The Extended Phenotype

by Richard Dawkins

  We began this chapter by wondering why replicators have ganged up into large, multicellular clones called organisms, and we initially gave a rather unsatisfactory answer. A more satisfying answer is now starting to emerge. An organism is the physical unit associated with one single life cycle. Replicators that gang up in multicellular organisms achieve a regularly recycling life history, and complex adaptations to aid their preservation, as they progress through evolutionary time.

  Some animals have a life cycle involving more than one distinct body. A butterfly is utterly different from the caterpillar which preceded it. It is hard to imagine a butterfly growing from a caterpillar by slow, within-organ change: caterpillar organ growing into corresponding butterfly organ. Instead, what happens is that the complex organ structure of the caterpillar is largely broken down and the tissues of the caterpillar are used as fuel for the development of a whole new body. The new butterfly body does not quite restart from a single cell, but the principle is the same. It develops a radically new bodily structure from simple, relatively undifferentiated imaginal discs. There is a partial return to the drawing board.

  Returning to the growth/reproduction distinction itself, Janzen was not actually wrong. Distinctions can be unimportant for some purposes, while they remain important for other purposes. For discussing certain kinds of ecological or economic questions, there may be no important distinction between growth and asexual reproduction. A sisterhood of aphids may indeed be analogous to a single bear. But for other purposes, for discussing the evolutionary putting together of complex organization, the distinction is crucial. A certain type of ecologist may gain illumination from comparing a field full of dandelions with a single tree. But for other purposes it is important to understand the differences, and to see the single dandelion ramet as analogous to the tree.

  But Janzen’s position is, in any case, a minority one. A more typical biologist might think it perverse of Janzen to regard asexual reproduction in aphids as growth, and equally perverse of Harper and me to regard vegetative propagation by multicellular runners, as growth and not reproduction. Our decision is based on the assumption that the runner is a multicellular meristem rather than a single-celled propagule, but why should we regard this as an important point? Again, the answer may be seen in a thought experiment involving two hypothetical species of plants, in this case strawberry-like plants called M and S (Dawkins in press).

  Both the hypothetical strawberry-like species propagate vegetatively, by runner. In both there is a population of what appear to be distinct and recognizable ‘plants’ connected by a network of runners. In both species, each ‘plant’ (i.e. ramet) can give rise to more than one daughter plant, so that we have the possibility of exponential growth of the ‘population’ (or growth of the ‘body’ depending on your point of view). Even though there is no sex, there can be evolution since mutations will sometimes occur in the mitotic cell divisions (Whitham & Slobodchikoff in press). Now comes the crucial difference between the two species. In species M (for many, or multicellular, or meristem), the runner is a broad-fronted multicellular meristem. This means that two cells in any one ‘plant’ may be the mitotic descendants of two different cells in the parent plant. In terms of mitotic descent, a cell may therefore be a closer cousin of a cell in another ‘plant’ than it is of another cell in its own plant. If mutation has introduced genetic heterogeneity into the cell population, this means that individual plants may be genetic mosaics, with some cells having closer genetic relatives in other plants than in their own. We will see the consequences of this for evolution in a moment. Meanwhile we turn to the other hypothetical species.

  Species S (for single) is exactly like M, except that each runner culminates in a single apical cell. This cell acts as the basal mitotic ancestor of all the cells of the new daughter plant. This means that all cells in a given plant are closer cousins to each other than they are to any cells in other plants. If mutation introduces genetic heterogeneity into the population of cells, there will be relatively few mosaic plants. Rather, each plant will tend to be a genetically uniform clone, but it may differ genetically from some other plants, while being genetically identical to yet other plants. There will be a true population of plants, each one of which has a genotype characteristic of all its cells. It is therefore possible to conceive of selection, in the sense which I have called ‘vehicle selection’, acting at the level of the whole plant. Some whole plants may be better than others, because of their superior genotypes.

