The Extended Phenotype

by Richard Dawkins

  Now, consider a gene in the D gene-pool which exerts an effect on D bodies to make them refuse to mate with H. Such a gene should be favoured by natural selection over its H-tolerating allele, since the latter will tend to end up in H bodies of the next generation, and will be discarded at meiosis. G genes are not discarded at meiosis, and they will tend to be selected if they influence H bodies so that they overcome the reluctance of D individuals to mate with them. We should therefore see an arms race between G genes acting on H bodies, and D genes acting on D bodies. In those respective bodies, both sets of genes are germ-line replicators. But what about D genes acting on H bodies? They should have just as powerful an influence over H phenotypes as G genes, since they constitute exactly half of the H genome. Naively, we might expect them to carry their arms race against G genes over into the H bodies which they share. But in H bodies, those D genes are in the same position as ants that have been taken as slaves. Any adaptation that they mediate in the H body cannot be passed on to the next generation; for the gametes produced by the H individual, regardless of how his developing phenotype, and indeed his survival, may have been influenced by D genes, are strictly G gametes. Just as would-be slave ants can be selected to resist being taken into captivity while they are still in their home nest, but cannot be selected to subvert the slavemaking nest once they are in it, so D genes can be selected to influence D bodies so that they resist being incorporated in H genomes in the first place, but once so incorporated they are no longer under selection, even though they can still have phenotypic effects. They lose the arms race because they are a dead-end. A similar argument could be made for fish of the hybrid ‘species’ Poeciliopsis monacha-occidentalis (Maynard Smith 1978a).

  The inability of slaves to evolve counter-adaptations was originally invoked by Trivers and Hare (1976), in their theory of the sex-ratio arms race in social Hymenoptera. This is one of the best known of recent discussions of a particular arms race, and it is worth considering further. Elaborating on ideas of Fisher (1930a) and Hamilton (1972), Trivers and Hare reasoned that the evolutionarily stable sex ratio in ant species with one singly mated queen per nest cannot be simply predicted. If the queen is assumed to have all power over the sex of reproductive offspring (young queens and males), the stable ratio of economic investment in male and female reproductives is 1:1. If, on the other hand, non-laying workers are assumed to hold all power over investment in young, the stable ratio will be 3:1 in favour of females, ultimately because of the haplodiploid genetic system. There is, therefore, a potential conflict between queen and workers. Trivers and Hare reviewed the, admittedly imperfect, available data, and reported a good average fit to the 3:1 prediction, from which they concluded that they had found evidence for worker power winning the battle against queen power. It was a clever attempt to use real data to test a hypothesis of a kind that is often criticized as untestable, but like other innovative first attempts it is easy to find fault with it. Alexander and Sherman (1977) complained about Trivers and Hare’s handling of the data, and also suggested an alternative explanation for the female-biased sex ratio common in ants. Their explanation (‘local mate competition’), like that of Trivers and Hare, was originally derived from Hamilton, in this case his paper on extraordinary sex ratios (1967).

  This controversy has had the good effect of stimulating further work. Especially illuminating in the context of arms races and manipulation is the paper by Charnov (1978), which is concerned with the origins of eusociality, and which introduces a potentially important version of the life/dinner principle. His argument works for diploid as well as haplodiploid organisms, and I shall consider the diploid case first. Consider a mother whose elder children have still not left the nest when the next brood hatches. When the time comes for them to leave the nest and begin their own reproduction, the young have the option, instead, of staying behind and helping to rear their young siblings. As is now well known, all other things being equal such a fledgling should be genetically indifferent between rearing offspring and rearing full siblings (Hamilton 1964a,b). But suppose the old mother could exert any manipulative power over the decision of her elder children: would she ‘prefer’ that they leave and rear families of their own, or stay and rear her next brood? Obviously that they should stay and rear her next brood, since grandchildren are half as valuable to her as children. (The argument as it stands is incomplete. If she manipulated all her children for the whole of her life into rearing yet more non-producing child labourers, her germ-line would peter out. We must assume that she manipulates some offspring of the same genetic type into developing into reproductives and others into developing as workers.) Selection will, then, favour such manipulative tendencies in parents.

