The Extended Phenotype

by Richard Dawkins

  We only have to accept the general plausibility of such asymmetries in order to answer the question raised by our discussion of manipulation. We agreed that if one organism could get away with manipulating the nervous system of another, and exploiting its muscle power, selection would favour such manipulation. But we were brought up short by the reflection that selection would also favour resistance to being manipulated. Should we, then, really expect to see effective manipulation in nature? The life/dinner principle, and other such principles as the rare-enemy effect, provide us with an answer. If the individual manipulator has more to lose by failing to manipulate than the individual victim has to lose by failing to resist manipulation, we should expect to see successful manipulation in nature. We should expect to see animals working in the interests of other animals’ genes.

  Brood parasitism provides what may be the most striking example. A parent bird such as a reed warbler sweeps a copious flow of food from a large catchment area into the narrow funnel of its nest. There is a living to be made by any creature that evolves the necessary adaptations to insert itself in the funnel and intercept the flow. That is what cuckoos and other brood parasites have done. But the reed warbler is not an uncomplaining cornucopia of free food. It is an active, complex machine, with sense organs, muscles and a brain. The brood parasite must not only have its body inserted in the host’s nest. It must also infiltrate the defences of the host’s nervous system, and its ports of entry are the host’s sense organs. The cuckoo uses key stimuli to unlock the host’s machinery of parental care and subvert it.

  The advantages of the brood-parasite way of life are so manifest that today we are taken aback to find Hamilton and Orians needing, in 1965, to defend the proposition that it has been favoured by natural selection, against theories of ‘degenerative breakdown’ of normal breeding behaviour. Hamilton and Orians went on to provide a satisfying discussion of the probable evolutionary origins of brood parasitism, the preadaptations which preceded its evolution, and the adaptations which have accompanied its evolution.

  One of these adaptations is egg mimicry. The perfection of egg mimicry, in at least some ‘gentes’ of cuckoo, shows that foster parents are potentially capable of keen-eyed discrimination against interlopers. This only underlines the mystery of why cuckoo hosts seem to be so poor at discriminating against cuckoo nestlings. Hamilton and Orians (1965) express the problem vividly: ‘Young brown-headed cowbirds and European Cuckoos, as they reach their maximum dimension, dwarf their foster parents. Consider the ludicrous sight of a tiny Garden Warbler … [Latin name leaves exact species intended unclear] standing atop a cuckoo to reach the mouth of the gaping parasite. Why does not the Garden Warbler take the adaptive measure of abandoning the nestling prematurely, especially when to the human observer it is so clearly identifiable?’ When a parent is a small fraction of the size of the nestling it is feeding, the most rudimentary eyesight should suffice to show that something has gone seriously wrong with the normal parental process. Yet, at the same time, the existence of egg mimicry shows that hosts are capable of fastidious discrimination, making use of keen eyesight. How are we to explain this paradox (see also Zahavi 1979)?

  One fact that helps to reduce the mystery is that there must be stronger selection pressure on hosts to discriminate against cuckoos at the egg stage than against cuckoos at the nestling stage, simply because eggs occur earlier. The benefit of detecting a cuckoo egg is the potential gaining of an entire breeding cycle in the future. The benefit of detecting a nearly fledged cuckoo is the saving of only a few days, at a time when it may be too late to breed again anyway. Another mitigating circumstance in the case of Cuculus canorus (Lack 1968) is that the host’s own young is usually not there for simultaneous comparison, having been tipped out by the baby cuckoo. It is well known that discrimination is easier if there is a model actually present for comparison.

