The Structure of Evolutionary Theory

by Stephen Jay Gould

  I certainly accept the numerous cases of well-documented gradualism in foraminiferal lineages, and I acknowledge MacLeod's argument (1991) that abrupt transitions without branching in some sequences (the “punctuated anagenesis” of Malmgren et al., 1983) may arise as artifacts of condensed intervals of sedimentation within truly gradualistic trends. But punctuated equilibrium has also been demonstrated with data of equal abundance and completeness (the elegant case of Wei and Kennett, 1988, stands out for thor­ough documentation of both the geography of origin and subsequent history of nontrending in Globorotalia (Globoconella) pliozea), and we have little idea, and no firm data, about overall relative frequencies for tempo and mode of evolution in this group.

  Stasis has been demonstrated in other microfossil “species” with equally dense documentation — e.g., Nichols (1982) on lower Tertiary pollen, Wig­gins (1986) on upper Miocene dinoflagellates, and Sorhannus (1990) on the Pliocene diatom Rhizosolenia praebergonii. Ross (1990) suggested that puta­tive differences in relative frequencies might be tested by comparing foraminiferal lineages with microfossils of sexual metazoans preserved in comparable abundance in the same sediments — and that forams vs. ostracodes might provide a good test. Indeed, published cases for ostracodes seem to speak strongly for stasis and punctuation as a predominant pattern (in contrast with foraminiferal data). I have already discussed Cronin's (1985) work on Cenozoic ostracodes (see p. 827), and now cite his general conclu­sion (p. 60):

  Morphologic and paleozoogeographic analysis of Cenozoic marine Ostracoda from the Atlantic, Caribbean, and Pacific indicates that climatic change modulates evolution by disrupting long-term stasis and catalyz­ing speciation during sustained, unidirectional climatic transitions and, conversely, by maintaining morphologic stasis during rapid, high-fre­quency climatic oscillations. In the middle Pliocene, 4 to 3 million years ago, at least six new species of Puriana suddenly appeared as the Isthmus of Panama closed, changing oceanographic circulation and global cli­mate. Since then morphologic stasis has characterized ancestral and de­scendant species during many glacial-interglacial cycles.

  The origin of new species by branching in response to geographic opportu­nity (rise of the Panamanian isthmus), rather than by anagenetic gradualism as a selective consequence of changing environments, matches the predictions of punctuated equilibrium. Of the contribution made by stasis to this conclu­sion, Cronin writes (p. 61): “Morphologic stasis characterizes most shallow marine ostracodes from the western Atlantic that were subjected to these climatic [Page 832] changes, suggesting a pattern predicted by the model of punctuated equilibrium.”

  Whatley's (1985) study of the common and speciose ostracode genera Poseidonamicus and Bradleya in Tertiary and Quaternary sediments of the southwest Pacific also match the expectations of punctuated equilibrium throughout. Whatley found some gradualism in size changes (a common pat­tern), but only stasis for all defining features of shape and ornament. Whatley concludes for Poseidonamicus (p. 108): “Although over some 55 million years, the ornament of the genus underwent considerable change, several of its species remained morphologically very stable over long periods of time: 10 to 15 million years being not uncommon ... This would seem to be evidence of virtual stasis between speciation events with respect to the evolution of the ornament of the various species.” In an interesting comment, relating stasis to the major prediction of punctuated equilibrium for evolutionary trends — the stairstep rather than the ball-up-the-inclined-plane model — Whatley writes (p. 109):

  Although the morphological change from the ornate P. rudis to the smooth P. nudus [I do love the rhyme as well] took place over a time span of more than 50 million years and, therefore, from a generic stand­point represents a very gradual change, it must be emphasized that the individual species within this evolutionary series are effectively invari­able with respect to their ornament. Morphological change was abrupt and coincided with speciation and further speciation was required to bring about yet further ornamental change. Ornamental change is clearly saltatory, very abrupt, and punctuated.

