Part 4: Sewall Wright
©Paul Handford, 1998
Wright's background was in animal breeding
Unlike Fisher, Wright had a substantial biological background by the time he began his evolutionary theorizing, so we might expect that, at least in some fashion, his approaches and concerns might differ from Fisher's. But although their eventual models of evolutionary populations and processes were indeed markedly divergent in many respects, we must be clear that, in many other regards, they were in explicit, direct, agreement. For example, we should emphasise that Wright was under no misapprehension about the importance of selection, despite his (understandable, if perhaps a little unfair) later reputation, in some quarters, as "Mr. Genetic Drift." But, though Wright indeed started out with a better first-hand biological grounding than did Fisher, it should be noted that this experience was with laboratory and domestic stocks, not with wild populations (that came later in collaborations with Dobzhansky.) His views of evolutionary populations were built directly on these experiences, as we shall see, but he was also markedly influenced by the attitudes of certain other biologists of the day, notably the systematists. On the other hand, he was but little influenced by that group of strongly selection-minded biologists and naturalists that had had such a profound impact on Fisher's thinking. But, like Fisher, he also had his influences (albeit very different) from outside of science.
Like Fisher, Wright expressed his ideas in difficult mathematics, leading to much misunderstanding
We have already noted that Fisher's ideas were much-misunderstood, partly through some opacity of self-expression on his own part and partly through the difficulty of his mathematical treatments. Similar impediments afflicted Wright, and it is one of life's ironies that they should have found themselves on opposite sides of the central debate of mid-century evolutionary theorizing (a debate concerning the relative significances of chance and determinism - drift and adaptationism - which still resonates to this day - see below.) Given all this, and abundant passion on all sides, there was ample opportunity for much misapprehension, misconstruction, and strife. And it happened.
Wright's intellectual origins were, then, in animal breeding. He was much impressed by some work on selection in rats: it showed that, although strong mass selection in a random-mating population could provoke a rapid and marked evolutionary response, this was often accompanied by deleterious side-effects. This result was well-known to the breeders of livestock, and the lesson which Wright drew was that such (what we might call Fisherian) mass selection had seemingly built-in limitations in producing sustained evolutionary advance - it was 'inefficient' in the long run.
Unlike Fisher, Wright came to see gene interactions & processes in small populations as important
Much of his own early work concerned inbreeding, which he investigated experimentally in guinea-pigs. This work persuaded him that genes often worked together in complex ways in producing phenotypic characters - seemingly a far cry from the "bean-bag" views of Fisher. This guinea-pig work, together with his analysis of the development of the Shorthorn cattle breed, also persuaded him of the potential importance of inbreeding in raising degrees of homozygosity, and thus exposing new variability to selection. He was also involved with the problem of correlations among relatives, as had been Fisher, but particularly as applied to systems of inbreeding; in the process, he invented the powerful general analytical technique of path coefficients, which permitted him to do many of the things which Fisher was doing by partitioning variance. He noted that, in his various inbred lines, different characteristics could become fixed by chance. This simple observation was to influence his entire vision of processes in natural populations. His animal breeding work also convinced him of the importance of interactions within and among gene systems.
Wright's evolutionary population was broken up into small sub-populations with inbreeding
Wright's developing ideas were thus clearly distinct from those of Fisher: he came to see populations as being effectively fragmented into small local groups, analogous to his inbred lines (what we now call demes), wherein inbreeding and small population size would increase homozygosity and enhance the importance of genetic drift in generating new combinations of interacting genes. These would then be exposed to natural selection, and thus either spread or decline. This interactive system is known as "the Shifting Balance Theory." Its essential nature is here expressed in Wright's own words, written in a review of Fisher's Genetical Theory of Natural Selection:
Wright searched for models which would give rapid and efficient evolutionary change
It is clear from his other writings that Wright was concerned to find a model which would generate more rapid and efficient adaptive advance and differentiation among geographic populations and among species than that possible under Fisher's mass selection. This, he thought, could be achieved by building in a role for drift at the level of the local population - non-adaptive drift quickly exposes new genetic combinations in local strains which may then be selected quickly, and so spread to characterize whole areas and, eventually perhaps, the whole species - there was no need to wait for the slow and "wasteful" grinding of Fisher's processes. Now, we have seen where he got his ideas about what can happen with inbreeding within local populations: it indeed had its empirical base; but whence came his convictions that nature ought to be swift, efficient and non-wasteful, or that non-adaptive change was indeed commonplace in wild organisms?
