So far we have recognized the possibility of a complete range of female preferences, from females with a taste for long-tailed males through to females with the opposite taste, for short-tailed males. But if we actually did a poll of the females in a particular population, we would probably find that a majority of females shared the same general tastes in males. We can express the range of female tastes in the population in the same units — inches — as we express the range of male tail lengths. And we can express the average female preference in the same units of inches. It could turn out that the average female preference was exactly the same as the average male tail length, 3 inches in both cases. In this case female choice will not be an evolutionary force tending to change male tail length. Or it could turn out that the average female preference was for a tail rather longer than the average tail that actually exists, say 4 inches rather than 3. Leaving open, for the moment, why there might be such a discrepancy, just accept that there is one and ask the next obvious question. Why, if most females prefer males with 4-inch tails, do the majority of males actually have 3-inch tails? Why doesn’t the average tail length in the population shift to 4 inches under the influence of female sexual selection? How can there be a discrepancy of 1 inch between the average preferred tail length and the actual average tail length?
The answer is that female taste is not the only kind of selection that bears upon male tail length. Tails have an important job to perform in flight, and a tail that is too long or too short will decrease the efficiency of flight. Moreover, a long tail costs more energy to carry around, and more to make it in the first place. Males with 4-inch tails might well pull the female birds, but the price the males would pay is their less-efficient flight, greater energy costs and greater vulnerability to predators. We can express this by saying that there is a utilitarian optimum tail length, which is different from the sexually selected optimum: an ideal tail length from the point of view of ordinary useful criteria; a tail length that is ideal from all points of view apart from attracting females.
Should we expect that the actual average tail length of males, 3 inches in our hypothetical example, will be the same as the utilitarian optimum? No, we should expect the utilitarian optimum to be less, say 2 inches. The reason is that the actual average tail length of 3 inches is the result of a compromise between utilitarian selection tending to make tails shorter, and sexual selection tending to make them longer. We may surmise that, if there were no need to attract females, average tail length would shrink towards 2 inches. If there were no need to worry about flying efficiency and energy costs, average tail length would shoot out towards 4 inches. The actual average of 3 inches is a compromise.
We left on one side the question of why females might agree in preferring a tail that departed from the utilitarian optimum. At first sight the very idea seems silly. Fashion-conscious females, with a taste for tails that are longer than they should be on good design criteria, are going to have poorly designed, inefficient, clumsily flying sons. Any mutant female who happened to have an unfashionable taste for shorter-tailed males, in particular a mutant female whose taste in tails happened to coincide with the utilitarian optimum, would produce efficient sons, well designed for flying, who would surely outcompete the sons of her more fashion-conscious rivals. Ah, but here is the rub. It is implicit in my metaphor of ‘fashion’. The mutant female’s sons may be efficient flyers, but they are not seen as attractive by the majority of females in the population. They will attract only minority females, fashion-defying females; and minority females, by definition, are harder to find than majority females, for the simple reason that they are thinner on the ground. In a society where only one in six males mates at all and the fortunate males have large harems, pandering to the majority tastes of females will have enormous benefits, benefits that are well capable of outweighing the utilitarian costs in energy and flight efficiency.
But even so, the reader may complain, the whole argument is based upon an arbitrary assumption. Given that most females prefer nonutilitarian long tails, the reader will admit, everything else follows. But why did this majority female taste come about in the first place? Why didn’t the majority of females prefer tails that are smaller than the utilitarian optimum, or exactly the same length as the utilitarian optimum? Why shouldn’t fashion coincide with utility? The answer is that any of these things might have happened, and in many species it probably did. My hypothetical case of females preferring long tails was, indeed, arbitrary. But whatever the majority female taste had happened to be, and no matter how arbitrary, there would have been a tendency for that majority to be maintained by selection or even, under some conditions, actually increased — exaggerated. It is at this point in the argument that the lack of mathematical justification in my account becomes really noticeable. I could invite the reader simply to accept that the mathematical reasoning of Lande proves the point, and leave it at that. This might be the wisest course for me to pursue, but I shall have one try at explaining part of the idea in words.
The key to the argument lies in the point we established earlier about ‘linkage disequilibrium’, the ‘togetherness’ of genes for tails of a given length — any length — and corresponding genes for preferring tails of that self-same length. We can think about the ‘togetherness factor’ as a measurable number. If the togetherness factor is very high, this means that knowledge about an individual’s genes for tail length enables us to predict, with great accuracy, his/her genes for preference, and vice versa. Conversely, if the togetherness factor is low, this means that knowledge about an individual’s genes in one of the two departments — preference or tail length — gives us only a slight hint about his/her genes in the other department.
The kind of thing that affects the magnitude of the togetherness factor is the strength of the females’ preference — how tolerant they are of what they see as imperfect males; how much of the variation in male tail length is governed by genes as opposed to environmental factors; and so on. If, as a result of all these effects, the togetherness factor — the tightness of binding of genes for tail length and genes for tail-length preference — is very strong, we can deduce the following consequence. Every time a male is chosen because of his long tail, not only are genes for long tails being chosen. At the same time, because of the ‘togetherness’ coupling, genes for preferring long tails are also being chosen. What this means is that genes that make females choose male tails of a particular length are, in effect, choosing copies of themselves. This is the essential ingredient of a selfreinforcing process: it has its own self-sustaining momentum. Evolution having started in a particular direction, this can, in itself, tend to make it persist in the same direction.
