The Structure of Evolutionary Theory

by Stephen Jay Gould

  If the rationale for a hierarchical theory of selection resides in the expan­sion of “individuality” to several levels of biological organization (see Gould and Lloyd, 1999), then we must specify a set of criteria that any material con­figuration must meet to merit designation as an “individual.” We may, I think (see Gould, 1994), most usefully divide these criteria into two categories: (1) requirements in ordinary language for granting individuality to any configu­ration (vernacular criteria); and (2) requirements in Darwinian theory for re­garding any entity as an evolutionary individual, or potential agent of selec­tion (evolutionary criteria). (I trust that, despite a traditional ethos contrary to such an admission, all thoughtful and self-introspective scientists no longer feel threatened or disloyal in acknowledging that all definitions must be the­ory laden — see Kuhn, 1962, for the classic statement.)

  We must also resolve one other terminological confusion before listing the criteria of individuality. What word shall we use as a general term for the dis­crete “thing” that can serve as a unit of selection at various hierarchical lev­els? In an important article, a manifesto for reviving interest in the power of group selection and the validity of the hierarchical model in general, D. S. Wilson and E. Sober (1994) suggest that we use the term “organism” for the generality (and therefore speak of “species organisms,” “gene organisms,” and so forth), while restricting the word “individual” to organic bodies (you, me, the oak tree, and the barnacle) at the conventional level of Darwinian concern. They choose this definition because they emphasize — I would say [Page 601] overemphasize (see Gould and Lloyd, 1999) — functional cohesiveness among their general criteria of “thingness,” a property better captured by “organ­ism” than by “individual” in vernacular English.

  I strongly urge the opposite and more conventional solution. This issue, I fully recognize, only concerns words, not the empirical world. But we get so muddled, and waste so much time, when we fail to be clear about words and definitions — especially when various scholars use the same word in different, or even opposite, ways — as in the classic confusion generated when molecu­lar biologists began to use “homology” for the percent of similarity in genetic sequence between two organisms, rather than for the well-established and en­tirely different concept of joint possession due to common ancestry. Fortu­nately, in this case, we evolutionists have apparently managed to persuade our molecular colleagues to respect conventional usage, and to call their im­portant concept “sequence similarity,” or some such.

  No one would create such a muddle on purpose, but this particular confu­sion already exists — and some common ground must therefore be established if we wish to address this growing and important literature without a peren­nial need to stop, translate, and bear linguistic idiosyncrasies continually in mind whenever we read a paper. At the moment, most authors use “organ­ism” for the Darwinian body (me and thee), and “individual” for the general­ized unit of selection at any hierarchical level — while others (like Wilson and Sober) employ reversed definitions.

  I strongly urge the former course — organisms as conventional vernacular bodies, and individual for the generalized term — for two reasons. First, this decision represents more common usage, both in vernacular English and among biologists. (Several academic departments include the phrase “organismic biology” in their title to defend a continuing focus on entire bodies against molecular claims for hegemony. But if genes are organisms as well, this ploy will not work!) Second, the technical definition of an “individual” in academic philosophy, and the spread of this terminology in the large litera­ture inspired by Ghiselin's arguments for species as individuals (summarized above), grants priority to “individual” as the general term, and “organism” as the restricted body. Ghiselin (1974b, p. 536) clearly defended this usage in his original definition: “In logic, 'individual' is not a synonym for 'organism.' Rather, it means a particular thing.” And Hull (1976, p. 175) explicitly la­beled the application of “organism” to higher-level objects as misleading because vernacular language so strongly equates “organism” with discrete bodies. He then urged “individual” as the general term, as advocated here: “From the point of view of human perception, organisms are paradigm indi­viduals. In fact, biologists tend to use the terms 'organism' and 'individual' interchangeably. Thus, biologists who wish to indicate the individualistic character of species are reduced to terming them 'superorganisms.' The same claim can be expressed less misleadingly by stating that both organisms and species are individuals.”

