The Structure of Evolutionary Theory

by Stephen Jay Gould

  In describing exceptions and fuzzinesses in the application of these vernac­ular criteria to organisms, and acknowledging that species face the same dif­ficulties of definition, Hull (1976, p. 177) wrote: “However, exactly the same questions arise for both. If organisms can count as individuals in the face of such difficulties, then so can species.” But Hull assumed that these common problems plague species far more intensely than they threaten organisms. I would suggest that the opposite situation may prevail in nature: species may be even better individuated than organisms when punctuated equilibrium ap­plies (and we consider species at their appropriate scales of geological time). This issue unites these two chapters in a crucial link between the theory of punctuated equilibrium (Chapter 9) and the classical debate about “units” or “levels” of selection (Chapter 8) — a conjunction that underlies my views on the importance and validation of macroevolutionary theory.

  Interestingly, albeit through a glass darkly, Hull (1976) grasped the logical link between the phenomenology of punctuated equilibrium and the defini­tion of species as individuals in his first important paper on this subject — even though he had not, by this time, encountered our empirical and theoretical arguments for such a pattern (Eldredge, 1971; Gould and Eldredge, 1971; Eldredge and Gould, 1972). (In his more inclusive review of 1980, Hull then explicitly joined our particular claims to the defense of species as individuals.) Hull begins by stating the problem (1976, p. 185): “Earlier I described indi­viduals as reasonably discrete, spatiotemporally continuous and unitary enti­ties individuated on the basis of spatiotemporal location rather than similar­ity of some kind. But one might object that species lack these characteristics. For example, in most cases new species arise gradually.”

  Hull then recognized that some neontological models of speciation acceler­ate the rate of branching relative to the supposedly standard rate of anagene­sis within species — and that such an acceleration will sharpen the definability of species by the criterion of discrete birth: “But there are processes in nature which serve to narrow the boundaries between ancestral and descendant spe­cies ... The end result is that the number of organisms intermediate between the ancestral and descendant species is reduced considerably” (1976, p. 185). Finally, Hull stresses the important point that all individuation, at any ap­propriate scale, entails some fuzziness at the boundaries — and that species therefore need not be construed as “worse” individuals than bodies (1976, p. 185). “If processes similar to those just described are common in nature, then the boundaries between ancestral and descendant species can be narrowed [Page 608] considerably, though not to a one-dimensional Euclidean line. But, of course, the replication of organisms does not happen instantaneously either. If absolutely discrete boundaries are required for individuals, then there are no individuals in nature. It is only our relative size and duration which make the boundaries between organisms look so much sharper than those between species.”

  But, to continue the Euclidean metaphor, and using an appropriate ruler with (say) a minimally noticeable geological increment equal to 10,000 years, the boundaries of many species do become momentary under punctuated equilibrium. Stasis persists for a long run of increments. At a commonly ob­served duration of 5 to 10 million years for marine invertebrate species in the fossil record (Raup, 1985; Stanley, 1985), one thousand increments of stasis would represent the geometry of a species lifetime, while even a million for the much shorter average duration of terrestrial mammalian species yields 100 increments. By comparison, many (probably most) events of speciation unroll within a single increment — leading to abrupt and momentary origin at geological scales, and the right-angle convention that has become stan­dard for plotting the emergence of species under punctuated equilibrium (see Fig. 8-3).

  Criteria for evolutionary individuality

  The vernacular criteria discussed above provide necessary, but obviously insufficient, conditions for identifying an entity as an evolutionary “individual” with the capacity to act as a causal agent in a process of Darwinian selection. Most unambiguous vernacular individuals cannot operate as Darwinian ac­tors. The earth, for example, surely merits designation as a well-defined indi­vidual — with a specifiable birth (perhaps attended by some initial fuzziness as a primordial fireball), sufficient stability over billions of years (including enough climatic homeostasis to provide a stage for the history of life), and a forthcoming rapid death (presumably by absorption after the sun burns out some five billion years from now, and expands in diameter at least to the orbit of Jupiter). But the earth remains “infertile” in the crucial Darwinian sense of reproductive potentiality. Planets do not have children, and therefore cannot function as Darwinian individuals.

  I do not cite this example to win an argument by ridicule, but rather to emphasize, once again, that all definitions must be embedded within theories. Mere vernacular individuality does not suffice for identification as a causal actor in Darwinian theory. Evolutionary individuality (or, more strictly, Dar­winian individuality, for different theories of biological change may entail other criteria) requires an additional set of attributes rooted in two features of Darwin's world: the genealogical basis of evolution as a branching tree, and the causal efficacy of selection as the leading process of evolutionary change.

