that the ... determinations of physical constants ... are about on a par with an upholsterer's measurements of carpets and curtains . . .'. From this Peirce concluded that we were free to conjecture that there was a certain looseness or imperfection in all clocks, and that this allowed an element of chance to enter. Thus Peirce conjectured that the world was not only ruled by the strict Newtonian laws, but that it was also at the same time ruled bylaws of chance, or of randomness, or of disorder: by laws of statistical probability. This made the world an interlocking system of clouds and clocks, so that even the best clock would, in its molecular structure, show some degree of cloudiness. So far as I know Peirce was the first. post-Newtonian physicist and philosopher who thus dared to adopt the view that to some degree all clocks are clouds; or in other words, that only clouds exist, though clouds of very different degrees of cloudiness.

Peirce supported this view by pointing out, no doubt correctly, that all physical bodies, even the jewels in a watch, were subject to molecular heat motion, a motion similar to that of the molecules of a gas, or of the individual gnats in a cluster of gnats.

These views of Peirce's were received by his contemporaries with little interest. Apparently only one philosopher noticed them; and he attacked them. Physicists seem to have ignored them; and even today most physicists believe that if we had to accept the classical mechanics of Newton as true, we should be compelled to accept physical determinism, and with it the proposition that all clouds are clocks. It was only with the downfall of classical physics and with the rise of the new quantum theory that physicists were prepared to abandon physical determinism.

Now the tables were turned. Indeterminism, which up to 1927 had been equated with obscurantism, became the ruling fashion; and some great scientists, such as Max Planck, Erwin Schrodinger, and Albert Einstein, who hesitated to abandon determinism, were considered old.fogies,15 although they had been in the forefront of the development of quantum theory. I myself once heard a brilliant young physicist describe Einstein, who was then still alive and hard at work, as 'antediluvian'. The deluge that was supposed to have swept Einstein away was the new quantum theory, which had risen during the years from 1925 to 1927, and to whose advent at most seven people had made contributions comparable to those of Einstein.


PERHAPS I may stop here for a moment to state my own view of the situation, and of scientific fashions. I believe that Peirce was right in holding that all clocks are clouds, to some considerable degree - even the most precise of clocks. This, I think, is a most important inversion of the mistaken determinist view that all clouds are clocks. I further believe that Peirce was right in holding that this view was compatible with the classical physics of Newton.16 I believe that this view is even more clearly compatible with Einstein's (special) relativity theory, and it is still more clearly compatible with the new quantum theory. In other words, I am an indeterminist-like Peirce, Compton, and most other contemporary physicists; and I believe, with most of them, that Einstein was mistaken in trying to hold fast to determinism. (I may perhaps say that I discussed this matter with him, and that I did not find him adamant.) But I also believe that those modern physicists were badly mistaken who pooh-poohed as antediluvian Einstein's criticism of the quantum theory. Nobody can fail to admire the quantum theory, and Einstein did so wholeheartedly; but his criticism of the fashionable interpretation of the theory-the Copenhagen interpretation-like the criticisms offered by de Broglie, Schrodinger, Bohm, Vigier, and more recently by Lande, have been too lightly brushed aside by most physicists.'" There are fashions in science, and some scientists climb on the band wagon almost as readily as do some painters and musicians. But although fashions and bandwagons may attract the weak, they should be resisted rather than encouraged;18 and criticism like Einstein's is always valuable: one can always learn something from it.


ARTHUR HOLLY COMPTON was among the first who welcomed the new quantum theory, and Heisenberg's new physical indeterminism of 1927. Compton invited Heisenberg to Chicago for a course of lectures which Heisenberg delivered in the spring of 1929. This course was Heisenberg's first full exposition of his theory, and his lectures were published as his first book a year later by the University of Chicago Press, with a preface by Arthur Compton.,,) In this preface Compton welcomed the new theory to whose advent his experiments had contributed by refuting its immediate predecessor;20 yet he also sounded a note of warning. Compton's warning anticipated some very similar warnings by Einstein, who always insisted that we should not consider the new quantum theory-'this chapter of the history of physics', as Compton called it generously and wisely-as being `complete'.21 And although this view was rejected by Bohr, we should remember the fact that the new theory failed, for example, to give even a hint of the neutron, discovered by Chadwick about a year later, which was to become the first of a long series of new elementary particles whose existence had not been foreseen by the new quantum theory (even though it 'is true that the existence of the positron could have been derived from the theory of Dirac).

In the same year, 1931, in his Terry Foundation Lectures, Compton became one of the first to examine the human and, more generally, the biological23 implications of the new indeterminism in physics. And now it became clear why he had welcomed the new theory so enthusiastically: it solved for him not only problems of physics but also biological and philosophical problems, and among the latter especially problems connected with ethics.


To show this, I shall now quote the striking opening passage of Compton's The Freedom of Man:

The-fundamental question of morality, a vital problem in religion, and a subject of active investigation in science: Is man a free agent? If... the atoms of our bodies follow physical laws as immutable as the motions of the planets, why try? What difference can it make how great the effort if our actions are already predetermined by mechanical laws ... ?24

Compton describes here what I shall call `the nightmare of the physical determinist'. A deterministic physical clockwork mechanism is, above all, completely self-contained: in the perfect deterministic physical world there is simply no room for any outside intervention. Everything that happens in such a world is physically predetermined, including all our movements and therefore all our actions. Thus all our thoughts, feelings, and efforts can have no practical influence upon what happens in the physical world: they are, if not mere illusions, at best superfluous byproducts ('epiphenomena') of physical events.

In this way, the daydream of the Newtonian physicist who hoped to prove all clouds to be clocks had threatened to turn into a nightmare; and the attempt to ignore this had led to something like an intellectual split personality. Compton, I think, was grateful to the new quantum theory for rescuing him from this difficult intellectual situation. Thus he writes, in The Freedom of Man: `The physicist has rarely ... bothered himself with the fact that if ... completely deterministic ... laws ... apply to man's actions, he is himself an automaton.' And in he Human Meaning of Science he expresses his relief:

In my own thinking on this vital subject I am thus in a much more satisfied state of mind than I could have been at any earlier stage of science. If the statements of the laws of physics were assumed correct, one would have had to suppose (as did most philosophers) that the feeling of freedom is illusory, or if [free] choice were considered effective, that the statements of the laws of physics were ... unreliable. The dilemma has been an uncomfortable one .. .

Later in the same book Compton sums up the situation crisply in the words: `... it is no longer justifiable to use physical law as evidence against human freedom'.27

These quotations from Compton show clearly that before Heisenberg he had been harassed by what I have here called the nightmare of the physical determinist, and that he had tried to escape from this nightmare by adopting something like an intellectual split personality. Or as he himself puts it: `We [physicists] have preferred merely to pay no attention to the difficulties.. .'

Compton welcomed the new theory which rescued him from all this. I believe that the only form of the problem of determinism which is worth discussing seriously is exactly that problem which worried Compton: the problem which arises from a physical theory which describes the world as a physically complete or a physically closed system. By a physically closed system I mean a set or system of physical entities, such as atoms or elementary particles or physical forces or fields of forces, which interact with each other- and only with each other -in accordance with definite laws of interaction that do not leave any room for interaction with, or interference by, anything outside that closed set or system of physical entities. It is this `closure' of the system that creates the deterministic nightmare.


I SHOULD like to digress here for a minute in order to contrast the problem of physical determinism, which I consider to be of fundamental importance, with the far from serious problem which many philosophers and psychologists, following Hume, have substituted for it. .

