The history of gene selectionism has provided a grand intellectual adventure for evolutionary theory — from inception as a manifesto (Williams, 1966), through numerous excursions into pop culture, to valiant (though doomed) attempts to work through the logical barriers and to develop a consistent and workable theory (Dawkins, 1982; Williams, 1992). “Pareto-errors” always inspire a good race. No one really loses — though false theories like gene selectionism must eventually yield — because the resulting clarifications can only strengthen a field, and interestingly fallacious ideas often yield important insights. Without this debate, evolutionary biologists might never have properly clarified the differing roles of replicators and interactors, items for bookkeeping and units of selection. And we might not have developed a consistent theory of hierarchical selection without the stimulus of an opposite claim that genes could function as exclusive causal agents.
Some evolutionists, largely perhaps in fealty to their own pasts, continue to use the language of gene selectionism, even while their revised accounts elucidate and unconsciously promote the hierarchical view (see, in particular, Williams's excellent fourth chapter, in his 1992 book, on selection upon multiple interactors at several levels). Williams, to use a locution of our times, may still be talking the talk of gene selectionism, but he is no longer walking the walk.
Nearly all-major participants in this discussion met at Ohio State University in the summer of 1988. There I witnessed a wonderful little vignette that may serve as an epitome for this section. George Williams presented his new views (the substance of his 1992 book), and surprised many people with his conceptual move towards hierarchy (within his unaltered terminology). I could not imagine two more different personalities in the brief and telling interchange that followed. Marjorie Grene — the great student of Aristotle, grande dame of philosophy, one of the feistiest and toughest people I have ever known, and a supporter of the hierarchical view — looked at Williams and simply said: “You've changed a lot.” George Williams, one of the calmest and most laconic of men, replied: “It's been a long time.”
Logical and Empirical Foundations for the
Theory of Hierarchical Selection
LOGICAL VALIDATION AND EMPIRICAL CHALLENGES
R. A. Fisher and the compelling logic of species selection
R. A. Fisher added a short section entitled “the benefit of species” to the second edition (1958) of his founding document for the Modern Synthesis: The Genetical Theory of Natural Selection (first published in 1930). I do not know why he did so, but the result could not be more favorable for fruitful debate — for Fisher, in these few additional paragraphs, grants to the concept of species selection the two requisite properties for any healthy and controversial theory. [Page 645] In presenting his argument, Fisher proclaims the logic of species selection unassailable, and then denies that this genuine phenomenon could have any substantial importance in the empirical record of evolution on our planet. No situation can be more propitious for useful debate about a scientific theory than validation in logic accompanied by controversy about actual evidence! (Obviously, I do not share Fisher's pessimism about empirical importance, and shall devote this section to explaining why.)
Fisher begins this interpolated passage by stating that Natural Selection (in his upper-case letters), in its conventional Darwinian mode of action among organisms, cannot explicitly build any features for “the benefit of the species” (though organismic adaptation may engender such effects as side consequences). Speaking of instinctual behaviors, Fisher writes (1958, p. 50): “Natural Selection can only explain these instincts in so far as they are individually beneficial, and leaves entirely open the question as to whether in the aggregate they are a benefit or an injury to the species.” But Fisher then recognizes that, in principle, selection among species could occur, and could lead to higher-level adaptations directly beneficial to species. However, lest this logical imperative derail his strict Darwinian commitments to the primacy of organismic selection, Fisher then adds that species selection — though clearly valid in logic and therefore subject to realization in nature — must be far too weak (relative to organismic selection) to have any practical effect upon evolution. I regard the following lines (Fisher, 1958, p. 50) as one of the “great quotations” in the history of evolutionary thought:
There would, however, be some warrant on historical grounds for saying that the term Natural Selection should include not only the selective survival of individuals of the same species, but of mutually competing species of the same genus or family. The relative unimportance of this as an evolutionary factor would seem to follow decisively from the small number of closely related species which in fact do come into competition, as compared to the number of individuals in the same species; and from the vastly greater duration of the species compared to the individual.
Fisher's theoretical validation of the logic behind species selection has never been effectively challenged. Even the most ardent gene selectionists have granted Fisher's point, and have then dismissed species selection from extensive consideration (as did Fisher) only for its presumed weakness relative to their favored genic level, and not because they doubt the theoretical validity, or even the empirical reality, of selection at this higher level. Dawkins (1982, pp. 106-107) has emphasized Fisher's argument about impotence by noting that, at most, species selection might accentuate some relatively “uninteresting” linear trends (like size increase among species in a lineage), but could not possibly “put together complex [organismal] adaptations such as eyes and brains.” Dawkins continues:
When we consider the species ... the replacement cycle time is the interval from speciation event to speciation event, and may be measured in [Page 646] thousands of years, tens of thousands, hundreds of thousands. In any given period of geological time, the number of selective species extinctions that can have taken place is many orders of magnitude less than the number of selective allele replacements that can have taken place ... We shall have to make a quantitative judgment taking into account the vastly greater cycle time between replicator deaths in the species selection case than in the gene selection case.