  Reproduction. Darwinian individuals must be able to bear children. Biological evolution is defined as a genealogical process. Darwinian evolution operates by the differential increase of your progeny (or whatever you pass [Page 609] into future generations) relative to the progeny of other individuals within the larger entity of your membership.

  Inheritance. Your children must, on average, be more like you than like other parents of your generation — so that evolution may proceed by the differential increase of your own heritable attributes (a requirement of Dar­winian systems, not of all conceivable evolutionary mechanisms). In other words, a principle of inheritance must prevail to permit the tracing of genea­logical patterns — so that the relative reproductive success of ancestors may be assessed.

  Variation. This criterion lies so deeply, and so fundamentally, within the constitution of Darwinism as a revolutionary ontology (and not just as a theory of evolution), that we should, perhaps, not even list variation as a sep­arate criterion, but merely state that this conception underlies all Darwinian thinking. We can hardly imagine a more radical restructuring of the material

  8-3. The original diagram of punctuated equilibrium published in Eldredge and Gould, 1972.

  [Page 610]

  world than the Darwinian shift to variation among members of a popula­tion as an ultimate and irreducible reality (see Mayr, 1982b; Gould, 1996a) — a reversal of the old Platonic notion that essences (approximated empiri­cally by measuring mean values, or by trying to construct an abstract ideal form and then searching for a closest actual embodiment) define the nature of things, and that variation among actual individuals (organisms in pop­ulations, in our most relevant example) can only be construed as “acci­dental,” and judged by relative departure from a materially unattainable ideal.

  Heredity and reproduction work in concert with variation to empower Darwinian selection in genealogically recognizable lineages. The failure of any criterion debars Darwinian evolution as a genealogical process. An ab­sence of reproduction, for example, enforces an oddly limited form of “evolu­tion” restricted to rules (or vagaries) of change within one or a number of in­dividuals, all separately constructed at the outset. Vernacular usage, in fact, does apply the term “evolution” to some nongenealogical systems of this sort — as in the “evolution” of stars along the H-R sequence. But the causes of such systematic temporal changes, unfolding predictably under laws of nature (and not by the contingencies of variational history), differ so pro­foundly from Darwinian evolution that we really should insist upon different words for these maximally disparate modes of history (Gould, 2000a). (A great burden of misunderstanding, in both popular and professional cultures
, must be ascribed to our confusing use of common terminology for such dif­ferent causes. Many interested laypeople feel that biological evolution must unfold by internal necessity just as stars follow their predictable sequences and as galaxies expand following the big bang. And many professional evolu­tionists, suffering from the common affliction of physics envy, and immured in the reductionistic biases of Western scientific culture, have tried to find pro­gressive patterns directly imposed by natural law, where Darwinian contingency actually reigns.)

  An absence of variation also stymies Darwinian change by eliminating the raw material or substrate for any selective mechanism. Evolution in non-varying populations might be treelike and genealogical, but such a process could not be Darwinian. One would have to imagine some very unearthlike way to generate change and diversity — for example, random dispersal of ini­tially identical creatures to varying environments, followed by a Lamarckian or directly inductive process of heritable environmental stamping upon all members of a population.

  Variation without heredity (that is, an absence of correlation between properties of offspring and parents) also stymies Darwinian causality. Selec­tion could occur in a single generation. That is, the biggest or the ugliest might outreproduce all others, or even ruthlessly murder all small and beauti­ful conspecifics — but to what evolutionary avail, if the offspring of survivors then reconstitute all the original variation in original proportions? If varia­tion occurred without correlation to parental constitution, but with inherent bias in a given direction — so that even random mortality produced a trend — [Page 611] then evolution would occur. But we have always labeled such styles of internally directed change as non-Darwinian, with Lamarckism as a primary and historically most influential example.

  Interaction. At each level, the varying individuals of an evolving population (organisms of a deme, demes of species, species of a clade) must interact with the environment in such a way that some individuals achieve relatively greater reproductive success as a causal result of heritable properties manifested by these fitter, and not manifested (or not as effectively expressed) by less fit individuals. This causal claim embodies the key feature of natural selection as an active process. In other words, we must be able to devise a testable causal scenario about why the differential possession of certain heri­table properties yields increased reproductive success.

  These statements inevitably engage the crucial issue of whether we should define selection by this causal interaction of individuals and environments, or by the product actually transmitted to future generations (see next section). The logic of Darwinism dictates that the form of heredity's product — how­ever fascinating in variety across nature's scales — cannot specify agency of selection. Interaction with environment defines agency (Lloyd and Gould, 1993; Gould and Lloyd, 1999) — and agents must be individuals (by both ver­nacular and evolutionary criteria). Some interacting individuals (like genes) usually pass faithful copies to the next generation. Others (like species) pass inevitably modified copies that are still more like themselves than like any other individual at their level. Still others (like sexual organisms) disaggregate their personhood and pass hereditary pieces and particles.