  Normally, when we postulate selection in favour of manipulation we dutifully pay lip service to counter-selection on the victim to resist manipulation. The beauty of Charnov’s point is that in this case there will be no counter-selection. The ‘arms race’ is a walk-over because one side, so to speak, doesn’t even try. The offspring being manipulated are, as we have already seen, indifferent to whether they rear young siblings or offspring of their own (again assuming all other things equal). Therefore, although we may postulate reverse manipulation by offspring of parents, this is bound, at least in the simple example visualized by Charnov, to be outweighed by parental manipulation of offspring. This is an asymmetry to be added to the list of parental advantages offered by Alexander (1974), but I find it more generally convincing than any others on the list.

  At first sight it might appear that Charnov’s argument does not apply to haplodiploid animals, and that would be a pity since most social insects are haplodiploid. But this view is mistaken. Charnov himself shows this for the special assumption that the population has an unbiased sex ratio, in which case even in haplodiploid species females are indifferent between rearing siblings (r = the average of ¾ and ¼) and rearing offspring (r = ½). But Craig (1980) and Grafen (in preparation) independently show that Charnov did not even need to assume an unbiased sex ratio. The potential worker is still indifferent between rearing siblings and rearing offspring at any conceivable population sex ratio. Thus suppose the population sex ratio is female-biased, even suppose it conforms to Trivers and Hare’s predicted 3:1. Since the worker is more closely related to her sister than to her brother or her offspring of either sex, it might seem that she would ‘prefer’ to rear siblings over offspring given such a female-biased sex ratio: is she not gaining mostly valuable sisters (plus only a few relatively worthless brothers) when she opts for siblings? But this reasoning neglects the relatively great reproductive value of males in such a population as a consequence of their rarity. The worker may not be closely related to each of her brothers, but if males are rare in the population as a whole each one of those brothers is correspondingly highly likely to be an ancestor of future generations.

  The mathematics confirm that Charnov’s conclusion is even more general than he suggested. In both diploid and haplodiploid species, at any population sex ratio, an individual female is theoretically indifferent whether she herself rears offspring or younger siblings. She is not, however, indifferent whether her offspring rear their own children or their siblings: she prefers them to rear their siblings (her offspring) over their offspring (her grandchildren). Therefore if there is any question of manipulation in this situation, parental manipulation of offspring is more likely than offspring manipulation of parents.

  It might appear that Charnov’s, Craig’s and Grafen’s conclusions radically contradict those of Trivers and Hare on sex ratios in social Hymenoptera. The statement that, at any sex ratio, a female hymenopteran is indifferent between rearing siblings and rearing offspring, sounds tantamount to saying that she is also indifferent to what the sex ratio in her nest is. But this is not so. It is still true that, given the assumption of worker control over investment in male and female reproductives, the resulting evolutionarily stable sex ratio will not be necessarily the same as the evolutiona
rily stable sex ratio given queen control. In this sense a worker is not indifferent to the sex ratio: she may well work to shift the sex ratio away from what the queen is ‘trying’ to achieve.

