The Structure of Evolutionary Theory

by Stephen Jay Gould

  3. In contrast with the initial pluralism of Haldane and Huxley (in the book that coined the Modern Synthesis), and of the first editions of founding documents for the second phase of the Synthesis (Dobzhansky's 1937 Genet­ics and the Origin of Species, Mayr's 1942 Systematics and the Origin of Spe­cies, and Simpson's 1944 Tempo and Mode in Evolution), later editions of these three documents encapsulated the hardening of this second phase, as initial pluralism yielded to an increasingly firm and exclusive commitment to adaptationist scenarios, and to natural selection as a virtually exclusive mech­anism of change. Even Sewall Wright's views on genetic drift and shifting [Page 71] balance altered from initial stress upon stochastic alternatives to selection to an auxiliary role for drift (as an impetus for the exploration of new, and poten­tially higher, adaptive peaks) as one aspect of a more inclusive and basically adaptationist process. The complex reasons for this hardening include some empirical documentations of selection, but also involve a set of basically so­cial and institutional factors not based on increasing factual adequacy.

  4. If this hardening on the second Darwinian branch of selection's efficacy reflects a general trend within evolutionary theory, then we should find a sim­ilar Darwinian strengthening (and narrowing) on the other two branches of selection's agency (organismal vs. higher levels) and scope (adequacy to ex­plain the entire geological record by extrapolated microevolution). The tri­umph (for good reasons at the time) of Williams over Wynne-Edwards af­firms this trend for agency, although Williams's important clarification then unfortunately hardened (among epigones) into a dogmatic and a priori re­jection of any hint of group selection. Similarly, the Synthesis's increasing confidence in the exclusivity of gradualistic microevolution deprived paleon­tology of any independent theoretical space, and relegated the field to docu­mentation of an admittedly underdetermined pageant, built by the exclusive agency of microevolutionary principles. Several synthesists even denied the efficacy of differential speciation as an input to macroevolutionary pattern (branding the speciosity of some clades as a “luxury” rather than a crucial in­put to survival and flourishing), and attributed all higher-level change to extensions of gradualistic and adaptive anagenesis within unbranched lineages.

  5. The trends to development, initial pluralism and later hardening of the Modern Synthesis win clearest expression in two sources of data: comparison of statements by leading scientists at the two contrasting centennial cele­brations of 1909 and 1959 (for Darwin's birth and for the publication of the Origin); and by documentation of hardening in the summary statements (and increasingly dogmatic dismissal of alternatives) in leading textbooks for sec­ondary and undergraduate courses in biology.

  Chapter 8: Species as individuals in the hierarchical theory of selection

  1. Selectionist mechanics, in the most abstract and general formulation, work by interaction of individuals and environments (broadly construed to include all biotic and abiotic elements), such that some individuals secure dif­ferential reproductive success as a consequence of higher fitness conferred by some of their distinctive features, leading to differential plurifaction of indi­viduals with these features (relative to other individuals with contrasting fea­tures), thus gradually transforming the population in adaptive ways. But the logic of this statement implies that organisms cannot be the only bio­logical entities that manifest the requisite properties of Darwinian individual­ity — properties that include both vernacular criteria (definite birth and death points, sufficient stability during a lifetime, to distinguish true entities from unboundable segments of continua), and more specifically Darwinian criteria (production of daughters, and inheritance of parental traits by daughters). In [Page 72] particular, and by these criteria, species must be construed not only as classes (as traditionally conceived), but also as distinct historical entities acting as good Darwinian individuals — and therefore potentially subject to selection. In fact, a full genealogical hierarchy of inclusion — with rising levels of genes, cell lineages, organisms, demes, species and clades — features clearly definable Darwinian individuals, subject to processes of selection, at each level, thus validating (in logic and theory, but not necessarily in the potency of actual practice in nature) an extension and reformulation of Darwin's exclusively organismal theory into a fully hierarchical theory of selection.

  2. The validity of the “interactor approach” to defining the mechanics of selection, and the fallacy of the “replicator approach” expose, as logically in­valid, all modern attempts to preserve Darwinian exclusivity of level, but to offer an even more reductionistic account in terms of genes, rather than or­ganisms, as agents — with organisms construed as passive containers for the genes that operate as exclusive agents of natural selection. This false argu­ment, based upon the true but irrelevant identification of genes as faithful replicators, must be replaced by the conceptually opposite formulation of a hierarchical theory of selection, with genes identified as only one valid, and lowest, level in a hierarchy of equally potent, and interestingly different, lev­els of Darwinian individuality: genes, cell lineages, organisms, demes, species and clades. Replication identifies a valid and important criterion for the cru­cial task of bookkeeping or tracing evolutionary change; but replicators can­not specify the causality of selectionist processes, which must be based upon the recognition and definition of interactors with environments. Even Wil­liams and Dawkins, the two leading exponents of exclusive gene selectionism, have acknowledged and properly described the hierarchical causality of inter­action (while proferring increasingly elaborate and implausible verbal de­fenses of gene selection in arguments about parallel hierarchies and Necker cubing of legitimate alternatives rooted in criteria of replication vs. interac­tion). Thus, Williams and Dawkins seem to grasp the validity of hierarchical selection through a glass darkly, while still trying explicitly to defend their in­creasingly indefensible preferences for exclusive gene selectionism.

