The Structure of Evolutionary Theory

by Stephen Jay Gould

  Suppression of cell lineage selection by the multicellular organism has greatly restricted a once vibrant and multifarious level. I must confess to my own parochialism in recognizing just one unit, the cell, as a surrogate for all entities that enclose genomes and form parts of organisms. Certainly [Page 697] organelles (or at least the mitochondria and chloroplasts that began their evolutionary history as symbiotic prokaryotes), and sometimes tissues and mod­ules of embryological development, can also act (in principle) as suborganismal units of selection. I make the amalgamation because this level has been largely suppressed, and therefore doesn't often come to our attention in stud­ies of modern multicellular organisms. In so doing, I feel some allegiance to the folk taxonomist who (as so often recorded for indigenous cultures) assid­uously names each species (much as a trained Linnaean systematist would do) for creatures important to his life, but then lumps into large categories (weeds, butterflies, bugs) the organisms of no great moment in his world.

  As the central premise of his fascinating and seminal book, Buss (1987) argues that the multicellular individual arose by “the interplay between selec­tion at the level of the individual and selection at the level of the cell lineage” (p. 29). More specifically, he attributes the distinctive features of metazoan development to an initial competition among cell lineages, eventually tamed and regulated by organismic selection in the interests of bodily integrity. Buss writes: “The thesis developed here is that the complex interdependent pro­cesses which we refer to as development are reflections of ancient interactions between cell lineages in their quest for increased replication. Those variants which had a synergistic effect and those variants which acted to limit subse­quent conflicts are seen today as patterns in metazoan cleavage, gastrulation, mosaicism, and epigenesis” (p. 29).

  Clearly, such a concept becomes intelligible only under the aegis of a hierarchical model of selection, as defended in this book's central thesis. Buss recognizes this conceptual link, of course, and his work becomes a strong confirmation of both the efficacy and necessity of this basic reconstruction in evolutionary theory. In terms similar to the views expressed here, Buss writes (pp. 5-6): “The logical structure of Darwin's argument allows any unit to evolve if it replicates with high fidelity, and if selection distinguishes be­tween the variants. Species, populations, and lineages of individuals, cells, organelles, and gene sequences can all potentially evolve. Yet we have been largely content to attribute the whole of biological diversity to selection upon individuals [organisms]. The once comfortable cloak of the Modern Synthesis has become restrictive.” (I am also grateful to Buss for recognizing the role of my profession, particularly in the work of Eldredge, Jablonski, Stanley, Vrba, and myself, in developing the hierarchical theory of selection. He writes (p. ix): “Indeed, hierarchical perspectives on evolution are undergoing a re­birth among paleontologists at the moment.”)

  In Buss's model of historical and sequential construction for nature's hierarchy, new levels arise to enclose the individuals of older levels by a two-step process. The initial features of the nascent level must originate in synergism, or positive interaction, with selection at the level just below, which formerly stood topmost, but will now be superseded (in the literal sense of “sat upon”) by the newly emerging style of organization. New levels must begin with such a helpful boost, for the initial tentative and unformed steps cannot yet possess enough power to suppress or regulate a well-established level beneath. But [Page 698] stabilization of the new level, implying a power to suppress at least some forms of harmful proliferation from within, then requires negative interac­tion, once the new and higher level achieves enough coherence to act in its own right.

  Since we have no direct data for key transitions that occurred so long ago and left no fossil evidence (so far as we know), Buss constructs some hypo­thetical examples of how such a process could work. (Such entirely specula­tive scenarios must be understood within their acknowledged limits — that is, as hypothetical stories, “cartoons” in Buss's words, invented to illuminate a potential mode, and not as claims about any historical actuality.) For exam­ple, if the first tentative multicellular organisms evolved as little more than spherical colonies of identical protists floating in the ocean, how might essen­tial organismic properties like cellular differentiation emerge? Suppose that a variant cell lineage arose in such a loosely knit, hollow sphere of cells, caus­ing members of the new line to enter the sphere's center, where proliferation could continue. In this way, a new cell lineage (and the beginning of cellular differentiation for the organism) could originate and proliferate by selection at the cell level. Buss then supposes that such an event might also be beneficial for the organism, and he draws an analogy to the ontogeny of some modern sponges:

