In other words, Necker cubers commit the same error in this case that Dawkins made in his original use of the metaphor to claim that all organismal selection can also be expressed in terms of gene selection. The metaphor of the Necker cube only applies when the same thing attains equal and alternative representation, not when the Cube's two versions represent genuinely different aspects of a common phenomenon. In Dawkins's original error, something can always be represented at the gene level — but that something counts as bookkeeping, not as the causality of selection, which remains organismal in his standard cases. Similarly, for aggregate species-level characters involved in selection, something can always be represented at the organismic level — but that something, in this case, only involves the composition of the character, not the causal process of selection, which occurs irreducibly at the species level as identified by emergent species-level fitnesses.
A UNIFIED PICTURE OF SPECIES SELECTION In advocating such an expanded role for species selection, we must guard against the ultimate fallacy of claiming too much — for if we turn all forms of species sorting into species selection by verbal legerdemain, then we falsely “win” by definition, but actually lose by an overly imperialistic extension that permits no distinctions [Page 668] and therefore sacrifices all utility as an empirical proposition in science. Fortunately, we can unite both criteria of emergent fitnesses and characters into a unified scheme that establishes two realms of species selection, one more inclusive than the other, but that also identifies a domain of species sorting leading us to reject causation by species selection.
Grantham (1995) has presented such a scheme, reproduced here as Figure 8-4. (I had independently developed the same system, almost with the same picture, in preparing to write this chapter. I mention this not to compromise Grantham's originality or priority in any way — for priority is chronology, and his cannot be gainsaid! — but to express the firm and almost eerie satisfaction that such a “multiple” formulation brings (see Merton, 1965), and to offer this example as proof that the inherent logic of a complex argument often drives independent researchers to a definite and almost ineluctable result-validating in this case the coherence of this “take” on species selection.)
Grantham's diagram circumscribes two realms of species selection, labeled as “hierarchical explanations.” The A realm contains Vrba's firmest examples based on emergent characters, while the B realm adds Lloyd's cases based on the emergent fitnesses associated with aggregate species-level characters. (Vrba, of course, would restrict species selection to the A realm, and ascribe the B realm to the “effect hypothesis” — but everyone seems to agree on the structure and relationships of the realms.) The A realm seems firmer because emergent characters count as adaptations of species, and maintain no expression
8-4. Grantham's 1995 epitome of criteria for invoking species selection in hierarchical models. The A domain includes rare best cases of species selection based on emergent species-level traits. The B domain adds aggregate traits that affect irreducible species-level fitness, and therefore also participate in species selection under the interactor approach. The aggregate and reducible traits of the C domain belong only to organisms and cannot figure in arguments for species selection.
[Page 669]
at lower levels. The B realm seems “looser” because these aggregate species characters can be represented at the organismic level, even though they may also rise by upward causation to become exaptations of species (Gould and Vrba, 1982; Vrba and Gould, 1986; Gould and Lloyd, 1999). But, in any case, the resulting species-level fitnesses are irreducible — so the B realm also represents species selection by standard definitions of selection as a causal process.
The C realm includes cases of species sorting based on aggregate species-level characters that impart only a reducible fitness at the species level — and therefore do not count as species selection. One might add a D realm at the base for cases describable as species sorting, but not associated with any higher-level character, either aggregate or emergent, and therefore not qualifying for consideration as species selection on any definition of species as evolutionary individuals and inter actors. The D realm, which may be quite large, includes several categories, most obviously species sorting based on the higher-level analog of drift — or random differentials in survival and death of species within a clade (see my summary chart, pp. 718–720).
As for any scientific theory, we want, most of all, to be able to make clear and testable distinctions at the crucial boundary between cases that affirm and cases that fall outside the hypothesis under consideration — in this case, between the B and C realms separating irreducible species selection from species sorting reducible to organismic selection. In these formative days for the theory of species selection, we have not yet developed a full set of firm criteria for making these crucial allocations. But let me suggest one guidepost at the outset. Ever since this literature began, astute workers have developed a strong intuition that species sorting based on events of differential birth (speciation rates) will usually represent true species selection, while species sorting based upon differential death (extinction) will often be reducible to organismic level (see Gilinsky, 1981; Arnold and Fristrup, 1982; Vrba and Eldredge, 1984; Grantham, 1995; Gould and Eldredge, 1977; Gould, 1983c).
