Species selection came to prominence in the 1970s through the work of Stanley (1975) and Eldredge and Gould (1972).1 These authors sought to 'decouple' macro- from microevolution, arguing that the long-term evolutionary trends revealed in the fossil record are not simply side effects of microevolutionary processes, as was traditionally assumed, but the product of autonomous macroevolutionary forces, such as species selection. The basic idea ofspecies selection is simple. Ifspecies vary, and the variation gives rise to differential extinction/speciation, then some
1 Though Gould (2002) argues that de Vries (1905) was the originator of the species selection concept. Other precursors of the modern discussion include Fisher (1930), who thought that species selection might have played a role in the evolution/maintenance of sexual reproduction, and Dobzhansky (1951), who argued that intra-specific genetic variation was an adaptation of species, permitting rapid evolution in response to environmental stress.
types of species will become more common than others. Therefore, species themselves are the 'focal' units in species selection theory (cf. Arnold and Fristrup 1982).
Eldredge and Gould (1972) presented species selection as a corollary of punctuated equilibrium theory, which says that most morphological evolution happens during the process of speciation; thereafter, species remain in a state of 'stasis'. Evolutionary history thus exhibits a punctuated pattern: millions of years of stasis broken by episodes of rapid evolution, concurrent with the formation of new species by cladogenesis, or lineage splitting.
If punctuated equilibrium is the norm, what explains long-term evolutionary trends, such as the increase in body size in many mammalian lineages, for example, horses, over geological time? Traditional neo-Darwinism attributed these trends to directional selection within species. But if most species are in stasis, this explanation cannot be right. An alternative is that the trends were due to the morphological changes occurring at speciation. When an ancestral horse species split, for example, it might have produced daughter species that were larger than it (a type of'directed mutation' at the species level). This is conceivable, but Eldredge and Gould argued that it is false. The morphological changes that occur at speciation are 'random' with respect to long-term trends, they claimed; a given horse species was as likely to have produced a smaller as a larger daughter. This generalization was dubbed 'Wright's rule', after S. Wright's views on speciation.
Widespread stasis combined with Wright's rule leaves just one explanation of evolutionary trends, Eldredge and Gould argued: species selection. If large horse species survive better than small ones, or leave more daughter species, and if body size is heritable at the species level, then over time there will be a trend towards increased size in the horse clade, despite the absence of any within-species change, and despite the 'randomness' of the morphological changes arising at speciation. Gould and Eldredge (1977) expressed this in the slogan 'punctuated equilibrium + Wright's rule = species selection'.
In their early work, Eldredge and Gould argued that species selection depends on both the ubiquity of stasis and the truth of Wright's rule.2 But there seems no reason why this should be. Even if most species did
2 Gould and Eldredge (1977) say that Wright's rule is a 'precondition' for species selection (p. 148).
undergo phyletic transformation, contrary to punctuated equilibrium, species-level selection could still operate, amplifying or counteracting the within-lineage changes (cf. Williams 1966). Similarly, even if Wright's rule were false, species selection could still occur—directed mutation does not imply the absence of Darwinian selection. In later work, Gould (2002) accepts that species selection is logically independent of these other aspects of punctuated equilibrium theory.
Both conceptually and historically, species selection is related to the 'individuality thesis' of Ghiselin (1974a) and Hull (1978), which says that species are individual entities, extended in space and time, rather than classes of organisms sharing a common property. Gene flow is the 'glue' that binds together the parts of a species into a genuine whole, on this view. The individuality thesis makes it plausible to think of species as entities with life cycles, that is, that are born, reproduce, and die, and thus as the right sorts of entity to function as units of selection. It is no accident that most advocates of species selection also endorse the Ghiselin/Hull view of species.3
Despite the plausibility of the individuality thesis, species are obviously not functionally organized the way other biological individuals, for example, cells, organisms, and insect colonies, are. These entities exhibit a division of labour between their parts, and have mechanisms for suppressing conflict among the parts, ensuring they work for the good of the whole. The same is not true of species. However, this does not necessarily invalidate the species selection concept. Since the turnover rate of species is much slower than that of colonies, organisms, and cells, it is unlikely that there has been sufficient time for species selection to produce comparably complex adaptations, even if it has operated.