  All these different strategies for hereditary passage permit us to recognize interacting individuals as causal agents of Darwinian selection. The special and unusual tactic of sexual organisms may seem curiously indirect (and we all know the enormous and confusing literature devoted to this subject), but disaggregation works as well as relatively faithful passage, so long as the es­sential Darwinian imperative remains in force: that is, so long as selectively successful individuals manage to bias the next generation with relatively more of their own hereditary material — however that material be passed or packaged. The “goal” of natural selection cannot be defined by faithful repli­cation, but rather by relative “plurifaction,” or “more-making.”* The indi­vidual that plurifies by increasing the percentage of its contribution to the he­redity of the next generation (however the units or items of heredity be constituted) gains in the evolutionary game. And we call the game Darwinian if plurifaction occurs by a causal interaction between properties of the suc­cessful individual and its environment. [Page 612]

  As for the vernacular criteria previously discussed (see pp. 602–603), these specifically evolutionary criteria teach us that organisms are not the only individuals capable of acting as units of Darwinian selection. In particular, and continuing to use species as a “type” example of individuality at higher lev­els, all evolutionary criteria apply to the species as a basic unit of macro-evolution. Species have children by branching (in our professional jargon, we even engender these offspring as “daughter species”). Speciation surely obeys principles of hereditary, for daughters, by strong constraints of homology, originate with phenotypes and genotypes closer to those of their parent than to any other species of a collateral lineage. Species certainly vary, for the defining property of reproductive isolation demands genetic differentiation from parents and collateral relatives. Finally, species interact with the envi­ronment in a causal way that can influence rates of birth (speciation) and death (extinction).

  As a further benefit for thus codifying the criteria of evolutionary individuality, we can immediately cut through the foolishness surrounding several dis­tressingly common, but artlessly and rather thoughtlessly contrived, claims (or, rather, loose metaphors) about the Darwinian character of large items in nature — an attractive idea for many people, particularly for romantics and “new-agers” who yearn for meaningful agency at the highest levels. We can dismiss these claims because the object hypothesized as an agent of selec­tion fails several crucial requirements for designation as an evolutionary in­dividual. As an obvious example, many proponents of the so-called Gaia hypothesis wax poetic about the earth and atmosphere as a homeostatic system robustly balanced by interaction with life to secure and stabilize the conditions required by organisms for diversification and geological persis­tence. Supporters often assume that such functional coherence must make the earth sufficiently like an organism to merit designation as a living entity. Some have even stated that the earth must therefore be recognized as the larg­est and most inclusive product of Darwinian selection — or even that the earth should, in fact, be viewed as a true Darwinian individual. This woolly notion confuses a gut feeling about functionality or adaptive “optimality” (for sup­port of life) with the requirements of Darwinian agency. The earth does not generate children, and did not arise by competitive prowess as the sole survi­vor among defeated brethren (who must have died or been expelled, I suppose, from the solar system long ago). Therefore, among a plethora of other rea­sons, the earth cannot be construed as a Darwinian agent or unit of selection.

  More plausibly, and more interestingly, communities and ecosystems have sometimes been designated as potential units of selection. In this instance, at least, a case could be conceived — for communities do maintain some func­tional coherence, some boundaries (however loose), and some potential for splitting off “daughter” communities with sufficient resemblance to a parent. But I can hardly imagine a set of circumstances that would allow such ecolog­ical units to express enough criteria of individuality to qualify for Darwinian agency. Communities are not (for the most part) genealogically constructed or filiated. They can rarely maintain sufficient coherence or persistence, for constituent species move in and out in relative independence. Williams (1992, [Page 613] p. 55) writes, for example: “The reason must be that communities lack the necessary high rates of reproduction and replacement and especially the high level of heritability required for effective selection. They change their make­up so rapidly that selection among communities must be overwhelmed by en­dogenous change.”

  But these principled exclusions leave us with a rich hierarchy of legitimate biological individuals, all related by the fascinating property of nested inclu­sion within evolution's genealogical system. In appropriate circumstances, bro
ad enough for vital agency in the evolution of life on earth, individuals at many levels — including genes, cell lineages, organisms, demes, species, and clades — can act as units of Darwinian selection. I doubt that we can defend any longer — or as any more than a convenient and parochial preference based on the happenstances of size and duration for a human body — the cen­tral Darwinian conviction that organisms represent the fundamental level of Darwinian individuality, with all other levels either nonexistent, impotent, entirely subservient, or operating only in odd and restricted circumstances.