Hume interpreted determinism (which he called `the doctrine of necessity', or `the doctrine of constant conjunction') as the doctrine that `like causes always produce like effects' and that `like effects necessarily follow from like causes'.3' Concerning human actions and volitions he held, more particularly, that `a spectator can commonly infer our actions from our motives and character; and even where he cannot, he concludes in general, that he might, were he perfectly acquainted with every circumstance of our situation and temper, and the most secret springs of our . . . disposition. Now this is the very essence of necessity ...'.3^ Hume's successors put it thus: our actions, or our volitions, or our tastes, or our preferences, are psychologically `caused' by preceding experiences ('motives'), and ultimately by our heredity and environment.

But this doctrine which we may call philosophical or psychological determinism is not only a very different affair from physical determinism, but it is also one which a physical determinist who understands this matter at all can hardly take seriously. For the thesis of philosophical determinism, that `Like effects have like causes' or that `Every event has a cause', is so vague that it is perfectly compatible with physical indeterminism.

Indeterminism-or more precisely, physical indeterminism-is merely the doctrine that not all events in the physical world are predetermined with absolute precision, in all their infinitesimal details. Apart from this, it is compatible with practically any degree of regularity you like, and it does not, therefore, entail the view that there are `events without causes'; simply because the terms `event' and `cause' are vague enough to make the doctrine that every event has a cause compatible with physical indeterminism. While physical determinism demands complete and infinitely precise physical predetermination and the absence of any exception whatever, physical indeterminism asserts no more than that determinism is false, and that there are at least some exceptions, here or there, to precise predetermination.

Thus even the formula `Every observable or measurable physical event has an observable or measurable physical cause' is still compatible with physical indeterminism, simply because no measurement can be infinitely precise: for the salient point about physical determinism is that, based on Newton's dynamics, it asserts the existence of a world of absolute mathematical precision. And although in so doing it goes beyond the realm of possible observation (as was seen by Peirce), it nevertheless is testable, in principle, with any desired degree of precision; and it actually withstood surprisingly precise tests.

By contrast, the formula `Every event has a cause' says nothing about precision; and if, more especially, we look at the laws of psychology, then there is not even a suggestion of precision. This holds for a `behaviourist' psychology as much as for an `introspective' or `mentalist' one. In the case of a mentalist psychology this is obvious. But even a behaviourist may at the very best predict that, under given conditions, a rat will take twenty to twentytwo seconds to run a maze: he will have no idea how, by specifying more and more precise experimental conditions, he could make predictions which become more and more precise-and, in principle, precise without limit. This is so because behaviourist `laws' are not, like those of Newtonian physics, differential equations, and because every attempt to introduce such differential equations would lead beyond behaviourism into physiology, and thus ultimately into physics; so it would lead us back to the problem of physical determinism.

As noted by Laplace, physical determinism implies that every physical event in the distant future (or in the distant past) is predictable (or retrodictable) with any desired degree of precision, provided we have sufficient knowledge about the present state of the physical world. The thesis of a philosophical (or psychological) determinism of Hume's type, on the other hand, asserts even in its strongest interpretation no more than that any observable difference between two events is related by some as yet perhaps unknown law to some difference-an observable difference perhaps-in the preceding state of the world; obviously a very much weaker assertion, and incidentally one which we could continue to uphold even if most of our experiments, performed under conditions which are, in appearance, `entirely equal', should yield different results. This was stated very clearly by Hume himself. `Even when these contrary experiments are entirely equal', he writes, `we remove not the notion of causes and necessity, but ... conclude, that the [apparent] chance ... lies only in ... our imperfect knowledge, not in the things themselves, which are in every case equally necessary [i.e., determined], tho' to appearance not equally constant or certain.'

This is why a Humean philosophical determinism and, more especially, a psychological determinism, lack the sting of physical determinism. For in Newtonian physics things really looked as if any apparent looseness in a system was in fact merely due to our ignorance, so that, should we be fully informed about the system, any appearance of looseness would disappear. Psychology, on the other hand, never had this character.

Physical determinism, we might say in retrospect, was a daydream of omniscience which seemed to become more real with every advance in physics until it became an apparently inescapable nightmare. But the corresponding daydreams of the psychologists were never more than castles in the air: they were Utopian dreams of attaining equality with physics, its mathematical methods, and its powerful applications; and perhaps even of attaining superiority, by moulding men and societies. (While these totalitarian dreams are not serious from a scientific point of view, they are very dangerous politically; but since I have dealt with these dangers elsewhere I do not propose to discuss the problem here.)


I HAVE called physical determinism a nightmare. It is a nightmare because it asserts that the whole world with everything in it is a huge automaton, and that we are nothing but little cogwheels, or at best sub-automata, within it.

It thus destroys, in particular, the idea of creativity. It reduces to a complete illusion the idea that in preparing this lecture I have used my brain to create something new. There was no more in it, according to physical determinism, than that certain parts of my body put down black marks on white paper: any physicist with sufficient detailed information could have written my lecture by the simple method of predicting the precise places on which the physical system consisting of my body (including my brain, of course, and my fingers) and my pen would put down those black marks. Or to use a more impressive example: if physical determinism is right, then a physicist who is completely deaf and who has never heard any music could write all the symphonies and concertos written by Mozart or Beethoven, by the simple method of studying the precise physical states of their bodies and predicting where they would put down black marks ^ on their lined paper. And our deaf physicist could do even more: by studying Mozart's or Beethoven's bodies with sufficient care he could write scores which were never actually written by Mozart or Beethoven, but which they would have written had certain external circumstances of their lives been different: if they had eaten lamb, say, instead of chicken, or drunk tea instead of coffee.

All this could be done by our deaf physicist if supplied with a sufficient knowledge of purely physical conditions. There would be no need for him to know anything about the theory of music-though he might be able to predict what answers Mozart or Beethoven would have written down under examination conditions if presented with questions on the theory of counterpoint.

I believe that all this is absurd; and its absurdity becomes even more obvious, I think, when we apply this method of physical prediction to a determinist.

For according to determinism, any theories-such as, say, determinism-are held because of a certain physical structure of the holder (perhaps of his brain). Accordingly we are deceiving ourselves (and are physically so determined as to deceive ourselves) whenever we believe that there are such things as arguments or reasons which make us accept determinism.

Or in other words, physical determinism is a theory which, if it is true, is not arguable, since it must explain all our reactions, including what appear to us as beliefs based on arguments, as due to. purely physical conditions. Purely physical conditions, including our physical environment, make us say or accept whatever we say or accept; and a well-trained physicist who does not know any`French, and who has never heard of determinism, would be able to predict what a French determinist would say in a French discussion on determinism; and of course also what his indeterminist opponent would say. But this means that if we believe that we have accepted a theory like determinism because we were swayed by the logical force of certain arguments, then we are deceiving ourselves, according to physical determinism; or more precisely, we are in a physical condition which determines us to deceive ourselves.

Hume saw much of this, even though it appears that he did not quite see what it meant for his own arguments; for he confined himself to comparing the determinism of `our judgements' with that of `our actions', saying that `zee have no more liberty in the one than in the other'.

Considerations such as these may perhaps be the reason why there are so many philosophers who refuse to take the problem of physical determinism seriously and dismiss it as a 'bogy'.37 Yet the doctrine that man is a machine was argued most forcefully and seriously in 1751, long before the theory of evolution became generally accepted, by de Lamettrie; and the theory of evolution gave the problem an even sharper edge, by suggesting that there may be no clear distinction between living matter and dead matter.38 And in spite of the victory of the new quantum theory, and the conversion of so many physicists to indeterminism, de Lamettrie's doctrine that man is a machine has today perhaps more defenders than ever before among physicists, biologists, and philosophers; especially in the form of the thesis that man is a computer.

For if we accept a theory of evolution (such as Darwin's) then even if we remain sceptical about the theory that life emerged from inorganic matter we can hardly deny that there must have been a time when abstract and. non-physical entities, such as reasons and arguments and scientific knowledge, and abstract rules, such as rules for building railways or bulldozers or sputniks or, say, rules of grammar or of counterpoint, did not exist, or at any rate had no effect upon the physical universe. It is difficult to understand how the physical universe could produce abstract entities such as rules, and then could come under the influence of these rules, so that these rules in their turn could exert very palpable effects upon the physical universe.

