The third argument of instability, while potent for demes, clearly does not apply to species. Sexual species are as well bounded as organisms. Just as genes and cell lineages generally do not wander from organism to organism (whereas organisms often move readily from deme to deme), neither can organisms or demes wander from species to species. The reasons for such tightness of bounding differ between organism and species, but these two evolutionary individuals probably exceed all others in the strength of this key criterion. Species maintain and “police” their borders just as well as organisms do.
The tight bounding of an organism arises from functional integration among constituent parts, including an impermeable outer covering in most cases, and often an internal immune system to keep out invaders. The tight bounding of a species (as classically defined for sexually reproducing eukaryotes) arises from reproductive interaction among parts (organisms), with firm exclusion of parts from any other species. Moreover, this exclusion is actively maintained, not merely passively propagated, by traits that became a favorite subject of study among founders of the Modern Synthesis, especially Dobzhansky and Mayr — so-called “isolating mechanisms.” Species may lack a literal skin, but they remain just as well bounded as organisms in the sense required by the theory of natural selection.
This discussion on the validity and centrality of species as units of selection highlights my only major unhappiness with Wilson and Sober's (1994) superb defense of hierarchical selection, otherwise followed closely in this book. They insist upon functional integration as the main criterion for identifying units of selection (vehicles in their terminology, interactors or evolutionary individuals for others). They insist that the following question “is and always was at the heart of the group selection controversy — can groups be like individuals in the harmony and coordination of their parts” (1994, p. 591).
I do not object to the invocation of functionality itself, but rather to their narrow definition, too parochially based upon the kind of functionality that organisms display. The cohesion (or “functionality”) of species does not lie in the style of interaction and homeostasis that unites organisms by the integration of their tissues and organs. Rather, the cohesion of species lies in their active maintenance of distinctive properties, achieved by joining their parts (organisms) through sexual reproduction, while excluding the parts of other species by evolution of isolating mechanisms.
I much prefer and support Wilson and Sober's more general definition (1994, p. 599): “Groups are real to the extent that they become functionally organized by natural selection at the group level.” Species meet this criterion by evolving species-level properties that maintain their cohesion as evolutionary individuals. The key to a broader concept of “functionality” (that is, the [Page 651] ability to operate discretely as a unit of selection) lies in the evolution of active devices for cohesion, not in any particular style of accomplishment — either the reproductive barriers that maintain species, or the homeostatic mechanisms that maintain organisms.
The second argument of weakness based on lack of sufficient variability among group mean values also doesn't apply to species. Demes of mice from separated but adjacent haystacks may differ so little in group properties that the survival of only one deme, with replenishment of all haystacks by migrants from this successful group, might scarcely alter either allelic frequencies across the entire species, or even average differences among demes. But new species must differ, by definition, from all others — at least to an extent that prevents the reproductive merging of members. Thus, the differential success of some species in a clade must alter — usually substantially — the average properties of the clade (whereas, one level down, the differential success of some demes need not change the average properties of the species very much, if at all).
The first argument about weakness due to long cycle time and small populations therefore remains as the only classical objection with potential force against species selection. And, at first glance, Fisher's argument would seem both potent and decisive. The basic observation cannot be faulted: billions of organism births usually occur for each species birth; and populations of organisms within a species almost always vastly exceed populations of species in a clade. How then could species selection, despite its impeccable logic, maintain any measurable importance when conventional organismal selection holds the tools for such greater strength?
The logic of Fisher's argument cannot be denied, but we must also consult the empirical world. Organismic selection must overwhelm species selection when both processes operate steadily and towards the same adaptive “goal” — for if both levels work in the same direction, then species selection can only add the merest increment to the vastly greater power of organismic selection; whereas, if the two levels work in opposite directions, organismic selection must overwhelm and cancel the effect of species selection.
But the empirical record of the great majority of well-documented fossil species affirms stasis throughout the geological range (see next chapter). The causes for observed nondirectionality within species have not been fully resolved, and the phenomenon remains compatible with the continuous operation of strong organismic selection — for two common explanations of stasis as a central component of punctuated equilibrium include general prevalence of stabilizing selection, and fluctuating directional selection with no overall linear component due to effectively random changes of relevant environments through time. In any case, however, the observation of general stasis seems well established at high relative frequency (I would say dominant, but I also must confess my partisanship).
In this factual circumstance, since change does not generally accumulate through time within a species, organismic selection in the conventional gradualistic and anagenetic mode cannot contribute much to the direction of [Page 652] trends within a clade. Change must therefore be concentrated in events of branching speciation, and trends must arise by the differential sorting of species with favored attributes. If new species generally arise in geological moments, as the theory of punctuated equilibrium holds, then trends owe their explanation even more clearly to higher-level sorting among species-individuals acting as discrete entities with momentary births and stable durations in geological time.