  Various authors have invoked the ‘supernormal stimulus’, in one form or another. Thus Lack remarks (p. 88) that ‘the young cuckoo, with its huge gape and loud begging call, has evidently evolved in exaggerated form the stimuli which elicit the feeding response of parent passerine birds. So much is this so that there are many records of adult passerine birds feeding a fledged young C. canorus raised by a different host species; this, like lipstick in the courtship of mankind, demonstrates successful exploitation by means of a “super-stimulus”.’ Wickler (1968) makes a similar point, quoting Heinroth as having referred to foster parents as behaving like ‘addicts’, and to the cuckoo nestling as a ‘vice of its foster parents’. As it stands, this kind of suggestion will strike many critics as unsatisfying, because it immediately prompts a question at least as big as the one it answers. Why doesn’t selection eliminate from the host species the tendency to be ‘addicted’ to ‘supernormal stimuli’?

  This, of course, is where the arms race concept comes in again. When a human behaves in a way that is manifestly bad for him, for instance when he continually takes poison, we may explain his behaviour in at least two ways. He may not realize that the substance that he is drinking is poison, so closely does it resemble a genuinely nutritous substance. This corresponds to the host bird’s being fooled by the cuckoo’s egg mimicry. Or he may be unable to save himself because of some direct subverting influence of the poison on his nervous system. Such is the case of the heroin addict who knows the drug is killing him, but who cannot stop taking it because the drug itself controls his nervous system. We have already seen that the cuckoo nestling’s lipstick-like gape is regarded as a supernormal stimulus, and that foster parents have been described as apparently ‘addicted’ to the supernormal stimulus. Could it be that the host can no more resist the supernormal manipulative power of the cuckoo nestling than the junkie can resist his fix, or than the brainwashed prisoner can resist the orders of his captor, however much it would benefit him to do so? Perhaps cuckoos have put their adaptive emphasis on mimetic deception at the egg stage, but on positive manipulation of the host’s nervous system at the late nestling stage.

  Any nervous system can be subverted if treated in the right way. Any evolutionary adaptation of the host nervous system to resist manipulation by cuckoo nestlings lays itself open to counter-adaptation by the cuckoos. Selection acting on cuckoos will work to find whatever chinks there may be in the hosts’ newly evolved psychological armour. Host birds may be very good at resisting psychological manipulation, but cuckoos might become even better at manipulating. All we need to postulate is that, for some reason such as that suggested by the life/dinner principle or the rare-enemy effect, cuckoos have won the arms race: a cuckoo in the nest has got to manipulate its host successfully or it will surely die; its individual foster-parent will benefit somewhat if it resists manipulation, but it still has a good chance of future reproductive success in other years even if it fails to resist this particular cuckoo. Moreover, cuckoos might be sufficiently rare that the risk of an individual of the foster species being parasitized is low; conversely the ‘risk’ of an individual cuckoo’s being a parasite is 100 per cent, no matter how common or rare either party to the arms race may be. The cuckoo is descended from a line of ancestors, every single one of whom has successfully fooled a host. The host is descended from a line of ancestors, many of whom may never have encountered a cuckoo in their lives, or may have reproduced successfully after being parasitized by a cuckoo. The arms race concept completes the classical supernormal stimulus explanation, by providing a functional account of the host’s maladaptive behaviour, instead of leaving it as an unexplained limitation of the nervous system.

  In one respect my treatment of cuckoos as manipulators may be found unsatisfying. The cuckoo is, after all, only diverting the normal parental behaviour of its host. It has not succeeded in building into the host’s behavioural repertoire a whole new behaviour pattern that was not there, in some form, before. Some might find analogies with drugs, hypnotism and electrical stimulation of the brain more persuasive if an example of this more extreme ki
nd of manipulation could be found. A possible case is the ‘preening invitation’ display of another brood parasite, the brown-headed cowbird Molothrus ater (Rothstein 1980). Allopreening, the preening of one individual by another, is not uncommon within various species of birds. It is not particularly surprising, therefore, that cowbirds should succeed in getting other birds of several species to preen them. Again, this can be seen as a simple diversion of intraspecific allopreening, with the cowbird providing a supernormal exaggeration of the normal eliciting stimuli for allopreening. What is rather surprising is that cowbirds manage to get themselves preened by species that never engage in intraspecific allopreening.

