CONTENTS

RICHARD DAWKINS

THE BLIND
WATCHMAKER

Apple Computer, Inc. makes no warranties, either express or implied, regarding the enclosed software package, its merchantability, or its fitness for any particular purpose. The exclusion of implied warranties is not permitted by some states. The above exclusion may not apply to you. This warranty provides you with specific legal rights. There may be other rights that you may have which vary from state to state.

Macintosh System Tools® are copyrighted programs of Apple Computer, Inc. licensed to W. W. Norton to distribute for use only in combination with Blind Watchmaker, Apple Software shall not be copied onto another diskette (except for archive purposes) or into memory unless as part of the execution of Blind Watchmaker. When Blind Watchmaker has completed execution Apple Software shall not be used by any other program.

Copyright © 1996, 1987, 1986 by Richard Dawkins

Illustrations by Liz Pyle

All rights reserved.

Printed in the United States of America.

First published as a Norton paperback 1987; reissued in a new edition 1996

Library of Congress Cataloging in Publication Data

Dawkins, Richard, 1941-

The blind watchmaker.

1. Evolution         2. Natural selection.      I. Title
QH366.2.D37         1985         575      85-4960

ISBN 0-393-31570-3

W. W. Norton & Company, Inc

500 Fifth Avenue, New York, N.Y 10110

www.wwnorton.com

W. W. Norton & Company Ltd.

Castle House, 75/76 Wells Street, London WIT 3QT

ABOUT THE AUTHOR

Richard Dawkins was born in Nairobi in 1941. He was educated at Oxford University, and after graduation remained there to work for his doctorate with the Nobel Prize-winning ethologist Niko Tinbergen. From 1967 to 1969 he was an Assistant Professor of Zoology at the University of California at Berkeley. In 1970 he became a Lecturer in Zoology at Oxford University and a Fellow of New College. In 1995 he became the first Charles Simonyi Professor of the Public Understanding of Science at Oxford University.

Richard Dawkins's first book, The Selfish Gene (1976; second edition, 1989), became an immediate international bestseller and, like The Blind Watchmaker, was translated into all the major languages. Its sequel, The Extended Phenotype, followed in 1982. His other best-sellers include River Out of Eden (1995) and Climbing Mount Improbable (1996; Penguin, 1997).

Richard Dawkins won both the Royal Society of Literature Award and the Los Angeles Times Literary Prize in 1987 for The Blind Watchmaker. The television film of the book, shown in the Horizon series, won the Sci-Tech Prize for the Best Science Programme of 1987. He has also won the 1989 Silver Medal of the Zoological Society of London and the 1990 Royal Society Michael Faraday Award for the furtherance of the public understanding of science. In 1994 he won the Nakayama Prize for Human Science and has been awarded an Honorary D.Litt. by the University of St Andrews and by the Australian National University, Canberra.

PREFACE

This book is written in the conviction that our own existence once presented the greatest of all mysteries, but that it is a mystery no longer because it is solved. Darwin and Wallace solved it, though we shall continue to add footnotes to their solution for a while yet. I wrote the book because I was surprised that so many people seemed not only unaware of the elegant and beautiful solution to this deepest of problems but, incredibly, in many cases actually unaware that there was a problem in the first place!

The problem is that of complex design. The computer on which I am writing these words has an information storage capacity of about 64 kilobytes (one byte is used to hold each character of text). The computer was consciously designed and deliberately manufactured. The brain with which you are understanding my words is an array of some ten million kiloneurones. Many of these billions of nerve cells have each more than a thousand ‘electric wires’ connecting them to other neurones. Moreover, at the molecular genetic level, every single one of more than a trillion cells in the body contains about a thousand times as much precisely-coded digital information as my entire computer. The complexity of living organisms is matched by the elegant efficiency of their apparent design. If anyone doesn't agree that this amount of complex design cries out for an explanation, I give up. No, on second thoughts I don’t give up, because one of my aims in the book is to convey something of the sheer wonder of biological complexity to those whose eyes have not been opened to it. But having built up the mystery, my other main aim is to remove it again by explaining the solution.  {xiv} 

Explaining is a difficult art. You can explain something so that your reader understands the words; and you can explain something so that the reader feels it in the marrow of his bones. To do the latter, it sometimes isn't enough to lay the evidence before the reader in a dispassionate way. You have to become an advocate and use the tricks of the advocate's trade. This book is not a dispassionate scientific treatise. Other books on Darwinism are, and many of them are excellent and informative and should be read in conjunction with this one. Far from being dispassionate, it has to be confessed that in parts this book is written with a passion which, in a professional scientific journal, might excite comment. Certainly it seeks to inform, but it also seeks to persuade and even — one can specify aims without presumption — to inspire. I want to inspire the reader with a vision of our own existence as, on the face of it, a spine-chilling mystery, and simultaneously to convey the full excitement of the fact that it is a mystery with an elegant solution which is within our grasp. More, I want to persuade the reader, not just that the Darwinian world-view happens to be true, but that it is the only known theory that could, in principle, solve the mystery of our existence. This makes it a doubly satisfying theory. A good case can be made that Darwinism is true, not just on this planet but all over the universe wherever life may be found.

In one respect I plead to distance myself from professional advocates. A lawyer or a politician is paid to exercise his passion and his persuasion on behalf of a client or a cause in which he may not privately believe. I have never done this and I never shall. I may not always be right, but I care passionately about what is true and I never say anything that I do not believe to be right. I remember being shocked when visiting a university debating society to debate with creationists. At dinner after the debate, I was placed next to a young woman who had made a relatively powerful speech in favour of creationism. She clearly couldn't be a creationist, so I asked her to tell me honestly why she had done it. She freely admitted that she was simply practising her debating skills, and found it more challenging to advocate a position in which she did not believe. Apparently it is common practice in university debating societies for speakers simply to be told on which side they are to speak. Their own beliefs don’t come into it. I had come a long way to perform the disagreeable task of public speaking, because I believed in the truth of the motion that I had been asked to propose. When I discovered that members of the society were using the motion as a vehicle for playing arguing games, I resolved to decline future invitations from debating societies that encourage insincere advocacy on issues where scientific truth is at stake.  {xv} 

For reasons that are not entirely clear to me, Darwinism seems more in need of advocacy than similarly established truths in other branches of science. Many of us have no grasp of quantum theory, or Einstein's theories of special and general relativity, but this does not in itself lead us to oppose these theories! Darwinism, unlike ‘Einsteinism’, seems to be regarded as fair game for critics with any degree of ignorance. I suppose one trouble with Darwinism is that, as Jacques Monod perceptively remarked, everybody thinks he understands it. It is, indeed, a remarkably simple theory; childishly so, one would have thought, in comparison with almost all of physics and mathematics. In essence, it amounts simply to the idea that non-random reproduction, where there is hereditary variation, has consequences that are far-reaching if there is time for them to be cumulative. But we have good grounds for believing that this simplicity is deceptive. Never forget that, simple as the theory may seem, nobody thought of it until Darwin and Wallace in the mid nineteenth century, nearly 200 years after Newton's Principia, and more than 2,000 years after Eratosthenes measured the Earth. How could such a simple idea go so long undiscovered by thinkers of the calibre of Newton, Galileo, Descartes, Leibnitz, Hume and Aristotle? Why did it have to wait for two Victorian naturalists? What was wrong with philosophers and mathematicians that they overlooked it? And how can such a powerful idea go still largely unabsorbed into popular consciousness?

It is almost as if the human brain were specifically designed to misunderstand Darwinism, and to find it hard to believe. Take, for instance, the issue of ‘chance’, often dramatized as blind chance. The great majority of people that attack Darwinism leap with almost unseemly eagerness to the mistaken idea that there is nothing other than random chance in it. Since living complexity embodies the very antithesis of chance, if you think that Darwinism is tantamount to chance you'll obviously find it easy to refute Darwinism! One of my tasks will be to destroy this eagerly believed myth that Darwinism is a theory of ‘chance’. Another way in which we seem predisposed to disbelieve Darwinism is that our brains are built to deal with events on radically different timescales from those that characterize evolutionary change. We are equipped to appreciate processes that take seconds, minutes, years or, at most, decades to complete. Darwinism is a theory of cumulative processes so slow that they take between thousands and millions of decades to complete. All our intuitive judgements of what is probable turn out to be wrong by many orders of magnitude. Our well-tuned apparatus of scepticism and subjective probability-theory misfires by huge margins, because it is tuned — ironically, by evolution  {xvi}  itself — to work within a lifetime of a few decades. It requires effort of the imagination to escape from the prison of familiar timescale, an effort that I shall try to assist.

A third respect in which our brains seem predisposed to resist Darwinism stems from our great success as creative designers. Our world is dominated by feats of engineering and works of art. We are entirely accustomed to the idea that complex elegance is an indicator of premeditated, Grafted design. This is probably the most powerful reason for the belief, held by the vast majority of people that have ever lived, in some kind of supernatural deity. It took a very large leap of the imagination for Darwin and Wallace to see that, contrary to all intuition, there is another way and, once you have understood it, a far more plausible way, for complex ‘design’ to arise out of primeval simplicity. A leap of the imagination so large that, to this day, many people seem still unwilling to make it. It is the main purpose of this book to help the reader to make this leap.

Authors naturally hope that their books will have lasting rather than ephemeral impact. But any advocate, in addition to putting the timeless part of his case, must also respond to contemporary advocates of opposing, or apparently opposing, points of view. There is a risk that some of these arguments, however hotly they may rage today, will seem terribly dated in decades to come. The paradox has often been noted that the first edition of The Origin of Species makes a better case than the sixth. This is because Darwin felt obliged, in his later editions, to respond to contemporary criticisms of the first edition, criticisms which now seem so dated that the replies to them merely get in the way, and in places even mislead. Nevertheless, the temptation to ignore fashionable contemporary criticisms that one suspects of being nine days’ wonders is a temptation that should not be indulged, for reasons of courtesy not just to the critics but to their otherwise confused readers. Though I have my own private ideas on which chapters of my book will eventually prove ephemeral for this reason, the reader — and time — must judge.

I am distressed to find that some women friends (fortunately not many) treat the use of the impersonal masculine pronoun as if it showed intention to exclude them. If there were any excluding to be done (happily there isn't) I think I would sooner exclude men, but when I once tentatively tried referring to my abstract reader as ‘she’, a feminist denounced me for patronizing condescension: I ought to say ‘he-or-she’, and ‘his-or-her’. That is easy to do if you don’t care about language, but then if you don’t care about language you don’t deserve readers of either sex. Here, I have returned to the normal conventions  {xvii}  of English pronouns. I may refer to the ‘reader’ as ‘he’, but I no more think of my readers as specifically male than a French speaker thinks of a table as female. As a matter of fact I believe I do, more often than not, think of my readers as female, but that is my personal affair and I'd hate to think that such considerations impinged on how I use my native language.

Personal, too, are some of my reasons for gratitude. Those to whom I cannot do justice will understand. My publishers saw no reason to keep from me the identities of their referees (not ‘reviewers’ — true reviewers, pace many Americans under 40, criticize books only after they are published, when it is too late for the author to do anything about it), and I have benefited greatly from the suggestions of John Krebs (again), John Durant, Graham Cairns-Smith, Jeffrey Levinton, Michael Ruse, Anthony Hallam and David Pye. Richard Gregory kindly criticized Chapter 12, and the final version has benefited from its complete excision. Mark Ridley and Alan Grafen, now no longer even officially my students, are, together with Bill Hamilton, the leading lights of the group of colleagues with whom I discuss evolution and from whose ideas I benefit almost daily. They, Pamela Wells, Peter Atkins and John Dawkins have helpfully criticized various chapters for me. Sarah Bunney made numerous improvements, and John Gribbin corrected a major error. Alan Grafen and Will Atkinson advised on computing problems, and the Apple Macintosh Syndicate of the Zoology Department kindly allowed their laser printer to draw biomorphs.

Once again I have benefited from the relentless dynamism with which Michael Rodgers, now of Longman, carries all before him. He, and Mary Cunnane of Norton, skilfully applied the accelerator (to my morale) and the brake (to my sense of humour) when each was needed. Part of the book was written during a sabbatical leave kindly granted by the Department of Zoology and New College. Finally — a debt I should have acknowledged in both my previous books — the Oxford tutorial system and my many tutorial pupils in zoology over the years have helped me to practise what few skills I may have in the difficult art of explaining.

Richard Dawkins
Oxford, 1986

CHAPTER 1

EXPLAINING
THE VERY IMPROBABLE

We animals are the most complicated things in the known universe. The universe that we know, of course, is a tiny fragment of the actual universe. There may be yet more complicated objects than us on other planets, and some of them may already know about us. But this doesn't alter the point that I want to make. Complicated things, everywhere, deserve a very special kind of explanation. We want to know how they came into existence and why they are so complicated. The explanation, as I shall argue, is likely to be broadly the same for complicated things everywhere in the universe; the same for us, for chimpanzees, worms, oak trees and monsters from outer space. On the other hand, it will not be the same for what I shall call ‘simple’ things, such as rocks, clouds, rivers, galaxies and quarks. These are the stuff of physics. Chimps and dogs and bats and cockroaches and people and worms and dandelions and bacteria and galactic aliens are the stuff of biology.

The difference is one of complexity of design. Biology is the study of complicated things that give the appearance of having been designed for a purpose. Physics is the study of simple things that do not tempt us to invoke design. At first sight, man-made artefacts like computers and cars will seem to provide exceptions. They are complicated and obviously designed for a purpose, yet they are not alive, and they are made of metal and plastic rather than of flesh and blood. In this book they will be firmly treated as biological objects.

The reader's reaction to this may be to ask, ‘Yes, but are they really biological objects?’ Words are our servants, not our masters. For different purposes we find it convenient to use words in different senses. Most cookery books class lobsters as fish. Zoologists can  {2}  become quite apoplectic about this, pointing out that lobsters could with greater justice call humans fish, since fish are far closer kin to humans than they are to lobsters. And, talking of justice and lobsters, I understand that a court of law recently had to decide whether lobsters were insects or ‘animals’ (it bore upon whether people should be allowed to boil them alive). Zoologically speaking, lobsters are certainly not insects. They are animals, but then so are insects and so are we. There is little point in getting worked up about the way different people use words (although in my nonprofessional life I am quite prepared to get worked up about people who boil lobsters alive). Cooks and lawyers need to use words in their own special ways, and so do I in this book. Never mind whether cars and computers are ‘really’ biological objects. The point is that if anything of that degree of complexity were found on a planet, we should have no hesitation in concluding that life existed, or had once existed, on that planet. Machines are the direct products of living objects; they derive their complexity and design from living objects, and they are diagnostic of the existence of life on a planet. The same goes for fossils, skeletons and dead bodies.

I said that physics is the study of simple things, and this, too, may seem strange at first. Physics appears to be a complicated subject, because the ideas of physics are difficult for us to understand. Our brains were designed to understand hunting and gathering, mating and child-rearing: a world of medium-sized objects moving in three dimensions at moderate speeds. We are ill-equipped to comprehend the very small and the very large; things whose duration is measured in picoseconds or gigayears; particles that don’t have position; forces and fields that we cannot see or touch, which we know of only because they affect things that we can see or touch. We think that physics is complicated because it is hard for us to understand, and because physics books are full of difficult mathematics. But the objects that physicists study are still basically simple objects. They are clouds of gas or tiny particles, or lumps of uniform matter like crystals, with almost endlessly repeated atomic patterns. They do not, at least by biological standards, have intricate working parts. Even large physical objects like stars consist of a rather limited array of parts, more or less haphazardly arranged. The behaviour of physical, nonbiological objects is so simple that it is feasible to use existing mathematical language to describe it, which is why physics books are full of mathematics.

Physics books may be complicated, but physics books, like cars and computers, are the product of biological objects — human brains. The objects and phenomena that a physics book describes are simpler than  {3}  a single cell in the body of its author. And the author consists of trillions of those cells, many of them different from each other, organized with intricate architecture and precision-engineering into a working machine capable of writing a book (my trillions are American, like all my units: one American trillion is a million millions; an American billion is a thousand millions). Our brains are no better equipped to handle extremes of complexity than extremes of size and the other difficult extremes of physics. Nobody has yet invented the mathematics for describing the total structure and behaviour of such an object as a physicist, or even of one of his cells. What we can do is understand some of the general principles of how living things work, and why they exist at all.

This was where we came in. We wanted to know why we, and all other complicated things, exist. And we can now answer that question in general terms, even without being able to comprehend the details of the complexity itself. To take an analogy, most of us don’t understand in detail how an airliner works. Probably its builders don’t comprehend it fully either: engine specialists don’t in detail understand wings, and wing specialists understand engines only vaguely. Wing specialists don’t even understand wings with full mathematical precision: they can predict how a wing will behave in turbulent conditions, only by examining a model in a wind tunnel or a computer simulation — the sort of thing a biologist might do to understand an animal. But however incompletely we understand how an airliner works, we all understand by what general process it came into existence. It was designed by humans on drawing boards. Then other humans made the bits from the drawings, then lots more humans (with the aid of other machines designed by humans) screwed, rivetted, welded or glued the bits together, each in its right place. The process by which an airliner came into existence is not fundamentally mysterious to us, because humans built it. The systematic putting together of parts to a purposeful design is something we know and understand, for we have experienced it at first hand, even if only with our childhood Meccano or Erector set.

What about our own bodies? Each one of us is a machine, like an airliner only much more complicated. Were we designed on a drawing board too, and were our parts assembled by a skilled engineer? The answer is no. It is a surprising answer, and we have known and understood it for only a century or so. When Charles Darwin first explained the matter, many people either wouldn't or couldn't grasp it. I myself flatly refused to believe Darwin's theory when I first heard about it as a child. Almost everybody throughout history, up to the second half of the nineteenth century, has firmly believed in the opposite — the  {4}  Conscious Designer theory. Many people still do, perhaps because the true, Darwinian explanation of our own existence is still, remarkably, not a routine part of the curriculum of a general education. It is certainly very widely misunderstood.

The watchmaker of my title is borrowed from a famous treatise by the eighteenth-century theologian William Paley. His Natural Theology — or Evidences of the Existence and Attributes of the Deity Collected from the Appearances of Nature, published in 1802, is the best-known exposition of the ‘Argument from Design’, always the most influential of the arguments for the existence of a God. It is a book that I greatly admire, for in his own time its author succeeded in doing what I am struggling to do now. He had a point to make, he passionately believed in it, and he spared no effort to ram it home clearly. He had a proper reverence for the complexity of the living world, and he saw that it demands a very special kind of explanation. The only thing he got wrong — admittedly quite a big thing! — was the explanation itself. He gave the traditional religious answer to the riddle, but he articulated it more clearly and convincingly than anybody had before. The true explanation is utterly different, and it had to wait for one of the most revolutionary thinkers of all time, Charles Darwin.

Paley begins Natural Theology with a famous passage:

In crossing a heath, suppose I pitched my foot against a stone, and were asked how the stone came to be there; I might possibly answer, that, for anything I knew to the contrary, it had lain there for ever: nor would it perhaps be very easy to show the absurdity of this answer. But suppose I had found a watch upon the ground, and it should be inquired how the watch happened to be in that place; I should hardly think of the answer which I had before given, that for anything I knew, the watch might have always been there.

Paley here appreciates the difference between natural physical objects like stones, and designed and manufactured objects like watches. He goes on to expound the precision with which the cogs and springs of a watch are fashioned, and the intricacy with which they are put together. If we found an object such as a watch upon a heath, even if we didn’t know how it had come into existence, its own precision and intricacy of design would force us to conclude

that the watch must have had a maker: that there must have existed, at some time, and at some place or other, an artificer or artificers, who formed it for the purpose which we find it actually to answer, who comprehended its construction, and designed its use.  {5} 

Nobody could reasonably dissent from this conclusion, Paley insists, yet that is just what the atheist, in effect, does when he contemplates the works of nature, for:

every indication of contrivance, every manifestation of design, which existed in the watch, exists in the works of nature; with the difference, on the side of nature, of being greater or more, and that in a degree which exceeds all computation.

Paley drives his point home with beautiful and reverent descriptions of the dissected machinery of life, beginning with the human eye, a favourite example which Darwin was later to use and which will reappear throughout this book. Paley compares the eye with a designed instrument such as a telescope, and concludes that ‘there is precisely the same proof that the eye was made for vision, as there is that the telescope was made for assisting it’. The eye must have had a designer, just as the telescope had.

Paley's argument is made with passionate sincerity and is informed by the best biological scholarship of his day, but it is wrong, gloriously and utterly wrong. The analogy between telescope and eye, between watch and living organism, is false. All appearances to the contrary, the only watchmaker in nature is the blind forces of physics, albeit deployed in a very special way. A true watchmaker has foresight: he designs his cogs and springs, and plans their interconnections, with a future purpose in his mind's eye. Natural selection, the blind, unconscious, automatic process which Darwin discovered, and which we now know is the explanation for the existence and apparently purposeful form of all life, has no purpose in mind. It has no mind and no mind's eye. It does not plan for the future. It has no vision, no foresight, no sight at all. If it can be said to play the role of watchmaker in nature, it is the blind watchmaker.

I shall explain all this, and much else besides. But one thing I shall not do is belittle the wonder of the living ‘watches’ that so inspired Paley. On the contrary, I shall try to illustrate my feeling that here Paley could have gone even further. When it comes to feeling awe over living ‘watches’ I yield to nobody. I feel more in common with the Reverend William Paley than I do with the distinguished modern philosopher, a well-known atheist, with whom I once discussed the matter at dinner. I said that I could not imagine being an atheist at any time before 1859, when Darwin's Origin of Species was published. ‘What about Hume?’, replied the philosopher. ‘How did Hume explain the organized complexity of the living world?’, I asked. ‘He didn’t’, said the philosopher. ‘Why does it need any special explanation?’  {6} 

Paley knew that it needed a special explanation; Darwin knew it, and I suspect that in his heart of hearts my philosopher companion knew it too. In any case it will be my business to show it here. As for David Hume himself, it is sometimes said that that great Scottish philosopher disposed of the Argument from Design a century before Darwin. But what Hume did was criticize the logic of using apparent design in nature as positive evidence for the existence of a God. He did not offer any alternative explanation for apparent design, but left the question open. An atheist before Darwin could have said, following Hume: ‘I have no explanation for complex biological design. All I know is that God isn't a good explanation, so we must wait and hope that somebody comes up with a better one.’ I can't help feeling that such a position, though logically sound, would have left one feeling pretty unsatisfied, and that although atheism might have been logically tenable before Darwin, Darwin made it possible to be an intellectually fulfilled atheist. I like to think that Hume would agree, but some of his writings suggest that he underestimated the complexity and beauty of biological design. The boy naturalist Charles Darwin could have shown him a thing or two about that, but Hume had been dead 40 years when Darwin enrolled in Hume's university of Edinburgh.

I have talked glibly of complexity, and of apparent design, as though it were obvious what these words mean. In a sense it is obvious — most people have an intuitive idea of what complexity means. But these notions, complexity and design, are so pivotal to this book that I must try to capture a little more precisely, in words, our feeling that there is something special about complex, and apparently designed things.

So, what is a complex thing? How should we recognize it? In what sense is it true to say that a watch or an airliner or an earwig or a person is complex, but the moon is simple? The first point that might occur to us, as a necessary attribute of a complex thing, is that it has a heterogeneous structure. A pink milk pudding or blancmange is simple in the sense that, if we slice it in two, the two portions will have the same internal constitution: a blancmange is homogeneous. A car is heterogeneous: unlike a blancmange, almost any portion of the car is different from other portions. Two times half a car does not make a car. This will often amount to saying that a complex object, as opposed to a simple one, has many parts, these parts being of more than one kind.

Such heterogeneity, or ‘many-partedness’, may be a necessary condition, but it is not sufficient. Plenty of objects are many-parted and heterogeneous in internal structure, without being complex in the sense in which I want to use the term. Mont Blanc, for instance, consists of many different kinds of rock, all jumbled together in such a  {7}  way that, if you sliced the mountain anywhere, the two portions would differ from each other in their internal constitution. Mont Blanc has a heterogeneity of structure not possessed by a blancmange, but it is still not complex in the sense in which a biologist uses the term.

Let us try another tack in our quest for a definition of complexity, and make use of the mathematical idea of probability. Suppose we try out the following definition: a complex thing is something whose constituent parts are arranged in a way that is unlikely to have arisen by chance alone. To borrow an analogy from an eminent astronomer, if you take the parts of an airliner and jumble them up at random, the likelihood that you would happen to assemble a working Boeing is vanishingly small. There are billions of possible ways of putting together the bits of an airliner, and only one, or very few, of them would actually be an airliner. There are even more ways of putting together the scrambled parts of a human.

This approach to a definition of complexity is promising, but something more is still needed. There are billions of ways of throwing together the bits of Mont Blanc, it might be said, and only one of them is Mont Blanc. So what is it that makes the airliner and the human complicated, if Mont Blanc is simple? Any old jumbled collection of parts is unique and, with hindsight, is as improbable as any other. The scrap-heap at an aircraft breaker's yard is unique. No two scrap-heaps are the same. If you start throwing fragments of aeroplanes into heaps, the odds of your happening to hit upon exactly the same arrangement of junk twice are just about as low as the odds of your throwing together a working airliner. So, why don’t we say that a rubbish dump, or Mont Blanc, or the moon, is just as complex as an aeroplane or a dog, because in all these cases the arrangement of atoms is ‘improbable’?

The combination lock on my bicycle has 4,096 different positions. Every one of these is equally ‘improbable’ in the sense that, if you spin the wheels at random, every one of the 4,096 positions is equally unlikely to turn up. I can spin the wheels at random, look at whatever number is displayed and exclaim with hindsight: ‘How amazing. The odds against that number appearing are 4,096:1. A minor miracle!’ That is equivalent to regarding the particular arrangement of rocks in a mountain, or of bits of metal in a scrap-heap, as ‘complex’. But one of those 4,096 wheel positions really is interestingly unique: the combination 1207 is the only one that opens the lock. The uniqueness of 1207 has nothing to do with hindsight: it is specified in advance by the manufacturer. If you spun the wheels at random and happened to hit 1207 first time, you would be able to steal the bike, and it would seem a minor miracle. If you struck lucky on one of those multi-dialled  {9}  combination locks on bank safes, it would seem a very major miracle, for the odds against it are many millions to one, and you would be able to steal a fortune.

Now, hitting upon the lucky number that opens the bank's safe is the equivalent, in our analogy, of hurling scrap metal around at random and happening to assemble a Boeing 747. Of all the millions of unique and, with hindsight equally improbable, positions of the combination lock, only one opens the lock. Similarly, of all the millions of unique and, with hindsight equally improbable, arrangements of a heap of junk, only one (or very few) will fly. The uniqueness of the arrangement that flies, or that opens the safe, is nothing to do with hindsight. It is specified in advance. The lock-manufacturer fixed the combination, and he has told the bank manager. The ability to fly is a property of an airliner that we specify in advance. If we see a plane in the air we can be sure that it was not assembled by randomly throwing scrap metal together, because we know that the odds against a random conglomeration's being able to fly are too great.

Now, if you consider all possible ways in which the rocks of Mont Blanc could have been thrown together, it is true that only one of them would make Mont Blanc as we know it. But Mont Blanc as we know it is defined with hindsight. Any one of a very large number of ways of throwing rocks together would be labelled a mountain, and might have been named Mont Blanc. There is nothing special about the particular Mont Blanc that we know, nothing specified in advance, nothing equivalent to the plane taking off, or equivalent to the safe door swinging open and the money tumbling out.

What is the equivalent of the safe door swinging open, or the plane flying, in the case of a living body? Well, sometimes it is almost literally the same. Swallows fly. As we have seen, it isn't easy to throw together a flying machine. If you took all the cells of a swallow and put them together at random, the chance that the resulting object would fly is not, for everyday purposes, different from zero. Not all living things fly, but they do other things that are just as improbable, and just as specifiable in advance. Whales don’t fly, but they do swim, and swim about as efficiently as swallows fly. The chance that a random conglomeration of whale cells would swim, let alone swim as fast and efficiently as a whale actually does swim, is negligible.

At this point, some hawk-eyed philosopher (hawks have very acute eyes — you couldn't make a hawk's eye by throwing lenses and light-sensitive cells together at random) will start mumbling something about a circular argument. Swallows fly but they don’t swim; and whales swim but they don’t fly. It is with hindsight that we decide  {9}  whether to judge the success of our random conglomeration as a swimmer or as a flyer. Suppose we agree to judge its success as an Xer, and leave open exactly what X is until we have tried throwing cells together. The random lump of cells might turn out to be an efficient burrower like a mole or an efficient climber like a monkey. It might be very good at wind-surfing, or at clutching oily rags, or at walking in ever decreasing circles until it vanished. The list could go on and on. Or could it?

If the list really could go on and on, my hypothetical philosopher might have a point. If, no matter how randomly you threw matter around, the resulting conglomeration could often be said, with hindsight, to be good for something, then it would be true to say that I cheated over the swallow and the whale. But biologists can be much more specific than that about what would constitute being ‘good for something’. The minimum requirement for us to recognize an object as an animal or plant is that it should succeed in making a living of some sort (more precisely that it, or at least some members of its kind, should live long enough to reproduce). It is true that there are quite a number of ways of making a living — flying, swimming, swinging through the trees, and so on. But, however many ways there may be of being alive, it is certain that there are vastly more ways of being dead, or rather not alive. You may throw cells together at random, over and over again for a billion years, and not once will you get a conglomeration that flies or swims or burrows or runs, or does anything, even badly, that could remotely be construed as working to keep itself alive.

This has been quite a long, drawn-out argument, and it is time to remind ourselves of how we got into it in the first place. We were looking for a precise way to express what we mean when we refer to something as complicated. We were trying to put a finger on what it is that humans and moles and earthworms and airliners and watches have in common with each other, but not with blancmange, or Mont Blanc, or the moon. The answer we have arrived at is that complicated things have some quality, specifiable in advance, that is highly unlikely to have been acquired by random chance alone. In the case of living things, the quality that is specified in advance is, in some sense, ‘proficiency’; either proficiency in a particular ability such as flying, as an aero-engineer might admire it; or proficiency in something more general, such as the ability to stave off death, or the ability to propagate genes in reproduction.

Staving off death is a thing that you have to work at. Left to itself — and that is what it is when it dies — the body tends to revert to a state of  {10}  equilibrium with its environment. If you measure some quantity such as the temperature, the acidity, the water content or the electrical potential in a living body, you will typically find that it is markedly different from the corresponding measure in the surroundings. Our bodies, for instance, are usually hotter than our surroundings, and in cold climates they have to work hard to maintain the differential. When we die the work stops, the temperature differential starts to disappear, and we end up the same temperature as our surroundings. Not all animals work so hard to avoid coming into equilibrium with their surrounding temperature, but all animals do some comparable work. For instance, in a dry country, animals and plants work to maintain the fluid content of their cells, work against a natural tendency for water to flow from them into the dry outside world. If they fail they die. More generally, if living things didn’t work actively to prevent it, they would eventually merge into their surroundings, and cease to exist as autonomous beings. That is what happens when they die.

With the exception of artificial machines, which we have already agreed to count as honorary living things, nonliving things don’t work in this sense. They accept the forces that tend to bring them into equilibrium with their surroundings. Mont Blanc, to be sure, has existed for a long time, and probably will exist for a while yet, but it does not work to stay in existence. When rock comes to rest under the influence of gravity it just stays there. No work has to be done to keep it there. Mont Blanc exists, and it will go on existing until it wears away or an earthquake knocks it over. It doesn't take steps to repair wear and tear, or to right itself when it is knocked over, the way a living body does. It just obeys the ordinary laws of physics.

Is this to deny that living things obey the laws of physics? Certainly not. There is no reason to think that the laws of physics are violated in living matter. There is nothing supernatural, no ‘life force’ to rival the fundamental forces of physics. It is just that if you try to use the laws of physics, in a naive way, to understand the behaviour of a whole living body, you will find that you don’t get very far. The body is a complex thing with many constituent parts, and to understand its behaviour you must apply the laws of physics to its parts, not to the whole. The behaviour of the body as a whole will then emerge as a consequence of interactions of the parts.

Take the laws of motion, for instance. If you throw a dead bird into the air it will describe a graceful parabola, exactly as physics books say it should, then come to rest on the ground and stay there. It behaves as a solid body of a particular mass and wind resistance ought to behave.  {11}  But if you throw a live bird in the air it will not describe a parabola and come to rest on the ground. It will fly away, and may not touch land this side of the county boundary. The reason is that it has muscles which work to resist gravity and other physical forces bearing upon the whole body. The laws of physics are being obeyed within every cell of the muscles. The result is that the muscles move the wings in such a way that the bird stays aloft. The bird is not violating the law of gravity. It is constantly being pulled downwards by gravity, but its wings are performing active work — obeying laws of physics within its muscles — to keep it aloft in spite of the force of gravity. We shall think that it defies a physical law if we are naive enough to treat it simply as a structureless lump of matter with a certain mass and wind resistance. It is only when we remember that it has many internal parts, all obeying laws of physics at their own level, that we understand the behaviour of the whole body. This is not, of course, a peculiarity of living things. It applies to all man-made machines, and potentially applies to any complex, many-parted object.

This brings me to the final topic that I want to discuss in this rather philosophical chapter, the problem of what we mean by explanation. We have seen what we are going to mean by a complex thing. But what kind of explanation will satisfy us if we wonder how a complicated machine, or living body, works? The answer is the one that we arrived at in the previous paragraph. If we wish to understand how a machine or living body works, we look to its component parts and ask how they interact with each other. If there is a complex thing that we do not yet understand, we can come to understand it in terms of simpler parts that we do already understand.

If I ask an engineer how a steam engine works, I have a pretty fair idea of the general kind of answer that would satisfy me. Like Julian Huxley I should definitely not be impressed if the engineer said it was propelled by ‘force locomotif’. And if he started boring on about the whole being greater than the sum of its parts, I would interrupt him: ‘Never mind about that, tell me how it works’. What I would want to hear is something about how the parts of an engine interact with each other to produce the behaviour of the whole engine. I would initially be prepared to accept an explanation in terms of quite large subcomponents, whose own internal structure and behaviour might be quite complicated and, as yet, unexplained. The units of an initially satisfying explanation could have names like fire-box, boiler, cylinder, piston, steam governor. The engineer would assert, without explanation initially, what each of these units does. I would accept this for the moment, without asking how each unit does its own particular  {12}  thing. Given that the units each do their particular thing, I can then understand how they interact to make the whole engine move.

Of course, I am then at liberty to ask how each part works. Having previously accepted the fact that the steam governor regulates the flow of steam, and having used this fact in my understanding of the behaviour of the whole engine, I now turn my curiosity on the steam governor itself. I now want to understand how it achieves its own behaviour, in terms of its own internal parts. There is a hierarchy of subcomponents within components. We explain the behaviour of a component at any given level, in terms of interactions between subcomponents whose own internal organization, for the moment, is taken for granted. We peel our way down the hierarchy, until we reach units so simple that, for everyday purposes, we no longer feel the need to ask questions about them. Rightly or wrongly for instance, most of us are happy about the properties of rigid rods of iron, and we are prepared to use them as units of explanation of more complex machines that contain them.

Physicists, of course, do not take iron rods for granted. They ask why they are rigid, and they continue the hierarchical peeling for several more layers yet, down to fundamental particles and quarks. But life is too short for most of us to follow them. For any given level of complex organization, satisfying explanations may normally be attained if we peel the hierarchy down one or two layers from our starting layer, but not more. The behaviour of a motor car is explained in terms of cylinders, carburettors and sparking plugs. It is true that each one of these components rests atop a pyramid of explanations at lower levels. But if you asked me how a motor car worked you would think me somewhat pompous if I answered in terms of Newton's laws and the laws of thermodynamics, and downright obscurantist if I answered in terms of fundamental particles. It is doubtless true that at bottom the behaviour of a motor car is to be explained in terms of interactions between fundamental particles. But it is much more useful to explain it in terms of interactions between pistons, cylinders and sparking plugs.

The behaviour of a computer can be explained in terms of interactions between semiconductor electronic gates, and the behaviour of these, in turn, is explained by physicists at yet lower levels. But, for most purposes, you would in practice be wasting your time if you tried to understand the behaviour of the whole computer at either of those levels. There are too many electronic gates and too many interconnections between them. A satisfying explanation has to be in terms of a manageably small number of interactions. This is why, if we want to  {13}  understand the workings of computers, we prefer a preliminary explanation in terms of about half a dozen major subcomponents — memory, processing mill, backing store, control unit, input-output handler, etc. Having grasped the interactions between the half-dozen major components, we then may wish to ask questions about the internal organization of these major components. Only specialist engineers are likely to go down to the level of AND gates and NOR gates, and only physicists will go down further, to the level of how electrons behave in a semiconducting medium.

For those that like ‘-ism’ sorts of names, the aptest name for my approach to understanding how things work is probably ‘hierarchical reductionism’. If you read trendy intellectual magazines, you may have noticed that ‘reductionism’ is one of those things, like sin, that is only mentioned by people who are against it. To call oneself a reductionist will sound, in some circles, a bit like admitting to eating babies. But, just as nobody actually eats babies, so nobody is really a reductionist in any sense worth being against. The nonexistent reductionist — the sort that everybody is against, but who exists only in their imaginations tries to explain complicated things directly in terms of the smallest parts, even, in some extreme versions of the myth, as the sum of the parts! The hierarchical reductionist, on the other hand, explains a complex entity at any particular level in the hierarchy of organization, in terms of entities only one level down the hierarchy; entities which, themselves, are likely to be complex enough to need further reducing to their own component parts; and so on. It goes without saying — though the mythical, baby-eating reductionist is reputed to deny this — that the kinds of explanations which are suitable at high levels in the hierarchy are quite different from the kinds of explanations which are suitable at lower levels. This was the point of explaining cars in terms of carburettors rather than quarks. But the hierarchical reductionist believes that carburettors are explained in terms of smaller units..., which are explained in terms of smaller units..., which are ultimately explained in terms of the smallest of fundamental particles. Reductionism, in this sense, is just another name for an honest desire to understand how things work.

We began this section by asking what kind of explanation for complicated things would satisfy us. We have just considered the question from the point of view of mechanism: how does it work? We concluded that the behaviour of a complicated thing should be explained in terms of interactions between its component parts, considered as successive layers of an orderly hierarchy. But another kind of question is how the complicated thing came into existence in the first place. This is the  {14}  question that this whole book is particularly concerned with, so I won't say much more about it here. I shall just mention that the same general principle applies as for understanding mechanism. A complicated thing is one whose existence we do not feel inclined to take for granted, because it is too ‘improbable’. It could not have come into existence in a single act of chance. We shall explain its coming into existence as a consequence of gradual, cumulative, step-by-step transformations from simpler things, from primordial objects sufficiently simple to have come into being by chance. Just as ‘big-step reductionism’ cannot work as an explanation of mechanism, and must be replaced by a series of small step-by-step peelings down through the hierarchy, so we can't explain a complex thing as originating in a single step. We must again resort to a series of small steps, this time arranged sequentially in time.

In his beautifully written book, The Creation, the Oxford physical chemist Peter Atkins begins:

I shall take your mind on a journey. It is a journey of comprehension, taking us to the edge of space, time, and understanding. On it I shall argue that there is nothing that cannot be understood, that there is nothing that cannot be explained, and that everything is extraordinarily simple... A great deal of the universe does not need any explanation. Elephants, for instance. Once molecules have learnt to compete and to create other molecules in their own image, elephants, and things resembling elephants, will in due course be found roaming through the countryside.

Atkins assumes the evolution of complex things — the subject matter of this book — to be inevitable once the appropriate physical conditions have been set up. He asks what the minimum necessary physical conditions are, what is the minimum amount of design work that a very lazy Creator would have to do, in order to see to it that the universe and, later, elephants and other complex things, would one day come into existence. The answer, from his point of view as a physical scientist, is that the Creator could be infinitely lazy. The fundamental original units that we need to postulate, in order to understand the coming into existence of everything, either consist of literally nothing (according to some physicists), or (according to other physicists) they are units of the utmost simplicity, far too simple to need anything so grand as deliberate Creation.

Atkins says that elephants and complex things do not need any explanation. But that is because he is a physical scientist, who takes for granted the biologists' theory of evolution. He doesn't really mean that elephants don’t need an explanation; rather that he is satisfied that biologists can explain elephants, provided they are allowed to take  {15}  certain facts of physics for granted. His task as a physical scientist, therefore, is to justify our taking those facts for granted. This he succeeds in doing. My position is complementary. I am a biologist. I take the facts of physics, the facts of the world of simplicity, for granted. If physicists still don’t agree over whether those simple facts are yet understood, that is not my problem. My task is to explain elephants, and the world of complex things, in terms of the simple things that physicists either understand, or are working on. The physicist's problem is the problem of ultimate origins and ultimate natural laws. The biologist's problem is the problem of complexity. The biologist tries to explain the workings, and the coming into existence, of complex things, in terms of simpler things. He can regard his task as done when he has arrived at entities so simple that they can safely be handed over to physicists.

I am aware that my characterization of a complex object — statistically improbable in a direction that is specified not with hindsight may seem idiosyncratic. So, too, may seem my characterization of physics as the study of simplicity. If you prefer some other way of defining complexity, I don’t care and I would be happy to go along with your definition for the sake of discussion. But what I do care about is that, whatever we choose to call the quality of being statistically-improbable-in-a-direction-specified-without-hindsight, it is an important quality that needs a special effort of explanation. It is the quality that characterizes biological objects as opposed to the objects of physics. The kind of explanation we come up with must not contradict the laws of physics. Indeed it will make use of the laws of physics, and nothing more than the laws of physics. But it will deploy the laws of physics in a special way that is not ordinarily discussed in physics textbooks. That special way is Darwin's way. I shall introduce its fundamental essence in Chapter 3 under the title of cumulative selection.

Meanwhile I want to follow Paley in emphasizing the magnitude of the problem that our explanation faces, the sheer hugeness of biological complexity and the beauty and elegance of biological design. Chapter 2 is an extended discussion of a particular example, ‘radar’ in bats, discovered long after Paley's time. And here, in this chapter, I have placed an illustration (Figure 1) — how Paley would have loved the electron microscope! — of an eye together with two successive ‘zoomings in’ on detailed portions. At the top of the figure is a section through an eye itself. This level of magnification shows the eye as an optical instrument. The resemblance to a camera is obvious. The iris diaphragm is responsible for constantly varying the aperture.  {16} 

the / stop

 {17}  The lens, which is really only part of a compound lens system, is responsible for the variable part of the focusing. Focus is changed by squeezing the lens with muscles (or in chameleons by moving the lens forwards or backwards, as in a man-made camera). The image falls on the retina at the back, where it excites photocells.

The middle part of Figure 1 shows a small section of the retina enlarged. Light comes from the left. The light-sensitive cells (‘photocells’) are not the first thing the light hits, but they are buried inside and facing away from the light. This odd feature is mentioned again later. The first thing the light hits is, in fact, the layer of ganglion cells which constitute the ‘electronic interface’ between the photocells and the brain. Actually the ganglion cells are responsible for preprocessing the information in sophisticated ways before relaying it to the brain, and in some ways the word ‘interface’ doesn't do justice to this. ‘Satellite computer’ might be a fairer name. Wires from the ganglion cells run along the surface of the retina to the ‘blind spot’, where they dive through the retina to form the main trunk cable to the brain, the optic nerve. There are about three million ganglion cells in the ‘electronic interface’, gathering data from about 125 million photocells.

At the bottom of the figure is one enlarged photocell, a rod. As you look at the fine architecture of this cell, keep in mind the fact that all that complexity is repeated 125 million times in each retina. And comparable complexity is repeated trillions of times elsewhere in the body as a whole. The figure of 125 million photocells is about 5,000 times the number of separately resolvable points in a good-quality magazine photograph. The folded membranes on the right of the illustrated photocell are the actual light-gathering structures. Their layered form increases the photocell's efficiency in capturing photons, the fundamental particles of which light is made. If a photon is not caught by the first membrane, it may be caught by the second, and so on. As a result of this, some eyes are capable of detecting a single photon. The fastest and most sensitive film emulsions available to photographers need about 25 times as many photons in order to detect a point of light. The lozenge-shaped objects in the middle section of the cell are mostly mitochondria. Mitochondria are found not just in photocells, but in most other cells. Each one can be thought of as a chemical factory which, in the course of delivering its primary product of usable energy, processes more than 700 different chemical substances, in long, interweaving assembly-lines strung out along the surface of its intricately folded internal membranes. The round globule at the left of Figure 1 is the nucleus. Again, this is characteristic of all animal and plant cells. Each nucleus, as we shall see in Chapter 5, contains a digitally coded  {18}  database larger, in information content, than all 30 volumes of the Encyclopaedia Britannica put together. And this figure is for each cell, not all the cells of a body put together.

The rod at the base of the picture is one single cell. The total number of cells in the body (of a human) is about 10 trillion. When you eat a steak, you are shredding the equivalent of more than 100 billion copies of the Encyclopaedia Britannica.

CHAPTER 2

GOOD DESIGN

Natural selection is the blind watchmaker, blind because it does not see ahead, does not plan consequences, has no purpose in view. Yet the living results of natural selection overwhelmingly impress us with the appearance of design as if by a master watchmaker, impress us with the illusion of design and planning. The purpose of this book is to resolve this paradox to the satisfaction of the reader, and the purpose of this chapter is further to impress the reader with the power of the illusion of design. We shall look at a particular example and shall conclude that, when it comes to complexity and beauty of design, Paley hardly even began to state the case.