  In discussing criteria of individuality, I will focus on species as paradigms for higher-level evolutionary entities for two reasons: (1) because I believe [Page 602] that a proper theory of macroevolution, the central concern of this book, rests upon such a proposition; and (2) because species seem so maximally un­like discrete “things” to many biologists, thus making the correction of this false impression a prerequisite for accepting the full hierarchical model of se­lection. But species can claim no favored status in the hierarchical model, and I use them here only as an example — so that the argument may then proceed to a full set of levels, each characterized by a valid kind of individual acting in a distinctive way.

  Criteria for vernacular individuality

  When we apply the term “individual” in ordinary English, we envisage a set of properties centered upon uniqueness, discreteness, functionality, and cohe­sion considered both spatially and temporally. To be a unique “thing,” and not just a part of a continuum, a named object must clearly begin and end — and must remain its definable self throughout a continuous existence. We may, I think, best summarize this intuition in three criteria. To be called an in­dividual, a material entity must have:

  • a discrete and definable beginning, or birth;

  • an equally discrete and definable ending, or death; and

  • sufficient stability (defined as coherence of substance and constancy of form) during its lifetime to merit continuous recognition as the same “thing.”

  I realize, of course, that the third criterion amalgamates several crucial notions into a single statement. We might specify at least four properties in­volved in our ordinary concept of “sufficient stability” for individuality:

  Change. An individual may undergo some, even substantial, change during its lifetime, but not so much either to become unrecognizable or to en­courage redefinition as a different thing — and, particularly, for temporal se­quences of individuals, not so much alteration that late stages come to resem­ble the next-named individual in a sequence more than the early stages of the same individual.

  Discreteness and cohesion. An individual must maintain clear and coherent boundaries during its lifetime. Parts should not “ooze out” into other individuals, while components of other individuals should not enter and become incorporated.

  Continuity. An individual cannot fade in and out of existence during its lifetime, but must maintain material continuity throughout. Members of classes, on the other hand, are not so constrained, for classes are defined by common properties, not by historical continuity. As Hull (1980) argues, the class of gold atoms does not require continuity or filiation. If all gold disap­peared, its position on the periodic table would remain — and an element later reconstituted with the right atomic particles and requisite properties would still, and legitimately, be called gold. But if all peacocks die, the species-indi­vidual Pavo cristatus disappears forever. Even if some human engineer retained [Page 603] an electronic record of the entire Pavo cristatus genome, and future technology permitted chemical reconstitution from nucleotides, we couldn't call the resurrected creature a member of Pavo cristatus, even if the reconsti­tuted object looked and acted like an extinct peacock of old.

  Functionality, or organization. We expect that, at least in some crucial ways, the parts of an individual will work together so that the individ­ual functions in a distinctive and cohesive way. This criterion, though crucial a
s we will see to the second set of evolutionary criteria, may be the least im­portant (perhaps even dispensable) for vernacular definitions. If a bounded object maintained all the other listed properties, but failed to do anything as an entity (and acted, instead, largely as a repository of separate parts), we would still call the object an individual, however inert and uninteresting.

  Conventional organisms certainly possess all these properties — as well they must, for the bodies of complex animals established our vernacular Western paradigm for the general concept of individuality. Yet note that, even here, at the point of maximal clarity, some fuzziness and indefiniteness plague every criterion. Consider human bodies, the inevitable exemplar of the paradigm. Our lives have reasonably discrete beginnings — but if a true moment could be defined without ambiguity, then our social and political debates about abor­tion would require new terms and engage different issues. Death might seem even more definable and momentary — but, again, fuzziness and ambiguity plague our definitions, leading to complex, and often heart wrenching, medi­cal and legal wrangles. Perhaps our bodies pass the criterion of “sufficient stability” with more clarity. We don't fuse with others, or rise again from the dead (at least in a material world that science can adjudicate). We are cer­tainly designed to operate discretely, even if our actions be dysfunctional. All our chemicals, and most of our cells, undergo periodic replacement — but I re­main myself and continue to look sufficiently like my baby pictures (though not much like my early embryonic form with tail and gill slits!).

