The Structure of Evolutionary Theory

by Stephen Jay Gould

  Leo Buss (1987), in a fascinating book on the role of hierarchical selection in the phylogenetic history of development (see pp. 696–700 for further dis­cussion), offers a compelling case for the vital importance of both synergistic and negative selection between levels in the history of life, which he views largely as a tale of sequential addition in hierarchical levels — so that nature's current hierarchy becomes a problem for historical explanation, not an inher­ent structure fully present throughout time. Buss argues that synergism must fuel the first steps in adding a new level atop a preexisting hierarchy (for ini­tial negativity against the previous highest level would preclude the origin of a new level). But, having once achieved a tentative foothold, the new level sta­bilizes best by imposing negative selection against differential proliferation of individuals at the level just below — for these individuals have now become parts of the new level's integrity, and selection at the new level will tend to check any dysfunctional imbalance caused by differential proliferation from below. [Page 680]

  2. Each hierarchical level differs from all others in substantial and interest­ing ways, both in the style and frequency of patterns in change and causal modes. Nature's hierarchy, for all the commonality of its unifying principles (selection, for example, acting at each level), does not display fractal structure with self-similarity across levels.

  As the theory of hierarchical selection develops, I predict that no subject within its aegis will prove more fascinating than the varying strengths and modalities among levels. Just as the study of allometry has recorded charac­teristic and predictable scale-dependent differences in structure and function of organisms at strongly contrasting sizes — a prominent subject in biology ever since Galileo formulated the principle of surfaces and volumes in 1638, and so elegantly codified in D'Arcy Thompson's masterpiece of both prose and concept, On Growth and Form (1917, second edition, 1942) — so too does individuality as a tiny gene imply substantially different properties for a unit of selection than “personhood” as a large species or an even larger clade. Allometric effects across hierarchical levels should greatly exceed the familiar (and extensive) differences between tiny and gigantic organisms for two un­surprising reasons (see Gould and Lloyd, 1999, for a detailed development of this argument). First, the size ranges among levels are far greater still. Second, organisms share many common properties simply by occupying a common level of evolutionary individuality despite an immense range of size; but the levels themselves differ strongly in basic modes of individuality, and therefore develop far greater disparity.

  But this promise also implies a corresponding danger. In some famous lines composed for a quite different, but interestingly related purpose, Alexander Pope explored the paradox of man's intermediary status between two such disparate extremes, both so desperately needed to know and to understand (the bestial and the godly in Pope's concern) — but both so inscrutable as so far from our own being:

  Placed on this isthmus of a middle state,

  A being darkly wise and rudely great...

  He hangs between; in doubt to act or rest;

  In doubt to deem himself a god, or beast...

  Created half to rise, and half to fall;

  Great lord of all things, yet a prey to all;

  Sole judge of truth, in endless error hurl'd;

  The glory, jest, and riddle of the world!

  I appreciate this image of an “isthmus of a middle state” — a narrow standing place linking two larger worlds of smaller and greater. Pope's dilemma may pack more emotional punch in its moral meaning (since his greater and lesser worlds define questions of value rather than geometry), but our problem fea­tures greater intellectual depth — for, surely, a larger conceptual chasm sepa­rates the gene from the clade in modes of evolutionary mechanics, than the bestial from the virtuous in styles of human behavior.

  The problem can be summarized with another, and much older, classical quotation. “Man is,” as Protagoras, wrote in his wonderfully ambiguous epigram, [Page 681] “the measure of all things” — ambiguous, that is, in embodying both positive and negative meanings: positive for humanistic reasons of ubiquitous self-valuing that might lead to some form of universal brotherhood and com­passion; but negative because our own “measure” can be so parochially limit­ing, and therefore so conducive to misunderstanding other scales if we must assess these various domains by the allometric properties of our limited es­tate.

  This issue becomes especially serious for the hierarchical theory of selec­tion. Humans hold status as both evolutionary individuals and organisms — yet all other “separate but equal” evolutionary individuals at other hierarchi­cal levels are not organisms. Unfortunately, organisms constitute a very spe­cial and distinctly odd kind of evolutionary individual, imbued with unique properties absent from (or much weaker in) other individuals (at other levels) that are equally potent as evolutionary agents. But if we mistakenly regard our own unique properties as indispensable traits for any kind of evolution­ary individual — the classic error of parochialism — then we will devalue, or even fail to identify, other individuals defined by different properties and resi­dent at other levels.

  I shall explore some of these crucial differences in the next two sections (disparate properties of the six major levels; and extensive comparison of organisms and species as evolutionary individuals). In this introductory com­ment, I only wish to emphasize that the uniqueness of the organism as a unit of selection lies in securing individuality by maximal homeostatic interaction among parts, an integration that ties each subpart to the fate of all, and there­fore strongly discourages any “breakout” or differential proliferation (by suborganismic selection) from within. To be sure, such integration represents a powerful strategy for individuation, but this strategy does not specify the only legitimate path, and other potent evolutionary individuals use other mechanisms. For this reason, I regret that Wilson and Sober (1994) so em­phasize these “organic” properties of individuality in their general definition, meant to apply to all levels. This parochial focus leads them to downplay the individuality of units of selection at other levels, where different defini­tional criteria predominate — in species, for example, where the maintenance of boundaries by reproductive isolating mechanisms, and the mixture of sub-parts in replenishment (sexual reproduction), maintain cohesion and stability just as well as organisms do by the different strategies of homeostasis and functional interaction of subparts.

