I believe it will be admitted that had Hamilton used his own ‘intelligent-gene’ thought experiment when calculating these coefficients of relationship, instead of thinking in terms of individuals as agents maximizing something, he would have got the right answer the first time. If these errors had been simple miscalculations it would obviously be pedantic to discuss them, once their original author had pointed them out. But they were not miscalculations, they were based on a highly instructive conceptual error. The same is true of the numbered ‘Misunderstandings of Kin Selection’ that I quoted before.
I have tried to show in this chapter that the concept of fitness as a technical term is a confusing one. It is confusing because it can lead to admitted error, as in the case of Hamilton’s original calculation of haplodiploid coefficients of relationship, and as in the case of several of my ‘12 Misunderstandings of Kin Selection’. It is confusing because it can lead philosophers to think the whole theory of natural selection is a tautology. And it is confusing even to biologists because it has been used in at least five different senses, many of which have been mistaken for at least one of the others.
Emerson, as we have seen, confused fitness[3] with fitness[1]. I now give an example of a confusion of fitness[3] with fitness[2]. Wilson (1975) provides a useful glossary of terms needed by sociobiologists. Under ‘fitness’ he refers us to ‘genetic fitness’. We turn to ‘genetic fitness’ and find it defined as ‘The contribution to the next generation of one genotype in a population relative to the contributions of other genotypes.’ Evidently ‘fitness’ is being used in the sense of the population geneticist’s fitness[2]. But then, if we look up ‘inclusive fitness’ in the glossary we find: ‘The sum of an individual’s own fitness plus all its influence on fitness in its relatives other than direct descendants …’ Here, ‘the individual’s own fitness’ must be ‘classical’ fitness[3] (since it is applied to individuals), not the genotypic fitness (fitness[2]) which is the only ‘fitness’ defined in the glossary. The glossary is, then, incomplete, apparently because of a confusion between fitness of a genotype at a locus (fitness[2]) and reproductive success of an individual (fitness[3]).
As if my fivefold list were not confusing enough already, it may need extending. For reasons concerned with an interest in biological ‘progress’, Thoday (1953) seeks the ‘fitness’ of a long-term lineage, defined as the probability that the lineage will continue for a very long time such as 108 generations, and contributed to by such ‘biotic’ factors (Williams 1966) as ‘genetic flexibility’. Thoday’s fitness does not correspond to any of my list of five. Then again, the fitness[2] of population geneticists is admirably clear and useful, but many population geneticists are, for reasons best known to themselves, very interested in another quantity which is called the mean fitness of a population. Within the general concept of ‘individual fitness’, Brown (1975; Brown & Brown 1981) wishes to make a distinction between ‘direct fitness’ and ‘indirect fitness’. Direct fitness is the same as what I am calling fitness[3]. Indirect fitness can be characterized as something like fitness[4] minus fitness[3], i.e. the component of inclusive fitness that results from the reproduction of collateral relatives as opposed to direct descendants (I presume grandchildren count in the direct component, though the decision is arbitrary). Brown himself is clear about the meaning of the terms, but I believe they have considerable power to confuse. For instance, they appear to lend weight to the view (not held by Brown, but held by a distressing number of other authors, e.g. Grant 1978, and several writers on ‘helpers at the nest’ in birds) that there is something unparsimonious about ‘kin selection’ (the ‘indirect component’) as compared with ‘individual selection’ (the ‘direct component’), a view which I have criticized sufficiently before (Dawkins 1976a, 1978a, 1979a).
The reader may have been bewildered and irritated at my drawn out list of five or more separate meanings of fitness. I have found this a painful chapter to write, and I am aware that it will not have been easy to read. It may be the last resort of a poor writer to blame his subject matter, but I really do believe it is the concept of fitness itself which is responsible for the agony in this case. The population geneticists’ fitness[2] aside, the concept of fitness as applied to individual organisms has become forced and contrived. 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 individual organism from its pedestal as notional agent of maximization, he exerted his genius in devising a means of rescuing the individual. He could have persisted in saying: gene survival is what matters; let us examine what a gene would have to do in order to propagate copies of itself. Instead he, in effect, said: gene survival is what matters; what is the minimum change we have to make to our old view of what individuals must do, in order that we may cling on to our idea of the individual as the unit of action? The result—inclusive fitness—was technically correct, but complicated and easy to misunderstand. I shall avoid mentioning fitness again in this book, which I trust will make for easier reading. The next three chapters develop the theory of the extended phenotype itself.
