The Extended Phenotype

by Richard Dawkins

  It may seem hard to believe that anyone could have been misled in the way that Wallace indicates, but Young (1971) provides ample evidence confirming that Darwin’s contemporaries often were so misled. Even today the confusion is not unknown, and an analogous muddle arises over the catch-phrase ‘selfish gene’: ‘This is an ingenious theory but far-fetched. There is no reason for imputing the complex emotion of selfishness to molecules’ (Bethell 1978); ‘Genes cannot be selfish or unselfish, any more than atoms can be jealous, elephants abstract or biscuits teleological’ (Midgley 1979; see reply in Dawkins 1981).

  Darwin (1866) was impressed by Wallace’s letter, which he found ‘as clear as daylight’, and he resolved to incorporate ‘survival of the fittest’ into his writings, although he cautioned that ‘the term Natural Selection has now been so largely used abroad and at home that I doubt whether it could be given up, and with all its faults I should be sorry to see the attempt made. Whether it will be rejected must now depend on the “survival of the fittest” …’ (Darwin clearly understood the ‘meme’ principle). ‘As in time the term must grow intelligible, the objections to its use will grow weaker and weaker. I doubt whether the use of any term would have made the subject intelligible to some minds … As for M. Janet, he is a metaphysician, and such gentlemen are so acute that I think they often misunderstand common folk.’

  What neither Wallace nor Darwin could have foreseen was that ‘survival of the fittest’ was destined to generate more serious confusion than ‘natural selection’ ever had. A familiar example is the attempt, rediscovered with almost pathetic eagerness by successive generations of amateur (and even professional) philosophers (‘so acute that they misunderstand common folk’?), to demonstrate that the theory of natural selection is a worthless tautology (an amusing variant is that it is unfalsifiable and therefore false!). In fact the illusion of tautology stems entirely from the phrase ‘survival of the fittest’, and not from the theory itself at all. The argument is a remarkable example of the elevation of words above their station, in which respect it resembles St Anselm’s ontological proof of the existence of God. Like God, natural selection is too big a theory to be proved or disproved by word-games. God and natural selection are, after all, the only two workable theories we have of why we exist.

  Briefly, the tautology idea is this. Natural selection is defined as the survival of the fittest, and the fittest are defined as those that survive. Therefore the whole of Darwinism is an unfalsifiable tautology and we don’t have to worry our heads about it any more. Fortunately, several authoritative replies to this whimsical little conceit are available (Maynard Smith 1969; Stebbins 1977; Alexander 1980), and I need not labour my own. I will, however, chalk up the tautology idea on my list of muddles attributable to the concept of fitness.

  It is, as I have said, a purpose of this chapter to show that fitness is a very difficult concept, and that there might be something to be said for doing without it whenever we can. One way I shall do this is to show that the word has been used by biologists in at least five different senses. The first and oldest meaning is the one closest to everyday usage.

  Fit the First

  When Spencer, Wallace and Darwin originally used the term ‘fitness’, the charge of tautology would not have occurred to anyone. I shall call this original usage fitness[1]. It did not have a precise technical meaning, and the fittest were not defined as those that survive. Fitness meant, roughly, the capacity to survive and reproduce, but it was not defined and measured as precisely synonymous with reproductive success. It had a range of specific meanings, depending upon the particular aspect of life that one was examining. If the subject of attention was efficiency in grinding vegetable food, the fittest individuals were those with the hardest teeth or the most powerful jaw muscles. In different contexts the fittest individuals would be taken to mean those with the keenest eyes, the strongest leg muscles, the sharpest ears, the swiftest reflexes. These capacities and abilities, along with countless others, were supposed to improve over the generations, and natural selection effected that improvement. ‘Survival of the fittest’ was a general characterization of these particular improvements. There is nothing tautological about that.