  In species M, especially if the runners are very broad-fronted meristems, a geneticist will not discern a population of plants at all. He will see a population of cells, each with its own genotype. Some cells will be genetically identical, others will have different genotypes. A form of natural selection might go on among cells, but it is hard to imagine selection among ‘plants’, because ‘the plant’ is not a unit that can be identified as having its own characteristic genotype. Rather, the whole mass of straggling vegetation will have to be regarded as a population of cells, with cells of any one genotype being untidily peppered across the different ‘plants’. The unit which I have called the ‘gene vehicle’, and which Janzen has called the ‘evolutionary individual’, will, in such a case, be no larger than the cell. It is cells that will be the genetic competitors. Evolution may take the form of improvements in cell structure and physiology, but it is hard to see how it could take the form of improvements in individual plants or their organs.

  It might be thought that improvements in organ structure could evolve, if it regularly happened that particular subpopulations of cells, in discrete areas of the plant, were a clone, descended from a single mitotic ancestor. For instance, the runner giving rise to a new ‘plant’ might be a broad-fronted meristem, but it might still be the case that each leaf sprang from a single cell at its own base. A leaf could therefore be a clone of cells more closely related to each other than to cells anywhere else in the plant. Given the commonness of somatic mutation in plants (Whitham & Slobodchikoff in press), might one not therefore imagine the evolution of improved complex adaptation at the level of the leaf, if not at the level of the whole plant? A geneticist could now discern a genetically heterogeneous population of leaves, each one made up of genetically homogenous cells, so might not natural selection go on between successful leaves and unsuccessful leaves? It would be tidy if the answer to this question could be yes; that is, if we could assert that vehicle selection will go on at any level in the hierarchy of multicellular units, provided that the cells within a unit tend to be genetically uniform compared with cells in other units at the same level. Unfortunately, however, something has been left out of the reasoning.

  It will be remembered that I classified replicators into germ-line replicators and dead-end replicators. Natural selection results in some replicators becoming more numerous at the expense of rival replicators, but this leads to evolutionary change only if the replicators are in germ-lines. A multicellular unit qualifies as a vehicle, in an evolutionarily interesting sense, only if at least some of its cells contain germ-line replicators. Leaves normally do not so qualify, for their nuclei contain only dead-end replicators. Leaf cells synthesize chemical substances which ultimately benefit other cells that do contain germ-line copies of the leaf genes, the genes which gave the leaves their characteristically leafy phenotypes. But we cannot accept the conclusion of the previous paragraph, that inter-leaf vehicle selection, and inter-organ selection generally, could go on if only the cells within an organ were closer mitotic cousins than cells in different organs. Inter-leaf selection could have evolutionary consequences only if leaves directly spawned daughter leaves. Leaves are organs, not organisms. For inter-organ selection to occur, it is necessary that the organs concerned should have their own germ lines and do their own reproducing, and this they normally do not. Organs are parts of organisms, and reproduction is the prerogative of organisms.

  For clarity I have been a bit extreme. There could be a range of int
ermediates between my two strawberry-like plants. Species M’s runner was said to be a broad-fronted meristem, while species S’s runner narrowed down to a one-celled bottleneck at the base of each new plant. But what if there was an intermediate species with a two-celled bottleneck at the base of each new plant? There are two main possibilities here. If the pattern of development is such that it is unpredictable which cells in the daughter plant will be descended from which of the two stem cells, the point I have made about developmental bottlenecks will simply be weakened quantitatively: genetic mosaics may occur in the population of plants, but there will still be a statistical tendency for cells to be genetically closer to fellow members of the same plant than to cells in other plants. Therefore we may still talk meaningfully about vehicle selection between plants in a population of plants, but the inter-plant selection pressure may have to be strong to outweigh selection among cells within plants. This is, incidentally, analogous to one of the conditions for ‘kin-group selection’ (Hamilton 1975a) to work. To make the analogy, we have only to see the plant as a ‘group’ of cells.

  The second possibility arising out of the assumption of a two-celled bottleneck at the base of each plant is that the pattern of development of the species might be such that certain organs of the plant are always the mitotic descendants of a designated one of the two cells. For instance, cells of the root system might develop from a cell in the lower part of the runner, while the rest of the plant developed from the other cell, in the upper part of the runner. If, further, the lower cell is always descended from a root cell in the parent plant, while the upper cell is recruited from an above-ground cell in the parent plant, we would have an interesting situation. Root cells would be closer cousins of other root cells in the population at large than they would be of stalk and leaf cells in their ‘own’ plant. Mutation would open up the possibility of evolutionary change, but it would be split-level evolution. Genotypes below ground could evolve away from genotypes above ground, irrespective of apparent joint membership in discrete ‘plants’. Theoretically we could even have a kind of within-organism ‘speciation’.

