Climbing Mount Improbable

by Richard Dawkins

  To return to the point about ‘chance’, Lord Kelvin used the prestigious platform of his Presidential Address to the British Association to quote, with approval, the words of another distinguished physical scientist, Sir John Herschel, who also, by the way, referred to Darwinism as ‘The Law of Higgledy-Piggledy’:

  We can no more accept the principle of arbitrary and casual variation and natural selection as a sufficient account, per se, of the past and present organic world, than we can receive the Laputan method of composing books (pushed à l’outrance) as a sufficient one of Shakespeare and the Principia.

  Herschel's allusion was to Gulliver's Travels in which Swift had mocked the Laputan method of writing books by combining words at random. Herschel and Kelvin, Hoyle and Wickramasinghe, my {76} anonymously quoted physical scientists and any number of Jehovah's Witness tracts all make the mistake of treating Darwinian natural selection as though it were tantamount to Laputan authorship. To this day, and in quarters where they should know better, Darwinism is widely regarded as a theory of ‘chance’.

  It is grindingly, creakingly, crashingly obvious that, if Darwinism were really a theory of chance, it couldn't work. You don't need to be a mathematician or physicist to calculate that an eye or a haemoglobin molecule would take from here to infinity to self-assemble by sheer higgledy-piggledy luck. Far from being a difficulty peculiar to Darwinism, the astronomic improbability of eyes and knees, enzymes and elbow joints and the other living wonders is precisely the problem that any theory of life must solve, and that Darwinism uniquely does solve. It solves it by breaking the improbability up into small, manageable parts, smearing out the luck needed, going round the back of Mount Improbable and crawling up the gentle slopes, inch by million-year inch. Only God would essay the mad task of leaping up the precipice in a single bound. And if we postulate him as our cosmic designer we are left in exactly the same position as when we started. Any Designer capable of constructing the dazzling array of living things would have to be intelligent and complicated beyond all imagining. And complicated is just another word for improbable — and therefore demanding of explanation. A theologian who ripostes that his god is sublimely simple has (not very) neatly evaded the issue, for a sufficiently simple god, whatever other virtues he might have, would be too simple to be capable of designing a universe (to say nothing of forgiving sins, answering prayers, blessing unions, transubstantiating wine, and the many other achievements variously expected of him). You cannot have it both ways. Either your god is capable of designing worlds and doing all the other godlike things, in which case he needs an explanation in his own right. Or he is not, in which case he cannot provide an explanation. God should be seen by Fred Hoyle as the ultimate Boeing 747.

  The height of Mount Improbable stands for the combination of perfection and improbability that is epitomized in eyes and enzyme molecules (and gods capable of designing them). To say that an {77} object like an eye or a protein molecule is improbable means something father precise. The object is made of a large number of parts arranged in a very special way. The number of possible ways in which those parts could have been arranged is exceedingly large. In the case of a protein molecule we can actually calculate that large number. Isaac Asimov did it for the particular protein haemoglobin, and called it the Haemoglobin Number. It has 190 noughts. That is the number of ways of rearranging the bits of haemoglobin such that the result would not be haemoglobin. In the case of the eye we can't do the equivalent calculation without fabricating lots of assumptions, but we can intuitively see that it is going to come to another stupefyingly large number. The actual, observed arrangement of parts is improbable in the sense that it is only one arrangement among trillions of possible arrangements.