We may say that a living body or organ is well designed if it has attributes that an intelligent and knowledgeable engineer might have built into it in order to achieve some sensible purpose, such as flying, swimming, seeing, eating, reproducing, or more generally promoting the survival and replication of the organism's genes. It is not necessary to suppose that the design of a body or organ is the best that an engineer could conceive of. Often the best that one engineer can do is, in any case, exceeded by the best that another engineer can do, especially another who lives later in the history of technology. But any engineer can recognize an object that has been designed, even poorly designed, for a purpose, and he can usually work out what that purpose is just by looking at the structure of the object. In Chapter 1 we bothered ourselves mostly with philosophical aspects. In this chapter, I shall develop a particular factual example that I believe would impress any engineer, namely sonar (‘radar’) in bats. In explaining each point, I shall begin by posing a problem that the living machine faces; then I shall consider possible solutions to the problem that a sensible  {22}  engineer might consider; I shall finally come to the solution that nature has actually adopted. This one example is, of course, just for illustration. If an engineer is impressed by bats, he will be impressed by countless other examples of living design.

Bats have a problem: how to find their way around in the dark. They hunt at night, and cannot use light to help them find prey and avoid obstacles. You might say that if this is a problem it is a problem of their own making, a problem that they could avoid simply by changing their habits and hunting by day. But the daytime economy is already heavily exploited by other creatures such as birds. Given that there is a living to be made at night, and given that alternative daytime trades are thoroughly occupied, natural selection has favoured bats that make a go of the night-hunting trade. It is probable, by the way, that the nocturnal trades go way back in the ancestry of all us mammals. In the time when the dinosaurs dominated the daytime economy, our mammalian ancestors probably only managed to survive at all because they found ways of scraping a living at night. Only after the mysterious mass extinction of the dinosaurs about 65 million years ago were our ancestors able to emerge into the daylight in any substantial numbers.

Returning to bats, they have an engineering problem: how to find their way and find their prey in the absence of light. Bats are not the only creatures to face this difficulty today. Obviously the night-flying insects that they prey on must find their way about somehow. Deep-sea fish and whales have little or no light by day or by night, because the sun's rays cannot penetrate far below the surface. Fish and dolphins that live in extremely muddy water cannot see because, although there is light, it is obstructed and scattered by the dirt in the water. Plenty of other modern animals make their living in conditions where seeing is difficult or impossible.

Given the question of how to manoeuvre in the dark, what solutions might an engineer consider? The first one that might occur to him is to manufacture light, to use a lantern or a searchlight. Fireflies and some fish (usually with the help of bacteria) have the power to manufacture their own light, but the process seems to consume a large amount of energy. Fireflies use their light for attracting mates. This doesn't require prohibitively much energy: a male's tiny pinprick can be seen by a female from some distance on a dark night, since her eyes are exposed directly to the light source itself. Using light to find one's own way around requires vastly more energy, since the eyes have to detect the tiny fraction of the light that bounces off each part of the scene. The light source must therefore be immensely brighter if it is to be used as a headlight to illuminate the path, than if it is to be used as a signal to  {23}  others. Anyway, whether or not the reason is the energy expense, it seems to be the case that, with the possible exception of some weird deep-sea fish, no animal apart from man uses manufactured light to find its way about.

What else might the engineer think of? Well, blind humans sometimes seem to have an uncanny sense of obstacles in their path. It has been given the name ‘facial vision’, because blind people have reported that it feels a bit like the sense of touch, on the face. One report tells of a totally blind boy who could ride his tricycle at a good speed round the block near his home, using ‘facial vision’. Experiments showed that, in fact, ‘facial vision’ is nothing to do with touch on the front of the face, although the sensation may be referred to the front of the face, like the referred pain in a phantom (severed) limb. The sensation of ‘facial vision’, it turns out, really goes in through the ears. The blind people, without even being aware of the fact, are actually using echoes, of their own footsteps and other sounds, to sense the presence of obstacles. Before this was discovered, engineers had already built instruments to exploit the principle, for example to measure the depth of the sea under a ship. After this technique had been invented, it was only a matter of time before weapons designers adapted it for the detection of submarines. Both sides in the Second World War relied heavily on these devices, under such code names as Asdic (British) and Sonar (American), as well as the similar technology of Radar (American) or RDF (British), which uses radio echoes rather than sound echoes.

The Sonar and Radar pioneers didn’t know it then, but all the world now knows that bats, or rather natural selection working on bats, had perfected the system tens of millions of years earlier, and their ‘radar’ achieves feats of detection and navigation that would strike an engineer dumb with admiration. It is technically incorrect to talk about bat ‘radar’, since they do not use radio waves. It is sonar. But the underlying mathematical theories of radar and sonar are very similar, and much of our scientific understanding of the details of what bats are doing has come from applying radar theory to them. The American zoologist Donald Griffin, who was largely responsible for the discovery of sonar in bats, coined the term ‘echolocation’ to cover both sonar and radar, whether used by animals or by human instruments. In practice, the word seems to be used mostly to refer to animal sonar.

It is misleading to speak of bats as though they were all the same. It is as though we were to speak of dogs, lions, weasels, bears, hyenas, pandas and others all in one breath, just because they are all carnivores. Different groups of bats use sonar in radically different ways, and they  {24}  seem to have ‘invented’ it separately and independently, just as the British, Germans and Americans all independently developed radar. Not all bats use echolocation. The Old World tropical fruit bats have good vision, and most of them use only their eyes for finding their way around. One or two species of fruit bats, however, for instance Rousettus, are capable of finding their way around in total darkness where eyes, however good, must be powerless. They are using sonar, but it is a cruder kind of sonar than is used by the smaller bats with which we, in temperate regions, are familiar. Rousettus clicks its tongue loudly and rhythmically as it flies, and navigates by measuring the time interval between each click and its echo. A good proportion of Rousettus's clicks are clearly audible to us (which by definition makes them sound rather than ultrasound: ultrasound is just the same as sound except that it is too high for humans to hear).

In theory, the higher the pitch of a sound, the better it is for accurate sonar. This is because low-pitched sounds have long wavelengths which cannot resolve the difference between closely spaced objects. All other things being equal therefore, a missile that used echoes for its guidance system would ideally produce very high-pitched sounds. Most bats do, indeed, use extremely high-pitched sounds, far too high for humans to hear — ultrasound. Unlike Rousettus, which can see very well and which uses unmodified relatively low-pitched sounds to do a modest amount of echolocation to supplement its good vision, the smaller bats appear to be technically highly advanced echo-machines. They have tiny eyes which, in most cases, probably can't see much. They live in a world of echoes, and probably their brains can use echoes to do something akin to ‘seeing’ images, although it is next to impossible for us to ‘visualize’ what those images might be like. The noises that they produce are not just slightly too high for humans to hear, like a kind of super dog whistle. In many cases they are vastly higher than the highest note anybody has heard or can imagine. It is fortunate that we can't hear them, incidentally, for they are immensely powerful and would be deafeningly loud if we could hear them, and impossible to sleep through.

These bats are like miniature spy planes, bristling with sophisticated instrumentation. Their brains are delicately tuned packages of miniaturized electronic wizardry, programmed with the elaborate software necessary to decode a world of echoes in real time. Their faces are often distorted into gargoyle shapes that appear hideous to us until we see them for what they are, exquisitely fashioned instruments for beaming ultrasound in desired directions.

Although we can't hear the ultrasound pulses of these bats directly,  {25}  we can get some idea of what is going on by means of a translating machine or ‘bat-detector’. This receives the pulses through a special ultrasonic microphone, and turns each pulse into an audible click or tone which we can hear through headphones. If we take such a ‘batdetector’ out to a clearing where a bat is feeding, we shall hear when each bat pulse is emitted, although we cannot hear what the pulses really ‘sound’ like. If our bat is Myotis, one of the common little brown bats, we shall hear a chuntering of clicks at a rate of about 10 per second as the bat cruises about on a routine mission. This is about the rate of a standard teleprinter, or a Bren machine gun.

Presumably the bat's image of the world in which it is cruising is being updated 10 times per second. Our own visual image appears to be continuously updated as long as our eyes are open. We can see what it might be like to have an intermittently updated world image, by using a stroboscope at night. This is sometimes done at discotheques, and it produces some dramatic effects. A dancing person appears as a succession of frozen statuesque attitudes. Obviously, the faster we set the strobe, the more the image corresponds to normal ‘continuous’ vision. Stroboscopic vision ‘sampling’ at the bat's cruising rate of about 10 samples per second would be nearly as good as normal ‘continuous’ vision for some ordinary purposes, though not for catching a ball or an insect.

This is just the sampling rate of a bat on a routine cruising flight. When a little brown bat detects an insect and starts to move in on an interception course, its click rate goes up. Faster than a machine gun, it can reach peak rates of 200 pulses per second as the bat finally closes in on the moving target. To mimic this, we should have to speed up our stroboscope so that its flashes came twice as fast as the cycles of mains electricity, which are not noticed in a fluorescent strip light. Obviously we have no trouble in performing all our normal visual functions, even playing squash or ping-pong, in a visual world ‘pulsed’ at such a high frequency. If we may imagine bat brains as building up an image of the world analogous to our visual images, the pulse rate alone seems to suggest that the bat's echo image might be at least as detailed and ‘continuous’ as our visual image. Of course, there may be other reasons why it is not so detailed as our visual image.

If bats are capable of boosting their sampling rates to 200 pulses per second, why don’t they keep this up all the time? Since they evidently have a rate control ‘knob’ on their ‘stroboscope’, why don’t they turn it permanently to maximum, thereby keeping their perception of the world at its most acute, all the time, to meet any emergency? One reason is that these high rates are suitable only for near targets. If a pulse  {26}  follows too hard on the heels of its predecessor it gets mixed up with the echo of its predecessor returning from a distant target. Even if this weren't so, there would probably be good economic reasons for not keeping up the maximum pulse rate all the time. It must be costly producing loud ultrasonic pulses, costly in energy, costly in wear and tear on voice and ears, perhaps costly in computer time. A brain that is processing 200 distinct echoes per second might not find surplus capacity for thinking about anything else. Even the ticking-over rate of about 10 pulses per second is probably quite costly, but much less so than the maximum rate of 200 per second. An individual bat that boosted its tickover rate would pay an additional price in energy, etc., which would not be justified by the increased sonar acuity. When the only moving object in the immediate vicinity is the bat itself, the apparent world is sufficiently similar in successive tenths of seconds that it need not be sampled more frequently than this. When the salient vicinity includes another moving object, particularly a flying insect twisting and turning and diving in a desperate attempt to shake off its pursuer, the extra benefit to the bat of increasing its sample rate more than justifies the increased cost. Of course, the considerations of cost and benefit in this paragraph are all surmise, but something like this almost certainly must be going on.

The engineer who sets about designing an efficient sonar or radar device soon comes up against a problem resulting from the need to make the pulses extremely loud. They have to be loud because when a sound is broadcast its wavefront advances as an ever-expanding sphere. The intensity of the sound is distributed and, in a sense, ‘diluted’ over the whole surface of the sphere. The surface area of any sphere is proportional to the radius squared. The intensity of the sound at any particular point on the sphere therefore decreases, not in proportion to the distance (the radius) but in proportion to the square of the distance from the sound source, as the wavefront advances and the sphere swells. This means that the sound gets quieter pretty fast, as it travels away from its source, in this case the bat.

When this diluted sound hits an object, say a fly, it bounces off the fly. This reflected sound now, in its turn, radiates away from the fly in an expanding spherical wavefront. For the same reason as in the case of the original sound, it decays as the square of the distance from the fly. By the time the echo reaches the bat again, the decay in its intensity is proportional, not to the distance of the fly from the bat, not even to the square of that distance, but to something more like the square of the square — the fourth power, of the distance. This means that it is very very quiet indeed. The problem can be partially overcome if the bat  {27}  beams the sound by means of the equivalent of a megaphone, but only if it already knows the direction of the target. In any case, if the bat is to receive any reasonable echo at all from a distant target, the outgoing squeak as it leaves the bat must be very loud indeed, and the instrument that detects the echo, the ear, must be highly sensitive to very quiet sounds — the echoes. Bat cries, as we have seen, are indeed often very loud, and their ears are very sensitive.

Now here is the problem that would strike the engineer trying to design a bat-like machine. If the microphone, or ear, is as sensitive as all that, it is in grave danger of being seriously damaged by its own enormously loud outgoing pulse of sound. It is no good trying to combat the problem by making the sounds quieter, for then the echoes would be too quiet to hear. And it is no good trying to combat that by making the microphone (‘ear’) more sensitive, since this would only make it more vulnerable to being damaged by the, albeit now slightly quieter, outgoing sounds! It is a dilemma inherent in the dramatic difference in intensity between outgoing sound and returning echo, a difference that is inexorably imposed by the laws of physics.

What other solution might occur to the engineer? When an analogous problem struck the designers of radar in the Second World War, they hit upon a solution which they called ‘send/receive’ radar. The radar signals were sent out in necessarily very powerful pulses, which might have damaged the highly sensitive aerials (American ‘antennas’) waiting for the faint returning echoes. The ‘send/receive’ circuit temporarily disconnected the receiving aerial just before the outgoing pulse was about to be emitted, then switched the aerial on again in time to receive the echo.

Bats developed ‘send/receive’ switching technology long long ago, probably millions of years before our ancestors came down from the trees. It works as follows. In bat ears, as in ours, sound is transmitted from the eardrum to the microphonic, sound-sensitive cells by means of a bridge of three tiny bones known (in Latin) as the hammer, the anvil and the stirrup, because of their shape. The mounting and hinging of these three bones, by the way, is exactly as a hi-fi engineer might have designed it to serve a necessary ‘impedance-matching’ function, but that is another story. What matters here is that some bats have well-developed muscles attached to the stirrup and to the hammer. When these muscles are contracted the bones don’t transmit sound so efficiently — it is as though you muted a microphone by jamming your thumb against the vibrating diaphragm. The bat is able to use these muscles to switch its ears off temporarily. The muscles contract immediately before the bat emits each outgoing pulse,  {28}  thereby switching the ears off so that they are not damaged by the loud pulse. Then they relax so that the ear returns to maximal sensitivity just in time for the returning echo. This send/receive switching system works only if split-second accuracy in timing is maintained. The bat called Tadarida is capable of alternately contracting and relaxing its switching muscles 50 times per second, keeping in perfect synchrony with the machine gun-like pulses of ultrasound. It is a formidable feat of timing, comparable to a clever trick that was used in some fighter planes during the First World War. Their machine guns fired ‘through’ the propeller, the timing being carefully synchronized with the rotation of the propeller so that the bullets always passed between the blades and never shot them off.

The next problem that might occur to our engineer is the following. If the sonar device is measuring the distance of targets by measuring the duration of silence between the emission of a sound and its returning echo — the method which Rousettus, indeed, seems to be using — the sounds would seem to have to be very brief, staccato pulses. A long drawn-out sound would still be going on when the echo returned, and, even if partially muffled by send/receive muscles, would get in the way of detecting the echo. Ideally, it would seem, bat pulses should be very brief indeed. But the briefer a sound is, the more difficult it is to make it energetic enough to produce a decent echo. We seem to have another unfortunate trade-off imposed by the laws of physics. Two solutions might occur to ingenious engineers, indeed did occur to them when they encountered the same problem, again in the analogous case of radar. Which of the two solutions is preferable depends on whether it is more important to measure range (how far away an object is from the instrument) or velocity (how fast the object is moving relative to the instrument). The first solution is that known to radar engineers as ‘chirp radar’.

We can think of radar signals as a series of pulses, but each pulse has a so-called carrier frequency. This is analogous to the ‘pitch’ of a pulse of sound or ultrasound. Bat cries, as we have seen, have a pulse-repetition rate in the tens or hundreds per second. Each one of those pulses has a carrier frequency of tens of thousands to hundreds of thousands of cycles per second. Each pulse, in other words, is a high-pitched shriek. Similarly, each pulse of radar is a ‘shriek’ of radio waves, with a high carrier frequency. The special feature of chirp radar is that it does not have a fixed carrier frequency during each shriek. Rather, the carrier frequency swoops up or down about an octave. If you think of it as its sound equivalent, each radar emission can be thought of as a swooping wolf-whistle. The advantage of chirp radar, as  {29}  opposed to the fixed pitch pulse, is the following. It doesn't matter if the original chirp is still going on when the echo returns. They won't be confused with each other. This is because the echo being detected at any given moment will be a reflection of an earlier part of the chirp, and will therefore have a different pitch.

Human radar designers have made good use of this ingenious technique. Is there any evidence that bats have ‘discovered’ it too, just as they did the send/receive system? Well, as a matter of fact, numerous species of bats do produce cries that sweep down, usually through about an octave, during each cry. These wolf-whistle cries are known as frequency modulated (FM). They appear to be just what would be required to exploit the ‘chirp radar’ technique. However, the evidence so far suggests that bats are using the technique, not to distinguish an echo from the original sound that produced it, but for the more subtle task of distinguishing echoes from other echoes. A bat lives in a world of echoes from near objects, distant objects and objects at all intermediate distances. It has to sort these echoes out from each other. If it gives downward-swooping, wolf-whistle chirps, the sorting is neatly done by pitch. When an echo from a distant object finally arrives back at the bat, it will be an ‘older’ echo than an echo that is simultaneously arriving back from a near object. It will therefore be of higher pitch. When the bat is faced with clashing echoes from several objects, it can apply the rule of thumb: higher pitch means farther away.

The second clever idea that might occur to the engineer, especially one interested in measuring the speed of a moving target, is to exploit what physicists call the Doppler Shift. This may be called the ‘ambulance effect’ because its most familiar manifestation is the sudden drop in pitch of an ambulance's siren as it speeds past the listener. The Doppler Shift occurs whenever a source of sound (or light or any other kind of wave) and a receiver of that sound move relative to one another. It is easiest to think of the sound source as motionless and the listener as moving. Assume that the siren on a factory roof is wailing continuously, all on one note. The sound is broadcast outwards as a series of waves. The waves can't be seen, because they are waves of air pressure. If they could be seen they would resemble the concentric circles spreading outwards when we throw pebbles into the middle of a still pond. Imagine that a series of pebbles is being dropped in quick succession into the middle of a pond, so that waves are continuously radiating out from the middle. If we moor a tiny toy boat at some fixed point in the pond, the boat will bob up and down rhythmically as the waves pass under it. The frequency with which the boat bobs is analogous to the pitch of a sound. Now suppose that the  {30}  boat, instead of being moored, is steaming across the pond, in the general direction of the centre from which the wave circles are originating. It will still bob up and down as it hits the successive wavefronts. But now the frequency with which it hits waves will be higher, since it is travelling towards the source of the waves. It will bob up and down at a higher rate. On the other hand, when it has passed the source of the waves and is travelling away the other side, the frequency with which it bobs up and down will obviously go down.

For the same reason, if we ride fast on a (preferably quiet) motorbike past a wailing factory siren, when we are approaching the factory the pitch will be raised: our ears are, in effect, gobbling up the waves at a faster rate than they would if we just sat still. By the same kind of argument, when our motorbike has passed the factory and is moving away from it, the pitch will be lowered. If we stop moving we shall hear the pitch of the siren as it actually is, intermediate between the two Doppler-shifted pitches. It follows that if we know the exact pitch of the siren, it is theoretically possible to work out how fast we are moving towards or away from it simply by listening to the apparent pitch and comparing it with the known ‘true’ pitch.

The same principle works when the sound source is moving and the listener is still. That is why it works for ambulances. It is rather implausibly said that Christian Doppler himself demonstrated his effect by hiring a brass band to play on an open railway truck as it rushed past his amazed audience. It is relative motion that matters, and as far as the Doppler Effect is concerned it doesn't matter whether we consider the sound source to be moving past the ear, or the ear moving past the sound source. If two trains pass in opposite directions, each travelling at 125 m.p.h., a passenger in one train will hear the whistle of the other train swoop down through a particularly dramatic Doppler Shift, since the relative velocity is 250 m.p.h.

The Doppler Effect is used in police radar speed-traps for motorists. A static instrument beams radar signals down a road. The radar waves bounce back off the cars that approach, and are registered by the receiving apparatus. The faster a car is moving, the higher is the Doppler shift in frequency. By comparing the outgoing frequency with the frequency of the returning echo the police, or rather their automatic instrument, can calculate the speed of each car. If the police can exploit the technique for measuring the speed of road hogs, dare we hope to find that bats use it for measuring the speed of insect prey?

The answer is yes. The small bats known as horseshoe bats have long been known to emit long, fixed-pitch hoots rather than staccato clicks or descending wolf-whistles. When I say long, I mean long by bat  {31}  standards. The ‘hoots’ are still less than a tenth of a second long. And there is often a ‘wolf-whistle’ tacked onto the end of each hoot, as we shall see. Imagine, first, a horseshoe bat giving out a continuous hum of ultrasound as it flies fast towards a still object, like a tree. The wavefronts will hit the tree at an accelerated rate because of the movement of the bat towards the tree. If a microphone were concealed in the tree, it would ‘hear’ the sound Doppler-shifted upwards in pitch because of the movement of the bat. There isn't a microphone in the tree, but the echo reflected back from the tree will be Doppler-shifted upwards in pitch in this way. Now, as the echo wavefronts stream back from the tree towards the approaching bat, the bat is still moving fast towards them. Therefore there is a further Doppler shift upwards in the bat's perception of the pitch of the echo. The movement of the bat leads to a kind of double Doppler shift, whose magnitude is a precise indication of the velocity of the bat relative to the tree. By comparing the pitch of its cry with the pitch of the returning echo, therefore, the bat (or rather its on-board computer in the brain) could, in theory, calculate how fast it was moving towards the tree. This wouldn't tell the bat how far away the tree was, but it might still be very useful information, nevertheless.

If the object reflecting the echoes were not a static tree but a moving insect, the Doppler consequences would be more complicated, but the bat could still calculate the velocity of relative motion between itself and its target, obviously just the kind of information a sophisticated guided missile like a hunting bat needs. Actually some bats play a trick that is more interesting than simply emitting hoots of constant pitch and measuring the pitch of the returning echoes. They carefully adjust the pitch of the outgoing hoots, in such a way as to keep the pitch of the echo constant after it has been Doppler-shifted. As they speed towards a moving insect, the pitch of their cries is constantly changing, continuously hunting for just the pitch needed to keep the returning echoes at a fixed pitch. This ingenious trick keeps the echo at the pitch to which their ears are maximally sensitive — important since the echoes are so faint. They can then obtain the necessary information for their Doppler calculations, by monitoring the pitch at which they are obliged to hoot in order to achieve the fixed-pitch echo. I don’t know whether man-made devices, either sonar or radar, use this subtle trick. But on the principle that most clever ideas in this field seem to have been developed first by bats, I don’t mind betting that the answer is yes.

It is only to be expected that these two rather different techniques, the Doppler shift technique and the ‘chirp radar’ technique, would be  {32}  useful for different special purposes. Some groups of bats specialize in one of them, some in the other. Some groups seem to try to get the best of both worlds, tacking an FM ‘wolf-whistle’ onto the end (or sometimes the beginning) of a long, constant-frequency ‘hoot’. Another curious trick of horseshoe bats concerns movements of their outer ear flaps. Unlike other bats, horseshoe bats move their outer ear flaps in fast alternating forward and backward sweeps. It is conceivable that this additional rapid movement of the listening surface relative to the target causes useful modulations in the Doppler shift, modulations that supply additional information. When the ear is flapping towards the target, the apparent velocity of movement towards the target goes up. When it is flapping away from the target, the reverse happens. The bat's brain ‘knows’ the direction of flapping of each ear, and in principle could make the necessary calculations to exploit the information.

Possibly the most difficult problem of all that bats face is the danger of inadvertent ‘jamming’ by the cries of other bats. Human experimenters have found it surprisingly difficult to put bats off their stride by playing loud artificial ultrasound at them. With hindsight one might have predicted this. Bats must have come to terms with the jamming-avoidance problem long ago. Many species of bats roost in enormous aggregations, in caves that must be a deafening babel of ultrasound and echoes, yet the bats can still fly rapidly about the cave, avoiding the walls and each other in total darkness. How does a bat keep track of its own echoes, and avoid being misled by the echoes of others? The first solution that might occur to an engineer is some sort of frequency coding: each bat might have its own private frequency, just like separate radio stations. To some extent this may happen, but it is by no means the whole story.

How bats avoid being jammed by other bats is not well understood, but an interesting clue comes from experiments on trying to put bats off. It turns out that you can actively deceive some bats if you play back to them their own cries with an artificial delay. Give them, in other words, false echoes of their own cries. It is even possible, by carefully controlling the electronic apparatus delaying the false echo, to make the bats attempt to land on a ‘phantom’ ledge. I suppose it is the bat equivalent of looking at the world through a lens.

It seems that bats may be using something that we could call a ‘strangeness filter’. Each successive echo from a bat's own cries produces a picture of the world that makes sense in terms of the previous picture of the world built up with earlier echoes. If the bat's brain hears an echo from another bat's cry, and attempts to incorporate it into the  {33}  picture of the world that it has previously built up, it will make no sense. It will appear as though objects in the world have suddenly jumped in various random directions. Objects in the real world do not behave in such a crazy way, so the brain can safely filter out the apparent echo as background noise. If a human experimenter feeds the bat artificially delayed or accelerated ‘echoes’ of its own cries, the false echoes will make sense in terms of the world picture that the bat has previously built up. The false echoes are accepted by the strangeness filter because they are plausible in the context of the previous echoes. They cause objects to seem to shift in position by only a small amount, which is what objects plausibly can be expected to do in the real world. The bat's brain relies upon the assumption that the world portrayed by any one echo pulse will be either the same as the world portrayed by previous pulses, or only slightly different: the insect being tracked may have moved a little, for instance.

There is a well-known paper by the philosopher Thomas Nagel called ‘What is it like to be a bat?’. The paper is not so much about bats as about the philosophical problem of imagining what it is ‘like’ to be anything that we are not. The reason a bat is a particularly telling example for a philosopher, however, is that the experiences of an echolocating bat are assumed to be peculiarly alien and different from our own. If you want to share a bat's experience, it is almost certainly grossly misleading to go into a cave, shout or bang two spoons together, consciously time the delay before you hear the echo, and calculate from this how far the wall must be.

That is no more what it is like to be a bat than the following is a good picture of what it is like to see colour: use an instrument to measure the wavelength of the light that is entering your eye: if it is long, you are seeing red, if it is short you are seeing violet or blue. It happens to be a physical fact that the light that we call red has a longer wavelength than the light that we call blue. Different wavelengths switch on the red-sensitive and the blue-sensitive photocells in our retinas. But there is no trace of the concept of wavelength in our subjective sensation of the colours. Nothing about ‘what it is like’ to see blue or red tells us which light has the longer wavelength. If it matters (it usually doesn't), we just have to remember it, or (what I always do) look it up in a book. Similarly, a bat perceives the position of an insect using what we call echoes. But the bat surely no more thinks in terms of delays of echoes when it perceives an insect, than we think in terms of wavelengths when we perceive blue or red.

Indeed, if I were forced to try the impossible, to imagine what it is like to be a bat, I would guess that echolocating, for them, might be  {34}  rather like seeing for us. We are such thoroughly visual animals that we hardly realize what a complicated business seeing is. Objects are ‘out there’, and we think that we ‘see’ them out there. But I suspect that really our percept is an elaborate computer model in the brain, constructed on the basis of information coming from out there, but transformed in the head into a form in which that information can be used. Wavelength differences in the light out there become coded as ‘colour’ differences in the computer model in the head. Shape and other attributes are encoded in the same kind of way, encoded into a form that is convenient to handle. The sensation of seeing is, for us, very different from the sensation of hearing, but this cannot be directly due to the physical differences between light and sound. Both light and sound are, after all, translated by the respective sense organs into the same kind of nerve impulses. It is impossible to tell, from the physical attributes of a nerve impulse, whether it is conveying information about light, about sound or about smell. The reason the sensation of seeing is so different from the sensation of hearing and the sensation of smelling is that the brain finds it convenient to use different kinds of internal model of the visual world, the world of sound and the world of smell. It is because we internally use our visual information and our sound information in different ways and for different purposes that the sensations of seeing and hearing are so different. It is not directly because of the physical differences between light and sound.

But a bat uses its sound information for very much the same kind of purpose as we use our visual information. It uses sound to perceive, and continuously update its perception of, the position of objects in three-dimensional space, just as we use light. The type of internal computer model that it needs, therefore, is one suitable for the internal representation of the changing positions of objects in three-dimensional space. My point is that the form that an animal's subjective experience takes will be a property of the internal computer model. That model will be designed, in evolution, for its suitability for useful internal representation, irrespective of the physical stimuli that come to it from outside. Bats and we need the same kind of internal model for representing the position of objects in three-dimensional space. The fact that bats construct their internal model with the aid of echoes, while we construct ours with the aid of light, is irrelevant. That outside information is, in any case, translated into the same kind of nerve impulses on its way to the brain.

My conjecture, therefore, is that bats ‘see’ in much the same way as we do, even though the physical medium by which the world ‘out there’ is translated into nerve impulses is so different — ultrasound  {35}  rather than light. Bats may even use the sensations that we call colour for their own purposes, to represent differences in the world out there that have nothing to do with the physics of wavelength, but which play a functional role, for the bat, similar to the role that colours play to us. Perhaps male bats have body surfaces that are subtly textured so that the echoes that bounce off them are perceived by females as gorgeously coloured, the sound equivalent of the nuptial plumage of a bird of paradise. I don’t mean this just as some vague metaphor. It is possible that the subjective sensation experienced by a female bat when she perceives a male really is, say, bright red: the same sensation as I experience when I see a flamingo. Or, at least, the bat's sensation of her mate may be no more different from my visual sensation of a flamingo, than my visual sensation of a flamingo is different from a flamingo's visual sensation of a flamingo.

Donald Griffin tells a story of what happened when he and his colleague Robert Galambos first reported to an astonished conference of zoologists in 1940 their new discovery of the facts of bat echolocation. One distinguished scientist was so indignantly incredulous that

he seized Galambos by the shoulders and shook him while complaining that we could not possibly mean such an outrageous suggestion. Radar and sonar were still highly classified developments in military technology, and the notion that bats might do anything even remotely analogous to the latest triumphs of electronic engineering struck most people as not only implausible but emotionally repugnant.

It is easy to sympathize with the distinguished sceptic. There is something very human in his reluctance to believe. And that, really, says it: human is precisely what it is. It is precisely because our own human senses are not capable of doing what bats do that we find it hard to believe. Because we can only understand it at a level of artificial instrumentation, and mathematical calculations on paper, we find it hard to imagine a little animal doing it in its head. Yet the mathematical calculations that would be necessary to explain the principles of vision are just as complex and difficult, and nobody has ever had any difficulty in believing that little animals can see. The reason for this double standard in our scepticism is, quite simply, that we can see and we can't echolocate.

I can imagine some other world in which a conference of learned, and totally blind, bat-like creatures is flabbergasted to be told of animals called humans that are actually capable of using the newly discovered inaudible rays called ‘light’, still the subject of top-secret  {36}  military development, for finding their way about. These otherwise humble humans are almost totally deaf (well, they can hear after a fashion and even utter a few ponderously slow, deep drawling growls, but they only use these sounds for rudimentary purposes like communicating with each other; they don’t seem capable of using them to detect even the most massive objects). They have, instead, highly specialized organs called ‘eyes’ for exploiting ‘light’ rays. The sun is the main source of light rays, and humans, remarkably, manage to exploit the complex echoes that bounce off objects when light rays from the sun hit them. They have an ingenious device called a ‘lens’, whose shape appears to be mathematically calculated so that it bends these silent rays in such a way that there is an exact one-to-one mapping between objects in the world and an ‘image’ on a sheet of cells called the ‘retina’. These retinal cells are capable, in some mysterious way, of rendering the light ‘audible’ (one might say), and they relay their information to the brain. Our mathematicians have shown that it is theoretically possible, by doing the right highly complex calculations, to navigate safely through the world using these light rays, just as effectively as one can in the ordinary way using ultrasound — in some respects even more effectively! But who would have thought that a humble human could do these calculations?

Echo-sounding by bats is just one of the thousands of examples that I could have chosen to make the point about good design. Animals give the appearance of having been designed by a theoretically sophisticated and practically ingenious physicist or engineer, but there is no suggestion that the bats themselves know or understand the theory in the same sense as a physicist understands it. The bat should be thought of as analogous to the police radar trapping instrument, not to the person who designed that instrument. The designer of the police radar speed-meter understood the theory of the Doppler Effect, and expressed this understanding in mathematical equations, explicitly written out on paper. The designer's understanding is embodied in the design of the instrument, but the instrument itself does not understand how it works. The instrument contains electronic components, which are wired up so that they automatically compare two radar frequencies and convert the result into convenient units — miles per hour. The computation involved is complicated, but well within the powers of a small box of modern electronic components wired up in the proper way. Of course, a sophisticated conscious brain did the wiring up (or at least designed the wiring diagram), but no conscious brain is involved in the moment-to-moment working of the box.

Our experience of electronic technology prepares us to accept the  {37}  idea that unconscious machinery can behave as if it understands complex mathematical ideas. This idea is directly transferable to the workings of living machinery. A bat is a machine, whose internal electronics are so wired up that its wing muscles cause it to home in on insects, as an unconscious guided missile homes in on an aeroplane. So far our intuition, derived from technology, is correct. But our experience of technology also prepares us to see the mind of a conscious and purposeful designer in the genesis of sophisticated machinery. It is this second intuition that is wrong in the case of living machinery. In the case of living machinery, the ‘designer’ is unconscious natural selection, the blind watchmaker.

I hope that the reader is as awestruck as I am, and as William Paley would have been, by these bat stories. My aim has been in one respect identical to Paley's aim. I do not want the reader to underestimate the prodigious works of nature and the problems we face in explaining them. Echolocation in bats, although unknown in Paley's time, would have served his purpose just as well as any of his examples. Paley rammed home his argument by multiplying up his examples. He went right through the body, from head to toe, showing how every part, every last detail, was like the interior of a beautifully fashioned watch. In many ways I should like to do the same, for there are wonderful stories to be told, and I love storytelling. But there is really no need to multiply examples. One or two will do. The hypothesis that can explain bat navigation is a good candidate for explaining anything in the world of life, and if Paley's explanation for any one of his examples was wrong we can't make it right by multiplying up examples. His hypothesis was that living watches were literally designed and built by a master watchmaker. Our modern hypothesis is that the job was done in gradual evolutionary stages by natural selection.

Nowadays theologians aren't quite so straightforward as Paley. They don’t point to complex living mechanisms and say that they are self-evidently designed by a creator, just like a watch. But there is a tendency to point to them and say ‘It is impossible to believe’ that such complexity, or such perfection, could have evolved by natural selection. Whenever I read such a remark, I always feel like writing ‘Speak for yourself’ in the margin. There are numerous examples (I counted 35 in one chapter) in a recent book called The Probability of God by the Bishop of Birmingham, Hugh Montefiore. I shall use this book for all my examples in the rest of this chapter, because it is a sincere and honest attempt, by a reputable and educated writer, to bring natural theology up to date. When I say honest, I mean honest. Unlike some of his theological colleagues, Bishop Montefiore is not  {38}  afraid to state that the question of whether God exists is a definite question of fact. He has no truck with shifty evasions such as ‘Christianity is a way of life. The question of God's existence is eliminated: it is a mirage created by the illusions of realism’. Parts of his book are about physics and cosmology, and I am not competent to comment on those except to note that he seems to have used genuine physicists as his authorities. Would that he had done the same in the biological parts. Unfortunately, he preferred here to consult the works of Arthur Koestler, Fred Hoyle, Gordon Rattray-Taylor and Karl Popper! The Bishop believes in evolution, but cannot believe that natural selection is an adequate explanation for the course that evolution has taken (partly because, like many others, he sadly misunderstands natural selection to be ‘random’ and ‘meaningless’).

He makes heavy use of what may be called the Argument from Personal Incredulity. In the course of one chapter we find the following phrases, in this order:

...there seems no explanation on Darwinian grounds... It is no easier to explain... It is hard to understand... It is not easy to understand... It is equally difficult to explain... I do not find it easy to comprehend... I do not find it easy to see... I find it hard to understand... it does not seem feasible to explain... I cannot see how... neo-Darwinism seems inadequate to explain many of the complexities of animal behaviour... it is not easy to comprehend how such behaviour could have evolved solely through natural selection... It is impossible... How could an organ so complex evolve?... It is not easy to see... It is difficult to see...

The Argument from Personal Incredulity is an extremely weak argument, as Darwin himself noted. In some cases it is based upon simple ignorance. For instance, one of the facts that the Bishop finds it difficult to understand is the white colour of polar bears.

As for camouflage, this is not always easily explicable on neo-Darwinian premises. If polar bears are dominant in the Arctic, then there would seem to have been no need for them to evolve a white-coloured form of camouflage.

This should be translated:

I personally, off the top of my head sitting in my study, never having visited the Arctic, never having seen a polar bear in the wild, and having been educated in classical literature and theology, have not so far managed to think of a reason why polar bears might benefit from being white.

In this particular case, the assumption being made is that only animals that are preyed upon need camouflage. What is overlooked is that  {39}  predators also benefit from being concealed from their prey. Polar bears stalk seals resting on the ice. If the seal sees the bear coming from far-enough away, it can escape. I suspect that, if he imagines a dark grizzly bear trying to stalk seals over the snow, the Bishop will immediately see the answer to his problem.

The polar bear argument turned out to be almost too easy to demolish but, in an important sense, this is not the point. Even if the foremost authority in the world can't explain some remarkable biological phenomenon, this doesn't mean that it is inexplicable. Plenty of mysteries have lasted for centuries and finally yielded to explanation. For what it is worth, most modern biologists wouldn't find it difficult to explain every one of the Bishop's 35 examples in terms of the theory of natural selection, although not all of them are quite as easy as the polar bears. But we aren't testing human ingenuity. Even if we found one example that we couldn't explain, we should hesitate to draw any grandiose conclusions from the fact of our own inability. Darwin himself was very clear on this point.

There are more serious versions of the argument from personal incredulity, versions which do not rest simply upon ignorance or lack of ingenuity. One form of the argument makes direct use of the extreme sense of wonder which we all feel when confronted with highly complicated machinery, like the detailed perfection of the echolocation equipment of bats. The implication is that it is somehow self-evident that anything so wonderful as this could not possibly have evolved by natural selection. The Bishop quotes, with approval, G. Bennett on spider webs:

It is impossible for one who has watched the work for many hours to have any doubt that neither the present spiders of this species nor their ancestors were ever the architects of the web or that it could conceivably have been produced step by step through random variation; it would be as absurd to suppose that the intricate and exact proportions of the Parthenon were produced by piling together bits of marble.

It is not impossible at all. That is exactly what I firmly believe, and I have some experience of spiders and their webs.

The Bishop goes on to the human eye, asking rhetorically, and with the implication that there is no answer, ‘How could an organ so complex evolve?’ This is not an argument, it is simply an affirmation of incredulity. The underlying basis for the intuitive incredulity that we all are tempted to feel about what Darwin called organs of extreme perfection and complication is, I think, twofold. First we have no intuitive grasp of the immensities of time available for evolutionary  {40}  change. Most sceptics about natural selection are prepared to accept that it can bring about minor changes like the dark coloration that has evolved in various species of moth since the industrial revolution. But, having accepted this, they then point out how small a change this is. As the Bishop underlines, the dark moth is not a new species. I agree that this is a small change, no match for the evolution of the eye, or of echolocation. But equally, the moths only took a hundred years to make their change. One hundred years seems like a long time to us, because it is longer than our lifetime. But to a geologist it is about a thousand times shorter than he can ordinarily measure!

Eyes don’t fossilize, so we don’t know how long our type of eye took to evolve its present complexity and perfection from nothing, but the time available is several hundred million years. Think, by way of comparison, of the change that man has wrought in a much shorter time by genetic selection of dogs. In a few hundreds, or at most thousands, of years we have gone from wolf to Pekinese, Bulldog, Chihuahua and Saint Bernard. Ah, but they are still dogs aren't they? They haven't turned into a different ‘kind’ of animal? Yes, if it comforts you to play with words like that, you can call them all dogs. But just think about the time involved. Let's represent the total time it took to evolve all these breeds of dog from a wolf, by one ordinary walking pace. Then, on the same scale, how far would you have to walk, in order to get back to Lucy and her kind, the earliest human fossils that unequivocally walked upright? The answer is about 2 miles. And how far would you have to walk, in order to get back to the start of evolution on Earth? The answer is that you would have to slog it out all the way from London to Baghdad. Think of the total quantity of change involved in going from wolf to Chihuahua, and then multiply it up by the number of walking paces between London and Baghdad. This will give some intuitive idea of the amount of change that we can expect in real natural evolution.

The second basis for our natural incredulity about the evolution of very complex organs like human eyes and bat ears is an intuitive application of probability theory. Bishop Montefiore quotes C. E. Raven on cuckoos. These lay their eggs in the nests of other birds, which then act as unwitting foster parents. Like so many biological adaptations, that of the cuckoo is not single but multiple. Several different facts about cuckoos fit them to their parasitic way of life. For instance, the mother has the habit of laying in other birds’ nests, and the baby has the habit of throwing the host's own chicks out of the nest. Both habits help the cuckoo succeed in its parasitic life. Raven goes on:  {41} 

It will be seen that each one of this sequence of conditions is essential for the success of the whole. Yet each by itself is useless. The whole opus perfectum must have been achieved simultaneously. The odds against the random occurrence of such a series of coincidences are, as we have already stated, astronomical.

Arguments such as this are in principle more respectable than the argument based on sheer, naked incredulity. Measuring the statistical improbability of a suggestion is the right way to go about assessing its believability. Indeed, it is a method that we shall use in this book several times. But you have to do it right! There are two things wrong with the argument put by Raven. First, there is the familiar, and I have to say rather irritating, confusion of natural selection with ‘randomness’. Mutation is random; natural selection is the very opposite of random. Second, it just isn't true that ‘each by itself is useless’. It isn't true that the whole perfect work must have been achieved simultaneously. It isn't true that each part is essential for the success of the whole. A simple, rudimentary, half-cocked eye/ear/echolocation system/cuckoo parasitism system, etc., is better than none at all. Without an eye you are totally blind. With half an eye you may at least be able to detect the general direction of a predator's movement, even if you can't focus a clear image. And this may make all the difference between life and death. These matters will be taken up again in more detail in the next two chapters.

CHAPTER 3

ACCUMULATING
SMALL CHANGE

We have seen that living things are too improbable and too beautifully ‘designed’ to have come into existence by chance. How, then, did they come into existence? The answer, Darwin's answer, is by gradual, step-by-step transformations from simple beginnings, from primordial entities sufficiently simple to have come into existence by chance. Each successive change in the gradual evolutionary process was simple enough, relative to its predecessor, to have arisen by chance. But the whole sequence of cumulative steps constitutes anything but a chance process, when you consider the complexity of the final end-product relative to the original starting point. The cumulative process is directed by nonrandom survival. The purpose of this chapter is to demonstrate the power of this cumulative selection as a fundamentally nonrandom process.

If you walk up and down a pebbly beach, you will notice that the pebbles are not arranged at random. The smaller pebbles typically tend to be found in segregated zones running along the length of the beach, the larger ones in different zones or stripes. The pebbles have been sorted, arranged, selected. A tribe living near the shore might wonder at this evidence of sorting or arrangement in the world, and might develop a myth to account for it, perhaps attributing it to a Great Spirit in the sky with a tidy mind and a sense of order. We might give a superior smile at such a superstitious notion, and explain that the arranging was really done by the blind forces of physics, in this case the action of waves. The waves have no purposes and no intentions, no tidy mind, no mind at all. They just energetically throw the pebbles around, and big pebbles and small pebbles respond differently to this treatment so they end up at different levels of the beach. A small amount of order has come out of disorder, and no mind planned it.  {44} 

The waves and the pebbles together constitute a simple example of a system that automatically generates non-randomness. The world is full of such systems. The simplest example I can think of is a hole. Only objects smaller than the hole can pass through it. This means that if you start with a random collection of objects above the hole, and some force shakes and jostles them about at random, after a while the objects above and below the hole will come to be nonrandomly sorted. The space below the hole will tend to contain objects smaller than the hole, and the space above will tend to contain objects larger than the hole. Mankind has, of course, long exploited this simple principle for generating non-randomness, in the useful device known as the sieve.

The Solar System is a stable arrangement of planets, comets and debris orbiting the sun, and it is presumably one of many such orbiting systems in the universe. The nearer a satellite is to its sun, the faster it has to travel if it is to counter the sun's gravity and remain in stable orbit. For any given orbit, there is only one speed at which a satellite can travel and remain in that orbit. If it were travelling at any other velocity, it would either move out into deep space, or crash into the Sun, or move into another orbit. And if we look at the planets of our solar system, look and behold, every single one of them is travelling at exactly the right velocity to keep it in its stable orbit around the Sun. A blessed miracle of provident design? No, just another natural ‘sieve’. Obviously all the planets that we see orbiting the sun must be travelling at exactly the right speed to keep them in their orbits, or we wouldn't see them there because they wouldn't be there! But equally obviously this is not evidence for conscious design. It is just another kind of sieve.

Sieving of this order of simplicity is not, on its own, enough to account for the massive amounts of nonrandom order that we see in living things. Nowhere near enough. Remember the analogy of the combination lock. The kind of non-randomness that can be generated by simple sieving is roughly equivalent to opening a combination lock with only one dial: it is easy to open it by sheer luck. The kind of non-randomness that we see in living systems, on the other hand, is equivalent to a gigantic combination lock with an almost uncountable number of dials. To generate a biological molecule like haemoglobin, the red pigment in blood, by simple sieving would be equivalent to taking all the amino-acid building blocks of haemoglobin, jumbling them up at random, and hoping that the haemoglobin molecule would reconstitute itself by sheer luck. The amount of luck that would be required for this feat is unthinkable, and has been used as a telling mind-boggler by Isaac Asimov and others.  {45} 

A haemoglobin molecule consists of four chains of amino acids twisted together. Let us think about just one of these four chains. It consists of 146 amino acids. There are 20 different kinds of amino acids commonly found in living things. The number of possible ways of arranging 20 kinds of thing in chains 146 links long is an inconceivably large number, which Asimov calls the ‘haemoglobin number’. It is easy to calculate, but impossible to visualize the answer. The first link in the 146-long chain could be any one of the 20 possible amino acids. The second link could also be any one of the 20, so the number of possible 2-link chains is 20 × 20, or 400. The number of possible 3-link chains is 20 × 20 × 20, or 8,000. The number of possible 146-link chains is 20 times itself 146 times. This is a staggeringly large number. A million is a 1 with 6 noughts after it. A billion (1,000 million) is a 1 with 9 noughts after it. The number we seek, the ‘haemoglobin number’, is (near enough) a 1 with 190 noughts after it! This is the chance against happening to hit upon haemoglobin by luck. And a haemoglobin molecule has only a minute fraction of the complexity of a living body. Simple sieving, on its own, is obviously nowhere near capable of generating the amount of order in a living thing. Sieving is an essential ingredient in the generation of living order, but it is very far from being the whole story. Something else is needed. To explain the point, I shall need to make a distinction between ‘single-step’ selection and ‘cumulative’ selection. The simple sieves we have been considering so far in this chapter are all examples of single-step selection. Living organization is the product of cumulative selection.