  It was only later that fitness was adopted as a technical term. Biologists thought they needed a word for that hypothetical quantity that tends to be maximized as a result of natural selection. They could have chosen ‘selective potential’, or ‘survivability’, or ‘W’ but in fact they lit upon ‘fitness’. They did the equivalent of recognizing that the definition they were seeking must be ‘whatever it takes to make the survival of the fittest into a tautology’. They redefined fitness accordingly.

  But the tautology is not a property of Darwinism itself, merely of the catchphrase we sometimes use to describe it. If I say that a train travelling at an average velocity of 120 m.p.h. will reach its destination in half the time it takes a train travelling at 60 m.p.h., the fact that I have uttered a tautology does not prevent the trains from running, nor does it stop us asking meaningful questions about what makes one train faster than the other: does it have a larger engine, superior fuel, a more streamlined shape, or what? The concept of velocity is defined in such a way as to make statements such as the one above tautologically true. It is this that makes the concept of velocity useful. As Maynard Smith (1969) witheringly put it: ‘Of course Darwinism contains tautological features: any scientific theory containing two lines of algebra does so.’ And when Hamilton (1975a), speaking of ‘survival of the fittest’, said that ‘accusations of tautology seem hardly fair on this small phrase itself’, he was putting it mildly. Given the purpose for which fitness was redefined, ‘survival of the fittest’ had to become a tautology.

  To redefine fitness in a special technical sense might have done no harm, other than to give some earnest philosophers a field day, but unfortunately its exact technical meaning has varied widely, and this has had the more serious effect of confusing some biologists too. The most precise and unexceptionable of the various technical meanings is that adopted by population geneticists.

  Fit the Second

  For population geneticists, fitness is an operational measure, exactly defined in terms of a measurement procedure. The word is applied not really to a whole individual organism but to a genotype, usually at a single locus. The fitness W of a genotype, say Aa, may be defined as 1 – s, where s is the coefficient of selection against the genotype (Falconer 1960). It may be regarded as a measure of the number of offspring that a typical individual of genotype Aa is expected to bring up to reproductive age, when all other variation is averaged out. It is usually expressed relative to the corresponding fitness of one particular genotype at the locus, which is arbitrarily defined as 1. Then there is said to be selection, at that locus, in favour of genotypes with higher fitness, relative to genotypes with lower fitness. I shall call this special population geneticists’ meaning of the term fitness[2]. When we say that brown-eyed individuals are fitter than blue-eyed individuals we are talking about fitness[2]. We assume that all other variation among the individuals is averaged out, and we are, in effect, applying the word fitness to two genotypes at a single locus.

  Fit the Third

  But while population geneticists are interested directly in changes in genotype frequencies and gene frequencies, ethologists and ecologists look at whole organisms as integrated systems that appear to be maximizing something. Fitness[3], or ‘classical fitness’, is a property of an individual organism, often expressed as the product of survival and fecundity. It is a measure of the individual’s reproductive success, or its success in passing its genes on to future generations. For instance, as mentioned in Chapter 7, Clutton-Brock et al. (1982) are conducting a long-term study of a red deer population on the island of Rhum, and part of their aim is to compare the lifetime reproductive successes or fitness[3] of identified individual stags and hinds.

  Notice the difference between the fitness[3] of an individual and the fitness[2]
of a genotype. The measured fitness[2] of the brown-eyed genotype will contribute to the fitness[3] of an individual who happens to have brown eyes, but so will the fitness[2] of his genotype at all other loci. Thus the fitness[2] of a genotype at a locus can be regarded as an average of the fitness[3]s of all individuals possessing that genotype. And the fitness[3] of an individual can be regarded as influenced by the fitness[2] of his genotype, averaged over all his loci (Falconer 1960).