  3. The logic of hierarchical selection cannot be gainsaid, and even Fisher admitted the consistency, even the theoretical necessity, while denying the empirical potency, of species selection. Fisher based his interesting and powerful argument on his assumption that low N for species in clades (relative to or­ganisms in populations) must debar any efficacy for species selection in a world of continuous and gradualistic anagenesis rooted in organismal selec­tion. However, Fisher's argument, although logically tight, fails empirically because species tend to be stable and directionally unchanging (however fluctuating) during their geological lifetimes, and the theoretically “weaker” force of species selection may therefore operate as the “only game in town” for macroevolution. The arguments for potency of species selection are stron­ger than corresponding assertions for interdemic selection (largely because species actively maintain their boundaries as Darwinian individuals, whereas demes remain subject to breakup and invasion). But, despite these intrinsic [Page 73] weaknesses and problems, interdemic selection has now been empirically val­idated as an important force in evolution — thus strengthening a prima facie case for the even greater importance of species selection in macroevolution.

  4. Two theoretical resolutions and clarifications have established both a sound theoretical basis, and a strong argument for the empirical potency, of species selection as an important component of macroevolution: first, the rec­ognition of differential proliferation rather than downward effect as the most operational criterion for defining and recognizing species selection; second, the acknowledgment that emergent fitnesses under the interactor approach, rather than emergent features treated as active adaptations of the species, constitute the proper criterion for identifying species selection. The former in­sistence upon emergent features (by me and other researchers, and in error), while logically sound and properly identifying a small subset of best and most interesting cases, relegated the subject to infrequent operational utility, and thus to relative impotence. The proper criterion (under the interactor ap­proach) of emergent
fitness universalizes the subject by permitting general identification in the immediacy of the current mechanics of selection, and not requiring knowledge — often unavailable given the limits of historical ar­chives — of adaptive construction and utility in ancestral states.

  5. The six levels recognized for convenience, and not accompanied by any claim of completion or exclusivity — gene, cell lineage, organism, deme, spe­cies and clade — feature two important principles that make the theory of hi­erarchical selection so different from, while still in the lineage and tradition of, exclusivistic Darwinian organismal selection. First, adjacent levels may in­teract in the full range of conceivable ways — in synergy, orthogonally, or in opposition. Opposition has been stressed in the existing literature, but only because this mode is easier to recognize, and not for any argument of greater importance in principle. Second, the levels operate non-fractally, with fas­cinating and distinguishing differences in mode of functioning, and relative importance of components, for each level. For example, the different mecha­nisms by which organisms and species maintain their equally strong individu­ality dictate that selection should dominate at the organismal level, while selection, drift, and drive should all play important and balanced roles at the species level.

  6. To cite just one difference (from conventions of the organismal level) for each nonstandard level, and to make the key point about distinctiveness of levels in an almost anecdotal manner: random change may be most promi­nent in relative frequency at the level of the gene-individual; true gene selec­tion also plays an important, if limited, role (largely in the mode that has been given the unfortunate name — for its implication of opposition, almost in ethical terms, to the supposed standard of proper organismal selection — of “selfish DNA”); however, the Dawkinsian argument for exclusivity of genic selection only records the confusion of a preferred level of bookkeeping with an erroneous claim for a privileged locus of selection. Selection among cell-lineages, although ancestrally important in the evolution of multicellular organisms, has largely been suppressed by the organismal level in the interests [Page 74] of its own integrity; failure of this suppression leads to the pyrrhic victory of cell-lineages that we call cancer. Interdemic selection, although once so widely rejected, probably plays an essential role in the evolution of social coopera­tion in general, and not only for such specific phenomena as human altruism. Species-level selection, combined with other species-level properties of drive and drift, establishes the independent basis for a distinctive speciational the­ory and reformulation of macroevolution. The highest level of clade selec­tion, although sometimes operative, may be relatively weak by an extension of Fisher's argument about low N.