I strongly support Dawkins's last statement, but will argue (see pages 703–712) that, when we factor punctuated equilibria into the equation, species selection emerges as a powerful force in macroevolution (though not as an architect of complex organismic adaptations).
Williams has also supported Fisher's argument about the logic of higher level selection — even in his gene selectionist manifesto of 1966, where he defends the possibility, but then denies the actuality: “If a group of adequately stable populations is available, group selection can theoretically produce biotic adaptations, for the same reason that genic selection can produce organic adaptations” (1966, p. 110). In his later book, however, Williams becomes much more positive about the importance and reality of selection at several hierarchical levels: “To Darwin and most of his immediate and later followers, the physical entities of interest for the theory of natural selection were discrete individual organisms. This restricted range of attention has never been logically defensible” (1992, p. 38).
The developing literature has added three “classical” arguments against higher-level selection to supplement Fisher's point that cycle times for species must be incomparably long relative to the lives of organisms. All these arguments share the favorable property of accepting a common logic but challenging the empirical importance of legitimate phenomena — a good substrate for productive debate in science, in contrast with the confusion about concepts and definitions that so often reigns. In the rest of this section, I shall summarize the four classical arguments (Fisher's original plus the three additions); note that they can all be effectively challenged at the level for which they were devised (“group,” or interdemic, selection); and then demonstrate that none has any strong force, in principle, against the empirical importance of the still higher level
of species selection.
The classical arguments against efficacy of higher-level selection
The usual arguments against higher-level selection admit that such phenomena must be possible in principle, but deny any meaningful efficacy on grounds of rarity and weakness relative to ordinary natural selection upon organisms.
Weakness (based on cycle time). R. A. Fisher's classical argument: How could species selection exert any measurable effect upon evolution? Rate and effect depend upon numbers and timings of births and deaths — to provide a sufficient population of items for differential sorting. But species endure for thousands or millions of years, and clades count their [Page 647] “populations” of component species in tens, or at most hundreds, and not as the millions or billions of organisms in many populations. How could species selection yield any measurable effect at all (relative to ordinary organismic selection) when, on average, billions of organismic births and deaths occur for each species origin or extinction, and when populations of organisms contain orders of magnitude more members than populations of related species in a clade?
Weakness (based on variability). Hamilton (1971, 1987), in devising arguments against interdemic selection, pointed out that variation among demic mean values for genetically relevant and selected aspects of organismic phenotypes will generally be lower than variation among organisms within a population for the same features. Group selection cannot become a strong force if the mean phenotypes of such higher-level individuals express such limited variation to serve as raw material for selective change.
Instability, as in Dawkins's metaphors of duststorms in the desert and clouds in the sky. This argument has also been most frequently advanced against interdemic selection. Demes, by definition, maintain porous borders because organisms in the same species can interbreed, and members of one deme can therefore, in principle, invade and join another in a reproductive role. If such invasions become frequent and numerous, the deme ceases to function as a discrete entity, and cannot be called an evolutionary individual. Moreover, many demes lack cohesion on their own account, and not only by susceptibility to incursion. Demes may arise as entirely temporary and adventitious aggregrates of organisms, devoid of any inherent mechanism for cohesion, and defined only by the transient and clumpy nature of appropriate habitats that may not even persist for a requisite generation — as in the deme of all mice in a haystack, or all cockroaches in an urban kitchen.
Invasibility from other more potent levels, usually from below. This standard argument, related both to Fisher's first point about cycle time and to the third point about invasibility discussed just above — and classically used to question the potential evolution of altruism by interdemic selection — asks how higher-level selection could possibly become effective if its operation inherently creates a situation where more powerful, lower-level invaders can cancel any result by working in the opposite direction. Suppose that interdemic selection, cranking along at its characteristic pace, increases the overall frequency of altruistic alleles in the entire species because demes with altruists enjoy differential success in competition against demes without altruists. This “leisurely” pace works well enough, but as soon as a selfish mutant arises in any deme with altruists, the advantage of this mutant in organismic selection against the altruistic allele should be so great that the frequency of altruistic genes must plummet within the deme, even while the deme profits in group selection from the presence of altruistic organisms. By Fisher's argument of cycle time, organismic selection of the self-serving should trump interdemic selection for altruism. [Page 648]
Overcoming these classical arguments, in practice for interdemic
selection, but in principle for species selection
Since the bulk of modern debate about higher-level selection has addressed interdemic (or so-called group) selection, the classical arguments have been framed mainly at the level just above our conventional focus upon organisms (though I predict that emphasis will shift to higher levels, particularly to species selection, as macroevolutionary theory develops). All four arguments have force, and do spell impotence for interdemic selection in many circumstances. But, as full generalities, these arguments have failed either to disprove interdemic selection as a meaningful process worthy of consideration at all, or to deny the efficacy of interdemic selection in several important circumstances.