  The drug analogy is especially apt for insect ‘cuckoos’ that use chemical means to coerce their hosts into acts that are profoundly damaging to their own inclusive fitness. Several species of ant have no workers of their own. The queens invade nests of other species, dispose of the host queen, and use the host workers to bring up their own reproductive young. The method of disposing of the host queen varies. In some species, such as the descriptively named Bothriomyrmex regicidus and B. decapitans, the parasite queen rides about on the back of the host queen and then, in Wilson’s (1971) delightful description, ‘begins the one act for which she is uniquely specialized: slowly cutting off the head of her victim’ (p. 363).

  Monomorium santschii achieves the same result by more subtle means. The host workers have weapons wielded by strong muscles, and nerves attached to the muscles; why should the parasite queen exert her own jaws if she can subvert the nervous systems controlling the numerous jaws of the host workers? It does not seem to be known how she achieves it, but she does: the host workers kill their own mother and adopt the usurper. A chemical secreted by the parasite queen seems the likely weapon, in which case it might be labelled a pheromone, but it is probably more illuminating to think of it as a formidably powerful drug. In line with this interpretation, Wilson (1971, p. 413) writes of symphylic substances as being ‘more than just elementary nutritive substances or even analogues of the natural host pheromones. Several authors have spoken of a narcotizing effect of symphylic substances.’ Wilson also uses the word ‘intoxicant’ and quotes a case in which worker ants under the influence of such a substance became temporarily disoriented and less sure of their footing.

  Those who have never been brainwashed or addicted to a drug find it hard to understand their fellow men who are driven by such compulsions. In the same naive way we cannot understand a host bird’s being compelled to feed an absurdly oversized cuckoo, or worker ants wantonly murdering the only being in the whole world that is vital to their genetic success. But such subjective feelings are misleading, even where the relatively crude achievements of human pharmacology are concerned. With natural selection working on the problem, who would be so presumptious as to guess what feats of mind control might not be achieved? Do not expect to see animals always behaving in such a way as to maximize their own inclusive fitness. Losers in an arms race may behave in some very odd ways indeed. If they appear disoriented and unsure of their footing, this may be only the beginning.

  Let me stress again what a feat of mind-control the Monomorium santschii queen achieves. To a sterile worker ant, her mother is a kind of genetic gold-mine. For a worker ant to kill her own mother is an act of genetic madness. Why do the workers do it? I am sorry I can do no more than, once again, vaguely talk about arms races. Any nervous system is vulnerable to manipulation by a clever-enough pharmacologist. There is no difficulty in believing that natural selection acting on M. santschii would seek out the weak points in the host workers’ nervous system, and insert a pharmacological key in the lock. Selection on the host species would soon have plugged those weak points, whereupon selection on the parasite would improve the drug, and the arms race was under way. If M. santschii is sufficiently rare, it is easy to see that it might ‘win’ the arms race, even though regicide is such a disastrous act for each host colony whose workers succumb to it. The overall risk of parasitization by M. santschii could be very low even though the marginal cost of regicide, given that an M. santschii queen has entered, is disastrously high. Each individual M. santschii queen is descended from a line of ancestors every one of whom has succeeded in manipulating host workers into regicide. Each host worker is descended from a line of ancestors whose colonies may seldom have been within 10 miles of an M. santschii queen. The costs of ‘bothering’ to be equipped to resist manipulation by an occasional M. santschii queen may outweigh the benefits. Reflections such as this lead me to believe that the hosts might well lose the arms race.

  Other species of parasitic ants use a different system. Instead of sending out queens to implant their eggs in host nests and using host labour there, they transport host labour back to their own nests. These are the so-called slavemaking ants. Slavemaking species have workers, but these workers devote part, or in some cases all, of their energy to going on slaving expeditions. They raid nests of other species and carry off larvae and pupae. These subsequently hatch in the slavers’ nest, where they work normally, foraging and tending brood, not ‘realizing’ that they are, in effect, slaves. The advantage of the slavemaking way of life is presumably that most of the cost of feeding the workforce in the larval stage is saved. That cost is borne by the home colony from which the slave pupae were taken.

