Life at the Edge of Chaos?

1.

Darwinism Evolving is a history of ideas about biological diversity and evolution, from Aristotle to the present day. The last part of the book is an account of some recent developments, and an attempt to forecast the future. Most of this review will be concerned with the final section, which seems to me mistaken. I must therefore start by saying that I found the historical part well informed, and full of valuable insights. The ideas discussed are fundamental, not only to biology, but also to our view of our relationship to the rest of the natural world. The last twenty years have seen an explosion of scholarship centered on Darwin, by historians and philosophers. The book is an admirable summary of, and addition to, that scholarship.

The historical thesis can be summarized by saying that Darwinism has a central core, the idea of natural selection, but that it has undergone three stages of evolution, according to the dynamic theories used to formulate it: first, the deterministic dynamics of Newton, then the probabilistic dynamics of Clerk Maxwell and Ludwig Boltzmann, and finally the dynamics of complex systems now being developed. Darwin himself, the authors suggest, formulated his theory in Newtonian terms. When I first met this claim, I was doubtful. Of course Darwin, like many of his contemporaries, wished to be seen as following the Newtonian method: Who would not claim to be following such a successful example. But how can one have “dynamics” without any mathematical equations? After reading their account, however, I was persuaded that the claim is reasonable, provided that one interprets dynamics in what Darwin himself might have called a loose and metaphorical sense.

The essential difference between Aristotle and Newton is that Aristotle thought that bodies move as they do because it is natural for them to do so, whereas Newton explained the elliptical orbits of planets as caused by an external force, gravity. A similar contrast exists between Lamarck and Darwin. Lamarck held that organisms evolve because they have an inherent tendency to become more complex. It was this idea that Darwin was rejecting when he said his theory had nothing in common with Lamarck’s. Instead of explaining evolution by an inherent tendency, Darwin thought that change was directed by an external force, natural selection.

Thus I think the authors make a good case for the claim that Newton is to Aristotle as Darwin is to Lamarck. Later in the nineteenth century, Newton’s deterministic dynamics was replaced in some fields of physics by the “stochastic” (i.e., probabilistic) dynamics of Maxwell and Boltzmann. They showed that the behavior of large aggregates of things (initially gases, which are aggregates of molecules) could be predicted by a dynamics which ignored the precise behavior of individuals, and took into account only the average behavior. Boltzmann once wrote that the nineteenth century should be seen as the century of Darwin, because Darwin explained evolution according to the chances of death or survival of millions of individuals. Boltzmann was giving Darwin the credit for inventing stochastic dynamics. Depew and Weber think that he was being overgenerous: the century, in their view, should be seen as Boltzmann’s. I cannot decide whether they are right. However that may be, Darwinism today is fully stochastic, but it did not become so until the 1920s with the invention of population genetics by R.A. Fisher, J.B.S. Haldane, and Sewall Wright.

It is a curious twist of history that when Mendel’s laws were rediscovered in 1901, Mendelism was at first seen as an alternative to Darwinism in explaining evolution. Darwinism held that change occurred by the natural selection of many minute variations, Mendelism that novelty arose suddenly, by mutation. The population geneticists showed that the two theories were complementary, not contradictory. They did so by developing a theory of the changes in the frequencies of genes, which is indeed analogous to the Maxwell-Boltzmann dynamics.

The authors rightly point out that the history of ideas about evolution is not simply a series of modulations on Darwin’s idea of natural selection. There was an alternative account, which saw the evolution of species and individual development as parallel processes, causally connected in ways that I find difficult to understand. Ernst Haeckel’s theory of recapitulation, according to which each individual during its development recapitulates the forms of its ancestors, is an example of such a theory. Although Haeckel’s ideas are hard to follow, it is clear that if a developmental change in one generation affects the nature of the next, then theories of evolution and development cannot be separated. Twentieth-century Darwinism was liberated from this difficulty by August Weismann. If, as he claimed, acquired characters are not inherited, then the processes by which organisms develop can be treated as if they were in a black box. The box will be opened one day, but in the meanwhile we can get on with genetics and evolution theory. I will return to the relation between development and evolution.

