Brief Candle in the Dark

by Richard Dawkins

  Having based his first book on the assumption that ‘if adaptations are to be treated as “for the good of” something, that something is the gene’, Dawkins was now mounting an attempt ‘to free the selfish gene from the individual organism which has been its conceptual prison’.

  One of the jailers of this Bastille was Bill Hamilton, who found himself cast in the role of the revolutionary deemed insufficiently radical by his disciple. Although Dawkins never wavered in his admiration for Hamilton, he felt that the idea of inclusive fitness was an obstacle to seeing biological facts through the gene’s eye. Inclusive fitness was about the selection of genes, but it complicated the issue by trying to fit into biology’s existing frame of reference. ‘Before Hamilton’s revolution, our world was peopled by individual organisms working single-mindedly to keep themselves alive and to have children. In those days it was natural to measure success in this undertaking at the level of the individual organism. Hamilton changed all that but unfortunately, instead of following his ideas through to their logical conclusion and sweeping the organism from its pedestal . . . he exerted his genius in devising a means of rescuing the individual.’

  John Maynard Smith, in my ‘Web of Stories’ interview with him in 1997 (see page 234),1 said very much the same thing.

  Extending the phenotype

  If St Peter were to twist my arm at the Pearly Gates and force me to come up with an answer to the question of how, if at all, I might justify having occupied a little space on this Earth and exchanged a fraction of its air, the best I could do would be to point to The Extended Phenotype. It isn’t really a new hypothesis which might be right or might be wrong, to be tested by experiment or observation. It’s more like a new way of looking at what is already familiar: a way of looking at biology that helps things fall into place and make sense. I suppose it’s a bit like ‘Today is the first day of the rest of your life.’ Trite, necessarily true, definitely not the sort of statement for which you would go out and seek supporting evidence: but nevertheless we recognize it as a truth that changes the way we see things. Obvious as it is, it can be worth bothering to say to somebody as a way of influencing the way they do things. That’s how I see the extended phenotype. But instead of being summarized in a pithy aphorism it needs explaining. One way to say it is that it amounts to a questioning of the assumed centrality of the ‘vehicle’.

  The phenotypic influence of a gene had been thought to end at the body wall of the individual containing that gene. Genes influence bodies via the processes of embryonic development. A mutant version of a gene subtly changes some detail of the shape of a swift’s wing. As a consequence the bird flies slightly faster for the same expenditure of energy, and this makes it slightly more likely to survive and therefore pass on the very same gene to future generations. Multiply the effect over many swifts and over many generations, and the result will be that the mutant gene comes to dominate the population, at the expense of alternative alleles.1

  All genes exert their immediate effects in ways that are buried deep in the internal biochemistry of the individual body, and these are usually invisible to all but specialist scientists. But the phenotypic effect which we finally recognize as the adaptation or survival tool is usually external and visible to the naked eye – as in the case of the swift’s wing. There is a cascade of internally buried causes and effects, often beginning with the synthesis of a protein precisely coded by the DNA sequence. We could arbitrarily identify a ‘phenotype’ at any point along the cascade – the protein itself, its immediate effect as a catalyst of cellular biochemistry, the consequent effect on the behaviour of cells interacting in tissues, many more downstream consequences – before we hit something visible on the outside of the animal: a webbier foot on a duck, perhaps, or a larger wing on a wasp, or a more stilted courtship gesture in an albatross. All these are properly called phenotypic effects of a gene.

  What I added in The Extended Phenotype was the thought that the sequence of causes and effects doesn’t have to stop at the body wall. Take, for instance, the set of tubes built by Jane Brockmann’s mud-daubing wasps, Trypoxylon politum (see pages 75–6 and picture section). Each tube is like an organ of the body: an external uterus for the nurturing of young. It has been shaped to a useful purpose by natural selection, in just the same kind of way as the wasp’s wings or legs or antennae. The genes exerted their influence via building behaviour, before that via a carefully rigged nervous system, before that by cell growth programs in embryology, before that by biochemical influences on cell growth, before that by protein synthesis under the influence of genes in the nuclei of cells. As with legs and wings, genes that influence for the better the shape and size of the (mud) ‘organ’ have been favoured by natural selection. And, as with legs and wings, genes, in interaction with lots of other genes, influence the shape and size of the mud ‘organ’ by indirect processes, beginning with an effect on cellular chemistry and running through a cascade of intermediate causes to the final phenotype.

  Yes, that’s right, phenotype. That’s the point I am making. This ‘phenotype’ is made of mud rather than living cells – hence extended phenotype – but is no less a true phenotype. The mud in the stream before the wasp fetches it is not phenotype. It becomes phenotype when it is shaped to a biological purpose, in this case the purpose of protecting a growing larva. It is phenotype because the shape and other properties of the tube have evolved, over many generations of increasing perfection. So there must be genes for tube length, genes for tube diameter, genes for thickness of tube wall, genes for distance between partitions down the length of the tube.

