The Structure of Evolutionary Theory

by Stephen Jay Gould

Among such geographically confined clades, Vrba's classic studies (1984a and b) of African antelopes stand out for detailed data on one of the most successful and speciose of vertebrate higher taxa. In the maximally diverse tribe Alcelaphini (including blesbucks, hartebeests, and wildebeests), the Quaternary record includes 25 species, all with a geologically sudden origin in recorded data, and with cladogenesis as a reliably inferred mode of origin for at least 18 species. Several species lived for 2 million years or longer in sta­sis, and no ancestors with incrementally transitional morphologies have been found for any of these forms.

  I continue to be amazed by the skewed interpretation often imposed by gradualistic expectations upon data for clades that seem, at least in my parti­san judgement, clearly dominated by punctuated equilibrium in overall rela­tive frequency. For example, in a well-known work, White and Harris (1977) used the Plio-Pleistocene record of African pigs for supposed validation of gradualism as a primary guide in biostratigraphic resolution (particularly of [Page 867] some important hominid-bearing strata). They did document one or two cases of gradual change, notably an increase in third molar length for Mesochoerus limnetes. But the clade includes 16 species during this short period of no more than 4 million years, 8 of which arise by punctuational cladogenesis even in White and Harris's own diagram (1977, p. 14). The authors' com­ments, unwittingly I suspect, frequently point to the domination of evolution­ary history in this clade by cladogenetic events and their consequences. They write (1977, p. 14), for example, that “Metridochoerus underwent a substan­tial adaptive radiation during the early Pleistocene, and at one point four dis­tinct metridochoere species existed contemporaneously.”

  Many of the best invertebrate examples fall into the same category of unique and endemic taxa confined to isolated places, and therefore forming, by strong inference, a complete and coherent phylogenetic unit — as in Wil­liamson's study (1981), cited several times previously, of speciation in pulmonate snails from separated African lakes. In another example from a fa­mous sequence of much greater temporal extent, Geary (1990, 1995) studied the evolution of melanopsid gastropods in the Middle to Late Miocene beds (spanning 5 to 10 million years) of the Pannonian Basin in Eastern Europe. In one case of gradualism, following a much longer interval of at least 7 million years in stasis for the ancestral form, Melanopsis impressa transformed to M. fossilis by directional increases in shell size and shouldering over a two mil­lion year interval. However, within the same Pannonian Stage, at least six new melanopsid species arose by punctuation: “their first appearances are abrupt, and preceded by no intermediate forms” (Geary, 1995, p. 68). Geary (p. 69) regards the stratigraphic resolution as “not particularly good,” but still fixes the origin of these species to within “tens of thousands of years” — a clear punctuation by the criterion of scaling against average species duration in stasis within the clade. Figure 9-27 (from Geary, 1995, p. 68) depicts Geary's results for this geographically isolated evolutionary radiation.

  As mentioned many times in several contexts within this chapter, Alan Cheetham's studies of the bryozoan Metrarabdotos (1986, 1987), now supplemented with the work of Jackson and Cheetham (1994, and Cheetham and Jackson, 1995) on Stylopoma, have set a standard of excellence and con­fidence for empirical studies of relative frequency. All major desiderata for such research have been realized in these genera — a group with a well-resolved phylogeny, in a clade restricted to a geographic region, and ex­haustively sampled in strata of unusually complete resolution over a long pe­riod. Moreover, Cheetham's multivariate morphometrics permit us to assess stasis and punctuation as a morphological totality, not only as a potentially biased impression based on a few preselected characters. Finally, studies with Jackson on the ecology and genetics of extant species demonstrate (see page 786) that the morphology of paleospecies almost surely provides a good sur­rogate and identifier for true biospecies in this clade. (As a personal note, I am also gratified that Cheetham began these studies with the intention of proving his suspicions for gradualism in the context of the developing debate about punctuated equilibrium — and ended up with the finest data-driven evidence [Page 868] ever gathered for the domination of a total evolutionary pattern by punctu­ated equilibrium.)