The easier question to approach is that of non-adaptationism. It turns out that, at the time when Wright was beginning to formulate his picture of things evolutionary, there was a widespread supposition among systematists that much species-level and sub-specific differentiation, including geographic forms or races, was of a non-adaptive nature and on this, as they say, hangs a tale, and an instructive one it is, too.........
Wright's models included a significant role for non-adaptive change, in marked contrast to Fisher
I have mentioned in my introductory remarks about him (pp. 8-9) that Wright was, in suggesting an important role for random drift in the evolutionary process, not in any way attempting to obscure or diminish the relevance of selection - he was under no misapprehension at all about its important and necessary involvement in things, despite the fact that, through the later arguments he had with Fisher and others, he came to be seen by many as advocating a view of "evolution primarily by drift." But we also need recognise that in suggesting some role for non-adaptive drift atall, he was in no way being perverse, for, as hinted in the preceding paragraph, it was part of the general intellectual climate of the 20s and 30s, especially (though not exclusively) in North America, that much character variation within and among species indeed was non-adaptive. It thus seemed perfectly reasonable to Wright to postulate some role for drift - and it would have seemed in no way a ridiculous or revolutionary notion to many others to do so. Of course, such a proposition did not "seem reasonable" to the likes of Fisher, and, as we have seen, this is because Fisher was operating in a very different intellectual cosmos: the strongly adaptationist tradition of England; he also felt that small, driftable, populations were doomed to stochastic extinction. So it was easy for Wright to envisage non-adaptive processes as probably widespread in nature, and this was very convenient for him, since it permitted the elaboration of a theory which generated an evolutionary engine with the desired properties of efficiency and speed.
It is now fair to demand: why were these systematists so persuaded of the commonness of non-adaptive character differences? The short answer is that they needed them to do their job of classifying organisms! - and this is the really interesting bit. To see why it is so interesting, we must recall some of the background to the whole idea of evolutionary descent itself.
One of the primary stimuli impelling Darwin towards his conviction of evolutionary descent was the overall hierarchical pattern of resemblances to be seen amongst organisms. This hierarchy was "natural" - it was clearly not some mere artefact of the human mind: there really were natural ways to group organisms, and group the the groups, and so on, according to the degrees of character resemblance among them. Why were these resemblances there? Why do organisms show these varying degrees of character-relationship with one another? Darwin's answer to this question was: it is because organisms really are related, by blood, by genealogy; and that the various taxa thus descended one from another through the ages: the more similar forms have a more recent common ancestor, while the more different ones have a common ancestor more distant in time. So: character-resemblance may be used to infer relationships of evolutionary descent - what we now call phylogeny. But there's a catch.
Darwin's other great idea was Natural Selection; he used it to explain the existence of the adaptive fit of organisms to their life's circumstances, and to (at least partially) explain how organisms became modified over time though descent. So there are two reasons why organisms resemble one another: close descent, and adaptation to similar circumstances. Here is the catch: if all characters reflected only the adaptive fit of organisms to their circumstances, how could we trace - no, more - how could we even see - their descent relationships? How would we know who was related to whom? How, in short, could we infer phylogenies? Darwin himself was sensitive to this problem and was careful to make clear that there were indeed such things as non-adaptive characters - how could he not? - for without them, patterns in propinquity of descent could not be seen, by him or by anyone else. You will perhaps recognise this as the problem of homology, and its distinction from analogy - the distinction of organismal similarity through close phylogenetic (descent) relationship from that due to evolutionary convergence.
There has always been this tension between the two sources of organismal similarity: descent and adaptation. Inference of descent requires the existence of at least some characters which are not there by virtue of adaptation. And, paradoxically, if we could not reliably infer descent, we would not be able to recognise true convergences as such! This tension is with us still - it is a persistent source of controversy and difficulty. Most recently, it has manifested itself in claims about the non-adaptiveness of the speciation process itself as part of the collection of ideas that go under the banner of Punctuated Equilibrium.
