The Extended Phenotype

by Richard Dawkins

  The second way in which the frequency of a gene B in a population can affect the survival prospects of a gene A is ‘between-body’ interaction. The vital influence here is on the probability that any body in which A sits will meet another body in which B sits. My hypothetical cicadas provided an example of this. So does Fisher’s sex ratio theory. As I have emphasized, it has been one of my purposes in this chapter to minimize the distinction between the two kinds of gene interaction, within-body and between-body interaction.

  But now consider interactions between genes in different gene-pools, different species. It will be seen that there is rather little distinction between a cross-species gene interaction and a within-species between-individual gene interaction. In neither case do the interacting genes share a body. In both cases the survival prospects of each may depend on the frequency, in its own gene-pool, of the other gene. Let me illustrate the point using the lupin thought experiment again. Suppose there is a species of beetle which is polymorphic like the cicadas. In some areas it turns out that the pink morphs of both species, cicadas and beetles, predominate, while in other areas the blue morphs of both species predominate. The two species differ in body size. They ‘cooperate’ in faking inflorescences, the smaller-bodied cicadas tending to sit near the tips of stems, where small flowers might be expected, the larger beetles tending to sit nearer the base of each fake inflorescence. A joint beetle/cicada ‘inflorescence’ fools birds more effectively than either a pure beetle or a pure cicada one.

  Model 2’s frequency-dependent selection will tend to lead to the evolution of one of two evolutionarily stable states, just as before, except that two species are now involved. If historical accident leads to one local area being dominated by pink morphs (regardless of species), selection within both species will favour pink morphs over blue; and vice versa. If the beetle species was relatively recently introduced into areas already colonized by the cicada species, the direction of selection within the beetle species will depend on the colour of the locally predominant morph of cicadas. Thus there will be frequency-dependent interaction between genes in two different gene-pools, the gene-pools of two non-interbreeding species. In faking the inflorescence of a lupin, cicadas might cooperate with spiders or snails just as effectively as with beetles or with cicadas of another species. Model 2 works across species and across phyla, as well as across individuals and within individuals.

  Across kingdoms, too. Consider the interaction between flax (Linum usitissimum) and the rust fungus Melampsora lini, although this is an antagonistic rather than a cooperative interaction. ‘There is essentially a one-to-one matching in which a specific allele in the flax confers resistance to a specific allele in the rust. This “gene-for-gene” system has since been found in numerous other plant species … Models of gene-for-gene interactions are not formulated in terms of ecological parameters because of the specific nature of the genetic systems. It is one case in which the genetic interactions between species can be understood without reference to the phenotypes. A model of a gene-for-gene system would necessarily have between-species frequency dependence … (Slatkin & Maynard Smith 1979, pp. 255–256).

  In this chapter, as in others, I have used hypothetical thought experiments to aid clear explanation. In case they are found too far-fetched, let me turn to Wickler again for an example of a real cicada which does something at least as far-fetched as anything I have invented. Ityraea nigrocincta, like I. gregorii, practises cooperative mimicry of lupin-like inflorescences, but it ‘possesses a further peculiarity in that both sexes have two morphs, a green form and a yellow form. These two morphs may squat together, and the green forms tend to sit at the top of the stem, especially on vertical stems, with the yellow forms below. The result is an extremely convincing “inflorescence”, because the true flowers of inflorescences often open progressively from base to apex, so that green buds are still present at the tip when the base is covered with open flowers’ (Wickler 1968).

  These three chapters have extended the concept of phenotypic expression of genes by easy stages. We began with the recognition that even within a body there are many degrees of distance of gene control over phenotypes. For a nuclear gene to control the shape of the cell in which it sits is presumably simpler than to control the shape of some other cell, or of the whole body in which the cell sits. Yet we conventionally lump the three together and call them all genetic control of phenotype. My thesis has been that the slight further conceptual step outside the immediate body is a comparatively minor one. Nevertheless it is an unfamiliar one, and I tried to develop the idea in stages, working through inanimate artefacts to internal parasites controlling their hosts’ behaviour. From internal parasites we moved via cuckoos to action at a distance. In theory, genetic action at a distance could include almost all interactions between individuals of the same or different species. The living world can be seen as a network of interlocking fields of replicator power.

  It is hard for me to imagine the kind of mathematics that the understanding of the details will eventually demand. I have a dim vision of phenotypic characters in an evolutionary space being tugged in different directions by replicators under selection. It is of the essence of my approach that the replicators tugging on any given phenotypic feature will include some from outside the body as well as those inside it. Some will obviously be tugging harder than others, so the arrows of force will have varying magnitude as well as direction. Presumably the theory of arms races—the rare-enemy effect, the life/dinner principle, etc.—will have a prominent role to play in the assignment of these magnitudes. Sheer physical proximity will probably play a role: genes seem likely, other things being equal, to exert more power over nearby phenotypic characters than over distant ones. As an important special case of this, cells are likely to be quantitatively more heavily influenced by genes inside them than by genes inside other cells. The same will go for bodies. But these will be quantitative effects, to be weighed in the balance with other considerations from arms race theory. Sometimes, say because of the rare-enemy effect, genes in other bodies may exert more power than the body’s ‘own’ genes, over particular aspects of its phenotype. My hunch is that almost all phenotypic characters will turn out to bear the marks of compromise between internal and external replicator forces.