There is, however, at least one perhaps somewhat evasive but at any rate easy way out of this difficulty. We can simply deny that these abstract entities exist and that they can influence the physical universe. And we can assert that what do exist are our brains, and that these are machines like computers; that the allegedly abstract rules are physical entities, exactly like the concrete physical punch-cards by which we `programme' our computers; and that the existence of anything non-physical is just `an illusion', perhaps, and at any rate unimportant, since everything would go on as it does even if there were no such illusions.

According to this way out, we need not worry about the `mental' status of these illusions. They may be universal properties of all things: the stone which I throw may have the illusion that it jumps, just as I have the illusion that I throw it; and my pen, or my computer, may have the illusion that it works because of its interest in the problems which it thinks that it is solving-and which I think that I am solving-while in fact there is nothing of any significance going on except purely physical interactions.

You may see from all this that the problem of physical determinism which worried Compton is indeed a serious problem. It is not just a philosophical puzzle, but it affects at least physicists, biologists, behaviourists, psychologists, and computer engineers.

Admittedly, quite a few philosophers have tried to show (following Hume or Schlick) that it is merely a verbal puzzle, a puzzle about the use of the word 'freedom'. But these philosophers have hardly seen the difference between the problem of physical determinism and that of philosophical determinism; and they are either determinists like Hume, which explains why for them `freedom' is `just a word', or they have never had that close contact with the physical sciences or with computer engineering which would have impressed upon them that we are faced with more than a merely verbal puzzle.


LIKE Compton I am among those who take the problem of physical determinism seriously, and like Compton I do not believe that we are mere computing machines (though I readily admit that we can learn a great deal from computing machines -even about ourselves). Thus, like Compton, I am a physical indeterminist: physical indeterminism, I believe, is a necessary prerequisite for any solution of our problem. We have to be indeterminists; yet I shall try to show that indeterminism is not enough.

With this statement, indeterminism is not enough, I have arrived, not merely at a new point, but at the very heart of my problem.

The problem may be explained as follows.

If determinism is true, then the whole world is a perfectly running flawless clock, including all clouds, all organisms, all animals, and all men. If, on the other hand, Peirce's or Heisenberg's or some other form of indeterminism is true, then sheer chance plays a major role in our physical world. But is chance really more satisfactory than determinism?

The question is well known. Determinists like Schlick have put it in this way: '... freedom of action, responsibility, and mental sanity, cannot reach beyond the realm of causality: they 6. Of Clouds and Clocks 227 stop where chance begins.... a higher degree of randomness ... [simply means] a higher degree of irresponsibility.'

I may perhaps put this idea of Schlick's in terms of an example I have used before: to say that the black marks made on white paper which I produced in preparation for this lecture were just the result of chance is hardly more satisfactory-than to say that they were physically predetermined. In fact, it is even less satisfactory. For some people may perhaps be quite ready to believe that the text of my lecture can be in principle completely explained by my physical heredity, and my physical environment, including my upbringing, the books I have been reading, and the talks I have listened to; but hardly anybody will believe that what I am reading to you is the result of nothing but chance -just a random sample of English words, or perhaps of letters, put together without any purpose, deliberation, plan, or intention.

The idea that the only alternative to determinism is just sheer chance was taken over by Schlick, together with many of his views on the subject, from Hume, who asserted that `the removal' of what he called `physical necessity' must always result in `the same thing with chance. As objects must either be conjoin'd or not, . . . 'tis impossible to admit of any medium betwixt chance and an absolute necessity'.

I shall later argue against this important doctrine according to which the only alternative to determinism is sheer chance. Yet I must admit that the doctrine seems to hold good for the quantum-theoretical models which have been designed to explain, or at least to illustrate, the possibility of human freedom. This seems to be the reason why these models are so very unsatisfactory.

Compton himself designed such a model, though he did not particularly like it. It uses quantum indeterminacy, and the unpredictability of a quantum jump, as a model of a human decision of great moment. It consists of an amplifier which amplifies the effect of a single quantum jump in such a way that it may either cause an explosion or destroy the relay necessary for bringing the explosion about. In this way one single quantum jump may be equivalent to a major decision. But in my opinion the model has no similarity to any rational decision. It is, rather, a model of a-kind of decision-making where people who cannot make up their minds say: 'Let us toss a penny.' In fact, the whole apparatus for amplifying a quantum jump seems rather unnecessary: tossing a penny, and deciding on the result of the toss whether or not to pull a trigger, would do just as well. And there are of course computers with built-in penny-tossing devices for producing random results, where such are needed.

It may perhaps be said that some of our decisions are like penny-tosses: they are snap-decisions, taken without deliberation, since we often do not have enough time to deliberate. A driver or a pilot has sometimes to take a snap-decision like this; and if he is well trained, or just lucky, the result may be satisfactory; otherwise not.

I admit that the quantum jump model may be a model for such snap-decisions; and I even admit that it is conceivable that something like the amplification of a quantum jump may actually happen in our brains if we make a snap-decision. But are snap-decisions really so very interesting? Are they characteristic of human behaviour-of rational human behaviour?

I do not think so; and I do not think that we shall get much further with quantum jumps. They are just the kind of examples which seem to lend support to the thesis of Hume and Schlick that perfect chance is the only alternative to perfect determinism. What we need for understanding rational human behaviour -and indeed, animal behaviour-is something intermediate in character between perfect chance and perfect determinismsomething intermediate between perfect clouds and perfect clocks.

Hume's and Schlick's ontological thesis that there cannot exist anything intermediate between chance and determinism seems to me not only highly dogmatic (not to say doctrinaire) but clearly absurd; and it is understandable only on the assumption that they believed in a complete determinism in which chance has no status except as a symptom of our ignorance. (But even then it seems to me absurd, for there is, clearly, something like partial knowledge, or partial ignorance.) For we know that even highly reliable clocks are not really perfect, and Schlick (if not Hume) must have known that this is largely due to factors such as friction-that is to say, to statistical or chance effects. And we also know that our clouds are not perfectly chance-like, since we can often predict the weather quite successfully, at least for short periods.


THUS we shall have to return to our old arrangement with clouds on the left and clocks on the right and animals and men somewhere in between. But even after we have done so (and there are some problems to be solved before we can say that this arrangement is in keeping with present-day physics), even then we have at best only made room for our main question.

For obviously what we want is to understand how such nonphysical things as purposes, deliberations, plans, decisions, theories, intentions, and values, can play a part in bringing about physical changes in the physical world. That they do this seems to be obvious, pace Hume and Laplace and'Schlick. It is clearly untrue that all those tremendous physical changes brought about hourly by our pens, or pencils, or bulldozers, can be explained in purely physical terms, either by a deterministic physical theory, or (by a stochastic theory) as due to chance.

Compton was well aware of this problem, as the following charming passage from his Terry Lectures shows: It was some time ago when I wrote to the secretary of Yale University agreeing to give a lecture on November io at 5 p.m. He had such faith in me that it was announced publicly that I should be there, and the audience had such confidence in his word that they came to the hall at the specified time. But consider the great physical improbability that their confidence was justified. In the meanwhile my work called me to the Rocky Mountains and across the ocean to sunny Italy. A phototropic organism [such as I happen to be, would not easily] ... tear himself away from there to go to chilly New Haven. The possibilities of my being elsewhere at this moment were infinite in number. Considered as a physical event, the probability of meeting my engagement would have been fantastically small. Why then was the audience's belief justified?

They knew my purpose, and it was my purpose [which] determined that I should be there.