  Redressing the tyranny of the organism: comments on

  characteristic features and differences among six

  primary levels

  I have little tolerance for numerical mysticism. I feel no special affinity for threes (as trinities), fours (Jung's primal archetype), fives (for fingers or echinoderms), sevens (for notes of the musical scale, planets in the Ptolemaic sys­tem, and so much else), or nines (the trinity of trinities). Similarly, in recogniz­ing six hierarchical levels for this discussion — genes, cells, organisms, demes, species, and clades — I only utilize a device of convenience, and do not make [Page 682] any assertion about a fixed number of units in the expanded hierarchy of Darwinian action.

  Any such claim of definity could only rank as both foolish and incoherent for at least two reasons. First, the hierarchy has not been set by structural or logical principles, but historically evolved in a contingent manner. Thus, be­fore the inventions of sexual reproduction and multicellular organisms, nei­ther species nor organisms (as a level distinct from cells) existed, and a quad­ripartite hierarchy held sway (and still does today in the dominant world of asexual unicells) — gene, cell, clone, and clade. Second, several of the levels discussed here coagulate numerous phenomena because they lie between two clear boundaries. As Buss (1987) points out, for example, we might, in cer­tain contexts, recognize several items that encase genes but serve as parts of multicellular organisms: chromosomes, organelles, cells, organs, etc. Before the multicellular organism evolved, and began to act with suc
h effectiveness as a suppressor of intraorganismic selection, we might have construed this domain of “proto-individuality” quite differently, and with finer resolution.

  As a second argument against granting necessary or inherent status to these six levels, I have followed nearly all students of this field in preferring a fully nested hierarchy of increasing inclusion, to other legitimate interactors that function only occasionally, transiently, or in special circumstances. This fully nested hierarchy operates with Linnaean logic in requiring that lower units amalgamate completely, and under strict genealogical constraint — so that no lower unit can belong to more than one higher unit, while no higher unit can “forage” outside its hereditary line to incorporate the lower units of other distinct evolutionary branches at the same level. Just as a genus can't belong to two families, a species of flies cannot incorporate some onychophores and a few myriapods to construct a more versatile species-individual.

  We logically require this property of nesting to correlate the nonhistorical process of selection with a set of quintessentially historical phenomena in evolutionary biology, including phylogenetics and the study of adaptation. Without such a fully inclusive hierarchy, for example, we could not use one level as a surrogate or convenient descriptor for events at other levels in the same nest — as when we choose the gene level for keeping the general books of evolution (see pp. 632–637 on the error of gene selectionism).

  Nature, of course, does not always obey this logical stricture, though we may appeal to the empirical success of this formulation as an indicator that nature does comply at a preponderant relative frequency. If life did not gener­ally work within a hierarchy of inclusion, the biotic world would present such a different appearance that our conventional ordering devices would not operate usefully, and would never have been proposed or accepted. (I am not a naive realist, and I have argued throughout this book that we impose our social preferences upon nature in constructing our theories. But nature does provide a strong input, and does impose a powerful constraint upon our formulations.) No one would ever have suggested a nested system like Linnaeus's, if common experience proclaimed that novel taxa generally arise by distant amalgamation — if, for example, each new mammal arose [Page 683] by a principle of “disparate thirds,” say with equal mixtures of dugong, aardvark, and howler monkey. (We all know, of course, though we rarely discuss the subject in polite company, that the Linnaean logic, which presupposes a to­pology of branching without amalgamation, cannot apply to groups that do show massive mixture, as in some families of plants with extensive hybridiza­tion, or especially in prokaryotes evolving with frequent lateral transfer — a phenomenon that, on accumulating evidence, may be common enough to truly discombobulate the Linnaean version for the pre-multicellular majority of life's tree (see Doolittle, 1999), with practical and theoretical consequences as broad as any revolutionary discoveries in the recent history of evolutionary biology.) Similarly, we all appreciate the conceptual difficulties imposed by some prominent cases in evolution, mostly at the genic or cellular level, that do violate the hierarchy of inclusion — most notably, the origin of some organelles as symbiotic prokaryotes.

