The Extended Phenotype

by Richard Dawkins

  The past experience of autosomal genetic fragments in a species is, statistically speaking, similar. It consists of an ensemble of typical species bodies, approximately 50 per cent male and 50 per cent female bodies, bodies which grew through a wide spectrum of ages at least up to reproductive age; and it includes a good random shuffling of ‘companion’ genes at other loci. The genes that exist today tend to be the ones that are good at surviving in that statistical ensemble of bodies, and in company with that statistical ensemble of companion genes. As we shall see, it is selection in favour of the qualities needed for survival in company with other, similarly selected, genes, that gives rise to the appearance of ‘coadapted genomes’. I shall show in Chapter 13 that this is a much more illuminating interpretation of the phenomenon of coadaptation than its alternative, that ‘the coadapted genome is the true unit of selection’.

  Probably no two genes in an organism have identical past experiences, though a linked pair may come close, and, mutants aside, all the genes on a Y chromosome have travelled together through the same set of bodies for a large number of generations. But the exact nature of a gene’s past experience is of less interest than the generalizations one can make about the past experiences of all genes that exist today. For instance, however variable the set of my ancestors may be, they all had in common that they survived at least to reproductive age, they copulated heterosexually and were fertile. The same generalization cannot be made about the historical set of bodies that were not my ancestors. The bodies that provided the past experience of existing genes are a non-random subset of all the bodies that have ever existed.

  The genes that exist today reflect the set of environments that they have experienced in the past. This includes the internal environments provided by the bodies the genes have inhabited, and also external environments, desert, forest, seashore, predators, parasites, social companions, etc. This is, of course, not because the environments have imprinted their qualities on the genes—that would be Lamarckism (see Chapter 9)—but because the genes that exist today are a selected set, and the qualities that made them survive reflect the qualities of the environments in which they survived.

  I said that a gene’s experience consists of time spent in approximately 50 per cent male and 50 per cent female bodies, but this is not, of course, true of genes on sex chromosomes. In mammals, assuming no Y-chromosome crossing-over, a gene on a Y chromosome has experienced only male bodies, and a gene on an X chromosome has spent two-thirds of its history in female bodies and one-third in male bodies. In birds, Y-chromosome genes have experienced only female bodies, and in particular cases such as cuckoos we can say something further. Female Cuculus canorus are divided into ‘gentes’, each gens parasitizing a different species of host (Lack 1968). Apparently each female learns the qualities of her own foster parents and their nest, and returns when adult to parasitize the same species. Males do not seem to discriminate as to gens in their choice of mate, and therefore they act as vehicles for gene flow between gentes. Of the genes in a female cuckoo, therefore, those on autosomes and the X chromosome have probably had recent experience of all gentes in the population, and been ‘reared’ by foster parents belonging to all the species parasitized by the cuckoo population. But the Y chromosome, uniquely, is confined for long sequences of generations to one gens and one foster species. Of all the genes sitting in a robin’s nest, one subset—robin genes and cuckoo Y-chromosome genes (and robin flea genes)—has sat in robins’ nests for many generations back. Another subset—cuckoo autosomal and X-chromosome genes—has experienced a mixture of nests. Of course the first subset share only part of their experience, a long series of robin nests. In other aspects of their experience, cuckoo Y-chromosome genes will have more in common with other cuckoo genes than with robin genes. But as far as certain particular selection pressures found in nests are concerned, cuckoo Y-chromosome genes have more in common with robin genes than with cuckoo autosome genes. It is natural, then, that cuckoo Y chromosomes should have evolved to reflect their peculiar experience, while other cuckoo genes simultaneously evolved to reflect their more general experience—a kind of incipient intragenomic ‘speciation’ at the chromosomal level. It is, indeed, widely assumed for this reason that genes for foster species-specific egg mimicry must be carried on the Y chromosome, while genes for general parasitic adaptations might be carried on any chromosome.

  I am not sure whether the fact has significance, but this backwards way of looking shows X chromosomes, too, to have a peculiar history. A gene on an autosome in a female cuckoo is as likely to have come from the father as from the mother, in which latter case it experiences the same host species for two generations running. A gene on an X chromosome in a female cuckoo is bound to have come from the father, and therefore is not especially likely to experience the same host species two generations running. A statistical ‘runs test’ on the sequence of foster species experienced by an autosomal gene would therefore reveal a slight runs effect, greater than that for a gene on an X chromosome, and much less than that for a gene on a Y chromosome.

  In any animal, an inverted portion of chromosome may resemble a Y chromosome in being unable to cross over. The ‘experience’ of any part of the ‘inversion supergene’ therefore repeatedly includes the other parts of the supergene and their phenotypic consequences. A habitat selection gene anywhere in such a supergene, say a gene that makes individuals choose dry microclimates, would then provide a consistent habitat ‘experience’ for successive generations of the whole supergene. Thus a given gene may ‘experience’ consistently dry habitats for the same kind of reason as a gene on a cuckoo’s Y chromosome consistently experiences meadow-pipit nests. This will provide a consistent selection pressure bearing upon that locus, favouring alleles that are adapted to a dry habitat in the same way as alleles for mimicking meadow-pipit eggs are favoured in Y chromosomes of female cuckoos of the meadow-pipit gens. This particular inversion supergene will tend to be found for generations in dry habitats, even though the rest of the genome may be randomly shuffled over the whole range of habitats available to the species. Many different loci on the inverted portion of chromosome may therefore come to be adapted to a dry climate, and again something akin to intragenomic incipient speciation may go on. I find this backwards way of looking at the past ‘experience’ of genetic replicators helpful.

