The Extended Phenotype

by Richard Dawkins

  In principle, we may consider any portion of chromosome as a potential candidate for the title of replicator. Natural selection may usually be safely regarded as the differential survival of replicators relative to their alleles. The word allele is nowadays customarily used of cistrons but it is clearly easy, and in the spirit of this chapter, to generalize it to any portion of chromosome. If we look at a portion of chromosome five cistrons long, its alleles are the alternative sets of five cistrons that exist at the homologous loci of all the chromosomes in the population. An allele of an arbitrary sequence of twenty-six codons is an alternative homologous sequence of twenty-six codons somewhere in the population. Any stretch of DNA, beginning and ending at arbitrarily chosen points on the chromosome, can be considered to be competing with allelomorphic stretches for the region of chromosome concerned. It further follows that we can generalize the terms homozygous and heterozygous. Having picked out an arbitrary length of chromosome as our candidate replicator, we look at the homologous chromosome in the same diploid individual. If the two chromosomes are identical over the whole length of the replicator, the individual is homozygous for that replicator, otherwise it is heterozygous.

  When I said ‘arbitrarily chosen portion of chromosome’, I really meant arbitrarily. The twenty-six codons that I chose might well span the border between two cistrons. The sequence still potentially fits the definition of a replicator, it is still possible to think of it as having alleles, and it still may be thought of as homozygous or heterozygous to the corresponding portion of the homologous chromosome in a diploid genotype. This, then, is our candidate replicator. But a candidate should be regarded as an actual replicator only if it possesses some minimum degree of longevity/fecundity/fidelity (there may be trade-offs among the three). Other things being equal, it is clear that larger candidates will have lower longevity/fecundity/fidelity than smaller ones because they are more vulnerable to being broken by recombination events. So, how large and how small a portion of chromosome is it useful to treat as a replicator?

  This depends on the answer to another question: ‘useful for what?’ The reason a replicator is interesting to Darwinians is that it is potentially immortal, or at least very long-lived in the form of copies. A successful replicator is one that succeeds in lasting, in the form of copies, for a very long time measured in generations, and succeeds in propagating many copies of itself. An unsuccessful replicator is one that potentially might have been long-lived, but in fact failed to survive, say because it caused the successive bodies in which it found itself to be sexually unattractive. We may apply the terms ‘successful’ and ‘unsuccessful’ to any arbitrarily defined portion of chromosome. Its success is measured relative to its alleles, and, if there is heterozygosity at the replicator locus in the population, natural selection will change the relative frequencies of the allelomorphic replicators in the population. But if the arbitrarily chosen portion of chromosome is very long it is not even potentially long-lived in its present form, for it is likely to be split apart by crossing-over in any given generation, regardless of how successful it may be in making a body survive and reproduce. To go to an extreme, if the potential replicator we consider is a whole chromosome, the difference between a ‘successful’ and an unsuccessful chromosome is of no significance, since both are almost bound to be split by crossing-over before the next generation in any case: their ‘fidelity’ is zero.

  This can be put in another way. An arbitrarily defined length of chromosome, or potential replicator, may be said to have an expected half-life, measured in generations. Two kinds of factor will affect this half-life. Firstly, replicators whose phenotypic effects render them successful at their business of propagating themselves will tend to have a long half-life. Replicators with longer half-lives than their alleles will come to predominate in the population, and this is the familiar process of natural selection. But if we set selection pressures on one side, we can say something about the half-life of a replicator on the basis of its length alone. If the stretch of chromosome we choose to define as our replicator of interest is long, it will tend to have a shorter half-life than a shorter replicator, simply because it is more likely to be broken by crossing-over. A very long portion of chromosome ceases to deserve the title of replicator at all.

  A corollary is that a long portion of chromosome, even if successful in terms of its phenotypic effects, will not be represented in many copies in the population. The Y chromosome excepted, and depending upon crossover rates, it seems unlikely that I share any whole chromosomes with any other individual. I assuredly share many small portions of chromosomes with others, and if we choose our portions small enough their likelihood of being shared becomes very high indeed. It is not normally useful to speak of inter-chromosome selection, therefore, since each chromosome is probably unique. Natural selection is the process by which replicators change in frequency in the population relative to their alleles. If the replicator under consideration is so large that it is probably unique, it cannot be said to have a ‘frequency’ to change. We must choose our arbitrary portion of chromosome so that it is small enough to last, at least potentially, for many generations before being split by crossing-over; small enough to have a ‘frequency’ that can be changed by natural selection. Is it possible to choose it too small? I shall return to this question below, after approaching it from another direction.