Compton shows here very beautifully that mere physical indeterminism is not enough. We have to be indeterminists, to be sure; but we also must try to understand how men, and perhaps animals, can be 'influenced' or 'controlled' by such things As aims, or purposes, or rules, or agreements.

This then is our central problem.


A C L o s E R look shows, however, that there are two problems in this story of Compton's journey from Italy to Yale. Of these two problems I shall here call the first Compton's problem, and the second Descartes's problem.

Compton's problem has rarely been seen by philosophers, and if at all, only dimly. It may be formulated as follows:

There are such things as letters accepting a proposal to lecture, and public announcements of intentions; publicly declared aims and purposes; general moral rules. Each of these documents or pronouncements or rules has a certain content, or meaning, which remains invariant if we translate it, or reformulate it. Thus this content or meaning is something quite abstract. Yet it can control-perhaps by way of a short cryptic entry in an engagement calendar-the physical movements of a man in such a way as to steer him back from Italy to Connecticut. How can that be?

This is what I shall call Compton's problem. It is important to note that in this form the problem is neutral with respect to the question whether we adopt a behaviourist or a mentalist psychology: in the formulation here given, and suggested by Compton's text, the problem is put in terms of Compton's behaviour in returning to Yale; but it would make very little difference if we included such mental events as volition, or the feeling of having grasped, or got hold of, an idea.

Retaining Compton's own behaviourist terminology, Compton's problem may be described as the problem of the influence of the universe of abstract meanings upon human behaviour (and thereby upon the physical universe). Here 'universe of meanings' is a shorthand term comprising such diverse things as promises, aims, and various kinds of rules, such as rules of grammar, or of polite behaviour, or of logic, or of chess, or of counterpoint; also such things as scientific publications (and other publications) ; appeals to our sense of justice or generosity; or to our artistic appreciation; and so on, almost ad infinitum.

I believe that what I have here called Compton's problem is one of the most interesting problems of philosophy, even though few philosophers have seen it. In my opinion it is a real key problem, and more important than the classical body-mind problem which I am calling here 'Descartes's problem'.

In order to avoid misunderstandings I may perhaps mention that by formulating his problem in behaviouristic terms, Compton certainly had no intention of subscribing to a full-fledged behaviourism. On the contrary, he did not doubt either the existence of his own mind, or that of other minds, or of experiences such as volitions, or deliberations, or pleasure, or pain. He would therefore have insisted that there is a second problem to be solved.

We may identify this second problem with the classical bodymind problem, or Descartes's problem. It may be formulated as follows: how can it be that such things as states of mind-volitions, feelings, expectations-influence or control the physical movements of our limbs? And (though this is less important in our context) how can it be that the physical states of an organism may influence its mental states?

Compton suggests that any satisfactory or acceptable solution of either of these two problems would have to comply with the following postulate which I shall call Compton's postulate of freedom: the solution must explain freedom; and it must also explain how freedom is not just chance but, rather, the result of a subtle interplay between something almost random or haphazard, and something like a restrictive or selective control-such as an aim or a standard- though certainly not a cast-iron control. For it is clear that the controls which guided Compton back from Italy allowed him plenty of freedom: freedom, say, to choose between an American and a French or Italian boat; or freedom to postpone his lecture, had some more important obligation arisen.

We may say that Compton's postulate of freedom restricts the acceptable solutions of our two problems by demanding that they should conform to the idea of combining freedom and control, and also to the idea of a 'plastic control', as I shall call it in contradistinction to a 'cast-iron control'.

Compton's postulate is a restriction which I accept gladly and freely; and my own free and deliberate though not uncritical acceptance of this restriction may be taken as an illustration of that combination of freedom and control which is the very content of Compton's postulate of freedom.


I HAVE explained our two central problems - Compton's problem and Descartes's problem. In order to solve them we need, I believe, a new theory; in fact, a new theory of evolution, and a new model of the organism.

This need arises because the existing indeterministic theories are unsatisfactory. They are indeterministic; but we know that indeterminism is not enough, and it is not clear how they escape from Schlick's objection, or whether they conform to Compton's postulate of freedom plus control. Again, Compton's problem is quite beyond them: they are hardly relevant to it. And although these theories are attempts to solve Descartes's problem, the solutions they propose do not appear to be satisfactory.

The theories I am alluding to may be called 'master-switch models of control' or, more briefly, 'master-switch theories'. Their underlying idea is that our body is a kind of machine which can be regulated by a lever or switch from one or more central control points. Descartes even went so far as to locate the control point precisely: it is in the pineal gland, he said, that mind acts upon body. Some quantum theorists suggested (and 6. Of Clouds and Clocks 233 Compton very tentatively accepted the suggestion) that our minds work upon our bodies by influencing or selecting some quantum jumps. These are then amplified by our central nervous system which acts like an electronic amplifier: the amplified quantum jumps operate a cascade of relays or masterswitches and ultimately effect muscular contractions. There are, I think, some indications in Compton's books that he did not much like this particular theory or model, and that he used it for one purpose only: to show that human indeterminism (or even `freedom') does not necessarily contradict quantum physics. I think he was right in all this, including his dislike of master-switch theories.

For these master-switch theories-whether the one of Descartes, or the amplifier theories of the quantum physicistsbelong to what I may perhaps call 'tiny baby theories'. They seem to me to be almost as unattractive as tiny babies.

I am sure you all know the story of the unmarried mother who pleaded: 'But it is only a very tiny one.' Descartes's pleading seems to me similar: `But it is such a tiny one: it is only an unextended mathematical point in which our mind may act upon our body.'

The quantum theorists hold a very similar tiny baby theory: `But it is only with one quantum jump, and just within the Heisenberg uncertainties-and these are very tiny indeed-that a mind can act upon a physical system.' I admit that there is perhaps a slight advance here, in so far as the size of the baby is specified. But I still do not love the baby.

For however tiny the master-switch may be, the masterswitch- cum-amplifier model strongly suggests that all our decisions are either snap-decisions (as I have called them in section x above) or else composed of snap-decisions. Now I admit that amplifier mechanisms are important characteristics of biological systems (for the energy of the reaction, released or triggered by a biological stimulus, usually exceeds greatly the energy of the triggering stimulus ;46 and I also admit, of course, that snapdecisions do occur. But they differ markedly from the kind of decision which Compton had in mind: they are almost like reflexes, and thus conform neither to the situation of Compton's problem of the influence of the universe of meanings upon our behaviour, nor to Compton's postulate of freedom (nor to the idea of a `plastic' control). Decisions which conform to all this are as a rule reached almost imperceptibly through lengthy deliberation. They are reached by a kind of maturing process which is not well represented by the master-switch model.

By considering this process ofdeliberation, we may get another hint for our new theory. For deliberation always works by trial and error or, more precisely, by the method of trial and of errorelimination: by tentatively proposing various possibilities, and eliminating those which do not seem adequate. This suggests that we might use in our new theory some mechanism of trial and error-elimination.

I shall now outline how I intend to proceed.

Before formulating my evolutionary theory in general terms I shall first show how it works in a particular case, by applying it to our first problem, that is, to Compton's problem of the influence of meaning upon behaviour.

After having in this way solved. Compton's problem, I shall formulate the theory in a general way. Then it will be found that it also contains-within the framework of our new, theory which creates a new problem-situation - a straightforward and almost trivial answer to Descartes 's classical body-mind problem.


LET us now approach our first problem-that is, Compton's problem of the influence of meaning upon behaviour-by way of some comments on the evolution of languages from animal languages to human languages.

Animal languages and human languages have many things in common, but there are also differences: as we all know, human languages do somehow transcend animal languages. Using and extending some ideas of my late teacher Karl Buhler47 I shall distinguish two functions which animal and human languages share, and two functions which human language alone possesses; or in other words, two lower functions, and two higher ones which evolved on the basis of the lower functions.

