The Extended Phenotype

by Richard Dawkins

  And how far afield can the phenotype extend? Is there any limit to action at a distance, a sharp cut-off, an inverse square law? The farthest action at a distance I can think of is a matter of several miles, the distance separating the extreme margins of a beaver lake from the genes for whose survival it is an adaptation. If beaver lakes could fossilize, we would presumably see a trend towards increased lake size if we arranged the fossils in chronological order. The increase in size was doubtless an adaptation produced by natural selection, in which case we have to infer that the evolutionary trend came about by allele replacement. In the terms of the extended phenotype, alleles for larger lakes replaced alleles for smaller lakes. In the same terms, beavers can be said to carry within themselves genes whose phenotypic expression extends many miles away from the genes themselves.

  Why not hundreds of miles, thousands of miles? Could an ectoparasite which stayed behind in England inject a swallow with a drug which affected that swallow’s behaviour on its arrival in Africa, and could the consequence in Africa be usefully regarded as the phenotypic expression of parasite genes in England? The logic of the extended phenotype might seem to favour the idea, but I think in practice it is unlikely, at least if we are talking about phenotypic expression as adaptation. I see a crucial practical difference from the case of the beaver dam. A gene in a beaver which, when compared with its alleles, causes a larger lake to come into existence, can directly benefit itself by means of its lake. Alleles causing smaller lakes are less likely to survive, as a direct consequence of their smaller phenotypes. It is, however, hard to see how a gene in an English ectoparasite could benefit itself, at the expense of its alleles in England, as a direct result of its African phenotypic expression. Africa is probably too far away for the consequences of the gene’s action to feed back and affect the welfare of the gene itself.

  By the same token, beyond a certain size of beaver lakes, it would become hard to regard further increases in size as adaptations. The reason is that, beyond a certain size, other beavers than the builders of the dam are just as likely to benefit from each increase in size as the dam-builders themselves. A big lake benefits all the beavers in the area, whether they created it or whether they just found it and exploited it. Similarly, even if a gene in an English animal could exert some phenotypic effect on Africa which directly benefited the survival of the gene’s ‘own’ animal, other English animals of the same kind would almost certainly benefit just as much. We must not forget that natural selection is all about relative success.

  It is admittedly possible to speak of a gene as having a particular phenotypic expression, even when its own survival is not influenced by that phenotypic expression. In this sense, then, a gene in England might indeed have phenotypic expression in a remote continent where its consequences do not feed back upon its own success in the English gene-pool. But I have already argued that in the world of the extended phenotype this is not a profitable way of speaking. I used the example of footprints in mud as phenotypic expression of genes for foot shape, and I gave my intention of using extended phenotype language only when the character concerned might conceivably influence, positively or negatively, the replication success of the gene or genes concerned.

  It is not plausible, but it helps to make the point if I construct a thought experiment in which it would indeed be useful to speak of a gene as having phenotypic expression extending to another continent. Swallows return, each year, to exactly the same nest. It follows that an ectoparasite, waiting dormant in a swallow’s nest in England, can expect to see the very same swallow both before and after the swallow’s journey to Africa. If the parasite could engineer some change in the swallow’s behaviour in Africa, it might indeed reap the consequences on the swallow’s return to England. Suppose, for instance, that the parasite needs a rare trace element which is not found in England, but which occurs in the fat of a particular African fly. Swallows normally have no preference for this fly, but the parasite, by injecting a drug into the swallow before it leaves for Africa, so changes its dietary preferences as to increase the likelihood of its eating specimens of this fly. When the swallow returns to England, its body contains enough of the trace element to benefit the individual parasite (or its children) waiting in the original nest, benefit them at the expense of rivals within the parasite species. Only in circumstances such as these would I wish to speak of a gene in one continent as having phenotypic expression in another continent.

  There is a risk, which I had better forestall, that such talk of adaptation on a global scale may call to the reader’s mind the fashionable image of the ecological ‘web’, of which the most extreme manifestation is the ‘Gaia’ hypothesis of Lovelock (1979). My web of interlocking extended phenotypic influences bears a superficial resemblance to the webs of mutual dependence and symbiosis that bulk so largely in the pop-ecology literature (e.g. The Ecologist) and in Lovelock’s book. The comparison could hardly be more misleading. Since Lovelock’s Gaia hypothesis has been enthusiastically espoused by no less a scientist then Margulis (1981), and extravagantly praised by Mellanby (1979) as the work of a genius, it cannot be ignored, and I must digress in order categorically to disclaim any connection with the extended phenotype.