  Trivers and Hare’s analysis of the exact nature of the conflict between queen and workers over the sex ratio can be extended in ways that further illuminate the concept of manipulation (e.g. Oster & Wilson 1978). The following account is derived from Grafen (in preparation). I shall not anticipate his conclusions in detail, but wish to emphasize one principle which is explicit in his analysis as well as implicit in that of Trivers and Hare. The question is not ‘Has the “best” sex ratio been successfully achieved?’ On the contrary, we make a working assumption that natural selection has produced a result, given some constraints, and then ask what those constraints are (see Chapter 3). In the present case we follow Trivers and Hare in recognizing that the evolutionarily stable sex ratio depends crucially upon which parties to the arms race have practical power, but we recognize a wider range of possible dispositions of power than they did. In effect, Trivers and Hare deduced the consequences of two alternative assumptions about practical power; firstly the assumption that the queen exerts all the power, and secondly the assumption that the workers exert all the power. But many other possible assumptions could be made, and each gives rise to a different prediction of the evolutionarily stable sex ratio. In other parts of their paper, indeed, Trivers and Hare consider some of these, for instance the assumption that workers are able to lay their own male eggs.

  Grafen, like Bulmer and Taylor (in preparation), has explored the consequences of assuming that power is divided as follows: the queen has absolute power over the sex of the eggs that she lays; the workers have absolute power over feeding the larvae. The workers can thus determine how many of the available female eggs shall develop into queens and how many into workers. They have the power to starve the young of one sex or the other, but they have to work within the constraint of what the queen gives them in the way of eggs. Queens have the power to lay eggs in any sex ratio they choose, including withholding, totally, eggs of one sex or the other. But, once laid, those eggs are at the mercy of the workers. A queen might, for instance, play the strategy (in the game theory sense) of laying only male eggs in a given year. Reluctant as we might expect them to be, the workers have no option but to rear their brothers. The queen, in this case, can preempt certain worker strategies, such as ‘preferentially feed sisters’, simply because she ‘plays’ first. But there are other things workers can do.

  Using game theory, Grafen shows that only certain queen strategies are evolutionarily stable replies to particular worker strategies, and only certain worker strategies are evolutionarily stable replies to particular queen strategies. The interesting question is, What are the evolutionarily stable combinations of worker and queen strategies? It turns out that there is more than one answer, and there can be as many as three evolutionarily stable states for a given set of parameters. Grafen’s particular conclusions are not my concern here, although I will remark that they are interestingly ‘counterintuitive’. What is my concern is that the evolutionarily stable state of the model population depends upon the assumptions we make about power. Trivers and Hare contrasted two possible absolute assumptions (absolute worker power versus absolute queen power). Grafen investigated one plausible division of power (queens have power over eggs, workers over larval feeding). But, as I have already noted, numerous other assumptions about power could be made. Each assumption generates different predictions about evolutionarily stable sex ratios, and tests of the predictions can therefore be regarded as providing evidence about the disposition of power in the nest.

  For instance, we might focus our research attention on the exact moment when a queen ‘decides’ whether to fertilize a given egg or not. It is plausible to assume that, since the event takes place within the queen’s own body, that particular decision is likely to have been selected to benefit the queen’s genes. Plausible it may be, but it is precisely this kind of assumption that the doctrine of the extended phenotype is going to call in question. For the moment, we simply note the possibility that workers might manipulate the queen’s nervous system, by pheromonal or other means, so as to subvert her behaviour in their genetic interests. Similarly, it is worker nerves and muscles that are immediately responsible for feeding the larvae, but we are not, therefore, necessarily entitled to assume that worker limbs move only in the interests of worker genes. As is well known, there is massive pheromonal traffic flow from queen to workers, and it is easy to imagine powerful manipulation of worker behaviour by queens. The point is that each assumption about power which we might make yields a testable prediction about sex ratios, and it is for this insight that we have to thank Trivers and Hare, not for the particular model whose predictions they happened to test.

  It is even conceivable, in some Hymenoptera, that males might exert power. Brockmann (1980) is making an intensive study of mud-daubing wasps Trypoxylon politum. These are ‘solitary’ (as opposed to truly social) wasps, but they are not always totally alone. As in other sphecids, each female builds her own nest (in this case out of mud), provisions it with paralysed prey (spiders), lays one egg on the prey, then seals up the nest and begins the cycle again. In many Hymenoptera, the female carries a lifetime’s supply of sperm from one brief period of insemination early in life. T. politum females, however, copulate frequently throughout adult life. Males haunt female nests, losing no opportunity to copulate with the female on each of her returns to the nest. A male may spend hours at a time sitting passively in the nest, probably helping to guard it against parasites, and fighting with other males who attempt to enter. Unlike most male Hymenoptera then, the male T. politum is present at the scene of the action. Might he not, therefore, be potentially in a position to influence the sex ratio, in the same kind of way as has been postulated for worker ants?

