The Blind Watchmaker

by Richard Dawkins

  Real-life & Caricature (in unison).

  Rank mysticism! Get back in the last Caricature century where you belong.

  I hope I am not being presumptuous when I take it that the reader’s sympathies are with neither the Mutationist nor with the caricature of a Darwinian. I assume that the reader agrees with the real-life Darwinian, as, of course, do I. The caricature does not really exist. Unfortunately some people think he exists, and think that, since they disagree with him, they are disagreeing with Darwinism itself. There is a school of biologists who have taken to saying something like the following. The trouble with Darwinism is that it neglects the constraints imposed by embryology. Darwinians (this is where the caricature comes in) think that, if selection would favour some conceivable evolutionary change, then the necessary mutational variation will turn out to be available. Mutational change in any direction is equally likely: selection provides the only bias.

  But any real-life Darwinian would acknowledge that, although any gene on any chromosome may mutate at any time, the consequences of mutation on bodies are severely limited by the processes of embryology. If I ever doubted this (I didn’t), my doubts would have been dispelled by my biomorph computer simulations. You can’t just postulate a mutation ‘for’ sprouting wings in the middle of the back. Wings, or anything else, can only evolve if the process of development allows them to. Nothing magically ‘sprouts’. It has to be made by the processes of embryonic development. Only a minority of the things that conceivably could evolve are actually permitted by the status quo of existing developmental processes. Because of the way arms develop, it is possible for mutations to increase the length of fingers and cause webs of skin to grow between them. But there may not be anything in the embryology of backs that lends itself to ‘sprouting’ angel wings. Genes can mutate till they are blue in the face, but no mammal will ever sprout wings like an angel unless mammalian embryological processes are susceptible to this kind of change.

  Now as long as we don’t know all the ins and outs of how embryos develop, there is room for disagreement over how likely it is that particular imagined mutations have or have not ever existed. It might turn out, for instance, that there is nothing in mammalian embryology to forbid angel wings, and that the caricature Darwinian was right, in this particular case, to suggest that angel wing-buds arose but were not favoured by selection. Or it might turn out that when we know more about embryology we shall see that angel wings were always a non-starter, and that therefore selection never had a chance to favour them. There is a third possibility, which we should list for completeness, that embryology never allowed the possibility of angel wings and that selection would never have favoured them even if it had. But what we must insist on is that we can’t afford to ignore the constraints on evolution that embryology imposes. All serious Darwinians would agree about this, yet some people portray Darwinians as denying it. It turns out that people who make a lot of noise about ‘developmental constraints’ as an alleged anti-Darwinian force are confusing Darwinism with the caricature of Darwinism that I parodied above.

  This all began with a discussion over what is meant when we say that mutation is ‘random’. I listed three respects in which mutation is not random: it is induced by X-rays, etc.; mutation rates are different for different genes; and forward mutation rates do not have to equal backward mutation rates. To this, we have now added a fourth respect in which mutation is not random. Mutation is non-random in the sense that it can only make alterations to existing processes of embryonic development. It cannot conjure, out of thin air, any conceivable change that selection might favour. The variation that is available for selection is constrained by the processes of embryology, as they actually exist.

  There is a fifth respect in which mutation might have been nonrandom. We can imagine (just) a form of mutation that was systematically biased in the direction of improving the animal’s adaptedness to its life. But although we can imagine it, nobody has ever come close to suggesting any means by which this bias could come about. It is only in this fifth respect, the ‘mutationist’ respect, that the true, real-life Darwinian insists that mutation is random. Mutation is not systematically biased in the direction of adaptive improvement, and no mechanism is known (to put the point mildly) that could guide mutation in directions that are non-random in this fifth sense. Mutation is random with respect to adaptive advantage, although it is non-random in all sorts of other respects. It is selection, and only selection, that directs evolution in directions that are non-random with respect to advantage. Mutationism is not just wrong in fact. It never could have been right. It is not in principle capable of explaining the evolution of improvement. Mutationism belongs with Lamarckism, not as a disproved rival to Darwinism but as no rival at all.

  The same is true of my next alleged rival to Darwinian selection, championed by the Cambridge geneticist Gabriel Dover under the odd name ‘molecular drive’ (since everything is made of molecules it is not obvious why Dover’s hypothetical process should deserve the name molecular drive any more than any other evolutionary process; it reminds me of a man I knew who complained of a gastric stomach, and worked things out using his mental brain). Motoo Kimura and the other proponents of the neutralist theory of evolution do not, as we saw, make any false claims for their theory. They have no illusions about random drift being a rival to natural selection as an explanation for adaptive evolution. They recognize that only natural selection can drive evolution in adaptive directions. Their claim is simply that a lot of evolutionary change (as a molecular geneticist sees evolutionary change) is not adaptive. Dover makes no such modest claims for his theory. He thinks that he can explain all of evolution without natural selection, although he generously concedes that there may be some truth in natural selection as well!