  Lovelock rightly regards homeostatic self-regulation as one of the characteristic activities of living organisms, and this leads him to the daring hypothesis that the whole Earth is equivalent to a single living organism. Whereas Thomas’s (1974) likening of the world to a living cell can be accepted as a throwaway poetic line, Lovelock clearly takes his Earth/organism comparison seriously enough to devote a whole book to it. He really means it. His explanations of the nature of the atmosphere are representative of his ideas. The Earth has much more oxygen than is typical of comparable planets. It has long been widely suggested that green plants are probably almost entirely responsible for this high oxygen content. Most people would regard oxygen production as a byproduct of plant activity, and a fortunate one for those of us who need to breathe oxygen (presumably, too, we have been selected to breathe oxygen partly because there is so much of it about). Lovelock goes further, and regards oxygen production by plants as an adaptation on the part of the Earth/organism or ‘Gaia’ (named after the Greek Earth goddess): plants produce oxygen because it benefits life as a whole.

  He uses the same kind of argument for other gases that occur in small amounts:

  What, then, is the purpose of methane and how does it relate to oxygen? One obvious function is to maintain the integrity of the anaerobic zones of its origin [p. 73].

  Another puzzling atmospheric gas is nitrous oxide … We may be sure that the efficient biosphere is unlikely to squander the energy required in making this odd gas unless it has some useful function. Two possible uses come to mind … [p. 74].

  Another nitrogenous gas made in large volumes in the soil and the sea and released to the air is ammonia … As with methane, the biosphere uses a great deal of energy in producing ammonia, which is now entirely of biological origin. Its function is almost certainly to control the acidity of the environment … [p. 77].

  The fatal flaw in Lovelock’s hypothesis would have instantly occurred to him if he had wondered about the level of natural selection process which would be required in order to produce the Earth’s supposed adaptations. Homeostatic adaptations in individual bodies evolve because individuals with improved homeostatic apparatus pass on their genes more effectively than individuals with inferior homeostatic apparatuses. For the analogy to apply strictly, there would have to have been a set of rival Gaias, presumably on different planets. Biospheres which did not develop efficient homeostatic regulation of their planetary atmospheres tended to go extinct. The Universe would have to be full of dead planets whose homeostatic regulation systems had failed, with, dotted around, a handful of successful, well-regulated planets of which Earth is one. Even this improbable scenario is not sufficient to lead to the evolution of planetary adaptations of the kind Lovelock proposes. In addition we w
ould have to postulate some kind of reproduction, whereby successful planets spawned copies of their life forms on new planets.

  I am not, of course, suggesting that Lovelock believes it happened like that. He would surely consider the idea of interplanetary selection as ludicrous as I do. Obviously he simply did not see his hypothesis as entailing the hidden assumptions that I think it entails. He might dispute that it does entail those assumptions, and maintain that Gaia could evolve her global adaptations by the ordinary processes of Darwinian selection acting within the one planet. I very much doubt that a model of such a selection process could be made to work: it would have all the notorious difficulties of ‘group selection’. For instance, if plants are supposed to make oxygen for the good of the biosphere, imagine a mutant plant which saved itself the costs of oxygen manufacture. Obviously it would outreproduce its more public-spirited colleagues, and genes for public-spiritedness would soon disappear. It is no use protesting that oxygen manufacture need not have costs: if it did not have costs, the most parsimonious explanation of oxygen production in plants would be the one the scientific world accepts anyway, that oxygen is a byproduct of something the plants do for their own selfish good. I do not deny that somebody may, one day, produce a workable model of the evolution of Gaia (possibly along the lines of ‘Model 2’ below), although I personally doubt it. But if Lovelock has such a model in mind he does not mention it. Indeed, he gives no indication that there is a difficult problem here.

  The Gaia hypothesis is an extreme form of what, for old times’ sake although it is now rather unfair, I shall continue to call the ‘BBC Theorem’. The British Broadcasting Corporation is rightly praised for the excellence of its nature photography, and it usually strings these admirable visual images together with a serious commentary. Things are changing now, but for years the dominant message of these commentaries was one that had been elevated almost to the status of a religion by pop ‘ecology’. There was something called the ‘balance of nature’, an exquisitely fashioned machine in which plants, herbivores, carnivores, parasites, and scavengers each played their appointed role for the good of all. The only thing that threatened this delicate ecological china shop was the insensitive bull of human progress, the bulldozer of …, etc. The world needs the patient, toiling dung beetles and other scavengers, but for whose selfless efforts as the sanitary workers of the world …, etc. Herbivores need their predators, but for whom their populations would soar out of control and threaten them with extinction, just as man’s population will unless …, etc. The BBC Theorem is often expressed in terms of the poetry of webs and networks. The whole world is a fine-meshed network of interrelationships, a web of connections which it has taken thousands of years to build up, and woe betide mankind if we tear it down …, etc.

  There is, no doubt, much merit in the moralistic exhortations that seem to flow from the BBC Theorem, but that does not mean its theoretical basis is sound. Its weakness is the one I have already exposed in the Gaia hypothesis. A network of relationships there may be, but it is made up of small, self-interested components. Entities that pay the costs of furthering the wellbeing of the ecosystem as a whole will tend to reproduce themselves less successfully than rivals that exploit their public-spirited colleagues, and contribute nothing to the general welfare. Hardin (1968) summed the problem up in his memorable phrase ‘The tragedy of the commons’, and more recently (Hardin 1978) in the aphorism, ‘Nice guys finish last’.