  To recapitulate the major conclusions of these studies (see also Figure 9-18 for the phylogeny of Metrarabdotos, with morphology expressed as multivariate Euclidian distances between samples based on all canonical scores of a discriminant analysis, and connecting nearest morphological neighbors in stratigraphic sequence), Cheetham measured 46 characters in 17 species of Metrarabdotos over a duration of 15 million years, with intense sampling for a 4.5 million year interval of Upper Miocene to Lower Pliocene sedi­ments (3.5 to 8.0 million years ago) in the Dominican Republic. Cheetham (1986, p. 195) specified the favorable features of Metrarabdotos on both geo­graphic and phylogenetic grounds: “The ascophoran genus Metrarabdotos is a favorable subject for detailed analysis of evolutionary pattern because of its diversity and wide distribution during much of Miocene and Pliocene time. Caribbean species . . . form an apparently monophyletic subset within which phylogenetic relationships can be inferred independently of evolutionary events in eastern Atlantic-Mediterranean congeneric species groups.”

  Moreover, the unusual resolution for the detailed sampling interval permits “a fine-scale comparison of successive populations similar to those made with oceanic planktonic groups in deep-sea cores” (1986, p. 195), thus dispelling the common argument of gradualists that stasis in metazoans from conven­tional sediments must arise as an artifact of coarseness of resolution, while

  9-27. From Geary, 1995. In the radiation of melanopsid gastropods in Middle to Late Miocene beds (spanning 5-10 million years) of the isolated Pannonian Ba­sin, Geary found one case of gradualism where, after at least 7 million years of stasis, ancestral M. impressa transformed to M. fossilis by directional increase in shell size and shouldering over a million year interval. However, during the same time, at least six new melanopsid species arose by punctuation, as also shown.

  [Page 869]

  gradualism of microfossils in deep-sea cores must record a general pattern that would be seen wherever stratigraphic sampling could attain such com­pleteness. Cheetham (1986) calculated 160,000 years for average spacing be­tween successively sampled populations in the intensely collected interval — for a stratigraphic completeness of 0.63 by Sadler's (1981) criteria.

  In Metrarabdotos (again see Fig. 9-18), 11 of the 17 species persist in stasis for 2-6 million years, and all originate punctuationally within the limit of resolution (at least in the intensely sampled interval) of 160,000 years — undoubtedly in far less time for many branching events, since 160,000 years represents a maximum figure based on the available unit of measurement. Again for the intensely sampled interval, Cheetham writes (1986, p. 190) that “nine comparisons of ancestor-descendant species pairs all show within-species rates of morphologic change that do not vary significantly from zero, hence accounting for none of the across-species difference. In all cases, the ra­tio of within-species fluctuation to across-species difference is low enough to allow the punctuated pattern to be distinguished with virtual certainty. In at least seven of the cases, ancestor species persisted after giving rise to descen­dants, in conformity with the punctuated equilibrium mode of evolution.”

  The morphometric details can only increase confidence in “the remarkably clear-cut evidence for a punctuated evolutionary pattern in these Metra­rabdotos species” (1986, p. 201). The reported central tendencies of samples integrate data from 46 measurements, providing a good assessment of gen­eral anatomical distance (based on characters considered important in the taxonomy and functional morphology of these organisms), and not on se­lected single characters (see Cheetham, 1987, for affirmation of punctuated equilibrium from analyses of temporal trends in individual characters as well). In supplementary affirmation, Cheetham studied the fine-scale pattern of temporal variation by computing autocorrela
tions between mean scores of stratigraphically successive pairs of populations: “In all cases, the autocor­relations of mean scores of successive populations are nonsignificant and near zero, and the autocorrelations of rate deviations are negative and (except in one case) nonsignificant. These autocorrelations clearly indicate that changes within species are fluctuations around a near zero, otherwise unchanging rate” (1986, p. 201).