  It is easy to measure the fitness[2] of a genotype at a locus, because each genotype, AA, Aa, etc., occurs a countable number of times in successive generations in a population. But the same is not true of the fitness[3] of an organism. You can’t count the number of times an organism occurs in successive generations, because he only occurs once, ever. The fitness[3] of an organism is often measured as the number of his offspring reared to adulthood, but there is some dispute over the usefulness of this. One problem is raised by Williams (1966) criticizing Medawar (1960) who had said: ‘The genetical usage of “fitness” is an extreme attenuation of the ordinary usage: it is, in effect, a system of pricing the endowments of organisms in the currency of offspring, i.e., in terms of net reproductive performance. It is a genetic valuation of goods, not a statement about their nature or quality.’ Williams is worried that this is a retrospective definition, suitable for particular individuals who have existed. It suggests a posthumous evaluation of particular animals as ancestors, not a way of evaluating the qualities that can be expected to make for success in general. ‘My main criticism of Medawar’s statement is that it focuses attention on the rather trivial problem of the degree to which an organism actually achieves reproductive survival. The central biological problem is not survival as such, but design for survival’ (Williams 1966, p. 158). Williams is, in a sense, hankering after the pre-tautological virtues of fitness[1], and there is much to be said in his favour. But the fact is that fitness[3] has become widely used by biologists in the sense described by Medawar. Medawar’s passage was addressed to laymen and was surely an attempt to enable them to follow standard biological terminology while avoiding the otherwise inevitable confusion with common-usage ‘athletic’ fitness.

  The concept of fitness has the power to confuse even distinguished biologists. Consider the following misunderstanding of Waddington (1957) by Emerson (1960). Waddington had used the word ‘survival’ in the sense of reproductive survival or fitness[3]: ‘… survival does not, of course, mean the bodily endurance of a single individual … that individual “survives” best which leaves most offspring’. Emerson quotes this, then goes on: ‘Critical data on this contention are difficult to find, and it is likely that much new investigation is needed before the point is either verified or refuted.’ For once, the ritual lip-service to the need for more research is utterly inappropriate. When we are talking about matters of definition, empirical research cannot help us. Waddington was clearly defining survival in a special sense (the sense of fitness[3]), not making a proposition of fact subject to empirical verification or falsification. Yet Emerson apparently thought Waddington was making the provocative statement that those individuals with the highest capacity to survive tend also to be the individuals with the largest number of offspring. His failure to grasp the technical concept of fitness[3], is indicated by another quotation from the same paper: ‘It would be extremely difficult to explain the evolution of the uterus and mammary glands in mammals … as the result of natural selection of the fittest individual.’ In accordance with the influential Chicago-based school of thought of which he was a leader (Allee, Emerson et al. 1949), Emerson used this as an argument in favour of group selection. Mammary glands and uteruses were, for him, adaptations for the continuation of the species.

  Workers who correctly use the concept of fitness[3] admit that it can be measured only as a crude approximation. If it is measured as the number of children born it neglects juvenile mortality and fails to account for parental care. If it is measured as number of offspring reaching reproductive age it neglects variation in reproductive success of the grown offspring. If it is measured as number of grandchildren it neglects … and so on ad infinitum. Ideally we might count the relative number of descendants alive after some very large number of generations. But such an ‘ideal’ measure has the curious property that, if carried to its logical conclusion, it can take only two values; it is an all-or-none measure. If we look far enough into the future, either I shall have no descendants at all, or all persons alive will be my descendants (Fisher 1930a). If I am descended from a particular individual male who lived a million years ago, it is virtually certain that you are descended from him too. The fitness of any particular long-dead individual, as measured in present-day descendants, is either zero or total. Williams would presumably say that if this is a problem it is so only for people that wish to measure the actual reproductive success of particular individuals. If, on the other hand, we are interested in qualities that tend, on average, to make individuals likely to end up in the set of ancestors, the problem does not arise. In any case a more biologically interesting shortcoming of the concept of fitness[3] has led to the development of two newer usages of the technical term fitness.