The two lower functions of language are these. First, language, like all other forms of behaviour, consists of symptoms or expressions; it is symptomatic or expressive of the state of the organism which makes the linguistic signs. Following Buhler, I call this the symptomatic or expressive function of language.

Secondly, for language or communication to take place, there must not only be a sign-making organism or a 'sender', but also a reacting one, a 'receiver'. The symptomatic expression of the first organism, the sender, releases or evokes or stimulates or triggers a reaction in the second organism, which responds to the sender's behaviour, thereby turning it into a signal. This function of language to act upon a receiver was called by Buhler the releasing or signalling function of language.

To take an example, a bird may be ready to fly away, and may express this by exhibiting certain symptoms. These may then release or trigger a certain response or reaction in a second bird, and as a consequence it too may get ready to fly away.

Note that the two functions, the expressive function and the release function, are distinct; for it is possible that instances of the first may occur without the second, though not the other way round: a bird may express by its behaviour that it is ready to fly away without thereby influencing another bird. So the first function may occur without the second; which shows that they can be disentangled in spite of the fact that, in any genuine instance of communication by language, they always occur together.

These two lower functions, the symptomatic or expressive function on the one hand, and the releasing or signalling function on the other, are common to the languages of animals and men; and these two lower functions are always present when any of the higher functions (which are characteristically human) are present.

For human language is very much richer. It has many functions, and dimensions, which animal languages do not have. Two of these new functions are most important for the evolution of reasoning and rationality: the descriptive function, and the argumentative function.

As an example of the descriptive function, I might now describe to you how two days ago a magnolia was flowering in my garden, and what happened when snow began to fall. I might thereby express my feelings, and also release or trigger some feeling in you: you may perhaps react by thinking of your magnolia trees. So the two lower functions would be present. But in addition to all this, I should have described to you some facts; I should have made some descriptive statements; and these statements of mine would be factually true, or factually false.

Whenever I speak I cannot help expressing myself; and if you listen to me you can hardly help reacting. So the lower functions are always present. The descriptive function need not be present, for I may speak to you without describing any fact. For example, in showing or expressing uneasiness-say, doubt about whether you will survive this long lecture-I need not describe anything. Yet description, including the description of conjectured states of affairs, which we formulate in the form of theories or hypotheses, is clearly an extremely important function of human language; and it is that function which distinguishes human language most clearly from the various animal languages (although there seems to be something approaching it in the language of the bees) . It is, of course, a function which is indispensable for science.

The last and highest of the four functions to be mentioned in this survey is the argumentative funnction of language, as it may be seen at work, in its highest form of development, in a welldisciplined critical discussion.

The argumentative function of language is not only the highest of the four functions I am here discussing, but it was also the latest of them to evolve. Its evolution has been closely connected with that of an argumentative, critical, and rational attitude; and since this attitude has led to the evolution of science, we may say that the argumentative function of language has created what is perhaps the most powerful tool for biological adaptation which has ever emerged in the course of organic evolution.

Like the other functions, the art of critical argument has developed by the method of trial and error-elimination, and it has had the most decisive influence on the human ability to think rationally. (Formal logic itself may be described as an 'organon of critical argument'. Like the descriptive use of language, the argumentative use has led to the evolution of ideal standards of control, or of 'regulative ideas' (using a Kantian term) : the main regulative idea of the descriptive use of language is truth (as distinct from falsity) ; and that of the argumentative use of language, in critical discussion, is validity (as distinct from invalidity).

Arguments, as a rule, are for or against some proposition or descriptive statement; this is why our fourth function - the argumentative function - must have emerged later than the descriptive function.
Even if I argue in a committee that the University ought not to authorize a certain expenditure because we cannot afford it, or because some alternative way of using the money would be more beneficial, I am arguing not only for or against a proposal but also for and,against some proposition - for the proposition, say, that the proposed use will not be beneficial, and against the proposition that the proposed use will be beneficial. So arguments, even arguments about proposals, as a rule bear on propositions, and very often on descriptive propositions.

Yet the argumentative use of language may be clearly distinguished from its descriptive use, simply because I can describe without arguing: I can describe, that is to say, without giving reasons for or against the truth of my description.

Our analysis of four functions of our language - the expressive, the signalling, the descriptive, and the argumentative functions - may be summed up by saying that, although it must be admitted that the two lower functions -the expressive and signalling functions-are always present whenever the higher functions are present, we must nevertheless distinguish the higher functions from the lower ones. Yet many behaviourists and many philosophers have overlooked the higher functions, apparently because the lower ones are always present, whether or not the higher ones are.


APART from the new functions of language which have evolved and emerged together with man, and with human rationality, we must consider another distinction of almost equal importance, the distinction between the evolution of organs and that of tools or machines, a distinction to be credited to one of the greatest of English philosophers, Samuel Butler, the author of Erewhon (1872).

Animal evolution proceeds largely, though not exclusively, by the modification of organs (or behaviour) or the emergence of new organs (or behaviour). Human evolution proceeds, largely, by developing new organs outside our bodies or persons: `exosomatically', as biologists call it, or `extra -personally'. These new organs are tools, or weapons, or machines, or houses.

The rudimentary beginnings of this exosomatic development can of course be found among animals. The making of lairs, or dens, or nests, is an early achievement. I may also remind you that beavers build very ingenious dams. But man, instead of growing better eyes and ears, grows spectacles, microscopes, telescopes, telephones, and hearing aids. And instead of growing swifter and swifter legs, he grows swifter and swifter motor cars. Yet the kind of extra-personal or exosomatic evolution which interests me here is this: instead of growing better memories and brains, we grow paper, pens, pencils, typewriters, dictaphones, the printing press, and libraries.

These add to our language-and especially to its descriptive and argumentative functions-what may be described as new dimensions. The latest development (used mainly in support of our argumentative abilities) is the growth of computers.


How are the higher functions and dimensions related to the lower ones? They do not replace the lower ones, as we have seen, but they establish a kind of plastic control over them-a control with feed-back.

Take, for example, a discussion at a scientific conference. It may be exciting and enjoyable, and give rise to expressions and symptoms of its being so; and these expressions in their turn may release similar symptoms in other participants. Yet there is no doubt that up to a point these symptoms and releasing signals will be due to, and controlled by, the scientific content of the discussion; and since this will be of a descriptive and of an argumentative nature, the lower functions will be controlled by the higher ones. Moreover, though a good joke or a pleasant grin may let the lower functions win in the short run, what counts in the long run is a good argument-a valid argument -and what it establishes or refutes. In other words, our discussion is controlled, though plastically, by the regulative ideas of truth and of validity.

All this is strengthened by the discovery and development of the new dimensions of printing and publishing, especially when these are used for printing and publishing scientific theories and hypotheses, and papers in which these are critically discussed.

I cannot do justice to the importance of critical arguments here: it is a topic on which I have written fairly extensively, and so I shall not raise it again here. I only wish to stress that critical arguments are a means of control: they are a means of eliminating errors, a means of selection. We solve our problems by tentatively proposing various competing theories and hypotheses, as trial balloons, as it were; and by submitting them to critical discussion and to empirical tests, for the purpose of error-elimination.

So the evolution of the higher functions of language which I have tried to describe may be characterized as the evolution of new means for problem-solving, by new kinds of trials, and by new methods of error-elimination; that is to say, new methods for controlling the trials.


I CAN now give my solution to our first main problem, that is, Compton's problem of the influence ofineaning upon behaviour. It is this.

The higher levels of language have evolved under the pressure of a need for the better control of two things: of our lower levels of language, and our adaptation to the environment, by the method of growing not only new tools, but also, for example, new scientific theories, and new standards of selection.

Now in developing its higher functions, our language has also grown abstract meanings and contents; that is to say, we have learned how to abstract from the various modes of formulating or expressing a theory, and how to pay attention to its invariant content or meaning (upon which its truth depends). And this holds not only for theories and other descriptive statements, but also for proposals, or aims, or whatever else may be submitted to critical discussion.