  If males did exert power, what would we expect the consequences to be? Since a male passes all his genes on to his daughters, and none to his mate’s sons, genes tending to make males favour daughters over sons would be favoured. If males exerted total power, completely determining the sex ratio of their mates’ offspring, the consequence would be odd. No males would be born in the first generation of male power. As a result, in the following year all eggs laid would be unfertilized and therefore male. The population would therefore oscillate violently and then go extinct (Hamilton 1967). If males exerted a limited amount of power, less drastic consequences would follow, the situation being formally analogous to that of the ‘driving X chromosome’ in the normal diploid genetic system (Chapter 8). In any case a male hymenopteran, if he found himself in a position to influence the sex ratio of his mate’s children, would be expected to try to do so in a female direction. He might do this by trying to influence his mate’s decision whether to release sperm from her spermatheca. It is not obvious how he might actually do this, but it is known that honeybee queens take longer over laying a female egg than a male one, perhaps using the extra time to achieve fertilization. It would be interesting to try experimentally interrupting a queen in the middle of egg-laying, to see if the delay increased the chance of a female egg emerging.

  Do male T. politum show any behaviour that we might suspect of being an attempt at such manipulation, for example do they behave as if trying to prolong egg-laying? Brockmann describes a curious behaviour pattern called ‘holding’. This is seen alternating with copulation during the final minutes before the female lays her egg. In addition to brief copulations throughout the provisioning phase of the nest, the final egg-laying and sealing up of the nest is heralded by a prolonged bout of repeated copulations, which lasts many minutes. The female goes head first into the vertical, organ pipe-shaped mud nest, and pushes her head up into the cluster of paralysed spiders lodged in the top of the nest. Her abdomen is facing the entrance at the bottom of the nest, and in this position the male copulates with her. The female then turns round so that she is head
downwards facing out of the nest, and probes the spiders with the tip of her abdomen, as if about to lay an egg. The male meanwhile ‘holds’ her head in his forelegs for about half a minute, grabs her antennae and pulls her downwards away from the spiders. She then turns around and they copulate again. She again turns to probe the spiders with her abdomen, the male again holds her head and drags her down. The whole cycle repeats some half dozen times. Finally, after one especially long bout of head-holding, the female lays her egg.

  Once the egg is laid, its sex is determined. We have already considered the hypothetical possibility of worker ants manipulating their mother’s nervous system, forcing her to change her fertilizing decision in their genetic interest. Brockmann’s suggestion is that male T. politum might attempt similar subversion, and that the head-holding and dragging behaviour may be a manifestation of their manipulation technique. When the male seizes the female’s antennae and drags her away from the spiders which she is probing with her abdomen, is he forcing a postponement of egg-laying as a means of increasing the chance of the egg’s being fertilized in the oviduct? The plausibility of this suggestion might depend on exactly where the egg is in the female’s body during the time of holding. Or is he, as Dr W. D. Hamilton has suggested to us, blackmailing the female by, in effect, threatening to bite her head off unless she postpones egg-laying until after further copulation? Perhaps he gains by repeated copulations, simply by flooding the female’s internal passages with his sperm, thereby raising the chance that the egg will encounter a sperm without one being deliberately released from the spermatheca by the female. Clearly these are just suggestions for further research, and Brockmann, together with Grafen and others, is following them up. Preliminary indications, I understand, do not support the hypothesis that males actually succeed in exerting power over the sex ratio.