When, as often happens, I find myself dissenting from something written by Stephen Jay Gould, I remind myself that we share a common childhood experience. We were both dinosaur nuts, at a time when to be interested in dinosaurs was to be an oddball. For both of us, that early passion has led to a life spent thinking about evolution, although in my case a world war, and a false start as an aircraft engineer, meant that I did not become a professional biologist until I was thirty; However, our paths diverged earlier. When I was eight, my father died, and we moved away from London to the country: for me, visiting the Natural History Museum was replaced by watching birds. I suspect that this switch may help to explain the disagreements between us, to which I shall return below. It also explains why I can review two books about dinosaurs only as an interested outsider, and not as an expert.
The two books are very different in intent. The Dinosauria is a joint effort by several experts, aimed at professionals, summarizing the current state of dinosaur research. Dinosaurs, Spitfires, and Sea Dragons is aimed at a wider readership, but one prepared to make some effort. It attempts to deduce how dinosaurs lived, mainly by comparing them with animals alive today.
Current research on dinosaurs, although it forms the basis of speculations on how they lived, is not in the main concerned with that question. Obviously, the major effort is devoted to discovering and describing the fossils themselves. I was impressed to learn that almost half the known genera of dinosaurs have been discovered during the last twenty years. The interpretation of these fossils is primarily aimed at recognizing anatomical similarities and differences, and arranging the specimens into a natural classification. These, of course, have been the main preoccupations of comparative anatomists for over two hundred years. Today, a “natural classification” is usually taken to mean one that reflects evolutionary history. If species are grouped together in a “taxon” (that is, a genus, family, order, or what have you), the implication is that the taxon includes all species, and only those species, descended from a common ancestor. In the jargon of the trade, the taxon is “monophyletic.” Anatomists, however, were seeking a natural classification long before they accepted a theory of evolution, and their methods were little altered when they did come to accept the theory. For some reason I do not understand, they usually assumed that a natural classification must be a hierarchical one—that is, species must be grouped into genera, genera into families, families into orders, and so on. Of course, natural classifications do not have to be like that: one need only mention Mendeleyev’s periodic table of the elements. But if species arose by a branching process of evolution—that is, if species split into two, but do not join again—then the natural classification is indeed hierarchical.
At this point, I must declare a strong prejudice against comparative anatomy, acquired when an undergraduate at University College London in the 1940s. At that time, at most British universities, the core of a degree in zoology consisted of comparative anatomy. This was often justified on the grounds that it provided the necessary background for understanding evolution. It seemed to me then, as it still does, that the obvious need, once Darwin had published the Origin of Species, was to understand heredity, and not to more anatomy. This was obvious to Darwin himself, as is clear from the immense, if largely unsuccessful, effort that he put into The Variation of Animals and Plants under Domestication. Sadly, it was not obvious to T.H. Huxley, nor to those responsible for training biologists in the English-speaking world, so that, a hundred years later, students were still being exposed to a course in nineteenth century anatomy.
This would have mattered less if the courses had been intellectually respectable, but they were not. It was accepted that some anatomical characteristics were important when classifying animals, and others were not. For example, it was important, when classifying land vertebrates, to ask how many openings there were in the side of the skull: lizards, dinosaurs, and crocodiles had two, mammals had one, and turtles had none. But why was it important? In those days, it was a matter of “judgment,” or “experience,” or “what I tell you three times is true.”
It is clear from reading The Dinosauria, edited by Weishampel, Dodson, and Osmólska, that an important change for the better has taken place since I was a student, although the new dispensation too will become an abuse if we do not watch it. The name of the change is “cladistics.” This is one of two very different approaches that have been made in recent years to devise a rationale for classification. The first was numerical taxonomy. Robert Sokal and Peter Sneath, the founders of numerical taxonomy, took the simple way out. If there is no good reason for preferring one trait to another, let us, like good democrats, treat all traits equally. When classifying a set of animals, write down everything about them, in as mindless a way as possible. Then write a computer program that will group together those species that are alike in most respects. It was a method only possible with the advent of computers. It had the great merit that there was no damned judgment about it. In practice, however, numerical taxonomy has been largely replaced by cladistics, essentially, I believe, because Sokal and Sneath insisted that classification should be independent of evolutionary considerations.