What I have called `Compton's problem' was the problem of explaining and understanding the controlling power of meanings, such as the contents of our theories, or of purposes, or aims; purposes or aims which in some cases we may have adopted after deliberation and discussion. But this is now no longer a problem. Their power of influencing us is part and parcel of these contents and meanings; for part of the function of contents and meanings is to control.

This solution of Compton's problem conforms to Compton's restricting postulate. For the control of ourselves and of our actions by our theories and purposes is a plastic control. We are not forced to submit ourselves to the control of our theories, for we can discuss them critically, and we can reject them freely if we think that they fall short of our regulative standards. So the control is far from one-sided. Not only do our theories control us, but we can control our theories (and even our standards) : there is a kind offeed-back here. And if we submit to our theories, then we do so freely, after deliberation; that is, after the critical discussion of alternatives, and after freely choosing between the competing theories, in the light of that critical discussion.

I submit this as my solution of Compton's problem; and before proceeding to solve Descartes's problem, I shall now briefly outline the more general theory of evolution which I have already used, implicitly, in my solution.


I OFFER my general theory with many apologies. It has taken me a long time to think it out fully, and to make it clear to myself. Nevertheless I still feel far from satisfied with it. This is partly due to the fact that it is an evolutionary theory, and one which adds only a little, I fear, to existing evolutionary theories, except perhaps a new emphasis.

I blush when I have to make this confession; for when I was younger I used to say very contemptuous things about evolutionary philosophies. When twenty-two years ago Canon Charles E. Raven, in his Science, Religion and the Future, described the Darwinian controversy as 'a storm in a Victorian teacup', I agreed, but criticized him for paying too much attention 'to the vapours still emerging from the cup', by which I meant the hot air of the evolutionary philosophies (especially those which told us that there were inexorable laws of evolution). But now I have to confess that this cup of tea has become, after all, my cup of tea; and with it I have to eat humble pie.

Quite apart from evolutionary philosophies, the trouble about evolutionary theory is its tautological, or almost tautological, character: the difficulty is that Darwinism and natural selection, though extremely important, explain evolution by 'the survival of the fittest' (a term due to Herbert Spencer). Yet there does not seem to be much difference, if any, between the assertion 'those that survive are the fittest' and the tautology 'those that survive are those that survive'. For we have, I am afraid, no other criterion of fitness than actual survival, so that we conclude from the fact that some organisms have survived that they were the fittest, or those best adapted to the conditions of life.

This shows that Darwinism, with all its great virtues, is by no means a perfect theory. It is in need of a restatement which makes it less vague. The evolutionary theory which I am going to sketch here is an attempt at such a restatement.

My theory may be described as an attempt to apply to the whole of evolution what we learned when we analysed the evolution from animal language to human language. And it consists of a certain view of evolution as a growing hierarchical system of plastic controls, and of a certain view of organisms as incorporating-or in the case of man, evolving exosomatically - this growing hierarchical system of plastic controls. The Neo-Darwinist theory of evolution is assumed; but it is restated by pointing out that its 'mutations' may be interpreted as more or less accidental trial-and-error gambits, and `natural selection' as one way of controlling them by error-elimination.

I shall now state the theory in the form of twelve short theses:

(i) All organisms are constantly, day and night, engaged in problem-solving; and so are all those evolutionary sequences of organisms-the phyla which begin with the most primitive forms and of which the now living organisms are the latest members.

(2) These problems are problems in an objective sense: they can be, hypothetically, reconstructed by hindsight, as it were. (I will say more about this later.) Objective problems in this sense need not have their conscious counterpart; and where they have their conscious counterpart, the conscious problem need not coincide with the objective problem.

(3) Problem-solving always proceeds by the method of trial and error: new reactions, new forms, new organs, new modes of behaviour, new hypotheses, are tentatively put forward and controlled by error-elimination.

(4) Error-elimination may proceed either by the complete elimination of unsuccessful forms (the killing-off of unsuccessful forms by natural selection) or by the (tentative) evolution of controls which modify or suppress unsuccessful organs, or forms of behaviour, or hypotheses.

(5) The single organism telescopes into one body, as it were, the controls developed during the evolution of its phylum-just as it partly recapitulates, in its ontogenetic development, its phylogenetic evolution.

(6) The single organism is a kind of spearhead of the evolutionary sequence of organisms to which it belongs (its phylum) : it is itself a tentative solution, probing into new environmental niches, choosing an environment and modifying it. It is thus related to its phylum almost exactly as the actions (behaviour) of the individual organism are related to this organism: the individual organism, and its behaviour, are both trials, which may be eliminated by error-elimination.

(7) Using 'P' for problem, 'TS' for tentative solutions, 'EE' for error-elimination, we can describe the fundamental evolutionary sequence of events as follows:

P-> TS -> EE -> P.

But this sequence is not a cycle: the second problem is, in general, different from the first: it is the result of the new situation which has arisen, in part, because of the tentative solutions which have been tried out, and the error-elimination which controls them. In order to indicate this, the above schema should be rewritten:

P.-> TS -> EE -> New Problem(s).

(8) But even in this form an important element is still missing: the multiplicity of the tentative solutions, the multiplicity of the trials. Thus our final schema becomes something like this:

Background Knowledge -> Problem -> Many TSs -> EE -> New Problem(s)

(9) In this form, our schema can be compared with that of Neo-Darwinism. According to Neo-Darwinism there is in the main one problem: the problem of survival. There is, as in our system, a multiplicity of tentative solutions-the variations or mutations. But there is only one way of error-elimination-the killing of the organism. And (partly for this reason) the fact that Pi and P2 will differ essentially is overlooked, or else its fundamental importance is not sufficiently clearly realized.

(10) In our system, not all problems are survival problems: there are many very specific problems and sub-problems (even though the earliest problems may have been sheer survival problems). For example an early problem Pi may be reproduc-- tion. Its solution may lead to a new problem, P2: the problem of getting rid of, or of spreading, the offspring-the children which threaten to suffocate not only the parent organism but each other.53 It is perhaps of interest to note that the problem of avoiding suffocation by one's of spring may be one of those problems which was solved by the evolution of multicellular organisms: instead of getting rid of one's offspring, one establishes a common economy, with various new methods of living together.

(11) The theory here proposed distinguishes between Pi and P2i and it shows that the problems (or the problem situations) which the organism is trying to deal with are often new, and arise themselves as products of the evolution. The theory thereby gives implicitly a rational account of what has usually been called by the somewhat dubious names of `creative evolution' or `emergent evolution'.54

(12) Our schema allows for the development of erroreliminating controls (warning organs like the eye; feed-back mechanisms) ; that is, controls which can eliminate errors without killing the organism; and it makes it possible, ultimately, for our hypotheses to die in our stead.


EACH organism can be regarded as a , hierarchical system of plastic controls-as a system of clouds controlled by clouds. The controlled subsystems make trial-and-error movements which are partly suppressed and partly restrained by the controlling system.

We have already met an example of this in the relation between the lower and higher functions of language. The lower ones continue to exist and to play their part; but they are constrained and controlled by the higher ones.

Another characteristic example is this. If I am standing quietly, without making any movement, then (according to the physiologists) my muscles are constantly at work, contracting and relaxing in an almost random fashion (see TSI to TS,, in thesis (8) of the preceding section), but controlled, without my being aware of it, by error-elimination (EE) so that every little deviation from my posture is almost at once corrected. So I am kept standing, quietly, by more or less the same method by which an automatic pilot keeps an aircraft steadily on its course.