The source of this intuition — which may turn out to be both wrong, and superficially based — arises from a sense that the extinction of a species may often be adequately explained simply as the summed deaths of all organisms, each for entirely organismal reasons and with no significant contribution from any species-level property. When the last reproductive organism dies, the species becomes extinct. But how could a new species originate without some involvement of population-level features? After all, individual organisms do not speciate; only populations do. But individual organisms die, and the extinction of a species might, at least in principle, represent no more than the summation of these deaths. Grantham expresses this common intuition particularly well (1995, pp. 309-310):
The concept of “speciation rate” cannot be expressed at the organismic level because there is no simple set of organismic traits that determine speciation rate. Rather, a diverse set of organismic and population-level [Page 670] traits (including dispersal ability, population structure, and behavioral compatibility between members of distant populations) affect gene flow and therefore affect speciation rates. Because of the large variety of factors affecting speciation rate . . . the higher level property of “speciation rate” is, at best, extraordinarily difficult to express in organismic terms. The speciation rate of a taxon is irreducible ... A species goes extinct if and only if every individual dies. Whereas differences in speciation rates cannot be expressed in organismic terms, differences in extinction rates will often be reducible (unless population-level traits such as variation matter).
Thus, I suspect that the A and B realms will be heavily populated with cases based on differential speciation, whereas the C realm will feature cases based on differential extinction.
A PERSONAL ODYSSEY Many historians of science, particularly feminists like Donna Harraway (1989, 1991), have forcefully argued that scholars can strike their most effective blows against the myth of pure objectivism by being candid about the interaction of their own autobiographies with their current claims — thus exposing the inevitable (and basically welcome) cultural and psychological embeddedness of science, while opening an author's prejudice both to his own scrutiny, and to the examination of his readers. To do so obsessively or promiscuously in a book of this sort would only clutter a text that would then become even more insufferably egocentric or idiosyncratic — so I have usually desisted (except for some parts of Chapter 1, and the dubious indulgence of my appendix to Chapter 9). But I will follow Harraway's recommendation in this particular case, because no other subject in evolutionary theory has so engaged and confused me, throughout my career, as the definition and elucidation of
species selection. For no other problem have I made so many published mistakes, and undergone so many changes of viewpoint before settling on what I now consider a satisfactory framework. Moreover, my basic reason for current satisfaction rests upon an interesting correction from within my own body of work — and, though I remain heartily embarrassed for not grasping both the inconsistency and the necessary resolution many years earlier, I do take some pleasure in my eventual arrival — and I do think that the story may help to illustrate the intellectual coherence of the framework now proposed in this book.
I made two sequential errors of opposite import. When Niles Eldredge and I first formulated punctuated equilibrium, I was most excited by the insight that trends would need to be reconceptualized as differential success of species, rather than anagenesis within lineages (a theme only dimly grasped in Eldredge and Gould, 1972, but fully developed in Gould and Eldredge, 1977, after much help from Stanley, 1975, and later from Vrba, 1980). I then committed the common fallacy of extending personal excitement too far — and I made the error (as we all did in these early days of “species selection” under punctuated equilibrium) of labelling as species selection any pattern that [Page 671] needed to be described in terms of differential success for species treated (under punctuated equilibrium) as stable entities. In other words, we failed to distinguish selection from sorting, and used the mere existence of sorting at the species level as a criterion for identifying species selection. This definition of species selection must be rejected as clearly wrong — particularly for the invalid “promotion” of several cases properly viewed as effects of causes fully reducible to conventional organismic selection.
In reaction to this previous excess, I then retreated too far in the other direction, by restricting species selection too severely — i.e., only to cases based on characters emergent at the species level (Gould, 1983c; Vrba and Gould, 1986). My later work with Elizabeth Lloyd (Lloyd and Gould, 1993; Gould and Lloyd, 1999) convinced me that emergent character, while properly identifying species selection, only identified a subset of genuine cases, and that emergent fitness, as defended in this section, provided a conceptually broader, and empirically more testable criterion.
In preparing this chapter, I finally realized why I had originally erred in restricting species selection to emergent characters. The source for amending Vrba and Gould (1986) lay in an earlier paper that I had written with Elisabeth Vrba (Gould and Vrba, 1982), particularly in the codification of adaptation (or the origin of a character directly for its current utility) and exaptation (or the cooptation of a preexisting character for an altered current utility) as subsets of the more inclusive phenomenon of aptation (any form of current utility, whatever the historical origin).
We developed this terminology, which has now been widely accepted (see extensive discussion in Chapter 11), in order to make a crucial, but often disregarded, distinction between “reasons for historical origin” and “basis of current utility.” The common conflation of these entirely separate notions has engendered enormous confusion in evolutionary theory — a situation that we documented and tried to correct in our paper (Vrba and Gould, 1986). Hardly any principle in general historical reasoning (not only in evolutionary theory) can be more important than clear separation between the historical basis of a phenomenon and its current operation. For example, crucial components of current utility often arose nonadaptively as spandrels, or side-consequences, of other features actively constructed or evolved (Gould and Lewontin, 1979).
I felt so enlightened by this distinction, and so committed (as a paleontologist and historian) to the special role of historical origin, that I longed to apply this notion to the important concept of species selection. I therefore concluded that we should not speak of species selection unless the character that imparted the relevant fitness could be identified as a true adaptation at the species level — that is, as a feature belonging to the species as a higher-level Darwinian individual, and evolved directly for current utility in promoting the differential success of the species. Emergent species characters qualify as adaptations — and I therefore felt drawn to this narrow criterion for identifying species selection.