  Now, there is an uninteresting sense in which, with hindsight, any particular arrangement of parts is just as improbable as any other. Even a junkyard is as improbable, with hindsight, as a 747, for its parts could have been arranged in so many other ways. The trouble is, most of those ways would also be junkyards. This is where the idea of quality comes in. The vast majority of arrangements of the parts of a Boeing junkyard would not fly. A small minority would. Of all the trillions of possible arrangements of the parts of an eye, only a tiny minority would see. The human eye forms a sharp image on a retina, corrected for spherical and chromatic aberration; automatically stops down or up with an iris diaphragm to keep the internal light intensity relatively constant in the face of large fluctuations in external light intensity; automatically changes the focal length of the lens depending upon whether the object being looked at is near or far; sorts out colour by comparing the firing rates of three different kinds of light-sensitive cell. Almost all random scramblings of the parts of an eye would fail to achieve any of these delicate and difficult tasks. There is something very special about the particular arrangement that exists. All particular arrangements are as improbable as each other. But of all particular arrangements, those that aren't useful hugely outnumber those that are. Useful devices are improbable and need a special explanation. {78}

  R. A. Fisher, the great mathematical geneticist and founder of the modern science of statistics, put the point in 1930, in his usual meticulous style (I never met him, but one can almost hear his fastidiously correct dictation to his long-suffering wife):

  An organism is regarded adapted partlcular situation, or to the totality of situations which constitute its environment, only in so far as we can imagine an assemblage of slightly different situations, or environments, to which the animal would on the whole be less well adapted; and equally only in so far as we can imagine an assemblage of slightly different organic forms, which would be less well adapted to that environment.

  Eyes, ears and hearts, the wing of a vulture, the web of a spider, these all impress us by their obvious perfection of engineering no matter where we see them: we dont need to have them presented to us in their natural surroundings to see that they are good for some purpose and that, if their parts were rearranged or altered in almost any way, they would be worse. They have ‘improbable perfection’ written all over them. An engineer can recognize them as the kind of thing that he would design, if called upon to solve a particular problem.

  This is another way of saying that objects such as these cannot be explained as coming into existence by chance. As we have seen, to invoke chance, on its own, as an explanation, is equivalent to vaulting from the bottom to the top of Mount Improbable's steepest cliff in one bound. And what corresponds to inching up the kindly, grassy llopes on the other side of the mountain? It is the slow, cumulative, one-step-at-a-time, non-random survival of random variants that Darwin called natural selection. The metaphor of Mount Improbable dramatizes the mistake of the sceptics quoted at the beginning of this chapter. Where they went wrong was to keep their eyes fixed on the vertical precipice and its dramatic height. They assumed that the ilteer cliff was the only way up to the summit on which are perched tyes and protein molecules and other supremely improbable arrangements of parts. It was Darwin's great achievement to discover the gentle gradients winding up the other side of the mountain. {79}

  But is this one of those rare cases where it is really true that there is no smoke without fire? Darwinism is widely misunderstood as a theory of pure chance. Mustn't it have done something to provoke this canard? Well, yes, there is something behind the misunderstood rumour, a feeble basis to the distortion. One stage in the Darwinian process is indeed a chance process — mutation. Mutation is the process by which fresh genetic variation is offered up for selection and it is usually described as random. But Darwinians make the fuss that they do about the ‘randomness’ of mutation only in order to contrast it to the non-randomness of selection, the other side of the process. It is not necessary that mutation should be random in order for natural selection to work. Selection can still do its work whether mutation is directed or not. Emphasizing that mutation can be random is our way. of calling attention to the crucial fact that, by contrast, selection is subli
mely and quintessentially non-random. It is ironic that this emphasis on the contrast between mutation and the non-randomness of selection has led people to think that the whole theory is a theory of chance.

  Even mutations are, as a matter of fact, non-random in various senses, although these senses aren't relevant to our discussion because they don't contribute constructively to the improbable perfection of organisms. For example, mutations have well-understood physical causes, and to this extent they are non-random. The reason X-ray machine operators take a step back before pressing the trigger, or wear lead aprons, is that X-rays cause mutations. Mutations are also more likely to occur in some genes than in others. There are ‘hot spots’ on chromosomes where mutation rates are markedly higher than the average. This is another kind of non-randomness. Mutations can be reversed (‘back mutations’). For most genes, mutation in either direction is equally probable. For some, mutation in one direction is more frequent than back mutation in the reverse direction. This gives rise to so-called ‘mutation pressure’ — a tendency to evolve in a particular direction regardless of selection. This is yet another sense in which mutation can be described as non-random. Notice that mutation pressure does not systematically drive in the direction of improvement. Nor do X-rays. Quite the contrary: the great majority of {80} mutations, however caused, are random with respect to quality, and that means they are usually bad because there are more ways of getting worse than of getting better.

