The Extended Phenotype

  From the viewpoint of this book an animal artefact, like any other phenotypic product whose variation is influenced by a gene, can be regarded as a phenotypic tool by which that gene could potentially lever itself into the next generation. A gene may so lever itself by adorning the tail of a male bird of paradise with a sexually attractive blue feather, or by causing a male bower bird to paint his bower with pigment crushed in his bill out of blue berries. The details may be different in the two cases but the effect, from the gene’s point of view, is the same. Genes that achieve sexually attractive phenotypic effects when compared with their alleles are favoured, and it is trivial whether those phenotypic effects are ‘conventional’ or ‘extended’. This is underlined by the interesting observation that bower bird species with especially splendid bowers tend to have relatively drab plumage, while those species with relatively bright plumage tend to build less elaborate and spectacular bowers (Gilliard 1963). It is as though some species have shifted part of the burden of adaptation from bodily phenotype to extended phenotype.

  So far the phenotypic effects we have been considering have extended only a few yards away from the initiating genes, but in principle there is no reason why the phenotypic levers of gene power should not reach out for miles. A beaver dam is built close to the lodge, but the effect of the dam may be to flood an area thousands of square metres in extent. As to the advantage of the pond from the beaver’s point of view, the best guess seems to be that it increases the distance the beaver can travel by water, which is safer than travelling by land, and easier for transporting wood. A beaver that lives by a stream quickly exhausts the supply of food trees lying along the stream bank within a reasonable distance. By building a dam across the stream the beaver creates a large shoreline which is available for safe and easy foraging without the beaver having to make long and difficult journeys overland. If this interpretation is right, the lake may be regarded as a huge extended phenotype, extending the foraging range of the beaver in a way which is somewhat analogous to the web of the spider. As in the case of the spider web, nobody has done a genetic study of beaver dams, but we really do not need to in order to convince ourselves of the rightness of regarding the dam, and the lake, as part of the phenotypic expression of beaver genes. It is enough that we accept that beaver dams must have evolved by Darwinian natural selection: this can only have come about if dams have varied under the control of genes (Chapter 2).

  Just by talking about a few examples of animal artefacts, then, we have pushed the conceptual range of the gene’s phenotype out to many miles. But now we come up against a complication. A beaver dam is usually the work of more than one individual. Mated pairs routinely work together, and successive generations of a family may inherit responsibility for the upkeep and extension of a ‘traditional’ dam-complex comprising a staircase of half a dozen dams stepping downstream, and maybe several ‘canals’ as well. Now it was easy to argue that a caddis house, or a spider web, was the extended phenotype of the genes of the single individual that built it. But what are we to make of an artefact that is the joint production of a pair of animals or a family? Worse, consider the mound built by a colony of compass termites, a tombstone-shaped slab, one of a vista of similar monoliths all oriented precisely north–south, and rising to a height that dwarfs its builders as a mile-high skyscraper would dwarf a man (von Frisch 1975). It is built by perhaps a million termites, separated by time into cohorts, like medieval masons who could work a lifetime on one cathedral and never meet their colleagues that would complete it. A partisan of the individual as the unit of selection might pardonably ask exactly whose extended phenotype the termite mound is supposed to be.

  If this consideration seems to complicate the idea of the extended phenotype beyond all reason, I can only point out that exactly the same problem has always arisen with ‘conventional’ phenotypes. We are thoroughly accustomed to the idea that a given phenotypic entity, an organ say, or a behaviour pattern, is influenced by a large number of genes whose effects interact in additive or more complex ways. Human height at a given age is affected by genes at many loci, interacting with each other and with dietary and other environmental effects. The height of a termite mound at a given mound-age is, no doubt, also controlled by many environmental factors and many genes, adding to or modifying each others’ effects. It is incidental that in the case of the termite mound the proximal theatre of within-body gene effects happens to be distributed among the cells of a large number of worker bodies.

  If we are going to worry about proximal effects, the genes influencing my height act in ways that are distributed among many separate cells. My body is full of genes, which happen to be identically distributed among my many somatic cells. Each gene exerts its effects at the cellular level, only a minority of genes expressing themselves in any one cell. The summed effects of all these effects on cells, together with similar effects from the environment, may be measured as my total height. Similarly, a termite mound is full of genes. These genes, too, are distributed among the nuclei of a large number of cells. It happens that the cells are not contained in quite such a compact single unit as the cells in my body, but even here the difference is not very great. Termites move around relative to each other more than human organs do, but it is not unknown for human cells to travel rapidly about the body in pursuance of their errands, for instance phagocytes hunting down and engulfing microscopic parasites. A more important difference (in the case of a termite mound, though not in the case of a coral reef built by a clone of individuals) is that the cells in the termite mound are gathered into genetically heterogeneous packages: each individual termite is a clone of cells but a different clone from all other individuals in the nest. This, however, is only a relative complication. Fundamentally what is going on is that genes, compared with their alleles, exert quantitative, mutually interacting, mutually modifying, effects on a shared phenotype, the mound. They do this proximally by controlling the chemistry of cells in worker bodies, and hence worker behaviour. The principle is the same, whether the cells happen to be organized into one large homogeneous clone, as in the human body, or into a heterogeneous collection of clones, as in the termite mound. I postpone until later the complicating point that a termite body itself is a ‘colony’, with a substantial fraction of its genetic replicators contained in symbiotic protozoa or bacteria.