  So organisms surely pass muster as individuals. But we encounter prob­lems, including several classic issues subject to endless discussion in the litera­ture, when we try to assign individuality at other (particularly higher) levels of the organic hierarchy. For example, the standard objection to interdemic selection (see pp. 648–652) holds that too many demes fail the criterion of “sufficient stability” — for they may not persist long enough to matter in evo­lution, and their borders may be too “leaky” as organisms move in and out in the absence of reproductive isolation between the parts (organisms) of differ­ent demes. All too frequently, demes may operate, in Dawkin's apt metaphor (1976, p. 36), “like clouds in the sky or dust storms in the desert . . . tempo­rary aggregations or federations.”

  Defenders of classical “group” (i.e. interdemic) selection recognize these problems of course, and all workable models have been purposely con­structed to overcome such objections by specifying conditions that will per­mit demes to fulfill the defining criteria listed above. In fact, the classical em­pirical issue of our literature on group selection asks whether demes can [Page 604] “hold together” long and discretely enough so that the differential prolifera­tion and survival of some denies vs. others can propel the general increase of an “altruistic” allele (promoting demic success), even while the allele's fre­quency declines within groups as “selfish” alternatives prevail in conven­tional natural selection among bodies. If demes can “hold together” by this operational criterion of evolutionary outcomes, then they possess “sufficient stability” to be regarded as individuals on functional grounds within selec­tionist theory.

  Traditionally, biologists have not been willing to imbue species with these requisite criteria for individuality. Species, in an argument dating to both Lamarck and Darwin, have often been construed as mere names of conve­nience attached to segments of evolving continua without clear borders. Un­der this gradualistic and anagenetic view (see Fig. 8-1), a species near the end of its arbitrary existence must be phenotypically more similar to a forthcom­ing descendant than to the initial ancestor. (Indeed, under strict gradualism, we even face the definitional absurdity that the last generation of an ancestor should be reproductively isolated from its own offspring — that is, the first generation of a new descendant. Some creatures may eschew such incest on moral or adaptive grounds, but no one would gainsay the biological possibil­ity.) Thus, on this traditional view, species cannot maintain sufficient tempo­ral stability to be called individuals. In addition, species do not have discrete birth points if they branch from their ancestors at rates no different from characteristic tempos of transformation during their subsequent anagenetic lifetimes (see Fig. 8-1). At most, some species display clear termination in extinction (but others evolve gradually to descendants.) Thus, species do not function as good vernacular individuals if gradualism and anagenesis pervade the history of life.

  Even so — or as long as most species arise by splitting of lineages rather

  8-1. The traditional view (as depicted, but not defended, in Lipps, 1993) of why species cannot be construed as proper biological individuals but only as arbitrary segments of a smooth and unbreakable continuum.

  [Page 605]

  than by wholesale transformation, no matter how gradual the tempo of branching — the individuality of species may be maintained in some technical sense, though only by violation of our vernacular intuitions. After all (see Fig. 8-2), so long as branching points (or fuzzy intervals) can be temporally lo­cated at all, then species do have definable intervals of existence, and can be individuated on this basis, even if their life courses violate our usual notions about sufficient morphological stability.

  Many evolutionary biologists have failed to recognize that the so-called cladistic revolution in systematics rests largely upon this insistence that spe­cies (and all taxa) be defined as discrete historical individuals by branching (leading to the rule of strict monophyly) — and not as classes with “essential” properties by appearance (leading to the acceptance of paraphyletic groups). Many biologists reject (and regard as nonsense) the cladistic principle that no species name can survive the branching off of a descendant — and that both branches must receive new names after such an event, even if the ancestral line remains phenotypically unchanged. But this counterintuitive rule makes sense within cladistic logic — for cladists define new entities only as products of branching (the word clade derives from a Greek term for branch). A trans­forming species that does not branch cannot receive a new name even if the final form bears no phenotypic resemblance or functional similarity to the original ancestor. Thus, if such extensive transformation occurs in un-branched lineages, a cladist, by failing to designate a truly different anatomy with a distinctive name, retains the technical individuality of species at the price of a severe assault against legitimate intuition.