  Throughout this book, our first recourse when considering such matters has been to the example of the eye, although it has, of course, been only a representative of the large set of organs that are too complex and well designed to have come about by chance. Only natural selection, I have repeatedly argued, even comes close to offering a plausible explanation for the human eye and comparable organs of extreme perfection and complexity. Fortunately, Dover has explicitly risen to the challenge, and has offered his own explanation of the evolution of the eye. Assume, he says, that 1,000 steps of evolution are needed to evolve the eye from nothing. This means that a sequence of 1,000 genetic changes were needed to transform a bare patch of skin into an eye. This seems to me to be an acceptable assumption for the sake of argument. In the terms of Biomorph Land, it means that the bare-skin animal is 1,000 genetic steps distant from the eyed animal.

  Now, how do we account for the fact that just the right set of 1,000 steps were taken to result in the eye as we know it? Natural selection’s explanation is well known. Reducing it to its simplest form, at each one of the 1,000 steps, mutation offered a number of alternatives, only one of which was favoured because it aided survival. The 1,000 steps of evolution represent 1,000 successive choice points, at each of which most of the alternatives led to death. The adaptive complexity of the modern eye is the end-product of 1,000 successful unconscious ‘choices’. The species has followed a particular path through the labyrinth of all possibilities. There were 1,000 branch-points along the path, and at each one the survivors were the ones that happened to take the turning that led to improved eyesight. The wayside is littered with the dead bodies of the failures who took the wrong turning at each one of the 1,000 successive choice points. The eye that we know is the end-product of a sequence of 1,000 successful selective ‘choices’.

  That was (one way of expressing) natural selection’s explanation of the evolution of the eye in 1,000 steps. Now, what of Dover’s explanation? Basically, he argues that it wouldn’t have mattered which choice the lineage took at each step: it would retrospectively have found a use for the organ that resulted. Each step that the lineage took, according to him, was a random step. At Step 1, for example, a random
mutation spread through the species. Since the newly evolved characteristic was functionally random, it didn’t aid the animals’ survival. So the species searched the world for a new place or new way of life in which they could use this new random feature that had been imposed upon their bodies. Having found a piece of environment that suited the random part of their bodies, they lived there for a while, until a new random mutation arose and spread through the species. Now the species had to scour the world for a new place or way of life where they could live with their new random bit. When they found it, Step 2 was completed. Now the Step 3 random mutation spread through the species, and so on for 1,000 steps, at the end of which the eye as we know it had been formed. Dover points out that the human eye happens to use what we call ‘visible’ light rather than infrared. But if random processes had happened to impose an infrared sensitive eye upon us, we would, doubtless, have made the most of it, and found a way of life that exploited infrared rays to the full.

  At first glance this idea has a certain seductive plausibility, but only at a very brief first glance. The seductiveness comes from the neatly symmetrical way in which natural selection is turned on its head. Natural selection, in its most simple form, assumes that the environment is imposed upon the species, and those genetic variants best fitted to that environment survive. The environment is imposed, and the species evolves to fit it. Dover’s theory turns this on its head. It is the nature of the species that is ‘imposed’, in this case by the vicissitudes of mutation, and other internal genetic forces in which he has a special interest. The species then locates that member of the set of all environments that best fits its imposed nature.

  But the seductiveness of the symmetry is superficial indeed. The wondrous cloud-cuckooism of Dover’s idea is displayed in all its glory the moment we begin to think in terms of numbers. The essence of his scheme is that, at each of the 1,000 steps, it didn’t matter which way the species turned. Each new innovation that the species came up with was functionally random, and the species then found an environment to suit it. The implication is that the species would have found a suitable environment, no matter which branch it had taken at every fork in the way. Now just think how many possible environments this lets us in for postulating. There were 1,000 branch points. If each branch point was a mere bifurcation (as opposed to a 3-way or 18-way branch, a conservative assumption), the total number of livable environments that must, in principle, exist, in order to allow Dover’s scheme to work, is 2 to the power 1,000 (the first branch gives two pathways; then each of those branches into two, making four in all, then each of these branches, giving 8; then 16, 32, 64, … all the way to 21,000). This number may be written as a 1 with 301 noughts after it. It is far far greater than the total number of atoms in the entire universe.

  Dover’s alleged rival to natural selection could never work, not just never in a million years but never in a million times longer than the universe has existed, never in a million universes each lasting a million times as long again. Notice that this conclusion is not materially affected if we change Dover’s initial assumption that 1,000 steps would be needed to make an eye. If we reduce it to only 100 steps, which is probably an underestimate, we still conclude that the set of possible livable environments that must be waiting in the wings, as it were, to cope with whatever random steps the lineage might take, is more than a million million million million million. This is a smaller number than the previous one, but it still means that the vast majority of Dover’s ‘environments’ waiting in the wings would each have to be made of less than a single atom.