Most biologists accept that species selection is possible, but there is considerable disagreement over its empirical significance.4 However, the issue is not solely empirical. There are also disagreements over exactly what species selection amounts to, what types of phenomena it might explain, and what it means for macroevolution to be 'irreducible' to microevolution. It is these conceptual issues that will occupy centre-stage here.
3 Though Williams (1992) is an exception.
4 See for example Maynard Smith (1983), Dawkins (1982), and Gould (2002) for opposing views on this matter.
7.2 GENUINE SPECIES SELECTION VERSUS 'CAUSATION FROM BELOW'
It is clear that species do differ in their fitness, that is, their rate of survival/speciation, but this does not necessarily imply a process of species-level selection. The extraordinary speciosity of Hawaiian drosophilids is apparently due to the physical environment of Hawaii, which is especially conducive to speciation, rather than to any biological properties of the species themselves; so this is not species selection (Hoffman and Hecht 1986). Similarly, Raup et al. (1973) argue that although extinction rates do vary among species, the variation is largely random; so again, variance in species fitness is not indicative of species selection.
Genuine species selection requires that differences in species fitness be caused by differences in a species character, rather than arising for some other reason. This causal ingredient is present in certain putative examples. Recall Jablonski's hypothesis that the average geographic range of late-Cretaceous mollusc species increased as a result of species selection, discussed in Chapter 2. Jablonski (1987) argues that species with larger geographic ranges had a greater tendency to speciate, that is, that differences in range caused differences in fitness. Further, he argues that geographic range was a heritable character, so selection on it produced a cross-generational response.
Conceptually this may seem unproblematic, but in fact there are substantive disagreements over how causality at the species level should be understood. Below I examine a number of proposals for how to distinguish genuine species selection from other causal processes that can lead to variance in species fitness.
In a series of publications, E. Vrba has argued that true species selection is much rarer than its advocates have thought, if indeed it has occurred at all (Vrba 1984, 1989; Lieberman and Vrba 1995). Most of the alleged examples are suspect, for they do not involve 'causation at the focal level of species' (1989 p. 130). Vrba illustrates this point with the example of two African antelope clades, discussed briefly in Chapter 3. Recall that ecological specialists (stenotopes) tended to speciate more frequently than generalists (eurytopes), resulting in an evolutionary trend. But the trend was a side effect of organism-level processes, according to Vrba, which produced greater local differentiation in the stenotopic species, thus restricting gene-flow and enhancing the probability of speciation. So Vrba categorizes this as 'effect macroevolution' rather than genuine species selection.5
According to Vrba, an 'acid test' for genuine species selection is that it must in principle be able to oppose selection at lower hierarchical levels, though it need not do (1989 p. 115). This is an attempt to capture the idea that a higher-level selection process requires 'autonomy' from selection at lower levels. Interestingly, Vrba regards her acid test as intimately linked to the emergent character requirement. If the species character than covaries with species fitness is aggregate rather than emergent, then the required autonomy from lower-level selection cannot obtain, she claims, so species selection cannot be the right verdict.
How should these ideas be assessed? The acid test requirement seems correct. In general, the direction of selection can be different at different hierarchical levels. So it has got to be possible for selection at the species level to counteract the effects of organism-level selection, as Vrba says. But this has nothing to do with emergent characters. As we saw in Chapter 3, whether a given character-fitness covariance is a by-product of lower-level selection is independent, in both directions, of whether the character in question is emergent. So Vrba's 'acid test' is correct, but she is wrong to link it to the emergent character requirement (cf. Grantham 1995; Stidd and Wade 1995).
To illustrate this point, consider the hypothesis that species selection was involved in the evolution/maintenance of sexual reproduction. The rationale for this hypothesis is that selection at the organismic level should favour asexuality, because of the well-known 'two-fold cost of sex' (Maynard Smith 1978). But sexuality might be advantageous at the species level, permitting a more rapid evolutionary response to environmental stress, so species selection could have favoured the sexual over the asexual lineages. This hypothesis clearly satisfies Vrba's acid test—higher- and lower-level selection oppose each other. But the species character subject to selection—'contains sexually reproducing organisms' —is presumably aggregate rather than emergent.