  Germ-line replicators, then, are units that actually survive or fail to survive, the difference constituting natural selection. Active replicators have some effect on the world, which influences their chances of surviving. It is the effects on the world of successful active germ-line replicators that we see as adaptations. Fragments of DNA qualify as active germ-line replicators. Where there is sexual reproduction, these fragments must not be defined too large if they are to retain the property of self-duplication. And they must not be defined too small if they are to be usefully regarded as active.

  If there were sex but no crossing-over, each chromosome would be a replicator, and we should speak of adaptations as being for the good of the chromosome. If there is no sex we can, by the same token, treat the entire genome of an asexual organism as a replicator. But the organism itself is not a replicator. This is for two quite distinct reasons which should not be confused with each other. The first reason follows from the arguments developed in this chapter, and applies only where there is sexual reproduction and meiosis: meiosis and sexual fusion see to it that not even our genomes are replicators, so we ourselves are not replicators either. The second reason applies to asexual as well as sexual reproduction. It will be explained in the next chapter, which goes on to discuss what organisms, and also groups of organisms, are, given that they are not replicators.

  6 Organisms, Groups and Memes: Replicators or Vehicles?

  I have made so much of the fragmenting effects of meiosis as a reason for not regarding sexually reproduced organisms as replicators, that it is tempting to see this as the only reason. If this were true, it should follow that as
exually reproduced organisms are true replicators, and that where reproduction is asexual we could legitimately speak of adaptations as ‘for the good of the organism’. But the fragmenting effect of meiosis is not the only reason for denying that organisms are true replicators. There is a more fundamental reason, and it applies to asexual organisms as much as to sexual ones.

  To regard an organism as a replicator, even an asexual organism like a female stick insect, is tantamount to a violation of the ‘central dogma’ of the non-inheritance of acquired characteristics. A stick insect looks like a replicator, in that we may lay out a sequence consisting of daughter, granddaughter, great-granddaughter, etc., in which each appears to be a replica of the preceding one in the series. But suppose a flaw or blemish appears somewhere in the chain, say a stick insect is unfortunate enough to lose a leg. The blemish may last for the whole of her lifetime, but it is not passed on to the next link in the chain. Errors that affect stick insects but not their genes are not perpetuated. Now lay out a parallel series consisting of daughter’s genome, granddaughter’s genome, great-granddaughter’s genome, etc. If a blemish appears somewhere along this series it will be passed on to all subsequent links in the chain. It may also be reflected in the bodies of all subsequent links in the chain, because in each generation there are causal arrows leading from genes to body. But there is no causal arrow leading from body to genes. No part of the stick insect’s phenotype is a replicator. Nor is her body as a whole. It is wrong to say that ‘just as genes can pass on their structure in gene lineages, organisms can pass on their structure in organism lineages’.

  I am sorry if I am about to labour this argument, but I fear it was my failure to be clear about it before that led to an unnecessary disagreement with Bateson, a disagreement which it is worth going to some trouble to sort out. Bateson (1978) made the point that genetic determinants of development are necessary but not sufficient. A gene may ‘program’ a particular bit of behaviour ‘without it being the only thing to do so’. He goes on:

  Dawkins accepts all this but then reveals his uncertainty about which language he is using by immediately giving special status back to the gene as the programmer. Consider a case in which the ambient environmental temperature during development is crucial for the expression of a particular phenotype. If the temperature changes by a few degrees the survival machine is beaten by another one. Would not that give as much status to a necessary temperature value as to a necessary gene? The temperature value is also required for the expression of a particular phenotype. It is also stable (within limits) from one generation to the next. It may even be transmitted from one generation to the next if the survival machine makes a nest for its offspring. Indeed, using Dawkins’ own style of teleological argument one could claim that the bird is the nest’s way of making another nest [Bateson 1978].

  I replied to Bateson, but did so too briefly, picking on the last remark about birds’ nests and saying, ‘A nest is not a true replicator because a [non-genetic] “mutation” which occurs in the construction of a nest, for example the accidental incorporation of a pine needle instead of the usual grass, is not perpetuated in future “generations of nests”. Similarly, protein molecules are not replicators, nor is messenger RNA’ (Dawkins 1978a). Bateson had taken the catchphrase about a bird being a gene’s way of making another gene, and inverted it, substituting ‘nest’ for ‘gene’. But the parallel is not a valid one. There is a causal arrow going from gene to bird, but none in the reverse direction. A changed gene may perpetuate itself better than its unmutated allele. A changed nest will do no such thing unless, of course, the change is due to a changed gene, in which case it is the gene that is perpetuated, not the nest. A nest, like a bird, is a gene’s way of making another gene.