In the early 20th. century, the strongest advocates of the adaptationist view were Wallace, Lankester and Poulton: those in whose tradition Fisher worked, but the non-adaptationist opposition was strong, and included most 19th. and early 20th. century taxonomists and, remember, the Mendelians. So we can see the matter of the adaptive or non-adaptive nature of geographic variation, or of differences among related taxa as central and pivotal to what "seemed reasonable" in the world of the 20s and 30s. After the Second World War, the adaptationist view became much stronger, eventually almost overwhelming the other view, but it did not die (and for good reason) and it is a prominent element in contemporary evolutionary biology: the question of the adaptiveness or otherwise of much population character differentiation is still actively debated, as is the long-term evolutionary importance of much of such variation. But that is another story.
Wright saw genetic drift, provided by his subdivided populations, as the source of non-adaptive character change
Returning to Wright himself, he was, then, strongly influenced by the non-adaptationist traditions of the taxonomy of his day. The question was: how did such non-adaptive "racial" or species differences arise? There were two main theoretical answers available: correlated characters (allometric relations), and sampling error (drift.) The first is the idea that, if several characters are necessarily, through developmental mechanisms, 'linked' to one another, then adaptive change in one might easily dictate non-adaptive change in the other characters as they necessarily change too. Wright was not a developmental biologist, but a geneticist with experience in sampling error phenomena, so it is perhaps no surprise that he seized on drift as his answer. Such a choice clearly chimed nicely with his other ideas on the structure of populations, and his experiences with guinea-pigs.
Now, Wright very well understood, just as well as did Fisher, that the consequences of drift were usually deleterious, and that small populations were typically doomed, one way or another. Hence his idea of a dynamic balance between the effects of drift and selection - available with a large population that is subdivided into small driftable units.
The notion of a subdivided population sounds entirely plausible - perhaps more so than the panmictic one suggested by Fisher - but we have to be clear that Wright's Shifting Balance model requires more than simple subdivision. It requires that these subdivisions be of a size small enough for drift; that they be sufficiently isolated from one another so as to prevent the powerfully homogenising effects of gene flow to overwhelm drift's effects; yet sufficiently often interconnected as to permit the occasionally-generated superior types to leak out and become selected positively in the whole species population. These are very particular strictures on the model; though experimental work shows that the system can work with the right conditions, it remains unclear to this day just how often, if ever, they are satisfied in nature. There is no doubt that populations often show structure - but whether it is of the requisite sort is still not known. Coyne, Barton & Turelli in 1997 published a general review and critique of the Shifting Balance Theory. Their main theoretical findings are as follows: 1) while sub-populations can move from one "adaptive peak" to another via drift and selection under some conditions, it is often unnecessary to invoke drift; 2) the spread of adaptations from a sub-population to the species as a whole faces two major difficulties. Their evaluation of empirical data shows that: although there is some evidence for separate phases of Wright's model, there are few cases where it provides a better explanation than simple mass (Fisherian) selection. Experiments fail to show that sub-divided populations show any greater response than does mass selection in large populations. Their overall conclusion: "In view of these problems, it seems unreasonable to consider the shifting balance process as an important explanation for the evolution of adaptations."
This has always been a contentious matter, and the argument continues: in 2000, in the journal Evolution (vol. 54: 306 et seq.), you will find further exchanges between Coyne, Barton & Turelli and Goodnight & Wade.
Reasons for the widely discrepant views of Wright & Fisher are rather obscure
So why did Wright favour a view of population phenomena and processes so different from those of Fisher? Well, he certainly did not share Fisher's selectionist enthusiasms (with some empirical justification); he had nothing of Fisher's eugenic concerns; he had no illusions about producing biological analogues of the great Physical Laws. The most that it seems one may say is that Wright very much found the ideas of balance and efficiency appealing; that he held a generally holistic, rather than reductionist, view of things - that there were important emergent properties of complex entities; and that he thought that Nature should be at least as capable as human stockbreeders. Of course, this insistence on efficiency, on a lack of wastage, implies a profoundly teleological view of the whole evolutionary process: after all, how can one assess efficiency and effectiveness except in the context of the achievement of some preconceived goal: if evolutionary trajectories aren't really going anywhere in particular (and Fisher explicitly disavowed such determinism), then the construct of "efficiency" in "getting there" becomes rather meaningless.