  Gingerich's (1974,1976) cases of gradualism in tooth size for several lower Eocene mammalian lineages in the Big Horn Basin of Wyoming instituted the empirical debate about punctuated equilibrium (see our response and critique in Gould and Eldredge, 1977). The tracing of gradualistic sequences for densely sampled series of small mammals (also based on dental evidence) then became an important research program for French paleontologists (see Godinot, 1985; and Chaline and Laurin, 1986, for sources more accessible to anglophonic readers). Large mammals have also furnished evidence for grad­ual anagenesis within species, as in Lister's study (1993a and b) of mammoths and moose — though he acknowledges that small sample sizes preclude a rig­orous distinction of this interpretation from an alternative reading of several cladogenetic events, each perhaps punctuated, and all leading in the same di­rection of change (Lister, 1993a, p. 77).

  But numerous examples of stasis in equally well-sampled strata have also been documented for mammals (see pp. 854–870 for commentary on relative frequencies). The rodent sequences that form the empirical basis for most gradualistic studies of the French school have also yielded several examples of stasis (Lich, 1990; Flynn, 1986). Summarizing his work on rhizomyid rodents from the Miocene Siwalik deposits in Pakistan, Flynn (1986, p. 273) wrote: “Most early rhizomyid species survive on the order of millions of years, with [Page 833] at least two spanning about five million years, and display apparent stasis in most characters.”

  Several analogs of Gingerich's classic studies on gradualism have provided strong evidence for stasis, thus proving diversity of modes, even where gradu­alism had been most strongly asserted as an exclusive pattern. Gingerich had studied the small condylarth Hyopsodus in early Eocene rocks from the Big Horn Basin of northwestern Wyoming. West (1979), however, found only stasis for the same genus from slightly younger Middle Eocene rocks from the Bridger Formation of southwestern Wyoming. West concluded (1979, p. 252): “Bridger Formation Hyopsodus data seems to show little size change through approximately one million years. This stasis or equilibrium condi­tion ... is the only well developed pattern in Bridger Hyopsodus.” Schankler (1981) then analyzed another genus, the condylarth Phenacodus, from the Big Horn Basin strata used by Gingerich to document gradualism in different taxa, and found only stasis within species (with abrupt transitions between species — a pattern that Schankler interpreted, correctly in my view, as a prob­able result of migration into a local area, rather than punctuational speciation in situ). He concluded (1981, p. 137): “The long-term stasis in mor­phology and size shown by the four species of Phenacodus conforms to the pattern expected in a model of evolution by punctuated equilibria.”

  As for the mammal we all love best (see pp. 908–916 for a more complete analysis), gradualism had long reigned as an unquestioned (and often quite unconscious) assumption in hominid evolution. An extensive, historically sanctioned set of dogmata, from ideas about “missing links” to the “single species hypothesis,” presupposed gradualism as a philosophical foundation. An early study by Cronin et al. (1980) — which would not be defended by several of its coauthors today — made the classic error of regarding a monotonically chang­ing set of mean values as virtual proof for anagenetic gradualism. (Such data cannot distinguish the stair steps of punctuated equilibrium from the same empirical pattern produced by gradualism in highly incomplete sections.)

  The spotty data of hominids offer little opportunity for adequate testing of such ideas (and we wouldn't even think of applying an apparatus of this kind to such a poor example if we didn't care so much about the particular case). Nonetheless, I am gratified by some strong hints of substantial stasis in several hominid species, especially for increasingly persuasive data on the importance of apparently punctuational speciation in this small clade dur­ing a crucial million year African interval (ca. 2-3 my B.P.) that featured the putative origin of at least half a dozen hominid species. Rightmire'
s early claims (1981,1986) for stasis in Homo erectus have been strongly challenged (Wolpoff, 1984), though the jury has surely not yet come in (despite a tenta­tive vote from this juror, despite his general biases in the other direction, for at least some fairly persuasive gradualism within this species).