  Since units of selection operate as interactors with the environment, and since entities obeying the criteria of “personhood” (see pages 602–613) do occasionally cohere by distant genealogical amalgamation, nature does pres­ent some exceptions to the principle of a fully nested hierarchy for evolution­ary individuals. But these exceptions truly function as the “rule provers” of our mottoes (in the sense of probing, or testing, our generalities), and not as falsifiers. The most widespread cases, including the origin of cellular organelles by endosymbiosis, represent “frozen” phenomena of history, not active amalgamations presently building evolutionary individuals by junction of disparate genealogical lines. (However, genic exceptions, as noted above, may be rife if lateral transport occurs as frequently as current theory and data are now beginning to suggest, especially for prokaryotes.) The most common, ac­tive cases involve symbiotic and coevolutionary unions tight enough to obey the Biblical rule of Naomi and Ruth: “whither thou goest, I will go.” Wilson and Sober (1989), for example, present a fascinating discussion of “phoretic associations,” or obligate carriage, by wingless insects as they move among resource patches, of various mites, nematodes, fungi, and microorganisms. In some cases, the load of these “hangers on” can disable or even kill the insect, and conventional Darwinism will then work in its usual, competitive, and organismic mode. But the phoretic associates may be limited to densities that do not affect the insect, and may also provide resources indispensable for suc­cessful colonization of new patches — in which case, the entire association may be evolving as a “superorganism.”

  With these caveats in mind — the somewhat arbitrary division of the evolutionary hierarchy into six levels, and the acknowledgment of interesting exceptions to full nesting among nature's various individuals — I shall try to specify some distinctive “allometric” properties of the levels and their inter­actions:

  THE GENE-INDIVIDUAL As we enter this first unfamiliar world of such great, and literally basic, importance to evolution, we encounter an initial rung of strong difference from the organism-individuals that, if only for psychological [Page 684] reasons, must stand as prototypes for our parochial concept of how a proper Darwinian unit must function in natural selection. If we could ever truly grasp the gene's world, with full sympathy and appreciation for rel­ative frequencies, hard-line selectionism would yield to a fascinating enlarge­ment that would actually strengthen selectionist theory by synergism with other (non-contradictory) forces — so this subject should therefore not intimi­date strict Darwinians. For the most part, however, the necessary acknowl­edgment of different gene-level processes has unfolded within the traditional perspective of organismic selection — with three basic categories of interpreta­tion as “good” for organisms, and acceptable on this basis; “bad” for organ­isms, and a destabilizing danger that must be conquered; or irrelevant to or­ganisms and therefore unimportant.* The implications for a hierarchical reconstruction of evolutionary theory have therefore been missed or downplayed. Consider the two major themes of recent literature:

  Motoo Kimura and the ‘neutral theory of molecular evolution.” Although I have called this book “the structure of evolution­ary theory,” I have propagated my own lamentable parochialism under a pre­tense of generality. For this book, despite its exuberant length, largely re­stricts itself to the Darwinian tradition of conventional causal explanations based on selection as a central mechanism. I do, to be sure, treat the major critiques of unbridled selectionism (constraints as channels, failure of pure extrapolationism into geological time), but I conduct this discussion within a Darwinian world, and do not adequately consider truly alternative mecha­nisms of change and their domains of operation. Since selection is a causal theory of change based on distinctive traits of definable individuals within specified environments (quite apart from any stochastic sources for the varia­tion that provides raw materials of change), the obvious first-line alternative to selection must lie in random reasons for change itself.

  As a basic statement in the logic of an argument, this point can hardly be denied, and therefore enjoys a long history of recognition in evolution­ary thought. But recognition scarcely implies acceptance. The Victorian age, basking in the triumph of an industrial and military might rooted in technol­ogy and mechanical engineering, granted little conceptual space to random events, so the issue barely arose in Darwin's own time. (Darwin got into enough trouble by invoking randomness for sources of raw material; he wasn't about to propose stochastic causes for change as well! To this day, a distressingly familiar vernacular misunderstanding of Darwinism rests upon confusing these two components (sources of raw material and causes of change) — as in the common charge that Darwinism must be wrong because human complexity couldn't arise by purely random processes. Nineteenth [Page 685] century theories o
f probability also eschewed ontological randomness in fa­vor of causal production by interaction of so many fundamentally orthogo­nal mechanisms that stochastic formulations would best fit the observed results — the philosophical solution traditionally adopted by the scientific determinists who invented probability theory, most notably by Laplace him­self.)

  For these primarily societal reasons, theories of random change enjoyed lit­tle currency before our own century, when for both external reasons of a new cultural context (spawned by such events as the breakup of colonial empires, the devastation of World Wars, and the consequent questioning of predict­able progress as time's direction), and internal prods from the mathematical apparatus of population genetics, random models of change became a major and controversial subject in evolutionary theory. I shall not review this well-known story, centering on the life and work of Sewall Wright (see Wright's own magnificent four-volume summing up, and Provine's fine biography). I only need to remind readers that genetic drift (often called “the Sewall Wright effect” in early literature), while unimpeachable in theory, and therefore surely operative in nature, received very short shrift, especially as the Modern Synthesis hardened around its adaptationist core (see Chapter 7). The Synthe­sis did not and could not deny genetic drift; instead, supporters resorted to the classical argument for dismissal in natural history — relegation to insig­nificant relative frequency. I learned the argument as a near mantra in all my graduate classes during the mid 1960's: fixation by genetic drift can only oc­cur in populations so tiny that most will already be on the brink of extinction.