I shall not review this enormous literature here (as my primary concern rests at still higher levels of selection), but I wish to note that two classes of argument grant interdemic selection sufficient strength and presence to count as a potentially major force in evolution (see Wade, 1978; Sober and Wilson, 1998). First, much mathematical modelling (and some experimental work) have adequately shown that, under reasonable conditions of potentially frequent occurrence in nature, group selection can assert its sway against the legitimate power of the four classical objections. In the cardinal example, under several plausible models, the frequency of altruistic alleles can increase within a species, so long as the rate of differential survival and propagation of demes with altruistic members (by group selection) overcome the admitted decline in frequency of altruists within successful demes by organismic selection. The overall frequency may rise within the species even while the frequency within each surviving deme declines.
Second, some well-documented patterns in nature seem hard to explain without a strong component of interdemic selection. Female-biased sex ratios, as discussed by Wilson and Sober (1994, pp. 640-641), provide the classic example because two adjacent levels make opposite and easily tested predictions: conventional organismic selection should favor a 1:1 ratio by Fisher's famous argument (1930); while interdemic selection should promote strongly female-biased ratios to enhance the productivity of groups. Williams (1966) accepted this framework, which he proposed as a kind of acid test for the existence of group selection. He allowed that female-biased ratios would point to group selection, but denied that any had, in fact, been documented, thus validating empirically the theoretical arguments he had developed for the impotence of group selection. Williams concluded (1966, p. 151): “Close conformity with the theory is certainly the rule, and there is no convincing evidence that sex ratios ever behave as a biotic adaptation.” But empirical examples of female-biased ratios were soon discovered aplenty (see Colwell, 1981, and numerous references in Wilson and Sober, 1994, p. 592). Some authors (Maynard Smith, 1987, for example) tried to interpret this evidence without invoking group selection, but I think that all major participants in the discussion now admit a strong component of interdemic selection in such results — and reported cases now number in the hundreds, so this phenomenon [Page 649] cannot be dismissed as an odd anomaly in a tiny corner of nature. Williams now accepts this interpretation (1992, p. 49), writing, “that selection in female-biased Mendelian populations favors males, and that it is only the selection among such groups that can favor the female bias.”
The primary appeal of this admirably documented example lies in the usual finding of only moderate female biases — more than organismic selection could allow (obviously, since any bias at all would establish the point), but less than models of purely interdemic selection predict. Thus, the empirical evidence suggests a balance between adjacent and opposing levels of selection — with alleles for female-biased sex ratios reduced in frequency by organismic selection within demes, but boosted above the Fisherian balance (across species as a whole) because they increase the productivity of demes in which they reside, however transiently, at high frequency.
When we move to the level of species selection, the most important for macroevolutionary theory, we encounter an even more favorable situation. For interdemic selection, the classical contrary arguments had legitimate force, but could be overcome under conditions broad enough to grant the phenomenon considerable importance. For species selection, on the other hand, three of the classical arguments don't even apply in principle — whi
le the fourth (weakness due to cycle time) becomes irrelevant if punctuated equilibrium prevails at a dominant relative frequency.
Proceeding through the classical objections in reverse order, the fourth argument about invasibility from below has strength only in particular contexts — when, in principle, a favored direction of higher-level selection will usually be opposed by stronger selection at the level immediately below. (In the classic case, selfish organismal “cheaters” derail group selection for altruism. Nonetheless, while the argument of invasibility may hold for this particular case — and while, for contingent reasons in the history of science, this example became the paradigm for discussion of interdemic selection — I see no reason in principle for thinking that organismal selection must always, or even usually, oppose interdemic selection. The two levels may operate simultaneously and in the same direction, or at least orthogonally — see Wade's (1978) classic work on this subject.)
In any case, no general reason has been advanced for thinking that organismic or interdemic selection should characteristically oppose species selection — and the argument of invasibility therefore collapses. Of course, organismic selection may operate contrary to the direction of species selection — and must frequently do so, particularly in the phenomenon that older textbooks called “overspecialization,” or the development of narrowly focussed and complex adaptations (the peacock's tail as a classic example) that enhance the reproductive success of individual organisms, but virtually guarantee a decreased geological life span for the species. Other equally common modes of organismic selection, however, either tend to increase geological longevity (improvements in general biomechanical design, for example) or to operate orthogonally, and therefore “beneath the notice” of species selection. Since our best examples of species selection work through differential rates of [Page 650] speciation rather than varying propensities for extinction, and since most organismal adaptations probably don't strongly influence a population's rate of speciation (or at least don't manifest any bias for decreasing the rate), essential orthogonality of the two levels will often prevail in evolution.
The Structure of Evolutionary Theory Page 103