Although I found the historical account in Darwinism Evolving illuminating, one concept which has become dominant in the twentieth century is hardly mentioned, either here or in most recent discussions of evolution by philosophers and historians. This is the concept of information. Its dominance in molecular biology is obvious from the vocabulary of the subject. Messenger, code, synonymous, transcription, translation, proofreading, editing are all technical terms: they are not metaphors but words whose precise meaning is well understood. Although the concept of information has become dominant only in the last half of this century, it is already present in August Weismann’s work at the end of the last. His great insight was to see that heredity is about the transmission, not of matter or energy, but of information. Since I will certainly be accused of an anachronism in saying this, let me quote his remark, discussing the inheritance of an acquired character, that it “is very like supposing that an English telegram to China is there received in the Chinese language.”

When we turn to the authors’ account of the present state of evolutionary biology, and its likely future development, I think they are mistaken; but their comments are well informed and thoughtful: if they are mistaken, it is not from ignorance or stupidity. And it must also be said that spotting what is important in current science, and predicting future developments, are matters of judgment; one cannot know that one is right.

The pieces of research that the authors pick out to discuss have one thing in common: they either deny natural selection or, more often, diminish its importance. I therefore start by explaining why I see natural selection as the necessary component of any satisfactory theory of evolution. Like better men before me—for example, Aristotle, Darwin, Alfred Russel Wallace, August Weismann—I start from the fact of adaptation. The most striking fact about living organisms is that their parts are adapted to ensure the survival and reproduction of the whole organism. Not only are structures like hearts, wings, eyes, and kidneys explicable only by their functions: so also are different behaviors, such as the migration of birds or the dances of bees. Adaptation is not only obvious on a large scale: the molecular structures that ensure the translation of DNA into protein are also to be understood only in functional terms. Of course not all features of organisms are adapted: no one who has suffered from myopia, appendicitis, cancer, or arthritis could think that. But myopia does not call for an explanation: it is to be expected in a universe in which entropy increases. What does need an explanation is how an eye can form an image and how a brain can interpret it.

Adaptation, then, is the phenomenon we have to explain. Natural selection is the only theory so far proposed that can do so. Lamarckism might seem to be a possible alternative. Thus if an animal acquires an adaptive characteristic during its lifetime, and passes it on to its offspring, that could lead to the evolution of adaptation. The snag is that most “acquired characters” are not adaptations: they are the results of injury, disease, and aging. For Lamarckism to work, there would have to be some process of selection whereby only those acquired characters that were adaptive were passed on. If we learn many ideas, but teach our children only the ones that worked, then our children would be better off than we are. That is what people mean when they speak of cultural inheritance being Lamarckian. But for genetically transmitted traits, there seems to be no way this can work. If we are to explain adaptation, we must choose between natural selection and creation.

Natural selection, in turn, requires that there be entities that multiply, and that have heredity. Like must beget like. In fact, more than heredity is needed. There must be entities that can exist in an indefinitely large number of forms, each of which can be accurately replicated. This is why the concept of information is central both to genetics and evolution theory.

With this as background, I turn to the particular topics that the authors pick out as significant for the future. I have space to discuss only three: the neutral theory of molecular evolution, genetic revolutions, and the dynamics of complex systems.

The neutral theory, first proposed by the Japanese geneticist Motoo Kimura, argues that most of the changes that occur in proteins and nucleic acids in the course of evolution do so because they are selectively “neutral”: they do not make their possessor either more or less likely to survive. The first thing to understand about this theory is that it does not deny either the phenomenon of adaptation or Darwin’s explanation of it. In Kimura’s own words, “the theory does not deny the role of natural selection in determining the course of adaptive evolution.” What it claims is that in addition to genetic changes caused by selection there are much more frequent changes that occur because they do not matter. I have never seen any reason why, as a naive Darwinist, I should reject this theory. It is mathematically elegant: I have even done a little of the mathematics myself, and would hate to see it wasted. For me, the only question is the empirical one: Is the theory actually true? At least as far as nucleic acids are concerned, I think that it is near enough true to be interesting and important. I was delighted when, perhaps somewhat ironically, Kimura was given the Darwin Medal of the Royal Society. But his ideas have not altered the way I think about the evolution of morphology or behavior.