Cladistics was invented by an East German taxonomist, Willi Hennig, but has since become the religion of many museum taxonomists, most notably those at the American Museum of Natural History in New York. There are two components to the method, one substantive and one semantic. The substantive component concerns which traits are important when classifying, if one wishes one’s classification to reflect evolution. The rule is that one should use “shared derived characters,” and not “shared primitive characters”; there are some impressive Greek words for these that I won’t bother with. For example, zebras and horses have a single toe, whereas opossums and humans have five toes. To have five toes is not evidence of close relationship, because it is the primitive condition in land vertebrates; but having a single toe is good evidence of relationship, because it is a recently evolved state.
The hard question, of course, is how to decide which is the primitive and which the derived state. If we had a perfect fossil record, there would be no problem: but if we had a perfect record, classification would be trivially easy anyway. In the absence of a useful fossil record, there are ways of deciding the question, of which the most useful is the “outgroup” comparison: when classifying the members of group A, the primitive state is that found in a closely related group, B. Thus one reason for thinking that having five toes is the primitive condition for mammals is that a related outgroup, for example the lizards, have five toes.
Weishampel, Dodson, and Osmólska insist that a classification must be derived from anatomical information by cladistic means. Their insistence gives an intellectual rigor to the book which it would otherwise lack. They are less consistent in the other, semantic component of cladistics. This asserts that any named taxon—for example, Eutheria (placental mammals), Dinosauria, Lacertilia (lizards)—must be monophyletic: that is, it must include all, and only, the descendants of a single common ancestor. Notice that, according to this requirement, Pisces (fish) is not a valid taxon, because it does not include the land vertebrates, which are descended from fish.
Consistent with this view, the editors include the birds among the dinosaurs, as they must on cladistic grounds, because the birds are thought to be descended from a particular group of small carnivorous dinosaurs. Yet Dodson’s chapter on dinosaur paleobiology opens, “The desire to understand dinosaurs as living creatures is compelling, despite the inherent problems of studying creatures that have not lived and breathed for at least sixty-five million years.” What, no bird song? Of course, I am nit-picking, but there is a serious point to be made. Cladistic methods may well be the best we have for deciding on evolutionary relationships, but no one in his senses would abide by the fine print of cladistic terminology.
It is of course true that we would like to know more of dinosaurs as living creatures. How much would I give to see Pteranodon flying? McGowan’s book, Dinosaurs, Spitfires, and Sea Dragons seems to me admirable. It is particularly good on locomotion, perhaps because this is the topic on which the skeleton is most informative. Two questions have been hotly debated in recent years. Were the dinosaurs warm-blooded? Why did they become extinct? The short answer to both these questions is that we do not know. The attractive feature of both questions is the wide range of kinds of information that may be relevant. For example, there are two good reasons for thinking that at least some dinosaurs may have been warm-blooded. The first is the platelike internal structure of their bones, which today is characteristic of the rapidly growing, warm-blooded mammals and birds. The second concerns the ratio of the numbers of herbivores to carnivores. Warm-blooded carnivores need more to eat, and therefore there are fewer of them, relative to their prey, than there are of cold-blooded carnivores. Fossil data suggest that there were relatively few carnivores. My impression, however, is that the question is still open.
On extinction, the suggestion that a meteorite collision with the earth produced climactic conditions that caused dinosaurs to disappear was put forward ten years ago by L. W. and W. Alvarez; it looks increasingly convincing. There are two main facts to bear in mind. The first is the presence, at the appropriate level in the rocks, between the Cretaceous and the Paleocene, of an iridium-rich layer: the presence of iridium is most easily explained by a meteorite collision. The second is that the extinction of the dinosaurs was not the only, or the most dramatic, extinction event at that time. For example, there was a widespread extinction of marine planktonic animals at the same time. It follows that a mechanism peculiar to dinosaurs will not do: we need something global and general. The very selective nature of the extinctions should provide important clues. For example, large land animals were wiped out, but small mammals, tortoises, and crocodiles did well: planktonic and filter-feeding marine animals did badly, but scavenging forms survived better.