In Chapter 3, we argued that the emergent character requirement stems from conflating the question whether lower-level selection is responsible for higher-level character-fitness covariance, with the question whether some lower-level processes or other are responsible. The
5 Effect macroevolution takes its name from Williams's (1966) distinction between adaptation and fortuitous side effects.
former, not the latter, is the salient question; for on plausible metaphysical assumptions the answer to the latter will always be 'yes'. The distinction between effect macroevolution and species selection, as Vrba draws it, commits this conflation. For Vrba diagnoses effect macroevolution wherever organism-level processes, of whatever sort, can be shown to be 'ultimately' responsible for the differential speciation/extinction. But taken to its logical conclusion, this threatens to make species selection impossible; it is no surprise that Vrba comes close to saying just this.
It follows that at least some of Vrba's examples of effect macroevolution should actually be classified as species selection, including the African antelope example. In that example, Vrba admits that the differences in species fitness are not side effects of differences in organismic fitness; on the contrary, 'organismal and species success are, to a large extent, independent' (Vrba and Gould 1986 p. 224). So the reason that stenotopic species were fitter than eurytopic ones is not that the former contained fitter organisms than the latter. Thus the evolutionary trend towards stenotopy was not a by-product of organismic selection; though it is presumably true that some causal processes at the organismic level were 'ultimately' responsible. So this is a bona fide case of species selection.
The same applies to the example of early-Tertiary marine gastropods, regarded as genuine species selection by some (Hansen 1983; Stanley 1979; Gilinsky 1986; Jablonski 1982, 1986); but not by others (Lieberman 1995; Eldredge 1989). Larval development in gastropods can either be planktotrophic (i.e. the larvae feed in the plankton), or not. The fossil record shows an increased frequency of species with non-planktotrophic larvae over time. This was apparently due to limited dispersal in non-planktotrophs, which restricts gene flow and hence raises the probability of speciation. Again, this should be classified as species selection. Differences in species' fitness were not caused by differences in the fitnesses of their constituent organisms, but by differences in the extent of within-species gene flow. These differences in turn stemmed from differences in mode of larval development, so can ultimately be explained by organism-level processes. But crucially, the lower-level explanation is not a selective explanation.
I turn now to a different proposal for how to distinguish 'real' species selection from its surrogates. A number of authors have suggested that the distinguishing mark lies in differential speciation versus differential extinction. For example, Grantham (1995) argues that 'the concept of ''speciation rate'' cannot be expressed at the organismic level'; so cases of differential speciation are not 'reducible' to the organismic level, while cases of differential extinction are (p. 309n). Similarly, Gould (1982) argues that true species selection is more likely to be by differential speciation, because 'propensity to speciate is not generally a property of individuals' (p. 92); see also Gilinsky (1986) and Sterelny (1996a).
This suggestion seems incorrect. In general, selection at any level can operate either on differences in survival or fecundity; it would be odd if the species level were an exception. Moreover, there are plausible examples of species selection by differential extinction, for example, the hypothesis that species selection favoured sexual reproduction, mentioned above. This hypothesis says that sexual lineages had better survival prospects than asexual ones, not that their intrinsic rate of cladogenesis was higher. So while it is true that extinction occurs when all the organisms in a species die, and so in that sense can be 'expressed in organismic terms', this does not mean that a species' probability of extinction is solely a function of organismic fitnesses, which is the critical issue.
Empirically, it may be true that differential species extinction is often a side effect of organismic selection. For example, in cases of competitive exclusion, where two species compete for resources and one drives the other extinct, the differential extinction is likely to be a side effect of the fitness disadvantage suffered by organisms in one species vis-a-vis organisms in the other. Conversely, most cases of differential speciation are probably not side effects of selection on organisms. There seems no particular reason why a species whose constituent organisms are especially fit should be more or less likely to speciate as a result. Evolutionists have discussed various mechanisms of speciation, none of which imply a link between a species' probability of speciating and the fitnesses of its constituent organisms. However, such a link cannot be ruled out a priori.
To conclude, it may be true that differential speciation is a more promising outlet for genuine species selection, but not for the reasons often alleged. The point is not that speciation is 'something that happens to species' rather than to organisms, nor that extinction rate is 'expressible in organismic terms' while speciation rate is not. Rather the point is that, empirically, a species' probability of speciating is unlikely to be affected by the fitnesses of its constituent organisms, while its probability of going extinct may very well be. So 'causation from below', as that notion was explicated in Chapter 3, is more likely to be the correct verdict in cases of differential extinction, for empirical reasons.