  What, then, would a genetics of termite mounds look like? Suppose we were to do a population survey of compass mounds in the Australian steppe, scoring a trait such as colour, basal length/width ratio, or some internal structural feature—for termite mounds are like bodies with a complex ‘organ’ structure. How could we do a genetic study of such group-manufactured phenotypes? We need not hope to find normal Mendelian inheritance with simple dominance. An obvious complication, as already mentioned, is that the genotypes of the individuals working on any one mound are not identical. For most of the life of an average colony, however, all the workers are full siblings, the children of the primary royal pair of alates who founded the colony. Like their parents the workers are diploid. We may presume that the king’s two sets of genes and the queen’s two sets of genes are permuted throughout the several million worker bodies. The ‘genotype’ of the aggregate of workers may therefore be regarded, in a sense, as a single tetraploid genotype consisting of all the genes which the original founding pair contributed. It is not quite as simple as that, for various reasons, for instance secondary reproductives often arise in older colonies and these may take on the full reproductive role if one of the original royal pair dies. This means that the workers building the later parts of a mound may not be full siblings of those that began the task, but their nephews and nieces (probably inbred and rather uniform, incidentally—Hamilton 1972; Bartz 1979). These later reproductives still draw their genes from the ‘tetraploid’ set introduced by the original royal pair, but their progeny will permute a particular subset of those original genes. One of the things a ‘mound-geneticist’ might look out for, then
, is a sudden change in details of mound-building after replacement of a primary reproductive by a secondary reproductive.

  Ignoring the problem introduced by secondary reproductives, let us confine our hypothetical genetic study to younger colonies whose workers consist entirely of full siblings. Some characters in which mounds vary may turn out to be largely controlled at one locus, while others will be polygenically controlled at many loci. This is no different from ordinary diploid genetics, but our new quasi-tetraploid genetics now introduces some complications. Suppose the behavioural mechanism involved in choosing the colour of the mud used in building varies genetically. (Again, colour is chosen for continuity with earlier thought experiments, although again it would be more realistic to avoid a visual trait, since termites make little use of vision. If necessary, we may suppose the choice to be made chemically, mud colour being incidentally correlated with the chemical cues. This is instructive, for it again emphasizes the fact that our way of labelling a phenotypic trait is a matter of arbitrary convenience.) For simplicity, assume that mud choice is influenced by the diploid genotype of the individual worker doing the choosing, in a simple one locus Mendelian fashion with choice of dark mud dominant over choice of light mud. Then a mound built by a colony containing some dark-preferring workers and some light preferring workers will consist of a mixture of dark and light muds and will presumably be intermediate in overall colour. Of course such simple genetic assumptions are highly improbable. They are equivalent to the simplifying assumptions we ordinarily make when explaining elementary conventional genetics, and I make them here to explain analogously the principles of how a science of ‘extended genetics’ might work.

  Using these assumptions, then, we can write down the expected extended phenotypes, considering mud colour only, resulting from crosses between the various possible founding pair genotypes. For instance, all colonies founded by a heterozygous king and a heterozygous queen will contain dark-building and light-building workers in the ratio 3:1. The resulting extended phenotype will be a mound built of three parts dark mud and one part light mud, therefore nearly, but not completely, dark in colour. If choice of mud colour is influenced by polygenes at many loci, the ‘tetraploid genotype’ of the colony may be expected to influence the extended phenotype, perhaps in an additive way. The colony’s immense size will lead to its acting as a statistical averaging device, making the mound as a whole become the extended phenotypic expression of the genes of the royal pair, manifested via the behaviour of several million workers each containing a different diploid sample of those genes.

  Mud colour was an easy character for us to choose, because mud itself blends in a simple additive manner: mix dark and light mud, and you get khaki mud. It was therefore easy for us to deduce the result of assuming that each worker goes its own way, choosing mud of its own preferred colour (or chemical associated with colour), as determined by its own diploid genotype. But what can we say about a characteristic of overall mound morphology, say basal width/length ratio? In itself, this is not a character that any single worker can determine. Each single worker must be obeying behavioural rules, the result of which, summed over thousands of individuals, is the production of a mound of regular shape and stated dimensions. Once again, the difficulty is one we have met before, in the embryonic development of an ordinary diploid multicellular body. Embryologists are still wrestling with problems of this kind, and very formidable they are. There do appear to be some close analogies with termite mound development. For instance, conventional embryologists frequently appeal to the concept of the chemical gradient, while in Macrotermes there is evidence that the shape and size of the royal cell is determined by a pheromonal gradient around the body of the queen (Bruinsma & Leuthold 1977). Each cell in a developing embryo behaves as if it ‘knows’ where it is in the body, and it grows to have a form and physiology appropriate to that part of the body (Wolpert 1970).