  Bateson is worried that I seem to give ‘special status’ to genetic determinants of behaviour. He fears that an emphasis on the gene as the entity for whose benefit organisms labour, rather than the other way around, leads to an undue emphasis on the importance of genetic as opposed to environmental determinants of development. The answer to this is that when we are talking about development it is appropriate to emphasize non-genetic as well as genetic factors. But when we are talking about units of selection a different emphasis is called for, an emphasis on the properties of replicators. The special status of genetic factors rather than non-genetic factors is deserved for one reason only: genetic factors replicate themselves, blemishes and all, but non-genetic factors do not.

  Let us grant with both hands that the temperature in the nest housing a developing bird is important both for its immediate survival and for the way it develops and therefore for its long-term success as an adult. The immediate effects of gene products on the biochemical springs of development may, indeed, closely resemble the effects of temperature changes (Waddington 1957). We could even imagine the enzyme products of genes as little Bunsen burners, selectively applied at crucial nodes of the branching biochemical tree of embryonic causation, controlling development by selective control of biochemical reaction rates. An embryologist rightly sees no fundamental distinction between genetic and environmental causal factors, and he correctly regards each as necessary but not sufficient. Bateson was putting the embryologist’s point of view, and no ethologist is better qualified to do so. But I was not talking embryology. I was not concerned with the rival claims of determinants of development. I was talking about replicators surviving in evolutionary time, and Bateson certainly agrees that neither a nest, nor the temperature inside it, nor the bird that built it, are replicators. We can quickly see that they are not replicators by experimentally altering one of them. The change may wreak havoc on the animal, on its development and its chances of survival, but the change will not be passed on to the next generation. Now make a similar mutilation (mutation) to a gene in the germ-line: the change may or may not affect the bird’s development and its survival, but it can be passed on to the next generation; it can be replicated.

  As is so often the case, an apparent disagreement turns out to be due to mutual misunderstanding. I thought that Bateson was denying proper respect to the Immortal Replicator. Bateson thought that I was denying proper respect to the Great Nexus of complex causal factors interacting in development. In fact, each of us was laying legitimate stress on considerations which are important for two different major fields of biology, the study of development and the study of natural selection.

  An organism, then, is not a replicator, not even (despite Lewontin 1970a—see Dawkins 1982) a crude replicator with poor copying fidelity. It is therefore better not to speak of adaptations as being for the good of the organism. What about larger units, groups of organisms, species, communities of species, etc.? Some of these larger groupings are clearly subject to a version of the ‘internal fragmentation destroys copying fidelity’ argument. The fragmenting agent in this case is not the recombining effects of meiosis, but immigration and emigration, the destruction of the integrity of groups by the movement of individuals into and out of them. As I have put it before, they are like clouds in the sky or dust-storms in the desert. They are temporary aggregations or federations. They are not stable through evolutionary time. Populations may last a long while, but they are constantly blending with other populations and so losing their identity. They are also subject to evolutionary change from within. A population is not a discrete enough entity to be a unit of natural selection, not stable and unitary enough to be ‘selected’ in preference to another population. But, just as the ‘fragmentation’ argument applied only to a subset of organisms, sexual ones, so it also applies only to a subset at the group level. It applies to groups capable of interbreeding, but it does not apply to reproductively isolated species.

  Let us, then, examine whether species behave sufficiently like coherent entities, multiplying and giving rise to other species, to deserve to be called replicators. Note that that is not the same as Ghiselin’s (1974b) logical claim that species are ‘individuals’ (see also Hull 1976). Organism
s, too, are individuals in Ghiselin’s sense, and I hope I have established that organisms are not replicators. Do species, or, to be more precise, do reproductively isolated gene-pools, really answer to the definition of replicators?

  It is important to remember that mere immortality is not a sufficient qualification. A lineage, such as a sequence of parents and offspring from the long-unchanged brachiopod genus Lingula, is unending in the same sense, and to the same extent, as a lineage of genes. Indeed, for this example we perhaps need not have chosen a ‘living fossil’ like Lingula. Even a rapidly evolving lineage can, in a sense, be treated as an entity which is either extinct or extant at any moment in geological time. Now, certain kinds of lineage may be more likely to go extinct than others, and we may be able to discern statistical laws of extinction. For example, lineages whose females reproduce asexually may be systematically more or less likely to go extinct than lineages whose females stick to sex (Williams 1975; Maynard Smith 1978a). It has been suggested that ammonite and bivalve lineages with a high rate of evolving larger size (i.e. with a high rate of obeying Cope’s Rule) are more likely to go extinct than more slowly evolving lineages (Hallam 1975). Leigh (1977) makes some excellent points about differential lineage extinction, and its relationship to lower levels of selection: ‘… those species are favored where selection within populations works more nearly for the good of the species’. Selection ‘… favors species that have, for whatever reason, evolved genetic systems where a gene’s selective advantage more nearly matches its contribution to fitness’. Hull (1980a,b) is particularly clear about the logical status of the lineage, and about its distinction from the replicator and the interactor (Hull’s name for what I am calling the ‘vehicle’).