Organismic selection may trump species selection in principle when both processes operate at maximal efficiency, but if change associated with speciation operates as “the only game in town,” then a weak force prevails while a potentially stronger force lies dormant. Nuclear bombs certainly make conventional firearms look risible as instruments of war, but if we choose not to employ the nukes, then bullets can be devastatingly effective. The empirical pattern of punctuated equilibrium therefore becomes the factual “weapon” that overcomes Fisher's strong theoretical objection to the efficacy of species selection.
(This argument provides a second example for the importance of punctuated equilibrium in validating the independence of macroevolutionary theory by failure of pure extrapolationism from microevolutionary dynamics. We saw previously (pp. 604–608) that punctuated equilibrium strongly fosters the argument for species as evolutionary individuals capable of operating as units of selection. We now note that punctuated equilibrium also affirms the potential strength of species selection against a cogent theoretical claim for its impotence.)
In summary, three of four classical arguments against higher-level selection do not apply to species, while the fourth loses its force in a world dominated by punctuated equilibrium. I see no barrier to the cardinal importance of species selection in the history of life.
EMERGENCE AND THE PROPER CRITERION
FOR SPECIES SELECTION
Differential proliferation or downward effect?
This subject and its literature, as I have noted throughout the chapter, have been plagued to an unusual degree by conceptual confusions and disputes about basic definitions and terminology. As an important example, and as many part
icipants have noted (see especially Damuth and Heisler, 1988; and Brandon, 1988,1990), two quite different criteria for the definition of higher-level selection have circulated through the literature. (In most cases, they yield the same conclusion, so this situation has not produced anarchy; but in a few cases, some crucial, they may lead to different assertions, so the situation has fostered confusion.)
In the first approach, one chooses a focal level of analysis (conventionally one of the two lower levels of organism or gene), and then considers the effect of membership in a higher-level group upon fitness values of the chosen lower-level unit. If, for an identical organism, life in one kind of deme yields a [Page 653] fitness different from life in another kind of deme, then selection includes a group effect from the deme level. (We invoke this formulation, for example, if we argue for group selection by showing that organisms in a deme with altruists do better than identical organisms in a group lacking altruists.)
In the second approach, strongly favored here, we hold firm to the classical bare-bones Darwinian definition, but recognize that selection can work on evolutionary individuals at many hierarchical levels. Selection has traditionally been defined as the differential reproductive success of evolutionary individuals based on the fitnesses of their traits in interaction with the environment. Thus, we recognize higher-level selection by the differential proliferation of some higher-level individuals (demes, species, clades) over others — just as we define conventional natural selection by the differential reproductive success of some organisms based on phenotypic traits that confer fitness.
These two approaches often yield concordant results for the obvious reason that differential proliferation of higher-level units (the second criterion) often defines the group effect that influences the fitness of lower-level individuals chosen as a focus (the first criterion). But the two criteria need not correspond, leading to situations where we would identify group selection by one criterion, but deny the same process by the other. For example, under the first criterion of group effects on lower-level fitness, some higher-level properties of groups can influence lower units without causing any differential reproduction of the groups themselves. On this criterion, for example, some biologists have held that frequency dependent selection must be viewed, ipso facto, as an example of group selection — a claim simply incomprehensible under the alternative criterion of differential group proliferation. (The unresolved, and perhaps largely semantic, issue of whether kin selection should be interpreted as a form of group selection, or only an extension of conventional lower-level selection, also presupposes this criterion of group effect upon lower-level fitness — see Wilson and Sober, 1994.)
A predominantly sociological issue has often set preferences between these criteria. Paleontologists, and other students of species selection, myself included, have strongly advocated the criterion of differential reproduction for higher-level individuals as a strict and obvious analog of ordinary natural selection as conventionally understood. Neontologists and students of group selection have generally (though not always) preferred the criterion of “group effect on gene or organismal fitness,” both from fealty to Darwinian traditions for using organisms as a primary focus, and because certain contentious issues, especially the evolution of altruism, have generally been posed in organismal terms — “why can saintly Joe be so nice if he loses reproductive success thereby?”