  Can we find any solution to this dilemma? Must we either deny that species can be viewed as individuals, or else accept a logically “pure” definition based on branching, but strongly in violation of vernacular usage? I suggest that this issue can be resolved empirically, and need not persist as a defini­tional or philosophical conundrum. If gradualism and anagenesis prevail in

  8-2. A repeat of Figure 7-2 to show that, even under the most gradualistic and anagenetic models, species can still be individuated under a conception of evolution as a branching process at the species level.

  [Page 606]

  nature, then all the aforementioned problems cannot be avoided and apply in force. But suppose, as Eldredge and I have long argued in our theory of punc­tuated equilibrium (see Chapter 9), that gradualistic anagenesis occurs only rarely in nature, and that the great majority of species remain essentially sta­ble throughout their geological lifetimes. (Our concept of stasis recognizes that species fluctuate mildly throughout time, to an extent no different from ordinary geographic variation among demes of a species at any one moment, but we hold that mean values of phenotypes generally do not change in a cu­mulative or directional manner.) Suppose also that species, on geological scales, branch in unresolvable “moments.” (In nearly all-geological circum­stances, single bedding planes amalgamate the events and accumulated results of several thousand years.) If species tend to originate in thousands to tens of thousands of years — that is, with glacial slowness by the inappropriate crite­rion of a human lifetime in potential observation — and then to persist in sta­sis for millions of years, th
eir origin becomes instantaneous in geological time, and species arise as discrete individuals at this proper macroevolutionary scale. Of course, some fuzziness must attend the origin of a species, for we acknowledge that macromutational beginnings in leaps of a single generation rarely, if ever, occur. But when “fuzziness” occupies only a thou­sand years in a million — that is one tenth of one percent of later existence in stasis — then geological indefiniteness surely does not exceed even the relative duration of the fuzziness (9 months in some 80 years) attending the embryo-logical beginning of human personhood!

  Under punctuated equilibrium, the remaining criterion of discrete death achieves even clearer definition — for nearly all species disappear by extinc­tion (“living on” only through their progeny of daughter species with new names and individualities), and not by gradual bodily transformation into something else. Species deaths, at geological scales, are surely more discrete and “momentary” than human deaths scaled against the lengths of our life­times.

  In summary, then, species that originate by branching can be individuated even under the assumption that gradualistic anagenesis prevails during the history of most species lifetimes (but only by violating our vernacular concep­tion of “personhood” or individuality). However, if punctuated equilibrium prevails as an empirical proposition (see Chapter 9 for defense of this conten­tion), then species are individuals — in some cases much “better” individuals than conventional bodies of organisms — by all vernacular criteria. Under punctuated equilibrium, species originate at points of birth with initial fuzzi­ness confined to an insignificant (usually unmeasurable) moment properly scaled against later existence in stasis. They experience even clearer moments of death, for nearly all species terminate by true extinction and not by trans­formational passage into a descendant that vernacular (non-cladistic) usage will wish to recognize with a different name (a phenomenon called “pseudo-extinction” by paleontologists). And species surely maintain “sufficient sta­bility” during their geological lifetimes by all criteria outlined on page 602. They remain discrete by reproductive isolation (conventionally cited, ever [Page 607] since Buffon, as the chief criterion of “specieshood”). They function as a unit and persist continuously. Above all, they do not change substantially in phenotype — the crucial concept of stasis. Surely, the average species in stasis un­dergoes less temporal change (with less directionality) than human bodies experience in our passage from babyhood through adult vigor and into senes­cence. If humans retain discrete personhood through all these slings and ar­rows of ontogeny, then species (under punctuated equilibrium) function as equally good or even better individuals by the same criteria of vernacular definition.