  One could imagine a theoretical world in which mutations were biased towards improvement. Mutations in this hypothetical world would be non-random not just in the sense that mutations induced by X-rays are non-random: these hypothetical mutations would be systematically biased to keep one jump ahead of selection and anticipate the needs of the organism. But this is the one kind of non-randomness which, contrary to numerous theoretical yearnings, almost certainly has no basis in fact: mutations are not systematically likely to anticipate the needs of the organism, nor is it clear how such anticipation could possibly work. What might ‘anticipation mean? Suppose a terrible ice age is closing in on a previously temperate region and the local deer are perishing in their lightweight coats. Most individuals will die anyway but the species will be saved if only, in the nick of time, it can evolve a thick coat like a musk-ox. It is in principle possible to imagine a mechanism tuned to switch on desirable mutations as and when they are needed. We know that X-rays increase the general mutation rate, indiscriminately making thinner coats or thicker coats. What if intense cold could somehow increase the mutation rate in one direction only: towards thicker coats? And, symmetrically, what if intense heat could induce mutations in the other direction, towards thinner coats?

  Darwinians wouldn't mind if such providential mutations were provided. It wouldn't undermine Darwinism, though it would put paid to its claims for exclusivity: a tailwind on a transatlantic flight can speed up your arrival in an agreeable way, and this doesn't undermine your belief that the primary force that got you home is the jet engine. But Darwinians would actually be pretty surprised (as well as intrigued) if any such beneficent mutational mechanism were discovered, for three reasons.

  First, despite energetic searching, no such mechanism has yet been discovered (at least in animals and plants: there is a very special and not generally relevant suggestion of a case in bacteria in which the j&cts are still controversial). Second, there is no existing theory that {81} could explain how the body might ‘know’ which sort of mutation to induce. I suppose one could imagine that, if there have been dozens of ice-age cycles stretching back over the millions of years, constituting a form of ‘race experience’, some as yet undiscovered kind of higher order natural selection could have built in a propensity to mutate in the right direction at the first inklings of the next ice age. But I repeat that there is no evidence for any such effect and, moreover, no theory so far worked out can handle it. Third — and this returns to my earlier point — some Darwinians, including me, find the proposed mechanism of directed mutation inelegantly superfluous. This is a largely aesthetic reaction which should therefore not be treated as overwhelming. But if we do react unsympathetically towards suggestions of directed mutation, it is because such suggestions are often made by people who mistakenly think that such a theory is needed: who don't understand that selection, on its own, is amply powerful enough to do the job, even if mutation is random. One way to dramatize the adequacy of non-random selection is to emphasize that mutation is allowed by the theory to be random. But, as I said before, it is not critical to the theory that mutation must be random, and it most certainly provides no excuse to tar the whole theory with the brush of randomness. Mutation may be random, but selection definitely is not.

  Before we leave our deer out there in the deepening cold, there is one variant of the theory of providential mutation that might have occurred to you as you read the last three paragraphs. It may indeed be hard to see how the body could ‘know’ that cold weather requires mutation towards a thicker coat, while hot weather requires mutation in the other direction. But it is slightly easier to imagine that mutation rates might be pre-programmed to inflate, indiscriminately in all directions, during times when the going gets tough. The intuitive rationale would be something like this. A new crisis, like an ice age or an age of intense heat, is felt by the body as stress. High stress on me, whether from cold, heat, drought or any cause unspecified, indicates that something is wrong with my bodily equipment for present conditions. It may be too late for me, but perhaps life could be improved for some of my children if I mutate the genes in my sex organs wildly, in all sorts of random directions. {82} Whatever the environmental crisis may be (cold, heat, drought, flood), those of my mutant children who contain mutations that turn out to be in the wrong direction (probably the majority) will die. But they would have died anyway if the crisis were sufficiently severe. Perhaps by producing a brood of mutant freaks and monsters an animal increases its chances of producing one child that is better at coping with the new crisis than it is itself.