  The idea of conflict and compromise between many selection pressures bearing on a given phenotypic character is, of course, familiar from ordinary biology. We often speak of, say, the size of a bird’s tail as a compromise between the needs of aerodynamics and the needs of sexual attractiveness. I do not know what kind of mathematics are considered suitable for describing this kind of within-body conflict and compromise, but whatever they are, they should be generalized to cope with the analogous problems of genetic action at a distance and extended phenotypes.

  But I have not the wings to fly in mathematical spaces. There must be a verbal message for those that study animals in the field. What difference will the doctrine of the extended phenotype make to how we actually see animals? Most serious field biologists now subscribe to the theorem, largely due to Hamilton, that animals are expected to behave as if maximizing the survival chances of all the genes inside them. I have amended this to a new central theorem of the extended phenotype: 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 the behaviour. The two theorems would amount to the same thing if animal phenotypes were always under the unadulterated control of their own genotypes, and uninfluenced by the genes of other organisms. In advance of a mathematical theory to handle the quantitative interactions between conflicting pressures, perhaps the simplest qualitative conclusion is that the behaviour we are looking at may be, at least partly, an adaptation for the preservation of some other animal’s or plant’s genes. It may therefore be positively maladaptive for the organism performing the behaviour.

  Once when I tried to persuade a colleague
of this—he is a staunch believer in the power of Darwinian selection, and a good field investigator of it—he thought that I was making an anti-adaptation point. He warned me that time and again people had written off some quirk of animal behaviour or morphology as functionless or maladaptive, only to discover that they had not fully understood it. He was right. But the point I am making is different. When I say here that a behaviour pattern is maladaptive, I only mean it is maladaptive for the individual animal performing it. I am suggesting that the individual performing the behaviour is not the entity for whose benefit the behaviour is an adaptation. Adaptations benefit the genetic replicators responsible for them, and only incidentally the individual organisms involved.

  This could have been the end of the book. We have extended the phenotype out as far as it can go. The past three chapters build to a climax of a sort, and we might have been content with this as consummation. But I prefer to end on an upbeat, to begin the arousing of a tentative new curiosity. I confessed at the outset to being an advocate, and an easy way for any advocate to prepare the ground for his case is to attack the alternative. Before advocating the doctrine of the extended phenotype of an active germ-line replicator, therefore, I tried to undermine the reader’s confidence in the individual organism as the unit of adaptive benefit. But, now that we have discussed the extended phenotype itself, it is time to reopen the question of the organism’s existence and obvious importance in the hierarchy of life, and see whether we see it any clearer in the light of the extended phenotype. Given that life did not have to be packaged into discrete organisms, and allowing that organisms are not always totally discrete, why, nevertheless, did active germ-line replicators so conspicuously opt for the organismal way of doing things?

  14 Rediscovering the Organism

  Having devoted most of this book to playing down the importance of the individual organism, and to building up an alternative image of a turmoil of selfish replicators, battling for their own survival at the expense of their alleles, reaching unimpeded through individual body walls as though those walls were transparent, interacting with the world and with each other without regard to organismal boundaries, we now hesitate. There really is something pretty impressive about individual organisms. If we actually could wear spectacles that made bodies transparent and displayed only DNA, the distribution of DNA that we would see in the world would be overwhelmingly non-random. If cell nuclei glowed like stars and all else was invisible, multicellular bodies would show up as close-packed galaxies with cavernous space between them. A million billion glowing pinpricks move in unison with each other and out of step with all the members of other such galaxies.

  The organism is a physically discrete machine, usually walled off from other such machines. It has an internal organization, often of staggering complexity, and it displays to a high degree the quality that Julian Huxley (1912) labelled ‘individuality’—literally indivisibility—the quality of being sufficiently heterogeneous in form to be rendered non-functional if cut in half. Genetically speaking, too, the individual organism is usually a clearly definable unit, whose cells have the same genes as each other but different genes from the cells of other organisms. To an immunologist the individual organism has a special kind of ‘uniqueness’ (Medawar 1957), in that it will readily accept grafts from other parts of its own body, but not from other bodies. To the ethologist—and this is really an aspect of Huxley’s indivisibility—the organism is a unit of behavioural action in a much stronger sense than, say, two organisms, or a limb of an organism. The organism has one coordinated central nervous system. It takes ‘decisions’ (Dawkins & Dawkins 1973) as a unit. All its limbs conspire harmoniously together to achieve one end at a time. On those occasions when two or more organisms try to coordinate their efforts, say when a pride of lions cooperatively stalks prey, the feats of coordination among individuals are feeble compared with the intricate orchestration, with high spatial and temporal precision, of the hundreds of muscles within each individual. Even a starfish, whose tube-feet enjoy a measure of autonomy and may tear the animal in two if the circum-oral nerve ring has been surgically cut, looks like a single entity, and in nature behaves as if it had a single purpose.