  I shall make no attempt to specify exactly how long a portion of chromosome can be permitted to be before it ceases to be usefully regarded as a replicator. There is no hard and fast rule, and we don’t need one. It depends on the strength of the selection pressure of interest. We are not seeking an absolutely rigid definition, but ‘a kind of fading-out definition, like the definition of “big” or “old”;. If the selection pressure we are discussing is very strong, that is if one replicator makes its possessors very much more likely to survive and reproduce than its alleles do, the replicator can be quite large and still be usefully regarded as a unit that is naturally selected. If, on the other hand, the difference in survival consequences between a putative replicator and its alleles is almost negligible, the replicators under discussion would have to be quite small if the difference in their survival values is to make itself felt. This is the rationale behind Williams’s (1966, p. 25) definition: ‘In evolutionary theory, a gene could be defined as any hereditary information for which there is a favorable or unfavorable selection bias equal to several or many times its rate of endogenous change.’

  The possibility of strong linkage disequilibrium (Clegg 1978) does not weaken the case. It simply increases the size of the chunk of genome that we can usefully treat as a replicator. If, which seems doubtful, linkage disequilibrium is so strong that populations contain ‘only a few gametic types’ (Lewontin 1974, p. 312), the effective replicator will be a very large chunk of DNA. When what Lewontin calls lc, the ‘characteristic length’ (the ‘distance over which coupling is effective’), is only ‘a fraction of the chromosome length, each gene is out of linkage equilibrium only with its neighbors but is assorted essentially independently of other genes farther away. The characteristic length is, in some sense, the unit of evolution since genes within it are highly correlated. The concept is a subtle one, however. It does not mean that the genome is broken up into discrete adjacent chunks of length lc. Every locus is the center of such a correlated segment and evolves in linkage with the genes near it’ (Lewontin 1974).

  Similarly, Slatkin (1972) wrote that ‘It is clear that when permanent linkage disequilibrium is maintained in a population, the higher order interactions are important and the chromosome tends to act as a unit. The degree to which this is true in any given system is a measure of whether the gene or the chromosome is the unit of selection, or, more accurately, what parts of the genome can be said to be acting in unison’. And Templeton et al. (1976) wrote that ‘… the unit of selection is a function in part of the intensity of selection: the more intense the selection, the more the whole genome tends to
hold together as a unit’. It was in this spirit that I playfully contemplated titling an earlier work The slightly selfish big bit of chromosome and the even more selfish little bit of chromosome (Dawkins 1976a, p. 35).

  It has often been suggested to me that a fatal objection to replicator selectionism is the existence of within-cistron crossing-over. If chromosomes were like bead necklaces, the argument runs, with crossing-over always breaking the necklace between beads and not within them, you might hope to define discrete replicators in the population, containing an integral number of cistrons. But since crossover can occur anywhere (Watson 1976), not just between beads, all hope of defining discrete units disappears.

  This criticism underestimates the elasticity that the replicator concept is permitted by the purpose for which it was coined. As I have just shown, we are not looking for discrete units, but for pieces of chromosome of indeterminate length which become more or less numerous than alternatives of exactly the same length. Moreover, as Mark Ridley reminds me, most within-cistron crossovers are, in any case, indistinguishable in their effects from between-cistron crossovers. Obviously, if the cistron concerned happens to be homozygous, paired at meiosis with an identical allele, the two sets of genetic material exchanged in a crossover will be identical, and the crossover might just as well never have happened. If the cistrons concerned are heterozygous, differing at one nucleotide locus, any within-cistron crossover that occurs ‘north’ of the heterozygous nucleotide will be indistinguishable from one at the northern boundary of the cistron; any within-cistron crossover ‘south’ of the heterozygous nucleotide will be indistinguishable from one at the southern boundary of the cistron. Only if the cistrons differ at two loci, and the crossover occurs between them, will it be identifiable as a within-cistron crossover. The general point is that it does not particularly matter where crossovers occur in relation to cistron boundaries. What matters is where crossovers occur in relation to heterozygous nucleotides. If, for instance, a sequence of six adjacent cistrons happens to be homozygous throughout an entire breeding population, a crossover anywhere within the six will be exactly equivalent in effect to a crossover at either end of the six.

  Natural selection can cause changes in frequency only at nucleotide loci that are heterozygous in the population. If there are large intervening chunks of nucleotide sequences that never differ among individuals, these cannot be subject to natural selection, for there is nothing to choose between them. Natural selection must focus its attention on heterozygous nucleotides. It is changes at the single nucleotide level that are responsible for evolutionarily significant phenotypic changes, although of course the unvarying remainder of the genome is necessary to produce a phenotype at all. Have we, then, arrived at an absurdly reductionistic reductio ad absurdum? Shall we write a book called The Selfish Nucleotide? Is adenine engaged in a remorseless struggle against cytosine for possession of locus number 30004?

  At the very least, this is not a helpful way to express what is going on. It becomes downright misleading if it suggests to the student that adenine at one locus is, in some sense, allied with adenine at other loci, pulling together for an adenine team. If there is any sense in which purines and pyrimidines compete with each other for heterozygous loci, the struggle at each locus is insulated from the struggle at other loci. The molecular biologist may, for his own important purposes (Chargaff cited in Judson 1979), count adenines and cytosines in the genome as a whole, but to do so is an idle exercise for the student of natural selection. If they are competitors at all, they are competitors for each locus separately. They are indifferent to the fate of their exact replicas at other loci (see also Chapter 8).