  But two apparently sound cases of stasis have attracted substantial atten­tion while we should also not neglect, if only for its radical meaning in the light of previous assumptions, the short-term stasis of Homo sapiens, at least from the earliest Cro-Magnon records in Europe (about 40,000 years B.P. to [Page 834] our present circumstances). When we realize that the cave painters of Chauvet, Lascaux, and Altamira do not differ from us in any phenotypic fea­tures, their stunning achievement seems less mysterious. For the two more substantial cases, the 0.9 to 1.0 million years of stasis in the first well docu­mented hominid species, Australopithecus afarensis (aka “Lucy”), has been presented with much data and commentary (Kimbel, Johanson and Rak, 1994; see discussion of popular misapprehensions in Gould, 1995). Grine (1993) has also recorded 0.8 million years of stasis in Australopithecus robustus from Swartkrans cave in South Africa.

  I am, in any case, gratified to note the changing presuppositions of this small, contentious and vital field of paleoanthropology. In early years of this debate, after refuting the Cronin et al. (1980) hypothesis, Jacobs and Godfrey (1982, p. 85) wrote: “The Hominidae can no longer be blissfully assumed to be safely above the punctuationist challenge to the gradualist orthodoxy.” Just twelve years later, McHenry could assert in the closing line of his review (1994): “It is interesting, however, how little change occurs within most hominid species through time.”

  This elevation of stasis to visibility, respectability and even to expecta­tion has generated subtle and interesting repercussions for gradualism. When gradualism enjoyed high status as a virtually definitional consequence of evo­lution itself, few researchers thought to question such an anticipated result (but simply rejoiced in any rare instance of affirmation). However, once stasis emerges as an alternative norm, with gradualism designated as uncommon by the same analysis, then gradualism itself must fall under scrutiny for the first time.

  With this shift of perspective, a paradox that should have been obvious from the start finally emerged into clear view: gradualism, prima facie, repre­sents a “weird” result, not an anticipated and automatic macroevolutionary expression of natural selection — thus, perhaps, accounting for its rarity. Geo­logical gradualism operates far too slowly to yield any workable effect at all when properly scaled down and translated to the immediacy of natural selec­tion in local populations! (See Jablonski, 1999, for a forceful assertion of this paradox.)

  Again, we encounter the major dilemma that I call (Gould, 1997f) “the paradox of the visibly irrelevant” — that is, phenomena prominent enough to be detectable and measurable at all in local populations during ordinary hu­man time must cascade to instantaneous completion when scaled into geolog­ical time, whereas truly gradual effects in geological time must be effectively invisible at scales of human observation in ecological time. Consequently, what we see in our world can't be the direct stuff, by simple extrapolation, of sustained macroevolutionary change — while what we view as slow and steady in the geological record can't be visible at all (in the same form) by the measuring rod of our own life's duration.

  Eldredge and I first raised this point explicitly in 1977 (Gould and Eldredge, 1977), for we had missed this implication in our original formula­tion of 1972. Here, on this issue, we finally caught the attention of many [Page 835] neontological colleagues who, before then, had been unmoved by punctuated equilibrium. How can geological gradualism be the extrapolated expression of natural selection within populations? Surely, if a doubling of tooth size (say) requires 2 million years to reach completion, then the process must be providing so small an increment of potential advantage in each generation that natural selection couldn't possibly “see” the effect in terms of reliably en­hanced reproductive success on a generational basis. Can a tooth elongated by a tiny fraction of a single millimeter possibly confer any evolutionary ad­vantage in a selective episode during one generation of a population's history? Conversely, if bigger teeth provide such sustained advantages, why stretch the process over millions of years? Neontological studies have amply confirmed that natural selection can be a powerful force — the lesson, after all, of our en­tire, and burgeoning, literature of measurable change in Darwin's finches, anolid lizards, peppered moths, etc. So why shouldn't such a doubling of tooth length be achieved over the palpable span of a few human generations? Of course we all recognize a host of standard arguments for reining in the speed of selective response: negative consequences through discoordination with other parts of the body, slowing by networks of correlated effects upon other anatomical features. But I doubt that even the summation of all such ef­fects could generate sufficient restraining power to spread the blessings of a moment over two million years of plodding achievement. (See, however, p. 540 for Mayr's confident assertion, a priori and without evidence, of this evolutionary style and rate as canonical).