  To recapitulate, the significance of the difference between growth and reproduction is that reproduction permits a new beginning, a new developmental cycle and a new organism which may be an improvement, in terms of the fundamental organization of complex structure, over its predecessor. Of course it may not be an improvement, in which case its genetic basis will be eliminated by natural selection. But growth without reproduction does not even allow the possibility of radical change at the organ level, either in the direction of improvement or the reverse. It allows only superficial tinkering. You may divert a developing Bentley into a fully grown Rolls Royce, simply by tinkering with the assembly process at the late point where the radiator is added. But if you want to change a Ford into a Rolls Royce you must start at the drawing board, before the car starts ‘growing’ on the assembly line at all. The point about recurrent reproduction life cycles, and hence, by implication, the point about organisms, is that they allow repeated returns to the drawing board during evolutionary time.

  We must beware here of the heresy of ‘biotic’ adaptationism (Williams 1966). We have seen that recurrent reproduction life cycles, i.e. ‘organisms’, make the evolution of complex organs possible. It is all too easy to treat this as a sufficient adaptive explanation for the existence of organismal life cycles, on the grounds that complex organs are, in some vague sense, a good idea. A related point is that repeated reproduction is possible only if individuals die (Maynard Smith 1969), but we should not therefore wish to say that individuals die as an adaptation to keep evolution going! The same could be said of mutation: its existence is a necessary precondition for evolution to occur, but it is nevertheless quite likely that natural selection has favoured evolution in the direction of a zero mutation rate—fortunately never attained (Williams 1966). The growth/reproduction/death type of life cycle—the multicellular clonal ‘organism’ type of life cycle—has had far-reaching consequences and was probably essential for the evolution of adaptive complexity, but this is not tantamount to an adaptive explanation for the existence of this type of life cycle. The Darwinian must begin by seeking immediate benefits to genes promoting this kind of life cycle, at the expense of their alleles. He may go on to acknowledge the possibility of other levels of selection, differential lineage extinction, say. But he must show the same circumspection in this difficult theoretical field as Fisher (1930a), Williams (1975) and Maynard Smith (1978a) brought to the analogous suggestions about sexual reproduction being there because it speeds up evolution.

  The organism has the following attributes. It is either a single cell, or if it is multicellular its cells are close genetic kin of each other: they are descended from a single stem cell, which means that they have a more recent common ancestor with each other than with the cells of any other organism. The organism is a unit with a life cycle which, however complicated it may be, repeats the essential characteristics of previous life cycles, and may be an improvement on previous life cycles. The organism either consists of germ-line cells, or it contains germ-line cells as a subset of its own cells, or, as in the case of a sterile social insect worker, it is in a position to work for the welfare of germ-line cells in closely related organisms.

  I have not aspired in this final chapter to give a completely satisfying answer to the question of why there are large multicellular organisms. I will be content if I can arouse new curiosity about the question. Instead of accepting that organisms exist and asking how adaptations benefit the organisms displaying them, I have tried to show that the very existence of organisms should be treated as a phenomenon deserving of explanation in its own right. Replicators exist. That is fundamental. Phenotypic manifestations of them, including extended phenotypic manifestations, may be expected to function as tools to keep replicators existing. Organisms are huge and complex assemblages of such tools, assemblages shared by gangs of replicators who in principle need not have gone around together but in fact do go around together and share a common interest in the survival and reproduction of the organism. As well as drawing attention to the phenomenon of the organism as one that needs explanation, I have tried in this last chapter to sketch the general direction in which we might proceed in seeking an explanation. It is only a preliminary sketch, but, for what it is worth, I summarize it here.

  The replicators that exist tend to be the ones that are good at manipulating the world to their own advantage. In doing this they exploit the opportunities offered by their environments, and an important aspect of the environment of a replicator is other replicators and their phenotypic manifestations. Those replicators are successful whose beneficial phenotypic effects are conditional upon the presence of other replicators which happen to be common. These other replicators are also successful, otherwise they would not be common. The world therefore tends to become populated by mutually compatible sets of successful replicators, replicators that get on well together. In principle this applies to replicators in different gene pools, different species, classes, phyla and kingdoms. But a relationship of specially intimate mutual compatibility has grown up between subsets of replicators that share cell nuclei and, where the existence of sexual reproduction makes the expression meaningful, share gene-pools.