  Fit the Fourth

  Hamilton (1964a,b), in a pair of papers which we can now see to have marked a turning point in the history of evolutionary theory, made us aware of an important deficiency in classical fitness[3], the measure based on the reproductive success of an organism. The reason reproductive success matters, as opposed to mere individual survival, is that reproductive success is a measure of success in passing on genes. The organisms we see around us are descended from ancestors, and they have inherited some of the attributes that made those individuals ancestors as opposed to non-ancestors. If an organism exists it contains the genes of a long line of successful ancestors. The fitness[3] of an organism is its success as an ancestor, or, according to taste, its capacity for success as an ancestor. But Hamilton grasped the central importance of what, previously, had been only glancingly referred to in stray sentences of Fisher (1930a) and Haldane (1955). This is that natural selection will favour organs and behaviour that cause the individual’s genes to be passed on, whether or not the individual is, himself, an ancestor. An individual that assists his brother to be an ancestor may thereby ensure the survival in the gene-pool of the genes ‘for’ brotherly assistance. Hamilton saw that parental care is really only a special case of caring for close relatives with a high probability of containing the genes for caring. Classical fitness[3], reproductive success, was too narrow. It had to be broadened to inclusive fitness, which will here be called fitness[4].

  It is sometimes supposed that the inclusive fitness of an individual is his own fitness[3] plus half the fitnesses[3] of each brother plus one-eighth of the fitness[3] of each cousin, etc. (e.g. Bygott et al. 1979). Barash (1980) explicitly defines it as ‘the sum of individual fitness (reproductive success) and the reproductive success of an individual’s relatives, with each relative devalued in proportion as it is more distantly related’. This would not be a sensible measure to try to use, and, as West-Eberhard (1975) emphasizes, it is not the measure Hamilton offered us. The reason it would not be sensible can be stated in various ways. One way of putting it is that it allows children to be counted many times, as though they had many existences (Grafen 1979). Then again, if a child is born to one of a set of brothers, the inclusive fitness of all the other brothers would, according to this view, immediately rise an equal notch, regardless whether any of them lifted a finger to feed the infant. Indeed the inclusive fitness of another brother yet unborn would theoretically be increased by the birth of his elder nephew. Further, this later brother could be aborted shortly after conception and still, according to this erroneous view, enjoy a massive ‘inclusive fitness’ through the descendants of his elder brothers. To push to the reductio ad absurdum, he needn’t even bother to be conceived, yet still have a high ‘inclusive fitness’!

  Hamilton clearly saw this fa
llacy, and therefore his concept of inclusive fitness was more subtle. The inclusive fitness of an organism is not a property of himself, but a property of his actions or effects. Inclusive fitness is calculated from an individual’s own reproductive success plus his effects on the reproductive success of his relatives, each one weighed by the appropriate coefficient of relatedness. Therefore, for instance, if my brother emigrates to Australia, so I can have no effect, one way or the other, on his reproductive success, my inclusive fitness does not go up each time he has a child!

  Now ‘effects’ of putative causes can only be measured by comparison with other putative causes, or by comparison with their absence. We cannot, then, think of the effects of individual A on the survival and reproduction of his relatives in any absolute sense. We could compare the effects of his choosing to perform act X rather than act Y. Or we could take the effects of his lifetime’s set of deeds and compare them with a hypothetical lifetime of total inaction—as though he had never been conceived. It is this latter usage that is normally meant by the inclusive fitness of an individual organism.

  The point is that inclusive fitness is not an absolute property of an organism in the same way as classical fitness[3] could in theory be, if measured in certain ways. Inclusive fitness is a property of a triple consisting of an organism of interest, an act or set of acts of interest, and an alternative set of acts for comparison. We aspire to measure, then, not the absolute inclusive fitness of the organism I, but the effect on I’s inclusive fitness of his doing act X in comparison with his doing act Y. If ‘act’ X is taken to be I’s entire life story, Y may be taken as equivalent to a hypothetical world in which I did not exist. An organism’s inclusive fitness, then, is defined so that it is not affected by the reproductive success of relatives on another continent whom he never meets and whom he has no way of affecting.