  Finally, some authors (see Marshall, 1995) have challenged Cheetham's phylogeny for its stratophenetic basis. But a purely cladistic analysis, as now preferred by many researchers, not only changes the previous scheme in only minor ways (Cheetham and Jackson, 1995, p. 192), but also — and the point becomes almost amusingly obvious once one grasps the different criteria used by the two methods — leads to an even stronger pattern of punctuated equilib­rium, for the stratophenetic phylogeny minimizes the mean morphologic dis­tances between putative ancestor-descendant pairs, while the cladistic phy­logeny makes no such assumption and must therefore yield a larger mean difference between species. Since the documented stasis within species is not affected in either case, the cladistic scheme must increase the average magni­tude of punctuational events, thus only decreasing the likelihood that between-species [Page 870] differences could be extrapolated from temporal variation within species (see Cheetham and Jackson, 1995, p. 192, for an elaboration of this argument with appropriate data).

  The corroborative study of a second bryozoan genus, Stylopoma, from the same beds yields an identical conclusion of overwhelming predominance — indeed, exclusivity — for punctuated equilibrium. Jackson and Cheetham (1994) used 12 morphological features to identify 19 species in this rarer ge­nus. Despite their more limited information, Cheetham and Jackson (1995, p. 195) found that “temporal overlap between putative ancestor-descendant species pairs is even greater than for Metrarabdotos, with 10 species surviv­ing beyond the detailed sampling interval more than 6 million years to the Holocene.” Moreover, “no evidence of morphologically intermediate forms” (Jackson and Cheetham, 1994, p. 420) has been found for any transition; all species origins are fully punctuational at the scale of detailed sampling.

  Finally, since Stylopoma provided Jackson and Cheetham's principal data for the correspondence of genetically defined biospecies with morphologi­cally designated paleospecies (modern specimens of Metrarabdotos are much less common and not so well suited for genetic work), this second study pro­vides strong additional support for punctuated equilibrium by coordinating several potentially independent indicators of evolutionary change with rapid events of branching speciation: “Moreover, the tight correlation between phenetic, cladistic, and genetic distances among living Stylopoma species sug­gests that changes in all three variables occurred together during speciation. All of these observations support the punctuated equilibrium model of spe­ciation.”

  I regard these empirical studies of relative frequencies as the strongest evidence now available for the most important and revisionary claim made by the theory of punctuated equilibrium: the overwhelming domination of evo­lutionary patterns in geological time by events at the species level (or higher), and the consequent need to explain macroevolution by patterns of sorting among species rather than by extrapolated trends of anagenetic transforma­tion within continuous lineages.

  Causal clues from differential patterns of relative frequencies

  Once we set our focus of inquiry on determining the relative frequencies of punctuated equilibrium in different times, places, environments, and taxa, we can ask the classic question of natural history, a subject rooted in the con­cept of variation: do we note characteristic differences in relative frequencies based on any of these factors and, if so, can we draw any causal inferences (useful to evolutionary theory) from these patterns. I have already raised this question in a number of preceding contexts in this chapter, most extensively for the observed higher frequency of gradualism in predominantly asexual oceanic protistan lineages, where I argued that this unusual result may not re­cord the greater completeness of strata in oceanic cores (as traditional views have assumed), but probably arises from interesting biological differences that have led us to look for a truly underlying punctuational pattern at the wrong scale in this case (see pp. 803–810). [Page 871]

  Most discussion on the linkage of differences in frequencies to distinctive structural and functional characteristics of organisms (rather than to types of environments per se) has focussed on claims for variation among broad taxonomic groups (high frequency for gradualism in rodents vs. rarity in bovids, for example), but I strongly suspect that this genealogical emphasis reflects traditions of specialization in research more than any inherent preference for taxonomic parsing in such a search. We should also consider more gen­eral features of organisms that cut across taxonomic lines, and we should therefore examine broader differentia potentially related to chosen environ­ments or tendencies to speciate. Schoch (1984), for example, suggested a link between high frequencies for punctuational speciation and intense so­cial competition, arguing that selection on such features tends to proceed so rapidly, even in ecological time, that speciation would almost surely oc­cur in a geological instant. Breton (1996) linked punctuational modes to evolution of “pioneer” structures (evolutionary novelties tied to morpho­logical reorganizations), and gradualism to “stabilization” and “settlement” structures (refinements and improvements in local adaptations). I suspect that such arguments may apply better to the different issue of average amounts of change per speciation event, than to questions about the rela­tive frequency of punctuational events (at whatever degree of alteration) per se.