There are many other questions about the life of dinosaurs that we would like to answer. The danger, of course, is that we can always dream up a reason for the presence of particular structures, but it is hard to test our speculations. All the same, I have never been able to resist speculating. My first published paper concerned the origin of flight. The earliest pterodactyls (not strictly dinosaurs, but close relatives) had long tails, with a fluke of skin at the end: the later ones did not. This was usually explained by saying that the pterodactyls evolved from bipedal reptiles that needed long tails for balance. The long tails, bordered by feathers, of the first birds were explained in the same way. I argued that stable flight requires a long tail: every child knows that one must put feathers on the back of an arrow if one wants it to fly straight. But with a sufficiently developed sensory and nervous system, it would be possible for an animal to fly, even though, in the absence of controls, it would be unstable. Such an animal would have advantages in landing and turning. I therefore argued that the first flying animals (Archaeopteryx, the first pterodactyls, the first flying insects) had long tails because they needed them as stabilizers. Later, with a highly evolved sensory system, the long tail could be dispensed with.
Such speculations are amusing, but difficult to test, especially when they concern extinct animals. Some paleontologists feel that adaptive stories of this kind have no place in science: scientists should confine themselves to accurate description and rational comparison of anatomical structures. The same objection has been made to adaptive storytelling about living animals, not only about dead ones. In 1979, I organized for the Royal Society a symposium on “The evolution of adaptation by natural selection.” Much the most influential paper at the meeting was read by Stephen Gould and Richard Lewontin. In it they criticized what they called the “adaptationist program”—that is, the habit of dividing an organism into traits, and inventing an adaptive explanation of each trait separately. Although neither would wish to deny the effectiveness of natural selection in adapting organisms to their ways of life, they emphasized the importance of other processes, in particular the constraints placed on the possibilities of evolutionary change by pre-existing developmental pathways, and chance events inevitable in a historical process. They complained, with much eloquence and some effect, of the habit of inventing “Just So Stories,” without any serious attempt to test the stories, or to consider alternatives.
The effect of the Gould–Lewontin paper has been considerable, and on the whole welcome. I doubt if many people have stopped trying to tell adaptive stories. Certainly I have not done so myself. I continue to ask why peacocks have long tails, or why some green finches are bright green and others dull brown, or why desert spiders bite one another when fighting over webs but woodland ones do not, or, more globally, why most animals and plants reproduce sexually. The answers I seek are adaptive ones: that is, they are answers in terms of the selective advantages and disadvantages of the particular structures or behaviors. In general, I don’t think that either historical accidents or developmental constraints have much to do with it, although it may well have been an accident that it was the peacock’s tail and not some other part of his anatomy that was exaggerated, and I think there is an important element of developmental constraint in the maintenance, if not in the origin, of sex. As we have sought for adaptive explanations, however, we have done so with an occasional glance over our shoulders to see if Gould or Lewontin were watching. After all, no one wants to be accused of telling Just So Stories.
There are two ways in which practitioners of the adaptationist program have cleaned up their act during the last ten years. There has been a change in the extent to which people consider alternative explanations of their data, and in the sophistication of their statistics. There has also been a revolution in the “comparative method”: that is, the comparison of species with different characteristics as a way of testing various hypotheses that might explain those differences. In what I am still tempted to think of as the good old days, one thought up some theory that predicted that animals in environment A ought to do X. One then searched the published literature for a few examples that fitted the prediction, and regarded one’s hypothesis as confirmed.
As a result of the work of such evolutionary biologists as Joe Felsenstein and Paul Harvey, among others, one can no longer get away with this. One is obliged to carry out a proper survey of the literature, and proper statistical tests. Interestingly, the tests require that one know the evolutionary relationships of the animals surveyed, and hence, in practice, depend on cladistic analysis. The reason is as follows. Suppose that one has a theory (as in fact I do) that predicts plants with seeds that are dispersed by animals should also have separate male and female individuals (as opposed to being hermaphrodite). It is no good finding one hundred species with both traits present, if all one hundred are junipers. The conjunction between animal-dispersed seeds and separate sexes may have evolved once only, so that we have one confirming instance, and not one hundred. Although it is easy to describe the fallacy, it is harder to devise ways of avoiding it. Such ways do now exist, however, and we are therefore in a better position to test adaptationist, or indeed nonadaptationist, explanations. Of course I cannot say that the Gould–Lewontin paper stimulated these changes in practice, but it may have helped. I hope Harvey will forgive me if I reveal that I recall him saying, after he completed a paper on allometry, “That’ll get Lewontin.”