The essential difference between single-step selection and cumulative selection is this. In single-step selection the entities selected or sorted, pebbles or whatever they are, are sorted once and for all. In cumulative selection, on the other hand, they ‘reproduce’; or in some other way the results of one sieving process are fed into a subsequent sieving, which is fed into... , and so on. The entities are subjected to selection or sorting over many ‘generations’ in succession. The end-product of one generation of selection is the starting point for the next generation of selection, and so on for many generations. It is natural to borrow such words as ‘reproduce’ and ‘generation’, which have associations with living things, because living things are the main examples we know of things that participate in cumulative selection. They may in practice be the only things that do. But for the moment I don’t want to beg that question by saying so outright.

Sometimes clouds, through the random kneading and carving of the winds, come to look like familiar objects. There is a much published photograph, taken by the pilot of a small aeroplane, of what looks a bit  {46}  like the face of Jesus, staring out of the sky. We have all seen clouds that reminded us of something — a sea horse, say, or a smiling face. These resemblances come about by single-step selection, that is to say by a Single coincidence. They are, consequently, not very impressive. The resemblance of the signs of the zodiac to the animals after which they are named, Scorpio, Leo, and so on, is as unimpressive as the predictions of astrologers. We don’t feel overwhelmed by the resemblance, as we are by biological adaptations — the products of cumulative selection. We describe as weird, uncanny or spectacular, the resemblance of, say, a leaf insect to a leaf or a praying mantis to a cluster of pink flowers. The resemblance of a cloud to a weasel is only mildly diverting, barely worth calling to the attention of our companion. Moreover, we are quite likely to change our mind about exactly what the cloud most resembles.

Hamlet. Do you see yonder cloud that's almost in shape of a camel?

Polonius. By the mass, and ‘tis like a camel, indeed.

Hamlet. Methinks it is like a weasel.

Polonius. It is backed like a weasel.

Hamlet. Or like a whale?

Polonius. Very like a whale.

I don’t know who it was first pointed out that, given enough time, a monkey bashing away at random on a typewriter could produce all the works of Shakespeare. The operative phrase is, of course, given enough time. Let us limit the task facing our monkey somewhat. Suppose that he has to produce, not the complete works of Shakespeare but just the short sentence ‘Methinks it is like a weasel’, and we shall make it relatively easy by giving him a typewriter with a restricted keyboard, one with just the 26 (capital) letters, and a space bar. How long will he take to write this one little sentence?

The sentence has 28 characters in it, so let us assume that the monkey has a series of discrete ‘tries’, each consisting of 28 bashes at the keyboard. If he types the phrase correctly, that is the end of the experiment. If not, we allow him another ‘try’ of 28 characters. I don’t know any monkeys, but fortunately my 11-month old daughter is an experienced randomizing device, and she proved only too eager to step into the role of monkey typist. Here is what she typed on the computer:

UMMK JK CDZZ F ZD DSDSKSM
S SS FMCV PU I DDRGLKDXRRDO
RDTE QDWFDVIOY UDSKZWDCCVYT
H CHVY NMGNBAYTDFCCVD D
RCDFYYYRM N DFSKD LD K WDWK
JJKAUIZMZI UXDKIDISFUMDKUDXI  {47} 

She has other important calls on her time, so I was obliged to program the computer to simulate a randomly typing baby or monkey:

WDLDMNLTDTJBKWIRZREZLMQCO P
Y YVMQKZPGJXWVHGLAWFVCHQYOPY
MWR SWTNUXMLCDLEUBXTQHNZVJQF
FU OVAODVYKDGXDEKYVMOGGS VT
HZQZDSFZIHIVPHZPETPWVOVPMZGF
GEWRGZRPBCTPGQMCKHFDBGW ZCCF

And so on and on. It isn't difficult to calculate how long we should reasonably expect to wait for the random computer (or baby or monkey) to type METHINKS IT IS LIKE A WEASEL. Think about the total number of possible phrases of the right length that the monkey or baby or random computer could type. It is the same kind of calculation as we did for haemoglobin, and it produces a similarly large result. There are 27 possible letters (counting ‘space’ as one letter) in the first position. The chance of the monkey happening to get the first letter — M — right is therefore 1 in 27. The chance of it getting the first two letters — ME — right is the chance of it getting the second letter — E — right (1 in 27) given that it has also got the first letter — M — right, therefore 1/27 × 1/27, which equals 1/729. The chance of it getting the first word METHINKS — right is 1/27 for each of the 8 letters, therefore (1/27) × (1/27) × (1/27) × (1/27)..., etc. 8 times, or (1/27) to the power 8. The chance of it getting the entire phrase of 28 characters right is (1/27) to the power 28, i.e. (1/27) multiplied by itself 28 times. These are very small odds, about 1 in 10,000 million million million million million million. To put it mildly, the phrase we seek would be a long time coming, to say nothing of the complete works of Shakespeare.

So much for single-step selection of random variation. What about cumulative selection; how much more effective should this be? Very very much more effective, perhaps more so than we at first realize, although it is almost obvious when we reflect further. We again use our computer monkey, but with a crucial difference in its program. It again begins by choosing a random sequence of 28 letters, just as before:

WDLMNLT DTJBKWIRZREZLMQCO P

It now ‘breeds from’ this random phrase. It duplicates it repeatedly, but with a certain chance of random error — ‘mutation’ — in the copying. The computer examines the mutant nonsense phrases, the ‘progeny’ of the original phrase, and chooses the one which, however slightly, most resembles the target phrase, METHINKS IT IS LIKE A  {48}  WEASEL. In this instance the winning phrase of the next ‘generation’ happened to be:

WDLTMNLT DTJBSWIRZREZLMQCO P

Not an obvious improvement! But the procedure is repeated, again mutant ‘progeny’ are ‘bred from’ the phrase, and a new ‘winner’ is chosen. This goes on, generation after generation. After 10 generations, the phrase chosen for ‘breeding’ was:

MDLDMNLS ITJISWHRZREZ MECS P

After 20 generations it was:

MELDINLS IT ISWPRKE Z WECSEL

By now, the eye of faith fancies that it can see a resemblance to the target phrase. By 30 generations there can be no doubt:

METHINGS IT ISWLIKE B WECSEL

Generation 40 takes us to within one letter of the target:

METHINKS IT IS LIKE I WEASEL

And the target was finally reached in generation 43. A second run of the computer began with the phrase:

Y YVMQKZPFJXWVHGLAWFVCHQXYOPY,

passed through (again reporting only every tenth generation):

Y YVMQKSPFTXWSHLIKEFV HQYSPY
YETHINKSPITXISHLIKEFA WQYSEY
METHINKS IT ISSLIKE A WEFSEY
METHINKS IT ISBLIKE A WEASES
METHINKS IT ISJLIKE A WEASEO
METHINKS IT IS LIKE A WEASEP

and reached the target phrase in generation 64. On a third run the computer started with:

GEWRGZRPBCTPGQMCKHFDBGW ZCCF

and reached METHINKS IT IS LIKE A WEASEL in 41 generations of selective ‘breeding’.

The exact time taken by the computer to reach the target doesn't matter. If you want to know, it completed the whole exercise for me, the first time, while I was out to lunch. It took about half an hour. (Computer enthusiasts may think this unduly slow. The reason is that  {49}  the program was written in BASIC, a sort of computer baby-talk. When I rewrote it in Pascal, it took 11 seconds.) Computers are a bit faster at this kind of thing than monkeys, but the difference really isn't significant. What matters is the difference between the time taken by cumulative selection, and the time which the same computer, working flat out at the same rate, would take to reach the target phrase if it were forced to use the other procedure of single-step selection: about a million million million million million years. This is more than a million million million times as long as the universe has so far existed. Actually it would be fairer just to say that, in comparison with the time it would take either a monkey or a randomly programmed computer to type our target phrase, the total age of the universe so far is a negligibly small quantity, so small as to be well within the margin of error for this sort of back-of-an-envelope calculation. Whereas the time taken for a computer working randomly but with the constraint of cumulative selection to perform the same task is of the same order as humans ordinarily can understand, between 11 seconds and the time it takes to have lunch.

There is a big difference, then, between cumulative selection (in which each improvement, however slight, is used as a basis for future building), and single-step selection (in which each new ‘try’ is a fresh one). If evolutionary progress had had to rely on single-step selection, it would never have got anywhere. If, however, there was any way in which the necessary conditions for cumulative selection could have been set up by the blind forces of nature, strange and wonderful might have been the consequences. As a matter of fact that is exactly what happened on this planet, and we ourselves are among the most recent, if not the strangest and most wonderful, of those consequences.

It is amazing that you can still read calculations like my haemoglobin calculation, used as though they constituted arguments against Darwin's theory. The people who do this, often expert in their own field, astronomy or whatever it may be, seem sincerely to believe that Darwinism explains living organization in terms of chance — ‘single-step selection’ — alone. This belief, that Darwinian evolution is ‘random’, is not merely false. It is the exact opposite of the truth. Chance is a minor ingredient in the Darwinian recipe, but the most important ingredient is cumulative selection which is quintessentially nonrandom.

Clouds are not capable of entering into cumulative selection. There is no mechanism whereby clouds of particular shapes can spawn daughter clouds resembling themselves. If there were such a mechanism, if a cloud resembling a weasel or a camel could give rise to  {50}  a lineage of other clouds of roughly the same shape, cumulative selection would have the opportunity to get going. Of course, clouds do break up and form ‘daughter’ clouds sometimes, but this isn't enough for cumulative selection. It is also necessary that the ‘progeny’ of any given cloud should resemble its ‘parent’ more than it resembles any old ‘parent’ in the ‘population’. This vitally important point is apparently misunderstood by some of the philosophers who have, in recent years, taken an interest in the theory of natural selection. It is further necessary that the chances of a given cloud's surviving and spawning copies should depend upon its shape. Maybe in some distant galaxy these conditions did arise, and the result, if enough millions of years have gone by, is an ethereal, wispy form of life. This might make a good science fiction story — The White Cloud, it could be called — but for our purposes a computer model like the monkey/Shakespeare model is easier to grasp.

Although the monkey/Shakespeare model is useful for explaining the distinction between single-step selection and cumulative selection, it is misleading in important ways. One of these is that, in each generation of selective ‘breeding’, the mutant ‘progeny’ phrases were judged according to the criterion of resemblance to a distant ideal target, the phrase METHINKS IT IS LIKE A WEASEL. Life isn't like that. Evolution has no long-term goal. There is no long-distance target, no final perfection to serve as a criterion for selection, although human vanity cherishes the absurd notion that our species is the final goal of evolution. In real life, the criterion for selection is always short-term, either simple survival or, more generally, reproductive success. If, after the aeons, what looks like progress towards some distant goal seems, with hindsight, to have been achieved, this is always an incidental consequence of many generations of short-term selection. The ‘watchmaker’ that is cumulative natural selection is blind to the future and has no long-term goal.

We can change our computer model to take account of this point. We can also make it more realistic in other respects. Letters and words are peculiarly human manifestations, so let's make the computer draw pictures instead. Maybe we shall even see animal-like shapes evolving in the computer, by cumulative selection of mutant forms. We shan't prejudge the issue by building-in specific animal pictures to start with. We want them to emerge solely as a result of cumulative selection of random mutations.

In real life, the form of each individual animal is produced by embryonic development. Evolution occurs because, in successive generations, there are slight differences in embryonic development.  {51}  These differences come about because of changes (mutations — this is the small random element in the process that I spoke of) in the genes controlling development. In our computer model, therefore, we must have something equivalent to embryonic development, and something equivalent to genes that can mutate. There are many ways in which we could meet these specifications in a computer model. I chose one and wrote a program that embodied it. I shall now describe this computer model, because I think it is revealing. If you don’t know anything about computers, just remember that they are machines that do exactly what you tell them but often surprise you in the result. A list of instructions for a computer is called a program (this is standard American spelling, and it is also recommended by the Oxford Dictionary: the alternative, ‘programme’, commonly used in Britain, appears to be a Frenchified affectation).

Embryonic development is far too elaborate a process to simulate realistically on a small computer. We must represent it by some simplified analogue. We must find a simple picture-drawing rule that the computer can easily obey, and which can then be made to vary under the influence of ‘genes’. What drawing rule shall we choose? Textbooks of computer science often illustrate the power of what they call ‘recursive’ programming with a simple tree-growing procedure. The computer starts by drawing a single vertical line. Then the line branches into two. Then each of the branches splits into two sub-branches. Then each of the sub-branches splits into sub-sub-branches, and so on. It is ‘recursive’ because the same rule (in this case a branching rule) is applied locally all over the growing tree. No matter how big the tree may grow, the same branching rule goes on being applied at the tips of all its twigs.

The ‘depth’ of recursion, means the number of sub-sub-... branches that are allowed to grow, before the process is brought to a halt. Figure 2 shows what happens when you tell the computer to obey exactly the same drawing rule, but going on to various depths of recursion. At high levels of recursion the pattern becomes quite elaborate, but you can easily see in Figure 2 that it is still produced by the same very simple branching rule. This is, of course, just what happens in a real tree. The branching pattern of an oak tree or an apple free looks complex, but it really isn't. The basic branching rule is very simple. It is because it is applied recursively at the growing tips all over the tree — branches make sub-branches, then each sub-branch makes sub-sub-branches, and so on — that the whole tree ends up large and bushy.

Recursive branching is also a good metaphor for the embryonic development of plants and animals generally. I don’t mean that animal  {52} 

Figure 2

embryos look like branching trees. They don’t. But all embryos grow by cell division. Cells always split into two daughter cells. And genes always exert their final effects on bodies by means of local influences on cells, and on the two-way branching patterns of cell division. An animal's genes are never a grand design, a blueprint for the whole body. The genes, as we shall see, are more like a recipe than like a blueprint; and a recipe, moreover, that is obeyed not by the developing embryo as  {53}  a whole, but by each cell or each local cluster of dividing cells. I'm not denying that the embryo, and later the adult, has a large-scale form. But this large-scale form emerges because of lots of little local cellular effects all over the developing body, and these local effects consist primarily of two-way branchings, in the form of two-way cell splittings. It is by influencing these local events that genes ultimately exert influences on the adult body.

The simple branching rule for drawing trees, then, looks like a promising analogue for embryonic development. Accordingly, we wrap it up in a little computer procedure, label it DEVELOPMENT, and prepare to embed it in a larger program labelled EVOLUTION. As a first step towards writing this larger program, we now turn our attention to genes. How shall we represent ‘genes’ in our computer model? Genes in real life do two things. They influence development, and they get passed on to future generations. In real animals and plants there are tens of thousands of genes, but we shall modestly limit our computer model to nine. Each of the nine genes is simply represented by a number in the computer, which will be called its value. The value of a particular gene might be, say 4, or –7.

How shall we make these genes influence development? There are lots of things they could do. The basic idea is that they should exert some minor quantitative influence on the drawing rule that is DEVELOPMENT. For instance, one gene might influence the angle of branching, another might influence the length of some particular branch. Another obvious thing for a gene to do is to influence the depth of the recursion, the number of successive branchings. I made Gene 9 have this effect. You can regard Figure 2, therefore, as a picture of seven related organisms, identical to each other except with respect to Gene 9. I shan't spell out in detail what each one of the other eight genes does. You can get a general idea of the kinds of things they do by studying Figure 3. In the middle of the picture is the basic tree, one of the ones from Figure 2. Encircling this central tree are eight others. All are the same as the central tree, except that one gene, a different gene in each of the eight, has been changed — ‘mutated’. For instance, the picture to the right of the central tree shows what happens when Gene 5 mutates by having +1 added to its value. If there'd been room, I'd have liked to print a ring of 18 mutants around the central tree. The reason for wanting 18 is that there are nine genes, and each one can mutate in an ‘upward’ direction (1 is added to its value) or in a ‘downward’ direction (1 is subtracted from its value). So a ring of 18 trees would be enough to represent all possible single-step mutants that you can derive from the one central tree.  {54} 

Figure 3

Each of these trees has its own, unique ‘genetic formula’, the numerical values of its nine genes. I haven't written the genetic formulae down, because they wouldn't mean anything to you, in themselves. That is true of real genes too. Genes only start to mean something when they are translated, via protein synthesis, into growing-rules for a developing embryo. And in the computer model too, the numerical values of the nine genes only mean something when they are translated into growing rules for the branching tree pattern. But you can get an idea of what each gene does by comparing the bodies of two organisms known to differ with respect to a certain gene. Compare, for instance, the basic tree in the middle of the picture with the two trees on either side, and you'll get some idea of what Gene 5 does.

This, too, is exactly what real-life geneticists do. Geneticists normally don’t know how genes exert their effects on embryos. Nor do they know the complete genetic formula of any animal. But by  {55}  comparing the bodies of two adult animals that are known to differ according to a single gene, they can see what effects that single gene has. It is more complicated than that, because the effects of genes interact with each other in ways that are more complicated than simple addition. Exactly the same is true of the computer trees. Very much so, as later pictures will show.

You will notice that all the shapes are symmetrical about a left/right axis. This is a constraint that I imposed on the DEVELOPMENT procedure. I did it partly for aesthetic reasons; partly to economize on the number of genes necessary (if genes didn’t exert mirror-image effects on the two sides of the tree, we'd need separate genes for the left and the right sides); and partly because I was hoping to evolve animal-like shapes, and most animal bodies are pretty symmetrical. For the same reason, from now on I shall stop calling these creatures ‘trees’, and shall call them ‘bodies’ or ‘biomorphs’. Biomorph is the name coined by Desmond Morris for the vaguely animal-like shapes in his surrealist paintings. These paintings have a special place in my affections, because one of them was reproduced on the cover of my first book. Desmond Morris claims that his biomorphs ‘evolve’ in his mind, and that their evolution can be traced through successive paintings.

Back to the computer biomorphs, and the ring of 18 possible mutants, of which a representative eight are drawn in Figure 3. Since each member of the ring is only one mutational step away from the central biomorph, it is easy for us to see them as children of the central parent. We have our analogue of REPRODUCTION, which, like DEVELOPMENT, we can wrap up in another small computer program, ready to embed in our big program called EVOLUTION. Note two things about REPRODUCTION. First, there is no sex; reproduction is asexual. I think of the biomorphs as female, therefore, because asexual animals like greenfly are nearly always basically female in form. Second, my mutations are all constrained to occur one at a time. A child differs from its parent at only one of the nine genes; moreover, all mutation occurs by +1 or –1 being added to the value of the corresponding parental gene. These are just arbitrary conventions: they could have been different and still remained biologically realistic.

The same is not true of the following feature of the model, which embodies a fundamental principle of biology. The shape of each child is not derived directly from the shape of the parent. Each child gets its shape from the values of its own nine genes (influencing angles, distances, and so on). And each child gets its nine genes from its parent's nine genes. This is just what happens in real life. Bodies don’t get passed down the generations; genes do. Genes influence embryonic  {56}  development of the body in which they are sitting. Then those same genes either get passed on to the next generation or they don’t. The nature of the genes is unaffected by their participation in bodily development, but their likelihood of being passed on may be affected by the success of the body that they helped to create. This is why, in the computer model, it is important that the two procedures called DEVELOPMENT and REPRODUCTION are written as two watertight compartments. They are watertight except that REPRODUCTION passes gene values across to DEVELOPMENT, where they influence the growing rules. DEVELOPMENT most emphatically does not pass gene values back to REPRODUCTION — that would be tantamount to ‘Lamarckism’ (see Chapter 11).

We have assembled our two program modules, then, labelled DEVELOPMENT and REPRODUCTION. REPRODUCTION passes genes down the generations, with the possibility of mutation. DEVELOPMENT takes the genes provided by REPRODUCTION in any given generation, and translates those genes into drawing action, and hence into a picture of a body on the computer screen. The time has come to bring the two modules together in the big program called EVOLUTION.

EVOLUTION basically consists of endless repetition of REPRODUCTION. In every generation, REPRODUCTION takes the genes that are supplied to it by the previous generation, and hands them on to the next generation but with minor random errors — mutations. A mutation simply consists in +1 or –1 being added to the value of a randomly chosen gene. This means that, as the generations go by, the total amount of genetic difference from the original ancestor can become very large, cumulatively, one small step at a time. But although the mutations are random, the cumulative change over the generations is not random. The progeny in any one generation are different from their parent in random directions. But which of those progeny is selected to go forward into the next generation is not random. This is where Darwinian selection comes in. The criterion for selection is not the genes themselves, but the bodies whose shape the genes influence through DEVELOPMENT.

In addition to being REPRODUCED, the genes in each generation are also handed to DEVELOPMENT, which grows the appropriate body on the screen, following its own strictly laid-down rules. In every generation, a whole ‘litter’ of ‘children’ (i.e. individuals of the next generation) is displayed. All these children are mutant children of the same parent, differing from their parent with respect to one gene each. This very high mutation rate is a distinctly unbiological feature of the  {57}  computer model. In real life, the probability that a gene will mutate is often less than one in a million. The reason for building a high mutation rate into the model is that the whole performance on the computer screen is for the benefit of human eyes, and humans haven't the patience to wait a million generations for a mutation!

The human eye has an active role to play in the story. It is the selecting agent. It surveys the litter of progeny and chooses one for breeding. The chosen one then becomes the parent of the next generation, and a litter of its mutant children are displayed simultaneously on the screen. The human eye is here doing exactly what it does in the breeding of pedigree dogs or prize roses. Our model, in other words, is strictly a model of artificial selection, not natural selection. The criterion for ‘success’ is not the direct criterion of survival, as it is in true natural selection. In true natural selection, if a body has what it takes to survive, its genes automatically survive because they are inside it. So the genes that survive tend to be, automatically, those genes that confer on bodies the qualities that assist them to survive. In the computer model, on the other hand, the selection criterion is not survival, but the ability to appeal to human whim. Not necessarily idle, casual whim, for we can resolve to select consistently for some quality such as ‘resemblance to a weeping willow’. In my experience, however, the human selector is more often capricious and opportunistic. This, too, is not unlike certain kinds of natural selection.

The human tells the computer which one of the current litter of progeny to breed from. The genes of the chosen one are passed across to REPRODUCTION, and a new generation begins. This process, like real-life evolution, goes on indefinitely. Each generation of biomorphs is only a single mutational step away from its predecessor and its successor. But after 100 generations of EVOLUTION, the biomorphs can be anything up to 100 mutational steps away from their original ancestor. And in 100 mutational steps, much can happen.

I never dreamed how much, when I first started to play with my newly written EVOLUTION program. The main thing that surprised me was that the biomorphs can pretty quickly cease to look like trees. The basic two-way branching structure is always there, but it is easily smothered as lines cross and recross one another, making solid masses of colour (only black and white in the printed pictures). Figure 4 shows one particular evolutionary history consisting of no more than 29 generations. The ancestor is a tiny creature, a single dot. Although the ancestor's body is a dot, like a bacterium in the primeval slime, hidden inside it is the potential for branching in exactly the pattern of the  {58} 

Figure 4

 {59}  central tree of Figure 3: it is just that its Gene 9 tells it to branch zero times! All the creatures pictured on the page are descended from the dot but, in order to avoid cluttering the page, I haven't printed all the descendants that I actually saw. I've printed only the successful child of each generation (i.e. the parent of the next generation) and one or two of its unsuccessful sisters. So, the picture basically shows just the one main line of evolution, guided by my aesthetic selection. All the stages in the main line are shown.

Let's briefly go through the first few generations of the main line of evolution in Figure 4. The dot becomes a Y in generation 2. In the next two generations, the Y becomes larger. Then the branches become slightly curved, like a well-made catapult. In generation 7, the curve is accentuated, so that the two branches almost meet. The curved branches get bigger, and each acquires a couple of small appendages in generation 8. In generation 9 these appendages are lost again, and the stem of the catapult becomes longer. Generation 10 looks like a section through a flower; the curved side-branches resemble petals cupping a central appendage or ‘stigma’. In generation 11, the same ‘flower’ shape has become bigger and slightly more complicated.

I won't pursue the narrative. The picture speaks for itself, on through the 29 generations. Notice how each generation is just a little different from its parent and from its sisters. Since each is a little different from its parent, it is only to be expected that each will be slightly more different from its grandparents (and its grandchildren), and even more different still from its great grandparents (and great grandchildren). This is what cumulative evolution is all about, although, because of our high mutation rate, we have speeded it up here to unrealistic rates. Because of this, Figure 4 looks more like a pedigree of species than a pedigree of individuals, but the principle is the same.

When I wrote the program, I never thought that it would evolve anything more than a variety of tree-like shapes. I had hoped for weeping willows, cedars of Lebanon, Lombardy poplars, seaweeds, perhaps deer antlers. Nothing in my biologist's intuition, nothing in my 20 years’ experience of programming computers, and nothing in my wildest dreams, prepared me for what actually emerged on the screen. I can't remember exactly when in the sequence it first began to dawn on me that an evolved resemblance to something like an insect was possible. With a wild surmise, I began to breed, generation after generation, from whichever child looked most like an insect. My incredulity grew in parallel with the evolving resemblance. You see the eventual results at the bottom of Figure 4. Admittedly they have  {60}  eight legs like a spider, instead of six like an insect, but even so! I still cannot conceal from you my feeling of exultation as I first watched these exquisite creatures emerging before my eyes. I distinctly heard the triumphal opening chords of Also sprach Zarathustia (the ‘2001 theme’) in my mind. I couldn't eat, and that night ‘my’ insects swarmed behind my eyelids as I tried to sleep.

There are computer games on the market in which the player has the illusion that he is wandering about in an underground labyrinth, which has a definite if complex geography and in which he encounters dragons, minotaurs or other mythic adversaries. In these games the monsters are rather few in number. They are all designed by a human programmer, and so is the geography of the labyrinth. In the evolution game, whether the computer version or the real thing, the player (or observer) obtains the same feeling of wandering metaphorically through a labyrinth of branching passages, but the number of possible pathways is all but infinite, and the monsters that one encounters are undesigned and unpredictable. On my wanderings through the backwaters of Biomorph Land, I have encountered fairy shrimps, Aztec temples, Gothic church windows, aboriginal drawings of kangaroos, and, on one memorable but unrecapturable occasion, a passable caricature of the Wykeham Professor of Logic. Figure 5 is another little collection from my trophy room, all of which developed in the same kind of way. I want to emphasize that these shapes are not artists’ impressions. They have not been touched-up or doctored in any way whatever. They are exactly as the computer drew them when they evolved inside it. The role of the human eye was limited to selecting, among randomly mutated progeny over many generations of cumulative evolution.

We now have a much more realistic model of evolution than the monkeys typing Shakespeare gave us. But the biomorph model is still deficient. It shows us the power of cumulative selection to generate an almost endless variety of quasi-biological form, but it uses artificial selection, not natural selection. The human eye does the selecting. Could we dispense with the human eye, and make the computer itself do the selecting, on the basis of some biologically realistic criterion? This is more difficult than it may seem. It is worth spending a little time explaining why.  {61} 

Figure 5

It is trivially easy to select for a particular genetic formula, so long as you can read the genes of all the animals. But natural selection doesn't choose genes directly, it chooses the effects that genes have on bodies, technically called phenotypic effects. The human eye is good at choosing phenotypic effects, as is shown by the numerous breeds of dogs, cattle and pigeons, and also, if I may say so, as is shown by Figure 5. To make the computer choose phenotypic effects directly, we should have to write a very sophisticated pattern-recognition program. Pattern-recognizing programs exist. They are used to read print and even handwriting. But they are difficult, ‘state of the art’ programs, needing very large and fast computers. Even if such a pattern-recognition program were not beyond my programming capabilities, and beyond the capacity of my little 64-kilobyte computer, I wouldn't bother with it. This is a task that is better done by the human eye, together with — and this is more to the point — the 10-giganeurone computer inside the skull.

It wouldn't be too difficult to make the computer select for vague general features like, say, tall-thinness, short-fatness, perhaps curvaceousness, spikiness, even rococo ornamentation. One method would be to program the computer to remember the kinds of qualities that humans have favoured in the past, and to exert continued selection of the same general kind in the future. But this isn't getting  {62}  us any closer to simulating natural selection. The important point is that nature doesn't need computing power in order to select, except in special cases like peahens choosing peacocks. In nature, the usual selecting agent is direct, stark and simple. It is the grim reaper. Of course, the reasons for survival are anything but simple — that is why natural selection can build up animals and plants of such formidable complexity. But there is something very crude and simple about death itself. And nonrandom death is all it takes to select phenotypes, and hence the genes that they contain, in nature.

To simulate natural selection in an interesting way in the computer, we should forget about rococo ornamentation and all other visually defined qualities. We should concentrate, instead, upon simulating nonrandom death. Biomorphs should interact, in the computer, with a simulation of a hostile environment. Something about their shape should determine whether or not they survive in that environment. Ideally, the hostile environment should include other evolving biomorphs: ‘predators’, ‘prey’, ‘parasites’, ‘competitors’. The particular shape of a prey biomorph should determine its vulnerability to being caught, for example, by particular shapes of predator biomorphs. Such criteria of vulnerability should not be built in by the programmer. They should emerge, in the same kind of way as the shapes themselves emerge. Evolution in the computer would then really take off, for the conditions would be met for a self-reinforcing ‘arms race’ (see Chapter 7), and I dare not speculate where it would all end. Unfortunately, I think it may be beyond my powers as a programmer to set up such a counterfeit world.

If anybody is clever enough to do it, it would be the programmers who develop those noisy and vulgar arcade games — ‘Space Invaders’ derivatives. In these programs a counterfeit world is simulated. It has a geography, often in three dimensions, and it has a fast-moving time dimension. Entities zoom around in simulated three-dimensional space, colliding with each other, shooting each other down, swallowing each other amid revolting noises. So good can the simulation be that the player handling the joystick receives a powerful illusion that he himself is part of the counterfeit world. I imagine that the summit of this kind of programming is achieved in the chambers used to train aeroplane and spacecraft pilots. But even these programs are small-fry compared to the program that would have to be written to simulate an emerging arms race between predators and prey, embedded in a complete, counterfeit ecosystem. It certainly could be done, however. If there is a professional programmer out there who feels like collaborating on the challenge, I should like to hear from him or her.  {63} 

Meanwhile, there is something else that is much easier, and which I intend trying when summer comes. I shall put the computer in a shady place in the garden. The screen can display in colour. I already have a version of the program which uses a few more ‘genes’ to control colour, in the same kind of way as the other nine genes control shape. I shall begin with any more-or-less compact and brightly coloured biomorph. The computer will simultaneously display a range of mutant progeny of the biomorph, differing from it in shape and/or colour pattern. I believe that bees, butterflies and other insects will visit the screen, and ‘choose’ by bumping into a particular spot on the screen. When a certain number of choices have been logged, the computer will wipe the screen clean, ‘breed’ from the preferred biomorph, and display the next generation of mutant progeny.

I have high hopes that, over a large number of generations, the wild insects will actually cause the evolution, in the computer, of flowers. If they do, the computer flowers will have evolved under exactly the same selection pressure as caused real flowers to evolve in the wild. I am encouraged in my hope by the fact that insects frequently visit bright blobs of colour on women's dresses (and also by more systematic experiments that have been published). An alternative possibility, which I would find even more exciting, is that the wild insects might cause the evolution of insect-like shapes. The precedent for this — and hence the reason for hope — is that bees in the past caused the evolution of bee-orchids. Male bees, over many generations of cumulative orchid evolution, have built up the bee-like shape through trying to copulate with flowers, and hence carrying pollen. Imagine the ‘bee-flower’ of Figure 5 in colour. Wouldn't you fancy it if you were a bee?

My main reason for pessimism is that insect vision works in a very different way from ours. Video-screens are designed for human eyes, not bee eyes. This could easily mean that, although both we and bees see bee-orchids, in our very different ways, as bee-like, bees might not see video-screen images at all. Bees might see nothing but 625 scanning lines! Still, it is worth a try. By the time this book is published, I shall know the answer.

There is a popular cliche, usually uttered in the tones Stephen Potter would have called ‘plonking’, which says that you cannot get out of computers any more than you put in. Other versions are that computers only do exactly what you tell them to, and that therefore computers are never creative. The cliche is true only in a crashingly trivial sense, the same sense in which Shakespeare never wrote anything except what his first schoolteacher taught him to write — words. I  {64}  programmed EVOLUTION into the computer, but I did not plan ‘my’ insects, nor the scorpion, nor the spitfire, nor the lunar lander. I had not the slightest inkling that they would emerge, which is why ‘emerge’ is the right word. True, my eyes did the selecting that guided their evolution, but at every stage I was limited to a small clutch of progeny offered up by random mutation, and my selection ‘strategy’, such as it was, was opportunistic, capricious and short-term. I was not aiming for any distant target, and nor does natural selection.

I can dramatize this by discussing the one time when I did try to aim for a distant target. First I must make a confession. You will have guessed it anyway. The evolutionary history of Figure 4 is a reconstruction. It was not the first time I had seen ‘my’ insects. When they originally emerged to the sound of trumpets, I had no means of recording their genes. There they were, sitting on the computer screen, and I couldn't get at them, couldn't decipher their genes. I delayed switching the computer off while I racked my brain trying to think of some way of saving them, but there was none. The genes were too deeply buried, just as they are in real life. I could print out pictures of the insects’ bodies, but I had lost their genes. I immediately modified the program so that in future it would keep accessible records of genetic formulae, but it was too late. I had lost my insects.

I set about trying to ‘find’ them again. They had evolved once, so it seemed that it must be possible to evolve them again. Like the lost chord, they haunted me. I wandered through Biomorph Land, moving through an endless landscape of strange creatures and things, but I couldn't find my insects. I knew that they must be lurking there somewhere. I knew the genes from which the original evolution had started. I had a picture of my insects’ bodies. I even had a picture of the evolutionary sequence of bodies leading up to my insects by slow degrees from a dot ancestor. But I didn’t know their genetic formula.

You might think that it would have been easy enough to reconstruct the evolutionary pathway, but it wasn't. The reason, which I shall come back to, is the astronomical number of possible biomorphs that a sufficiently long evolutionary pathway can offer, even when there are only nine genes varying. Several times on my pilgrimage through Biomorph Land I seemed to come close to a precursor of my insects, but, then, in spite of my best efforts as a selecting agent, evolution went off on what proved to be a false trail. Eventually, during my evolutionary wanderings through Biomorph Land — the sense of triumph was scarcely less than on the first occasion — I finally cornered them again. I didn’t know (still don’t) if these insects were exactly the same as my original, ‘lost chords of Zarathustra’ insects, or whether  {65}  they were superficially ‘convergent’ (see next chapter), but it was good enough. This time there was no mistake: I wrote down the genetic formula, and now I can ‘evolve’ insects whenever I want.

Yes I am piling on the drama a bit, but there is a serious point being made. The point of the story is that even though it was I that programmed the computer, telling it in great detail what to do, nevertheless I didn’t plan the animals that evolved, and I was totally surprised by them when I first saw their precursors. So powerless was I to control the evolution that, even when I very much wanted to retrace a particular evolutionary pathway it proved all but impossible to do so. I don’t believe I would ever have found my insects again if I hadn't had a printed picture of the complete set of their evolutionary precursors, and even then it was difficult and tedious. Does the powerlessness of the programmer to control or predict the course of evolution in the computer seem paradoxical? Does it mean that something mysterious, even mystical was going on inside the computer? Of course not. Nor is there anything mystical going on in the evolution of real animals and plants. We can use the computer model to resolve the paradox, and learn something about real evolution in the process.

To anticipate, the basis of the resolution of the paradox will turn out to be as follows. There is a definite set of biomorphs, each permanently sitting in its own unique place in a mathematical space. It is permanently sitting there in the sense that, if only you knew its genetic formula, you could instantly find it; moreover, its neighbours in this special kind of space are the biomorphs that differ from it by only one gene. Now that I know the genetic formula of my insects, I can reproduce them at will, and I can tell the computer to ‘evolve’ towards them from any arbitrary starting point. When you first evolve a new creature by artificial selection in the computer model, it feels like a creative process. So it is, indeed. But what you are really doing is finding the creature, for it is, in a mathematical sense, already sitting in its own place in the genetic space of Biomorph Land. The reason it is a truly creative process is that finding any particular creature is extremely difficult, simply and purely because Biomorph Land is very very large, and the total number of creatures sitting there is all but infinite. It isn't feasible just to search aimlessly and at random. You have to adopt some more efficient — creative — searching procedure.

Some people fondly believe that chess-playing computers work by internally trying out all possible combinations of chess moves. They find this belief comforting when a computer beats them, but their belief is utterly false. There are far too many possible chess moves: the search-space is billions of times too large to allow blind stumbling to  {66}  succeed. The art of writing a good chess program is thinking of efficient short cuts through the search-space. Cumulative selection, whether artificial selection as in the computer model or natural selection out there in the real world, is an efficient searching procedure, and its consequences look very like creative intelligence. That, after all, is what William Paley's Argument from Design was all about. Technically, all that we are doing, when we play the computer biomorph game, is finding animals that, in a mathematical sense, are waiting to be found. What it feels like is a process of artistic creation. Searching a small space, with only a few entities in it, doesn't ordinarily feel like a creative process. A child's game of hunt the thimble doesn't feel creative. Turning things over at random and hoping to stumble on the sought object usually works when the space to be searched is small. As the search-space gets larger, more and more sophisticated searching procedures become necessary. Effective searching procedures become, when the search-space is sufficiently large, indistinguishable from true creativity.

The computer biomorph models make these points well, and they constitute an instructive bridge between human creative processes, such as planning a winning strategy at chess, and the evolutionary creativity of natural selection, the blind watchmaker. To see this, we must develop the idea of Biomorph Land as a mathematical ‘space’, an endless but orderly vista of morphological variety, but one in which every creature is sitting in its correct place, waiting to be discovered. The 17 creatures of Figure 5 are arranged in no special order on the page. But in Biomorph Land itself each occupies its own unique position, determined by its genetic formula, surrounded by its own particular neighbours. All the creatures in Biomorph Land have a definite spatial relationship one to another. What does that mean? What meaning can we attach to spatial position?

The space we are talking about is genetic space. Each animal has its own position in genetic space. Near neighbours in genetic space are animals that differ from one another by only a single mutation. In Figure 3, the basic tree in the centre is surrounded by 8 of its 18 immediate neighbours in genetic space. The 18 neighbours of an animal are the 18 different kinds of children that it can give rise to, and the 18 different kinds of parent from which it could have come, given the rules of our computer model. At one remove, each animal has 324 (18 × 18, ignoring back-mutations for simplicity) neighbours, the set of its possible grandchildren, grandparents, aunts or nieces. At one remove again, each animal has 5,832 (18 × 18 × 18) neighbours, the set of possible great grandchildren, great grandparents, first cousins, etc.  {67} 

What is the point of thinking in terms of genetic space? Where does it get us? The answer is that it provides us with a way to understand evolution as a gradual, cumulative process. In any one generation, according to the rules of the computer model, it is possible to move only a single step through genetic space. In 29 generations it isn't possible to move farther than 29 steps, in genetic space, away from the starting ancestor. Every evolutionary history consists of a particular pathway, or trajectory, through genetic space. For instance, the evolutionary history recorded in Figure 4 is a particular winding trajectory through genetic space, connecting a dot to an insect, and passing through 28 intermediate stages. It is this that I mean when I talk metaphorically about ‘wandering’ through Biomorph Land.

I wanted to try to represent this genetic space in the form of a picture. The trouble is, pictures are two-dimensional. The genetic space in which the biomorphs sit is not two-dimensional space. It isn't even three-dimensional space. It is nine-dimensional space! (The important thing to remember about mathematics is not to be frightened. It isn't as difficult as the mathematical priesthood sometimes pretends. Whenever I feel intimidated, I always remember Silvanus Thompson's dictum in Calculus Made Easy: ‘What one fool can do, another can’.) If only we could draw in nine dimensions we could make each dimension correspond to one of the nine genes. The position of a particular animal, say the scorpion or the bat or the insect, is fixed in genetic space by the numerical value of its nine genes. Evolutionary change consists of a step by step walk through nine-dimensional space. The amount of genetic difference between one animal and another, and hence the time taken to evolve, and the difficulty of evolving from one to the other, is measured as the distance in nine-dimensional space from one to the other.

Alas, we can't draw in nine dimensions. I sought a way of fudging it, of drawing a two-dimensional picture that conveyed something of what it feels like to move from point to point in the nine-dimensional genetic space of Biomorph Land. There are various possible ways in which this could be done, and I chose one that I call the triangle trick. Look at Figure 6. At the three corners of the triangle are three arbitrarily chosen biomorphs. The one at the top is the basic tree, the one on the left is one of ‘my’ insects, and the one on the right has no name but I thought it looked pretty. Like all biomorphs, each of these three has its own genetic formula, which determines its unique position in nine-dimensional genetic space.

The triangle lies on a flat two-dimensional ‘plane’ that cuts through the nine-dimensional hypervolume (what one fool can do, another  {68} 

Figure 6

can). The plane is like a flat piece of glass stuck through a jelly. On the glass is drawn the triangle, and also some of the biomorphs whose genetic formulae entitle them to sit on that particular flat plane. What is it that entitles them? This is where the three biomorphs at the corners of the triangle come in. They are called the anchor biomorphs.

Remember that the whole idea of ‘distance’ in genetic ‘space’ is that genetically similar biomorphs are near neighbours, genetically different biomorphs are distant neighbours. On this particular plane, the distances are all calculated with reference to the three anchor biomorphs. For any given point on the sheet of glass, whether inside the triangle or outside it, the appropriate genetic formula for that point is calculated as a ‘weighted average’ of the genetic formulae of the three anchor biomorphs. You will already have guessed how the weighting is done. It is done by the distances on the page, more precisely the nearnesses, from the point in question to the three anchor biomorphs. So, the nearer you are to the insect on the plane, the more insect-like are the local biomorphs. As you move along the glass towards the tree, the ‘insects’ gradually become less insect-like and more tree-like. If you walk into the centre of the triangle the animals that you find there, for instance the spider with a Jewish seven-branched candelabra on its head, will be various ‘genetic compromises’ between the three anchor biomorphs.

But this account gives altogether too much prominence to the three anchor biomorphs. Admittedly the computer did use them to calculate  {69}  the appropriate genetic formula for every point on the picture. But actually any three anchor points on the plane would have done the trick just as well, and would have given identical results. For this reason, in Figure 7 I haven't actually drawn the triangle. Figure 7 is exactly the same kind of picture as Figure 6. It just shows a different plane. The same insect is one of the three anchor points, this time the right-hand one. The other anchor points, in this case, are the spitfire and the bee-flower, both as seen in Figure 5. On this plane, too, you will notice that neighbouring biomorphs resemble each other more than distant biomorphs. The spitfire, for instance, is part of a squadron of similar aircraft, flying in formation. Because the insect is on both sheets of glass, you can think of the two sheets as passing, at an angle, through each other. Relative to Figure 6, the plane of Figure 7 is said to be ‘rotated about’ the insect.

Figure 7

The elimination of the triangle is an improvement to our method, because it was a distraction. It gave undue prominence to three particular points in the plane. We still have one further improvement to make. In Figure 6 and 7, spatial distance represents genetic distance, but the scaling is all distorted. One inch upwards is not necessarily equivalent to one inch across. To remedy this, we must choose our three anchor biomorphs carefully, so that their genetic distances, one from the other, are all the same. Figure 8 does just this. Again the triangle is not actually drawn. The three anchors are the scorpion from Figure 5, the insect again (we have yet another ‘rotation about’ the  {70}  insect), and the rather nondescript biomorph at the top. These three biomorphs are all 30 mutations distant from each other. This means that it is equally easy to evolve from any one to any other one. In all three cases, a minimum of 30 genetic steps must be taken. The little blips along the lower margin of Figure 8 represent units of distance measured in genes. You can think of it as a genetic ruler. The ruler doesn't only work in the horizontal direction. You can tilt it in any direction, and measure the genetic distance, and hence the minimum evolution time, between any point on the plane and any other (annoyingly, that is not quite true on the page, because the computer's printer distorts proportions, but this effect is too trivial to make a fuss about, although it does mean that you will get slightly the wrong answer if you simply count blips on the scale).

Figure 8

These two-dimensional planes cutting through nine-dimensional genetic space give some feeling for what it means to walk through Biomorph Land. To improve that feeling, you have to remember that evolution is not restricted to one flat plane. On a true evolutionary walk you could ‘drop through’, at any time, to another plane, for instance from the plane of Figure 6 to the plane of Figure 7 (in the vicinity of the insect, where the two planes come close to each other).

I said that the ‘genetic ruler’ of Figure 8 enables us to calculate the  {71}  minimum time it would take to evolve from one point to another. So it does, given the restrictions of the original model, but the emphasis is on the word minimum. Since the insect and the scorpion are 30 genetic units distant from one another, it takes only 30 generations to evolve from one to the other if you never take a wrong turning, if, that is, you know exactly what genetic formula you are heading towards, and how to steer towards it. In real-life evolution there is nothing that corresponds to steering towards some distant genetic target.