  7. I explore the distinctive differences between levels of selection by trying to exemplify and “play out” the detailed disparities in a “grand analogy” be­tween the conventional operation of organismic selection and the relative conceptual novelty of species selection. As an idiosyncratic sample of poten­tial reforms and surprises, consider the following claims: First, the formula­tion of a general taxonomy for sources of change in hierarchically ordered systems, based on a primary distinction of “drive” for directed changes aris­ing within an individual, based on change among lower-level individuals as constituent parts; and “sorting,” with two causally distinct subcategories of “selection” and “drift” for change based on alterations of relative frequencies among individuals at the focal level itself. Second, the recognition, by follow­ing the logic of the analogy, of some strikingly counterintuitive comparisons that become both interesting and revealing upon subsequent reflection — in­cluding the likeness of Lamarckian change, construed as ontogenetic drive at the organismal level, with standard anagenetic transformation as organismal drive at the species level (transformation by directional change of constituent parts of a higher-level individual, in this case the organisms of a species); this similarity may also highlight the rather different reasons for general unimpor­tance of both levels of drive — Lamarckism for the well-known reason of the­oretical non-occurrence in a Mendelian world, and anagenesis based on the controversial claim for its evident plausibility in theory (as a basic Darwinian process), but rarity in fact, given the dominant relative frequency of punctu­ated equilibrium. Third, the establishment of a framework for distinguishing directional speciation as a form of reproductive drive (inherently biased dif­ferences in autapomorphies of descendant species vs. ancestral states) from true species selection as a higher order sorting among daughter species that arise with phenotypic differences randomly distributed about parental means. I believe that we have missed this crucial distinction because the ana­log of directional speciation at the organismal level — drives induced by muta­tion pressure — occur so rarely (for conventional reasons of organismal selec­tion's power to suppress them) that we haven't considered the greater potency of analogous processes at other levels. Fourth, the importance of testing “Wright's Rule” — the claim that speciation is random with respect to the di­rection of evolutionary trends within clades — because the major alternative of directional speciation as the cause of trends holds such potential power at the species level, whereas its analog (drives of mutation pressure) assumes so little importance at the organismal level. Fifth, the potentially far greater [Page 75] importance of drift (both species drift and founder drift) vs. selection as a mech­anism of sorting at the species level, but not at the organismal level, where se­lection predominates in standard formulations. Sixth, the identification of an intrinsically, and probably unbreakable (in most cases), negative correlation between speciation and extinction propensities as the primary constraint op­erating to prevent the takeover of life by a few megaclades (which might dom­inate by enhancing speciation while retarding extinction among constituent species — or perhaps the Coleoptera have prevailed by this means). Seventh, the recognition that the organismal level operates uniquely in securing the in­tegrity of its individuals by devices (physiological homeostasis among organs, and spatial bounding by an external surface) that “clear out” both drive from below and drift at its own level as mechanisms operating at high relative fre­quency — thus leaving selection in its most dominant position at this level. Perhaps our Darwinian prejudice for regarding selection as by far the most effective, or virtually the only important, process of evolutionary change arises more from the parochialism of our organismal focus (given our own personal residence in this category) than from any universal characterization of all levels in evolution.

  Chapter 9: Punctuated equilibrium and the validation of macroevolutionary theory

  1. The clear predominance of an empirical pattern of stasis and abrupt geological appearance as the history of most fossil species has always been acknowledged by paleontologists, and remains the standard testimony (as documented herein) of the best specialists in nearly every taxonomic group. In Darwinian traditions, this pattern has been attributed to imperfections of the geological record that impose this false signal upon the norm of a truly gradualistic history. Darwin's argument may work in principle for punctuational origin, but stasis is data and cannot be so encompassed.

  2. This traditional argument from imperfection has stymied the study of evolution by paleontologists because the record's primary (and operational) signal has been dismissed as misleading, or as “no data.” Punctuated equilib­rium, while not denying imperfection, regards this signal as a basically accu­rate record of evolution's standard mode at the level of the origin of species. In particular, before the formulation of punctuated equilibrium, stasis had been read as an embarrassing indication of absence of evidence for the de­sired subject of study — that is, of data for evolution itself, falsely defined as gradual change — and this eminently testable, fully operational, and intellec­tually fascinating (and positive) subject of stasis had never been subjected to quantitative empirical study, a situation that has changed dramatically during the last 25 years.

  3. The key empirical ingredients of
punctuated equilibrium — punctuation, stasis, and their relative frequencies — can be made testable and defined operationally. The theory only refers to the origin and development of species in geological time, and must not be misconstrued (as so often done) as a claim for true saltation at a lower organismal level, or for catastrophic mass extinction [Page 76] at a higher faunal level. Punctuation must be scaled relative to the later duration of species in stasis, and we suggest 1-2 percent (analogous to human gestation vs. the length of human life) as an upper bound. Punctuated equi­librium can be distinguished from other causes of rapid change (including anagenetic passage through bottlenecks and the traditional claim of imperfect preservation for a truly gradualistic event) by the criterion of ancestral sur­vival following the branching of a descendant. Punctuations can be revealed by positive evidence (rather than inferred from compression on a single bed­ding plane) in admittedly rare situations, but not so infrequent in absolute number, of unusual fineness of stratigraphic resolution or ability to date the individual specimens of a single bedding plane. Stasis is not defined as abso­lute phenotypic immobility, but as fluctuation of means through time at a magnitude not statistically broader than the range of geographic variation among modern populations of similar species, and not directional in any pre­ferred way, especially not towards the phenotype of descendants. Punctuated equilibrium will be validated, as all such theories in natural history must be (including natural selection itself), by predominant relative frequency, not by exclusivity. Gradualism certainly can and does occur, but at very low relative frequencies when all species of a fauna are tabulated, and when we overcome our conventional bias for studying only the small percentage of species quali­tatively recognized beforehand as having changed through time.