7.3 SPECIES VERSUS AVATARS: DAMUTH'S CHALLENGE
Damuth (1985) argues that the idea of species selection is conceptually flawed, for species are not the right sorts of thing for selection to act on. He notes that most species are subdivided into a number of local populations, which can be widely distributed geographically. As a result, species do not interact with each other, or with the environment, in an ecologically meaningful way, so cannot be subject to selection. Of course, Damuth does not deny that rates of speciation and extinction vary; his claim is that this variation does not reflect a causal process of selection at the species level.
In place of species selection, Damuth recommends the concept of avatar selection. An 'avatar' is a local population of conspecific organisms, integrated into a particular ecological community. So an avatar of a baboon species, for example, might consist of all those baboons inhabiting a particular rain forest. The avatars within a community do interact and compete with each other, so can be subject to a selection process, Damuth argues.
One argument against Damuth's proposal is that avatars are much less well-defined than species. If the world's biota were divided into discrete ecosystems, between which there was little flow of energy or matter, avatars could be quite easily picked out, or individuated. But the world is not like this, which is why the 'reality' of ecosystems is a controversial issue in ecology. By contrast, species can be picked out by the criterion of reproductive isolation in a reasonably determinate way.
This means that fitness is more clearly defined for species than for avatars. Damuth sees no difficulty in treating avatars as fitness-bearing entities, since 'avatars may go extinct or produce other avatars' (1985 p. 1136). But since the local populations of a species normally exchange migrants to some extent, and sometimes fuse with each other completely, it is not obvious how avatar identity is to be judged. What determines when one avatar has become two? For species, by contrast, the onset of reproductive isolation provides a fairly clear criterion for when one species has become two, though there are problem cases.
The choice between species and avatars as focal units reflects the tension between the 'replicator—interactor' conception of evolution and Lewontin's 'heritable variation in fitness' conception, discussed in Chapter 1. For in effect, Damuth's point is that species are not 'interactors' in the sense of Hull (1981), while avatars are. This may be true. But on the other hand, fitness and reproduction are better defined for species than for avatars, so the Lewontin criteria apply more naturally to species. Previously we argued that the Lewontin approach is theoretically superior to the Dawkins/Hull approach. If this is right, then Damuth's revisionist proposal should not be endorsed. It is true that avatars come closer than species to being 'interactors' in Hull's sense; but the fundamental requirement for an entity to be a unit of selection is that it be capable of reproduction, hence form determinate parent—offspring lineages.
In reply to Damuth, Sterelny (1996a) makes the point that selective forces need not be spatially local. At the organismic level this is clear. For example, conspecific organisms are often affected by the same parasites wherever they are found, so inhabit a similar selective environment in that respect. So if there is selection for parasite-resistance, the organisms subject to selection may not be in direct competition, nor in physical contact with each other. The same is true of species selection. Even though species are often distributed across many separate communities, selection can still act at the species level. So long as species vary with respect to a character that causally affects species fitness, then species selection can operate; the selective forces need not be spatially local. Indeed, Sterelny argues that the property of being geographically widespread, or fragmented into many local populations, might itself be a character that affects a species' probability of extinction/speciation.
Damuth's argument is reminiscent of an earlier argument due to Fisher (1930). Fisher argued that species selection is unlikely to be a significant force due to 'the small number of closely related species which in fact do come into competition' (p. 121). But Sterelny's point about the non-locality of selective forces also shows that Fisher's argument is mistaken. Neither competition nor direct contact is needed for a set of species to be subject to a common selection pressure, any more than it is necessary for organisms. All that is necessary, at both levels, is that character differences should cause fitness differences.
The example of sexual reproduction illustrates this point.6 Suppose it is true that asexual lineages have gone extinct because of their reduced capacity to evolve. The set of species whose composition was modified
6 Ironically, Fisher himself allowed that species selection for sexual reproduction might have occurred; see footnote 1 on page 203.
by species selection would then include all those species, sexual and asexual, which faced environmental stresses such that the capacity to evolve affected their survival prospects. Clearly, most of these species will never have come into contact, and may be found anywhere on earth. For environmental deterioration is not restricted to a particular area; nor does it only affect certain taxa. So the species in question would have been geographically widespread and taxonomically diverse.
Was this article helpful?