  In other words, gradualism should be viewed as a problem and a potential anomaly, not as an expectation. In an important early recognition of this principle, Lande (1976), who (to say the least) is no friend of punctuated equilibrium, calculated that Gingerich's measured trends confer such a small effect upon the immediacy of ecological moments that, for one case, Lande calculated an advantage corresponding to elimination of individuals four or more standard deviations from the mean in regimes of truncation selection! However unrealistic one might deem such a model, no one should miss the “bottom line”: most populations don't include any viable individuals four standard deviations from the mean — and one can hardly imagine that the re­moval of such occasional misfits or anomalies could slowly move the mean value of a population to new adaptive heights over a million sustained years.

  I do not mean to say that this paradox cannot be resolved to make gradual­ism intelligible once again, but I do hold that any revalidation demands a sub­stantial reconceptualization for this venerable phenomenon. The obvious so­lution lies embedded in results such as Bell and Haglund (1982) on the fine-scale structure of stasis. Selection in the immediacy of ecological moments cannot be measured as either the net nontrending of stasis or the steady accu­mulations of changing means in anagenetic gradualism. Any local population constantly jiggles to and fro in selective accommodation to changing local en­vironment (as when mean coloration for peppered moths becomes darker for a few centuries, but then lighter again, and back to previous values, when li­chens return to trees after abatement of industrial pollution). The extent of [Page 836] selection in an anagenetic sequence must be cumulated through each and ev­ery one of these jiggles, not measured by calculating the coefficients needed simply to change one endpoint into another. (Such a tactic would lead to the evidently false conclusion that little or no selection had ever occurred in pep­pered moths.) In other words, perhaps we must construe gradualism itself as a “higher level” phenomenon of net accumulation through the jiggles, not as an expression of ordinary directional selection summed through the ages.

  But such a conclusion then raises a different (and broader) question: what, then, is ordinary geological gradualism after all? How can such a minuscule directional effect persist through all the swings and jiggles? And what does such a phenomenon represent? Must we interpret such slow net change as caused by drift, as Lande's models made conceivable? Such a conclusion would seem unlikely given the common impression that certain features, size increase in particular, occur preferentially and nonrandomly in gradualistic sequences (but see Jablonski, 1997, and Gould, 1997b, on the apparent fal­sity, and status as a psychological artifact, of this venerable claim known as “Cope's Rule”). Can we even argue for natural selection as the primary cause of classical gradualism at all? I am confident that selection remains a good candidate, but of what sort, and at wh
at level? The selective basis of gradual­ism surely cannot be ascribed to the extrapolated advantage at every given moment of traits so enhanced over the long run. Rather, the selective edge must lie in some form of more general benefit not consistently visible in eco­logical moments, but somehow skewed to a higher probability of immediate occurrence that can then cumulate to a consistent trend in macroevolutionary time.

  One might be tempted to equate this skewing agent with some form of gen­eral biomechanical improvement that might hold cumulative sway above the jiggling of momentary advantage in any direction. But then the kinds of fea­tures that seem to prevail in gradual anagenesis do not stand out for potential membership in this category. Perhaps we need to consider selection on supra-organismal units, or perhaps we should entertain nonselectionist alternatives, especially in the light of Lande's modelling for drift. (Such hypotheses of ran­dom change would require a far better knowledge of relative frequencies, both for characters within a taxon and among taxa themselves, than we now possess or even know how to generate.) In any case, I do not think we have even begun to explore the range of potential explanations for the puzzling phenomenon of anagenetic gradualism. I, at least, find the subject very con­fusing and challenging.