11 The Genetical Evolution of Animal Artefacts
What do we really mean by the phenotypic effect of a gene? A smattering of molecular biology may suggest one kind of answer. Each gene codes for the synthesis of one protein chain. In a proximal sense that protein is its phenotypic effect. More distal effects like eye colour or behaviour are, in their turn, effects of the protein functioning as an enzyme. Such a simple account does not, however, bear much searching analysis. The ‘effect’ of any would-be cause can be given meaning only in terms of a comparison, even if only an implied comparison, with at least one alternative cause. It is strictly incomplete to speak of blue eyes as ‘the effect’ of a given gene G1. If we say such a thing, we really imply the potential existence of at least one alternative allele, call it G2, and at least one alternative phenotype, P2, in this case, say, brown eyes. Implicitly we are making a statement about a relation between a pair of genes {G1, G2} and a pair of distinguishable phenotypes {P1, P2}, in an environment which either is constant or varies in a non-systematic way so that its contribution randomizes out. ‘Environment’, in that last clause, is taken to include all the genes at other loci that must be present in order for P1 or P2 to be expressed. Our statement is that there is a statistical tendency for individuals with G1 to be more likely than individuals with G2 to show P1 (rather than P2). Of course there is no need to demand that P1 should always be associated with G1, nor that G1 should always lead to P1: in the real world outside logic textbooks, the simple concepts of ‘necessary’ and ‘sufficient’ must usually be replaced by statistical equivalents.
Such an insistence that phenotypes are not caused by genes, but only phenotypic differences caused by gene differences (Jensen 1961; Hinde 1975) may seem to weaken the concept of genetic determination to the point where it ceases to be interesting. This is far from the case, at least if the subject of our interest is natural selection, because natural selection too is concerned with differences (Chapter 2). Natural selection is the process by which some alleles out-propagate their alternatives, and the instruments by which they achieve this are their phenotypic effects. It follows that phenotypic effects can always be thought of as relative to alternative phenotypic effects.
It is customary to speak as if differences always mean differences between individual bodies or other discrete ‘vehicles’. The purpose of the next three chapters is to show that we can emancipate the concept of the phenotypic difference from that of the discrete vehicle altogether, and this is the meaning of the title ‘extended phenotype’. I shall show that the ordinary logic of genetic terminology leads inevitably to the conclusion that genes can be said to ha
ve extended phenotypic effects, effects which need not be expressed at the level of any particular vehicle. Following an earlier paper (Dawkins 1978a) I shall take a step-by-step approach to the extended phenotype, beginning with conventional examples of ‘ordinary’ phenotypic effects and gradually extending the concept of the phenotype outwards so that the continuity is easy to accept. The idea of the genetic determination of animal artefacts is a didactically useful intermediate example, and this will be the main topic of this chapter.
But first, consider a gene A whose immediate molecular effect is the synthesis of a black protein which directly colours the skin of an animal black. Then the gene’s only proximal effect, in the molecular biologist’s simple sense, is the synthesis of this one black protein. But is A a gene ‘for being black’? The point I want to make is that, as a matter of definition, that depends on how the population varies. Assume that A has an allele A′, which fails to synthesize the black pigment, so that individuals homozygous for A′ tend to be white. In this case A is truly a gene ‘for’ being black, in the sense in which I wish to use the phrase. But it may alternatively be that all the variation in skin colour that actually occurs in the population is due to variation at a quite different locus, B. B’s immediate biochemical effect is the synthesis of a protein which is not a black pigment, but which acts as an enzyme, one of whose indirect effects (in comparison with its allele B′), at some distant remove, is the facilitation of the synthesis by A of the black pigment in skin cells.
To be sure, A, the gene whose protein product is the black pigment, is necessary in order for an individual to be black: so are thousands of other genes, if only because they are necessary to make the individual exist at all. But I shall not call A a gene for blackness unless some of the variation in the population is due to lack of A. If all individuals, without fail, have A, and the only reason individuals are not black is that they have B′ rather than B, we shall say that B, but not A, is a gene for blackness. If there is variation at both loci affecting blackness, we shall refer to both A and B as genes for blackness. The point that is relevant here is that both A and B are potentially entitled to be called genes for blackness, depending on the alternatives that exist in the population. The fact that the causal chain linking A to the production of the black pigment molecule is short, while that for B is long and tortuous, is irrelevant. Most gene effects seen by whole animal biologists, and all those seen by ethologists, are long and tortuous.
A geneticist colleague has argued that there are virtually no behaviour-genetic traits, because all those so-far discovered have turned out to be ‘byproducts’ of more fundamental morphological or physiological effects. But what on earth does he think any genetic trait is, morphological, physiological or behavioural, if not a ‘byproduct’ of something more fundamental? If we think the matter through we find that all genetic effects are ‘byproducts’ except protein molecules.
Returning to the black skin example, it is even possible that the chain of causation linking a gene such as B to its black-skinned phenotype might involve a behavioural link. Suppose that A can synthesize black pigment only in the presence of sunlight, and suppose that B works by making individuals seek sunlight, in comparison with B′ which makes them seek shade. B individuals will then tend to be blacker than B′ individuals, because they spend more time in the sun. B is still, by existing terminological convention, a gene ‘for blackness’, no less than it would be if its causal chain involved internal biochemistry only, rather than an ‘external’ behavioural loop. Indeed, a geneticist in the pure sense of the word need not care about the detailed pathway from gene to phenotypic effect. Strictly speaking, a geneticist who concerns himself with these interesting matters is temporarily wearing the hat of an embryologist. The pure geneticist is concerned with end products, and in particular with differences between alleles in their effects on end products. Natural selection’s concerns are precisely the same, for natural selection ‘works on outcomes’ (Lehrman 1970). The interim conclusion is that we are already accustomed to phenotypic effects being attached to their genes by long and devious chains of causal connection, therefore further extensions of the concept of phenotype should not overstretch our credulity. This chapter takes the first step towards such further extension, by looking at animal artefacts as examples of the phenotypic expression of genes.