Let's now use the biomorphs to return to the point made by the monkeys typing Hamlet, the importance of gradual, step-by-step change in evolution, as opposed to pure chance. Begin by relabelling the graticules along the bottom of Figure 8, but in different units. Instead of measuring distance as ‘number of genes that have to change in evolution’, we are going to measure distance as ‘odds of happening to jump the distance, by sheer luck, in a single hop’. To think about this, we now have to relax one of the restrictions that I built into the computer game: we shall end by seeing why I built that restriction in in the first place. The restriction was that children were only ‘allowed’ to be one mutation distant from their parents. In other words, only one gene was allowed to mutate at a time, and that gene was allowed to change its ‘value’ only by +1 or –1. By relaxing the restriction, we are now allowing any number of genes to mutate simultaneously, and they can add any number, positive or negative, to their current value. Actually, that is too great a relaxation, since it allows genetic values to range from minus infinity to plus infinity. The point is adequately made if we restrict gene values to single figures, that is if we allow them to range from –9 to +9.

So, within these wide limits, we are theoretically allowing mutation, at a stroke, in a single generation, to change any combination of the nine genes. Moreover, the value of each gene can change any amount, so long as it doesn't stray into double figures. What does this mean? It means that, theoretically, evolution can jump, in a single generation, from any point in Biomorph Land to any other. Not just any point on one plane, but any point in the entire nine-dimensional hypervolume. If, for instance, you should want to jump in one fell swoop from the insect to the fox in Figure 5, here is the recipe. Add the following numbers to the values of Genes 1 to 9, respectively: –2,2,2,–2,2,0,–4,–1,1. But since we are talking about random jumps, all points in Biomorph Land are equally likely as destinations for one of these jumps. So, the odds against jumping to any particular destination, say the fox, by sheer luck, are easy to calculate. They are simply the total number of biomorphs in the space. As you can see, we  {72}  are embarking on another of those astronomical calculations. There are nine genes, and each of them can take any of 19 values. So the total number of biomorphs that we could jump to in a single step is 19 times itself 9 times over: 19 to the power 9. This works out as about half a trillion biomorphs. Paltry compared with Asimov's ‘haemoglobin number’, but still what I would call a large number. If you started from the insect, and jumped like a demented flea half a trillion times, you could expect to arrive at the fox once.

What is all this telling us about real evolution? Once again, it is ramming home the importance of gradual, step-by-step change. There have been evolutionists who have denied that gradualism of this kind is necessary in evolution. Our biomorph calculation shows us exactly one reason why gradual, step-by-step change is important. When I say that you can expect evolution to jump from the insect to one of its immediate neighbours, but not to jump from the insect directly to the fox or the scorpion, what I exactly mean is the following. If genuinely random jumps really occurred, then a jump from insect to scorpion would be perfectly possible. Indeed it would be just as probable as a jump from insect to one of its immediate neighbours. But it would also be just as probable as a jump to any other biomorph in the land. And there's the rub. For the number of biomorphs in the land is half a trillion, and if no one of them is any more probable as a destination than any other, the odds of jumping to any particular one are small enough to ignore.

Notice that it doesn't help us to assume that there is a powerful nonrandom ‘selection pressure’. It wouldn't matter if you'd been promised a king's ransom if you achieved a lucky jump to the scorpion. The odds against your doing so are still half a trillion to one. But if, instead of jumping you walked, one step at a time, and were given one small coin as a reward every time you happened to take a step in the right direction, you would reach the scorpion in a very short time. Not necessarily in the fastest possible time of 30 generations, but very fast, nevertheless. Jumping could theoretically get you the prize faster — in a single hop. But because of the astronomical odds against success, a series of small steps, each one building on the accumulated success of previous steps, is the only feasible way.

The tone of my previous paragraphs is open to a misunderstanding which I must dispel. It sounds, once again, as though evolution deals in distant targets, homing in on things like scorpions. As we have seen, it never does. But if we think of our target as anything that would improve survival chances, the argument still works. If an animal is a parent, it must be good enough to survive at least to adulthood. It is  {73}  possible that a mutant child of that parent might be even better at surviving. But if a child mutates in a big way, so that it has moved a long distance away from its parent in genetic space, what are the odds of its being better than its parent? The answer is that the odds against are very large indeed. And the reason is the one we have just seen with our biomorph model. If the mutational jump we are considering is a very large one, the number of possible destinations of that jump is astronomically large. And because, as we saw in Chapter 1, the number of different ways of being dead is so much greater than the number of different ways of being alive, the chances are very high that a big random jump in genetic space will end in death. Even a small random jump in genetic space is pretty likely to end in death. But the smaller the jump the less likely death is, and the more likely is it that the jump will result in improvement. We shall return to this theme in a later chapter.

That is as far as I want to go in drawing morals from Biomorph Land. I hope that you didn’t find it too abstract. There is another mathematical space filled, not with nine-gened biomorphs but with flesh and blood animals made of billions of cells, each containing tens of thousands of genes. This is not biomorph space but real genetic space. The actual animals that have ever lived on Earth are a tiny subset of the theoretical animals that could exist. These real animals are the products of a very small number of evolutionary trajectories through genetic space. The vast majority of theoretical trajectories through animal space give rise to impossible monsters. Real animals are dotted around here and there among the hypothetical monsters, each perched in its own unique place in genetic hyperspace. Each real animal is surrounded by a little cluster of neighbours, most of whom have never existed, but a few of whom are its ancestors, its descendants and its cousins.

Sitting somewhere in this huge mathematical space are humans and hyenas, amoebas and aardvarks, flatworms and squids, dodos and dinosaurs. In theory, if we were skilled enough at genetic engineering, we could move from any point in animal space to any other point. From any starting point we could move through the maze in such a way as to recreate the dodo, the tyrannosaur and trilobites. If only we knew which genes to tinker with, which bits of chromosome to duplicate, invert or delete. I doubt if we shall ever know enough to do it, but these dear dead creatures are lurking there forever in their private corners of that huge genetic hypervolume, waiting to be found if we but had the knowledge to navigate the right course through the maze. We might even be able to evolve an exact reconstruction of a  {74}  dodo by selectively breeding pigeons, though we'd have to live a million years in order to complete the experiment. But when we are prevented from making a journey in reality, the imagination is not a bad substitute. For those, like me, who are not mathematicians, the computer can be a powerful friend to the imagination. Like mathematics, it doesn't only stretch the imagination. It also disciplines and controls it.

CHAPTER 4

MAKING TRACKS THROUGH
ANIMAL SPACE

As we saw in Chapter 2, many people find it hard to believe that something like the eye, Paley's favourite example, so complex and well designed, with so many interlocking working parts, could have arisen from small beginnings by a gradual series of step-by-step changes. Let's return to the problem in the light of such new intuitions as the biomorphs may have given us. Answer the following two questions:

The answer to Question 1 is clearly a decisive no. The odds against a ‘yes’ answer for questions like Question 1 are many billions of times greater than the number of atoms in the universe. It would need a gigantic and vanishingly improbable leap across genetic hyperspace. The answer to Question 2 is equally clearly yes, provided only that the difference between the modern eye and its immediate predecessor X is sufficiently small. Provided, in other words, that they are sufficiently close to one another in the space of all possible structures. If the answer to Question 2 for any particular degree of difference is no, all we have to do is repeat the question for a smaller degree of difference. Carry on doing this until we find a degree of difference sufficiently small to give us a ‘yes’ answer to Question 2.

X is defined as something very like a human eye, sufficiently similar that the human eye could plausibly have arisen by a single alteration in X. If you have a mental picture of X and you find it  {78}  implausible that the human eye could have arisen directly from it, this simply means that you have chosen the wrong X. Make your mental picture of X progressively more like a human eye, until you find an X that you do find plausible as an immediate predecessor to the human eye. There has to be one for you, even if your idea of what is plausible may be more, or less, cautious than mine!

Now, having found an X such that the answer to Question 2 is yes, we apply the same question to X itself. By the same reasoning we must conclude that X could plausibly have arisen, directly by a single change, from something slightly different again, which we may call X'. Obviously we can then trace X' back to something else slightly different from it, X", and so on. By interposing a large enough series of Xs, we can derive the human eye from something not slightly different from itself but very different from itself. We can ‘walk’ a large distance across ‘animal space’, and our move will be plausible provided we take small-enough steps. We are now in a position to answer a third question.

It seems to me clear that the answer has to be yes, provided only that we allow ourselves a sufficiently large series of Xs. You might feel that 1,000 Xs is ample, but if you need more steps to make the total transition plausible in your mind, simply allow yourself to assume 10,000 Xs. And if 10,000 is not enough for you, allow yourself 100,000, and so on. Obviously the available time imposes an upper ceiling on this game, for there can be only one X per generation. In practice the question therefore resolves itself into: Has there been enough time for enough successive generations? We can't give a precise answer to the number of generations that would be necessary. What we do know is that geological time is awfully long. Just to give you an idea of the order of magnitude we are talking about, the number of generations that separate us from our earliest ancestors is certainly measured in the thousands of millions. Given, say, a hundred million Xs, we should be able to construct a plausible series of tiny gradations linking a human eye to just about anything!

So far, by a process of more-or-less abstract reasoning, we have concluded that there is a series of imaginable Xs, each sufficiently similar to its neighbours that it could plausibly turn into one of its neighbours, the whole series linking the human eye back to no eye at all. But we still haven't demonstrated that it is plausible that this series of Xs actually existed. We have two more questions to answer.  {79} 

This is really a question about embryology, not genetics; and it is an entirely separate question from the one that worried the Bishop of Birmingham and others. Mutation has to work by modifying the existing processes of embryonic development. It is arguable that certain kinds of embryonic process are highly amenable to variation in certain directions, recalcitrant to variation in others. I shall return to this matter in Chapter 11, so here I'll just stress again the difference between small change and large. The smaller the change you postulate, the smaller the difference between X" and X', the more embryologically plausible is the mutation concerned. In the previous chapter we saw, on purely statistical grounds, that any particular large mutation is inherently less probable than any particular small mutation. Whatever problems may be raised by Question 4, then, we can at least see that the smaller we make the difference between any given X' and X", the smaller will be the problems. My feeling is that, provided the difference between neighbouring intermediates in our series leading to the eye is sufficiently small, the necessary mutations are almost bound to be forthcoming. We are, after all, always talking about minor quantitative changes in an existing embryonic process. Remember that, however complicated the embryological status quo may be in any given generation, each mutational change in the status quo can be very small and simple.

We have one final question to answer:

Rather oddly, some people have thought that the answer to this question is a self-evident ‘no’. For instance, I quote from Francis Hitching's book of 1982 called The Neck of the Giraffe or Where Darwin Went Wrong. I could have quoted basically the same words from almost any Jehovah's Witness tract, but I choose this book because a reputable publisher (Pan Books Ltd) saw fit to publish it, despite a very large number of errors which would quickly have been spotted if an unemployed biology graduate, or indeed undergraduate, had been asked to glance through the manuscript. (My favourites, if you'll indulge me just two in-jokes, are the conferring of a knighthood  {80}  on Professor John Maynard Smith, and the description of Professor Ernst Mayr, that eloquent and most unmathematical arch-critic of mathematical genetics, as ‘the high priest’ of mathematical genetics.)

For the eye to work the following minimum perfectly coordinated steps have to take place (there are many others happening simultaneously, but even a grossly simplified description is enough to point up the problems for Darwinian theory). The eye must be clean and moist, maintained in this state by the interaction of the tear gland and movable eyelids, whose eyelashes also act as a crude filter against the sun. The light then passes through a small transparent section of the protective outer coating (the cornea), and continues via a lens which focuses it on the back of the retina. Here 130 million light-sensitive rods and cones cause photochemical reactions which transform the light into electrical impulses. Some 1,000 million of these are transmitted every second, by means that are not properly understood, to a brain which then takes appropriate action.

Now it is quite evident that if the slightest thing goes wrong en route — the cornea is fuzzy, or the pupil fails to dilate, or the lens becomes opaque, or the focussing goes wrong — then a recognizable image is not formed. The eye either functions as a whole, or not at all. So how did it come to evolve by slow, steady, infinitesimally small Darwinian improvements? Is it really plausible that thousands upon thousands of lucky chance mutations happened coincidentally so that the lens and the retina, which cannot work without each other, evolved in synchrony? What survival value can there be in an eye that doesn't see?

This remarkable argument is very frequently made, presumably because people want to believe its conclusion. Consider the statement that ‘if the slightest thing goes wrong... if the focusing goes wrong... a recognizable image is not formed’. The odds cannot be far from 50/50 that you are reading these words through glass lenses. Take them off and look around. Would you agree that ‘a recognizable image is not formed’? If you are male, the odds are about 1 in 12 that you are colourblind. You may well be astigmatic. It is not unlikely that, without glasses, your vision is a misty blur. One of today's most distinguished (though not yet knighted) evolutionary theorists so seldom cleans his glasses that his vision is probably a misty blur anyway, but he seems to get along pretty well and, by his own account, he used to play a mean game of monocular squash. If you have lost your glasses, it may be that you upset your friends by failing to recognize them in the street. But you yourself would be even more upset if somebody said to you: ‘Since your vision is now not absolutely perfect, you might as well go around with your eyes tight shut until you find your glasses again.’ Yet that is essentially what the author of the passage I have quoted is suggesting.  {81} 

He also states, as though it were obvious, that the lens and the retina cannot work without each other. On what authority? Someone close to me has had a cataract operation in both eyes. She has no lenses in her eyes at all. Without glasses she couldn't even begin to play lawn tennis or aim a rifle. But she assures me that you are far better off with a lensless eye than with no eye at all. You can tell if you are about to walk into a wall or another person. If you were a wild creature, you could certainly use your lensless eye to detect the looming shape of a predator, and the direction from which it was approaching. In a primitive world where some creatures had no eyes at all and others had lensless eyes, the ones with lensless eyes would have all sorts of advantages. And there is a continuous series of Xs, such that each tiny improvement in sharpness of image, from swimming blur to perfect human vision, plausibly increases the organism's chances of surviving.

The book goes on to quote Stephen Jay Gould, the noted Harvard palaeontologist, as saying:

We avoid the excellent question, ‘What good is 5 percent of an eye?’ by arguing that the possessor of such an incipient structure did not use it for sight.

An ancient animal with 5 per cent of an eye might indeed have used it for something other than sight, but it seems to me at least as likely that it used it for 5 per cent vision. And actually I don’t think it is an excellent question. Vision that is 5 per cent as good as yours or mine is very much worth having in comparison with no vision at all. So is 1 per cent vision better than total blindness. And 6 per cent is better than 5, 7 per cent better than 6, and so on up the gradual, continuous series.

This kind of problem has worried some people interested in animals that gain protection from predators by ‘mimicry’. Stick insects look like sticks and so are saved from being eaten by birds. Leaf insects look like leaves. Many edible species of butterfly gain protection by resembling noxious or poisonous species. These resemblances are far more impressive than the resemblance of clouds to weasels. In many cases they are more impressive than the resemblance of ‘my’ insects to real insects. Real insects, after all, have six legs, not eight! Real natural selection has had at least a million times as many generations as I had, in which to perfect the resemblance.

We use the word ‘mimicry’ for these cases, not because we think that the animals consciously imitate other things, but because natural selection has favoured those individuals whose bodies were mistaken for other things. To put it another way, ancestors of stick insects that did not resemble sticks did not leave descendants. The German-American geneticist Richard Goldschmidt is the most distinguished of those who  {82}  have argued that the early evolution of such resemblances could not have been favoured by natural selection. As Gould, an admirer of Goldschmidt, said of dung-mimicking insects: ‘can there be any edge in looking 5 per cent like a turd?’ Largely through Gould's influence, it has recently become fashionable to say that Goldschmidt was underrated in his own lifetime, and that he really has much to teach us. Here is a sample of his reasoning.

Ford speaks... of any mutation which chances to give a ‘remote resemblance’ to a more protected species, from which some advantage, however slight, might accrue. We must ask how remote the resemblance can be to have selective value. Can we really assume that the birds and monkeys and also mantids are such wonderful observers (or that some very clever ones among them are) to notice a ‘remote’ resemblance and be repelled by it? I think that this is asking too much.

Such sarcasm ill becomes anybody on the shaky ground that Goldschmidt here treads. Wonderful observers? Very clever ones among them? Anybody would think the birds and monkeys benefited from being fooled by the remote resemblance! Goldschmidt might rather have said: ‘Can we really assume that the birds, etc. are such poor observers (or that some very stupid ones among them are)?’ Nevertheless, there is a real dilemma here. The initial resemblance of the ancestral stick insect to a stick must have been very remote. A bird would need extremely poor vision to be fooled by it. Yet the resemblance of a modern stick insect to a stick is marvellously good, down to the last fine details of fake buds and leaf-scars. The birds whose selective predation put the finishing touches to their evolution must, at least collectively, have had excellently good vision. They must have been extremely hard to fool, otherwise the insects would not have evolved to become as perfect mimics as they are: they would have remained relatively imperfect mimics. How can we resolve this apparent contradiction?

One kind of answer suggests that bird vision has been improving over the same evolutionary timespan as insect camouflage. Perhaps, to be a little facetious, an ancestral insect that looked only 5 per cent like a turd would have fooled an ancestral bird with only 5 per cent vision. But that is not the kind of answer I want to give. I suspect, indeed, that the whole process of evolution, from remote resemblance to near perfect mimicry, has gone on, rather rapidly, many times over in different insect groups, during the whole long period that bird vision has been just about as good as it is today.

Another kind of answer that has been offered to the dilemma is the following. Perhaps each species of bird or monkey has poor vision and  {83}  latches onto just one limited aspect of an insect. Maybe one predator species notices only the colour, another only the shape, another only the texture, and so on. Then an insect that resembles a stick in only one limited respect will fool one kind of predator, even though it is eaten by all other kinds of predators. As evolution progresses, more and more features of resemblance are added to the repertoire of the insects. The final multifaceted perfection of mimicry has been put together by the summed natural selection provided by many different species of predators. No one predator sees the whole perfection of mimicry, only we do that.

This seems to imply that only we are ‘clever’ enough to see the mimicry in all its glory. Not only because of this human snobbishness, I prefer yet another explanation. This is that, no matter how good any one predator's vision may be under some conditions, it can be exceedingly poor under other conditions. We can easily, in fact, appreciate from our own familiar experience the whole spectrum from exceedingly poor vision to excellent vision. If I am looking directly at a stick insect, 8 inches in front of my nose and in strong daylight, I shall not be fooled by it. I shall notice the long legs hugging the line of the trunk. I may spot the unnatural symmetry which a real stick would not have. But if I, with the very same eyes and brain, am walking through a forest at dusk, I may well fail to distinguish almost any dull-coloured insect from the twigs that abound everywhere. The image of the insect may pass over the edge of my retina rather than the more acute central region. The insect may be 50 yards away, and so make only a tiny image on my retina. The light may be so poor that I can hardly see anything at all anyway.

In fact, it doesn't matter how remote, how poor is the resemblance of an insect to a stick, there must be some level of twilight, or some degree of distance away from the eye, or some degree of distraction of the predator's attention, such that even a very good eye will be fooled by the remote resemblance. If you don’t find that plausible for some particular example that you have imagined, just turn down the imaginary light a bit, or move a bit further away from the imaginary object! The point is that many an insect was saved by an exceedingly slight resemblance to a twig or a leaf or a fall of dung, on occasions when it was far away from a predator, or on occasions when the predator was looking at it at dusk, or looking at it through a fog, or looking at it while distracted by a receptive female. And many an insect was saved, perhaps from the very same predator, by an uncannily close resemblance to a twig, on occasions when the predator happened to be seeing it at relatively close range and in a good light.  {84}  The important thing about light intensity, distance of insect from predator, distance of image from centre of retina, and similar variables, is that they are all continuous variables. They vary by insensible degrees all the way from the extreme of invisibility to the extreme of visibility. Such continuous variables foster continuous and gradual evolution.

Richard Goldschmidt's problem — which was one of a set that made him resort, for most of his professional life, to the extreme belief that evolution takes great leaps rather than small steps — turns out to be no problem at all. And incidentally, we have also demonstrated to ourselves, yet again, that 5 per cent vision is better than no vision at all. The quality of my vision right at the edge of my retina is probably even poorer than 5 per cent of the quality at the centre of my retina, however you care to measure quality. Yet I can still detect the presence of a large lorry or bus out of the extreme corner of my eye. Since I ride a bicycle to work every day this fact has quite probably saved my life. I notice the difference on those occasions when it is raining and I wear a hat. The quality of our vision on a dark night must be far poorer than 5 per cent of what it is at midday. Yet many an ancestor was probably saved through seeing something that really mattered, a sabre-tooth ‘tiger’ perhaps, or a precipice, in the middle of the night.

Every one of us knows from personal experience, for example on dark nights, that there is an insensibly graded continuous series running all the way from total blindness up to perfect vision, and that every step along this series confers significant benefits. By looking at the world through progressively defocused and focused binoculars, we can quickly convince ourselves that there is a graded series of focusing quality, each step in the series being an improvement over the previous one. By progressively turning the colour-balance knob of a colour television set, we can convince ourselves that there is a graded series of progressive improvement from black and white to full colour vision. The iris diaphragm that opens and shuts the pupil prevents us from being dazzled in bright light, while allowing us to see in dim light. We all experience what it is like not to have an iris diaphragm, when we are momentarily dazzled by oncoming car headlights. Unpleasant, and even dangerous, as this dazzling can be, it still doesn't mean that the whole eye ceases to work! The claim that ‘The eye either functions as a whole, or not at all’ turns out to be, not merely false but self-evidently false to anybody who thinks for 2 seconds about his own familiar experience.

Let us return to our Question 5. Considering each member of the series of Xs connecting the human eye to no eye at all, is it plausible  {85}  that every one of them worked sufficiently well that it assisted the survival and reproduction of the animals concerned? We have now seen the silliness of the anti-evolutionist's assumption that the answer is an obvious no. But is the answer yes? It is less obvious, but I think that it is. Not only is it clear that part of an eye is better than no eye at all. We also can find a plausible series of intermediates among modern animals. This doesn't mean, of course, that these modern intermediates really represent ancestral types. But it does show that intermediate designs are capable of working.

Some single-celled animals have a light-sensitive spot with a little pigment screen behind it. The screen shields it from light coming from one direction, which gives it some ‘idea’ of where the light is coming from. Among many-celled animals, various types of worm and some shellfish have a similar arrangement, but the pigment-backed light-sensitive cells are set in a little cup. This gives slightly better direction-finding capability, since each cell is selectively shielded from light rays coming into the cup from its own side. In a continuous series from flat sheet of light-sensitive cells, through shallow cup to deep cup, each step in the series, however small (or large) the step, would be an optical improvement. Now, if you make a cup very deep and turn the sides over, you eventually make a lensless pinhole camera. There is a continuously graded series from shallow cup to pinhole camera (see, for illustration, the first seven generations of the evolutionary series in Figure 4).

A pinhole camera forms a definite image, the smaller the pinhole the sharper (but dimmer) the image, the larger the pinhole the brighter (but fuzzier) the image. The swimming mollusc Nautilus, a rather strange squid-like creature that lives in a shell like the extinct ammonites (see the ‘shelled cephalopod’ of Figure 5), has a pair of pinhole cameras for eyes. The eye is basically the same shape as ours, but there is no lens and the pupil is just a hole that lets the seawater into the hollow interior of the eye. Actually, Nautilus is a bit of a puzzle in its own right. Why, in all the hundreds of millions of years since its ancestors first evolved a pinhole eye, did it never discover the principle of the lens? The advantage of a lens is that it allows the image to be both sharp and bright. What is worrying about Nautilus is that the quality of its retina suggests that it would really benefit, greatly and immediately, from a lens. It is like a hi-fi system with an excellent amplifier fed by a gramophone with a blunt needle. The system is crying out for a particular simple change. In genetic hyperspace, Nautilus appears to be sitting right next door to an obvious and immediate improvement, yet it doesn't take the small step necessary. Why not?  {86}  Michael Land of Sussex University, our foremost authority on invertebrate eyes, is worried, and so am I. Is it that the necessary mutations cannot arise, given the way Nautilus embryos develop? I don’t want to believe it, but I don’t have a better explanation. At least Nautilus dramatizes the point that a lensless eye is better than no eye at all.

When you have a cup for an eye, almost any vaguely convex, vaguely transparent or even translucent material over its opening will constitute an improvement, because of its slight lens-like properties. It collects light over its area and concentrates it on a smaller area of retina. Once such a crude proto-lens is there, there is a continuously graded series of improvements, thickening it and making it more transparent and less distorting, the trend culminating in what we would all recognize as a true lens. Nautilus's relatives, the squids and octopuses, have a true lens, very like ours although their ancestors certainly evolved the whole camera-eye principle completely independently of ours. Incidentally, Michael Land reckons that there are nine basic principles for image-forming that eyes use, and that most of them have evolved many times independently. For instance, the curved dish-reflector principle is radically different from our own camera-eye (we use it in radiotelescopes, and also in our largest optical telescopes because it is easier to make a large mirror than a large lens), and it has been independently ‘invented’ by various molluscs and crustaceans. Other crustaceans have a compound eye like insects (really a bank of lots of tiny eyes), while other molluscs, as we have seen, have a lensed camera-eye like ours, or a pinhole camera-eye. For each of these types of eye, stages corresponding to evolutionary intermediates exist as working eyes among other modern animals.

Anti-evolution propaganda is full of alleged examples of complex systems that ‘could not possibly’ have passed through a gradual series of intermediates. This is often just another case of the rather pathetic ‘Argument from Personal Incredulity’ that we met in Chapter 2. Immediately after the section on the eye, for example, The Neck of the Giraffe goes on to discuss the bombardier beetle, which

squirts a lethal mixture of hydroquinone and hydrogen peroxide into the face of its enemy. These two chemicals, when mixed together, literally explode. So in order to store them inside its body, the Bombardier Beetle has evolved a chemical inhibitor to make them harmless. At the moment the beetle squirts the liquid out of its tail, an anti-inhibitor is added to make the mixture explosive once again. The chain of events that could have led to the evolution of such a complex, coordinated and subtle process is beyond biological explanation on a simple step-by-step basis. The slightest  {87}  alteration in the chemical balance would result immediately in a race of exploded beetles.

A biochemist colleague has kindly provided me with a bottle of hydrogen peroxide, and enough hydroquinone for 50 bombardier beetles. I am now about to mix the two together. According to the above, they will explode in my face. Here goes...

Well, I'm still here. I poured the hydrogen peroxide into the hydroquinone, and absolutely nothing happened. It didn’t even get warm. Of course I knew it wouldn't: I'm not that foolhardy! The statement that ‘these two chemicals, when mixed together, literally explode’, is, quite simply, false, although it is regularly repeated throughout creationist literature. If you are curious about the bombardier beetle, by the way, what actually happens is as follows. It is true that it squirts a scaldingly hot mixture of hydrogen peroxide and hydroquinone at enemies. But hydrogen peroxide and hydroquinone don’t react violently together unless a catalyst is added. This is what the bombardier beetle does. As for the evolutionary precursors of the system, both hydrogen peroxide and various kinds of quinones are used for other purposes in body chemistry. The bombardier beetle's ancestors simply pressed into different service chemicals that already happened to be around. That's often how evolution works.

On the same page of the book as the bombardier beetle passage is the question: ‘What use would be... half a lung? Natural selection would surely eliminate creatures with such oddities, not preserve them.’ In a healthy adult human, each of the two lungs is divided into about 300 million tiny chambers, at the tips of a branching system of tubes. The architecture of these tubes resembles the biomorph tree at the bottom of Figure 2 in the previous chapter. In that tree, the number of successive branchings, determined by ‘Gene 9’, is eight, and the number of twig tips is 2 to the power 8, or 256. As you go down the page in Figure 2, the number of twig tips successively doubles. In order to provide 300 million twig tips, only 29 successive doublings would be required. Note that there is a continuous gradation from a single chamber to 300 million tiny chambers, each step in the gradation being provided by another two-way branching. This transition can be accomplished in 29 branchings, which we may naively think of as a stately walk of 29 steps across genetic space.

In the lungs, the result of all this branching is that the surface area inside each lung is rather more than 70 square yards. Area is the important variable for a lung, for it is area that determines the rate at which oxygen can be taken in, and waste carbon dioxide pushed out. Now, the thing about area is that it is a continuous variable. Area is  {88}  not one of those things that you either have or you don’t. It is a thing that you can have a little bit more of, or a little bit less of. More than most things, lung area lends itself to gradual, step-by-step change, all the way from 0 square yards up to 70 square yards.

There are plenty of surgical patients walking around with only one lung, and some of them are down to a third of normal lung area. They may be walking, but they aren't walking very far, nor very fast. That is the point. The effect of gradually reducing lung area is not an absolute, all-or-none effect on survival. It is a gradual, continuously varying effect on how far you can walk, and how fast. A gradual, continuously varying effect, indeed, on how long you can expect to live. Death doesn't suddenly arrive below a particular threshold lung area! It becomes gradually more probable as lung area decreases below an optimum (and as it increases above the same optimum, for different reasons connected with economic waste).

The first of our ancestors to develop lungs almost certainly lived in water. We can get an idea of how they might have breathed by looking at modern fish. Most modern fish breathe in water with gills, but many species living in foul, swampy water supplement this by gulping air at the surface. They use the internal chamber of the mouth as a kind of crude proto-lung, and this cavity is sometimes enlarged into a breathing pocket rich in blood vessels. As we've seen, there is no problem in imagining a continuous series of Xs connecting a single pocket to a branching set of 300 million pockets as in a modern human lung.

Interestingly, many modern fish have kept their pocket single, and use it for a completely different purpose. Although it probably began as a lung, over the course of evolution it has become the swimbladder, an ingenious device with which the fish maintains itself as a hydrostat in permanent equilibrium. An animal without an air bladder inside it is normally slightly heavier than water, so sinks to the bottom. This is why sharks have to swim continuously to stop themselves sinking. An animal with large air pockets inside it, like us with our great lungs, tends to rise to the surface. Somewhere in the middle of this continuum, an animal with an air bladder of exactly the right size neither sinks nor rises, but floats steadily in effortless equilibrium. This is the trick that modern fish, other than sharks, have perfected. Unlike sharks, they don’t waste energy preventing themselves from sinking. Their fins and tail are freed for guidance and rapid propulsion. They no longer rely on outside air to fill the bladder, but have special glands for manufacturing gas. Using these glands and other means, they accurately regulate the volume of gas in the bladder, and hence keep themselves in precise hydrostatic equilibrium.  {89} 

Several species of modern fish can leave the water. An extreme is the Indian climbing perch, which hardly ever goes into the water. It has independently evolved a quite different kind of lung from that of our ancestors — an air chamber surrounding the gills. Other fish live basically in water but make brief forays out of it. This is probably what our ancestors did. The thing about forays is that their duration can vary continuously, all the way down to zero. If you are a fish who basically lives and breathes in water, but who occasionally ventures on land, perhaps to cross from one mud puddle to another thereby surviving a drought, you might benefit not just from half a lung but from one-hundredth of a lung. It doesn't matter how small your primordial lung is, there must be some time out of water that you can just endure with the lung, which is a little bit longer than you could have endured without the lung. Time is a continuous variable. There is no hard-and-fast divide between water-breathing and air-breathing animals. Different animals may spend 99 per cent of their time in water, 98 per cent, 97 per cent, and so on all the way to 0 per cent. At every step of the way, some fractional increase in lung area will be an advantage. There is continuity, gradualism, all the way.

What use is half a wing? How did wings get their start? Many animals leap from bough to bough, and sometimes fall to the ground. Especially in a small animal, the whole body surface catches the air and assists the leap, or breaks the fall, by acting as a crude aerofoil. Any tendency to increase the ratio of surface area to weight would help, for example flaps of skin growing out in the angles of joints. From here, there is a continuous series of gradations to gliding wings, and hence to flapping wings. Obviously there are distances that could not have been jumped by the earliest animals with proto-wings. Equally obviously, for any degree of smallness or crudeness of ancestral air-catching surfaces, there must be some distance, however short, which can be jumped with the flap and which cannot be jumped without the flap.

Or, if prototype wingflaps worked to break the animal's fall, you cannot say ‘Below a certain size the flaps would have been of no use at all’. Once again, it doesn't matter how small and un-winglike the first wingflaps were. There must be some height, call it h, such that an animal would just break its neck if it fell from that height, but would just survive if it fell from a slightly lower height. In this critical zone, any improvement in the body surface's ability to catch the air and break the fall, however slight that improvement, can make the difference between life and death. Natural selection will then favour slight, prototype wingflaps. When these small wingflaps have become the norm, the critical height h will become slightly greater. Now a  {90}  slight further increase in the wingflaps will make the difference between life and death. And so on, until we have proper wings.

There are animals alive today that beautifully illustrate every stage in the continuum. There are frogs that glide with big webs between their toes, tree-snakes with flattened bodies that catch the air, lizards with flaps along their bodies; and several different kinds of mammals that glide with membranes stretched between their limbs, showing us the kind of way bats must have got their start. Contrary to the creationist literature, not only are animals with ‘half a wing’ common, so are animals with a quarter of a wing, three quarters of a wing, and so on. The idea of a flying continuum becomes even more persuasive when we remember that very small animals tend to float gently in air, whatever their shape. The reason this is persuasive is that there is an infinitesimally graded continuum from small to large.

The idea of tiny changes cumulated over many steps is an immensely powerful idea, capable of explaining an enormous range of things that would be otherwise inexplicable. How did snake venom get its start? Many animals bite, and any animal's spit contains proteins which, if they get into a wound, may cause an allergic reaction. Even so-called non-venomous snakes can give bites that cause a painful reaction in some people. There is a continuous, graded series from ordinary spit to deadly venom.

How did ears get their start? Any piece of skin can detect vibrations if they come in contact with vibrating objects. This is a natural outgrowth of the sense of touch. Natural selection could easily have enhanced this faculty by gradual degrees until it was sensitive enough to pick up very slight contact vibrations. At this point it would automatically have been sensitive enough to pick up airborne vibrations of sufficient loudness and/or sufficient nearness of origin. Natural selection would then favour the evolution of special organs — ears — for picking up airborne vibrations originating from steadily increasing distances. It is easy to see that there would have been a continuous trajectory of step-by-step improvement, all the way. How did echolocation get its start? Any animal that can hear at all may hear echoes. Blind humans frequently learn to make use of these echoes. A rudimentary version of such a skill in ancestral mammals would have provided ample raw material for natural selection to build upon, leading up by gradual degrees to the high perfection of bats.

Five per cent vision is better than no vision at all. Five per cent hearing is better than no hearing at all. Five per cent flight efficiency is better than no flight at all. It is thoroughly believable that every organ or apparatus that we actually see is the product of a smooth trajectory  {91}  through animal space, a trajectory in which every intermediate stage assisted survival and reproduction. Wherever we have an X in a real live animal, where X is some organ too complex to have arisen by chance in a single step, then according to the theory of evolution by natural selection it must be the case that a fraction of an X is better than no X at all; and two fractions of an X must be better than one; and a whole X must be better than nine-tenths of an X. I have no trouble at all in accepting that these statements are true of eyes, ears including bat ears, wings, camouflaged and mimicking insects, snake jaws, stings, cuckoo habits and all the other examples trotted out in antievolution propaganda. No doubt there are plenty of conceivable Xs for which these statements would not be true, plenty of conceivable evolutionary pathways for which the intermediates would not be improvements on their predecessors. But those Xs are not found in the real world.

Darwin wrote (in The Origin of Species):

If it could be demonstrated that any complex organ existed which could not possibly have been formed by numerous, successive, slight modifications, my theory would absolutely break down.

One hundred and twenty five years on, we know a lot more about animals and plants than Darwin did, and still not a single case is known to me of a complex organ that could not have been formed by numerous successive slight modifications. I do not believe that such a case will ever be found. If it is — it'll have to be a really complex organ, and, as we'll see in later chapters, you have to be sophisticated about what you mean by ‘slight’ — I shall cease to believe in Darwinism.

Sometimes the history of gradual, intermediate stages is clearly written into the shape of modern animals, even taking the form of outright imperfections in the final design. Stephen Gould, in his excellent essay on The Panda's Thumb, has made the point that evolution can be more strongly supported by evidence of telling imperfections than by evidence of perfection. I shall give just two examples.

Fish living on the sea bottom benefit by being flat and hugging the contours. There are two very different kinds of flat fish living on the sea bottom, and they have evolved their flatness in quite different ways. The skates and rays, relatives of sharks, have become flat in what might be called the obvious way. Their bodies have grown out sideways to form great ‘wings’. They are like sharks that have passed under a steam roller, but they remain symmetrical and ‘the right way up’. Plaice, sole, halibut and their relatives have become flat in a different way. They are bony fish (with swimbladders) related to  {92}  herrings, trout, etc., and are nothing to do with sharks. Unlike sharks, bony fish as a rule have a marked tendency to be flattened in a vertical direction. A herring, for instance, is much ‘taller’ than it is wide. It uses its whole, vertically flattened body as a swimming surface, which undulates through the water as it swims. It was natural, therefore, that when the ancestors of plaice and sole took to the sea bottom, they should have lain on one side rather than on the belly like the ancestors of skates and rays. But this raised the problem that one eye was always looking down into the sand and was effectively useless. In evolution this problem was solved by the lower eye ‘moving’ round to the upper side.

We see this process of moving round re-enacted in the development of every young bony flatfish. A young flatfish starts life swimming near the surface, and it is symmetrical and vertically flattened just like a herring. But then the skull starts to grow in a strange, asymmetrical, twisted fashion, so that one eye, for instance the left, moves over the top of the head to finish up on the other side. The young fish settles on the bottom, with both its eyes looking upwards, a strange Picasso-like vision. Incidentally, some species of flatfish settle on the right side, others on the left, and others on either side.

The whole skull of a bony flatfish retains the twisted and distorted evidence of its origins. Its very imperfection is powerful testimony of its ancient history, a history of step-by-step change rather than of deliberate design. No sensible designer would have conceived such a monstrosity if given a free hand to create a flatfish on a clean drawing board. I suspect that most sensible designers would think in terms of something more like a skate. But evolution never starts from a clean drawing board. It has to start from what is already there. In the case of the ancestors of skates this was free-swimming sharks. Sharks in general aren't flattened from side to side as free-swimming bony fish like herrings are. If anything, sharks are already slightly flattened from back to belly. This meant that when some ancient sharks first took to the sea bottom, there was an easy smooth progression to the skate shape, with each intermediate being a slight improvement, given bottom conditions, over its slightly less flattened predecessor.

On the other hand, when the free-swimming ancestor of plaice and halibut, being, like a herring, vertically flattened from side to side, took to the bottom, it was better off lying on its side than balancing precariously on its knife edge of a belly! Even though its evolutionary course was eventually destined to lead it into the complicated and probably costly distortions involved in having two eyes on one side, even though the skate way of being a flat fish might ultimately have  {93}  been the best design for bony fish too, the would-be intermediates that set out along this evolutionary pathway apparently did less well in the short term than their rivals lying on their side. The rivals lying on their side were so much better, in the short term, at hugging the bottom. In genetic hyperspace, there is a smooth trajectory connecting free-swimming ancestral bony fish to flatfish lying on their side with twisted skulls. There is not a smooth trajectory connecting these bony fish ancestors to flatfish lying on their belly. This speculation cannot be the whole truth, because there are some bony fish that have evolved flatness in a symmetrical, skate-like way. Perhaps their free-swimming ancestors were already slightly flattened for some other reason.

My second example of an evolutionary progression that didn’t happen because of disadvantageous intermediates, even though it might ultimately have turned out better if it had, concerns the retina of our eyes (and all other vertebrates). Like any nerve, the optic nerve is a trunk cable, a bundle of separate ‘insulated’ wires, in this case about three million of them. Each of the three million wires leads from one cell in the retina to the brain. You can think of them as the wires leading from a bank of three million photocells (actually three million relay stations gathering information from an even larger number of photocells) to the computer that is to process the information in the brain. They are gathered together from all over the retina into a single bundle, which is the optic nerve for that eye.

Any engineer would naturally assume that the photocells would point towards the light, with their wires leading backwards towards the brain. He would laugh at any suggestion that the photocells might point away from the light, with their wires departing on the side nearest the light. Yet this is exactly what happens in all vertebrate retinas. Each photocell is, in effect, wired in backwards, with its wire sticking out on the side nearest the light. The wire has to travel over the surface of the retina, to a point where it dives through a hole in the retina (the so-called ‘blind spot’) to join the optic nerve. This means that the light, instead of being granted an unrestricted passage to the photocells, has to pass through a forest of connecting wires, presumably suffering at least some attenuation and distortion (actually probably not much but, still, it is the principle of the thing that would offend any tidy-minded engineer!).

I don’t know the exact explanation for this strange state of affairs. The relevant period of evolution is so long ago. But I am ready to bet that it had something to do with the trajectory, the pathway through the real-life equivalent of Biomorph Land, that would have to be traversed in order to turn the retina the right way round, starting from  {94}  whatever ancestral organ preceded the eye. There probably is such a trajectory, but that hypothetical trajectory, when realized in actual bodies of intermediate animals, proved disadvantageous — temporarily disadvantageous only, but that is enough. Intermediates could see even less well than their imperfect ancestors, and it is no consolation that they are building better eyesight for their remote descendants! What matters is survival in the here and now.

‘Dollo's Law’ states that evolution is irreversible. This is often confused with a lot of idealistic nonsense about the inevitability of progress, often coupled with ignorant nonsense about evolution ‘violating the Second Law of Thermodynamics’ (those that belong to the half of the educated population that, according to the novelist C. P. Snow, know what the Second Law is, will realize that it is no more violated by evolution than it is violated by the growth of a baby). There is no reason why general trends in evolution shouldn't be reversed. If there is a trend towards large antlers for a while in evolution, there can easily be a subsequent trend towards smaller antlers again. Dollo's Law is really just a statement about the statistical improbability of following exactly the same evolutionary trajectory twice (or, indeed, any particular trajectory), in either direction. A single mutational step can easily be reversed. But for larger numbers of mutational steps, even in the case of the biomorphs with their nine little genes, the mathematical space of all possible trajectories is so vast that the chance of two trajectories ever arriving at the same point becomes vanishingly small. This is even more true of real animals with their vastly larger numbers of genes. There is nothing mysterious or mystical about Dollo's Law, nor is it something that we go out and ‘test’ in nature. It follows simply from the elementary laws of probability.

For just the same reason, it is vanishingly improbable that exactly the same evolutionary pathway should ever be travelled twice. And it would seem similarly improbable, for the same statistical reasons, that two lines of evolution should converge on exactly the same endpoint from different starting points.

It is all the more striking a testimony to the power of natural selection, therefore, that numerous examples can be found in real nature, in which independent lines of evolution appear to have converged, from very different starting points, on what looks very like the same endpoint. When we look in detail we find — it would be worrying if we didn’t — that the convergence is not total. The different lines of evolution betray their independent origins in numerous points of detail. For instance, octopus eyes are very like ours, but the wires leading  {95}  from their photocells don’t point forwards towards the light, as ours do. Octopus eyes are, in this respect, more ‘sensibly’ designed. They have arrived at a similar endpoint, from a very different starting point. And the fact is betrayed in details such as this.

Such superficially convergent resemblances are often extremely striking, and I shall devote the rest of the chapter to some of them. They provide most impressive demonstrations of the power of natural selection to put together good designs. Yet the fact that the superficially similar designs also differ, testifies to their independent evolutionary origins and histories. The basic rationale is that, if a design is good enough to evolve once, the same design principle is good enough to evolve twice, from different starting points, in different parts of the animal kingdom. This is nowhere better illustrated than in the case we used for our basic illustration of good design itself echolocation. Most of what we know about echolocation comes from bats (and human instruments), but it also occurs in a number of other unrelated groups of animals. At least two separate groups of birds do it, and it has been carried to a very high level of sophistication by dolphins and whales. Moreover, it was almost certainly ‘discovered’ independently by at least two different groups of bats. The birds that do it are the oil birds of South America, and the cave swiftlets of the Far East, the ones whose nests are used for birds’ nest soup. Both types of bird nest deep in caves where little or no light penetrates, and both navigate through the blackness using echoes from their own vocal clicks. In both cases the sounds are audible to humans, not ultrasonic like the more specialized bat clicks. Indeed, neither bird species seems to have developed echolocation to such a pitch of sophistication as bats have. Their clicks are not FM, nor do they appear suitable for Doppler-shift speed metering. Probably, like the fruit bat Rousettus, they just time the silent interval between each click and its echo.

In this case we can be absolutely certain that the two bird species have invented echolocation independently of bats, and independently of each other. The line of reasoning is of a kind that evolutionists frequently use. We look at all the thousands of species of birds, and observe that the vast majority of them don’t use echolocation. Just two isolated little genera of birds do it, and those two have nothing else in common with each other except that both live in caves. Although we believe that all birds and bats must have a common ancestor if we trace their lineages back far enough, that common ancestor was also the common ancestor of all mammals (including ourselves) and all birds. The vast majority of mammals and the vast majority of birds don’t use  {96}  echolocation, and it is highly probable that their common ancestor didn’t either (nor did it fly — that is another technology that has been independently evolved several times). It follows that the echolocation technology has been independently developed in bats and birds, just as it was independently developed by British, American and German scientists. The same kind of reasoning, on a smaller scale, leads to the conclusion that the common ancestor of the oil-bird and the cave swiftlet also did not use echolocation, and that these two genera have developed the same technology independently of each other.

Within the mammals too, bats are not the only group to have independently developed the echolocation technology. Several different kinds of mammals, for instance shrews, rats and seals, seem to use echoes to a small extent, as blind humans do, but the only animals to rival bats in sophistication are whales. Whales are divided into two main groups, toothed whales and baleen whales. Both, of course, are mammals descended from land-dwelling ancestors, and they may well have ‘invented’ the whale way of life independently of one another, starting from different land-dwelling ancestors. The toothed whales include sperm whales, killer whales and the various species of dolphins, all of which hunt relatively large prey such as fish and squids, which they catch in their jaws. Several toothed whales, of which only dolphins have been thoroughly studied, have evolved sophisticated echo-sounding equipment in their heads.