  How do I know those genes exist? I don’t. Not in the sense that anyone has ever done a genetic study of the phenotypic traits I have just listed. But I am confident that, if such a genetic study were undertaken – and it certainly could be – all those phenotypic traits would be found to vary under genetic control. Why so confident? Because the tubes constructed by the mud-dauber have obviously been shaped to their well-designed form by natural selection, and the logic of natural selection implies the involvement of genes. How else than by favouring certain genes over other genes could natural selection have shaped mud tubes towards increasing suitability to their function of protecting larvae? Of course, to repeat, the genes affect the tubes only indirectly, via the building behaviour of wasps. And before that in the causal chain via wasp nervous systems. And before that via cellular processes making wasp nervous systems. But all phenotypic effects are indirect, anyway. The influence of genes on mud tubes is indirect in exactly the same kind of way as the influence of genes on wings, legs and antennae. And the extended phenotype we are looking at may not be the last point on the cascade of causes and effects. Anything caused by it, ‘downstream’ in the cascade, could be seen as a further extended phenotype, provided only that genes responsible for it have been favoured as a result by natural selection.

  The picture reproduced in this book shows variation in tube colour. Are there, then, genes for tube colour? Maybe. Here I am less confident, but only because it is not obvious that tube colour has been favoured by natural selection. It is possible that some colours are better than others, and possible that genes make wasps fussy about the colour of mud they gather. On the other hand it could well be that the wasps are indifferent to colour of mud, simply gathering it from whatever is available in a local stream, which might happen to be light brown, dark brown or reddish brown. Why doesn’t the same ‘indifference’ argument apply to length of tube or thickness of wall? It might, but it seems unlikely in those cases. It is easy to see that the tube wall could be too thin for purpose (offering inadequate protection for the larva, or even falling apart). And it could be too thick (using too much mud, requiring more time-consuming trips to the stream). It is hard to see how thickness of tube wall could not be subject to natural selection. I personally suspect that tube colour is also subject to natural selection (some colours might be easier for predators to see), but it is entirely plausible that
the time-saving advantages of fetching mud from the nearest stream regardless of colour (rather than searching far and wide for a stream with better-coloured mud) could be paramount.

  These are hypothetical details for illustration only. The point is that the logic of natural selection (its choice of genes by virtue of their phenotypic effects) compels us to recognize that such functional phenotypes are not limited to the individual body, the ‘vehicle’. Animal artefacts provide the clearest and simplest example, and here I benefited from my close friendship as a graduate student at Oxford, where we shared a flat, with Michael Hansell. Mike is now the world’s leading authority on animal artefacts and author of several books on the subject, including the beautiful Built by Animals, which cleverly uses the subject of artefacts as a platform to talk about many aspects of animal behaviour more generally. The Extended Phenotype has a whole chapter on animal artefacts, from caddis larva houses to bird nests to termite mounds to beaver dams. Even the lake created by a beaver dam is properly regarded as (extended) phenotypic expression of beaver genes, probably the largest phenotype in the world.

  If The Extended Phenotype had restricted its scope to artefacts such as Jane Brockmann’s mud-dauber tubes or Mike Hansell’s caddis mobile homes, I wouldn’t have bothered to say (and the publishers wouldn’t have bothered to print on the paperback cover) ‘If you never read anything else of mine, please at least read this.’ But the extension goes further. The chapter on animal artefacts softens the reader up for the more radical ideas of parasite manipulation of hosts and ‘action at a distance’. A fluke lives inside its snail shell, as a caddis larva lives inside its stone house. The fluke doesn’t ‘build’ its shell as the caddis builds its house. But if the fluke could find a way to modify the snail shell to its own advantage, and if we can be sure the modification was favoured by natural selection, neo-Darwinian logic forces us to recognize a fluke gene ‘for’ snail-shell characteristics. The logic of the extended phenotype, if you buy the analogy with the caddis house (and how could you not?), concludes that fluke genomes contain genes ‘for’ snail phenotypes, at least in the same sense as they contain genes ‘for’ fluke phenotypes.

  The shell of a snail is its protective dwelling, just as the stone house is that of the caddis. We would not be surprised to find that parasitic infection debilitates a snail, for example causing the shell to be thinner than it should be and the snail more vulnerable. But what shall we say if a snail’s shell is thicker when it has a parasite? Such is indeed the case for snails parasitized by some flukes. Is the snail better protected as a result of some parasitic influence? Is the fluke doing the snail an altruistic good turn? Is the snail actually better off for housing a parasite?

  In one sense probably yes, but not a good Darwinian sense. Here’s what I think. Everything about an animal is a compromise between conflicting pressures. Just as a shell can be too thin for the snail’s own good, so it can also be too thick. How so? It’s a question of economics, as so often in evolutionary theory. The wherewithal to make a shell, calcium for example, is expensive. As we saw in the chapter on digger wasp economics, too much investment in one part of the body’s economy has to be paid for in the form of too little investment in some other part. A snail that invests too heavily in its shell must skimp on something else, and will be less successful than a rival snail that invests a bit less in the shell (and therefore more somewhere else). We may presume that the average shell thickness of unparasitized snails is an optimum. When a fluke forces its snail to thicken its shell, it is pushing that snail away from the snail optimum and in the direction of a different and more costly optimum, which is the fluke optimum.