Three major reasons lead to my strong preference for the criterion of differential proliferation correlated with properties of relevant evolutionary individuals that confer fitness in interaction with their environment. First, we thereby follow standard definitions of selection, which have always been based on causal plurifaction, not on mere effect. Second, why would we ever [Page 654] prefer an elaborate and indirect definition — in terms of effects on something else at a scale far removed from the causal interaction — over a simpler account rooted in the direct result of the causal process itself? Considered in these terms, the criterion of “group effect on organismal fitness” seems downright peculiar. We only entertain such a standard for contingent reasons of history and philosophical preference — the Darwinian tradition for focusing on organisms, and our larger scientific allegiance to reductionism. Third, we can too easily lose the force and location of causality when we study a phenomenon through indirect effects expressed elsewhere, rather than by immediate operation. True, we are supposed to assess the separate effects upon lower-level focal units — from deme membership, or species membership, for example. But since several higher levels may simultaneously affect a lower focal unit, we may not be able to untangle the differences, and we may end up with an account of consequences, rather than causes.
As an obvious example of these pitfalls, I point out that gene-selectionism, with all its fallacies, arises from an erroneous inversion in the criterion of “group effect on lower-level fitness.” One begins with the basic statement that membership in higher-level units affects the fitness of genes. So far, so good. But if one then makes the error of assuming that replicators, rather than interactors, should be units of selection — and then chooses genes as fundamental replicators both by general reductionistic preference, and by allegiance to faithfulness in replication as a necessary criterion — then one becomes tempted to misidentify effects as causes. The gene selectionist then slides down the following slippery slope: why should I talk about higher-level interactors affecting gene fitness? why don't I just consider higher-level interactors as one aspect of the gene's environment? in that case, why should I talk about higher-level interactors as entities at all? environment is environment, however constituted, and whether clumped into interactors housing the genes or not? in fact, why even try to identify the environment's forms of dumpiness? why not, instead, simply average the gene's fitness over all aspects of environment to achieve a single measure of the gene's evolutionary prowess?
This line of argument, as its least attractive feature, relentlessly dissolves causality. We begin with the causal agents of selection — interactors at various hierarchical levels. (Even the most ardent gene selectionists, as I show on pages 631–632, cannot avoid discussing the causal process of selection in terms of these interactors.) We then represent interactors by their effects on genes. Next, we decide to consider interactors only as environments of genes. Then we lose interest in their nature and action because “environmental clumping” (the “expression” of interactors in this view) does not appear to represent an important issue. Finally, we dissolve the interactors entirely by deciding to average the fitness of genes across all aspects of the environment. And, before we notice what we have really done, causality has disappeared.
In a vigorous defense of gene selection against the hierarchical view of Wilson and Sober (1994), Dawkins (tongue-in-cheek to be sure) pretends to be “baffled” by “the sheer, wanton, head-in-bag perversity of the position that they [Page 655] champion” (commentary in Wilson and Sober, 1994, p. 617). Such a sense of strong psychological frustration must arise when you and your opponents seem to be saying the same thing, but in such utterly different ways, and to such radically different effect. Thus, Dawkins presents his gene-selectionist reformulation of Wilson and Sober's Weltanschauung (mine as well, by the way):
Selection chooses only replicators . . . Replicators are judged by their phenotypic effect. Phenotypic effects may happen to be bundled, together with the phenotypic effects of other replicators, in vehicles. Those vehicles often turn out to be the objects that we recognize as organisms, but this did not have to be so ... There did not have to be any vehicles at all . . . The environment of a replicator includes the outside world, but it also includes, most importantly, other replicators, other genes in the same organism and in different organisms, and their phenotypic products.
(Note that I did not exaggerate or caricature in my previous summary; gene selectionists do regard “clumping” into vehicles as beside the point, and they do dissolve these
vehicles — the true units of selection — into “environment” considered as the sum of contexts for any gene.)
Wilson and Sober (1994, p. 641) responded to Dawkins with their own frustration:
Dawkins remains so near, yet so far ... We could not ask for a better summary of the gene-centered view. The question is, are vehicles of selection absent from this account or have they merely been reconceptualized as environments of the genes. The answer to this question is obvious at the individual [organism] level, because Dawkins acknowledged long ago that individuals [organisms] can be vehicles of selection . . . despite the fact that they are also environments of the genes. The answer is just as obvious at the group level... [Dawkins's] passage does not refute the existence of vehicles, but merely assumes that the vehicle concept can be dispensed with and that natural selection can be studied entirely in terms of average genic effects.
Is this brouhaha much ado about nothing? Are the two views — selection on a hierarchy of interactors, and representation of all selective forces in terms of gene fitnesses, with interactors treated as environments of genes — truly equivalent, and our decision just a matter of preference, or a question of psychological judgment about superior sources of insight? Is this twofold choice just another manifestation of Dawkins's old Necker Cube (see p. 640) — a flipping between two equivalent facets of reality, an example of conventionalism in philosophy?
The Structure of Evolutionary Theory Page 104