Dolphins emit rapid trains of high-pitched clicks, some audible to us, some ultrasonic. It is probable that the ‘melon’, the bulging dome on the front of a dolphin's head, looking — pleasing coincidence — like the weirdly bulging radar dome of a Nimrod ‘advance-warning’ surveillance aircraft, has something to do with beaming the sonar signals forwards, but its exact workings are not understood. As in the case of bats, there is a relatively slow ‘cruising rate’ of clicking, rising to a high-speed (400 clicks per second) buzz when the animal is closing in on prey. Even the ‘slow’ cruising rate is pretty fast. The river dolphins that live in muddy water are probably the most skilled echolocators, but some open-sea dolphins have been shown in tests to be pretty good too. An Atlantic bottlenose dolphin can discriminate circles, squares and triangles (all of the same standardized area), using only its sonar. It can tell which of two targets is the nearer, when the difference is only 1¹/₄ inches at an overall distance of about 7 yards. It can detect a steel sphere half the size of a golf ball, at a range of 70 yards. This performance is not quite as good as human vision in a good light, but probably better than human vision in moonlight.

The intriguing suggestion has been made that dolphins, if they  {97}  chose to use it, have a potentially effortless means of communicating ‘mental pictures’ to one another. All that they would have to do is use their highly versatile voices to mimic the pattern of sound that would be produced by echoes from a particular object. In this way they could convey to one another mental pictures of such objects. There is no evidence for this delightful suggestion. Theoretically, bats could do the same thing, but dolphins seem more likely candidates because they are in general more social. They are also probably ‘cleverer’, but this isn't necessarily a relevant consideration. The instruments that would be needed for communicating echo pictures are no more sophisticated than the instruments that both bats and dolphins already have for echolocating in the first place. And there would seem to be an easy, gradual continuum between using the voice to make echoes and using it to mimic echoes.

At least two groups of bats then, two groups of birds, toothed whales, and probably several other kinds of mammals to a smaller extent, have all independently converged on the technology of sonar, at some time during the last hundred million years. We have no way of knowing whether any other animals now extinct — pterodactyls perhaps? — also evolved the technology independently.

No insects and no fish have so far been found to use sonar, but two quite different groups of fish, one in South America and one in Africa, have developed a somewhat similar navigation system, which appears to be just about as sophisticated and which can be seen as a related, but different, solution to the same problem. These are so-called weakly electric fish. The word ‘weakly’ is to differentiate them from strongly electric fish, which use electric fields, not to navigate, but to stun their prey. The stunning technique, incidentally, has also been independently invented by several unrelated groups of fish, for example electric ‘eels’ (which are not true eels but whose shape is convergent on true eels) and electric rays.

The South American and the African weakly electric fish are quite unrelated to each other, but both live in the same kinds of waters in their respective continents, waters that are too muddy for vision to be effective. The physical principle that they exploit — electric fields in water — is even more alien to our consciousness than that of bats and dolphins. We at least have a subjective idea of what an echo is, but we have almost no subjective idea of what it might be like to perceive an electric field. We didn’t even know of the existence of electricity until a couple of centuries ago. We cannot as subjective human beings empathize with electric fish, but we can, as physicists, understand them.  {98} 

It is easy to see on the dinner plate that the muscles down each side of any fish are arranged as a row of segments, a battery of muscle units. In most fish they contract successively to throw the body into sinuous waves, which propel it forwards. In electric fish, both strongly and weakly electric ones, they have become a battery in the electric sense. Each segment (‘cell’) of the battery generates a voltage. These voltages are connected up in series along the length of the fish so that, in a strongly electric fish such as an electric eel, the whole battery generates as much as 1 amp at 650 volts. An electric eel is powerful enough to knock a man out. Weakly electric fish don’t need high voltages or currents for their purposes, which are purely information-gathering ones.

The principle of electrolocation, as it has been called, is fairly well understood at the level of physics though not, of course, at the level of what it feels like to be an electric fish. The following account applies equally to African and South American weakly electric fish: the convergence is that thorough. Current flows from the front half of the fish, out into the water in lines that curve back and return to the tail end of the fish. There are not really discrete ‘lines’ but a continuous ‘field’, an invisible cocoon of electricity surrounding the fish's body. However, for human visualization it is easiest to think in terms of a family of curved lines leaving the fish through a series of portholes spaced along the front half of the body, all curving round in the water and diving into the fish again at the tip of its tail. The fish has what amounts to a tiny voltmeter monitoring the voltage at each ‘porthole’. If the fish is suspended in open water with no obstacles around, the lines are smooth curves. The tiny voltmeters at each porthole all register the voltage as ‘normal’ for their porthole. But if some obstacle appears in the vicinity, say a rock or an item of food, the lines of current that happen to hit the obstacle will be changed. This will change the voltage at any porthole whose current line is affected, and the appropriate voltmeter will register the fact. So in theory a computer, by comparing the pattern of voltages registered by the voltmeters at all the portholes, could calculate the pattern of obstacles around the fish. This is apparently what the fish brain does. Once again, this doesn't have to mean that the fish are clever mathematicians. They have an apparatus that solves the necessary equations, just as our brains unconsciously solve equations every time we catch a ball.

It is very important that the fish's own body is kept absolutely rigid. The computer in the head couldn't cope with the extra distortions that would be introduced if the fish's body were bending and twisting like an ordinary fish. Electric fish have, at least twice independently, hit  {99}  upon this ingenious method of navigation, but they have had to pay a price: they have had to give up the normal, highly efficient, fish method of swimming, throwing the whole body into serpentine waves. They have solved the problem by keeping the body stiff as a poker, but they have a single long fin all the way along the length of the body. Then instead of the whole body being thrown into waves, just the long fin is. The fish's progress through the water is rather slow, but it does move, and apparently the sacrifice of fast movement is worth it: the gains in navigation seem to outweigh the losses in speed of swimming. Fascinatingly, the South American electric fish have hit upon almost exactly the same solution as the African ones, but not quite. The difference is revealing. Both groups have developed a single long fin that runs the whole length of the body, but in the African fish it runs along the back whereas in the South American fish it runs along the belly. This kind of difference in detail is very characteristic of convergent evolution, as we have seen. It is characteristic of convergent designs by human engineers too, of course.

Although the majority of weakly electric fish, in both the African and the South American groups, give their electric discharges in discrete pulses and are called ‘pulse’ species, a minority of species in both groups do it a different way and are called ‘wave’ species. I shall not discuss the difference further. What is interesting for this chapter is that the pulse/wave split has evolved twice, independently, in the unrelated New World and Old World groups.

One of the most bizarre examples of convergent evolution that I know concerns the so-called periodical cicadas. Before getting to the convergence, I must fill in some background information. Many insects have a rather rigid separation between a juvenile feeding stage, in which they spend most of their lives, and a relatively brief adult reproducing stage. Mayflies, for instance, spend most of their lives as underwater feeding larvae, then emerge into the air for a single day into which they cram the whole of their adult lives. We can think of the adult as analogous to the ephemeral winged seed of a plant like a sycamore, and the larva as analogous to the main plant, the difference being that sycamores make many seeds and shed them over many successive years, while a mayfly larva gives rise to only one adult right at the end of its own life. Anyway, periodical cicadas have carried the mayfly trend to an extreme. The adults live for a few weeks, but the ‘juvenile’ stage (technically ‘nymphs’ rather than larvae) lasts for 13 years (in some varieties) or 17 years (in other varieties). The adults emerge at almost exactly the same moment, having spent 13 (or 17) years cloistered underground. Cicada plagues, which occur in any  {100}  given area exactly 13 (or 17) years apart, are spectacular eruptions that have led to their incorrectly being called ‘locusts’ in vernacular American speech. The varieties are known, respectively, as 13-year cicadas and 17-year cicadas.

Now here is the really remarkable fact. It turns out that there is not just one 13-year cicada species and one 17-year species. Rather, there are three species, and each one of the three has both a 17-year and a 13-year variety or race. The division into a 13-year race and a 17-year race has been arrived at independently, no fewer than three times. It looks as though the intermediate periods of 14, 15 and 16 years have been shunned convergently, no fewer than three times. Why? We don’t know. The only suggestion anyone has come up with is that what is special about 13 and 17, as opposed to 14, 15 and 16, is that they are prime numbers. A prime number is a number that is not exactly divisible by any other number. The idea is that a race of animals that regularly erupts in plagues gains the benefit of alternately ‘swamping’ and starving its enemies, predators or parasites. And if these plagues are carefully timed to occur a prime number of years apart, it makes it that much more difficult for the enemies to synchronize their own life cycles. If the cicadas erupted every 14 years, for instance, they could be exploited by a parasite species with a 7-year life cycle. This is a bizarre idea, but no more bizarre than the phenomenon itself. We really don’t know what is special about 13 and 17 years. What matters for our purposes here is that there must be something special about those numbers, because three different species of cicada have independently converged upon them.

Examples of convergence on a large scale occur when two or more continents are isolated from one another for a long time, and a parallel range of ‘trades’ is adopted by unrelated animals on each of the continents. By ‘trades’ I mean ways of making a living, such as burrowing for worms, digging for ants, chasing large herbivores, eating leaves up trees. A good example is the convergent evolution of a whole range of mammal trades in the separate continents of South America, Australia, and the Old World.

These continents weren't always separate. Because our lives are measured in decades, and even our civilizations and dynasties are measured only in centuries, we are accustomed to thinking of the map of the world, the outlines of the continents, as fixed. The theory that continents drifted about was proposed long ago by the German geophysicist Alfred Wegener, but most people laughed at him until well after the Second World War. The admitted fact that South America and Africa look a bit like separated pieces of a jigsaw puzzle  {101}  was assumed to be just an amusing coincidence. In one of the most rapid and complete revolutions science has known, the formerly controversial theory of ‘continental drift’ has now become universally accepted under the name of plate tectonics. The evidence that the continents have drifted, that South America did indeed break away from Africa for instance, is now literally overwhelming, but this is not a book about geology and I shall not spell it out. For us the important point is that the timescale on which continents have drifted about is the same slow timescale on which animal lineages have evolved, and we cannot ignore continental drift if we are to understand the patterns of animal evolution on those continents.

Up until about 100 million years ago, then, South America was joined to Africa in the east and to Antarctica in the south. Antarctica was joined to Australia, and India was joined to Africa via Madagascar. There was in fact one huge southern continent, which we now call Gondwanaland, consisting of what is now South America, Africa, Madagascar, India, Antarctica and Australia all rolled into one. There was also a single large northern continent called Laurasia consisting of what is now North America, Greenland, Europe and Asia (apart from India). North America was not connected to South America. About 100 million years ago there was a big break-up of the land masses, and the continents have been slowly moving towards their present positions ever since (they will, of course, continue to move in the future). Africa joined up with Asia via Arabia and became part of the huge continent that we now speak of as the Old World. North America drifted away from Europe, Antartica drifted south to its present icy location. India detached itself from Africa and set off across what is now called the Indian Ocean, eventually to crunch into south Asia and raise the Himalayas. Australia drifted away from Antarctica into the open sea to become an island continent miles from anywhere else.

It happens that the break-up of the great southern continent of Gondwanaland began during the age of the dinosaurs. When South America and Australia broke away to begin their long periods of isolation from the rest of the world, they each carried their own cargo of dinosaurs, and also of the less-prominent animals that were to become the ancestors of modern mammals. When, rather later, for reasons that are not understood and are the subject of much profitable speculation, the dinosaurs (with the exception of the group of dinosaurs that we now call birds) went extinct, they went extinct all over the world. This left a vacuum in the ‘trades’ open to land-dwelling animals. The vacuum was filled, over a period of millions of years of evolution, mostly by mammals. The interesting point for us here is  {102}  that there were three independent vacuums, and they were independently filled by mammals in Australia, South America and the Old World.

The primitive mammals that happened to be around in the three areas when the dinosaurs more or less simultaneously vacated the great life trades, were all rather small and insignificant, probably nocturnal, previously overshadowed and overpowered by the dinosaurs. They could have evolved in radically different directions in the three areas. To some extent this is what happened. There is nothing in the Old World that resembles the giant ground sloth of South America, alas now extinct. The great range of South American mammals included an extinct giant guinea-pig, the size of a modern rhinoceros but a rodent (I have to say ‘modern’ rhinoceros because the Old World fauna included a giant rhinoceros the size of a two-storey house). But although the separate continents each produced their unique mammals, the general pattern of evolution in all three areas was the same. In all three areas the mammals that happened to be around at the start fanned out in evolution, and produced a specialist for each trade which, in many cases, came to bear a remarkable resemblance to the corresponding specialist in the other two areas. Each trade, the burrowing trade, the large hunter trade, the plains-grazing trade, and so on, was the subject of independent convergent evolution in two or three separate continents. In addition to these three major sites of independent evolution, smaller islands such as Madagascar have interesting parallel stories of their own, which I shall not go into.

Setting aside the strange egg-laying mammals of Australia — the duck-billed platypus and the spiny anteaters — modern mammals all belong to one of two great groups. These two are the marsupials (whose young are born very small and are then kept in a pouch) and the placentals (all the rest of us). The marsupials came to dominate the Australian story and the placentals the Old World, while the two groups played important roles alongside each other in South America. The South American story is complicated by the fact that it was subject to sporadic waves of invasion by mammals from North America.

Having set the scene, we can now look at some of the trades and convergences themselves. An important trade is concerned with the exploitation of the great grasslands variously known as prairie, pampas, savannah, etc. Practitioners of this trade include horses (of which the main African species are called zebras and the desert models are called donkeys), and cattle, such as the North American bison, now hunted to near-extinction. Herbivores typically have very long guts  {103}  containing various kinds of fermenting bacteria, since grass is a poor-quality food and needs a lot of digesting. Rather than break their eating up into discrete meals, they typically eat more or less continuously. Huge volumes of plant material flow through them like a river, all the day long. The animals are often very large, and they frequently go about in great herds. Each one of these big herbivores is a mountain of valuable food to any predator that can exploit it. As a consequence of this there is, as we shall see, a whole trade devoted to the difficult task of catching and killing them. These are the predators. Actually, when I say ‘a trade’, I really mean a whole lot of ‘sub-trades’: lions, leopards, cheetahs, wild dogs and hyenas all hunt in their own specialized ways. The same kind of subdivision is found in the herbivores, and in all the other ‘trades’.

The herbivores have keen senses with which they are continuously alert for predators, and they are usually capable of running very fast to escape them. To this end they often have long, spindly legs, and they typically run on the tips of their toes, which have become specially elongated and strengthened in evolution. The nails at the ends of these specialized toes have become large and hard, and we call them hooves. Cattle have two enlarged toes at the extremities of each leg: the familiar ‘cloven’ hooves. Horses do much the same thing except that, probably for reasons of historical accident, they run on only one toe instead of two. It is derived from what was originally the middle one of the five toes. The other toes have almost completely disappeared over evolutionary time, although they occasionally reappear in freakish ‘throwbacks’.

Now South America, as we have seen, was isolated during the period in which horses and cattle were evolving in other parts of the world. But South America has its own great grasslands, and it evolved its own separate groups of large herbivores to exploit the resource. There were massive rhino-like Leviathans that had no connection with true rhinos. The skulls of some of the early South American herbivores suggest that they ‘invented’ the trunk independently of the true elephants. Some resembled camels, some looked like nothing on earth (today), or like weird chimeras of modern animals. The group called the litopterns are almost unbelievably similar to horses in their legs, yet they were utterly unrelated to horses. The superficial resemblance fooled a nineteenth-century Argentinian expert who thought, with pardonable national pride, that they were the ancestors of all horses in the rest of the world. In fact their resemblance to horses was superficial, and convergent. Grassland life is much the same the world over, and horses and litopterns independently evolved the same qualities to  {104}  cope with the problems of grassland life. In particular, the litopterns, like the horses, lost all their toes except the middle one on each leg, which became enlarged as the bottom joint of the leg and developed a hoof. The leg of a litoptern is all but indistinguishable from the leg of a horse, yet the two animals are only distantly related.

In Australia the large grazers and browsers are very different kangaroos. Kangaroos have the same need to move rapidly, but they have done it in a different way. Instead of developing four-legged galloping to the high pitch of perfection that horses (and presumably litopterns) did, kangaroos have perfected a different gait: two-legged hopping with a large balancing tail. There is little point in arguing over which of these two gaits is ‘better’. They are each highly effective if the body evolves in such a way as to exploit them to the full. Horses and litopterns happened to exploit four-legged galloping, and so ended up with almost identical legs. Kangaroos happened to exploit two-legged hopping, and so ended up with their own uniquely (at least since the dinosaurs) massive hind legs and tail. Kangaroos and horses arrived at different endpoints in ‘animal space’, probably because of some accidental difference in their starting points.

Turning now to the meat-eaters that the great grazers were running away from, we find some more fascinating convergences. In the Old World we are familiar with such large hunters as wolves, dogs, hyenas, and the big cats — lions, tigers, leopards and cheetahs. A big cat that has only recently gone extinct is the sabre-tooth (‘tiger’), named after its colossal canine teeth which jutted down from the upper jaw in the front of what must have been a terrifying gape. Until recent times there were no true cats or dogs in Australia or the New World (pumas and jaguars are recently evolved from Old World cats). But in both those continents there were marsupial equivalents. In Australia the thylacine, or marsupial wolf (often called the Tasmanian wolf because it survived in Tasmania for a little longer than in mainland Australia), was tragically driven extinct within living memory, slaughtered in enormous numbers as a ‘pest’ and for ‘sport’ by humans (there is a slight hope that it may still survive in remote parts of Tasmania, areas which themselves are now threatened with destruction in the interests of providing ‘employment’ for humans). It is not to be confused with the dingo, by the way, which is a true dog, introduced to Australia more recently by (aboriginal) man. A cine film made in the 1930s of the last known thylacine, restlessly pacing its lonely zoo cage, shows an uncannily dog-like animal, its marsupial nature betrayed only by its slightly undog-like way of holding its pelvis and back legs, presumably something to do with accommodating its  {105}  pouch. To any dog-lover, the contemplation of this alternative approach to the dog design, this evolutionary traveller along a parallel road separated by 100 million years, this part-familiar yet part utterly alien other-worldly dog, is a moving experience. Maybe they were pests to humans, but humans were much bigger pests to them; now there are no thylacines left and a considerable surplus of humans.

In South America, too, there were no true dogs or cats during the long period of isolation that we are discussing but, as in Australia, there were marsupial equivalents. Probably the most spectactular was Thylacosmilus, which looked exactly like the recently extinct sabre-tooth ‘tiger’ of the Old World, only more so if you see what I mean. Its daggered gape was even wider, and I imagine that it was even more terrifying. Its name records its superficial affinity with the sabre-tooth (Smilodon) and the Tasmanian ‘wolf’ (Thylacinus), but in terms of ancestry it is very remote from both. It is slightly closer to the thylacine since both are marsupials, but the two have evolved their big carnivore design independently on different continents; independently of each other and of the placental carnivores, the true cats and dogs of the Old World.

Australia, South America and the Old World offer numerous further examples of multiple convergent evolution. Australia has a marsupial ‘mole’, superficially almost indistinguishable from the familiar moles of other continents, but pouched, making its living in the same way as other moles and with the same enormously strengthened forepaws for digging. There is a pouched mouse in Australia, though in this case the resemblance is not so close and it does not make its living in quite the same way. Anteating (where ‘ants’ are deemed for convenience to include termites — another convergence as we shall see) is a ‘trade’ that is filled by a variety of convergent mammals. They may be subdivided into anteaters that burrow, anteaters that climb trees and anteaters that wander over the ground. In Australia, as we might expect, there is a marsupial anteater. Called Myrmecobius, it has a long thin snout for poking into ants’ nests, and a long sticky tongue with which it mops up its prey. It is a ground-dwelling anteater. Australia also has a burrowing anteater, the spiny anteater. This is not a marsupial, but a member of the group of egg-laying mammals, the monotremes, so remote from us that marsupials are our close cousins by comparison. The spiny anteater, too, has a long pointed snout, but its spines give it a superficial resemblance to a hedgehog rather than to another typical anteater.

South America could easily have had a marsupial anteater, alongside its marsupial sabre-tooth ‘tiger’, but as it happens the anteater trade was early filled by placental mammals instead. The  {106}  largest of today's anteaters is Myrmecophaga (which just means anteater in Greek), the large ground-wandering anteater of South America and probably the most extreme anteating specialist in the world. Like the Australian marsupial Myrmecobius, it has a long and pointed snout, extremely long and pointed in this case, and an extremely long sticky tongue. South America also has a small tree-climbing anteater, which is a close cousin of Myrmecophaga and looks like a miniature and less extreme version of it, and a third, intermediate form. Although placental mammals, these anteaters are very far from any Old World placentals. They belong to a uniquely South American family, which also includes armadillos and sloths. This ancient placental family coexisted with the marsupials from the early days of the continent's isolation.

The Old World anteaters include various species of pangolin in Africa and Asia, ranging from tree-climbing forms to digging forms, all looking a bit like fircones with pointed snouts. Also in Africa is the weird ant-bear or aardvark, which is partially specialized for digging. A feature that characterizes all anteaters, whether marsupial, monotreme or placental, is an extremely low metabolic rate. The metabolic rate is the rate at which their chemical ‘fires’ burn, most easily measured as the blood temperature. There is a tendency for metabolic rate to depend on body size in mammals generally. Smaller animals tend to have higher metabolic rates, just as the engines of small cars tend to turn over at a higher rate than those of larger cars. But some animals have high metabolic rates for their size, and anteaters, of whatever ancestry and affinities, tend to have very low metabolic rates for their size. It is not obvious why this is, but it is so strikingly convergent among animals that have nothing else in common but their anteating habit, that it almost certainly is somehow related to this habit.

As we have seen, the ‘ants’ that anteaters eat are often not true ants at all, but termites. Termites are often known as ‘white ants’, but they are related to cockroaches, rather than to true ants, which are related to bees and wasps. Termites resemble ants superficially because they have convergently adopted the same habits. The same range of habits, I should say, because there are many different branches of the ant/termite trade, and both ants and termites have independently adopted most of them. As so often with convergent evolution, the differences are revealing as well as the similarities.

Both ants and termites live in large colonies consisting mostly of sterile, wingless workers, dedicated to the efficient production of winged reproductive castes which fly off to found new colonies. An  {107}  interesting difference is that in ants the workers are all sterile females, whereas in termites they are sterile males and sterile females. Both ant and termite colonies have one (or sometimes several) enlarged ‘queens’, sometimes (in both ants and termites) grotesquely enlarged. In both ants and termites the workers can include specialist castes such as soldiers. Sometimes these are such dedicated fighting machines, especially in their huge jaws (in the case of ants, but ‘gunturrets’ for chemical warfare in the case of termites), that they are incapable of feeding themselves and have to be fed by non-soldier workers. Particular species of ants parallel particular species of termites. For example, the habit of fungus-farming has arisen independently in ants (in the New World) and termites (in Africa). The ants (or termites) forage for plant material that they do not digest themselves but make into compost on which they grow fungi. It is the fungi that they themselves eat. The fungi, in both cases, grow nowhere else than in the nests of ants or termites, respectively. The fungus-farming habit has also been discovered independently and convergently (more than once) by several species of beetles.

There are also interesting convergences within the ants. Although most ant colonies live a settled existence in a fixed nest, there seems to be a successful living to be made by wandering in enormous pillaging armies. This is called the legionary habit. Obviously all ants walk about and forage, but most kinds return to a fixed nest with their booty, and the queen and the brood are left behind in the nest. The key to the wandering legionary habit, on the other hand, is that the armies take the brood and the queen with them. The eggs and larvae are carried in the jaws of workers. In Africa the legionary habit has been developed by the so-called driver ants. In Central and South America the parallel ‘army ants’ are very similar to driver ants in habit and appearance. They are not particularly closely related. They have certainly evolved the characteristics of the ‘army’ trade independently and convergently.

Both driver ants and army ants have exceptionally large colonies, up to a million in army ants, up to about 20 million in driver ants. Both have nomadic phases alternating with ‘statary’ phases, relatively stable encampments or ‘bivouacs’. Army ants and driver ants, or rather their colonies taken together as amoeba-like units, are both ruthless and terrible predators of their respective jungles. Both cut to pieces anything animal in their path, and both have acquired a mystique of terror in their own land. Villagers in parts of South America are reputed traditionally to vacate their villages, lock, stock and barrel when a large ant army is approaching, and to return when the legions have  {108}  marched through, having cleaned out every cockroach, spider and scorpion even from the thatched roofs. I remember as a child in Africa being more frightened of driver ants than of lions or crocodiles. It is worth getting this formidable reputation into perspective by quoting the words of Edward O. Wilson, the world's foremost authority on ants as well as the author of Sociobiology:

In answer to the single question I am asked most frequently about ants, I can give the following answer: No, driver ants are not really the terror of the jungle. Although the driver ant colony is an ‘animal’ weighing in excess of 20 kg and possessing on the order of 20 million mouths and stings and is surely the most formidable creation of the insect world, it still does not match up to the lurid stories told about it. After all, the swarm can only cover about a metre of ground every three minutes. Any competent bush mouse, not to mention man or elephant, can step aside and contemplate the whole grass-roots frenzy at leisure, an object less of menace than of strangeness and wonder, the culmination of an evolutionary story as different from that of mammals as it is possible to conceive in this world.

As an adult in Panama I have stepped aside and contemplated the New World equivalent of the driver ants that I had feared as a child in Africa, flowing by me like a crackling river, and I can testify to the strangeness and wonder. Hour after hour the legions marched past, walking as much over each others’ bodies as over the ground, while I waited for the queen. Finally she came, and hers was an awesome presence. It was impossible to see her body. She appeared only as a moving wave of worker frenzy, a boiling peristaltic ball of ants with linked arms. She was somewhere in the middle of the seething ball of workers, while all around it the massed ranks of soldiers faced threateningly outwards with jaws agape, every one prepared to kill and to die in defence of the queen. Forgive my curiosity to see her: I prodded the ball of workers with a long stick, in a vain attempt to flush out the queen. Instantly 20 soldiers buried their massively muscled pincers in my stick, possibly never to let go, while dozens more swarmed up the stick causing me to let go with alacrity.

I never did glimpse the queen, but somewhere inside that boiling ball she was, the central data bank, the repository of the master DNA of the whole colony. Those gaping soldiers were prepared to die for the queen, not because they loved their mother, not because they had been drilled in the ideals of patriotism, but simply because their brains and their jaws were built by genes stamped from the master die carried in the queen herself. They behaved like brave soldiers because they had inherited the genes of a long line of ancestral queens whose lives, and whose genes, had been saved by soldiers as brave as themselves. My  {109}  soldiers had inherited the same genes from the present queen as those old soldiers had inherited from the ancestral queens. My soldiers were guarding the master copies of the very instructions that made them do the guarding. They were guarding the wisdom of their ancestors, the Ark of the Covenant. These strange statements will be made plain in the next chapter.

I felt the strangeness then, and the wonder, not unmixed with revivals of half-forgotten fears, but transfigured and enhanced by a mature understanding, which I had lacked as a child in Africa, of what the whole performance was for. Enhanced, too, by the knowledge that this story of the legions had reached the same evolutionary culmination not once but twice. These were not the driver ants of my childhood nightmares, however similar they might be, but remote, New World cousins. They were doing the same thing as the driver ants, and for the same reasons. It was night now and I turned for home, an awestruck child again, but joyful in the new world of understanding that had supplanted the dark, African fears.

CHAPTER 5

THE POWER
AND THE ARCHIVES

It is raining DNA outside. On the bank of the Oxford canal at the bottom of my garden is a large willow tree, and it is pumping downy seeds into the air. There is no consistent air movement, and the seeds are drifting outwards in all directions from the tree. Up and down the canal, as far as my binoculars can reach, the water is white with floating cottony flecks, and we can be sure that they have carpeted the ground to much the same radius in other directions too. The cotton wool is mostly made of cellulose, and it dwarfs the tiny capsule that contains the DNA, the genetic information. The DNA content must be a small proportion of the total, so why did I say that it was raining DNA rather than cellulose? The answer is that it is the DNA that matters. The cellulose fluff, although more bulky, is just a parachute, to be discarded. The whole performance, cotton wool, catkins, tree and all, is in aid of one thing and one thing only, the spreading of DNA around the countryside. Not just any DNA, but DNA whose coded characters spell out specific instructions for building willow trees that will shed a new generation of downy seeds. Those fluffy specks are, literally, spreading instructions for making themselves. They are there because their ancestors succeeded in doing the same. It is raining instructions out there; it's raining programs; it's raining tree-growing, fluff-spreading, algorithms. That is not a metaphor, it is the plain truth. It couldn't be any plainer if it were raining floppy discs.

It is plain and it is true, but it hasn't long been understood. A few years ago, if you had asked almost any biologist what was special about living things as opposed to nonliving things, he would have told you about a special substance called protoplasm. Protoplasm wasn't like any other substance; it was vital, vibrant, throbbing, pulsating,  {112}  ‘irritable’ (a schoolmarmish way of saying responsive). If you took a living body and cut it up into ever smaller pieces, you would eventually come down to specks of pure protoplasm. At one time in the last century, a real-life counterpart of Arthur Conan Doyle's Professor Challenger thought that the ‘globigerina ooze’ at the bottom of the sea was pure protoplasm. When I was a schoolboy, elderly textbook authors still wrote about protoplasm although, by then, they really should have known better. Nowadays you never hear or see the word. It is as dead as phlogiston and the universal aether. There is nothing special about the substances from which living things are made. Living things are collections of molecules, like everything else.

What is special is that these molecules are put together in much more complicated patterns than the molecules of nonliving things, and this putting together is done by following programs, sets of instructions for how to develop, which the organisms carry around inside themselves. Maybe they do vibrate and throb and pulsate with ‘irritability’, and glow with ‘living’ warmth, but these properties all emerge incidentally. What lies at the heart of every living thing is not a fire, not warm breath, not a ‘spark of life’. It is information, words, instructions. If you want a metaphor, don’t think of fires and sparks and breath. Think, instead, of a billion discrete, digital characters carved in tablets of crystal. If you want to understand life, don’t think about vibrant, throbbing gels and oozes, think about information technology. It is this that I was hinting at in the previous chapter, when I referred to the queen ant as the central data bank.

The basic requirement for an advanced information technology is some kind of storage medium with a large number of memory locations. Each location must be capable of being in one of a discrete number of states. This is true, anyway, of the digital information technology that now dominates our world of artifice. There is an alternative kind of information technology based upon analogue information. The information on an ordinary gramophone record is analogue. It is stored in a wavy groove. The information on a modern laser disc (often called ‘compact disc’, which is a pity, because the name is uninformative and also usually mispronounced with the stress on the first syllable) is digital, stored in a series of tiny pits, each of which is either definitely there or definitely not there: there are no half measures. That is the diagnostic feature of a digital system: its fundamental elements are either definitely in one state or definitely in another state, with no half measures and no intermediates or compromises.

The information technology of the genes is digital. This fact was  {113}  discovered by Gregor Mendel in the last century, although he wouldn't have put it like that. Mendel showed that we don’t blend our inheritance from our two parents. We receive our inheritance in discrete particles. As far as each particle is concerned, we either inherit it or we don’t. Actually, as R. A. Fisher, one of the founding fathers of what is now called neo-Darwinism, has pointed out, this fact of paniculate inheritance has always been staring us in the face, every time we think about sex. We inherit attributes from a male and a female parent, but each of us is either male or female, not hermaphrodite. Each new baby born has an approximately equal probability of inheriting maleness or femaleness, but any one baby inherits only one of these, and doesn't combine the two. We now know that the same goes for all our particles of inheritance. They don’t blend, but remain discrete and separate as they shuffle and reshuffle their way down the generations. Of course there is often a powerful appearance of blending in the effects that the genetic units have on bodies. If a tall person mates with a short person, or a black person with a white person, their offspring are often intermediate. But the appearance of blending applies only to effects on bodies, and is due to the summed small effects of large numbers of particles. The particles themselves remain separate and discrete when it comes to being passed on to the next generation.

The distinction between blending inheritance and particulate inheritance has been of great importance in the history of evolutionary ideas. In Darwin's time everybody (except Mendel who, tucked away in his monastery, was unfortunately ignored until after his death) thought that inheritance was blending. A Scottish engineer called Fleeming Jenkin pointed out that the fact (as it was thought to be) of blending inheritance all but ruled out natural selection as a plausible theory of evolution. Ernst Mayr rather unkindly remarks that Jenkin's article ‘is based on all the usual prejudices and misunderstandings of the physical scientists’. Nevertheless, Darwin was deeply worried by Jenkin's argument. It was most colourfully embodied in a parable of a white man shipwrecked on an island inhabited by ‘negroes’:

grant him every advantage which we can conceive a white to possess over the native; concede that in the struggle for existence his chance of a long life will be much superior to that of the native chiefs, yet from all these admissions, there does not follow the conclusion that, after a limited or unlimited number of generations, the inhabitants of the island will be white. Our shipwrecked hero would probably become king; he would kill a great many blacks in the struggle for existence; he would have a great many wives and children, while many of his subjects would live and die as bachelors... Our white's qualities would certainly tend very much to  {114}  preserve him to a good old age, and yet he would not suffice in any number of generations to turn his subjects’ descendants white... In the first generation there will be some dozens of intelligent young mulattoes, much superior in average intelligence to the negroes. We might expect the throne for some generations to be occupied by a more or less yellow king; but can any one believe that the whole island will gradually acquire a white, or even a yellow population, or that the islanders would acquire the energy, courage, ingenuity, patience, self-control, endurance, in virtue of which qualities our hero killed so many of their ancestors, and begot so many children, these qualities, in fact, which the struggle for existence would select, if it could select anything?

Don't be distracted by the racist assumptions of white superiority. These were as unquestioned in the time of Jenkin and Darwin as our speciesist assumptions of human rights, human dignity, and the sacredness of human life are unquestioned today. We can rephrase Jenkin's argument in a more neutral analogy. If you mix white paint and black paint together, what you get is grey paint. If you mix grey paint and grey paint together, you can't reconstruct either the original white or the original black. Mixing paints is not so far from the pre-Mendelian vision of heredity, and even today popular culture frequently expresses heredity in terms of a mixing of ‘bloods’. Jenkin's argument is an argument about swamping. As the generations go by, under the assumption of blending inheritance, variation is bound to become swamped. Greater and greater uniformity will prevail. Eventually there will be no variation left for natural selection to work upon.

Plausible as this argument must have sounded, it is not only an argument against natural selection. It is more an argument against inescapable facts about heredity itself! It manifestly isn't true that variation disappears as the generations go by. People are not more similar to each other today than they were in their grandparents’ time. Variation is maintained. There is a pool of variation for selection to work on. This was pointed out mathematically in 1908 by W. Weinberg, and independently by the eccentric mathematician G. H. Hardy, who incidentally, as the betting book of his (and my) college records, once took a bet from a colleague of ‘One half penny to his fortune till death, that the sun will rise tomorrow’. But it took R. A. Fisher and his colleagues, the founders of modern population genetics, to develop the full answer to Fleeming Jenkin in terms of Mendel's theory of particle genetics. This was an irony at the time, because, as we shall see in Chapter 11, the leading followers of Mendel in the early twentieth century thought of themselves as anti-Darwinian. Fisher and his colleagues showed that Darwinian selection made sense, and  {115}  Jenkin's problem was elegantly solved, if what changed in evolution was the relative frequency of discrete hereditary particles, or genes, each of which was either there or not there in any particular individual body. Darwinism post-Fisher is called neo-Darwinism. Its digital nature is not an incidental fact that happens to be true of genetic information technology. Digitalness is probably a necessary precondition for Darwinism itself to work.

In our electronic technology the discrete, digital locations have only two states, conventionally represented as 0 and 1 although you can think of them as high and low, on and off, up and down: all that matters is that they should be distinct from one another, and that the pattern of their states can be ‘read out’ so that it can have some influence on something. Electronic technology uses various physical media for storing 1s and 0s, including magnetic discs, magnetic tape, punched cards and tape, and integrated ‘chips’ with lots of little semiconductor units inside them.

The main storage medium inside willow seeds, ants and all other living cells is not electronic but chemical. It exploits the fact that certain kinds of molecule are capable of ‘polymerizing’, that is joining up in long chains of indefinite length. There are lots of different kinds of polymer. For example, ‘polythene’ is made of long chains of the small molecule called ethylene — polymerized ethylene. Starch and cellulose are polymerized sugars. Some polymers, instead of being uniform chains of one small molecule like ethylene, are chains of two or more different kinds of small molecule. As soon as such heterogeneity enters into a polymer chain, information technology becomes a theoretical possibility. If there are two kinds of small molecule in the chain, the two can be thought of as 1 and 0 respectively, and immediately any amount of information, of any kind, can be stored, provided only that the chain is long enough. The particular polymers used by living cells are called polynucleotides. There are two main families of polynucleotides in living cells, called DNA and RNA for short. Both are chains of small molecules called nucleotides. Both DNA and RNA are heterogeneous chains, with four different kinds of nucleotides. This, of course, is where the opportunity for information storage lies. Instead of just the two states 1 and 0, the information technology of living cells uses four states, which we may conventionally represent as A, T, C and G. There is very little difference, in principle, between a two-state binary information technology like ours, and a four-state information technology like that of the living cell.

As I mentioned at the end of Chapter 1, there is enough information  {116}  capacity in a single human cell to store the Encyclopaedia Britannica, all 30 volumes of it, three or four times over. I don’t know the comparable figure for a willow seed or an ant, but it will be of the same order of staggeringness. There is enough storage capacity in the DNA of a single lily seed or a single salamander sperm to store the Encyclopaedia Britannica 60 times over. Some species of the unjustly called ‘primitive’ amoebas have as much information in their DNA as 1,000 Encyclopaedia Britannicas.

Amazingly, only about 1 per cent of the genetic information in, for example, human cells, seems to be actually used: roughly the equivalent of one volume of the Encyclopaedia Britannica. Nobody knows why the other 99 per cent is there. In a previous book I suggested that it might be parasitic, freeloading on the efforts of the 1 per cent, a theory that has more recently been taken up by molecular biologists under the name of ‘selfish DNA’. A bacterium has a smaller information capacity than a human cell, by a factor of about 1,000, and it probably uses nearly all of it: there is little room for parasites. Its DNA could ‘only’ hold one copy of the New Testament!

Modern genetic engineers already have the technology to write the New Testament or anything else into a bacterium's DNA. The ‘meaning’ of the symbols in any information technology is arbitrary, and there is no reason why we should not assign combinations, say triplets, from DNA's 4-letter alphabet, to letters of our own 26-letter alphabet (there would be room for all the upper and lower-case letters with 12 punctuation characters). Unfortunately, it would take about five man-centuries to write the New Testament into a bacterium, so I doubt if anybody will bother. If they did, the rate of reproduction of bacteria is such that 10 million copies of the New Testament could be run off in a single day, a missionary's dream if only people could read the DNA alphabet but, alas, the characters are so small that all 10 million copies of the New Testament could simultaneously dance upon the surface of a pin's head.

Electronic computer memory is conventionally classified into ROM and RAM. ROM stands for ‘read only’ memory. More strictly it is ‘write once, read many times’ memory. The pattern of 0s and 1s is ‘burned’ into it once and for all on manufacture. It then remains unchanged throughout the life of the memory, and the information can be read out any number of times. Other electronic memory, called RAM, can be ‘written to’ (one soon gets used to this inelegant computer jargon) as well as read. RAM can therefore do everything that ROM can do, and more. What the letters RAM actually stand for is misleading, so I won't mention it. The point about RAM is that you  {117}  can put any pattern of 1s and 0s into any part of it that you like, on as many occasions as you like. Most of a computer's memory is RAM. As I type these words they are going straight into RAM, and the word-processing program controlling things is also in RAM, although it could theoretically be burned into ROM and then never subsequently altered. ROM is used for a fixed repertoire of standard programs, which are needed again and again, and which you can't change even if you wanted to.

DNA is ROM. It can be read millions of times over, but only written to once — when it is first assembled at the birth of the cell in which it resides. The DNA in the cells of any individual is ‘burned in’, and is never altered during that individual's lifetime, except by very rare random deterioration. It can be copied, however. It is duplicated every time a cell divides. The pattern of A,T,C and G nucleotides is faithfully copied into the DNA of each of the trillions of new cells that are made as a baby grows. When a new individual is conceived, a new and unique pattern of data is ‘burned into’ his DNA ROM, and he is then stuck with that pattern for the rest of his life. It is copied into all his cells (except his reproductive cells, into which a random half of his DNA is copied, as we shall see).

All computer memory, whether ‘ROM’ or ‘RAM’, is addressed. This means that every location in the memory has a label, usually a number but this is an arbitrary convention. It is important to understand the distinction between the address and the contents of a memory location. Each location is known by its address. For instance the first two letters of this chapter, ‘It’, are at this moment sitting in RAM locations 6446 and 6447 of my computer, which has 65,536 RAM locations altogether. At another time, the contents of those two locations will be different. The contents of a location is whatever was most recently written in that location. Each ROM location also has an address and a contents. The difference is that each location is stuck with its contents, once and for all.

The DNA is arranged along stringy chromosomes, like long computer tapes. All the DNA in each of our cells is addressed in the same sense as computer ROM, or indeed computer tape, is addressed. The exact numbers or names that we use to label a given address are arbitrary, just as they are for computer memory. What matters is that a particular location in my DNA corresponds precisely to one particular location in your DNA: they have the same address. The contents of my DNA location 321762 may or may not be the same as the contents of your location 321762. But my location 321762 is in precisely the same position in my cells as your location 321762 is in your cells. ‘Position’  {118}  here means position along the length of a particular chromosome. The exact physical position of a chromosome in a cell doesn't matter. Indeed, it floats about in fluid so its physical position varies, but every location along the chromosome is precisely addressed in terms of linear order along the length of the chromosome, just as every location along a computer tape is precisely addressed, even if the tape is strewn around the floor rather than being neatly rolled up. All of us, all human beings, have the same set of DNA addresses, but not necessarily the same contents of those addresses. That is the main reason why we are all different from each other.

Other species don’t have the same set of addresses. Chimpanzees, for instance, have 48 chromosomes compared to our 46. Strictly speaking it is not possible to compare contents, address by address, because addresses don’t correspond to each other across species barriers. Closely related species, however, like chimps and humans, have such large chunks of adjacent contents in common that we can easily identify them as basically the same, even though we can't use quite the same addressing system for the two species. The thing that defines a species is that all members have the same addressing system for their DNA. Give or take a few minor exceptions, all members have the same number of chromosomes, and every location along the length of a chromosome has its exact opposite number in the same position along the length of the corresponding chromosome in all other members of the species. What can differ among the members of a species is the contents of those locations.

The differences in contents in different individuals come about in the following manner, and here I must stress that I am talking about sexually reproducing species such as our own. Our sperms or eggs each contain 23 chromosomes. Each addressed location in one of my sperms corresponds to a particular addressed location in every other one of my sperms, and in every one of your eggs (or sperms). All my other cells contain 46 — a double set. The same addresses are used twice over in each of these cells. Every cell contains two chromosome 9s, and two versions of location 7230 along chromosome 9. The contents of the two may or may not be the same, just as they may or may not be the same in other members of the species. When a sperm, with its 23 chromosomes, is made from a body cell with its 46 chromosomes, it only gets one of the two copies of each addressed location. Which one it gets can be treated as random. The same goes for eggs. The result is that every sperm produced and every egg produced is unique in terms of the contents of their locations, although their addressing system is identical in all members of one species (with minor exceptions that  {119}  need not concern us). When a sperm fertilizes an egg, a full complement of 46 chromosomes is, of course, made up; and all 46 are then duplicated in all the cells of the developing embryo.

I said that ROM cannot be written to except when it is first manufactured, and that is true also of the DNA in cells, except for occasional random errors in copying. But there is a sense in which the collective data bank consisting of the ROMs of an entire species can be constructively written to. The nonrandom survival and reproductive success of individuals within the species effectively ‘writes’ improved instructions for survival into the collective genetic memory of the species as the generations go by. Evolutionary change in a species largely consists of changes in how many copies there are of each of the various possible contents at each addressed DNA location, as the generations pass. Of course, at any particular time, every copy has to be inside an individual body. But what matters in evolution is changes in frequency of alternative possible contents at each address in populations. The addressing system remains the same, but the statistical profile of location contents changes as the centuries go by.

Once in a blue moon the addressing system itself changes. Chimpanzees have 24 pairs of chromosomes and we have 23. We share a common ancestor with chimpanzees, so at some point in either our ancestry or chimps’ there must have been a change in chromosome number. Either we lost a chromosome (two merged), or chimps gained one (one split). There must have been at least one individual who had a different number of chromosomes from his parents. There are other occasional changes in the entire genetic system. Whole lengths of code, as we shall see, may occasionally be copied to completely different chromosomes. We know this because we find, scattered around the chromosomes, long strings of DNA text that are identical.

When the information in a computer memory has been read from a particular location, one of two things may happen to it. It can either simply be written somewhere else, or it can become involved in some ‘action’. Being written somewhere else means being copied. We have already seen that DNA is readily copied from one cell to a new cell, and that chunks of DNA may be copied from one individual to another individual, namely its child. ‘Action’ is more complicated. In computers, one kind of action is the execution of program instructions. In my computer's ROM, location numbers 64489, 64490 and 64491, taken together, contain a particular pattern of contents — 1s and 0s which — when interpreted as instructions, result in the computer's little loudspeaker uttering a blip sound. This bit pattern is 101011010011000011000000. There is nothing inherently blippy or noisy about that bit pattern. Nothing about it tells you that it will  {120}  have that effect on the loudspeaker. It has that effect only because of the way the rest of the computer is wired up. In the same way, patterns in the DNA four-letter code have effects, for instance on eye colour or behaviour, but these effects are not inherent in the DNA data patterns themselves. They have their effects only as a result of the way the rest of the embryo develops, which in turn is influenced by the effects of patterns in other parts of the DNA. This interaction between genes will be a main theme of Chapter 7.

Before they can be involved in any kind of action, the code symbols of DNA have to be translated into another medium. They are first transcribed into exactly corresponding RNA symbols. RNA also has a four-letter alphabet. From here, they are translated into a different kind of polymer called a polypeptide or protein. It might be called a polyamino acid, because the basic units are amino acids. There are 20 kinds of amino acids in living cells. All biological proteins are chains made of these 20 basic building-blocks. Although a protein is a chain of amino acids, most of them don’t remain long and stringy. Each chain coils up into a complicated knot, the precise shape of which is determined by the order of amino acids. This knot shape therefore never varies for any given sequence of amino acids. The sequence of amino acids in turn is precisely determined by the code symbols in a length of DNA (via RNA as an intermediary). There is a sense, therefore, in which the three-dimensional coiled shape of a protein is determined by the one-dimensional sequence of code symbols in the DNA.

The translation procedure embodies the celebrated three-letter ‘genetic code’. This is a dictionary, in which each of the 64 (4 × 4 × 4) possible triplets of DNA (or RNA) symbols is translated into one of the 20 amino acids or a ‘stop reading’ symbol. There are three of these ‘stop reading’ punctuation marks. Many of the amino acids are coded by more than one triplet (as you might have guessed from the fact that there are 64 triplets and only 20 amino acids). The whole translation, from strictly sequential DNA ROM to precisely invariant three-dimensional protein shape, is a remarkable feat of digital information technology. Subsequent steps by which genes influence bodies are a little less obviously computer-like.

Every living cell, even a single bacterial cell, can be thought of as a gigantic chemical factory. DNA patterns, or genes, exert their effects by influencing the course of events in the chemical factory, and they do this via their influence on the three-dimensional shape of protein molecules. The word gigantic may seem surprising for a cell, especially when you remember that 10 million bacterial cells could sit on the surface of a pin's head. But you will also remember that each of these  {121}  cells is capable of holding the whole text of the New Testament and, moreover, it is gigantic when measured by the number of sophisticated machines that it contains. Each machine is a large protein molecule, put together under the influence of a particular stretch of DNA. Protein molecules called enzymes are machines in the sense that each one causes a particular chemical reaction to take place. Each kind of protein machine churns out its own particular chemical product. To do this it uses raw materials that are drifting around in the cell, being, very probably, the products of other protein machines. To get an idea of the size of these protein machines, each one is made of about 6,000 atoms, which is very large by molecular standards. There are about a million of these large pieces of apparatus in a cell, and there are more than 2,000 different kinds of them, each kind specialized to do a particular operation in the chemical factory — the cell. It is the characteristic chemical products of such enzymes that give a cell its individual shape and behaviour.

Since all body cells contain the same genes, it might seem surprising that all body cells aren't the same as each other. The reason is that a different subset of genes is read in different kinds of cells, the others being ignored. In liver cells, those parts of the DNA ROM specifically relevant to the building of kidney cells are not read, and vice versa. The shape and behaviour of a cell depend upon which genes inside that cell are being read and translated into their protein products. This in turn depends on the chemicals already in the cell, which depends partly on which genes have previously been read in the cell, and partly on neighbouring cells. When one cell divides into two, the two daughter cells aren't necessarily the same as each other. In the original fertilized egg, for instance, certain chemicals congregate at one end of the cell, others at the other end. When such a polarized cell divides, the two daughter cells receive different chemical allocations. This means that different genes will be read in the two daughter cells, and a kind of self-reinforcing divergence gets going. The final shape of the whole body, the size of its limbs, the wiring up of its brain, the timing of its behaviour patterns, are all the indirect consequences of interactions between different kinds of cells, whose differences in their turn arise through different genes being read. These diverging processes are best thought of as locally autonomous in the manner of the ‘recursive’ procedure of Chapter 3, rather than as coordinated in some grand central design.

‘Action’, in the sense used in this chapter, is what a geneticist is talking about when he mentions the ‘phenotypic effect’ of a gene. DNA has effects upon bodies, upon eye colour, hair crinkliness,  {122}  strength of aggressive behaviour and thousands of other attributes, all of which are called phenotypic effects. DNA exerts these effects initially locally, after being read by RNA and translated into protein chains, which then affect cell shape and behaviour. This is one of the two ways in which the information in the pattern of DNA can be read out. The other way is that it can be duplicated into a new DNA strand. This is the copying that we discussed earlier.

There is a fundamental distinction between these two routes of transmission of the DNA information, vertical and horizontal transmission. The information is transmitted vertically to other DNA in cells (that make other cells) that make sperms or eggs. Hence it is transmitted vertically to the next generation and then, vertically again, to an indefinite number of future generations. I shall call this ‘archival DNA’. It is potentially immortal. The succession of cells along which archival DNA travels is called the germ line. The germ line is that set of cells, within a body, which is ancestral to sperms or eggs and hence ancestral to future generations. DNA is also transmitted sideways or horizontally: to DNA in non-germ-line cells such as liver cells or skin cells; within such cells to RNA, thence to protein and various effects on embryonic development and therefore on adult form and behaviour. You can think of horizontal transmission and vertical transmission as corresponding to the two sub-programs called DEVELOPMENT and REPRODUCTION in Chapter 3.

Natural selection is all about the differential success of rival DNA in getting itself transmitted vertically in the species archives. ‘Rival DNA’ means alternative contents of particular addresses in the chromosomes of the species. Some genes are more successful than rival genes at remaining in the archives. Although vertical transmission down the archives of the species is ultimately what ‘success’ means, the criterion for success is normally the action that the genes have on bodies, by means of their sideways transmission. This, too, is just like the biomorph computer model. For instance, suppose that in tigers there is a particular gene which, by means of its sideways influence in cells of the jaw, causes the teeth to be a little sharper than those that would be grown under the influence of a rival gene. A tiger with extra-sharp teeth can kill prey more efficiently than a normal tiger; hence it has more offspring; hence it passes on, vertically, more copies of the gene that makes sharp teeth. It passes on all its other genes at the same time, of course, but only the specific ‘sharp-teeth gene’ will find itself, on average, in the bodies of sharp-toothed tigers. The gene itself benefits, in terms of its vertical transmission, from the average effects that it has on a whole series of bodies.  {123} 

DNA's performance as an archival medium is spectacular. In its capacity to preserve a message it far outdoes tablets of stone. Cows and pea plants (and, indeed, all the rest of us) have an almost identical gene called the histone H4 gene. The DNA text is 306 characters long. We can't say that it occupies the same addresses in all species, because we can't meaningfully compare address labels across species. But what we can say is that there is a length of 306 characters in cows, which is virtually identical to a length of 306 characters in peas. Cows and peas differ from each other in only two characters out of these 306. We don’t know exactly how long ago the common ancestor of cows and peas lived, but fossil evidence suggests that it was somewhere between 1,000 and 2,000 million years ago. Call it 1.5 billion years ago. Over this unimaginably (for humans) long time, each of the two lineages that branched from that remote ancestor has preserved 305 out of the 306 characters (on average: it could be that one lineage has preserved all 306 of them and the other has preserved 304). Letters carved on gravestones become unreadable in mere hundreds of years.

In a way the conservation of the histone-H4 DNA document is even more impressive because, unlike tablets of stone, it is not the same physical structure that lasts and preserves the text. It is repeatedly being copied and recopied as the generations go by, like the Hebrew scriptures which were ritually copied by scribes every 80 years to forestall their wearing-out. It is hard to estimate exactly how many times the histone H4 document has been recopied in the lineage leading to cows from the common ancestor with peas, but it is probably as many as 20 billion times. It is also hard to find a yardstick with which to compare the preservation of more than 99 per cent of information in 20 billion successive copyings. We can try using a version of the game of grandmothers’ whispers. Imagine 20 billion typists sitting in a row. The line of typists would reach right round the Earth 500 times. The first typist writes a page of a document and hands it to his neighbour. He copies it and hands his copy to the next one. He copies it again and hands it on to the next, and so on. Eventually, the message reaches the end of the line, and we read it (or rather our 12,000th great grandchildren do, assuming that all the typists have a speed typical of a good secretary). How faithful a rendering of the original message would it be?

To answer this we have to make some assumption about the accuracy of the typists. Let's twist the question round the other way. How good would each typist have to be, in order to match the DNA's performance? The answer is almost too ludicrous to express. For what it is worth, every typist would have to have an error rate of about one  {124}  in a trillion; that is, he would have to be accurate enough to make only a single error in typing the Bible 250,000 times at a stretch. A good secretary in real life has an error rate of about one per page. This is about half a billion times the error rate of the histone H4 gene. A line of real-life secretaries would degrade a text to 99 per cent of its original letters by the 20th member of the line of 20 billion. By the 10,000th member of the line, less than 1 per cent of the original text would survive. This point of near total degradation would be reached before 99.9995 per cent of the typists had even seen it.

This whole comparison has been a bit of a cheat, but in an interesting and revealing respect. I gave the impression that what we are measuring is copying errors. But the histone H4 document hasn't just been copied, it has been subjected to natural selection. Histone is vitally important for survival. It is used in the structural engineering of chromosomes. Maybe lots more mistakes in copying the histone H4 gene occurred, but the mutant organisms did not survive, or at least did not reproduce. To make the comparison fair, we should have to assume that built into each typist's chair is a gun, wired up so that if he makes a mistake he is summarily shot, his place being taken by a reserve typist (squeamish readers may prefer to imagine a spring-loaded ejector seat gently catapulting miscreant typists out of the line, but the gun gives a more realistic picture of natural selection).

So, this method of measuring the conservatism of DNA, by looking at the number of changes that have actually occurred during geological time, compounds genuine copying fidelity with the filtering effects of natural selection. We see only the descendants of successful DNA changes. The ones that led to death are obviously not with us. Can we measure the actual copying fidelity on the ground, before natural selection gets to work on each new generation of genes? Yes, this is the inverse of what is known as the mutation rate, and it can be measured. The probability of any particular letter being miscopied on any one copying occasion turns out to be a little more than one in a billion. The difference between this, the mutation rate, and the lower rate at which change has actually been incorporated in the histone gene during evolution, is a measure of the effectiveness of natural selection in preserving this ancient document.

The histone gene's conservatism over the aeons is exceptional by genetic standards. Other genes change at a higher rate, presumably because natural selection is more tolerant of variations in them. For instance, genes coding the proteins known as fibrinopeptides change in evolution at a rate that closely approximates the basic mutation rate. This probably means that mistakes in the details of these proteins  {125}  (they are produced during the clotting of blood) don’t matter much for the organism. Haemoglobin genes have a rate of changing that is intermediate betwen histones and fibrinopeptides. Presumably natural selection's tolerance of their errors is intermediate. Haemoglobin is doing an important job in the blood, and its details really matter, but several alternative variants of it seem capable of doing the job equally well.

Here we have something that seems a little paradoxical, until we think about it further. The slowest-evolving molecules, like histones, turn out to be the ones that have been most subject to natural selection. Fibrinopeptides are the fastest-evolving molecules because natural selection almost completely ignores them. They are free to evolve at the mutation rate. The reason this seems paradoxical is that we place so much emphasis on natural selection as the driving force of evolution. If there is no natural selection, therefore, we might expect that there would be no evolution. Conversely, strong ‘selection pressure’, we could be forgiven for thinking, might be expected to lead to rapid evolution. Instead, what we find is that natural selection exerts a braking effect on evolution. The baseline rate of evolution, in the absence of natural selection, is the maximum possible rate. That is synonymous with the mutation rate.

This isn't really paradoxical. When we think about it carefully, we see that it couldn't be otherwise. Evolution by natural selection could not be faster than the mutation rate, for mutation is, ultimately, the only way in which new variation enters the species. All that natural selection can do is accept certain new variations, and reject others. The mutation rate is bound to place an upper limit on the rate at which evolution can proceed. As a matter of fact, most of natural selection is concerned with preventing evolutionary change rather than with driving it. This doesn't mean, I hasten to insist, that natural selection is a purely destructive process. It can construct too, in ways that Chapter 7 will explain.

Even the mutation rate is pretty slow. This is another way of saying that, even without natural selection, the performance of the DNA code in accurately preserving its archive is very impressive. A conservative estimate is that, in the absence of natural selection, DNA replicates so accurately that it takes five million replication generations to miscopy 1 per cent of the characters. Our hypothetical typists are still hopelessly outclassed by DNA, even if there is no natural selection. To match DNA with no natural selection, the typists would each have to be able to type the whole of the New Testament with only one error. That is, they would each have to be about 450 times more accurate  {126}  than a typical real-life secretary. This is obviously much less than the comparable figure of half a billion, which is the factor by which the histone H4 gene after natural selection is more accurate than a typical secretary; but it is still a very impressive figure.

But I have been unfair to the typists. I assumed, in effect, that they are not capable of noticing their mistakes and correcting them. I have assumed a complete absence of proofreading. In reality, of course, they do proofread. My line of billions of typists wouldn't, therefore, cause the original message to degenerate in quite the simple way that I portrayed. The DNA-copying mechanism does the same kind of error-correction automatically. If it didn’t, it wouldn't achieve anything like the stupendous accuracy that I have described. The DNA-copying procedure incorporates various ‘proofreading’ drills. This is all the more necessary because the letters of the DNA code are by no means static, like hieroglyphs carved in granite. On the contrary, the molecules involved are so small — remember all those New Testaments fitting on a pin's head — that they are under constant assault from the ordinary jostling of molecules that goes on due to heat. There is a constant flux, a turnover of letters in the message. About 5,000 DNA letters degenerate per day in every human cell, and are immediately replaced by repair mechanisms. If the repair mechanisms weren't there and ceaselessly working, the message would steadily dissolve. Proofreading of newly copied text is just a special case of normal repair work. It is mainly proofreading that is responsible for DNA's remarkable accuracy and fidelity of information storage.

We have seen that DNA molecules are the centre of a spectacular information technology. They are capable of packing an immense amount of precise, digital information into a very small space; and they are capable of preserving this information — with astonishingly few errors, but still some errors — for a very long time, measured in millions of years. Where are these facts leading us? They are leading us in the direction of a central truth about life on Earth, the truth that I alluded to in my opening paragraph about willow seeds. This is that living organisms exist for the benefit of DNA rather than the other way around. This won't be obvious yet, but I hope to persuade you of it. The messages that DNA molecules contain are all but eternal when seen against the time scale of individual lifetimes. The lifetimes of DNA messages (give or take a few mutations) are measured in units ranging from millions of years to hundreds of millions of years; or, in other words, ranging from 10,000 individual lifetimes to a trillion individual lifetimes. Each individual organism should be seen as a temporary  {127}  vehicle, in which DNA messages spend a tiny fraction of their geological lifetimes.

The world is full of things that exist...! No disputing that, but is it going to get us anywhere? Things exist either because they have recently come into existence or because they have qualities that made them unlikely to be destroyed in the past. Rocks don’t come into existence at a high rate, but once they exist they are hard and durable. If they were not they wouldn't be rocks, they would be sand. Indeed, some of them are, which is why we have beaches! It is the ones that happen to be durable that exist as rocks. Dewdrops, on the other hand, exist, not because they are durable, but because they have only just come into existence and have not yet had time to evaporate. We seem to have two kinds of ‘existenceworthiness’: the dewdrop kind, which can be summed up as ‘likely to come into existence but not very durable’; and the rock kind, which can be summed up as ‘not very likely to come into existence but likely to last for a long time once there’. Rocks have durability and dewdrops have ‘generatability’. (I've tried to think of a less ugly word but I can't.)

DNA gets the best of both worlds. DNA molecules themselves, as physical entities, are like dewdrops. Under the right conditions they come into existence at a great rate, but no one of them has existed for long, and all will be destroyed within a few months. They are not durable like rocks. But the patterns that they bear in their sequences are as durable as the hardest rocks. They have what it takes to exist for millions of years, and that is why they are still here today. The essential difference from dewdrops is that new dewdrops are not begotten by old dewdrops. Dewdrops doubtless resemble other dewdrops, but they don’t specifically resemble their own ‘parent’ dewdrops. Unlike DNA molecules, they don’t form lineages, and therefore can't pass on messages. Dewdrops come into existence by spontaneous generation, DNA messages by replication.

Truisms like ‘the world is full of things that have what it takes to be in the world’ are trivial, almost silly, until we come to apply them to a special kind of durability, durability in the form of lineages of multiple copies. DNA messages have a different kind of durability from that of rocks, and a different kind of generatability from that of dewdrops. For DNA molecules, ‘what it takes to be in the world’ comes to have a meaning that is anything but obvious and tautological. ‘What it takes to be in the world’ turns out to include the ability to build machines like you and me, the most complicated things in the known universe. Let us see how this can be so.

Fundamentally, the reason is that the properties of DNA that we  {128}  have identified turn out to be the basic ingredients necessary for any process of cumulative selection. In our computer models in Chapter 3, we deliberately built into the computer the basic ingredients of cumulative selection. If cumulative selection is really to happen in the world, some entities have got to arise whose properties constitute those basic ingredients. Let us look, now, at what those ingredients are. As we do so, we shall keep in mind the fact that these very same ingredients, at least in some rudimentary form, must have arisen spontaneously on the early Earth, otherwise cumulative selection, and therefore life, would never have got started in the first place. We are talking here not specifically about DNA, but about the basic ingredients needed for life to arise anywhere in the universe.

When the prophet Ezekiel was in the valley of bones he prophesied to the bones and made them join up together. Then he prophesied to them and made flesh and sinews come around them. But still there was no breath in them. The vital ingredient, the ingredient of life, was missing. A dead planet has atoms, molecules and larger lumps of matter, jostling and nestling against each other at random, according to the laws of physics. Sometimes the laws of physics cause the atoms and molecules to join up together like Ezekiel's dry bones, sometimes they cause them to split apart. Quite large accretions of atoms can form, and they can crumble and break apart again. But still there is no breath in them.

Ezekiel called upon the four winds to put living breath into the dry bones. What is the vital ingredient that a dead planet like the early Earth must have, if it is to have a chance of eventually coming alive, as our planet did? It is not breath, not wind, not any kind of elixir or potion. It is not a substance at all, it is a property, the property of self-replication. This is the basic ingredient of cumulative selection. There must somehow, as a consequence of the ordinary laws of physics, come into being self-copying entities or, as I shall call them, replicators. In modern life this role is filled, almost entirely, by DNA molecules, but anything of which copies are made would do. We may suspect that the first replicators on the primitive Earth were not DNA molecules. It is unlikely that a fully fledged DNA molecule would spring into existence without the aid of other molecules that normally exist only in living cells. The first replicators were probably cruder and simpler than DNA.

There are two other necessary ingredients, which will normally arise automatically from the first ingredient, self-replication itself. There must be occasional errors in the self-copying; even the DNA system very occasionally makes mistakes, and it seems likely that the  {129}  first replicators on Earth were much more erratic. And at least some of the replicators should exert power over their own future. This last ingredient sounds more sinister than it actually is. All it means is that some properties of the replicators should have an influence over their probability of being replicated. At least in a rudimentary form, this is likely to be an inevitable consequence of the basic facts of self-replication itself.

Each replicator, then, has copies of itself made. Each copy is the same as the original, and has the same properties as the original. Among these properties, of course, is the property of making (sometimes with errors) more copies of itself. So each replicator is potentially the ‘ancestor’ of an indefinitely long line of descendant replicators, stretching into the distant future, and branching to produce, potentially, an exceedingly large number of descendant replicators. Each new copy must be made from raw materials, smaller building blocks knocking around. Presumably the replicators act as some kind of mould or template. Smaller components fall together into the mould in such a way that a duplicate of the mould is made. Then the duplicate breaks free and is able to act as a mould in its own right. Hence we have a potentially growing population of replicators. The population will not grow indefinitely, because eventually the supply of raw materials, the smaller elements that fall into the moulds, will become limiting.

Now we introduce our second ingredient into the argument. Sometimes the copying will not be perfect. Mistakes will happen. The possibility of errors can never be totally eliminated from any copying process, although their probability can be reduced to low levels. This is what the manufacturers of hi-fi equipment are striving towards all the time, and the DNA-replication process, as we have seen, is spectactularly good at reducing errors. But modern DNA replication is a high-technology affair, with elaborate proofreading techniques that have been perfected over many generations of cumulative selection. As we have seen, the first replicators probably were relatively crude, low-fidelity contraptions in comparison.

Now go back to our population of replicators, and see what the effect of erratic copying will be. Obviously, instead of there being a uniform population of identical replicators, we shall have a mixed population. Probably many of the products of erratic copying will be found to have lost the property of self-replication that their ‘parent’ had. But a few will retain the property of self-replication, while being different from the parent in some other respect. So we shall have copies of errors being duplicated in the population.  {130} 

When you read the word ‘error’, banish from your mind all pejorative associations. It simply means an error from the point of view of high-fidelity copying. It is possible for an error to result in an improvement. I dare say many an exquisite new dish has been created because a cook made a mistake while trying to follow a recipe. Insofar as I can claim to have had any original scientific ideas, these have sometimes been misunderstandings, or misreadings, of other peoples’ ideas. To return to our primeval replicators, while most miscopyings probably resulted in diminished copying effectiveness, or total loss of the self-copying property, a few might actually have turned out to be better at self-replication than the parent replicator that gave rise to them.

What does ‘better’ mean? Ultimately it means more efficient at self-replication, but what might this mean in practice? This brings us to our third ‘ingredient’. I referred to this as ‘power’, and you'll see why in a moment. When we discussed replication as a moulding process, we saw that the last step in the process must be the new copy's breaking free of the old mould. The time that this occupies may be influenced by a property which I shall call the ‘stickiness’ of the old mould. Suppose that in our population of replicators, which vary because of old copying errors back in their ‘ancestry’, some varieties happen to be more sticky than others. A very sticky variety clings to each new copy for an average time of more than an hour before it finally breaks free and the process can begin again. A less-sticky variety lets go of each new copy within a split second of its formation. Which of these two varieties will come to predominate in the population of replicators? There is no doubt about the answer. If this is the only property by which the two varieties differ, the sticky one is bound to become far less numerous in the population. The non-sticky one is churning out copies of non-sticky ones at thousands of times the rate that the sticky one is making copies of sticky ones. Varieties of intermediate stickiness will have intermediate rates of self-propagation. There will be an ‘evolutionary trend’ towards reduced stickiness.

Something like this kind of elementary natural selection has been duplicated in the test-tube. There is a virus called Q-beta which lives as a parasite of the gut bacterium Escherichia coli. Q-beta has no DNA but it does contain, indeed it largely consists of, a single strand of the related molecule RNA. RNA is capable of being replicated in a similar way to DNA.

In the normal cell, protein molecules are assembled to the specification of RNA plans. These are working copies of plans, run off  {131}  from the DNA masters held in the cell's precious archives. But it is theoretically possible to build a special machine — a protein molecule like the rest of the cellular machines — that runs off RNA copies from other RNA copies. Such a machine is called an RNA-replicase molecule. The bacterial cell itself normally has no use for these machines, and doesn't build any. But since the replicase is just a protein molecule like any other, the versatile protein-building machines of the bacterial cell can easily turn to building them, just as the machine tools in a car factory can quickly be turned over in time of war to making munitions: all they need is to be fed the right blueprints. This is where the virus comes in.

The business part of the virus is an RNA plan. Superficially it is indistinguishable from any of the other RNA working blueprints that are floating around, after being run off the bacterium's DNA master. But if you read the small print of the viral RNA you will find something devilish written there. The letters spell out a plan for making RNA-replicase: for making machines that make more copies of the very same RNA plans, that make more machines that make more copies of the plans, that make more...

So the factory is hijacked by these self-interested blueprints. In a sense it was crying out to be hijacked. If you fill your factory with machines so sophisticated that they can make anything that any blueprint tells them to make, it is hardly surprising if sooner or later a blueprint arises that tells these machines to make copies of itself. The factory fills up with more and more of these rogue machines, each churning out rogue blueprints for making more machines that will make more of themselves. Finally, the unfortunate bacterium bursts and releases millions of viruses that infect new bacteria. So much for the normal life cycle of the virus in nature.

I have called RNA-replicase and RNA respectively a machine and a blueprint. So they are, in a sense (to be disputed on other grounds in a later chapter), but they are also molecules, and it is possible for human chemists to purify them, bottle them and store them on a shelf. This is what Sol Spiegelman and his colleagues did in America in the 1960s. Then they put the two molecules together in solution, and a fascinating thing happened. In the test-tube, the RNA molecules acted as templates for the synthesis of copies of themselves, aided by the presence of the RNA-replicase. The machine tools and the blueprints had been extracted and put into cold storage, separately from one another. Then, as soon as they were given access to each other, and also to the small molecules needed as raw materials, in water, both got back to their old tricks even though they were no longer in a living cell but in a test tube.

It is but a short step from this to natural selection and evolution in the  {132}  laboratory. It is just a chemical version of the computer biomorphs. The experimental method is basically to lay out a long row of test-tubes each containing a solution of RNA-replicase, and also of raw materials, small molecules that can be used for RNA synthesis. Each test-tube contains the machine tools and the raw material, but so far it is sitting idle, doing nothing because it lacks a blueprint to work from. Now a tiny amount of RNA itself is dropped into the first test-tube. The replicase apparatus immediately gets to work and manufactures lots of copies of the newly introduced RNA molecules, which spread through the test-tube. Now a drop of the solution in the first test-tube is removed and put into the second test-tube. The process repeats itself in the second test-tube and then a drop is removed and used to seed the third test-tube, and so on.

Occasionally, because of random copying errors, a slightly different, mutant RNA molecule spontaneously arises. If, for any reason, the new variety is competitively superior to the old one, superior in the sense that, perhaps because of its low ‘stickiness’, it gets itself replicated faster or otherwise more effectively, the new variety will obviously spread through the test-tube in which it arose, outnumbering the parental type that gave rise to it. Then, when a drop of solution is removed from that test-tube to seed the next test-tube, it will be the new mutant variety that does the seeding. If we examine the RNAs in a long succession of test-tubes, we see what can only be called evolutionary change. Competitively superior varieties of RNA produced at the end of several test-tube ‘generations’ can be bottled and named for future use. One variety for example, called V2, replicates much more rapidly than normal Q-beta RNA, probably because it is smaller. Unlike Q-beta RNA, it doesn't have to ‘bother’ to contain the plans for making replicase. Replicase is provided free by the experimenters. V2 RNA was used as the starting point for an interesting experiment by Leslie Orgel and his colleagues in California, in which they imposed a ‘difficult’ environment.

They added to their test-tubes a poison called ethidium bromide which inhibits the synthesis of RNA: it gums up the works of the machine tools. Orgel and colleagues began with a weak solution of the poison. At first, the rate of synthesis was slowed down by the poison, but after evolving through about nine test-tube transfer ‘generations’, a new strain of RNA that was resistant to the poison had been selected. Rate of RNA synthesis was now comparable to that of normal V2 RNA in the absence of poison. Now Orgel and his colleagues doubled the concentration of poison. Again the rate of RNA replication dropped, but after another 10 or so test-tube transfers a strain of RNA had  {133}  evolved that was immune even to the higher concentration of poison. Then the concentration of the poison was doubled again. In this way, by successive doublings, they managed to evolve a strain of RNA that could self-replicate in very high concentrations of ethidium bromide, 10 times as concentrated as the poison that had inhibited the original ancestral V2 RNA. They called the new, resistant RNA V40. The evolution of V40 from V2 took about 100 test-tube transfer ‘generations’ (of course, many actual RNA-replication generations go on between each test-tube transfer).

Orgel has also done experiments in which no enzyme was provided. He found that RNA molecules can replicate themselves spontaneously under these conditions, albeit very slowly. They seem to need some other catalyzing substance, such as zinc. This is important because, in the early days of life when replicators first arose, we cannot suppose that there were enzymes around to help them to replicate. There probably was zinc, though.

The complementary experiment was carried out a decade ago in the laboratory of the influential German school working on the origin of life under Manfred Eigen. These workers provided replicase and RNA building blocks in the test-tube, but they did not seed the solution with RNA. Nevertheless, a particular large RNA molecule evolved spontaneously in the test-tube, and the same molecule re-evolved itself again and again in subsequent independent experiments! Careful checking showed that there was no possibility of chance infection by RNA molecules. This is a remarkable result when you consider the statistical improbability of the same large molecule spontaneously arising twice. It is very much more improbable than the spontaneous typing of METHINKS IT IS LIKE A WEASEL. Like that phrase in our computer model, the particular favoured RNA molecule was built up by gradual, cumulative evolution.

The variety of RNA produced, repeatedly, in these experiments was of the same size and structure as the molecules that Spiegelman had produced. But whereas Spiegelman's had evolved by ‘degeneration’ from naturally occurring, larger, Q-beta viral RNA, those of the Eigen group had built themselves up from almost nothing. This particular formula is well adapted to an environment consisting of test-tubes provided with ready-made replicase. It therefore is converged upon by cumulative selection from two very different starting points. The larger, Q-beta RNA molecules are less well adapted to a test-tube environment but better adapted to the environment provided by E.coli cells. Experiments such as these help us to appreciate the entirely  {134}  automatic and non-deliberate nature of natural selection. The replicase ‘machines’ don’t ‘know’ why they make RNA molecules: it is just a byproduct of their shape that they do. And the RNA molecules themselves don’t work out a strategy for getting themselves duplicated. Even if they could think, there is no obvious reason why any thinking entity should be motivated to make copies of itself. If I knew how to make copies of myself, I'm not sure that I would give the project high priority in competition with all the other things I want to do: why should I? But motivation is irrelevant for molecules. It is just that the structure of the viral RNA happens to be such that it makes cellular machinery churn out copies of itself. And if any entity, anywhere in the universe, happens to have the property of being good at making more copies of itself, then automatically more and more copies of that entity will obviously come into existence. Not only that but, since they automatically form lineages and are occasionally miscopied, later versions tend to be ‘better’ at making copies of themselves than earlier versions, because of the powerful processes of cumulative selection. It is all utterly simple and automatic. It is so predictable as to be almost inevitable.

A ‘successful’ RNA molecule in a test-tube is successful because of some direct, intrinsic property of itself, something analogous to the ‘stickiness’ of my hypothetical example. But properties like ‘stickiness’ are rather boring. They are elementary properties of the replicator itself, properties that have a direct effect on its probability of being replicated. What if the replicator has some effect upon something else, which affects something else, which affects something else, which... eventually, indirectly affects the replicator's chance of being replicated? You can see that, if long chains of causes like this existed, the fundamental truism would still hold. Replicators that happen to have what it takes to get replicated would come to predominate in the world, no matter how long and indirect the chain of causal links by which they influence their probability of being replicated. And, by the same token, the world will come to be filled with the links in this causal chain. We shall see those links, and marvel at them.

In modern organisms we see them all the time. They are eyes and skins and bones and toes and brains and instincts. These things are the tools of DNA replication. They are caused by DNA, in the sense that differences in eyes, skins, bones, instincts, etc. are caused by differences in DNA. They exert an influence over the replication of the DNA that caused them, in that they affect the survival and reproduction of their bodies — which contain that same DNA, and whose fate is therefore shared by the DNA. Therefore, the DNA itself exerts  {135}  an influence over its own replication, via the attributes of bodies. DNA can be said to exert power over its own future, and bodies and their organs and behaviour patterns are the instruments of that power.

When we talk about power, we are talking about consequences of replicators that affect their own future, however indirect those consequences might be. It doesn't matter how many links there are in the chain from cause to effect. If the cause is a self-replicating entity, the effect, be it ever so distant and indirect, can be subject to natural selection. I shall summarize the general idea by telling a particular story about beavers. In detail it is hypothetical, but it certainly cannot be far from the truth. Although nobody has done research upon the development of brain connections in the beaver, they have done this kind of research on other animals, like worms. I am borrowing the conclusions and applying them to beavers, because beavers are more interesting and congenial to many people than worms.

A mutant gene in a beaver is just a change in one letter of the billion-letter text; a change in a particular gene G. As the young beaver grows, the change is copied, together with all the other letters in the text, into all the beaver's cells. In most of the cells the gene G is not read; other genes, relevant to the workings of the other cell types, are. G is read, however, in some cells in the developing brain. It is read and transcribed into RNA copies. The RNA working copies drift around the interior of the cells, and eventually some of them bump into protein-making machines called ribosomes. The protein-making machines read the RNA working plans, and turn out new protein molecules to their specification. These protein molecules curl up into a particular shape determined by their own amino-acid sequence, which in turn is governed by the DNA code sequence of the gene G. When G mutates, the change makes a crucial difference to the amino-acid sequence normally specified by the gene G, and hence to the coiled-up shape of the protein molecule.

These slightly altered protein molecules are mass-produced by the protein-making machines inside the developing brain cells. They in turn act as enzymes, machines that manufacture other compounds in the cells, the gene products. The products of the gene G find their way into the membrane surrounding the cell, and are involved in the processes whereby the cell makes connections with other cells. Because of the slight alteration in the original DNA plans, the production-rate of certain of these membrane compounds is changed. This in turn changes the way in which certain developing brain cells connect up with one another. A subtle alteration in the wiring diagram of a particular part of the beaver's brain has occurred, the indirect, indeed far-removed, consequence of a change in the DNA text.  {136} 

Now it happens that this particular part of the beaver's brain, because of its position in the total wiring diagram, is involved in the beaver's dam-building behaviour. Of course, large parts of the brain are involved whenever the beaver builds a dam but, when the G mutation affects this particular part of the brain's wiring diagram, the change has a specific effect on the behaviour. It causes the beaver to hold its head higher in the water while swimming with a log in its jaws. Higher, that is, than a beaver without the mutation. This makes it a little less likely that mud, attached to the log, will wash off during the journey. This increases the stickiness of the log, which in turn means that, when the beaver thrusts it into the dam, the log is more likely to stay there. This will tend to apply to all the logs placed by any beaver bearing this particular mutation. The increased stickiness of the logs is a consequence, again a very indirect consequence, of an alteration in the DNA text.

The increased stickiness of the logs makes the dam a sounder structure, less likely to break up. This in turn increases the size of the lake created by the dam, which makes the lodge in the centre of the lake more secure against predators. This tends to increase the number of offspring successfully reared by the beaver. If we look at the whole population of beavers, those that possess the mutated gene will, on average, tend therefore to rear more offspring than those not possessing the mutated gene. Those offspring will tend to inherit archive copies of the selfsame altered gene from their parents. Therefore, in the population, this form of the gene will become more numerous as the generations go by. Eventually it will become the norm, and will no longer deserve the title ‘mutant’. Beaver dams in general will have improved another notch.

The fact that this particular story is hypothetical, and that the details may be wrong, is irrelevant. The beaver dam evolved by natural selection, and therefore what happened cannot be very different, except in practical details, from the story I have told. The general implications of this view of life are explained and elaborated in my book The Extended Phenotype, and I shan't repeat the arguments here. You will notice that in this hypothetical story there were no fewer than 11 links in the causal chain linking altered gene to improved survival. In real life there might be even more. Every one of those links, whether it is an effect on the chemistry inside a cell, a later effect on how brain cells wire themselves together, an even later effect on behaviour, or a final effect on lake size, is correctly regarded as caused by a change in the DNA. It wouldn't matter if there were 111 links. Any effect that a change in a gene has on its own replication probability is fair game for  {137}  natural selection. It is all perfectly simple, and delightfully automatic and unpremeditated. Something like it is well-nigh inevitable, once the fundamental ingredients of cumulative selection — replication, error and power — have come into existence in the first place. But how did this happen? How did they come into existence on Earth, before life was there? We shall see how this difficult question might be answered, in the next chapter.

CHAPTER 6

ORIGINS AND MIRACLES

Chance, luck, coincidence, miracle. One of the main topics of this chapter is miracles and what we mean by them. My thesis will be that events that we commonly call miracles are not supernatural, but are part of a spectrum of more-or-less improbable natural events. A miracle, in other words, if it occurs at all, is a tremendous stroke of luck. Events don’t fall neatly into natural events versus miracles.

There are some would-be events that are too improbable to be contemplated, but we can't know this until we have done a calculation. And to do the calculation, we must know how much time was available, more generally how many opportunities were available, for the event to occur. Given infinite time, or infinite opportunities, anything is possible. The large numbers proverbially furnished by astronomy, and the large timespans characteristic of geology, combine to turn topsy-turvy our everyday estimates of what is expected and what is miraculous. I shall build up to this point using a specific example which is the other main theme of this chapter. This example is the problem of how life originated on Earth. To make the point clearly, I shall arbitrarily concentrate on one particular theory of the origin of life, although any one of the modern theories would have served the purpose.

We can accept a certain amount of luck in our explanations, but not too much. The question is, how much? The immensity of geological time entitles us to postulate more improbable coincidences than a court of law would allow but, even so, there are limits. Cumulative selection is the key to all our modern explanations of life. It strings a series of acceptably lucky events (random mutations) together in a nonrandom sequence so that, at the end of the sequence, the finished  {140}  product carries the illusion of being very very lucky indeed, far too improbable to have come about by chance alone, even given a timespan millions of times longer than the age of the universe so far. Cumulative selection is the key but it had to get started, and we cannot escape the need to postulate a single-step chance event in the origin of cumulative selection itself.

And that vital first step was a difficult one because, at its heart, there lies what seems to be a paradox. The replication processes that we know seem to need complicated machinery to work. In the presence of a replicase ‘machine tool’, fragments of RNA will evolve, repeatedly and convergently, towards the same endpoint, an endpoint whose ‘probability’ seems vanishingly small until you reflect on the power of cumulative selection. But we have to assist this cumulative selection to get started. It won't go unless we provide a catalyst, such as the replicase ‘machine tool’ of the previous chapter. And that catalyst, it seems, is unlikely to come into existence spontaneously, except under the direction of other RNA molecules. DNA molecules replicate in the complicated machinery of the cell, and written words replicate in Xerox machines, but neither seem capable of spontaneous replication in the absence of their supporting machinery. A Xerox machine is capable of copying its own blueprints, but it is not capable of springing spontaneously into existence. Biomorphs readily replicate in the environment provided by a suitably written computer program, but they can't write their own program or build a computer to run it. The theory of the blind watchmaker is extremely powerful given that we are allowed to assume replication and hence cumulative selection. But if replication needs complex machinery, since the only way we know for complex machinery ultimately to come into existence is cumulative selection, we have a problem.

Certainly the modern cellular machinery, the apparatus of DNA replication and protein synthesis, has all the hallmarks of a highly evolved, specially fashioned machine. We have seen how staggeringly impressive it is as an accurate data storage device. At its own level of ultra-miniaturization, it is of the same order of elaborateness and complexity of design as the human eye is at a grosser level. All who have given thought to the matter agree that an apparatus as complex as the human eye could not possibly come into existence through single-step selection. Unfortunately, the same seems to be true of at least parts of the apparatus of cellular machinery whereby DNA replicates itself, and this applies not just to the cells of advanced creatures like ourselves and amoebas, but also to relatively more primitive creatures like bacteria and blue-green algae.  {141} 

So, cumulative selection can manufacture complexity while single-step selection cannot. But cumulative selection cannot work unless there is some minimal machinery of replication and replicator power, and the only machinery of replication that we know seems too complicated to have come into existence by means of anything less than many generations of cumulative selection! Some people see this as a fundamental flaw in the whole theory of the blind watchmaker. They see it as the ultimate proof that there must originally have been a designer, not a blind watchmaker but a far-sighted supernatural watchmaker. Maybe, it is argued, the Creator does not control the day-to-day succession of evolutionary events; maybe he did not frame the tiger and the lamb, maybe he did not make a tree, but he did set up the original machinery of replication and replicator power, the original machinery of DNA and protein that made cumulative selection, and hence all of evolution, possible.

This is a transparently feeble argument, indeed it is obviously self-defeating. Organized complexity is the thing that we are having difficulty in explaining. Once we are allowed simply to postulate organized complexity, if only the organized complexity of the DNA/protein replicating engine, it is relatively easy to invoke it as a generator of yet more organized complexity. That, indeed, is what most of this book is about. But of course any God capable of intelligently designing something as complex as the DNA/protein replicating machine must have been at least as complex and organized as that machine itself. Far more so if we suppose him additionally capable of such advanced functions as listening to prayers and forgiving sins. To explain the origin of the DNA/protein machine by invoking a supernatural Designer is to explain precisely nothing, for it leaves unexplained the origin of the Designer. You have to say something like ‘God was always there’, and if you allow yourself that kind of lazy way out, you might as well just say ‘DNA was always there’, or ‘Life was always there’, and be done with it.

The more we can get away from miracles, major improbabilities, fantastic coincidences, large chance events, and the more thoroughly we can break large chance events up into a cumulative series of small chance events, the more satisfying to rational minds our explanations will be. But in this chapter we are asking how improbable, how miraculous, a single event we are allowed to postulate. What is the largest single event of sheer naked coincidence, sheer unadulterated miraculous luck, that we are allowed to get away with in our theories, and still say that we have a satisfactory explanation of life? In order for a monkey to write ‘Methinks it is like a weasel’ by chance, it needs a  {142}  very large amount of luck, but it is still measurable. We calculated the odds against it as about 10 thousand million million million million million million (10⁴⁰) to 1 against. Nobody can really comprehend or imagine such a large number, and we just think of this degree of improbability as synonymous with impossible. But although we can't comprehend these levels of improbability in our minds, we shouldn't just run away from them in terror. The number 10⁴⁰ may be very large but we can still write it down, and we can still use it in calculations. There are, after all, even larger numbers: 10⁴⁶, for instance, is not just larger; you must add 10⁴⁰ to itself a million times in order to obtain 10⁴⁶. What if we could somehow muster a gang of 10⁴⁶ monkeys each with its own typewriter? Why, lo and behold, one of them would solemnly type ‘Methinks it is like a weasel’, and another would almost certainly type ‘I think therefore I am’. The problem is, of course, that we couldn't assemble that many monkeys. If all the matter in the universe were turned into monkey flesh, we still couldn't get enough monkeys. The miracle of a monkey typing ‘Methinks it is like a weasel’ is quantitatively too great, measurably too great, for us to admit it to our theories about what actually happens. But we couldn't know this until we sat down and did the calculation.

So, there are some levels of sheer luck, not only too great for puny human imaginations, but too great to be allowed in our hard-headed calculations about the origin of life. But, to repeat the question, how great a level of luck, how much of a miracle, are we allowed to postulate? Don't let's run away from this question just because large numbers are involved. It is a perfectly valid question, and we can at least write down what we would need to know in order to calculate the answer.

Now here is a fascinating thought. The answer to our question — of how much luck we are allowed to postulate — depends upon whether our planet is the only one that has life, or whether life abounds all around the universe. The one thing we know for certain is that life has arisen once, here on this very planet. But we have no idea at all whether there is life anywhere else in the universe. It is entirely possible that there isn't. Some people have calculated that there must be life elsewhere, on the following grounds (I won't point out the fallacy until afterwards). There are probably at least 10²⁰ (i.e. 100 billion billion) roughly suitable planets in the universe. We know that life has arisen here, so it can't be all that improbable. Therefore it is almost inescapable that at least some among all those billions of billions of other planets have life.

The flaw in the argument lies in the inference that, because life has  {143}  arisen here, it can't be too terribly improbable. You will notice that this inference contains the built-in assumption that whatever went on on Earth is likely to have gone on elsewhere in the universe, and this begs the whole question. In other words, that kind of statistical argument, that there must be life elsewhere in the universe because there is life here, builds in, as an assumption, what it is setting out to prove. This doesn't mean that the conclusion that life exists all around the universe is necessarily wrong. My guess is that it is probably right. It simply means that that particular argument that led up to it is no argument at all. It is just an assumption.

Let us, for the sake of discussion, entertain the alternative assumption that life has arisen only once, ever, and that was here on Earth. It is tempting to object to this assumption on the following emotional grounds. Isn't there something terribly medieval about it? Doesn't it recall the time when the church taught that our Earth was the centre of the universe, and the stars just little pinpricks of light set in the sky for our delight (or, even more absurdly presumptuous, that the stars go out of their way to exert astrological influences on our little lives)? How very conceited to assume that, out of all the billions of billions of planets in the universe, our own little backwater of a world, in our own local backwater of a solar system, in our own local backwater of a galaxy, should have been singled out for life? Why, for goodness sake, should it have been our planet?

I am genuinely sorry, for I am heartily thankful that we have escaped from the small-mindedness of the medieval church and I despise modern astrologers, but I am afraid that the rhetoric about backwaters in the previous paragraph is just empty rhetoric. It is entirely possible that our backwater of a planet is literally the only one that has ever borne life. The point is that if there were only one planet that had ever borne life, then it would have to be our planet, for the very good reason that ‘we’ are here discussing the question! If the origin of life is such an improbable event that it happened on only one planet in the universe, then our planet has to be that planet. So, we can't use the fact that Earth has life to conclude that life must be probable enough to have arisen on another planet. Such an argument would be circular. We have to have some independent arguments about how easy or difficult it is for life to originate on a planet, before we can even begin to answer the question of how many other planets in the universe have life.

But that isn't the question we set out with. Our question was, how much luck are we allowed to assume in a theory of the origin of life on Earth? I said that the answer depends upon whether life has arisen only  {144}  once, or many times. Begin by giving a name to the probability, however low it is, that life will originate on any randomly designated planet of some particular type. Call this number the spontaneous generation probability or SGP. It is the SGP that we shall arrive at if we sit down with our chemistry textbooks, or strike sparks through plausible mixtures of atmospheric gases in our laboratory, and calculate the odds of replicating molecules springing spontaneously into existence in a typical planetary atmosphere. Suppose that our best guess of the SGP is some very very small number, say one in a billion. This is obviously such a small probability that we haven't the faintest hope of duplicating such a fantastically lucky, miraculous event as the origin of life in our laboratory experiments. Yet if we assume, as we are perfectly entitled to do for the sake of argument, that life has originated only once in the universe, it follows that we are allowed to postulate a very large amount of luck in a theory, because there are so many planets in the universe where life could have originated. If, as one estimate has it, there are 100 billion billion planets, this is 100 billion times greater than even the very low SGP that we postulated. To conclude this argument, the maximum amount of luck that we are allowed to assume, before we reject a particular theory of the origin of life, has odds of one in N, where N is the number of suitable planets in the universe. There is a lot hidden in that word ‘suitable’, but let us put an upper limit of 1 in 100 billion billion for the maximum amount of luck that this argument entitles us to assume.