  Is it plausible that the fluke optimum should be thicker than the snail optimum? Yes, actually very plausible. Any animal has to balance the needs of individual survival against those of reproduction. Peacocks and peahens sit at different optima along the sex–survival continuum. Hens ‘care’ more about survival; cocks more about reproduction, even at the cost of a shorter life. This is because, not having to lay large, costly eggs, a cock can potentially pack far more reproduction into a short life than a hen can. Most cocks are less successful in terms of passing on genes than the average hen, but a few ‘elite’ cocks are far more successful than the average hen and are so even if they die young. Cocks tend to inherit their characteristics from ancestors belonging to the elite minority who died young after a bonanza of reproduction. So a tendency to shift the bodily economy away from the individual survival optimum and towards the reproduction optimum is favoured in cocks.

  The snail ‘cares’ about its reproduction: the end towards which its survival is just the means. The fluke doesn’t ‘care’ at all about the reproductive success of the particular snail which is its present domicile. As between snail survival and snail reproduction, the fluke genes reach a different compromise from the snail genes. The snail genes ‘want’ to save some resources for snail reproduction and so compromise on survival. The fluke genes ‘want’ the snail to put all its resources into preserving the protective house in which the fluke rides – snail survival, and to hell with snail reproduction. Things would be different if flukes were passed directly from parent snail to offspring snail: in that case the flukes too would ‘care’ about snail reproduction, not just snail survival. This is one of the most important lessons from The Extended Phenotype. Parasites become gentler to their hosts, more symbiotic, to the extent that their offspring infect the offspring of their particular hosts, rather than getting passed on to random members of the host species.

  Parasite genes, then, can have ‘extended’ effects on host phenotypes. The parasitology literature is filled with fascinating, even macabre, accounts of hosts whose habits are manipulated to facilitate the life-cycles of the parasites that ride inside them, and my chapter ‘Host phenotypes of parasite genes’ lists lots of examples. It’s almost as though the parasite is pulling the host’s puppet strings, and the logic of natural selection compels us to carry that image down to the level of the parasite’s genes. In 2012, David Hughes and his colleagues published a splendid book on Host Manipulation by Parasites,1 which takes a very ‘extended phenotype’ view of the facts.

  Action at a distance

  But parasites don’t necessarily ride in (or on) their hosts. Empty space supervenes between cuckoo and host, but the cuckoo is no less a parasite, and the distorted parental behaviour of the foster parent is no less a naturally selected adaptation of the cuckoo nestling. By what black arts of seduction does the monstrous cuckoo nestling cajole the nervous system of the tiny wren? We don’t know, but it is surely the product of an evolutionary arms race (see page 340). In that arms race, natural selection of cuckoos implies selection of cuckoo genes ‘for’ manipulating hosts. And that’s just another way of saying there are cuckoo genes ‘for’ host behaviour, cuckoo genes whose phenotypic effects are manifested as complaisant changes in host behaviour. So the extended phenotype extends beyond the body wall, beyond the stone house enclosing the caddis, beyond the snail shell enclosing the fluke, right outside the body and across the space between cuckoo and host – a space in which something is transmitted from the one and picked up by the other. This is the meaning of ‘action at a distance’, which is the title of the penultimate chapter of The Extended Phenotype. And it doesn’t apply only to parasites and hosts.

  If a physiologist wants to bring a female canary into reproductive condition, increase the size of her functional ovary and cause her to start nest-building and other reproductive behaviour patterns, there are various things he can do. He can inject her with gonadotropins or oestrogens. He can use electric light to increase the day-length that she experiences. Or, most interestingly from our point of view, he can play her a tape recording of male canary song. It apparently has to be canary song; budgerigar song will not do, although budgerigar song has a similar effect on female budgerigars.

  That quotation is from an earlier chapter of The Extended Phenotype, the one on ‘Arms races and ma
nipulation’, but it serves equally to illustrate action at a distance. Genes in male canaries have been naturally selected for their extended phenotypic effect – at a distance – on female canaries.

  This theme was foreshadowed in a 1978 paper which I wrote jointly with my friend John Krebs (see page 340), called ‘Animal signals: information or manipulation?’ That paper could be credited with importing the ‘selfish gene’ revolution into the study of animal signals such as bird song. Hitherto, under the influence of Niko Tinbergen, Mike Cullen, Desmond Morris and other ethologists of the Tinbergen–Lorenz school, animal signals had been treated in a cooperative spirit: both parties to the communication would benefit from a flow of accurate information between them (‘I am informing you, for our mutual benefit, that I am a male of your species, possessing a territory, and I am ready to mate’). John Krebs and I turned that on its head by regarding the sender of the signal as manipulating the receiver, as if flooding her nervous system with a drug, or as if stimulating her brain electrically with micro-electrodes. I put the point with calculated bathos in The Extended Phenotype:

  The snort of a pig-frog Rana grylio may affect another pig-frog as the nightingale affected Keats, or the skylark Shelley.

  And much later, in Unweaving the Rainbow (whose title was paraphrased from Keats) I said something like it again, after quoting the ‘Ode to a Nightingale’: