The Extended Phenotype
A weird example of a driving replicator which is probably not a gene in the ordinary sense of the word is given by Werren, Skinner and Charnov (1981). They studied the parasitoid wasp Nasonia vitripennis, in which there is a variety of males called Dl, or ‘daughterless’. Wasps being haplodiploid, males pass their genes only to daughters: a male’s mate may have sons, but those sons are haploid and fatherless. When Dl males mate with females, they cause them to produce all male broods. Most of the sons of females mated to Dl males are themselves Dl males. Although no nuclear genes pass from father to son, therefore, the Dl factor somehow does pass from father to ‘son’. The Dl factor rapidly spreads, in exactly the same way as a driving Y chromosome would. It is not known what the Dl factor physically consists of. It is certainly not nuclear genetic material, and it is theoretically possible that it is not even composed of nucleic acid, although Werren et al. suspect that it probably is cytoplasmically borne nucleic acid. Theoretically, any kind of physical or chemical influence of a Dl male on his mate, which causes her to have Dl sons, would spread like a driving Y chromosome, and would qualify as an active germ-line replicator in the sense of Chapter 5. It is also an outlaw par excellence, for it spreads itself at the total expense of all the nuclear genes in the males that bear it.
Selfish sperm
With some exceptions, all the diploid cells of an organism are genetically identical, but the haploid gametes it produces are all different. Only one out of many sperms in an ejaculate can fertilize an egg, and there is therefore potential for competition among them. Any gene that found phenotypic expression when in the haploid state in a sperm cell could be favoured over its alleles if it improved the competitive ability of the sperm. Such a gene would not necessarily be sex-linked: it could be found on any chromosome. If it was sex-linked it would have the effect of biasing the sex ratio, and would be an outlaw. If it was on an autosome it would still qualify as an outlaw for the general kind of reason already given for any segregation distorter: ‘… if there were genes affecting sperm-cell function there would be competition among sperm cells, and a gene that improved the ability to fertilize would increase in the population. If such a gene happened to cause, say, malfunction of the liver, that would be just too bad; the gene would increase anyway, since selection for good health is much less effective than selection by competition among sperm cells’ (Crow 1979). There is, of course, no particular reason why a sperm competition gene should happen to cause malfunction of the liver but, as already pointed out, most mutations are deleterious, so some undesirable side effect is pretty likely.
Why does Crow assert that selection for good health is much less effective than selection by competition among sperm cells? There must inevitably be a quantitative trade-off involving the magnitude of the effect on health. But, that aside, and even allowing for the controversial possibility that only a minority of sperms are viable (Cohen 1977), the argument appears to have force because the competition between sperm cells in an ejaculate would seem to be so fierce.
A million million spermatozoa,
All of them alive:
Out of their cataclysm but one poor Noah
Dare hope to survive.
And of that billion minus one
Might have chanced to be
Shakespeare, another Newton, a new Donne—
But the One was Me.
Shame to have ousted your betters thus.
Taking ark while the others remained outside!
Better for all of us, froward Homunculus,
If you’d quietly died!
Aldous Huxley
One might imagine that a mutant gene that expressed itself when in the haploid genotype of a spermatozoon, causing increased competitive ability, say an improved swimming tail or the secretion of a spermicide to which the sperm itself was immune, would be immediately favoured by a selection pressure gigantic enough to outweigh all but the most catastrophic of deleterious side effects on the diploid body. But although it may be true that only one in hundreds of millions of sperms ‘dare hope to survive’, the calculation looks very different from the point of view of a single gene. If we forget linkage groups and brand-new mutations for a moment, however rare a gene may be in the gene-pool, if a given male has it in his diploid genotype, at least 50 per cent of his sperms must have it. If one sperm has received a gene giving it competitive advantage, 50 per cent of its rivals in the same ejaculate will have received the same gene. Only if the mutation has arisen de novo during the genesis of a single sperm will the selection pressure be astronomical in magnitude. Usually it will be a more modest selection pressure, not of millions to one but only two to one. If we take linkage effects into consideration the calculation is more complicated, and the selection pressure in favour of competitive sperms will increase somewhat.
In any case, it is a strong enough pressure for us to expect that, if genes expressed themselves when in the haploid genotype of the sperm, outlaws would be favoured, to the detriment of the rest of the genes in the diploid father’s genome. It seems, to say the least, fortunate that sperm phenotypes are, as a matter of fact, usually not under the control of their own haploid genotypes (Beatty & Gluecksohn-Waelsch 1972). Of course sperm phenotypes must be under some genetic control, and natural selection has doubtless worked on the genes controlling sperm phenotypes to perfect sperm adaptation. But those genes seem to express themselves when in the diploid genotype of the father, not when in the haploid genotype of the sperm. When in the sperm they are passively carried.
The passivity of their genotypes may be an immediate consequence of the lack of cytoplasm in spermatozoa: a gene cannot achieve phenotypic expression except via cytoplasm. This is a proximal explanation. But it is at least worth toying with reversing the proposition to obtain an ultimate functional explanation: sperms are made small, as an adaptation to prevent the phenotypic expression of the haploid genotype. On this hypothesis we are proposing an arms race between (haploid-expressed) genes for increased competitive ability among spermatozoa on the one hand, and genes expressing themselves when in the diploid genotype of the father on the other hand, causing sperms to become smaller and therefore unable to give phenotypic expression to their own haploid genotypes. This hypothesis does not explain why eggs are larger than sperms; it assumes the basic fact of anisogamy, and therefore does not aspire to be an alternative to theories of the origins of anisogamy (Parker 1978b; Maynard Smith 1978a; Alexander & Borgia 1979). Moreover not all sperms are small, as Sivinski (1980) reminds us in a most intriguing review. But the present explanation still deserves consideration as an ancillary to others. It is analogous to Hamilton’s (1967) explanation for the inertness of Y chromosomes, to which I have already alluded.
Green beards and armpits
Some of the outlaws I have been considering have been realistic, and are actually known to geneticists. Some of the suggested outlaws that I shall now come to are, frankly, pretty improbable. I make no apology for this. I see them as thought experiments. They play the same role in helping me to think straight about reality as imaginary trains travelling at nearly the speed of light do for physicists.
So, in this spirit of thought experiment, imagine a gene on a Y chromosome which makes its possessor kill his daughters and feed them to his sons. This is clearly a behavioural version of a driving Y-chromosome effect. If it arose it would tend to spread for the same reason, and it would be an outlaw in the same sense that its phenotypic effect would be detrimental to the rest of the male’s genes. Modifiers, on any chromosome other than the Y chromosome, which tended to reduce the phenotypic effect of the daughter-killing gene would be favoured over their alleles. In a sense the outlaw gene is using the sex of the male’s children as a convenient label for the presence or absence of itself: all sons are labelled as definite possessors of the gene, all daughters as definite non-possessors of it.
A similar argument can be made for X chromosomes. Hamilton (1972, p. 201) pointed out that in normal diploid species a gene
on an X chromosome in the homogametic sex has three-quarters of a chance of being identical by descent with a gene in a sibling of the homogametic sex. Thus the ‘X-chromosome relatedness’ of human sisters is as high as the overall relatedness of hymenopteran sisters, and higher than the overall relatedness of human sisters. Hamilton went so far as to wonder whether an X-chromosome effect might account for the fact that helpers at the nest in birds seem, usually, to be elder brothers, rather than sisters, of the nestlings (the male sex is homogametic in birds). He noted that the X chromosome in birds accounts for some 10 per cent of the whole genome, and that it is therefore not too improbable that the genetic basis for brotherly care might lie on the X chromosome. If so, brotherly care might be favoured by the same kind of selection pressure as Hamilton had earlier suggested for sisterly care in Hymenoptera. Perhaps significantly, Syren and Luyckx (1977) point out that in some termites, the only non-haplodiploid group to have achieved full eusociality, ‘approximately half the genome is maintained as a linkage group with the sex chromosome’ (Lacy 1980).
Wickler (1977), commenting on Whitney’s (1976) rediscovery of Hamilton’s X-chromosome idea, suggests that Y-chromosome effects are potentially even more powerful than X-chromosome effects, but Y chromosomes as a rule don’t make up such a high proportion of the genome. In any case ‘sex-linked altruism’ must be discriminating: individuals acting under the influence of their sex chromosomes should tend to show favouritism towards close relatives of the same sex rather than of the opposite sex. Genes for sex-blind sibling care would not be outlaws.
The value of the outlawed sex-chromosome thought experiment does not lie in its plausibility—which, like Hamilton, I do not rate highly—but in the fact that it focuses our attention on the importance of this discrimination. The sex of another individual is used as a label to identify it as a member of a class of individuals about whose genetics something is known. In the ordinary theory of kin selection, relatedness (or rather some proximate correlate of relatedness such as presence in one’s own nest) is used as a label indicating higher than average probability of sharing a gene. From the point of view of a gene on a Y chromosome, the sex of a sibling is a label which signifies the difference between certainty of sharing the gene and certainty of not sharing it.
Note, by the way, the ineptness of notions of individual fitness, or even inclusive fitness as ordinarily understood, at dealing with situations like this. The normal calculation of inclusive fitness makes use of a coefficient of relationship which is some measure of the probability that a pair of relatives will share a particular gene, identical by descent. This is a good approximation provided the genes concerned have no better way of ‘recognizing’ copies of themselves in other individuals. If a gene is on a sex chromosome and can use the sex of relatives as a label, however, its best ‘estimate’ of the probability that a relative shares a copy of itself will be a better estimate than that provided by the coefficient of relationship. In its most general form, the principle of genes appearing to ‘recognize’ copies of themselves in other individuals has been dubbed the ‘green-beard effect’ (Dawkins 1976a, p. 96, following Hamilton 1964b, p. 25). Green-beard or ‘recognition alleles’ have been described in the literature as outlaws (Alexander & Borgia 1978; Alexander 1980), and they should therefore be discussed in this chapter, even though, as we shall see, their status as outlaws needs careful examination (Ridley & Grafen 1981).
The green-beard effect reduces the principle of ‘gene self-recognition’ to its bare essentials in an unrealistically hypothetical, but nevertheless instructive, way. A gene is postulated which has two pleiotropic effects. One effect is to confer a conspicuous label, the ‘green beard’. The other effect is to confer a tendency to behave altruistically towards bearers of the label. Such a gene, if it ever arose, could easily be favoured by natural selection, although it would be vulnerable to a mutant arising which conferred the label without the altruism.
Genes are not conscious little devils, able to recognize copies of themselves in other individuals and to act accordingly. The only way for the green-beard effect to arise is by incidental pleiotropy. A mutation must arise which just happens to have two complementary effects: the label or ‘green beard’, and the tendency to behave altruistically towards labelled individuals. I have always thought such a fortuitous conjunction of pleiotropic effects too good to be true. Hamilton also noted the idea’s inherent implausibility, but he went on ‘… exactly the same a priori objections might be made to the evolution of assortative mating which manifestly has evolved, probably many times independently and despite its obscure advantages’ (Hamilton 1964b, p. 25). It is worth briefly examining this comparison with assortative mating, which for present purposes I shall take to mean the tendency of individuals to prefer to mate with individuals that genetically resemble them.
Why is it that the green-beard effect seems so much more far-fetched than assortative mating? It is not just that assortative mating is positively known to occur. I suggest another reason. This is that when we think of assortative mating we implicitly assume self-inspection as a means of facilitating the effect. If black individuals prefer to mate with black individuals, and white with white, we do not find this hard to believe because we tacitly assume that individuals perceive their own colour. Each individual, whatever his colour, is assumed to be obeying the same rule: inspect yourself (or members of your own family) and choose a mate of the same colour. This principle does not stretch our credulity by demanding that two specific effects—colour and behavioural preference—are controlled pleiotropically by the same gene. If there is a general advantage in mating with similar partners, natural selection will favour the self-inspection rule regardless of the exact nature of the recognition character used. It does not have to be skin colour. Any conspicuous and variable character would work with the identical behavioural rule. No far-fetched pleiotropism need be postulated.