The "drift-selection" debate was, and remains, very stimulating to research in evolution
The arguments between Wright and Fisher and their colleagues over the relative importance of drift in natural populations were extremely stimulating of research activity; in this respect at least then, this argument differed dramatically from the preceding Grand Evolutionary Debate, that over Gradualism and Saltationism. Although the adaptation - drift debates were often carried on with an almost religious fervour at times, the effects were more often illuminating than stultifying. In particular, the drift-adaptation debate directly stimulated, especially in Britain initially, but rapidly elsewhere, including North America, the serious attempt to investigate and understand the ecological and evolutionary meaning of genetically-based variation in real wild populations, studied "on their own terms" in their natural fields - an effort that became dubbed "ecological genetics." This was in large part the single-minded construction of a somewhat eccentric Englishman, E.B. Ford. It was from his school and others like it that grew some of the work that was seminal in the developing New Synthesis of Evolutionary Biology (see later) which swiftly left the non-adaptationism of the early systematists, and eagerly embraced the adaptationism that had been so long typical of English natural history-based evolutionary thinking. But the debate continues, for its themes are contemporary ones: we still argue over the extent of drift's effects and the pervasiveness of adaptation in the organic world.
One of Wright's most influential ideas was that of "the adaptive topography"
Before we leave Wright, we must make some mention of his notion of Adaptive Topographies, since it has become, despite its serious ambiguities (albeit not widely acknowledged), and possibly fatal flaws (some of them pointed out by Fisher) one of the more influential images in modern evolutionary thinking. Adaptive Topographies seem to be an intuitively straightforward way of understanding evolutionary dynamics; probably because it seems to offer a simple way of visualising and grasping a very complex issue, it has been very seductive and, therefore, generally accepted. But it may be seriously misleading. Essentially, the notion concerns the idea that fitness is a complex and interactive function (rather than being simply additive) of the alleles present at many loci in any given individual; that the mean fitness of any given population depends in a comparably complex way on the multi-locus allele-frequencies of the whole population; and that all possible population genetic compositions thus generate a multi-dimensional surface representing the array of possible population mean fitnesses - with a complex of peaks and troughs representing higher and lower mean population fitness. Simple theory, as we have seen, shows that selection can only increase mean population fitness - can only make the population climb the fitness peaks. This seems to mean that selection alone cannot guarantee that any given population can, by selection alone, reach the globally highest fitness peak if that population has the ill fortune of beginning with a genetic constitution which places it on the slopes of one of the lower peaks. On such a model, something else is required to jump to the slopes of one of the highest peaks. Wright believed that this was the role of drift: in the little demes of his Shifting Balance Theory, drift and inbreeding (not selection) sooner or later will generate the necessary genetic combinations which, after being positively selected locally, then expanding outwards, can put the population on the track of the loftiest of fitness peaks given by the available genic variance.
This all seems consistent with Fisher's Fundamental Theorem of Natural Selection: that the rate of increase of population mean fitness due to selection is proportional to the population's additive variance in fitness at that time, and that selection will always raise such mean fitness. As we have seen, Fisher's Theorem is a very tricky one, sufficiently subtle and complex as to have fooled a great many of the most able population geneticists down to the present. Apparently it fooled Wright too, and he thought that it helped support his Adaptive Surfaces idea.
The image of an Adaptive Surface is indeed a powerfully compelling one, seemingly of great heuristic value in "visualizing what's going on" in population evolution. Accordingly, representations of such surfaces have appeared in most texts in evolutionary biology, and get mentioned in many, if not most, courses in evolution. The whole thing seems easy to grasp, and it seems to be underwritten by the impeccable mathematics of Fisher's Heavy Theorem. But there is a very real likelihood that the value of the model is severely compromised, because multidimensional surfaces do not have the same geometric properties as the 3-D surfaces which are readily intuited by us mere mortals. Fisher himself, in his persistent criticisms of Wright's ideas, pointed out that, with many dimensions, fitness peaks cease to be disconnected from one another by "fitness valleys or troughs" - rather, there is probably a way, albeit a tortuous one, of getting from anywhere to anywhere else, yet always increasing population mean fitness as you go. Let's say the matter is not settled; but clearly a wise counsel would be to make sure those slippery fitness surfaces are well-salted. For an extensive discussion of all this, and more, see Provine 1986.
Note 1 Coyne, J.A., N.H. Barton & M.Turelli 1997 Perspective: A critique of Sewall Wright's Shifting Balance Theory of Evolution. Evolution 51: 643-671. Return to text location