Think about what this means. We go to a chemist and say: get out your textbooks and your calculating machine; sharpen your pencil and your wits; fill your head with formulae, and your flasks with methane and ammonia and hydrogen and carbon dioxide and all the other gases that a primeval nonliving planet can be expected to have; cook them all up together; pass strokes of lightning through your simulated atmospheres, and strokes of inspiration through your brain; bring all your clever chemist's methods to bear, and give us your best chemist's estimate of the probability that a typical planet will spontaneously generate a self-replicating molecule. Or, to put it another way, how long would we have to wait before random chemical events on the planet, random thermal jostling of atoms and molecules, resulted in a self-replicating molecule?

Chemists don’t know the answer to this question. Most modern chemists would probably say that we'd have to wait a long time by the standards of a human lifetime, but perhaps not all that long by the standards of cosmological time. The fossil history of earth suggests that we have about a billion years — one ‘aeon’, to use a convenient  {145}  modern definition — to play with, for this is roughly the time that elapsed between the origin of the Earth about 4.5 billion years ago and the era of the first fossil organisms. But the point of our ‘numbers of planets’ argument is that, even if the chemist said that we'd have to wait for a ‘miracle’, have to wait a billion billion years — far longer than the universe has existed, we can still accept this verdict with equanimity. There are probably more than a billion billion available planets in the universe. If each of them lasts as long as Earth, that gives us about a billion billion billion planet-years to play with. That will do nicely! A miracle is translated into practical politics by a multiplication sum.

There is a concealed assumption in this argument. Well, actually there are lots, but there's one in particular that I want to talk about. This is that, once life (i.e. replicators and cumulative selection) originates at all, it always advances to the point where its creatures evolve enough intelligence to speculate about their origins. If this is not so, our estimate of the amount of luck that we are allowed to postulate must be reduced accordingly. To be more precise, the maximum odds against the origin of life on any one planet that our theories are allowed to postulate, is the number of available planets in the universe divided by the odds that life, once started, will evolve sufficient intelligence to speculate about its own origins.

It may seem a little strange that ‘sufficient intelligence to speculate about its own origins’ is a relevant variable. To understand why it is, consider an alternative assumption. Suppose that the origin of life was quite a probable event, but the subsequent evolution of intelligence was exceedingly improbable, demanding a huge stroke of luck. Suppose the origin of intelligence is so improbable that it has happened on only one planet in the universe, even though life has started on many planets. Then, since we know we are intelligent enough to discuss the question, we know that Earth must be that one planet. Now suppose that the origin of life, and the origin of intelligence given that life is there, are both highly improbable events. Then the probability of any one planet, such as Earth, enjoying both strokes of luck is the product of the two low probabilities, and this is a far smaller probability.

It is as though, in our theory of how we came to exist, we are allowed to postulate a certain ration of luck. This ration has, as its upper limit, the number of eligible planets in the universe. Given our ration of luck, we can then ‘spend’ it as a limited commodity over the course of our explanation of our own existence. If we use up almost all our ration of luck in our theory of how life gets started on a planet in  {146}  the first place, then we are allowed to postulate very little more luck in subsequent parts of our theory, in, say, the cumulative evolution of brains and intelligence. If we don’t use up all our ration of luck in our theory of the origin of life, we have some left over to spend on our theories of subsequent evolution, after cumulative selection has got going. If we want to use up most of our ration of luck in our theory of the origin of intelligence, then we haven't much left over to spend on our theory of the origin of life: we must come up with a theory that makes the origin of life almost inevitable. Alternatively, if we don’t need our whole luck ration for these two stages of our theory, we can, in effect, use the surplus to postulate life elsewhere in the universe.

My personal feeling is that, once cumulative selection has got itself properly started, we need to postulate only a relatively small amount of luck in the subsequent evolution of life and intelligence. Cumulative selection, once it has begun, seems to me powerful enough to make the evolution of intelligence probable, if not inevitable. This means that we can, if we want to, spend virtually our entire ration of postulatable luck in one big throw, in our theory of the origin of life on a planet. Therefore we have at our disposal, if we want to use it, odds of 1 in 100 billion billion as an upper limit (or 1 in however many available planets we think there are) to spend in our theory of the origin of life. This is the maximum amount of luck we are allowed to postulate in our theory. Suppose we want to suggest, for instance, that life began when both DNA and its protein-based replication machinery spontaneously chanced to come into existence. We can allow ourselves the luxury of such an extravagant theory, provided that the odds against this coincidence occurring on a planet do not exceed 100 billion billion to one.

This allowance may seem large. It is probably ample to accommodate the spontaneous arising of DNA or RNA. But it is nowhere near enough to enable us to do without cumulative selection altogether. The odds against assembling a well-designed body that flies as well as a swift, or swims as well as a dolphin, or sees as well as a falcon, in a single blow of luck — single-step selection — are stupendously greater than the number of atoms in the universe, let alone the number of planets! No, it is certain that we are going to need a hefty measure of cumulative selection in our explanations of life.

But although we are entitled, in our theory of the origin of life, to spend a maximum ration of luck amounting, perhaps, to odds of 100 billion billion to one against, my hunch is that we aren't going to need more than a small fraction of that ration. The origin of life on a planet can be a very improbable event indeed by our everyday standards, or  {147}  indeed by the standards of the chemistry laboratory, and still be sufficiently probable to have occurred, not just once but many times, all over the universe. We can regard the statistical argument about numbers of planets as an argument of last resort. At the end of the chapter I shall make the paradoxical point that the theory we are looking for may actually need to seem improbable, even miraculous, to our subjective judgement (because of the way our subjective judgement has been made). Nevertheless, it is still sensible for us to begin by seeking that theory of the origin of life with the least degree of improbability. If the theory that DNA and its copying machinery arose spontaneously is so improbable that it obliges us to assume that life is very rare in the universe, and may even be unique to Earth, our first resort is to try to find a more probable theory. So, can we come up with any speculations about relatively probable ways in which cumulative selection might have got its start?

The word ‘speculate’ has pejorative overtones, but these are quite uncalled for here. We can hope for nothing more than speculation when the events we are talking about took place four billion years ago and took place, moreover, in a world that must have been radically different from that which we know today. For instance, there almost certainly was no free oxygen in the atmosphere. Though the chemistry of the world may have changed, the laws of chemistry have not changed (that's why they are called laws), and modern chemists know enough about those laws to make some well-informed speculations, speculations that have to pass rigorous tests of plausibility imposed by the laws. You can't just speculate wildly and irresponsibly, allowing your imagination to run riot in the manner of such unsatisfying space fiction panaceas as ‘hyperdrives’, ‘time warps’ and ‘infinite improbability drives’. Of all possible speculations about the origin of life, most run foul of the laws of chemistry and can be ruled out, even if we make full use of our statistical fall-back argument about numbers of planets. Careful selective speculation is therefore a constructive exercise. But you do have to be a chemist to do it.

I am a biologist not a chemist, and I must rely on chemists to get their sums right. Different chemists prefer different pet theories, and there is no shortage of theories. I could attempt to lay all these theories before you impartially. That would be the proper thing to do in a student textbook. This isn't a student textbook. The basic idea of The Blind Watchmaker is that we don’t need to postulate a designer in order to understand life, or anything else in the universe. We are here concerned with the kind of solution that must be found, because of the kind of problem we are faced with. I think that this is best explained,  {148}  not by looking at lots of particular theories, but by looking at one as an example of how the basic problem — how cumulative selection got its start — might be solved.

Now, which theory to choose as my representative sample? Most textbooks give greatest weight to the family of theories based on an organic ‘primeval soup’. It seems probable that the atmosphere of Earth before the coming of life was like that of other planets which are still lifeless. There was no oxygen, plenty of hydrogen and water, carbon dioxide, very likely some ammonia, methane and other simple organic gases. Chemists know that oxygen-free climates like this tend to foster the spontaneous synthesis of organic compounds. They have set up in flasks miniature reconstructions of conditions on the early Earth. They have passed through the flasks electric sparks simulating lightning, and ultraviolet light, which would have been much stronger before the Earth had an ozone layer shielding it from the sun's rays. The results of these experiments have been exciting. Organic molecules, some of them of the same general types as are normally only found in living things, have spontaneously assembled themselves in these flasks. Neither DNA nor RNA has appeared, but the building blocks of these large molecules, called purines and pyrimidines, have. So have the building blocks of proteins, amino acids. The missing link for this class of theories is still the origin of replication. The building blocks haven't come together to form a self-replicating chain like RNA. Maybe one day they will.

But, in any case, the organic primeval-soup theory is not the one I have chosen for my illustration of the kind of solution that we must look for. I did choose it in my first book, The Selfish Gene, so I thought that here I would fly a kite for a somewhat less-fashionable theory (although it recently has started gaining ground), which seems to me to have at least a sporting chance of being right. Its audacity is appealing, and it does illustrate well the properties that any satisfying theory of the origin of life must have. This is the ‘inorganic mineral’ theory of the Glasgow chemist Graham Cairns-Smith, first proposed 20 years ago and since developed and elaborated in three books, the latest of which, Seven Clues to the Origin of Life, treats the origin of life as a mystery needing a Sherlock Holmes solution.

Cairns-Smith's view of the DNA/protein machinery is that it probably came into existence relatively recently, perhaps as recently as three billion years ago. Before that there were many generations of cumulative selection, based upon some quite different replicating entities. Once DNA was there, it proved to be so much more efficient as a replicator, and so much more powerful in its effects on its own  {149}  replication, that the original replication system that spawned it was cast off and forgotten. The modern DNA machinery, according to this view, is a late-comer, a recent usurper of the role of fundamental replicator, having taken over that role from an earlier and cruder replicator. There may even have been a whole series of such usurpations, but the original replication process must have been sufficiently simple to have come about through what I have dubbed ‘single-step selection’.

Chemists divide their subject into two main branches, organic and inorganic. Organic chemistry is the chemistry of one particular element, carbon. Inorganic chemistry is all the rest. Carbon is important and deserves to have its own private branch of chemistry, partly because life chemistry is all carbon-chemistry, and partly because those same properties that make carbon-chemistry suitable for life also make it suitable for industrial processes, such as those of the plastics industry. The essential property of carbon atoms that makes them so suitable for life and for industrial synthetics, is that they join together to form a limitless repertoire of different kinds of very large molecules. Another element that has some of these same properties is silicon. Although the chemistry of modern Earth-bound life is all carbon-chemistry, this may not be true all over the universe, and it may not always have been true on this Earth. Cairns-Smith believes that the original life on this planet was based on self-replicating inorganic crystals such as silicates. If this is true, organic replicators, and eventually DNA, must later have taken over or usurped the role.

He gives some arguments for the general plausibility of this idea of ‘takeover’. An arch of stones, for instance, is a stable structure capable of standing for many years even if there is no cement to bind it. Building a complex structure by evolution is like trying to build a mortarless arch if you are allowed to touch only one stone at a time. Think about the task naively, and it can't be done. The arch will stand once the last stone is in place, but the intermediate stages are unstable. It's quite easy to build the arch, however, if you are allowed to subtract stones as well as add them. Start by building a solid heap of stones, then build the arch resting on top of this solid foundation. Then, when the arch is all in position, including the vital keystone at the top, carefully remove the supporting stones and, with a modicum of luck, the arch will remain standing. Stonehenge is incomprehensible until we realize that the builders used some kind of scaffolding, or perhaps ramps of earth, which are no longer there. We can see only the end-product, and have to infer the vanished scaffolding. Similarly, DNA and protein are two pillars of a stable and elegant arch, which persists  {150}  once all its parts simultaneously exist. It is hard to imagine it arising by any step-by-step process unless some earlier scaffolding has completely disappeared. That scaffolding must itself have been built by an earlier form of cumulative selection, at whose nature we can only guess. But it must have been based upon replicating entities with power over their own future.

Cairns-Smith's guess is that the original replicators were crystals of inorganic materials, such as those found in clays and muds. A crystal is just a large orderly array of atoms or molecules in the solid state. Because of properties that we can think of as their ‘shape’, atoms and small molecules tend naturally to pack themselves together in a fixed and orderly manner. It is almost as though they ‘want’ to slot together in a particular way, but this illusion is just an inadvertent consequence of their properties. Their ‘preferred’ way of slotting together shapes the whole crystal. It also means that, even in a large crystal such as a diamond, any part of the crystal is exactly the same as any other part, except where there are flaws. If we could shrink ourselves to the atomic scale, we would see almost endless rows of atoms, stretching to the horizon in straight lines — galleries of geometric repetition.

Since it is replication we are interested in, the first thing we must know is, can crystals replicate their structure? Crystals are made of myriads of layers of atoms (or equivalent), and each layer builds upon the layer below. Atoms (or ions; the difference needn't concern us) float around free in solution, but if they happen to encounter a crystal they have a natural tendency to slot into position on the surface of the crystal. A solution of common salt contains sodium ions and chloride ions jostling about in a more or less chaotic fashion. A crystal of common salt is a packed, orderly array of sodium ions alternating with chloride ions at right angles to one another. When ions floating in the water happen to bump into the hard surface of the crystal, they tend to stick. And they stick in just the right places to cause a new layer to be added to the crystal just like the layer below. So once a crystal gets started it grows, each layer being the same as the layer below.

Sometimes crystals spontaneously start to form in solution. At other times they have to be ‘seeded’, either by particles of dust or by small crystals dropped in from elsewhere. Cairns-Smith invites us to perform the following experiment. Dissolve a large quantity of photographer's ‘hypo’ fixer in very hot water. Then let the solution cool down, being careful not to let any dust drop in. The solution is now ‘supersaturated’, ready and waiting to make crystals, but with no  {151}  seed crystals to start the process going. I quote from Cairns-Smith's Seven Clues to the Origin of Life:

Carefully take the lid off the beaker, drop one tiny piece of ‘hypo’ crystal onto the surface of the solution, and watch amazed at what happens. Your crystal grows visibly: it breaks up from time to time and the pieces also grow... Soon your beaker is crowded with crystals, some several centimetres long. Then after a few minutes it all stops. The magic solution has lost its power — although if you want another performance just re-heat and re-cool the beaker... to be supersaturated means to have more dissolved than there ought to be... the cold supersaturated solution almost literally did not know what to do. It had to be ‘told’ by adding a piece of crystal that already had its units (billions and billions of them) packed together in the way that is characteristic for ‘hypo’ crystals. The solution had to be seeded.

Some chemical substances have the potential to crystallize in two alternative ways. Graphite and diamonds, for instance, are both crystals of pure carbon. Their atoms are identical. The two substances differ from each other only in the geometric pattern with which the carbon atoms are packed. In diamonds, the carbon atoms are packed in a tetrahedral pattern which is extremely stable. This is why diamonds are so hard. In graphite the carbon atoms are arranged in flat hexagons layered on top of each other. The bonding between layers is weak, and they therefore slide over each other, which is why graphite feels slippery and is used as a lubricant. Unfortunately you can't crystallize diamonds out of a solution by seeding them, as you can with hypo. If you could, you'd be rich; no on second thoughts you wouldn't, because any fool could do the same.

Now suppose we have a supersaturated solution of some substance, like hypo in that it was eager to crystallize out of solution, and like carbon in that it was capable of crystallizing in either of two ways. One way might be somewhat like graphite, with the atoms arranged in layers, leading to little flat crystals; while the other way gives chunky, diamond-shaped crystals. Now we simultaneously drop into our supersaturated solution a tiny flat crystal and a tiny chunky crystal. We can describe what would happen in an elaboration of Cairns-Smith's description of his hypo experiment. You watch amazed at what happens. Your two crystals grow visibly: they break up from time to time and the pieces also grow. Flat crystals give rise to a population of flat crystals. Chunky crystals give rise to a population of chunky crystals. If there is any tendency for one type of crystal to grow and split more quickly than the other, we shall have a simple kind of natural selection. But the process still lacks a vital ingredient in order to give  {152}  rise to evolutionary change. That ingredient is hereditary variation, or something equivalent to it. Instead of just two types of crystal, there must be a whole range of minor variants that form lineages of like shape, and that sometimes ‘mutate’ to produce new shapes. Do real crystals have something corresponding to hereditary mutation?

Clays and muds and rocks are made of tiny crystals. They are abundant on Earth and probably always have been. When you look at the surface of some types of clay and other minerals with a scanning electron microscope you see an amazing and beautiful sight. Crystals grow like rows of flowers or cactuses, gardens of inorganic rose petals, tiny spirals like cross-sections of succulent plants, bristling organ pipes, complicated angular shapes folded as if in miniature crystalline origami, writhing growths like worm casts or squeezed toothpaste. The ordered patterns become even more striking at greater levels of magnification. At levels that betray the actual position of atoms, the surface of a crystal is seen to have all the regularity of a machine-woven piece of herringbone tweed. But — and here is the vital point — there are flaws. Right in the middle of an expanse of orderly herringbone there can be a patch, identical to the rest except that it is twisted round at a different angle so that the ‘weave’ goes off in another direction. Or the weave may lie in the same direction, but each row has ‘slipped’ half a row to one side. Nearly all naturally occurring crystals have flaws. And once a flaw has appeared, it tends to be copied as subsequent layers of crystal encrust themselves on top of it.

Flaws can occur anywhere over the surface of a crystal. If you like thinking about capacity for information storage (I do), you can imagine the enormous number of different patterns of flaws that could be created over the surface of a crystal. All those calculations about packing the New Testament into the DNA of a single bacterium could be done just as impressively for almost any crystal. What DNA has over normal crystals is a means by which its information can be read. Leaving aside the problem of read-out, you could easily devise an arbitrary code whereby flaws in the atomic structure of the crystal denote binary numbers. You could then pack several New Testaments into a mineral crystal the size of a pin's head. On a larger scale, this is essentially how music information is stored on the surface of a laser (‘compact’) disc. The musical notes are converted, by computer, into binary numbers. A laser is used to etch a pattern of tiny flaws in the otherwise glassy smooth surface of the disc. Each little hole etched corresponds to a binary 1 (or a 0, the labels are arbitrary). When you play the disc, another laser beam ‘reads’ the pattern of flaws, and a special-purpose computer built into the player turns the binary  {153}  numbers back into sound vibrations, which are amplified so that you can hear them.

Although laser discs are used today mainly for music, you could pack the whole Encyclopaedia Britannica onto one of them, and read it out using the same laser technique. Flaws in crystals at the atomic level are far smaller than the pits etched in a laser disc's surface, so crystals can potentially pack more information into a given area. Indeed DNA molecules, whose capacity for storing information has already impressed us, are something close to crystals themselves. Although clay crystals theoretically could store the same prodigious quantities of information as DNA or laser discs can, nobody is suggesting that they ever did. The role of clay and other mineral crystals in the theory is to act as the original ‘low-tech’ replicators, the ones that were eventually replaced by high-tech DNA. They form spontaneously in the waters of our planet without the elaborate ‘machinery’ that DNA needs; and they develop flaws spontaneously, some of which can be replicated in subsequent layers of crystal. If fragments of suitably flawed crystal later broke away, we could imagine them acting as ‘seeds’ for new crystals, each one ‘inheriting’ its ‘parent's’ pattern of flaws.

So, we have a speculative picture of mineral crystals on the primeval Earth showing some of the properties of replication, multiplication, heredity and mutation that would have been necessary in order for a form of cumulative selection to get started. There is still the missing ingredient of ‘power’: the nature of the replicators must somehow have influenced their own likelihood of being replicated. When we were talking about replicators in the abstract, we saw that ‘power’ might simply be direct properties of the replicator itself, intrinsic properties like ‘stickiness’. At this elementary level, the name ‘power’ seems scarcely justified. I use it only because of what it can become in later stages of evolution: the power of a snake's fang, for instance, to propagate (by its indirect consequences on snake survival) DNA coding for fangs. Whether the original low-tech replicators were mineral crystals or organic direct forerunners of DNA itself, we may guess that the ‘power’ they exercised was direct and elementary, like stickiness. Advanced levers of power, like a snake's fang or an orchid's flower, came far later.

What might ‘power’ mean to a clay? What incidental properties of the clay could influence the likelihood that it, the same variety of clay, would be propagated around the countryside? Clays are made from chemical building blocks such as silicic acid and metal ions, which are in solution in rivers and streams having been dissolved — ‘weathered’ —  {154}  out of rocks further upstream. If conditions are right they crystallize out of solution again downstream, forming clays. (Actually the ‘stream’, in this case, is more likely to mean the seeping and trickling of the groundwater than a rushing open river. But, for simplicity, I shall continue to use the general word stream.) Whether or not a particular type of clay crystal is allowed to build up depends, among other things, upon the rate and pattern of flow of the stream. But deposits of clay can also influence the flow of the stream. They do this inadvertently by changing the level, shape and texture of the ground through which the water is flowing. Consider a variant of clay that just happens to have the property of reshaping the structure of the soil so that the flow speeds up. The consequence is that the clay concerned gets washed away again. This kind of clay, by definition, is not very ‘successful’. Another unsuccessful clay would be one that changed the flow in such a way that a rival variant of clay was favoured.

We aren't, of course, suggesting that clays ‘want’ to go on existing. Always we are talking only about incidental consequences, events which follow from properties that the replicator just happens to have. Consider yet another variant of clay. This one happens to slow down the flow in such a way that future deposition of its own kind of clay is enhanced. Obviously this second variant will tend to become common, because it happens to manipulate streams to its own ‘advantage’. This will be a ‘successful’ variant of clay. But so far we are dealing only with single-step selection. Could a form of cumulative selection get going?

To speculate a little further, suppose that a variant of a clay improves its own chances of being deposited, by damming up streams. This is an inadvertent consequence of the peculiar defect structure of the clay. In any stream in which this kind of clay exists, large, stagnant shallow pools form above dams, and the main flow of water is diverted into a new course. In these still pools, more of the same kind of clay is laid down. A succession of such shallow pools proliferates along the length of any stream that happens to be ‘infected’ by seeding crystals of this kind of clay. Now, because the main flow of the stream is diverted, during the dry season the shallow pools tend to dry up. The clay dries and cracks in the sun, and the top layers are blown off as dust. Each dust particle inherits the characteristic defect structure of the parent clay that did the damming, the structure that gave it its damming properties. By analogy with the genetic information raining down on the canal from my willow tree, we could say that the dust carries ‘instructions’ for how to dam streams and eventually make more dust. The dust spreads far and wide in the wind, and there is a good chance  {155}  that some particles of it will happen to land in another stream, hitherto not ‘infected’ with the seeds of this kind of dam-making clay. Once infected by the right sort of dust, a new stream starts to grow crystals of dam-making clay, and the whole depositing, damming, drying, eroding cycle begins again.

To call this a ‘life’ cycle would be to beg an important question, but it is a cycle of a sort, and it shares with true life cycles the ability to initiate cumulative selection. Because streams are infected by dust ‘seeds’ blown from other streams, we can arrange the streams in an order of ‘ancestry’ and ‘descent’. The clay that is damming up pools in stream B arrived there in the form of dust crystals blown from stream A. Eventually, the pools of stream B will dry up and make dust, which will infect streams F and P. With respect to the source of their dam-making clay, we can arrange streams into ‘family trees’. Every infected stream has a ‘parent’ stream, and it may have more than one ‘daughter’ stream. Each stream is analogous to a body, whose ‘development’ is influenced by dust seed ‘genes’, a body that eventually spawns new dust seeds. Each ‘generation’ in the cycle starts when seed crystals break away from the parent stream in the form of dust. The crystalline structure of each particle of dust is copied from the clay in the parent stream. It passes on that crystalline structure to the daughter stream, where it grows and multiplies and finally sends ‘seeds’ out again.

The ancestral crystal structure is preserved down the generations unless there is an occasional mistake in crystal growth, an occasional alteration in the pattern of laying down of atoms. Subsequent layers of the same crystal will copy the same flaw, and if the crystal breaks in two it will give rise to a sub-population of altered crystals. Now if the alteration makes the crystal either less or more efficient in the damming/drying/erosion cycle, this will affect how many copies it has in subsequent ‘generations’. Altered crystals might, for instance, be more likely to split (‘reproduce’). Clay formed from altered crystals might have greater damming power in any of a variety of detailed ways. It might crack more readily in a given amount of sun. It might crumble into dust more readily. The dust particles might be better at catching the wind, like fluff on a willow seed. Some crystal types might induce a shortening of the ‘life cycle’, consequently a speeding up of their ‘evolution’. There are many opportunities for successive ‘generations’ to become progressively ‘better’ at getting passed to subsequent generations. In other words, there are many opportunities for rudimentary cumulative selection to get going.

These little flights of fancy, embellishments of Cairns-Smith's own, concern only one of several kinds of mineral ‘life cycle’ that could have  {156}  started cumulative selection along its momentous road. There are others. Different varieties of crystals might earn their passage to new streams, not by crumbling into dust ‘seeds’, but by dissecting their streams into lots of little streamlets that spread around, eventually joining and infecting new river systems. Some varieties might engineer waterfalls that wear down the rocks faster, and hence speed into solution the raw materials needed to make new clays further downstream. Some varieties of crystal might better themselves by making conditions hard for ‘rival’ varieties that compete for raw materials. Some varieties might become ‘predatory’, breaking up rival varieties and using their elements as raw materials. Keep holding in mind that there is no suggestion of ‘deliberate’ engineering, either here or in modern, DNA-based life. It is just that the world automatically tends to become full of those varieties of clay (or DNA) that happen to have properties that make them persist and spread themselves about.

Now to move on to the next stage of the argument. Some lineages of crystals might happen to catalyse the synthesis of new substances that assist in their passage down the ‘generations’. These secondary substances would not (not at first, anyway) have had their own lineages of ancestry and descent, but would have been manufactured anew by each generation of primary replicators. They could be seen as tools of the replicating crystal lineages, the beginnings of primitive ‘phenotypes’. Cairns-Smith believes that organic molecules were prominent among non-replicating ‘tools’ of his inorganic crystalline replicators. Organic molecules frequently are used in the commercial inorganic chemical industry because of their effects on the flow of fluids, and on the break-up or growth of inorganic particles: just the sorts of effects, in short, that could have influenced the ‘success’ of lineages of replicating crystals. For instance, a clay mineral with the lovely name montmorillonite tends to break up in the presence of small amounts of an organic molecule with the less-lovely name carboxymethyl cellulose. Smaller quantities of carboxymethyl cellulose, on the other hand, have just the opposite effect, helping to stick montmorillonite particles together. Tannins, another kind of organic molecule, are used in the oil industry to make muds easier to drill. If oil-drillers can exploit organic molecules to manipulate the flow and drillability of mud, there is no reason why cumulative selection should not have led to the same kind of exploitation by self-replicating minerals.

At this point Cairns-Smith's theory gets a sort of free bonus of added plausibility. It so happens that other chemists, supporting more conventional organic ‘primeval soup’ theories, have long accepted that  {157}  clay minerals would have been a help. To quote one of them (D. M. Anderson), ‘It is widely accepted that some, perhaps many, of the abiotic chemical reactions and processes leading to the origin on Earth of replicating micro-organisms occurred very early in the history of Earth in close proximity to the surfaces of clay minerals and other inorganic substrates.’ This writer goes on to list five ‘functions’ of clay minerals in assisting the origin of organic life, for instance ‘Concentration of chemical reactants by adsorption’. We needn't spell the five out here, or even understand them. From our point of view, what matters is that each of these five ‘functions’ of clay minerals can be twisted round the other way. It shows the close association that can exist between organic chemical synthesis and clay surfaces. It is therefore a bonus for the theory that clay replicators synthesized organic molecules and used them for their own purposes.

Cairns-Smith discusses, in more detail than I can accommodate here, early uses that his clay-crystal replicators might have had for proteins, sugars and, most important of all, nucleic acids like RNA. He suggests that RNA was first used for purely structural purposes, as oil drillers use tannins or we use soap and detergents. RNA-like molecules, because of their negatively charged backbones, would tend to coat the outsides of clay particles. This is getting us into realms of chemistry that are beyond our scope. For our purposes what matters is that RNA, or something like it, was around for a long time before it became self-replicating. When it finally did become self-replicating, this was a device evolved by the mineral crystal ‘genes’ to improve the efficiency of manufacture of the RNA (or similar molecule). But, once a new self-replicating molecule had come into existence, a new kind of cumulative selection could get going. Originally a side-show, the new replicators turned out to be so much more efficient than the original crystals that they took over. They evolved further, and eventually perfected the DNA code that we know today. The original mineral replicators were cast aside like worn-out scaffolding, and all modern life evolved from a relatively recent common ancestor, with a single, uniform genetic system and a largely uniform biochemistry.

In The Selfish Gene I speculated that we may now be on the threshold of a new kind of genetic takeover. DNA replicators built ‘survival machines’ for themselves — the bodies of living organisms including ourselves. As part of their equipment, bodies evolved on-board computers — brains. Brains evolved the capacity to communicate with other brains by means of language and cultural traditions. But the new milieu of cultural tradition opens up new possibilities for self-replicating entities. The new replicators are not DNA and they are not  {158}  clay crystals. They are patterns of information that can thrive only in brains or the artificially manufactured products of brains — books, computers, and so on. But, given that brains, books and computers exist, these new replicators, which I called memes to distinguish them from genes, can propagate themselves from brain to brain, from brain to book, from book to brain, from brain to computer, from computer to computer. As they propagate they can change — mutate. And perhaps ‘mutant’ memes can exert the kinds of influence that I am here calling ‘replicator power’. Remember that this means any kind of influence affecting their own likelihood of being propagated. Evolution under the influence of the new replicators — memic evolution — is in its infancy. It is manifested in the phenomena that we call cultural evolution. Cultural evolution is many orders of magnitude faster than DNA-based evolution, which sets one even more to thinking of the idea of ‘takeover’. And if a new kind of replicator takeover is beginning, it is conceivable that it will take off so far as to leave its parent DNA (and its grandparent clay if Cairns-Smith is right) far behind. If so, we may be sure that computers will be in the van.

Could it be that one far-off day intelligent computers will speculate about their own lost origins? Will one of them tumble to the heretical truth, that they have sprung from a remote, earlier form of life, rooted in organic, carbon chemistry, rather than the silicon-based electronic principles of their own bodies? Will a robotic Cairns-Smith write a book called Electronic Takeover? Will he rediscover some electronic equivalent of the metaphor of the arch, and realize that computers could not have sprung spontaneously into existence but must have originated from some earlier process of cumulative selection? Will he go into detail and reconstruct DNA as a plausible early replicator, victim of electronic usurpation? And will he be far-sighted enough to guess that even DNA may itself have been a usurper of yet more remote and primitive replicators, crystals of inorganic silicates? If he is of a poetic turn of mind, will he even see a kind of justice in the eventual return to silicon-based life, with DNA no more than an interlude, albeit one that lasted longer than three aeons?

That is science fiction, and it probably sounds far-fetched. That doesn't matter. Of more immediate moment is that Cairns-Smith's own theory, and indeed all other theories of the origin of life, may sound far-fetched to you and hard to believe. Do you find both Cairns-Smith's clay theory, and the more orthodox organic primeval-soup theory, wildly improbable? Does it sound to you as though it would need a miracle to make randomly jostling atoms join together into a self-replicating molecule? Well, at times it does to me too. But  {159}  let's look more deeply into this matter of miracles and improbability. By doing so, I shall demonstrate a point which is paradoxical but all the more interesting for that. This is that we should, as scientists, be even a little worried if the origin of life did not seem miraculous to our own human consciousness. An apparently (to ordinary human consciousness) miraculous theory is exactly the kind of theory we should be looking for in this particular matter of the origin of life. This argument, which amounts to a discussion of what we mean by a miracle, will occupy the rest of this chapter. In a way it is an extension of the argument we made earlier about billions of planets.

So, what do we mean by a miracle? A miracle is something that happens, but which is exceedingly surprising. If a marble statue of the Virgin Mary suddenly waved its hand at us we should treat it as a miracle, because all our experience and knowledge tells us that marble doesn't behave like that. I have just uttered the words ‘May I be struck by lightning this minute’. If lightning did strike me in the same minute, it would be treated as a miracle. But actually neither of these two occurrences would be classified by science as utterly impossible. They would simply be judged very improbable, the waving statue much more improbable than the lightning. Lightning does strike people. Any one of us might be struck by lightning, but the probability is pretty low in any one minute (although the Guinness Book of Records has a charming picture of a Virginian man, nicknamed the human lightning conductor, recovering in hospital from his seventh lightning strike, with an expression of apprehensive bewilderment on his face). The only thing miraculous about my hypothetical story is the coincidence between my being struck by lightning and my verbal invocation of the disaster.

Coincidence means multiplied improbability. The probability of my being struck by lightning in any one minute of my life is perhaps 1 in 10 million as a conservative estimate. The probability of my inviting a lightning strike in any particular minute is also very low. I have just done it for the only time in the 23,400,000 minutes of my life so far, and I doubt if I'll do it again, so call these odds one in 25 million. To calculate the joint probability of the coincidence occurring in any one minute we multiply the two separate probabilities. For my rough calculation this comes to about one in 250 trillion. If a coincidence of this magnitude happened to me, I should call it a miracle and would watch my language in future. But although the odds against the coincidence are extremely high, we can still calculate them. They are not literally zero.

In the case of the marble statue, molecules in solid marble are  {160}  continuously jostling against one another in random directions. The jostlings of the different molecules cancel one another out, so the whole hand of the statue stays still. But if, by sheer coincidence, all the molecules just happened to move in the same direction at the same moment, the hand would move. If they then all reversed direction at the same moment the hand would move back. In this way it is possible for a marble statue to wave at us. It could happen. The odds against such a coincidence are unimaginably great but they are not incalculably great. A physicist colleague has kindly calculated them for me. The number is so large that the entire age of the universe so far is too short a time to write out all the noughts! It is theoretically possible for a cow to jump over the moon with something like the same improbability. The conclusion to this part of the argument is that we can calculate our way into regions of miraculous improbability far greater than we can imagine as plausible.

Let's look at this matter of what we think is plausible. What we can imagine as plausible is a narrow band in the middle of a much broader spectrum of what is actually possible. Sometimes it is narrower than what is actually there. There is a good analogy with light. Our eyes are built to cope with a narrow band of electromagnetic frequencies (the ones we call light), somewhere in the middle of the spectrum from long radio waves at one end to short X-rays at the other. We can't see the rays outside the narrow light band, but we can do calculations about them, and we can build instruments to detect them. In the same way, we know that the scales of size and time extend in both directions far outside the realm of what we can visualize. Our minds can't cope with the large distances that astronomy deals in or with the small distances that atomic physics deals in, but we can represent those distances in mathematical symbols. Our minds can't imagine a time span as short as a picosecond, but we can do calculations about picoseconds, and we can build computers that can complete calculations within picoseconds. Our minds can't imagine a timespan as long as a million years, let alone the thousands of millions of years that geologists routinely compute.

Just as our eyes can see only that narrow band of electromagnetic frequencies that natural selection equipped our ancestors to see, so our brains are built to cope with narrow bands of sizes and times. Presumably there was no need for our ancestors to cope with sizes and times outside the narrow range of everyday practicality, so our brains never evolved the capacity to imagine them. It is probably significant that our own body size of a few feet is roughly in the middle of the range of sizes we can imagine. And our own lifetime of a few decades is roughly in the middle of the range of times we can imagine.

We can say the same kind of thing about improbabilities and miracles.  {161}  Picture a graduated scale of improbabilities, analogous to the scale of sizes from atoms to galaxies, or to the scale of times from picoseconds to aeons. On the scale we mark off various landmark points. At the far left-hand end of the scale are events which are all but certain, such as the probability that the sun will rise tomorrow — the subject of G. H. Hardy's halfpenny bet. Near this left-hand end of the scale are things that are only slightly improbable, such as shaking a double six in a single throw of a pair of dice. The odds of this happening are 1 in 36. I expect we've all done it quite often. Moving towards the right-hand end of the spectrum, another landmark point is the probability of a perfect deal in bridge, where each of the four players receives a complete suit of cards. The odds against this happening are 2,235,197,406,895,366,368,301,559,999 to 1. Let us call this one dealion, the unit of improbability. If something with an improbability of one dealion was predicted and then happened, we should diagnose a miracle unless, which is more probable, we suspected fraud. But it could happen with a fair deal, and it is far far far more probable than the marble statue's waving at us. Nevertheless, even this latter event, as we have seen, has its rightful place along the spectrum of events that could happen. It is measurable, albeit in units far larger than gigadealions. Between the double-six dice throw, and the perfect deal at bridge, is a range of more or less improbable events that do sometimes happen, including any one individual's being struck by lightning, winning a big prize on the football pools, scoring a hole-in-one at golf, and so on. Somewhere in this range, too, are those coincidences that give us an eerie spine-tingling feeling, like dreaming of a particular person for the first time in decades, then waking up to find that they died in the night. These eerie coincidences are very impressive when they happen to us or to one of our friends, but their improbability is measured in only picodealions.

Having constructed our mathematical scale of improbabilities, with its benchmark or landmark points marked on it, let us now turn a spotlight on that subrange of the scale with which we, in our ordinary thought and conversation, can cope. The width of the spotlight's beam is analogous to the narrow range of electromagnetic frequencies that our eyes can see, or to the narrow range of sizes or times, close to our own size and longevity, that we can imagine. On the spectrum of improbabilities, the spotlight turns out to illuminate only the narrow range from the left-hand end (certainty) up to minor miracles, like a hole-in-one or a dream that comes true. There is a vast range of mathematically calculable improbabilities way outside the range of the spotlight.  {162} 

Our brains have been built by natural selection to assess probability and risk, just as our eyes have been built to assess electromagnetic wavelength. We are equipped to make mental calculations of risk and odds, within the range of improbabilities that would be useful in human life. This means risks of the order of, say, being gored by a buffalo if we shoot an arrow at it, being struck by lightning if we shelter under a lone tree in a thunderstorm, or drowning if we try to swim across a river. These acceptable risks are commensurate with our lifetimes of a few decades. If we were biologically capable of living for a million years, and wanted to do so, we should assess risks quite differently. We should make a habit of not crossing roads, for instance, for if you crossed a road every day for half a million years you would undoubtedly be run over.

Evolution has equipped our brains with a subjective consciousness of risk and improbability suitable for creatures with a lifetime of less than one century. Our ancestors have always needed to take decisions involving risks and probabilities, and natural selection has therefore equipped our brains to assess probabilities against a background of the short lifetime that we can, in any case, expect. If on some planet there are beings with a lifetime of a million centuries, their spotlight of comprehensible risk will extend that much farther towards the right-hand end of the continuum. They will expect to be dealt a perfect bridge hand from time to time, and will scarcely trouble to write home about it when it happens. But even they will blench if a marble statue waves at them, for you would have to live dealions of years longer than even they do to see a miracle of this magnitude.

What has all this to do with theories of the origin of life? Well, we began this argument by agreeing that Cairns-Smith's theory, and the primeval-soup theory, sound a bit far-fetched and improbable to us. We naturally feel inclined to reject these theories for that reason. But ‘we’, remember, are beings whose brains are equipped with a spotlight of comprehensible risk that is a pencil-thin beam illuminating the far left-hand end of the mathematical continuum of calculable risks. Our subjective judgement of what seems like a good bet is irrelevant to what is actually a good bet. The subjective judgement of an alien with a lifetime of a million centuries will be quite different. He will judge as quite plausible an event, such as the origin of the first replicating molecule as postulated by some chemist's theory, which we, kitted up by evolution to move in a world of a few decades’ duration, would judge to be an astounding miracle. How can we decide whose point of view is the right one, ours or the long-lived alien's?

There is a simple answer to this question. The long-lived alien's  {163}  point of view is the right one for looking at the plausibility of a theory like Cairns-Smith's or the primeval-soup theory. This is because those two theories postulate a particular event — the spontaneous arising of a self-replicating entity — as occurring only once in about a billion years, once per aeon. One and a half aeons is about the time that elapsed between the origin of the Earth and the first bacteria-like fossils. For our decade-conscious brains, an event that happens only once per aeon is so rare as to seem a major miracle. For the long-lived alien, it will seem less of a miracle than a golf hole-in-one seems to us — and most of us probably know somebody who knows somebody who has scored a hole-in-one. In judging theories of the origin of life, the long-lived alien's subjective timescale is the relevant one, because it is approximately the timescale involved in the origin of life. Our own subjective judgement about the plausibility of a theory of the origin of life is likely to be wrong by a factor of a hundred million.

In fact our subjective judgement is probably wrong by an even greater margin. Not only are our brains equipped by nature to assess risks of things in a short time; they are also equipped to assess risks of things happening to us personally, or to a narrow circle of people that we know. This is because our brains didn’t evolve under conditions dominated by mass media. Mass reporting means that, if an improbable thing happens to anybody, anywhere in the world, we shall read about it in our newspapers or in the Guinness Book of Records. If an orator, anywhere in the world, publicly challenged the lightning to strike him if he lied, and it promptly did so, we should read about it and be duly impressed. But there are several billion people in the world to whom such a coincidence could happen, so the apparent coincidence is actually not as great as it seems. Our brains are probably equipped by nature to assess the risks of things happening to ourselves, or to a few hundred people in the small circle of villages within drum-range that our tribal ancestors could expect to hear news about. When we read in a newspaper about an amazing coincidence happening to somebody in Valparaiso or Virginia, we are more impressed by it than we should be. More impressed by a factor of perhaps a hundred million, if that is the ratio between the world population surveyed by our newspapers, and the tribal population about whom our evolved brains ‘expect’ to hear news.

This ‘population calculation’ is also relevant to our judgement of the plausibility of theories of the origin of life. Not because of the population of people on Earth, but because of the population of planets in the universe, the population of planets where life could have originated. This is just the argument we met earlier in this chapter, so  {164}  there is no need to dwell on it here. Go back to our mental picture of a graduated scale of improbable events with its benchmark coincidences of bridge hands and dice throws. On this graduated scale of dealions and microdealions, mark the following three new points. Probability of life arising on a planet (in, say, a billion years), if we assume that life arises at a rate of about once per solar system. Probability of life arising on a planet if life arises at a rate of about once per galaxy. Probability of life on a randomly selected planet if life arose only once in the universe. Label these three points respectively the Solar System Number, the Galaxy Number and the Universe Number. Remember that there are about 10,000 million galaxies. We don’t know how many solar systems there are in each galaxy because we can only see stars, not planets, but we earlier used an estimate that there may be 100 billion billion planets in the universe.

When we assess the improbability of an event postulated by, for instance the Cairns-Smith theory, we should assess it, not against what we subjectively think of as probable or improbable, but against numbers like these three numbers, the Solar System Number, the Galaxy Number and the Universe Number. Which of these three numbers is the most appropriate depends upon which of the following three statements we think is nearest the truth:

1. Life has arisen in only one planet in the entire universe (and that planet, as we saw earlier, then has to be Earth).

2. Life has arisen on about one planet per galaxy (in our galaxy, Earth is the lucky planet).

3. The origin of life is a sufficiently probable event that it tends to arise about once per solar system (in our solar system Earth is the lucky planet).

These three statements represent three benchmark views about the uniqueness of life. The actual uniqueness of life probably lies somewhere between the extremes represented by Statement 1 and Statement 3. Why do I say that? Why, in particular, should we rule out a fourth possibility, that the origin of life is a far more probable event than is suggested by Statement 3? It isn't a strong argument, but, for what it is worth, it goes like this. If the origin of life were a much more probable event than is suggested by the Solar System Number we should expect, by now, to have encountered extraterrestrial life, if not in (whatever passes for) the flesh, at least by radio.

It is often pointed out that chemists have failed in their attempts to duplicate the spontaneous origin of life in the laboratory. This fact is  {165}  used as if it constituted evidence against the theories that those chemists are trying to test. But actually one can argue that we should be worried if it turned out to be very easy for chemists to obtain life spontaneously in the test-tube. This is because chemists’ experiments last for years, not thousands of millions of years, and because only a handful of chemists, not thousands of millions of chemists, are engaged in doing these experiments. If the spontaneous origin of life turned out to be a probable enough event to have occurred during the few man-decades in which chemists have done their experiments, then life should have arisen many times on Earth, and many times on planets within radio range of Earth. Of course all this begs important questions about whether chemists have succeeded in duplicating the conditions of the early Earth but, even so, given that we can't answer these questions, the argument is worth pursuing.

If the origin of life were a probable event by ordinary human standards, then a substantial number of planets within radio range should have developed a radio technology long enough ago (bearing in mind that radio waves travel at 186,000 miles per second) for us to have picked up at least one transmission during the decades that we have been equipped to do so. There are probably about 50 stars within radio range if we assume that they have had radio technology for only as long as we have. But 50 years is just a fleeting instant, and it would be a major coincidence if another civilization were so closely in step with us. If we embrace in our calculation those civilizations that had radio technology 1,000 years ago, there will be something like a million stars within radio range (together with however many planets circle round each one of them). If we include those whose radio technology goes back 100,000 years, the whole trillion-star galaxy would be within radio range. Of course, broadcast signals would become pretty attenuated over such huge distances.

So we have arrived at the following paradox. If a theory of the origin of life is sufficiently ‘plausible’ to satisfy our subjective judgement of plausibility, it is then too ‘plausible’ to account for the paucity of life in the universe as we observe it. According to this argument, the theory we are looking for has got to be the kind of theory that seems implausible to our limited, Earth-bound, decade-bound imaginations. Seen in this light, both Cairns-Smith's theory and the primeval-soup theory seem if anything in danger of erring on the side of being too plausible! Having said all this I must confess that, because there is so much uncertainty in the calculations, if a chemist did succeed in creating spontaneous life I would not actually be disconcerted!

We still don’t know exactly how natural selection began on Earth.  {166}  This chapter has had the modest aim of explaining only the kind of way in which it must have happened. The present lack of a definitely accepted account of the origin of life should certainly not be taken as a stumbling block for the whole Darwinian world view, as it occasionally — probably with wishful thinking — is. The earlier chapters have disposed of other alleged stumbling blocks, and the next chapter takes up yet another one, the idea that natural selection can only destroy, never construct.