  I am grateful to Dr J. P. Hailman for not withholding from me the sarcastic reaction of a colleague to the paper that was a brief trial-run for this book (Dawkins 1978a): ‘Richard Dawkins has rediscovered the organism.’ The irony was not lost on me, but there are wheels within wheels. We agree that there is something special about the individual organism as a level in the hierarchy of life, but it is not something obvious, to be accepted without question. My hope is that this book has revealed that there is a second side to the Necker Cube. But Necker Cubes have a habit of flipping back again to their original orientation, and then continuing to alternate. Whatever it is that is special about the individual organism as a unit of life, we should at least see it more clearly for having viewed the other side of the Necker Cube, for having trained our eyes to see through body walls into the world of replicators, and out and beyond to their extended phenotypes.

  So, what is it that is special about the individual organism? Given that life can be viewed as consisting of replicators with their extended phenotypic tools of survival, why in practice have replicators chosen to group themselves together by the hundreds of thousands in cells, and why have they influenced those cells to clone themselves by the millions of billions in organisms?

  One kind of answer is suggested by the logic of complex systems. Simon (1962) has written a stimulating essay on ‘The architecture of complexity’, which suggests (using a parable of two watchmakers called Tempus and Hora, which has become well known), a general functional reason why complex organization of any kind, biological or artificial, tends to be organized into nested hierarchies of repeated subunits. I have developed his argument in the ethological context, concluding that the evolution of statistically ‘improbable assemblies proceeds more rapidly if there is a succession of intermediate stable sub-assemblies. Since the argument can be applied to each sub-assembly, it follows that highly complex systems which exist in the world are likely to have a hierarchical architecture’ (Dawkins 1976b). In the present context the hierarchy consists of genes within cells and cells within organisms. Margulis (1981) makes a persuasive and fascinating case for an old idea that the hierarchy contains yet another intermediate level: eukaryotic ‘cells’ are themselves, in a sense, multi-cellular clusters, symbiotic unions of entities such as mitochondria, plastids and cilia, which are homologous to, and descended from, prokaryotic cells. I will not pursue the matter further here. Simon’s point is a very general one, and we need a more specific answer to the question of why replicators chose to organize their phenotypes into functional units, especially at the two levels of the cell and the multicellular organism.

  In order to ask questions about why the world is the way it is, we have to imagine how it might have been. We invent possible worlds in which life might have been organized differently, and ask what would have happened if it had been. What instructive alternatives to the way life is, then, can we imagine? First, in order to see why replicating molecules ganged up in cells, we imagine a world in which there are replicating molecules floating freely in the sea. There are different varieties of replicator, and they are competing with each other for space and for the chemical resources needed to build copies of themselves, but they are not grouped together in chromosomes or nuclei. Each solitary replicator exerts phenotypic power to make copies of itself, and selection favours those with the most effective phenotypic power. It is easy to believe that this form of life would not be evolutionarily stable. It would be invaded by mutant replicators that ‘gang up’. Certain replicators would have chemical effects that complement those of other replicators, complement them in the sense that when the two chemical effects are put together replication of both is facilitated (Model 2 in the previous chapter). I have already used the example of genes coding for enzymes that catalyse success
ive stages of a biochemical chain reaction. The same principle may be applied to larger groups of mutually complementary replicating molecules Indeed, earthly biochemistry suggests that the minimal unit of replication, except possibly in the food-rich environment of a total parasite, is about fifty cistrons (Margulis 1981). It makes no difference to the argument whether new genes arise by duplication of old ones and remain close together, or whether previously independent genes positively come together. We can still discuss the evolutionary stability of the state of being ‘ganged up’.

  Ganging up of genes together into cells, then, is easily understood, but why did cells gang together into multicellular clones? In this case it might be thought that we do not have to invent thought experiments, because unicellular, or acellular, organisms abound on our world. These, however, are all small, and it may be useful to imagine a possible world in which there exist large and complex unicellular or mononucleate organisms. Could there be a form of life in which one single set of genes, enthroned in a single central nucleus, directed the biochemistry of a macroscopic body with complex organs, either a single gigantic ‘cell’, or a multicellular body in which all but one of the cells lacked their own private copies of the genome? I think such a form of life could only exist if its embryology followed principles very different from those with which we are familiar. In the embryologies that we know, only a minority of genes are ‘turned on’ in any one type of differentiating tissue at any one time (Gurdon 1974). It is admittedly a weak argument at this stage, but if there were only one set of genes in the entire body, it is not easy to see how the appropriate gene products could be conveyed to the various parts of the differentiating body at the appropriate times.