Of course it is not new that biologists should argue about the relative importance of adaptive (functional) and developmental (archetypes, bauplanes, or basic structural plans) considerations in determining anatomy: the debate goes back at least to that between Cuvier and Geoffroy Saint-Hilaire. But I am still tempted to speculate on the factors that lead particular scientists to take particular positions. This is especially so because, when I meet Gould or Lewontin, we rarely disagree about particular cases. It is more a matter of where we choose to put our research effort.
I think that the prevalence of adaptationism among the English can be traced to two causes: their escape from natural theology, and their continued love of natural history. In the early eighteenth century an influential group of scientists associated with the Royal Society sought for evidence of a creator in the manifest design of the universe. It is a curious fact that the first use of a statistical significance test was in a paper by Arbuthnot in the Transactions of the Royal Society in 1710 arguing that the human sex ratio demonstrated the existence of a benificent creator. He was typical of his time. The movement culminated in Paley’s watchmaker. Darwin himself was much influenced as a young man by natural theology. In his theory of natural selection, he did not abandon his conviction that there is design in nature: he merely provided an alternative, non-theistic, explanation for it. In Richard Dawkin’s phrase, he invented the blind watchmaker, and in so doing liberated himself, and us, from the need to believe in God.
This will seem a naive statement. After all, there are Christians who accept natural selection and evolutionists who have never been Christians, and there are reasons other than the argument from design for believing in God. Yet the path that starts with the argument from design goes on to see that the main problem for any theory of evolution is to explain adaptation, and concludes by seeing natural selection as the major cause of evolutionary change, is a common one. It accurately describes my own intellectual development as a boy, and I think it is widespread among evolutionary biologists: in England and America, at least, it is surprising how many of them are lapsed Christians. However, the recognition of adaptation as the central problem need not have its origins in natural theology. I was much struck, recently, when reading the work of the great German biologist August Weismann to find that he, too, saw adaptation as central, and that it was this that led him to accept, not only the central role of selection, but also a theory of heredity that denied the inheritance of acquired characters. I do not know what influences, other than Darwin, led him to this position.
The second influence leading to adaptationism is, I think, natural history. Lewontin once complained that ecological genetics (a genetical version of the adaptationist program) was an invention of English country gentlemen. The point about English country gentlemen, of course, is that many of them were naturalists. To be fair, it was not necessary to be a gentleman: Darwin might have claimed to be one, but Wallace would not. It is easy to see how the structure of English, and indeed European, society could encourage an interest in natural history. But why should naturalists be adaptationists? I never met a bird watcher who was not a naive adaptationist, but why? I think it may be that, if one watches an animal doing something, it is hard not to identify with it, and hence to ascribe a purpose to its behavior.
This is true even when what the animal is doing seems to make no sense. I remember sitting, with my wife and two friends, watching the vast flocks of flamingos on the lake in the center of the Ngoro-Ngoro crater. Most were feeding, but a group of several thousand were marching and counter-marching through the water, with their heads in the air. Occasionally a bird would join the marchers, or drop out and resume feeding. As we watched, we speculated on what they were up to. Were they stirring up food from the mud, or suggesting to the flock that the time had come to migrate or advertising their virtues as potential mates? Probably none of these explanations is correct, but the one thing that did not occur to any of us was that the marching served no purpose. I’m sure they were not suffering from a developmental constraint, or the victims of an historical contingency. Of course, there is a simpler reason why naturalists tend to be adaptationists: it is an approach that usually works.
I suggested earlier that Gould and I might today hold different views if he had lived in the country and watched birds, and I had remained true to my first love of fossils. Fossils have a romance of their own, but they don’t do much. If watching birds makes one an adaptationist, studying fossils often has the opposite effect.
I fear that what started out as a review of two books about dinosaurs has wandered off into a discussion of the functional and adaptationist approaches to anatomy. There is, however, some excuse. The two books admirably reflect these two approaches. Weishampel, Dodson, and Osmólska show what comparative anatomy can be like at its best. They also show how cladistic methods have introduced some rigor into biological classification. In contrast, McGowan’s approach is functional. If we are prepared to apply principles learned from the study of living organisms to interpreting extinct ones, we can build a picture of how the dinosaurs lived.
April 25, 1991