CHAPTER 7

CONSTRUCTIVE EVOLUTION

People sometimes think that natural selection is a purely negative force, capable of weeding out freaks and failures, but not capable of building up complexity, beauty and efficiency of design. Does it not merely subtract from what is already there, and shouldn't a truly creative process add something too? One can partially answer this by pointing to a statue. Nothing is added to the block of marble. The sculptor only subtracts, but a beautiful statue emerges nevertheless. But this metaphor can mislead, for some people leap straight to the wrong part of the metaphor — the fact that the sculptor is a conscious designer — and miss the important part: the fact that the sculptor works by subtraction rather than addition. Even this part of the metaphor should not be taken too far. Natural selection may only subtract, but mutation can add. There are ways in which mutation and natural selection together can lead, over the long span of geological time, to a building up of complexity that has more in common with addition than with subtraction. There are two main ways in which this build-up can happen. The first of these goes under the name of ‘coadapted genotypes’; the second under the name of ‘arms races’. The two are superficially rather different from one another, but they are united under the headings of ‘coevolution’ and ‘genes as each others' environments’.

First, the idea of ‘coadapted genotypes’. A gene has the particular effect that it does only because there is an existing structure upon which to work. A gene can't affect the wiring up of a brain unless there is a brain being wired up in the first place. There won't be a brain being wired up in the first place, unless there is a complete developing embryo. And there won't be a complete developing embryo unless  {170}  there is a whole program of chemical and cellular events, under the influence of lots and lots of other genes, and lots and lots of other, non-genetic, causal influences. The particular effects that genes have are not intrinsic properties of those genes. They are properties of embryological processes, existing processes whose details may be changed by genes, acting in particular places and at particular times during embryonic development. We saw this message demonstrated, in elementary form, by the development of the computer biomorphs.

In a sense, the whole process of embryonic development can be looked upon as a cooperative venture, jointly run by thousands of genes together. Embryos are put together by all the working genes in the developing organism, in collaboration with one another. Now comes the key to understanding how such collaborations come about. In natural selection, genes are always selected for their capacity to flourish in the environment in which they find themselves. We often think of this environment as the outside world, the world of predators and climate. But from each gene's point of view, perhaps the most important part of its environment is all the other genes that it encounters. And where does a gene ‘encounter’ other genes? Mostly in the cells of the successive individual bodies in which it finds itself. Each gene is selected for its capacity to cooperate successfully with the population of other genes that it is likely to meet in bodies.

The true population of genes, which constitutes the working environment of any given gene, is not just the temporary collection that happens to have come together in the cells of any particular individual body. At least in sexually reproducing species, it is the set of all genes in the population of interbreeding individuals — the gene ‘pool’. At any given moment, any particular copy of a gene, in the sense of a particular collection of atoms, must be sitting in one cell of one individual. But the set of atoms that is any one copy of a gene is not of permanent interest. It has a life-expectancy measured only in months. As we have seen, the long-lived gene as an evolutionary unit is not any particular physical structure but the textual archival information that is copied on down the generations. This textual replicator has a distributed existence. It is widely distributed in space among different individuals, and widely distributed in time over many generations. When looked at in this distributed way, any one gene can be said to ‘meet’ another when they find themselves sharing a body. It can ‘expect’ to meet a variety of other genes in different bodies at different times in its distributed existence, and in its march through geological time. A successful gene will be one that does well in the environments provided by these other genes that it is likely to meet in lots of different  {171}  bodies. ‘Doing well’ in such environments will turn out to be equivalent to ‘collaborating’ with these other genes. It is most directly seen in the case of biochemical pathways.

Biochemical pathways are sequences of chemicals that constitute successive stages in some useful process, like the release of energy or the synthesis of an important substance. Each step in the pathway needs an enzyme — one of those large molecules that is shaped to act like a machine in a chemical factory. Different enzymes are needed for different steps in the chemical pathway. Sometimes there are two, or more, alternative chemical pathways to the same useful end. Although both pathways culminate in the identical useful result, they have different intermediate stages leading up to that end, and they normally have different starting points. Either of the two alternative pathways will do the job, and it doesn't matter which one is used. The important thing for any particular animal is to avoid trying to do both at once, for chemical confusion and inefficiency would result.

Now suppose that Pathway 1 needs the succession of enzymes A1, B1 and C1, in order to synthesize a desired chemical D, while Pathway 2 needs enzymes A2, B2 and C2 in order to arrive at the same desirable end-product. Each enzyme is made by a particular gene. So, in order to evolve the assembly line for Pathway 1, a species needs the genes coding for A1, B1 and C1 all to coevolve together. In order to evolve the alternative assembly line for Pathway 2, a species would need the genes coding for A2, B2 and C2 to coevolve with one another. The choice between these two coevolutions doesn't come about through advance planning. It comes about simply through each gene being selected by virtue of its compatibility with the other genes that already happen to dominate the population. If the population happens to be already rich in genes for B1 and C1, this will set up a climate favouring the Al gene rather than the A2 gene. Conversely, if the population is already rich in genes for B2 and C2 this will set up a climate in which the A2 gene is favoured by selection rather than the A1 gene.

It will not be as simple as that, but you will have got the idea: one of the most important aspects of the ‘climate’ in which a gene is favoured or disfavoured is the other genes that are already numerous in the population; the other genes, therefore, with which it is likely to have to share bodies. Since the same will obviously be true of these ‘other’ genes themselves, we have a picture of teams of genes all evolving towards cooperative solutions to problems. The genes themselves don’t evolve, they merely survive or fail to survive in the gene pool. It is the ‘team’ that evolves. Other teams might have done the job just as  {172}  well, or even better. But once one team has started to dominate the gene pool of a species it thereby has an automatic advantage. It is difficult for a minority team to break in, even a minority team which would, in the end, have done the job more efficiently. The majority team has an automatic resistance to being displaced, simply by virtue of being in the majority. This doesn't mean that the majority team can never be displaced. If it couldn't, evolution would grind to a halt. But it does mean that there is a kind of built-in inertia.

Obviously this kind of argument is not limited to biochemistry. We could make the same kind of case for clusters of compatible genes building the different parts of eyes, ears, noses, walking limbs, all the cooperating parts of an animal's body. Genes for making teeth suitable for chewing meat tend to be favoured in a ‘climate’ dominated by genes making guts suitable for digesting meat. Conversely, genes for making plant-grinding teeth tend to be favoured in a climate dominated by genes that make guts suitable for digesting plants. And vice versa in both cases. Teams of ‘meat-eating genes’ tend to evolve together, and teams of ‘plant-eating genes’ tend to evolve together. Indeed, there is a sense in which most of the working genes in a body can be said to cooperate with each other as a team, because over evolutionary time they (i.e. ancestral copies of themselves) have each been part of the environment in which natural selection has worked on the others. If we ask why the ancestors of lions took to meat-eating, while the ancestors of antelopes took to grass-eating, the answer could be that originally it was an accident. An accident, in the sense that it could have been the ancestors of lions that took up grass-eating, and the ancestors of antelopes that took up meat-eating. But once one lineage had begun to build up a team of genes for dealing with meat rather than grass, the process was self-reinforcing. And once the other lineage had begun to build up a team of genes for dealing with grass rather than meat, that process was self-reinforcing in the other direction.

One of the main things that must have happened in the early evolution of living organisms was an increase in the numbers of genes participating in such cooperatives. Bacteria have far fewer genes than animals and plants. The increase may have come about through various kinds of gene duplication. Remember that a gene is just a length of coded symbols, like a file on a computer disc; and genes can be copied to different parts of the chromosomes, just as files can be copied to different parts of the disc. On my disc that holds this chapter there are officially just three files. By ‘officially’ I mean that the computer's operating system tells me that there are just three files. I can ask it to read one of these three files, and it presents me with a  {173}  one-dimensional array of alphabetical characters, including the characters that you are now reading. All very neat and orderly, it seems. But in fact, on the disc itself, the arrangement of the text is anything but neat and orderly. You can see this if you break away from the discipline of the computer's own official operating system, and write your own private programs to decipher what is actually written on every sector of the disc. It turns out that fragments of each of my three files are dotted around, interleaved with each other and with fragments of old, dead files that I erased long ago and had forgotten. Any given fragment may turn up, word for word the same, or with minor differences, in half a dozen different places all around the disc.

The reason for this is interesting, and worth a digression because it provides a good genetic analogy. When you tell a computer to delete a file, it appears to obey you. But it doesn't actually wipe out the text of that file. It simply wipes out all pointers to that file. It is as though a librarian, ordered to destroy Lady Chatterley's Lover, simply tore up the card from the card index, leaving the book itself on the shelf. For the computer, this is a perfectly economical way to do things, because the space formerly occupied by the ‘deleted’ file is automatically available for new files, as soon as the pointers to the old file have been removed. It would be a waste of time actually to go to the trouble of filling the space itself with blanks. The old file won't itself be finally lost until all its space happens to be used for storing new files.

But this re-using of space occurs piecemeal. New files aren't exactly the same size as old ones. When the computer is trying to save a new file to a disc, it looks for the first available fragment of space, writes as much of the new file as will fit, then looks for another available fragment of space, writes a bit more, and so on until all the file is written somewhere on the disc. The human has the illusion that the file is a single, orderly array, only because the computer is careful to keep records ‘pointing’ to the addresses of all the fragments dotted around. These ‘pointers’ are like the ‘continued on page 94’ pointers used by the New York Times. The reason many copies of any one fragment of text are found on a disc is that if, like all my chapters, the text has been edited and re-edited many dozens of times, each edit will result in a new saving to the disc of (almost) the same text. The saving may ostensibly be a saving of the same file. But as we have seen, the text will in fact be repeatedly scattered around the available ‘gaps’ on the disc. Hence multiple copies of a given fragment of text can be found all around the surface of the disc, the more so if the disc is old and much used.

Now the DNA operating system of a species is very very old indeed,  {174}  and there is evidence that it, seen in the long term, does something a bit like the computer with its disc files. Part of the evidence comes from the fascinating phenomenon of ‘introns’ and ‘exons’. Within the last decade, it has been discovered that any ‘single’ gene, in the sense of a single continuously read passage of DNA text, is not all stored in one place. If you actually read the code letters as they occur along the chromosome (i.e. if you do the equivalent of breaking out of the discipline of the ‘operating system’) you find fragments of ‘sense’, called exons, separated by portions of ‘nonsense’ called introns. Any one ‘gene’ in the functional sense, is in fact split up into a sequence of fragments (exons) separated by meaningless introns. It is as if each exon ended with a pointer saying ‘continued on page 94’. A complete gene is then made up of a whole series of exons, which are actually strung together only when they are eventually read by the ‘official’ operating system that translates them into proteins.

Further evidence comes from the fact that the chromosomes are littered with old genetic text that is no longer used, but which still makes recognizable sense. To a computer programmer, the pattern of distribution of these ‘genetic fossil’ fragments is uncannily reminiscent of the pattern of text on the surface of an old disc that has been much used for editing text. In some animals, a high proportion of the total number of genes is in fact never read. These genes are either complete nonsense, or they are outdated ‘fossil genes’.

Just occasionally, textual fossils come into their own again, as I experienced when writing this book. A computer error (or, to be fair, it may have been human error) caused me accidentally to ‘erase’ the disc containing Chapter 3. Of course the text itself hadn't literally all been erased. All that had been definitely erased were the pointers to where each ‘exon’ began and ended. The ‘official’ operating system could read nothing, but ‘unofficially’ I could play genetic engineer and examine all the text on the disc. What I saw was a bewildering jigsaw puzzle of textual fragments, some of them recent, others ancient ‘fossils’. By piecing together the jigsaw fragments, I was able to recreate the chapter. But I mostly didn’t know which fragments were recent and which were fossil. It didn’t matter for, apart from minor details that necessitated some new editing, they were the same. At least some of the ‘fossils’, or outdated ‘introns’, had come into their own again. They rescued me from my predicament, and saved me the trouble of rewriting the entire chapter.

There is evidence that, in living species too, ‘fossil genes’ occasionally come into their own again, and are re-used after lying dormant for a million years or so. To go into detail would carry us too  {175}  far from the main pathway of this chapter, for you will remember that we are already out on a digression. The main point was that the total genetic capacity of a species may increase due to gene duplication. Reusing of old ‘fossil’ copies of existing genes is one way in which this can happen. There are other, more immediate, ways in which genes may be copied to widely distributed parts of the chromosomes, like files being duplicated to different parts of a disc, or different discs.

Humans have eight separate genes called globin genes (used for making haemoglobin, among other things), on various different chromosomes. It seems certain that all eight have been copied, ultimately from a single ancestral globin gene. About 1,100 million years ago, the ancestral globin gene duplicated, forming two genes. We can date this event because of independent evidence about how fast globins habitually evolve (see Chapters 5 and 11). Of the two genes produced by this original duplication, one became the ancestor of all the genes that make haemoglobin in vertebrates. The other became the ancestor of all the genes that make myoglobins, a related family of proteins that work in muscles. Various subsequent duplications have given rise to the so-called alpha, beta, gamma, delta, epsilon and zeta globins. The fascinating thing is that we can construct a complete family tree of all the globin genes, and even put dates on all the divergence points (delta and beta globin parted company, for example, about 40 million years ago, epsilon and gamma globins 100 million years ago). Yet the eight globins, descendants as they are of these remote branchings in distant ancestors, are still all present inside every one of us. They diverged to different parts of an ancestor's chromosomes, and we have each inherited them on our different chromosomes. Molecules are sharing the same body with their remote molecular cousins. It is certain that a great deal of such duplication has gone on, all over the chromosomes, and throughout geological time. This is an important respect in which real life is more complicated than the biomorphs of Chapter 3. They all had only nine genes. They evolved by changes in those nine genes, never by increasing the number of genes to ten. Even in real animals, such duplications are rare enough not to invalidate my general statement that all members of a species share the same DNA ‘addressing’ system.

Duplication within the species isn't the only means by which the number of cooperating genes has increased in evolution. An even rarer, but still possibly very important occurrence, is the occasional incorporation of a gene from another species, even an extremely remote species. There are, for example, haemoglobins in the roots of plants of the pea family. They don’t occur in any other plant families, and it  {176}  seems almost certain that they somehow got into the pea family by cross-infection from animals, viruses perhaps acting as intermediaries.

An especially important event along these lines, according to the increasingly favoured theory of the American biologist Lynn Margulis, took place at the origin of the so-called eukaryotic cell. Eukaryotic cells include all cells except those of bacteria. The living world is divided, fundamentally, into bacteria versus the rest. We are part of the rest, and are collectively called the eukaryotes. We differ from bacteria mainly in that our cells have discrete little mini-cells inside them. These include the nucleus, which houses the chromosomes; the tiny bomb-shaped objects called mitochondria (which we briefly met in Figure 1), filled with intricately folded membranes; and, in the (eukaryotic) cells of plants, chloroplasts. Mitochondria and chloroplasts have their own DNA, which replicates and propagates itself entirely independently of the main DNA in the chromosomes of the nucleus. All the mitochondria in you are descended from the small population of mitochondria that travelled from your mother in her egg. Sperms are too small to contain mitochondria, so mitochondria travel exclusively down the female line, and male bodies are dead ends as far as mitochondrial reproduction is concerned. Incidentally, this means that we can use mitochondria to trace our ancestry, strictly down the female line.

Margulis's theory is that mitochondria and chloroplasts, and a few other structures inside cells, are each descended from bacteria. The eukaryotic cell was formed, perhaps 2 billion years ago, when several kinds of bacteria joined forces because of the benefits that each could obtain from the others. Over the aeons they have become so thoroughly integrated into the cooperative unit that became the eukaryotic cell, that it has become almost impossible to detect the fact, if indeed it is a fact, that they were once separate bacteria.

It seems that, once the eukaryotic cell had been invented, a whole new range of designs became possible. Most interestingly from our point of view, cells could manufacture large bodies comprising many billions of cells. All cells reproduce by splitting into two, both halves getting a full set of genes. As we saw in the case of the bacteria on a pin's head, successive splittings into two can generate a very large number of cells in rather a short time. You start with one and it splits into two. Then each of the two splits, making four. Each of the four splits, making eight. The numbers go up by successive doublings, from 8 to 16, 32, 64, 128, 256, 512, 1,024, 2,048, 4,096, 8,192. After only 20 doublings, which doesn't take very long, we are up in the millions. After only 40 doublings the number of cells is more than a trillion. In  {177}  the case of bacteria, the enormous numbers of cells produced by successive doublings go their separate ways. The same is true of many eukaryotic cells, for instance protozoa such as amoebas. A major step in evolution was taken when cells that had been produced by successive splittings stuck together instead of going off independently. Higher-order structure could now emerge, just as it did, on an incomparably smaller scale, in the two-way branching computer biomorphs.

Now, for the first time, large body size became a possibility. A human body is a truly colossal population of cells, all descended from one ancestor, the fertilized egg; and all therefore cousins, children, grandchildren, uncles, etc. of other cells in the body. The 10 trillion cells that make up each one of us are the product of a few dozens of generations of cell doublings. These cells are classified into about 210 (according to taste) different kinds, all built by the same set of genes but with different members of the set of genes turned on in different kinds of cells. This, as we have seen, is why liver cells are different from brain cells, and bone cells are different from muscle cells.

Genes working through the organs and behaviour patterns of many-celled bodies can achieve methods of ensuring their own propagation that are not available to single cells working on their own. Many-celled bodies make it possible for genes to manipulate the world, using tools built on a scale that is orders of magnitude larger than the scale of single cells. They achieve these large-scale indirect manipulations via their more direct effects on the miniature scale of cells. For instance, they change the shape of the cell membrane. The cells then interact with one another in huge populations to produce large-scale group effects such as an arm or a leg or (more indirectly) a beaver's dam. Most of the properties of an organism that we are equipped to see with our naked eyes are so-called ‘emergent properties’. Even the computer biomorphs, with their nine genes, had emergent properties. In real animals they are produced at the whole-body level by interactions between cells. An organism works as an entire unit, and its genes can be said to have effects on the whole organism, even though each copy of any one gene exerts its immediate effects only within its own cell.

We have seen that a very important part of a gene's environment is the other genes that it is likely to meet in successive bodies as the generations go by. These are the genes that are permuted and combined within the species. Indeed, a sexually reproducing species can be thought of as a device that permutes a discrete set of mutually accustomed genes in different combinations. Species, according to this view, are continually shuffling collections of genes that meet each  {178}  other within the species, but never meet genes of other species. But there is a sense in which the genes of different species, even if they don’t meet at close quarters inside cells, nevertheless constitute an important part of each others’ environment. The relationship is often hostile rather than cooperative, but this can be treated as just a reversal of sign. This is where we come to the second major theme of this chapter, ‘arms races’. There are arms races between predators and prey, parasites and hosts, even — though the point is a more subtle one and I shan't discuss it further — between males and females within one species.

Arms races are run in evolutionary time, rather than on the timescale of individual lifetimes. They consist of the improvement in one lineage's (say prey animals’) equipment to survive, as a direct consequence of improvement in another (say predators’) lineage's evolving equipment. There are arms races wherever individuals have enemies with their own capacity for evolutionary improvement. I regard arms races as of the utmost importance because it is largely arms races that have injected such ‘progressiveness’ as there is in evolution. For, contrary to earlier prejudices, there is nothing inherently progressive about evolution. We can see this if we consider what would have happened if the only problems animals had had to face had been those posed by the weather and other aspects of the nonliving environment.

After many generations of cumulative selection in a particular place, the local animals and plants become well fitted to the conditions, for instance the weather conditions, in that place. If it is cold the animals come to have thick coats of hair, or feathers. If it is dry they evolve leathery or waxy waterproof skins to conserve what little water there is. The adaptations to local conditions affect every part of the body, its shape and colour, its internal organs, its behaviour, and the chemistry in its cells.

If the conditions in which a lineage of animals lives remain constant, say it is dry and hot and has been so without a break for 100 generations, evolution in that lineage is likely to come to a halt, at least as far as adaptations to temperature and humidity are concerned. The animals will become as well fitted as they can be to the local conditions. This doesn't mean that they couldn't be completely redesigned to be even better. It does mean that they can't improve themselves by any small (and therefore likely) evolutionary step: none of their immediate neighbours in the local equivalent of ‘biomorph space’ would do any better.

Evolution will come to a standstill until something in the  {179}  conditions changes: the onset of an ice age, a change in the average rainfall of the area, a shift in the prevailing wind. Such changes do happen when we are dealing with a timescale as long as the evolutionary one. As a consequence, evolution normally does not come to a halt, but constantly ‘tracks’ the changing environment. If there is a steady downward drift in the average temperature in the area, a drift that persists over centuries, successive generations of animals will be propelled by a steady selection ‘pressure’ in the direction, say, of growing longer coats of hair. If, after a few thousand years of reduced temperature the trend reverses and average temperatures creep up again, the animals will come under the influence of a new selection pressure, and will be pushed towards growing shorter coats again.

But so far we have considered only a limited part of the environment, namely the weather. The weather is very important to animals and plants. Its patterns change as the centuries go by, so this keeps evolution constantly in motion as it ‘tracks’ the changes. But weather patterns change in a haphazard, inconsistent way. There are other parts of an animal's environment that change in more consistently malevolent directions, and that also need to be ‘tracked’. These parts of the environment are living things themselves. For a predator such as a hyena, a part of its environment that is at least as important as the weather is its prey, the changing populations of gnus, zebras and antelopes. For the antelopes and other grazers that wander the plains in search of grass, the weather may be important, but the lions, hyenas and other carnivores are important too. Cumulative selection will see to it that animals are well fitted to outrun their predators or outwit their prey, no less than it sees to it that they are well fitted to the prevailing weather conditions. And, just as long-term fluctuations in the weather are ‘tracked’ by evolution, so long-term changes in the habits or weaponry of predators will be tracked by evolutionary changes in their prey. And vice versa, of course.

We can use the general term ‘enemies’ of a species, to mean other living things that work to make life difficult. Lions are enemies of zebras. It may seem a little callous to reverse the statement to ‘Zebras are enemies of lions’. The role of the zebra in the relationship seems too innocent and wronged to warrant the pejorative ‘enemy’. But individual zebras do everything in their power to resist being eaten by lions, and from the lions’ point of view this is making life harder for them. If zebras and other grazers all succeeded in their aim, the lions would die of starvation. So by our definition zebras are enemies of lions. Parasites such as tapeworms are enemies of their hosts, and hosts are enemies of parasites since they tend to evolve measures to  {180}  resist them. Herbivores are enemies of plants, and plants are enemies of herbivores, to the extent that they manufacture thorns, and poisonous or nasty-tasting chemicals.

Lineages of animals and plants will, in evolutionary time, ‘track’ changes in their enemies no less assiduously than they track changes in average weather conditions. Evolutionary improvements in cheetah weaponry and tactics are, from the gazelles’ point of view, like a steady worsening of the climate, and they are tracked in the same kind of way. But there is one enormously important difference between the two. The weather changes over the centuries, but it does not change in a specifically malevolent way. It is not out to ‘get’ gazelles. The average cheetah will change over the centuries, just like the mean annual rainfall changes. But whereas mean annual rainfall will drift up and down, with no particular rhyme or reason, the average cheetah, as the centuries go by, will tend to become better equipped to catch gazelles than his ancestors were. This is because the succession of cheetahs, unlike the succession of annual weather conditions, is itself subject to cumulative selection. Cheetahs will tend to become fleeter of foot, keener of eye, sharper of tooth. However ‘hostile’ the weather and other inanimate conditions may seem to be, they have no necessary tendency to get steadily more hostile. Living enemies, seen over the evolutionary timescale, have exactly that tendency.

The tendency for carnivores to get progressively ‘better’ would soon run out of steam, as do human arms races (for reasons of economic cost which we shall come to), were it not for the parallel tendency in the prey. And vice versa. Gazelles, no less than cheetahs, are subject to cumulative selection, and they too will tend, as the generations go by, to improve their ability to run fast, to react swiftly, to become invisible by blending into the long grass. They too are capable of evolving in the direction of becoming better enemies, in this case enemies of cheetahs. From the cheetahs’ point of view the mean annual temperature does not get systematically better or worse as the years go by, except in so far as any change for a well-adapted animal is a change for the worse. But the mean annual gazelle does tend to get systematically worse — more difficult to catch because better adapted to evade cheetahs. Again, the tendency towards progressive improvement in gazelles would slow to a halt, were it not for the parallel tendency to improvement shown by their predators. One side gets a little better because the other side has. And vice versa. The process goes into a vicious spiral, on a timescale of hundreds of thousands of years.

In the world of nations on their shorter timescale, when two enemies each progressively improve their weaponry in response to the  {181}  other side's improvements, we speak of an ‘arms race’. The evolutionary analogy is close enough to justify borrowing the term, and I make no apology to my pompous colleagues who would purge our language of such illuminating images. I have introduced the idea here in terms of a simple example, gazelles and cheetahs. This was to get across the important difference between a living enemy, which itself is subject to evolutionary change, and an inanimate non-malevolent condition such as the weather, which is subject to change, but not systematic, evolutionary change. But the time has come to admit that in my efforts to explain this one valid point I may have misled the reader in other ways. It is obvious, when you come to think about it, that my picture of an ever-advancing arms race was too simple in at least one respect. Take running speed. As it stands so far, the arms-race idea seems to suggest that cheetahs and gazelles should have gone on, generation after generation, getting ever faster until both travelled faster than sound. This has not happened and it never will. Before resuming the discussion of arms races, it is my duty to forestall misunderstandings.

The first qualification is this. I gave an impression of a steady upward climb in the prey-catching abilities of cheetahs, and the predator-avoiding abilities of gazelles. The reader might have come away with a Victorian idea of the inexorability of progress, each generation better, finer and braver than its parents. The reality in nature is nothing like that. The timescale over which significant improvement might be detected is, in any case, likely to be far longer than could be detected by comparing one typical generation with its predecessor. The ‘improvement’, moreover, is far from continuous. It is a fitful affair, stagnating or even sometimes going ‘backwards’, rather than moving solidly ‘forwards’ in the direction suggested by the arms-race idea. Changes in conditions, changes in the inanimate forces I have lumped under the general heading of ‘the weather’, are likely to swamp the slow and erratic trends of the arms race, as far as any observer on the ground could be aware. There may well be long stretches of time in which no ‘progress’ in the arms race, and perhaps no evolutionary change at all, takes place. Arms races sometimes culminate in extinction, and then a new arms race may begin back at square one. Nevertheless, when all this is said, the arms-race idea remains by far the most satisfactory explanation for the existence of the advanced and complex machinery that animals and plants possess. Progressive ‘improvement’ of the kind suggested by the arms-race image does go on, even if it goes on spasmodically and interruptedly; even if its net rate of progress is too slow to be detected within the lifetime of a man, or even within the timespan of recorded history.

The second qualification is that the relationship that I am calling  {182}  ‘enemy’ is more complicated than the simple bilateral relationship suggested by the stories of cheetahs and gazelles. One complication is that a given species may have two (or more) enemies which are even more severe enemies of each other. This is the principle behind the commonly expressed half-truth that grass benefits by being grazed (or mown). Cattle eat grass, and might therefore be thought of as enemies of grass. But grasses also have other enemies in the plant world, competitive weeds, which, if allowed to grow unchecked, might turn out to be even more severe enemies of grasses than cattle. Grasses suffer somewhat from being eaten by cattle, but the competitive weeds suffer even more. Therefore the net effect of cattle on a meadow is that the grasses benefit. The cattle turn out to be, in this sense, friends of grasses rather than enemies.

Nevertheless, cattle are enemies of grass in that it is still true that an individual grass plant would be better off not being eaten by a cow than being eaten, and any mutant plant that possessed, say, a chemical weapon that protected it against cows, would set more seed (containing genetic instructions for making the chemical weapon) than rival members of its own species that were more palatable to cows. Even if there is a special sense in which cows are ‘friends’ of grasses, natural selection does not favour individual grass plants that go out of their way to be eaten by cows! The general conclusion to this paragraph is as follows. It may be convenient to think of an arms race between two lineages such as cattle and grass, or gazelles and cheetahs, but we should never lose sight of the fact that both participants have other enemies against whom they are simultaneously running other arms races. I shall not pursue the point here, but it can be developed into one of the explanations for why particular arms races stabilize and do not go on for ever — do not lead to predators pursuing their prey at Mach 2 and so on.

The third ‘qualification’ to the simple arms-race is not so much a qualification as an interesting point in its own right. In my hypothetical discussion of cheetahs and gazelles I said that cheetahs, unlike the weather, had a tendency as the generations go by to become ‘better hunters’, to become more severe enemies, better equipped to kill gazelles. But this does not imply that they become more successful at killing gazelles. The kernel of the arms-race idea is that both sides in the arms race are improving from their own point of view, while simultaneously making life more difficult for the other side in the arms race. There is no particular reason (or at least none in anything that we have discussed so far) to expect either side in the arms race to become steadily more successful or less successful than the other. In  {183}  fact the arms-race idea, in its purest form, suggests that there should be absolutely zero progress in the success rate on both sides of the arms race, while there is very definite progress in the equipment for success on both sides. Predators become better equipped for killing, but at the same time prey become better equipped to avoid being killed, so the net result is no change in the rate of successful killings.

The implication is that if, by the medium of a time machine, predators from one era could meet prey from another era, the later, more ‘modern’ animals, whether predators or prey, would run rings round the earlier ones. This is not an experiment that can ever be done, although some people assume that certain remote and isolated faunas, such as those of Australia and Madagascar, can be treated as if they were ancient, as if a trip to Australia were like a trip backwards in a time machine. Such people think that native Australian species are usually driven extinct by superior competitors or enemies introduced from the outside world, because the native species are ‘older’, ‘out of date’ models, in the same position vis-a-vis invading species as a Jutland battleship contending with a nuclear submarine. But the assumption that Australia has a ‘living fossil’ fauna is hard to justify. Perhaps a good case for it might be made, but it seldom is. I'm afraid it may be no more than the zoological equivalent of chauvinistic snobbery, analogous to the attitude that sees every Australian as an uncouth swagman with not much under his hat and corks dangling round the brim.

The principle of zero change in success rate, no matter how great the evolutionary progress in equipment, has been given the memorable name of the ‘Red Queen effect’ by the American biologist Leigh van Valen. In Through the Looking Glass, you will remember, the Red Queen seized Alice by the hand and dragged her, faster and faster, on a frenzied run through the countryside, but no matter how fast they ran they always stayed in the same place. Alice was understandably puzzled, saying, ‘Well in our country you'd generally get to somewhere else — if you ran very fast for a long time as we've been doing.’ ‘A slow sort of country!’ said the Queen. ‘Now, here, you see, it takes all the running you can do, to keep in the same place. If you want to get somewhere else, you must run at least twice as fast as that!’

The Red Queen label is amusing, but it can be misleading if taken (as it sometimes is) to mean something mathematically precise, literally zero relative progress. Another misleading feature is that in the Alice story the Red Queen's statement is genuinely paradoxical, irreconcilable with common sense in the real physical world. But van Valen's evolutionary Red Queen effect is not paradoxical at all. It is  {184}  entirely in accordance with common sense, so long as common sense is intelligently applied. If not paradoxical, however, arms races can give rise to situations that strike the economically minded human as wasteful.

Why, for instance, are trees in forests so tall? The short answer is that all the other trees are tall, so no one tree can afford not to be. It would be overshadowed if it did. This is essentially the truth, but it offends the economically minded human. It seems so pointless, so wasteful. When all the trees are the full height of the canopy, all are approximately equally exposed to the sun, and none could afford to be any shorter. But if only they were all shorter; if only there could be some sort of trade-union agreement to lower the recognized height of the canopy in forests, all the trees would benefit. They would be competing with each other in the canopy for exactly the same sunlight, but they would all have ‘paid’ much smaller growing costs to get into the canopy. The total economy of the forest would benefit, and so would every individual tree. Unfortunately, natural selection doesn't care about total economies, and it has no room for cartels and agreements. There has been an arms race in which forest trees became larger as the generations went by. At every stage of the arms race there was no intrinsic benefit in being tall for its own sake. At every stage of the arms race the only point in being tall was to be relatively taller than neighbouring trees.

As the arms race wore on, the average height of trees in the forest canopy went up. But the benefit that the trees obtained from being tall did not go up. It actually deteriorated because of the enhanced costs of growing. Successive generations of trees got taller and taller, but at the end they might better, in one sense, have stayed where they started. Here, then, is the connection with Alice and the Red Queen, but you can see that in the case of the trees it is not really paradoxical. It is generally characteristic of arms races, including human ones, that although all would be better off if none of them escalated, so long as one of them escalates none can afford not to. Once again, by the way, I should stress that I have told the story too simply. I do not mean to suggest that in every literal generation trees are taller than their counterparts in the previous generation, nor that the arms race is necessarily still going on.

Another point illustrated by the trees is that arms races do not necessarily have to be between members of different species. Individual trees are just as likely to be harmfully overshadowed by members of their own species as by members of other species. Probably more so in fact, for all organisms are more seriously threatened by  {185}  competition from their own species than from others. Members of one's own species are competitors for the same resources, to a much more detailed extent, than members of other species. There are also arms races within species between male roles and female roles, and between parent roles and offspring roles. I have discussed these in The Selfish Gene, and will not pursue them further here.

The tree story allows me to introduce an important general distinction between two kinds of arms race, called symmetric and asymmetric arms races. A symmetric arms race is one between competitors trying to do roughly the same thing as each other. The arms race between forest trees struggling to reach the light is an example. The different species of trees are not all making their livings in exactly the same way, but as far as the particular race we are talking about is concerned — the race for the sunlight above the canopy — they are competitors for the same resource. They are taking part in an arms race in which success on one side is felt by the other side as failure. And it is a symmetric arms race because the nature of the success and failure on the two sides is the same: attainment of sunlight and being overshadowed, respectively.

The arms race between cheetahs and gazelles, however, is asymmetric. It is a true arms race in which success on either side is felt as failure by the other side, but the nature of the success and failure on the two sides is very different. The two sides are ‘trying’ to do very different things. Cheetahs are trying to eat gazelles. Gazelles are not trying to eat cheetahs, they are trying to avoid being eaten by cheetahs. From an evolutionary point of view asymmetric arms races are more interesting, since they are more likely to generate highly complex weapons systems. We can see why this is by taking examples from human weapons technology.

I could use the USA and the USSR as examples, but there is really no need to mention specific nations. Weapons manufactured by companies in any of the advanced industrial countries may end up being bought by any of a wide variety of nations. The existence of a successful offensive weapon, such as the Exocet type of surface skimming missile, tends to ‘invite’ the invention of an effective counter, for instance a radio jamming device to ‘confuse’ the control system of the missile. The counter is more likely than not to be manufactured by an enemy country, but it could be manufactured by the same country, even by the same company! No company, after all, is better equipped to design a jamming device for a particular missile than the company that made the missile in the first place. There is nothing inherently improbable about the same company producing both and selling them  {186}  to opposite sides in a war. I am cynical enough to suspect that it probably happens, and it vividly illustrates the point about equipment improving while its net effectiveness stands still (and its costs increase).

From my present point of view the question of whether the manufacturers on opposite sides of a human arms race are enemies of each other or identical with each other is irrelevant, and interestingly so. What matters is that, regardless of their manufacturers, the devices themselves are enemies of each other in the special sense I have defined in this chapter. The missile, and its specific jamming device, are enemies of each other in that success in one is synonymous with failure in the other. Whether their designers are also enemies of each other is irrelevant, although it will probably be easier to assume that they are.

So far I have discussed the example of the missile and its specific antidote without stressing the evolutionary, progressive aspect, which is, after all, the main reason for bringing it into this chapter. The point here is that not only does the present design of a missile invite, or call forth, a suitable antidote, say a radio jamming device. The antimissile device, in its turn, invites an improvement in the design of the missile, an improvement that specifically counters the antidote, an anti-anti-missile device. It is almost as though each improvement in the missile stimulates the next improvement in itself, via its effect on the antidote. Improvement in equipment feeds on itself. This is a recipe for explosive, runaway evolution.

At the end of some years of this ding-dong invention and counter-invention, the current version of both the missile and its antidote will have attained a very high degree of sophistication. Yet at the same time — here is the Red Queen effect again — there is no general reason for expecting either side in the arms race to be any more successful at doing its job than it was at the beginning of the arms race. Indeed if both the missile and its antidote have been improving at the same rate, we can expect that the latest, most advanced and sophisticated versions, and the earliest, most primitive and simplest versions will be exactly as successful as each other, against their contemporary counter-devices. There has been progress in design, but no progress in accomplishment, specifically because there has been equal progress in design on both sides of the arms race. Indeed, it is precisely because there has been approximately equal progress on both sides that there has been so much progress in the level of sophistication of design. If one side, say the antimissile jamming device, pulled too far ahead in the design race, the other side, the missile in this case, would simply  {187}  cease to be used and manufactured: it would go ‘extinct’. Far from being paradoxical like Alice's original example, the Red Queen effect in its arms-race context turns out to be fundamental to the very idea of progressive advancement.

I said that asymmetric arms races were more likely to lead to interesting progressive improvements than symmetric ones, and we can now see why this is, using human weapons to illustrate the point. If one nation has a 2-megaton bomb, the enemy nation will develop a 5-megaton bomb. This provokes the first nation into developing a 10-megaton bomb, which in turn provokes the second into making a 20-megaton bomb, and so on. This is a true progressive arms race: each advance on one side provokes the counter-advance on the other, and the result is a steady increase in some attribute as time goes by — in this case, explosive power of bombs. But there is no detailed, one-to-one correspondence between the designs in such a symmetric arms race, no ‘meshing’ or ‘interlocking’ of design details as there is in an asymmetric arms race such as that between missile and missile-jamming device. The missile-jamming device is designed specificially to overcome particular detailed features of the missile; the designer of the antidote takes into account minute details of the design of the missile. Then in designing a counter to the antidote, the designer of the next generation of missiles makes use of his knowledge of the detailed design of the antidote to the previous generation. This is not true of the bombs of ever-increasing megatonnage. To be sure, designers on one side may pirate good ideas, may imitate design features, from the other side. But if so, this is incidental. It is not a necessary part of the design of a Russian bomb that it should have detailed, one-to-one correspondences with specific details of an American bomb. In the case of an asymmetric arms race, between a lineage of weapons and the specific antidotes to those weapons, it is the one-to-one correspondences that, over the successive ‘generations’, lead to ever greater sophistication and complexity.

In the living world too, we shall expect to find complex and sophisticated design wherever we are dealing with the end-products of a long, asymmetric arms race in which advances on one side have always been matched, on a one-to-one, point-for-point basis, by equally successful antidotes (as opposed to competitors) on the other. This is conspicuously true of the arms races between predators and their prey, and, perhaps even more, of arms races between parasites and hosts. The electronic and acoustic weapons systems of bats, which we discussed in Chapter 2, have all the finely tuned sophistication that we expect from the end-products of a long arms race. Not surprisingly, we  {188}  can trace this same arms race on the other side. The insects that bats prey upon have a comparable battery of sophisticated electronic and acoustic gear. Some moths even emit bat-like (ultra-) sounds that seem to put the bats off. Almost all animals are either in danger of being eaten by other animals or in danger of failing to eat other animals, and an enormous number of detailed facts about animals makes sense only when we remember that they are the end-products of long and bitter arms races. H. B. Cott, author of the classic book Animal Coloration, put the point well in 1940, in what may be the first use in print of the arms-race analogy in biology:

Before asserting that the deceptive appearance of a grasshopper or butterfly is unnecessarily detailed, we must first ascertain what are the powers of perception and discrimination of the insects’ natural enemies. Not to do so is like asserting that the armour of a battle-cruiser is too heavy, or the range of her guns too great, without inquiring into the nature and effectiveness of the enemy's armament. The fact is that in the primeval struggle of the jungle, as in the refinements of civilized warfare, we see in progress a great evolutionary armament race — whose results, for defence, are manifested in such devices as speed, alertness, armour, spinescence, burrowing habits, nocturnal habits, poisonous secretions, nauseous taste, and (camouflage and other kinds of protective coloration), and for offence, in such counter-attributes as speed, surprise, ambush, allurement, visual acuity, claws, teeth, stings, poison fangs, and [lures]. Just as greater speed in the pursued has developed in relation to increased speed in the pursuer; or defensive armour in relation to aggressive weapons; so the perfection of concealing devices has evolved in response to increased powers of perception.

Arms races in human technology are easier to study than their biological equivalents because they are so much faster. We can actually see them going on, from year to year. In the case of a biological arms race, on the other hand, we can usually see only the end-products. Very rarely a dead animal or plant fossilizes, and it is then sometimes possible to see progressive stages in an animal arms race a little more directly. One of the most interesting examples of this concerns the electronic arms race, as shown in the brain sizes of fossil animals.

Brains themselves do not fossilize but skulls do, and the cavity in which the brain was housed — the braincase — if interpreted with care, can give a good indication of brain size. I said ‘if interpreted with care’, and the qualification is an important one. Among the many problems is the following. Big animals tend to have big brains partly just because they are big, but this doesn't necessarily mean that they are, in any interesting sense, ‘cleverer’. Elephants have bigger brains than humans  {189}  but, probably with some justice, we like to think that we are cleverer than elephants and that our brains are ‘really’ bigger if you make allowance for the fact that we are much smaller animals. Certainly our brains occupy a much larger proportion of our body than elephants’ brains do, as is evident from the bulging shape of our skulls. This is not just species vanity. Presumably a substantial fraction of any brain is needed to perform routine caretaking operations around the body, and a big body automatically needs a big brain for this. We must find some way of ‘taking out’ of our calculations that fraction of brain that can be attributed simply to body size, so that we can compare what is left over as the true ‘braininess’ of animals. This is another way of saying that we need some good way of defining exactly what we mean by true braininess. Different people are at liberty to come up with different methods of doing the calculations, but probably the most authoritative index is the ‘encephalization quotient’ or EQ used by Harry Jerison, a leading American authority on brain history.

The EQ is actually calculated in a somewhat complicated way, taking logarithms of brain weight and body weight, and standardizing against the average figures for a major group such as the mammals as a whole. Just as the ‘intelligence quotient’ or IQ used (or it may be misused) by human psychologists is standardized against the average for a whole population, the EQ is standardized against, say, the whole of the mammals. Just as an IQ of 100 means, by definition, an IQ identical to the average for a whole population, so an EQ of 1 means, by definition, an EQ identical to the average for, say, mammals of that size. The details of the mathematical technique don’t matter. In words, the EQ of a given species such as a rhino or a cat, is a measure of how much bigger (or smaller) the animal's brain is than we should expect it to be, given the animal's body size. How that expectation is calculated is certainly open to debate and criticism. The fact that humans have an EQ of 7 and hippos an EQ of 0.3 may not literally mean that humans are 23 times as clever as hippos! But the EQ as measured is probably telling us something about how much ‘computing power’ an animal has in its head, over and above the irreducible minimum of computing power needed for the routine running of its large or small body.

Measured EQs among modern mammals are very varied. Rats have an EQ of about 0.8, slightly below the average for all mammals. Squirrels are somewhat higher, about 1.5. Perhaps the three-dimensional world of trees demands extra computing power for controlling precision leaps, and even more for thinking about efficient paths through a maze of branches that may or may not connect farther on. Monkeys are well above average, and apes (especially ourselves) even higher. Within the monkeys it turns out that some types have higher EQs than others and that, interestingly, there is some connection with how they make their living: insect-eating and fruit-eating monkeys have bigger brains, for their size, than leaf-eating monkeys. It makes some sense to argue that an animal needs less computing power to find leaves, which are abundant all around, than to find fruit, which may have to be searched for, or to catch insects, which take active steps to get away. Unfortunately, it is now looking as though the true story is more complicated, and that other variables, such as metabolic rate, may be more important. In the mammals as a whole, carnivores typically have a slightly higher EQ than the herbivores upon which they prey. The reader will probably have some ideas about why this might be, but it is hard to test such ideas. Anyway, whatever the reason, it seems to be a fact.

So much for modern animals. What Jerison has done is to reconstruct the probable EQs of extinct animals that now exist only as fossils. He has to estimate brain size by making plaster casts of the insides of braincases. Quite a lot of guesswork and estimation has to go into this, but the margins of error are not so great as to nullify the whole enterprise. The methods of taking plaster casts can, after all, be checked for their accuracy, using modern animals. We make-believe that the dried skull is all that we have from a modern animal, use a plaster cast to estimate how big its brain was from the skull alone, and then check with the real brain to see how accurate our estimate was. These checks on modern skulls encourage confidence in Jerison's estimates of long-dead brains. His conclusion is, firstly, that there is a tendency for brains to become bigger as the millions of years go by. At any given time, the current herbivores tended to have smaller brains than the contemporary carnivores that preyed on them. But later herbivores tended to have larger brains than earlier herbivores, and later carnivores larger brains than earlier carnivores. We seem to be seeing, in the fossils, an arms race, or rather a series of restarting arms races, between carnivores and herbivores. This is a particularly pleasing parallel with human armament races, since the brain is the on-board computer used by both carnivores and herbivores, and electronics is probably the most rapidly advancing element in human weapons technology today.

How do arms races end? Sometimes they may end with one side going extinct, in which case the other side presumably stops evolving in that particular progressive direction, and indeed it will probably even ‘regress’ for economic reasons soon to be discussed. In other cases, economic pressures may impose a stable halt to an arms race,  {191}  stable even though one side in the race is, in a sense, permanently ahead. Take running speed, for instance. There must be an ultimate limit to the speed at which a cheetah or a gazelle can run, a limit imposed by the laws of physics. But neither cheetahs nor gazelles have reached that limit. Both have pushed up against a lower limit which is, I believe, economic in character. High-speed technology is not cheap. It demands long leg bones, powerful muscles, capacious lungs. These things can be had by any animal that really needs to run fast, but they must be bought. They are bought at a steeply increasing price. The price is measured as what economists call ‘o