Well then, will the same kind of mechanism work for the green-beard effect? Might animals obey a behavioural rule of the form: ‘Inspect yourself and behave altruistically towards other individuals that resemble you’? The answer is that they might, but this would not be a true example of the green-beard effect. Instead, I call it the ‘armpit effect’. In the paradigm hypothetical example, the animal is supposed to smell its own armpits, and behave altruistically towards other individuals with a similar smell. (The reason for choosing an olfactory name is that police dog trials have shown that dogs presented with handkerchiefs held under human armpits can distinguish the sweats of any two individual humans except identical twins (Kalmus 1955). This suggests that there is an enormously variable richness of genetic labelling in molecules of sweat. In the light of the result with identical twins, one feels inclined to bet that police dogs could be trained to sniff out the coefficient of relationship between pairs of humans and that, for instance, they could be trained to track down a criminal if given a sniff of his brother. Be that as it may, ‘armpit effect’ is here being used as a general name for any case of an animal inspecting himself, or a known close relative, and discriminating in favour of other individuals with a similar smell or with some other perceived similarity.)
The essential difference between the green-beard effect and the armpit self-inspection effect is as follows. The armpit self-inspection behavioural rule will lead to the detection of other individuals that are similar in some respect, perhaps in many respects, but it will not specifically lead to the detection of individuals that possess copies of the gene mediating the behavioural rule itself. The armpit rule might provide an admirable means of detecting true kin from non-kin, or of detecting whether a brother was a full brother or only a half brother. This could be very important, and it might provide the basis for selection in favour of self-inspecting behaviour, but the selection would be conventional, familiar kin selection. The self-inspection rule would be functioning simply as a kin-recognition device, analogous to a rule like: ‘Behave altruistically towards individuals that gre
w up in your own nest.’
The green-beard effect is quite different. Here the important thing is that a gene (or close linkage group) programs the recognition specifically of copies of itself. The green-beard effect is not a mechanism for the recognition of kin. Rather, kin recognition and ‘green-beard’ recognition are alternative ways in which genes could behave as if discriminating in favour of copies of themselves.
To return to Hamilton’s comparison with assortative mating, we can see that it does not really provide good grounds for optimism over the plausibility of the green-beard effect. Assortative mating is much more likely to involve self-inspection. If, for whatever reason, it is an advantage in general for like to mate with like, selection would favour an armpit type of behavioural rule: Inspect yourself, and choose a mate that resembles you. This will achieve the desired result—an optimal balance between outbreeding and inbreeding (Bateson 1983) or whatever the advantage may be—regardless of the exact nature of the characteristics by which individuals differ.
Assortative mating is not the only analogy Hamilton might have chosen. Another one is the case of cryptic moths choosing to sit on a background that matches their own colour. Kettlewell (1955) gave dark-coloured carbonaria and light-coloured typical morphs of the peppered moth Biston betularia the opportunity to sit on dark- or light-coloured backgrounds. There was a statistically significant tendency for moths to choose a background matching their own body colour. This could be due to pleiotropism (or genes for colour being closely linked to genes for background choice). If this were so, as Sargent (1969a) believes, it might, by analogy, reduce our scepticism about the inherent plausibility of the green-beard effect. Kettlewell believed, however, that the moths achieved the matching by the simpler mechanism of ‘contrast conflict’. He suggested that a moth could see a small portion of its own body, and that it moved around until the observed contrast between its own body and the background reached a minimum. It is easy to believe that natural selection might favour the genetic basis for such a contrast-minimizing behavioural rule, because it would work automatically in conjunction with any colour, including a newly mutated colour. It is of course analogous to the ‘armpit self-inspection’ effect, and is plausible for the same reason.