The second issue is that of genetic revolutions. It is now known that changes in the genetic information—DNA—can occur not only by “point mutations” that alter a single base at a time, but by a variety of processes in which pieces of DNA are duplicated or moved to new sites in the chromosome. Perhaps inevitably, this has led to the idea that such genomic shuffling can, without the need for selection, give rise to evolutionary novelty, and in particular to new species. This is nonsense. It would take too long to clear up all the confusions involved, but I will address three points: the non-adaptive nature of mutation, the significance of gene duplication, and “transposition.”

The founders of population genetics, Fisher, Haldane, and Wright, knew nothing of DNA or of the nature of mutation. What they assumed was that mutations that reduced fitness would be much commoner than those that increased it. If this were not the case, then natural selection would not be needed to explain an increase in fitness: the increase would happen automatically, because of mutation. They knew nothing of the nature of genes, but they assumed, reasonably enough, that they must be complicated, and that random changes in their structure would be far more likely to do harm than good. There is nothing in molecular biology to make us alter that judgment. Changes in DNA, whether by point mutation or by rearrangement, are more likely to lower fitness than to increase it. If evolution leads to adaptation, it is because natural selection picks out the small proportion of favorable mutations.

Genes are sometimes duplicated, and such duplications are occasionally incorporated in evolution. To take the example discussed in Darwinism Evolving, the hemoglobin that transports oxygen in our blood is coded for by several genes: two cooperate to produce adult hemoglobin, and a third is active in the fetus. These genes arose by the duplication of a single ancestral gene. This is fascinating, but in what way does it remove the need for selection? The original duplication occurred in a single individual. To become established in all members of a species, selection was needed. The genes are subtly different, and their functional success depends on these differences. The differences arose by mutations—again in individuals—and again had to spread by selection. Merely to duplicate a gene gives you two copies of the same information. That does not help. Without selection, duplicated copies of genes deteriorate until they make no sense.

Duplication is important, but it does not obviate the need for selection. But what if a gene could spread through the population by repeated acts of duplication, combined with jumping to new places in the chromosomes? Such a gene could spread, even if the individuals that carried it were not favored by selection. In the author’s words, “variation can pass to fixation without passing through Mendel’s laws.” Such genes do exist: they are called “transposons,” or, more colloquially, jumping genes. Such a gene, the “p” element, has in fact spread in recent years through wild populations of the geneticist’s favorite organism, the fruit fly Drosophila melanogaster. But no one thinks it has contributed to the fitness of the flies. It made them very sick while it was spreading but now seems to have settled down to being relatively harmless. It is best thought of as a piece of parasitic DNA that has characteristics favoring its own spread at the expense of its host. “Selfish DNA” of this kind is precisely what a Darwinist would expect, but not as an explanation of the adaptation of organisms.

To avoid misunderstanding, I must add that there are two processes that can add significant novelty in a single step. The first is the origin of a new species by hybridization between existing species, a process that has been common in plants. More important is the origin of novelty by symbiosis: two independently evolved organisms come to live together in intimate union. Symbiosis was involved, for example, in the evolution of nitrogen-fixing plants, of the hydroids that build coral reefs, of the animals that live in deep-sea vents, and, most important of all, of the cells ancestral to animals, plants, and fungi. I am surprised that the authors do not discuss symbiosis, because it has been presented, for reasons that escape me, as a refutation of Darwinism. The motorbike is a great invention, but it requires that someone first invent the bicycle and the internal combustion engine.

2.