  There are, indeed, genes whose effect is to control the mutation rate in other genes. In theory one might argue that these ‘mutator genes’ could be triggered by stress, and such a tendency could be favoured by some sort of high-level natural selection. But alas, this theory turns out to have no more support than our previous theory of beneficently directed mutation. First, there is no evidence for it. More seriously, there are profound theoretical difficulties with any view of increased mutation rates that has them being positively favoured by natural selection. The argument is a general one which leads to the conclusion that mutator genes will always tend to disappear from the population, and this will apply to our hypothetical stressed animals.

  Briefly, the general argument is as follows. Any animal that has succeeded in reaching an age to become a parent must already be pretty good. If you start with something pretty good and change it at random, the chances are that you'll make it worse. And, as a matter of fact, the great majority of mutations do make things worse. It is true that a small minority of mutations may make things better — that's ultimately why evolution by natural selection is possible at all. It is also true that a mutator gene, by increasing the total mutation rate, can help its possessor to come up with that precious rarity, a mutation that is an improvement. When this happens, that particular copy of the mutator gene itself will temporarily flourish, for it will share a body with the improved mutation that it has helped to create. You might think that this constitutes positive natural selection in favour pf the mutator gene and therefore, by this means, mutation rates iould increase. Alas, mark the sequel.

  In future generations, sexual reproduction will do its shuffling , rearranging and recombining the genes that share individual {83} bodies. As the generations pass, there is nothing to stop the mutator gene becoming detached from the good gene that it created: some individuals will be born with only the good gene, others with only the mutator. The good gene itself will go on being rewarded by
natural selection and may go from strength to strength in future populations. But the unfortunate mutator gene that created it has been cast aside by sexual reshuffling. Like any other gene, the mutator gene's long-term fate depends on its average effects: its effects averaged over all the different bodies in which it finds itself, over the long term. The average effects of the good gene that the mutator created are good, and the good gene will survive in more and more bodies in the population. But the average effects of the mutator itself are bad and, in spite of its occasional flashes of benefit, on average the mutator is bound to be penalized by natural selection. Most of the bodies in which it finds itself will be freaks or dead.

  This argument against the possibility of mutator genes being positively selected depends upon the assumption that reproduction is sexual. If reproduction is asexual, the ‘shuffling’ step of the argument is missing. Mutator genes can be favoured by natural selection over long periods because, when there is no sex, they are not detached from the occasional good genes that they create, and they can ‘hitchhike’ down the generations on the backs of good genes. When reproduction is asexual, a new good mutation will initiate a new clone of prospering individuals. A new bad mutation will quickly disappear, dragging its subclone of freaks down with it. If a good mutation is sufficiently good, the clone will continue to prosper, and all the genes in it will cash in — even the bad ones. The bad ones prosper because, in spite of their ill-effects, the average quality of the genes in the clone is positive. And among these prospering coat-tail-hangers will be the mutator gene responsible for creating the good mutation in the first place. As far as the good mutation is concerned, it will ‘wish’ it could shake off the dead weight of bad genes, and the mutator that created it is no exception. The good mutation, if it could think, would yearn for some good, cleansing sexual reproduction. If only my bodies would go in for some sex, it would say, I could shake off this crowd of ill-favoured hitch-hikers. I could be valued for my own virtues {84} alone. Some of the bodies I found myself would be bad, others good, but on average I'd be free to benefit from my own good effects. The bad genes, on the other hand, have no ‘desire’ for sexual reproduction: they are on to a good thing. If they had to go it alone in the genetic free-for-all which is sex, they'd soon go under.