Another way to see this is in terms of what has become known as the ‘green-beard effect’. The green-beard effect is a kind of academic biological joke. It is purely hypothetical, but it is instructive nevertheless. It was originally proposed as a way of explaining the fundamental principle underlying W. D. Hamilton’s important theory of kin selection, which I discussed at length in The Selfish Gene. Hamilton, now my colleague at Oxford, showed that natural selection would favour genes for behaving altruistically towards close kin, simply because copies of those selfsame genes had a high probability of being in the bodies of kin. The ‘green-beard’ hypothesis puts the same point more generally, if less practically. Kinship, the argument runs, is only one possible way in which genes can, in effect, locate copies of themselves in other bodies. Theoretically, a gene could locate copies of itself by more direct means. Suppose a gene happened to arise that had the following two effects (genes with two or more effects are common): it makes its possessors have a conspicuous ‘badge’ such as a green beard, and it also affects their brains in such a way that they behave altruistically towards green-bearded individuals. A pretty improbable coincidence, admittedly, but if it ever did arise the evolutionary consequence is clear. The green-beard altruism gene would tend to be favoured by natural selection, for exact
ly the same kinds of reason as genes for altruism towards offspring or brothers. Every time a green-bearded individual helped another, the gene for giving this discriminating altruism would be favouring a copy of itself. The spread of the green-beard gene would be automatic and inevitable.
Nobody really believes, not even I, that the green-beard effect, in this ultra-simple form, will ever be found in nature. In nature, genes discriminate in favour of copies of themselves by means of less specific but more plausible labels than green beards. Kinship is just such a label. ‘Brother’ or, in practice, something like ‘he who has just hatched in the nest from which I have just fledged’, is a statistical label. Any gene that makes individuals behave altruistically towards bearers of such a label has a good statistical chance of aiding copies of itself: for brothers have a good statistical chance of sharing genes. Hamilton’s theory of kin selection can be seen as one way in which a green-beard type of effect can be made plausible. Remember, by the way, that there is no suggestion here that genes ‘want’ to help copies of themselves. It is just that any gene that happens to have the effect of helping copies of itself will tend, willy nilly, to become more numerous in the population.
Kinship, then, can be seen as a way in which something like the green-beard effect can be made plausible. The Fisher theory of sexual selection can be explained as yet another way in which the green-beard can be made plausible. When the females of a population have strong preferences for male characteristics, it follows, by the reasoning we have been through, that each male body will tend to contain copies of genes that make females prefer his own characteristics. If a male has inherited a long tail from his father, the chances are that he has also inherited from his mother the genes that made her choose the long tail of his father. If he has a short tail, the chances are that he contains genes for making females prefer short tails. So, when a female exercises her choice of male, whichever way her preference lies, the chances are that the genes that bias her choice are choosing copies of themselves in the males. They are choosing copies of themselves using male tail length as a label, in a more complicated version of the way the hypothetical green-beard gene uses the green beard as a label.
If half the females in the population preferred long-tailed males, and the other half short-tailed males, genes for female choice would still be choosing copies of themselves, but there would be no tendency for one or other tail type to be favoured in general. There might be a tendency for the population to split into two — a longtailed, long-preferring faction, and a short-tailed, short-preferring faction. But any such two-way split in female ‘opinion’ is an unstable state of affairs. The moment a majority, however slight, started to accrue among females for one type of preference rather than the other, that majority would be reinforced in subsequent generations. This is because males preferred by females of the minority school of thought would have a harder time finding mates; and females of the minority school of thought would have sons who had a relatively hard time finding mates, so minority females would have fewer grandchildren. Whenever small minorities tend to become even smaller minorities, and small majorities tend to become bigger majorities, we have a recipe for positive feedback: ‘Unto every one that hath shall be given, and he shall have abundance: but from him that hath not shall be taken away even that which he hath.’ Whenever we have an unstable balance, arbitrary, random beginnings are self-reinforcing. Just so, when we cut through a tree trunk, we may be uncertain whether the tree will fall to the north or the south; but, after remaining poised for a time, once it starts to fall in one direction or the other, nothing can bring it back.
Lacing our climbing boots even more securely we prepare to hammer in another piton. Remember that selection by females is pulling male tails in one direction, while ‘utilitarian’ selection is pulling them in the other (‘pulling’ in the evolutionary sense, of course), the actual average tail length being a compromise between the two pulls. Let us now recognize a quantity called the ‘choice discrepancy’. This is the difference between the actual average tail length of males in the population, and the ‘ideal’ tail length that the average female in the population would really prefer. The units in which the choice discrepancy is measured are arbitrary, just as the Fahrenheit and Centigrade scales of temperature are arbitrary. Just as the Centigrade scale finds it convenient to fix its zero point at the freezing point of water, we shall find it convenient to fix our zero at the point where the pull of sexual selection exactly balances the opposite pull of utilitarian selection. In other words, a choice discrepancy of zero means that evolutionary change comes to a halt because the two opposite kinds of selection exactly cancel each other out.