This example also illustrates the thesis (i) of the preceding section-that each organism is all the time engaged in problemsolving by trial and error; that it reacts to new and old problems by more or less chance-like,55 or cloud-like, trials which are eliminated if unsuccessful. (If successful, they increase the probability of the survival of mutations which `simulate' the solutions so reached, and tend to make the solution hereditary,56 by incorporating it into the spatial structure or form of the new organism.) 

The method of trial and error-elimination does not operate with completely chancelike or random trials (as has been sometimes suggested), even though the trials may look pretty random; there must be at least an `after-effect' (in the sense of my The Logic of Scientific Discovery, pp. 162 ff.). For the organism is constantly learning from its mistakes, that is, it establishes controls which suppress or eliminate, or at least reduce the frequency of, certain possible trials (which were perhaps actual ones in its evolutionary past).


THIS is a very brief outline of the theory. It needs, of course, much elaboration. But I wish to explain one point a little more fully-the use I have made (in theses (I) to (3) of section xvin) of the terms `problem' and `problem-solving' and, more particularly, my assertion that we can speak of problems in an objective, or non-psychological sense.

The point is important, for evolution is clearly not a conscious process. Many biologists say that the evolution of certain organs solves certain problems; for example, that the evolution of the eye solves the problem of giving a moving animal a timely warning to change its direction before bumping into something hard. Nobody suggests that this kind of solution to this kind of problem is consciously sought. Is it not, then, just a metaphor if we speak of problem-solving?

I do not think so; rather, the situation is this: when we speak of a problem, we do so almost always from hindsight. A man who works on a problem can seldom say clearly what his problem is (unless he has found a solution) ; and even if he can explain his problem, he may mistake it. And this may even hold of scientists-though scientists are among those few who consciously try to be fully aware of their problems. For example, Kepler's conscious problem was to discover the harmony of the world order; but we may say that the problem he solved was the mathematical description of motion in a set of two-body planetary systems. Similarly, Schrodinger was mistaken about the problem he had solved by finding the (time-independent) Schrodinger equation: he thought his waves were chargedensity waves, of a changing continuous field of electric charge. Later Max Born gave a statistical interpretation of the Schrodinger wave amplitude; an interpretation which shocked Schrodinger and which he disliked as long as he lived. He had solved a problem-but it was not the one he thought he had solved. This we know now, by hindsight.

Yet clearly it is in science that we are most conscious of the problems we try to solve. So it should not be inappropriate to use hindsight in other cases, and to say that the amoeba solves some problems (though we need not assume that it is in any sense aware of its problems) : from the amoeba to Einstein is just one step.


BUT Compton tells us that the amoeba's actions are not rational, while we may assume that Einstein's actions are. So there should be some difference, after all.

I admit that there is a difference: even though their methods of almost random or cloud-like trial and error movements are fundamentally not very different,58, 55 there is a great difference in their attitudes towards error. Einstein, unlike the amoeba, consciously tried his best, whenever a new solution occurred to him, to fault it and detect an error in it: he approached his own solutions critically.

I believe that this consciously critical attitude towards his own ideas is the one really important difference between the method of Einstein and that of the amoeba. It made it possible for Einstein to reject, quickly, hundreds of hypotheses as inadequate before examining one or another hypothesis more carefully, if it appeared to be able to stand up to more serious criticism.

As the physicist John Archibald Wheeler said recently, 'Our whole problem is to make the mistakes as fast as possible'. This problem of Wheeler's is solved by consciously adopting the critical attitude. This, I believe, is the highest form so far of the rational attitude, or of rationality.

The scientist's trials and errors consist of hypotheses. He formulates them in words, and often in writing. He can then try to find flaws in any one of these hypotheses, by criticizing it, and by testing it experimentally, helped by his fellow scientists who will be delighted if they can find a flaw in it. If the hypothesis does not stand up to these criticisms and to these tests at least as well as-its competitors,60 it will be eliminated.

It is different with primitive man, and with the amoeba. Here there is no critical attitude, and so it happens more often than not that natural selection eliminates a mistaken hypothesis or expectation by eliminating those organisms which hold it, or believe in it. So we can say that the critical or rational method consists in letting our hypotheses die in our stead: it is a case of exosomatic evolution.


HERE I may perhaps turn to a question which has given me much trouble although in the end I arrived at a very simple solution.

The question is: Can we show that plastic controls exist? Are there inorganic physical systems in nature which may be taken as examples or as physical models of plastic controls?

It seems that this question was implicitly answered in the negative by many physicists who, like Descartes or Compton, operate with master-switch models, and by many philosophers who, like Hume or Schlick, deny that anything intermediate between complete determinism and pure chance can exist. Admittedly, cyberneticists and computer engineers have more recently succeeded in constructing computers made of hardware but incorporating highly plastic controls; for example, computers with built-in mechanism for chance-like trials, checked or evaluated by feed-back (in the manner of an automatic pilot or a self-homing device) and eliminated if erroneous. But these systems, although incorporating what I have called plastic controls, consist essentially of complex relays of master-switches. What I was seeking, however, was a simple physical model of Peircean indeterminism; a purely physical system resembling a very cloudy cloud in heat motion, controlled by some other cloudy clouds-though by somewhat less cloudy ones.

If we return to our old arrangement of clouds and clocks, with a cloud on the left and a clock on the right, then we could say that what we are looking for is something intermediate, like an organism or like our cloud of gnats, but not alive: a pure physical system, controlled plastically and 'softly', as it were.

Let us assume that the cloud to be controlled is a gas. Then we can put on the extreme left an uncontrolled gas which will soon diffuse and so cease to constitute a physical system. We put on the extreme right an iron cylinder filled with gas: this is our example of a `hard' control, a `cast-iron' control. In between, but far to the left, are many more or less 'softly' controlled systems, such as our cluster of gnats, and huge balls of particles, such as a gas kept together by gravity, somewhat like the sun. (We do not mind if the control is far from perfect, and many particles escape.) The planets may perhaps be said to be castiron controlled in their movements-comparatively speaking, of course, for even the planetary system is a cloud, and so are all the milky ways, star clusters, and clusters of clusters. But are there, apart from organic systems and those huge systems of particles, examples of any 'softly' controlled small physical systems?

I think there are, and I propose to put in the middle of our diagram a child's balloon or, perhaps better, a soap bubble; and this, indeed, turns out to be a very primitive and in many respects an excellent example or model of a Peircean system and of a 'soft' kind of plastic control.

The soap bubble consists of two subsystems which are both clouds and which control each other: without the air, the soapy film would collapse, and we should have only a drop of soapy water. Without the soapy film, the air would be uncontrolled: it would diffuse, ceasing to exist as a system. Thus the control is mutual; it is plastic, and of a feed-back character. Yet it is possible to make a distinction between the controlled system (the air) and the controlling systems (the film) : the enclosed air is not only more cloudy than the enclosing film, but it also ceases to be a physical (self-interacting) system if the film is removed. As against this, the film, after removal of the air, will form- a droplet which, though of a different shape, may still be said to be a physical system.

Comparing the bubble with a `hardware' system like a precision clock or a computer, we should of course say (in accordance with Peirce's point of view) that even these hardware systems are clouds controlled by clouds. But these `hard' systems are built with the purpose of minimizing, so far as it is possible, the cloud-like effects of molecular heat motions and fluctuations: though they are clouds, the controlling mechanisms are designed to suppress, or compensate for, all cloud-like effects as far as possible. This holds even for computers with mechanisms simulating chance-like trial-and-error mechanisms.

Our soap bubble is different in this respect and, it seems, more similar to an organism: the molecular effects are not eliminated but contribute essentially to the working of the system which is enclosed by a skin-a permeable wall61 that leaves the system `open', and able to `react' to environmental influences in a manner which is built, as it were, into its `organization': the soap bubble, when struck by a heat ray, absorbs the heat (much like a hot-house), and so the enclosed air will expand, keeping the bubble floating.