  It is worth explaining why the theory of natural selection is not susceptible to a symmetrical destruction by a version of the ‘largenumbers argument’. In Chapter 3 we thought of all real and conceivable animals as sitting in a gigantic hyperspace. We are doing a similar thing here, but simplifying it by considering evolutionary branch points as 2-way, rather than 18-way branches. So the set of all possible animals that might have evolved in 1,000 evolutionary steps are perched on a gigantic tree, which branches and branches so that the total number of final twigs is 1 followed by 301 noughts. Any actual evolutionary history can be represented as a particular pathway through this hypothetical tree. Of all conceivable evolutionary pathways, only a minority actually ever happened. We can think of most of this ‘tree of all possible animals’ as hidden in the darkness of non-existence. Here and there, a few trajectories through the darkened tree are illuminated. These are the evolutionary pathways that actually happened, and, numerous as these illuminated branches are, they are still an infinitesimal minority of the set of all branches. Natural selection is a process that is capable of picking its way through the tree of all conceivable animals, and finding just that minority of pathways that are viable. The theory of natural selection cannot be attacked by the kind of large-numbers argument with which I attacked Dover’s theory, because it is of the essence of the theory of natural selection that it is continually cutting down most of the branches of the tree. That is precisely what natural selection does. It picks its way, step by step, through the tree of all conceivable animals, avoiding the almost infinitely large majority of sterile branches — animals with eyes in the soles of their feet, etc. — which the Dover theory is obliged, by the nature of its peculiar inverted logic, to countenance.

  We have dealt with all the alleged alternatives to the theory of natural selection except the oldest one. This is the theory that life was created, or its evolution master-minded, by a conscious designer. It would obviously be unfairly easy to demolish some particular version of this theory such as the one (or it may be two) spelled out in Genesis. Nearly all peoples have developed their own creation myth, and the Genesis story is just the one that happened to have been adopted by one particular tribe of Middle Eastern herders. It has no more special status than the belief of a particular West African tribe that the world was created from the excrement of ants. All these myths have in common that they depend upon the deliberate intentions of some kind of supernatural being.

  At first sight there is an important distinction to be made between what might be called ‘instantaneous creation’ and ‘guided evolution’. Modern theologians of any sophistication have given up believing in instantaneous creation. The evidence for some sort of evolution has become too overwhelming. But many theologians who call themselves evolutionists, for instance the Bishop of Birmingham quoted in Chapter 2, smuggle God in by the back door: they allow him some sort of supervisory role over the course that evolution has taken, either influencing key moments in evolutionary history (especially, of course, human evolutionary history), or even meddling more comprehensively in the day-to-day events that add up to evolutionary change.

  We cannot disprove beliefs like these, especially if it is assumed that God took care that his interventions always closely mimicked what would be expected from evolution by natural selection. All that we can say about such beliefs is, firstly, that they are superfluous and, secondly, that they assume the existence of the main thing we want to explain, namely organized complexity. The one thing that makes evolution such a neat theory is that it explains how organized complexity can arise out of primeval simplicity.

  If we want to postulate a deity capable of engineering all the organized complexity in the world, either instantaneously or by guiding evolution, that deity must already have been vastly complex in the first place. The creationist, whether a naive Bible-thumper or an educated bishop, simply postulates an already existing being of prodigious intelligence and complexity. If we are going to allow ourselves the luxury of postulating organized complexity without offering an explanation, we might as well make a job of it and simply postulate the existence of life as we know it! In short, divine creation, whether instantaneous or in the form of guided evolution, joins the list of other theories we have considered in this chapter. All give some superficial appearance of being alternatives to Darwinism, whose merits might be tested by an appeal to evidence. All turn out, on closer
inspection, not to be rivals of Darwinism at all. The theory of evolution by cumulative natural selection is the only theory we know of that is in principle capable of explaining the existence of organized complexity. Even if the evidence did not favour it, it would still be the best theory available! In fact the evidence does favour it. But that is another story.

  Let us hear the conclusion of the whole matter. The essence of life is statistical improbability on a colossal scale. Whatever is the explanation for life, therefore, it cannot be chance. The true explanation for the existence of life must embody the very antithesis of chance. The antithesis of chance is nonrandom survival, properly understood. Nonrandom survival, improperly understood, is not the antithesis of chance, it is chance itself. There is a continuum connecting these two extremes, and it is the continuum from single-step selection to cumulative selection. Single-step selection is just another way of saying pure chance. This is what I mean by nonrandom survival improperly understood. Cumulative selection, by slow and gradual degrees, is the explanation, the only workable explanation that has ever been proposed, for the existence of life’s complex design.

  The whole book has been dominated by the idea of chance, by the astronomically long odds against the spontaneous arising of order, complexity and apparent design. We have sought a way of taming chance, of drawing its fangs. ‘Untamed chance’, pure, naked chance, means ordered design springing into existence from nothing in a single leap. It would be untamed chance if once there was no eye, and then, suddenly, in the twinkling of a generation, an eye appeared, fully fashioned, perfect and whole. This is possible, but the odds against it will keep us busy writing noughts till the end of time. The same applies to the odds against the spontaneous existence of any fully fashioned, perfect and whole beings, including — I see no way of avoiding the conclusion — deities.