  The slavemaking habit is interesting from the present point of view, because it raises an unusual arms race asymmetry. Presumably there is an arms race between slavemaking species and slave species. Adaptations to counter slavery, for instance enlarged soldier jaws for driving off slave-raiders, should be expected in species that are victims of slave raids. But surely the more obvious countermeasure the slaves could take would simply be to withhold their labour in the slavers’ nest, or to kill slavemaker brood instead of feeding them? It seems the obvious countermeasure, but there are formidable obstacles to its evolution. Consider an adaptation to ‘go on strike’, to refuse to work in the slavemakers’ nest. The slave workers would of course have to have some means of recognizing that they had hatched in a foreign nest, but that should not be difficult in principle. The problem arises when we think in detail of how the adaptations would be passed on.

  Since workers don’t reproduce, all worker adaptations, in any social insect species, have to be passed on by reproductive relatives of the workers. This normally presents no insuperable problems, because workers directly assist their own reproductive relatives, so genes giving rise to worker adaptations directly assist copies of themselves in reproductives. But take the example of a mutant gene causing slave workers to go on strike. It may very effectively sabotage the slavemakers’ nest, possibly wipe it out altogether. And to what effect? The area now contains one less slavemaking nest, presumably a good thing for all potential victim nests in the area, not just the nest from which the rebel slaves came, but nests containing non-striking genes as well. The same kind of problem arises in the general case of the spread of ‘spiteful’ behaviour (Hamilton 1970; Knowlton & Parker 1979).

  The only easy way the genes for striking can be preferentially passed on is for striking to benefit, selectively, the strikers’ own home nest, the nest they left behind and in which their own reproductive relatives are being reared. This could happen if slavemakers habitually returned to make repeat raids on the same nest, but otherwise we must conclude that anti-slavery adaptations must be confined to the period before the slave pupae have left their home nest. Once the slaves have arrived in the slavemakers’ nest they effectively drop out of the arms race since they no longer have any power to influence the success of their reproductive relatives. The slavemakers can develop manipulative adaptations of any degree of sophistication, physical or chemical, pheromones or powerful drugs, and the slaves cannot evolve countermeasures.

  Actually, the very fact that the slaves cannot evolve countermeasures will tend to reduce the likelihood that the manipulative techniques evolved by the slavemakers will be very sophisticated: the fact t
hat the slaves cannot retaliate, in an evolutionary sense, means that the slavemakers do not need to spend costly resources on elaborate and sophisticated manipulation adaptations, because simple and cheap ones will do. The example of slavery in ants is rather a special one, but it illustrates a particularly interesting sense in which one side in an arms race can be said to lose completely.

  A case could be made for drawing an analogy here with the hybrid frog Rana esculenta (White 1978). This common European frog, the edible frog of French restaurants, is not a species in the normal sense of the word. Individuals of the ‘species’ are really various kinds of hybrids between two other species, Rana ridibunda and R. lessonae. There are two different diploid forms and two different triploid forms of R. esculenta. For simplicity I shall consider only one of the diploid forms, but the argument holds for all the varieties. These frogs coexist with R. lessonae. Their diploid karyotype consists of one set of lessonae chromosomes and one set of ridibunda chromosomes. At meiosis they discard the lessonae chromosomes and produce pure ridibunda gametes. They mate with lessonae individuals, thereby restoring the hybrid genotype in the next generation. In this race of Rana esculenta bodies, therefore, ridibunda genes are germ-line replicators, lessonae genes dead-end ones. Dead-end replicators can exert phenotypic effects. They can even be naturally selected. But the consequences of that natural selection are irrelevant to evolution (see Chapter 5). To make the next paragraph easier to follow, I shall call R. esculenta H (for hybrid), R. ridibunda G (for germ-line) and R. lessonae D (for dead-end, although it should be remembered that ‘D’ genes are dead-end replicators only when in H frogs; when in ‘D’ frogs they are normal germ-line replicators).