I turn now to the issue which, for the authors, is central. They predict that, in the future, the theory of evolution will be expressed in terms of a new dynamics: no longer the dynamics of Newton or of Boltzmann, but of complex systems theory. First, I must say something about this new dynamics. The advent of computers has made it possible to study the behavior of systems too complex to study analytically. Exciting, and at least to me unexpected, types of behavior have been found by computer simulation, some of which have rather charming names, such as “strange attractors,” “self-organized criticality,” “life at the edge of chaos.” Since life is certainly complex, perhaps this new dynamics is the way to describe it.

I certainly think that some familiarity with complex systems dynamics should be part of the mental furniture of scientists in the future, just as a familiarity with simple harmonic motion, or limit cycles, is part of my mental furniture. It is also true that some of the models we make of biological systems are too complicated to treat analytically, so that we use computer simulation to study them. Sometimes their behavior turns out to be, in a mathematical sense, chaotic. But the authors are making a much larger claim for the new dynamics than this. It is that complex systems give rise to structures which either can explain evolution without natural selection or at the very least can explain how the structures between which natural selection can choose come to exist in the first place. Stuart Kauffman, for example, has argued that complex dynamics will give structure “for free,” instead of each detail of a structure having to be independently evolved. Still, he would not deny the role of selection in stabilizing those structures that are adaptive.

I discuss below a particular example of a dynamic system—Turing’s morphogenetic waves—which gives rise to just the kind of structure that, as a biologist, I want to see. But first I must explain why I have a general feeling of unease when contemplating complex systems dynamics. Its devotees are practicing fact-free science. A fact for them is, at best, the output of a computer simulation: it is rarely a fact about the world.

General ideas in evolutionary biology have usually emerged from the study of particular biological observations. Darwin has told us that the origin of all his ideas lay in the peculiarities of the animals of the Galapagos islands and of the fossils of South America. Alfred Russel Wallace also became an evolutionist through exposure to the animals of the tropics. Many other evolutionists—for example August Weismann, Ernst Mayr, Theodosius Dobzhansky—started out as naturalists.

Even if an idea does not have its origins in particular observations, it is certainly expected to explain something. I do not know what led W.D. Hamilton to explain animal societies by examining their kinship relations. I do know that the moment when I thought he might be on to something was when he pointed out that in bees, ants, and wasps, females are more closely related to their sisters than to their daughters. My own main contribution has been the introduction of evolutionary game theory. I was led to the idea when trying to explain ritualized behavior in animal fights. I did not take the idea seriously until I found that it can be used to explain peculiar sex ratios (as Fisher and Hamilton knew before me), plant growth, animal migration, male mating behavior, and even the evolution of viruses. The facts, of course, need not be about natural history. It is hard to think of a theory more divorced from natural history than Kimura’s neutral theory of molecular evolution, but Kimura has been almost obsessively concerned to show that his theory will explain the molecular evidence.

My difficulty, then, is that I do not know what observations complex systems dynamics is trying to explain. It is a theory looking for a question to answer.

The question that most urgently needs an answer concerns the mechanism of development: How does an egg turn into an adult? At first sight, this is a topic to which modern dynamics seems to have much to contribute. It is therefore a fair test of the usefulness of notions of self-organization. It turns out, however, that so far a thoroughly reductionist approach, based not on systems theory but on molecular genetics and the concept of information, is proving more effective. Until recently, the position of evolutionary biologists could be caricatured as follows. We can treat development as a black box because, if acquired characters are not inherited, novelty must start with mutation. To be sure, mutations, to be selected, must influence development, but how they do so hardly matters. The ways by which organisms develop in one way and not in others enter into evolutionary biology only in the form of “developmental constraints”: not every novelty is possible. But we do not know enough about development to say what the constraints are. All we can really say is that evolution is constrained by the laws of physics and chemistry: if a species would benefit by being able to fly, it must evolve wings—levitation is not an option.