  Sometimes the effects of a mutation are easy to interpret at the cellular level. For instance, a mutation that affects skin pigmentation has a rather obvious local effect on each skin cell. But other mutations affect complex characters in drastic ways. Well-known examples are the ‘homeotic’ mutants of Drosophila, such as the one that grows a complete and well-formed leg in the socket where an antenna ought to be. For a change in one single gene to wreak such a major, yet orderly, change in the phenotype, it must make its lesion rather high in a hierarchical chain of command. By analogy, if a single infantryman goes off his head he alone runs amok; but if a general loses his reason an entire army behaves crazily on a grand scale—invades an ally instead of an enemy, say—while each individual soldier in that army is obeying orders normally and sensibly, and his individual behaviour as he puts one foot in front of the other will be indistinguishable from that of a soldier in an army with a sane general.

  Presumably an individual termite working on a little corner of a big mound is in a similar position to a cell in a developing embryo, or a single soldier tirelessly obeying orders whose purpose in the larger scheme of things he does not understand. Nowhere in the single termite’s nervous system is there anything remotely equivalent to a complete image of what the finished mound will look like (Wilson 1971, p. 228). Each worker is equipped with a small toolkit of behavioural rules, and he/she is probably stimulated to choose an item of behaviour by local stimuli emanating from the work already accomplished, no matter whether he/she or other workers accomplished it—stimuli emanating from the present state of the nest in the worker’s immediate vicinity (‘stigmergie’, Grassé 1959). For my purposes it doesn’t matter exactly what the behavioural rules are, but they would be something like: ‘If you come upon a heap of mud with a certain pheromone on it, put another dollop of mud on the top.’ The important point about such rules is that they have a purely local effect. The grand design of the whole mound emerges only as the summed consequences of thousands of obeyings of micro-rules (Hansell 1984). Particular interest attaches to the local rules that are responsible for determining global properties such as the base length of the compass mound. How do the individual workers on the ground ‘know’ that they have reached the boundary of the ground plan? Perhaps in something like the same way as the cells at the boundary of a liver ‘know’ that they are not in the middle of the liver. In any case, whatever the local behavioural rules may be that determine the overall shape and size of a termite mound, they are presumably subject to genetic variation in the population at large. It is entirely plausible, indeed almost inevitable, that both the shape and size of compass termite mounds have evolved by natural selection, just like any feature of bodily morphology. This can only have come about through the selection of mutations acting at the local level on the building behaviour of individual worker termites.

  Now our special problem arises, which would not arise in the ordinary embryogenesis of a multicellular body, nor in the example of mixing light and dark muds. Unlike the cells of a multicellular body, the workers are not genetically identical. In the example of the dark and light mud, it was easy to suppose that a genetically heterogeneous work-force would simply construct a mound of mixed mud. But a work-force which was genetically heterogeneous with respect to one of the behavioural rules that affected overall mound shape might produce curious results. By analogy with our simple Mendelian model of mud selection, a colony might contain workers favouring two different rules for determining the boundary of the mound, say in the ratio three to one. It is pleasing to imagine that such a bimodal colony might produce a mound with a strange double wall and a moat between! More probably, the rules obeyed by individuals would include provision for the minority to allow themselves to be overruled by majority decisions, so that only one clean-cut wall would emerge. This could work in a similar way to the ‘democratic’ choice of a new nest site in honeybee swarms, observed by Lindauer (1961).

  Scout bees leave the swarm hanging in a tree, and prospect for suitable new permanent sites such as hollow trees. Each scout returns and dances
on the surface of the swarm, using the well-known von Frisch code to indicate the direction and distance of the prospective site that she has just investigated. The ‘vigour’ of the dance indicates the individual scout’s assessment of the virtues of the site. New bees are recruited to go out and examine it for themselves, and if they ‘approve’ they too dance ‘in its favour’ on returning. After some hours, then, the scouts have formed themselves into a few ‘parties’, each one ‘advocating’ a different nest site. Finally, minority ‘opinions’ become even smaller minorities, as allegiances are transferred to majority dances. When an overwhelming majority has been achieved for one site, the whole swarm takes off and flies there to set up home.

  Lindauer observed this procedure for nineteen different swarms, and in only two of these cases was a consensus not soon reached. I quote his account of one of these:

  In the first case two groups of messengers had got into competition; one group announced a nesting place to the northwest, the other to the northeast. Neither of the two wished to yield. The swarm then finally flew off and I could scarcely believe my eyes—it sought to divide itself. The one half wanted to fly to the northwest, the other to the northeast. Apparently each group of scouting bees wanted to abduct the swarm to the nesting place of its own choice. But that was naturally not possible, for one group was always without the queen, and there resulted a remarkable tug of war in the air, once 100 meters to the northwest, then again 150 meters to the northeast, until finally after half an hour the swarm gathered together at the old location. Immediately both groups began again with their soliciting dances, and it was not until the next day that the northeast group finally yielded; they ended their dance and thus an agreement was reached on the nesting place in the northwest [Lindauer 1961, p. 43].