  But there is a more interesting reason for rejecting the concept of the selfish nucleotide, in favour of some larger replicating entity. The whole purpose of our search for a ‘unit of selection’ is to discover a suitable actor to play the leading role in our metaphors of purpose. We look at an adaptation and want to say, ‘It is for the good of …’. Our quest in this chapter is for the right way to complete that sentence. It is widely admitted that serious error follows from the uncritical assumption that adaptations are for the good of the species. I hope I shall be able to show, in this book, that yet other theoretical dangers, albeit lesser ones, attend the assumption that adaptations are for the good of the individual organism. I am suggesting here that, since we must speak of adaptations as being for the good of something, the correct something is the active, germ-line replicator. And while it may not be strictly wrong to say that an adaptation is for the good of the nucleotide, i.e. the smallest replicator responsible for the phenotypic differences concerned in the evolutionary change, it is not helpful to do so.

  We are going to use the metaphor of power. An active replicator is a chunk of genome that, when compared to its alleles, exerts phenotypic power over its world, such that its frequency increases or decreases relative to that of its alleles. While it is undoubtedly meaningful to speak of a single nucleotide as exerting power in this sense, it is much more useful, since the nucleotide only exerts a given type of power when embedded in a larger unit, to treat the larger unit as exerting power and hence altering the frequency of its copies. It might be thought that the same argument could be used to justify treating an even larger unit, such as the whole genome, as the unit that exerts the power. This is not so, at least for sexual genomes.

  We reject the whole sexual genome as a candidate replicator, because of its high risk of being fragmented at meiosis. The single nucleotide does not suffer from this problem but, as we have just seen, it raises another problem. It cannot be said to have a phenotypic effect except in the context of the other nucleotides that surround it in its cistron. It is meaningless to speak of the phenotypic effect of adenine. But it is entirely sensible to speak of the phenotypic effect of substituting adenine for cytosine at a named locus within a named cistron. The case of a cistron within a genome is not analogous. Unlike a nucleotide, a cistron is large enough to have a consistent phenotypic effect, relatively, though not completely, independently of where it lies on the chromosome (but not regardless of what other genes share its genome). For a cistron, its sequential context vis-à-vis other cistrons is not overwhelmingly important in determining its phenotypic effect in comparison with its alleles. For a nucleotide’s phenotypic effect, on the other hand, its sequential context is everything.

  Bateson (1981) expresses the following misgiving about ‘replicator selection’.

  A winning character is defined in relation to another one while genetic replicators are thought about in absolute and atomistic terms. The difficulty is brought home if you ask yourself, what exactly is Dawkins’ replicator? You might answer: ‘That bit of genetic material making the difference between the winning and losing characters.’ You would have stated that a replicator must be defined in relation to something else. Alternatively, your reply might be: ‘A replicator consists of all the genes required for the expression of the surviving character’. In that case you are saddled with a complex and unwieldy concept. Either way your answer would show how misleading it is to think of replicators as the atoms of evolution.

  I certainly would join Bateson in rejecting the second of his two alternative answers, the unwieldly one. The first of his alternatives, on the other hand, expresses exactly my position, and I do not share Bateson’s misgivings about it. For my purposes a genetic replicator is defined by reference to its alleles, but this is not a weakness of the concept. Or, if it is deemed to be a weakness, it is a weakness that afflicts the whole science of population genetics, not just the particular idea of genetic units of selection. It is a fundamental truth, though it is not always realized, that whenever a geneticist studies a gene ‘for’ any phenotypic character, he is always referring to a difference between two alleles. This is a recurring theme throughout this book.

  To prove the pudding, let me show how easy it is to use the gene as a conceptual unit of selection, while admitting that it is only defined by comp
arison with its alleles. It is now accepted that a particular major gene for dark coloration in the peppered moth Biston betularia has increased in frequency in industrial areas because it produces phenotypes that are superior in industrial areas (Kettlewell 1973). At the same time, we have to admit that this gene is only one of thousands that are necessary in order for the dark coloration to show itself. A moth cannot have dark wings unless it has wings, and it cannot have wings unless it has hundreds of genes and hundreds of equally necessary environmental factors. But this is all irrelevant. The difference between the carbonaria and the typica phenotype can still be due to a difference at one locus, even though the phenotypes themselves could not exist without the participation of thousands of genes. And it is the same difference that is the basis of the natural selection. Both geneticists and natural selection are concerned with differences! However complex the genetic basis of features that all members of a species have in common, natural selection is concerned with differences. Evolutionary change is a limited set of substitutions at identifiable loci.

  Further difficulties will be dealt with in the next chapter. Meanwhile, I end this chapter with a small diversion which may be helpful in illustrating the replicator or ‘gene’s-eye’ view of evolution. An appealing aspect of the view appears if we look backwards in time. The replicators in frequent existence today constitute a relatively successful subset of those that have existed in the past. A given replicator in me could, in theory, be traced backwards through a straight line of ancestors. These ancestors, and the environments that they provided for the replicator, can be regarded as the replicator’s ‘past experience’.