  The origin of a variant cell line which entered the center of such a sphere to continue cell division . . . may have produced a structure which was sufficiently negatively buoyant to fall to the sea floor. Many modern sponges ... do just this. A flagellated sphere populated by amoeboid cells simply drops to the ocean bottom . . . The pelago-benthic life cycle of sponges may have arisen as a consequence of variants which, in pursu­ing their own replication, fortuitously presented the individual with a benthic existence and all the attendant opportunities inherent in the in­vasion of a new adaptive zone.

  This move toward a more complex and better-integrated organism begins with an initial synergism between cellular and organismic selection (origin of a new cell lineage by invasion and proliferation in the organism's hollow cen­ter, leading to organismal advantages through an imposed change of habitat). But later stabilization of this innovation requires the suppression of cell lin­eage selection by the organismic level — for if the two cell lineages (at the sphere's periphery and center) engage in an anarchic battle for ever greater representation in cellular percentages, either the organism will lose coherence and die, or one lineage will win and the organism will return to its previous state of minimal differentiation.

  Moving away from speculation and towards an explanation of metazoan development, Buss interprets several defining features of many (but not all) metazoan phyla as records of successful suppression of cell-lineage selection by organismal selection from above. In particular, he views early germ-line sequestration (Weismann's crucial criterion in his defense of Darwin against resurgent late 19th century Lamarckism, see Chapter 3), and maternal predestination, [Page 699] as organismal devices evolved to set and stabilize the course of development as early in ontogeny as possible, thus greatly reducing the po­tential for new forms of differential cellular proliferation either to arise at all in later ontogeny, or to reach the germ line and act in cell-lineage selection even if they do manage to originate. Buss sums up his thesis:

  Selection at the level of the individual has opposed selection at the level of the cell lineage by acting to set the timing of terminal somatic differentiation as far back in ontogeny as possible — whenever possible into the maternal cytoplasm itself, (p. 5). ... The release of the totipotent germinative lineage from the task of producing somatic tissues meant that the number of divisions made by the totipotent lineage could be re­duced and, consequently, the opportunity for variants to arise to become severely restricted (p. 100) . . . Metazoans, by the twin devices of mater­nal predestination and germ-line sequestration, have effectively closed their ontogenies to heritable intrusion arising in the course of that ontog­eny. A novel epigenetic program can only arise if a mutation of extraor­dinarily improbable precision and autonomy occurs in the germ cells themselves (p. 102).

  But nothing can be won without a price in our complex world of interact­ing levels, either in evolution or in human society. In stabilizing the organismic level with such effective devices to suppress cellular and other forms of suborganismic selection, organisms have greatly reduced their flexibility for future evolutionary change of more than a superficial nature. For these mech­anisms of development do not suppress only the forms of cell-lineage selec
­tion that would harm the organism; rather, they impede any effective cellular selection at all, whether beneficial or harmful. These policing devices of the organism therefore close off an avenue once open for substantial change in basic designs, thus restricting maximal potency to the iteration of essentially similar species (as in such famous examples as the cichlids of African lakes, or the Galapagos finches), now representing evolution in its most vigorous con­temporary mode. Ou sont les neiges d'antan? “The clear implication is that evolution of cellular differentiation fueled the evolution of controls over vari­ants which fail to behave altruistically. The mechanisms, which metazoans employ to limit the heritability of variants, which fail to contribute to somatic functions, are blind to the traits, which a variant might express. Potentially beneficial variants are as limited as are potentially detrimental ones” (p. 103, Buss's italics).