  Most arguments about patterns of differences in relative frequencies have invoked “externalist” claims about characteristic environments, rather than “internalist” correlations with structural features of organisms (although the two subjects may, of course, be correlated and need not stand in antithesis). In a first attempt, Johnson (1975, 1982), working with Devonian brachiopods and conodonts but generalizing more widely in an important set of pa­pers, linked higher frequencies of gradualism to pelagic environments and greater prevalence of punctuated equilibrium to benthic habitats. He then justified the ecological correlations by linking characteristic evolutionary modes with relative stability of environments: “Among marine invertebrates, pelagic organisms are the most likely to have inhabited extensive, gradually changing environments and are therefore the most likely to have evolved by a rate and pattern that can be described as phyletic gradualism . . . Post-larval, attached and stationary benthic organisms are the most likely to have inhab­ited environments that are subject to relatively abrupt changes and are there­fore the most likely to have evolved by a rate and pattern that could be de­scribed as punctuated equilibria.”

  In a series of papers, Parsons (well summarized in 1993) suggested a simi­lar linkage, while proposing a different, but generally concordant, explana­tion based on a putative correlation between environmental “stress” and pat­terns of genetic variation and available “metabolic energy.” (I put Parsons's last factor in quotation marks because I have trouble grasping both the defini­tion and operationality of such a concept.) Parsons writes (1993, p. 328): “In moderately stressed and narrowly fluctuating environments, sufficient genetic variability and metabolic energy should be available to permit adaptation. In these environments, phyletic gradualism is expected. In highly stressed and [Page 872] widely fluctuating environments, a punctuated evolutionary pattern is ex­pected whereby stasis occurs most of the time.”

  In one of the most important recent papers on punctuated equilibrium, Sheldon (1996) has generalized a superficially paradoxical link of morpho­logical stasis to highly fluctuating environments, and gradualism to more sta­ble and narrowly fluctuating environments, as the “plus ca change” model — citing the sardonic French motto that “the more things change, the more ev­erything's the same,” a reference to the proposed link of morphological stasis with highly variable e
nvironments. Sheldon (1996, p. 772) explained the ba­sis and resolution of the paradox: “One might expect a changing environ­ment to lead to changing morphology, and a stable environment to stable morphology. But over long intervals the opposite may often occur ... Perhaps gradual phyletic evolution can only be sustained by organisms living in or able to track narrowly fluctuating, slowly changing environments, whereas stasis, almost paradoxically, seems to prevail in more widely-fluctuating, rap­idly changing environments.” Sheldon's figure (reproduced here as 9-28) will make his argument clear. Species in highly variable habitats must adapt to pervasive and rapid fluctuations, and generally do so by evolving a stable and generalized morphology suited to the full environmental range. But when external fluctuation exceeds a certain limit of internal toleration, rapid speciation may be the only viable response. On the other hand, mildly fluctuat­ing environments may enhance selection for more precisely tuned adaptations capable of tracking long-term climatic trends by gradual adjustment.

  Sheldon (1987) began his work by publishing one of the most widely discussed empirical defenses of gradualism, based on several lineages of Ordovician trilobites from the Builth Inlier, an environment interpreted as generally stable and only narrowly fluctuating. (I appreciate the richness of Sheldon's data, but regard his interpretations as ambiguous, for most of his published trajectories seem to me — from my partisan standpoint (as I keep repeating to

  9-28. An epitome of Sheldon's argument (1996) that, paradoxically, highly fluctuating environments may induce stasis and punctuation, with gradualism more commonly found in environments undergoing slower but more steady change.

  [Page 873]

  remind readers to be especially critical) — more consistent with expectations of fluctuating stasis with little or no net change, see Eldredge and Gould, 1988.) I am particularly grateful that Sheldon, even while developing one of the most famous data sets against the theory, has always accepted the impor­tance of establishing relative frequencies for different groups and situations, and has consistently regarded punctuated equilibrium as a valuable theory (to which he has made major contributions), with important implications for our understanding of macroevolution.