  The cell nucleus as a population of uneasily cohabiting replicators is a remarkable phenomenon in itself. Just as remarkable, and quite distinct, is the phenomenon of multicellular cloning, the phenomenon of the multicellular organism. Replicators whose effects interact with those of other replicators to produce multicellular organisms achieve for themselves vehicles with complex organs and behaviour patterns. Complex organs and behaviour patterns are favoured in arms races. The evolution of complex organs and behaviour patterns is possible because the organism is an entity with a recurrent life cycle, each cycle beginning with a single cell. The fact that each cycle restarts in every generation from a single cell permits mutations to achieve radical
evolutionary changes by going ‘back to the drawing board’ of embryological engineering. It also, by concentrating the efforts of all cells in the organism on the welfare of a narrow, shared germ-line, partly removes the ‘temptation’ for outlaws to work for their own private good at the expense of the other replicators with a stake in the same germ-line. The integrated multicellular organism is a phenomenon which has emerged as a result of natural selection on primitively independent selfish replicators. It has paid replicators to behave gregariously. The phenotypic power by which they ensure their survival is in principle extended and unbounded. In practice the organism has arisen as a partially bounded local concentration, a shared knot of replicator power.

  Afterword by Daniel Dennett

  Why is a philosopher writing an Afterword for this book? Is The Extended Phenotype science or philosophy? It is both; it is science, certainly, but it is also what philosophy should be, and only intermittently is: a scrupulously reasoned argument that opens our eyes to a new perspective, clarifying what had been murky and ill-understood, and giving us a new way of thinking about topics we thought we already understood. As Richard Dawkins says at the outset, ‘The extended phenotype may not constitute a testable hypothesis in itself, but it so far changes the way we see animals and plants that it may cause us to think of testable hypotheses that we would otherwise never have dreamed of’(p. 2). And what is this new way of thinking? It is not just the ‘gene’s-eye point of view’ made famous in Dawkins’s 1976 book, The Selfish Gene. Building here on that foundation, he shows how our traditional way of thinking about organisms should be replaced by a richer vision in which the boundary between organism and environment first dissolves and then gets (partially) rebuilt on a deeper foundation. ‘I shall show that the ordinary logic of genetic terminology leads inevitably to the conclusion that genes can be said to have extended phenotypic effects, effects which need not be expressed at the level of any particular vehicle’ (p. 196). Dawkins is not declaring revolution; he is using ‘the ordinary logic of genetic terminology’, to prove a striking implication of the biology already securely in hand, a new ‘central theorem’: ‘An animal’s behaviour tends to maximize the survival of the genes “for” that behaviour, whether or not those genes happen to be in the body of the particular animal performing it’ (p. 233). Dawkins’s earlier eye-opener, the recommendation that biologists adopt the gene’s-eye point of view, was also not presented as revolutionary, but rather as clarifying a shift of attention that had already begun to sweep through biology in 1976. There has been so much anxious and misguided criticism of Dawkins’s earlier idea that many laypeople and even some biologists may fail to appreciate how bountiful this shift of attention has been. We now know that a genome, such as the human genome, consists of, and depends on, mechanisms of breathtaking deviousness and ingenuity—not just molecular copyists and proof-reading editors, but outlaws and vigilantes to combat them, chaperones and escape artists and protection rackets and addicts and other devious nano-agents, out of whose robotic conflicts and projects emerge the marvels of visible nature. The fruits of this new vision extend far beyond the almost daily headlines about striking new discoveries about one bit of DNA or another. Why and how do we age? Why do we get sick? How does HIV work? How do brains get wired up in the course of embryological development? Can we use parasites instead of poisons to control agricultural pests? Under what conditions is cooperation not just possible, but likely to arise and persist? All these vital questions and many more are illuminated by re-thinking the issues in terms of the processes by which the opportunities for replicators to replicate, and their associated costs and benefits, sort themselves out.