  This perspective implies a striking limitation upon the strictly Darwinian style of extrapolative and gradualistic selection that the Modern Synthesis promulgated as an adequate explanation for evolution at all scales of time and effect (see quotation from Wilson et al. on p. 583). If Buss's views are valid, then conventional Neo-Darwinian evolution must work within stric­tures of essentially established ontogenies that can surely generate exuberant adaptive variations upon set themes, but may be effectively unable to con­struct major innovations that establish the outlines of macroevolution. Once [Page 700] again, we grasp the need for independent macroevolutionary theory — and Buss has supplied an important piece of the general argument with his con­cept of a correlation between such major innovations and the origin of new hierarchical levels, a theme that obviously requires the hierarchical model and cannot be encompassed within the strict Darwinism of the Modern Syn­thesis. Buss concludes (p. 188, his italics): “Synergisms between the units drove the elaboration of a higher unit and conflicts arising between units were minimized by adaptations limiting further variation. This conclusion has the fascinating and crucial corollary that the major features of evolution were shaped during periods of transition between units of selection.”

  THE ORGANISM-INDIVIDUAL As virtually the entire history of Darwinian thought has unfolded under the assumption that organisms act as nearly ex­clusive agents of selection (or at least that our interest in evolution centers upon the alterations and fates of organisms), I shall not dwell upon this ca­nonical individual here. I want only to reemphasize the unique and decidedly peculiar features of our kind of entity (in contrast to the characteristic prop­erties of individuals at other levels): maximal cohesion based on functional integration, including relatively inflexible spatial orientations of subparts (spatiotemporal if we include embryogenesis). This style of integrity enables the organism to be particularly effective in suppressing selection against its interests by potential evolutionary individuals dwelling within and forming its parts. As noted above, the virtual “extinction” of effective cell lineage se­lection in complex metazoan phyla occurred as a historical result of the evo­lutionary “invention” of the intricate organism — perhaps the only example of an “endangered level” in the entire history of evolution!

  As another portentous implication of individuality in this mode, organisms become chock full of adaptations as a consequence, under natural selection, of building coherence by functional integration. This local phenomenon at one level of Darwinian individuality has generated an understandable and commanding concern with adaptation, leading to doctrines of exclusivism in extreme cases (all too common, given our psychological preferences for simple and unifying worldviews — a need traditionally met theologically, but sometimes, particularly in our increasingly secular age, scientistically). If, as some strict Darwinians believe, “organized adaptive complexity” repre­sents both the primary result of evolution and the cause of all other patterns in the history of life, then we will fail to understand nature for two cardinal reasons: (1) because we have adopted a criterion too strict even for its organismal level of most promising application (see Chapters 10 and 11); and (2) because the criterion of “organized adaptive complexity” does not strongly characterize the nature or definition of individuality at most other levels of the hierarchy.

  Nature's hierarchy is not fractal; each level, to express the point metaphorically, does some things well, and other things poorly or not at all — and the evolutionary pattern of nature features many essential things. In our mother's house — the Earth — are many mansions. Gene selection is “good” at iterating [Page 701] elements — an important input of raw material for generating “organized adaptive complexity” at a higher level. Organisms are good at building com­plex adaptations. Species are good at forging temporal trends of geological duration, and their efforts largely regulate the relative diversity among phyla (why so many beetles, and so few pogonophorans). To say (as Dawkins, Wil­liams, and other detractors often do) that species selection must be unimpor­tant because such a process can't build organismal complexity reminds me of the cook who didn't like opera because singing couldn't boil water.

  THE DEME-INDIVIDUAL This kind of individual has borne the brunt of the general argument about higher-level selection ever since Darwin awarded the idea a strictly limited amount of conceptual space in trying to puzzle out the origins of human altruism (see pp. 133-137). The subject has been ex­tensively reviewed and controverted (Wynne-Edwards, 1962, vs. Williams, 1966, for an early and generally unacceptable version; Wade, 1978, 1985; D. S. Wilson, 1980, 1983,1989; Wilson and Sober, 1994; Sober and Wilson, 1998, for reviews). I shall therefore provide only an idiosyncratic sketch here, for the terms and concepts of this discussion permeate the chapter, while my own interest as a paleontologist flows to the still higher levels that have not been extensively studied.