In so doing, I committed a basic logical error about the nature of selection. [Page 672] However much I may love history, selection cannot be, and has never been, defined as a historical relationship of character and result. Selection must be defined by present operation, as identified by an observable differential in reproductive success based on the current interaction of a trait of a Darwinian individual with its environment. This definition includes no reference to the historical origin of any relevant trait, which may be either an adaptation or an exaptation. Damuth and Heisler (1988) emphasize this crucial point, with an apt literary flourish at the end to note the irrelevancy of a trait's “aristocracy” (depth of historical origin, or “blue-blooded” continuity) to the hierarchy of selection:
The historical origin of a character is irrelevant to the way that it functions in a selection process. Thus, the issue of whether a character is a group or individual “adaptation,” and whether it has been shaped for its present role by any particular process, is of no importance in the study of the selection mechanism. There may certainly be historical significance in such observations about the origin of characters. Nevertheless, selection evaluates characters in terms of their current relationship to fitness only, not in terms of their history. There is hierarchy in the world of natural selection, but no aristocracy.
Once I recognized the irrelevancy of historical origin to the identification of selection — my only previous rationale for insisting that characters for species selection must be species-level adaptations, and therefore emergent at the species level — I understood that the “emergent character” criterion must be rejected as too restrictive (while correctly identifying the firmest subset of cases for species selection), whereas the “emergent fitness” criterion must be preferred, as not only legitimately broader in scope, but also properly formulated in terms of conventional definitions of selection. In my own preferred nomenclature, species-level characters that are exaptations rather than adaptations can function perfectly well in species selection. Aggregate species-level characters originate as exaptations of species because they arise at the organismal level and pass upwards as effects to the species level. When I mistakenly thought that characters for species selection had to be species-level adaptations, I had excluded aggregate characters (as species-level exaptations), and therefore falsely rejected the emergent fitness approach (see Gould and Lloyd, 1999, for an elaboration of this argument).
In the early 1980's, my own students Tony Arnold and Kurt Fristrup had strongly urged the criterion of emergent fitness upon me, and I well remember my bitter disappointment that I could not convince them to use the restrictive criterion of emergent characters! (I had not yet developed the nomenclature of adaptation and exaptation, and therefore did not yet possess the personal tools for a conceptual resolution.) Thus, my error reflected an active commitment (not a passive consequence of inattention), maintained in the face of an available correction that I now regard as one of the finest papers ever published on the subject (Arnold and Fristrup, 1982). I did not grasp, for another decade, how the terminology developed by Vrba and me also derailed the criterion [Page 673] that we both preferred. To sum up: selection operates on current utilities, and remains agnostic about historical origins in utilizing both adaptations and exaptations with equal facility. Emergent species-level characters will generally count as adaptations, thus clearly available for species selection. But all aggregate species-level characters represent potential exaptations, and therefore become equally available for species selection under the proper criterion of emergent fitness.
I would, however, salvage a lesson from this odyssey of errors. Vrba and I were not wrong in identifying emergent characters as especially interesting (we only erred in deeming them necessary for spec
ies selection). Emergent characters belong exclusively to the species. As adaptations, they become part of the defining cohesion that permits a species to function as an evolutionary individual. Emergent characters thus stand out in designating the style of individuality maintained by species. Aggregate characters, on the other hand, do not clearly define a species as a functional entity (variability, for example, represents an attribute, not an “organ,” of a species) — for aggregate characters belong as much to the component organisms, as to the entire species. Thus, emergent characters are special and fascinating (though not essential to the definition and recognition of species as legitimate Darwinian individuals — see Gould and Lloyd, 1999). Emergent characters do deserve primary consideration in discussions about the structural basis of species both as natural entities in general, and as the basic individuals of macroevolution in particular. But we do not require emergent characters to identify a process of selection.
As a final note, and as one contribution to recognizing the crucial and characteristic differences among Darwinian individuals at the six primary levels of the evolutionary hierarchy, we should suspect that species selection will emphasize exaptations, whereas organismal selection employs a higher relative frequency of adaptations — for species, as more loosely organized in functional terms than organisms, probably possess far fewer emergent characters than organisms. But species “make up” for their relative paucity of adaptations by developing a higher frequency of exaptations. Most of these exaptations derive their raw material from adaptations at the organismal level that cascade upwards to effects at the species level. By joining fewer adaptations (emergent characters) with more exaptations (usually based on aggregate characters), species may become just as rich as organisms in features that can serve as a basis for selection. Species selection may therefore become just as strong and decisive as conventional Darwinian selection at the level of organisms — a process whose power we do not doubt, and whose range we once falsely extended to encompass all of nature.
The Structure of Evolutionary Theory Page 107