As in all uses of similarity or analogy we should, however, look out for limitations; and here we might point out that, at least in some organisms, molecular fluctuations are apparently amplified and so used to release trial-and-error movements. At any rate, amplifiers seem to play important roles in all organisms (which in this respect resemble some computers with their master-switches and cascades of amplifiers and relays). Yet there are no amplifiers in the soap bubble.

However this may be, our bubble shows that natural physical cloud-like systems which are plastically and softly controlled by other cloud-like systems do exist. (Incidentally, the film of the bubble need not, of course, be derived from organic matter, though it will have to contain large molecules.)


THE evolutionary theory here proposed yields an immediate solution to our second main problem-the classical Cartesian body-mind problem. It does so (without saying what `mind' or `consciousness' is) by saying something about the evolution, and thereby about the functions, of mind or consciousness.

We must assume that consciousness grows from small beginnings; perhaps its first form is a vague feeling of irritation, experienced when the organism has a problem to solve such as getting away from an irritant substance. However this may be, consciousness will assume evolutionary significance-and increasing significance-when it begins to anticipate possible ways of reacting: possible trial -and-error movements, and their possible outcomes.

We can say now that conscious states, or sequences of conscious states, may function as systems of control, of errorelimination: the elimination, as a rule, of (incipient) behaviour, that is (incipient) movement. Consciousness, from this point of view, appears as just one of many interacting kinds of control; and if we remember the control systems incorporated for example in books-theories, systems of law, and all that constitutes the 'universe of meanings'-then consciousness can hardly be said to be the highest control system in the hierarchy. For it is to a considerable extent controlled by these exosomatic linguistic systems-even though they may be said to be produced by consciousness. Consciousness in turn is, we may conjecture, produced by physical states; yet it controls them to a considerable extent. Just as a legal or social system is produced by us, yet controls us, and is in no reasonable sense `identical' to or `parallel' with us, but interacts with us, so states of consciousness (the `mind') control the body, and interact with it.

Thus there is a whole set of analogous relationships. As our exosomatic world of meanings is related to consciousness, so consciousness is related to the behaviour of the acting individual organism. And the behaviour of the individual organism is similarly related to its body, to the individual organism taken as a physiological system. The latter is similarly related to the evolutionary sequence of organisms-the phylum of which it forms the latest spearhead, as it were: as the individual organism is thrown up experimentally as a probe by the phylum and yet largely controls the fate of the phylum, so the behaviour of the organism is thrown up experimentally as a probe by the physiological system and yet controls, largely, the fate of this system. Our conscious states are similarly related to our behaviour. They anticipate our behaviour, working out, by trial and error, its likely consequences; thus they not only control but they try out, deliberate.

We now see that this theory offers us an almost trivial answer to Descartes's problem. Without saying what `the mind' is, it leads immediately to the conclusion that our mental states control (some of) our physical movements, and that there is some give-and-take, some feed-back, and so some interaction, between mental activity and the other functions of the organism.

The control will again be of the `plastic' kind; in fact all of us -especially those who play a musical instrument such as the piano or the violin-know that the body does not always do what we want it to do; and that we have to learn, from our ill-success, how to modify our aims, making allowances for those limitations which beset our control: though we are free, to some considerable extent, there are always conditions-physical or otherwise-which set limits to what we can do. (Of course, before giving in, we are free to try to transcend these limits.)

Thus, like Descartes, I propose the adoption of a dualistic outlook, though I do not of course recommend talking of two kinds of interacting substances. But I think it is helpful and legitimate to distinguish two kinds of interacting states (or events), physio-chemical and mental ones. Moreover, I suggest that if we distinguish only these two kinds of states we still take too narrow a view of our world: at the very least we should also distinguish those artifacts which are products of organisms, and especially the products of our minds, and which can interact with our minds and thus with the state of our physical environment. Although these artifacts are often 'mere bits of matter', `mere tools' perhaps, they are even on the animal level sometimes consummate works of art; and on the human level, the products of our minds are often very much more than 'bits of matter'-marked bits of paper, say; for these bits of paper may represent states of a discussion, states of the growth of knowledge, which may transcend (sometimes with serious consequences) the grasp of most or even all of the minds that helped to produce them. Thus we have to be not merely dualists, but pluralists; and we have to recognize that the great changes which we have brought about, often unconsciously, in our physical universe show that abstract rules and abstract ideas, some of which are perhaps only partially grasped by human minds, may move mountains.


As an afterthought, I should like to add one last point.

It would be a mistake to think that, because of natural selection, evolution can only lead to what may be called `utilitarian' results: to adaptations which are useful in helping us to survive.

Just as, in a system with plastic controls, the controlling and controlled subsystems interact, so our tentative solutions interact with our problems and also with our aims. This means that our aims can change and that the choice of an aim may become a problem; different aims may compete, and new aims may be invented and controlled by the method of trial and error-elimination.

Admittedly, if a new aim clashes with the aim of surviving, then this new aim may be eliminated by natural selection. It is well known that many mutations are lethal and thus suicidal; and there are many examples of suicidal aims. Others are perhaps neutral with respect to survival.

Many aims that at first are subsidiary to survival may later become autonomous, and even opposed to survival; for example, the ambition to excel in courage, to climb Mount Everest, to discover a new continent, or to be the first on the Moon; or the ambition to discover some new truth.

Other aims may from the very beginning be autonomous departures, independent of the aim to survive. Artistic aims are perhaps of this kind, or some religious aims, and to those who cherish them they may become much more important than survival.

All this is part of the superabundance of life-the almost excessive abundance of trials and errors upon which the method of trial and error-elimination depends.

It is perhaps not uninteresting to see that artists, like scientists, use this trial-and-error method. a painter may put down, tentatively, a speck of colour and step back for a critical assessment of its efect in order to alter it if it does not solve the problem he wants to solve. And it may happen that an unexpected or accidental effect of his tentative trial - a colour speck or brush stroke-may change his problem, or create a new subproblem, or a new aim: the evolution of artistic aims and of artistic standards (which, like the rules of logic, may become exosomatic systems of control) proceeds also by the trial-and-error method.

We may perhaps here look back for a moment to the problem of physical determinism, and to our example of the deaf physicist who had never experienced music but would be able to 'compose' a Mozart opera or a Beethoven symphony, simply by studying Mozart's or Beethoven's bodies and their environments as physical systems, and predicting where their pens would put down black marks on lined paper. I presented these as unacceptable consequences of physical determinism. Mozart and Beethoven are, partly, controlled by their `taste', their system of musical evaluation. Yet this system is not cast iron but rather plastic. It responds to new ideas, and it can be modified by new trials and errors-perhaps even by an accidental mistake, an unintended discord.

In conclusion, let me sum up the situation.
We have seen that it is unsatisfactory to look upon the world as a closed physical system-whether a strictly deterministic system or a system in which whatever is not strictly determined is simply due to chance: on such a view of the world human creativeness and human freedom can only be illusions. The attempt to make use of quantum-theoretical indeterminacy is also unsatisfactory, because it leads to chance rather than freedom, and to snap-decisions rather than deliberate decisions.

I have therefore offered here a different view of the worldone in which the physical world is an open system. This is compatible with the view of the evolution of life as a process of trial and error-elimination; and it allows us to understand rationally, though far from fully, the emergence of biological novelty and the growth of human knowledge and human freedom.

I have tried to outline an evolutionary theory which takes account of all this and which offers solutions to Compton's and Descartes's problems. It is, I am afraid, a theory which manages to be too humdrum and too speculative at the same time; and even though I think that testable consequences can be derived from it, I am far from suggesting that my proposed solution is what philosophers have been looking for. But I feel that Compton might have said that it presents, in spite of its faults, a possible answer to his problem-and one which might lead to further advance.  

PDF  Part 1PDF Part 2

Rathouse IndexFull IndexPopper articles