  In a curious way, the development and acceptance of hierarchy theory has been impeded because the classical treatment of this subject has been focussed so strongly, indeed almost exclusively, on this level — and demes are the hard­est of all individuals to validate and justify within the evolutionary hierarchy. All other individuals build better boundaries (to retain their own subparts, or lower-level individuals, and to exclude the subparts of other individuals at their level), and experience less difficulty in remaining sufficiently stable for the requisite time until reproduction. But demes are especially vulnerable to the classic objection (see p. 647) that, lacking strong internal mechanisms for coherence, their individuality may be too fleeting and subject to change by loss or invasion — as in Dawkins's well-formulated and memorable image of dust storms in the desert or clouds in the sky. Indeed, as I argued previously (p. 648), the classic defense of interdemic selection depends upon the identi­fication of plausible conditions that would allow such adventitious groups to remain stable long enough to act as units of selection. The centering of the general argument for higher-level selection upon demes has, by false and un­fortunate implication, led to the widespread impression that any kind of supraorganismal selection must face the same difficulties — perhaps with problems growing ever more intense as individuals become more inclusive. But this argument, based on illogical assumptions about linear extrapolation, does not hold because demes (in most circumstances) are uniquely unstable in the evolutionary hierarchy. Species, for example, usually attain as much sta­bility and coherence as organisms, though by different mechanisms (see pp. 703–705).

  Group selection has traditionally been invoked under our organismic bi­ases as an explanation for bodily behaviors — with altruism as a paradigm, [Page 702] ever since Darwin himself (see Chapter 2, pp. 133–136) — that seem, prima facie, difficult to explain as beneficial to organisms, but can easily be con­strued as valuable for groups. But we should recognize such restricted invoca­tion (only for cases that trouble organismic traditionalists) as yet another pa­rochial limitation, and we should acknowledge a potentially general role for interdemic selection within any species of appropriate population structure. (Under such a criterion of judgment by relative frequency, we mus
t ask a dif­ferent, and quite unanswered, fundamental question: how many higher taxa generally maintain population structures that promote interdemic selection; in what environments; and with what correlations to such factors as phylogenetic status, body size, behavioral complexity, etc.)

  If various arguments for the rarity of extensive evolution within large panmictic populations hold merit, and if Sewall Wright's shifting balance the­ory applies to a high percentage of populations, then interdemic selection may become a major mechanism for evolution within species through time. However, if punctuated equilibrium generally holds (see Chapter 9 for a de­fense of this view), then anagenesis within species will be rare in any case (whether by transformation via organismic selection under panmixia, or by shifting balance via interdemic selection in appropriately subdivided popula­tions). Or perhaps, as an intermediate position, panmictic transformation is rare, but shifting balance frequent, in species that meet the criteria for appro­priate population structure. The high relative frequency of punctuated equi­librium would then measure the relative rarity of such population structures, and the few groups that show extensive gradualism within species may gener­ally subdivide their populations according to Wrightian criteria. This conjec­ture has not been tested, but could be, and with an interesting mixture of paleontological data on the history of species and neontological information on population structures within modern representatives of the same groups.

  In any case, even if Wright's criteria don't hold often enough within the central range of species during the heart of their geological life, Mayr's peripatric model of speciation suggests that the origin of most species may occur by a process close to interdemic selection, and operating near a blurred bor­derline with species selection. If many species spawn large numbers of periph­erally isolated demes, but only a few of these demes become species; and if the small class of successful speciators possess traits at the population level that encourage full speciation in interaction with the environment; then species will arise by selection and differential preservation on a just a few “winners” within a set of populations that begin as demes of an ancestral species (as best illustrated by the probable main reason for failure of others to speciate — reincorporation of a peripherally isolated deme into the larger parental popu­lation).