Obviously, the larger the choice discrepancy, the stronger the evolutionary ‘pull’ exerted by females against the counteracting pull of utilitarian natural selection. What we are interested in is not the absolute value of the choice discrepancy at any particular time, but how the choice discrepancy changes in successive generations. As a result of a given choice discrepancy, tails get longer, and at the same time (remember that genes for choosing long tails are being selected in concert with genes for having long tails) the females’ ideal preferred tail gets longer too. After a generation of this dual selection, both average tail length and average preferred tail length have become longer, but which has increased the most? This is another way of asking what will happen to the choice discrepancy.
The choice discrepancy could have stayed the same (if average tail length and average preferred tail length both increased by the same amount). It could have become smaller (if average tail length increased more than preferred tail length did). Or, finally, it could have become larger (if average tail length increased somewhat, but average preferred tail length increased even more). You can begin to see that, if the choice discrepancy gets smaller as tails get larger, tail length will evolve towards a stable equilibrium length. But if the choice discrepancy gets larger as tails get larger, future generations should theoretically see tails shooting out at ever increasing speed. This is, without any doubt, what Fisher must have calculated before 1930, although his brief published words were not clearly understood by others at the time.
Let us first take the case where the choice discrepancy becomes ever smaller as the generations go by. It will eventually become so small that the pull of female preference in one direction is exactly balanced by the pull of utilitarian selection in the other. Evolutionary change will then come to a halt, and the system is said to be in a state of equilibrium. The interesting thing Lande proved about this is that, at least under some conditions, there is not just one point of equilibrium, but many (theoretically an infinite number arranged in a straight line on a graph, but there’s mathematics for you!). There is not just one balance point but many: for any strength of utilitarian selection pulling in one direction, the strength of female preference evolves in such a way as to reach a point where it balances it exactly.
So, if conditions are such that the choice discrepancy tends to become smaller as the generations go by, the population will come to rest at the ‘nearest’ point of equilibrium. Here utilitarian selection pulling in one direction will be exactly counteracted by female selection pulling in the other, and the tails of the males will stay the same length, regardless of how long that is. The reader may recognize that we have here a negative-feedback system, but it is a slightly weird kind of negative-feedback system. You can always tell a negativefeedback system by what happens if you ‘perturb’ it away from its ideal, ‘set point’. If you perturb a room’s temperature by opening the window, for instance, the thermostat responds by turning on the heater to compensate.
How might the system of sexual selection be perturbed? Remember that we are talking about the evolutionary timescale here, so it is difficult for us to do experiments — the equivalent of opening the window — and live to see the results. But no doubt, in nature, the system is perturbed often, for instance by spontaneous, random fluctuations in numbers of males due to
chance, lucky or unlucky, events. Whenever this happens, given the conditions we have so far discussed, a combination of utilitarian selection and sexual selection will return the population to the nearest one of the set of equilibrium points. This probably will not be the same equilibrium point as before, but will be another point a little bit higher, or lower, along the line of equilibrium points. So, as time goes by, the population can drift up or down the line of equilibrium points. Drifting up the line means that tails get longer — theoretically there is no limit to how long. Drifting down the line means that tails get shorter — theoretically all the way down to a length of zero.
The analogy of a thermostat is often used to explain the idea of a point of equilibrium. We can develop the analogy to explain the more difficult idea of a line of equilibria. Suppose that a room has both a heating device and a cooling device, each with its own thermostat. Both thermostats are set to keep the room at the same fixed temperature, 70 degrees F. If the temperature drops below 70, the heater turns itself on and the refrigerator turns itself off. If the temperature rises above 70, the refrigerator turns itself on and the heater turns itself off. The analogue of the widow bird’s tail length is not the temperature (which remains approximately constant at 70°) but the total rate of consumption of electricity. The point is that there are lots of different ways in which the desired temperature can be achieved. It can be achieved by both devices working very hard, the heater belting out hot air and the refrigerator working flat out to neutralize the heat. Or it can be achieved by the heater putting out a bit less heat, and the cooler working correspondingly less hard to neutralize it. Or it can be achieved by both devices working scarcely at all. Obviously, the latter is the most desirable solution from the point of view of the electricity bill but, as far as the object of maintaining the fixed temperature of 70 degrees is concerned, every one of a large series of working rates is equally satisfactory. We have a line of equilibrium points, rather than a single point. Depending upon details of how the system was set up, upon delays in the system and other things of the kind that preoccupy engineers, it is theoretically possible for the room’s electricity-consumption rate to drift up and down the line of equilibrium points, while the temperature remains the same. If the room’s temperature is perturbed a little below 70 degrees it will return, but it won’t necessarily return to the same combination of work rates of the heater and the cooler. It may return to a different point along the line of equilibria.