  I have dealt with the BBC Theorem and the Gaia hypothesis, because of the danger that my own language of the extended phenotype and action at a distance may sound like some of the more exuberantly extended networks and webs of the TV ‘ecologists’. To emphasize the difference, then, let me borrow the rhetoric of webs and networks, but use it in a very different way, to explain the idea of the extended phenotype and of genetic action at a distance.

  Loci in germ-line chromosomes are hotly contested pieces of property. The contestants are allelomorphic replicators. Most of the replicators in the world have won their place in it by defeating all available alternative alleles. The weapons with which they won, and the weapons with which their rivals lost, are their respective phenotypic consequences. These phenotypic consequences are conventionally thought of as being restricted to a small field around the replicator itself, its boundaries being defined by the body wall of the individual organism in whose cells the replicator sits. But the nature of the causal influence of gene on phenotype is such that it makes no sense to think of the field of influence as being limited in this arbitrary way, any more than it makes sense to think of it as limited to intracellular biochemistry. We must think of each replicator as the centre of a field of influence on the world at large. Causal influence radiates out from the replicator, but its power does not decay with distance according to any simple mathematical law. It travels wherever it can, far or near, along available avenues, avenues of intracellular biochemistry, of intercellular chemical and physical interaction, of gross bodily form and physiology. Through a variety of physical and chemical media it radiates out beyond the individual body to touch objects in the world outside, inanimate artefacts and even other living organisms.

  Just as every gene is the centre of a radiating field of influence on the world, so every phenotypic character is the centre of converging influences from many genes, both within and outside the body of the individual organism. The whole biosphere—recognize the superficial affinity with the BBC Theorem—the whole world of plant and animal matter is criss-crossed with an intricate network of fields of genetic influence, a web of phenotypic power. I can almost hear the television commentary: ‘Imagine being shrunk to the size of a mitochondrion stationed at a convenient vantage point outside the nuclear membrane of a human zygote. Watch the molecules of messenger RNA as they stream, by the millions, out into the cytoplasm on their errands of phenotypic power-play. Now grow to the size of a cell in the developing limb-bud of a chick embryo. Feel the wafts of chemical inducers as they roll down the gentle slopes of their axial gradients! Now grow again to your full size, and stand in the middle of a wood, at dawn in spring. Birdsong surges round you. Male syrinxes pour out sound, and all round the wood female ovaries swell. Here the influence travels in the form of pressure waves in the open air, rather than molecules in cytoplasm, but the principle is the same. At all three levels of the Lilliputian/Brobdingnagian thought experiment, you are privileged to stand in the midst of uncountable interlocking fields of replicator power.’

  The reader will gather that it was the message of the BBC Theorem that I wanted to criticize, not its rhetoric! Nevertheless, a more restrained brand of eloquence is often more effective. A master of restrained eloquence in biological writing is Ernst Mayr. His chapter on ‘The unity of the genotype’ (Mayr 1963) is often held up to me in conversation as deeply antithetical to my replicator-based viewpoint. Since, on the contrary, I find myself enthusiastically endorsing almost every word of that chapter, something must have been misunderstood, somewhere.

  Much the same could be said of Wright’s (1980) equally eloquent article on ‘Genic and organismic selection’, which purports to be a repudiation of the genic selection view that I hold, yet almost none of which I find myself disagreeing with. I think this is a valuable paper, even though its ostensible purpose is to attack the view that ‘with respect to natural selection … it is the gene, not the individual or group, that is the unit’. Wright concludes that ‘The likelihood of organismic, instead of merely genic, selection goes far toward meeting one of the most serious objections to the theory of natural selection encountered by Darwin.’ He attributes the ‘genic selection’ view to Williams, Maynard Smith and me, and traces it back to R. A. Fisher, I think correctly. All of which might lead him to be somewhat bemused by the following accolade from Medawar (1981): ‘The most important single innovation in the modern synthesis was however the new conception that a population that was deemed to undergo evolution could best be thought of as a population of fundame
ntal replicating units—of genes—rather than as a population of individual animals or of cells. Sewall Wright … was a principal innovator in this new way of thinking—a priority for which R. A. Fisher, an important but lesser figure, never forgave him …’

  In the rest of this chapter, I hope to show that the version of ‘genic selectionism’ that can be attacked as naively atomistic and reductionistic is a straw man; that it is not the view that I am advocating; and that if genes are correctly understood as being selected for their capacity to cooperate with other genes in the gene-pool, we arrive at a theory of genic selection which Wright and Mayr will recognize as fully compatible with their own views. Not only compatible but, I would claim, a truer and a clearer expression of their views. I shall quote key passages from the summary of Mayr’s chapter (pp. 295–296), showing how they may be adapted to the world of the extended phenotype.

  The phenotype is the product of the harmonious interaction of all genes. The genotype is a ‘physiological team’ in which a gene can make a maximum contribution to fitness by elaborating its chemical ‘gene product’ in the needed quantity and at the time when it is needed in development [Mayr 1963].