In effect, this has been the position taken by most evolutionary biologists for fifty years. Occasionally someone has said that our science will be transformed when it incorporates development: I have even said it myself. The snag has been that we have not had a theory of development to incorporate. I think that situation is changing.

The problem of development is as follows. How do cells of many different kinds—muscle cells, brain cells, epithelial cells, and so on—come to be arranged in a particular way in space? Weismann thought that different genes (he called them ids) passed to different organs—leg genes to the developing leg, brain genes to the brain, and so on. We now know that virtually all genes pass to all cells. But this is no problem. Muscle cells are different from neurons not because they contain different genes but because different genes are switched on. Thanks to the work of Jacques Monod and François Jacob, and a vast body of research since, we have a good idea of how this switching occurs. Although there is much still to discover, cell differentiation is not, I think, a problem requiring a new dynamics.

The spatial pattern chosen remains a problem: How are the right genes switched on in the right places? There are many possibilities, but I will mention only two. An old idea is that of a “gradient,” which can best be explained by a recently discovered example. In the fruit fly, the first difference between the front and back ends of the egg is caused by cells in the mother’s ovary, external to the egg, that release at the anterior end a specific chemical, which then diffuses backward, giving rise to a gradient of concentration, high at one end and low at the other. This in turn causes different genes to be switched on in different places. Of course, a single gradient cannot set up the whole pattern, but a succession of such processes might do so.

An alternative mechanism was proposed by the English mathematician Alan Turing in 1952. This was a rare example of an idea emerging from purely dynamical considerations which seemed to offer biologists just what they needed. Turing showed that if one starts with a uniformly distributed set of chemicals, able to react and diffuse, a possible result is the development of a standing wave, with regularly arranged peaks and troughs of concentration. This could explain the development of regularly repeated structures—the segments of a millipede, the petals of a flower, the rays of a starfish. I was fascinated by Turing’s model, and wrote several papers using Turing’s ideas to explain various quantitative features of repeated patterns.

It was difficult to take the idea much further without knowing what the chemicals were. Recently things have changed. Using the techniques of molecular genetics, many genes active early in development have been identified, and pictures have been produced showing where each of them is first active in the embryo. A recent picture showed the egg of Drosophila with seven regular stripes indicating the activity of one such gene: these were interspersed with other stripes where a second gene is active. These fourteen stripes are the basis of the segments of the adult insect.

When I saw these pictures, I was delighted. It was hard not to see them as a Turing wave. (I was particularly pleased because I had predicted the number seven, but that is another story.) Sadly, it seems that they are nothing of the kind. As explained in an elegantly titled paper in Nature, “Making stripes inelegantly,” each stripe is induced in a different way: if Turing was right, each stripe would be induced by the same chemical. I have not quite given up hope. It may be that the data can be differently interpreted. Perhaps the petals of a flower will prove to be induced by a Turing wave. I refuse to believe that the segments of a millipede are induced inelegantly.

All the same, the story has a moral for the authors of this book. Turing was, in Kaufmann’s words, offering us something for free—a way of getting a pattern without having to specify each part of it separately. This is precisely what, it is claimed, complex dynamics will do for us. The moral, of course, is that whether development works like that is, ultimately, an empirical question.

At last we are getting an account of development sufficiently precise to illuminate evolution. The account is emerging from the most reductionist kind of molecular genetics. There is nothing holist about it, and I doubt whether its practitioners have heard about complex systems dynamics. But it is telling us something about evolution. For me, the most remarkable fact about biology to have emerged in recent years is this. In the early embryos of many different kinds of animals—vertebrates, insects, various kinds of worm—the same spatial pattern of activity of the same eight or nine genes is visible, before a particular morphological structure is apparent. It seems that a common pattern of positional information evolved before the evolution of the different body plans that characterize the major groups of animals.

Depew and Weber have written a thoughtful and often illuminating book. I found the historical sections in the main convincing. But the authors’ account of contemporary evolutionary biology omits most of the things I find exciting, and their guess about what the future will hold is very different from mine.