The fascinating subject of animal artefacts is reviewed by Hansell (1984). He shows that artefacts provide useful case studies for several principles of general ethological importance. This chapter uses the example of artefacts in the service of explaining another principle, that of the extended phenotype. Consider a hypothetical species of caddis-fly whose larvae build houses out of stones which they select from those available on the bottom of the stream. We might observe that the population contains two rather distinct colours of house, dark and light. By breeding experiments we establish that the characters ‘dark house’ and ‘light house’ breed true in some simple Mendelian fashion, say with dark house dominant to light house. In principle it ought to be possible to discover, by analysing recombination data, where the genes for house colour sit on the chromosomes. This is, of course, hypothetical. I do not know of any genetic work on caddis houses, and it would be difficult to do because adults are difficult to breed in captivity (M. H. Hansell, personal communication). But my point is that, if the practical difficulties could be overcome, nobody would be very surprised if house colour did turn out to be a simple Mendelian character in accordance with my thought experiment. (Actually, colour is a slightly unfortunate example to have chosen, since caddis vision is poor and they almost certainly ignore visual cues in choosing stones. Rather than use a more realistic example like stone shape (Hansell), I stay with colour for the sake of the analogy with the black pigment discussed above.)
The interesting sequel is this. House colour is determined by the colour of the stones chosen from the stream bed by the larva, not by the biochemical synthesis of a black pigment. The genes determining house colour must work via the behavioural mechanism that chooses stones, perhaps via the eyes. So much would be agreed by any ethologist. All that this chapter adds is a logical point: once we have accepted that there are genes for building behaviour, the rules of existing terminology imply that the artefact itself should be treated as part of the phenotypic expression of genes in the animal. The stones are outside the body of the organism, yet logically such a gene is a gene ‘for’ house colour, in exactly as strong a sense as the hypothetical gene B was for skin colour. And B was indeed a gene for skin colour, even though it worked by mediating sun-seeking behaviour, in exactly as strong a sense as a gene ‘for’ albinism is called a gene for skin colour. The logic is identical in all three cases. We have taken the first step in extending the concept of a gene’s phenotypic effect outside the individual body. It was not a difficult step to take, because we had already softened up our resistance by realizing that even normal ‘internal’ phenotypic effects may lie at the end of long, ramified, and indirect, causal chains. Let us now step out a little further.
The house of a caddis is strictly not a part of its cellular body, but it does fit snugly round the body. If the body is regarded as a gene vehicle, or survival machine, it is easy to see the stone house as a kind of extra protective wall, in a functional sense the outer part of the vehicle. It just happens to be made of stone rather than chitin. Now consider a spider sitting at the centre of her web. If she is regarded as a gene vehicle, her web is not a part of that vehicle in quite the same obvious sense as a caddis house, since when she turns round the web does not turn with her. But the distinction is clearly a frivolous one. In a very real sense her web is a temporary functional extension of her body, a huge extension of the effective catchment area of her predatory organs.
Once again, I know of no genetic analysis of spider web morphology, but there is nothing difficult in principle about imagining such an analysis. It is known that individual spiders have consistent idiosyncrasies which are repeated
in web after web. One female Zygiella-x-notata, for instance, was seen to build more than 100 webs, all lacking a particular concentric ring (Witt, Read & Peakall 1968). Nobody familiar with the literature on behaviour genetics (e.g. Manning 1971) would be surprised if the observed idiosyncrasies of individual spiders turned out to have a genetic basis. Indeed, our belief that spiders’ webs have evolved their efficient shape through natural selection necessarily commits us to a belief that, at least in the past, web variation must have been under genetic influence (Chapter 2). As in the case of the caddis houses, the genes must have worked via building behaviour, before that in embryonic development perhaps via neuroanatomy, before that perhaps via cell membrane biochemistry. By whatever embryological routes the genes may work in detail, the small extra step from behaviour to web is not any more difficult to conceive of than the many causal steps which preceded the behavioural effect, and which lie buried in the labyrinth of neuroembryology.
Nobody has any trouble understanding the idea of genetic control of morphological differences. Nowadays few people have trouble understanding that there is, in principle, no difference between genetic control of morphology and genetic control of behaviour, and we are unlikely to be misled by unfortunate statements such as ‘Strictly speaking, it is the brain (rather than the behaviour) that is genetically inherited’ (Pugh in press). The point here is, of course, that if there is any sense in which the brain is inherited, behaviour may be inherited in exactly the same sense. If we object to calling behaviour inherited, as some do on tenable grounds, then we must, to be consistent, object to calling brains inherited too. And if we do decide to allow that both morphology and behaviour may be inherited, we cannot reasonably at the same time object to calling caddis house colour and spider web shape inherited. The extra step from behaviour to extended phenotype, in this case the stone house or the web, is as conceptually negligible as the step from morphology to behaviour.