The Structure of Evolutionary Theory

by Stephen Jay Gould

  The other obvious explanation, in a gradualistic and anagenetic world ruled by conventional selection, held that living fossils had stagnated because they lacked genetic variation, and therefore presented insufficient fuel for Darwinian change. This more plausible idea seemed sufficiently intriguing that Selander et al. (1970), in the early days of electrophoresis as a novel method for measuring overall genetic variation, immediately applied the tech­nique to Limulus, the horseshoe crab — and found no lowering of genetic variability relative to known levels for other arthropods. This negative pat­tern has held, and no standard lineage of living fossils exhibits depauperate levels of genetic variability.

  But punctuated equilibrium suggests another, remarkably simple, explana­tion once you begin to think in this alternative mode — an insight that ranks in the exhilarating, yet frustrating, category of obvious “scales falling from eyes” propositions, once one grasps the new phrasing of a basic question. If evolutionary rate correlates primarily with frequency of speciation — the car­dinal prediction of punctuated equilibrium — then living fossils may simply represent those groups at the left tail of the distribution for numbers of speciation events through time. In other words, living fossils may be groups that have persisted through geological time at consistently and unvaryingly low species diversity. (Average species longevity need not be particularly high, [Page 817] for low species numbers, if consistently maintained without geological bursts of radiation, will yield the full effect.) Such groups cannot be common — for consistently low diversity makes a taxon maximally subject to extinction in our contingent world of unpredictable fortune, where spread and number represent the best hedges against disappearance, especially in episodes of mass extinction — but every bell curve has a left tail.

  This explanation holds remarkably well, and probably provides a basic explanation of “living fossils.” Such groups are neither mysteriously optimal, nor unfortunately devoid of variability. They simply represent the few higher taxa of life's history that have persisted for a long time at consistently low species number — and have therefore never experienced substantial opportu­nity for extensive change in modal morphology because species provide the raw material for change at this level, and these groups have never contained many species.

  Westoll (1949), for example, published a classic study, summarized again and again in treatises and textbooks (Fig. 9-12), showing that lungfishes evolved very rapidly during their early history, but have stagnated ever since. The literature abounds in hypothesized explanations based on adaptation and ecological opportunity in an anagenetic world. The obvious alternative stares us in the face, but rises to consciousness only when theories like punc­tuated equilibrium encourage us to reconceptualize macroevolution in speciational terms: in their early period of rapid evolution, lungfishes maintained high species diversity, and could therefore change quickly in modal mor­phology. Their epoch of later stagnation correlates perfectly with a sharp re­duction of diversity to very low levels (only three genera living today, for example) with little temporal fluctuation in numbers — thus depriving macroevolution of fuel for selection (at the species level), and relegating lungfishes to the category of living fossils.

  A breakthrough in the application of quantitative modelling to cladistic patterns of evolution directly recorded in the fossil record has been achieved by Wagner (1995 and 1999) and Wagner and Erwin (1995). These authors show, first of all, the pitfalls of working only with cladistic information from living organisms, and they illustrate the benefits of incorporating stratophenetic data from the fossil record into any complete analysis (see Wagner and Erwin, 1995, pp. 96-98, in a section entitled “why cladistic topology is insufficient for discerning patterns of speciation”). They then build models based on three alternative modes of evolution, and characterize the differ­ences in cladistic pattern expected from each: anagenetic gradualism, specia­tion by “bifurcation” (where, after branching, the two descendant species both accumulate differences from an ancestor then recorded as extinct), and speciation by “cladogenesis” (where one daughter species arises with autapomorphic differences, but the ancestral species persists in stasis). Cladogenesis is usually defined — both in this book and in the evolutionary literature in general — as any style of evolution by branching of lineages rather than by transformation of a single lineage (anagenesis). Wagner and Erwin restrict the term “cladogenesis” to the mode of speciation predicted by punctuated equilibrium —

  [Page 818]

  9-12A. The famous figure from Westoll (1949) showing rapid morphological change early in the history of lungfishes, followed by prolonged stagnation thereafter.

  branching off of a descendant, leaving a persisting and unaltered ancestor. They contrast this mode with bifurcation — the style of speciation predicted by gradualism: splitting of an ancestral population into two descen­dant species, both diverging steadily from the ancestor (which becomes ex­tinct). I follow Wagner and Erwin's restricted use of “cladogenesis” only in discussing their work, and use the broader definition throughout the rest of this book.

  The last two modes of bifurcation and cladogenesis both depict branching speciation in the definitional sense that two species emerge, where only one existed before. But note the crucial difference: bifurcation represents the op­eration of speciation in a gradualistic world, where an event of branching may be considered equivalent to two cases of gradualism following a separa­tion of populations, and where the separation itself need not correlate with any acceleration in rate of evolutionary transformation. Cladogenesis, on the other hand, represents the predictions and expectations of punctuated equi­librium. Therefore, if we can model the differences between bifurcation and cladogenesis, and test these distinctive expectations against real patterns in nature, we may achieve our best and fairest potential evaluation for the rela­tive frequency of punctuated equilibrium — for punctuated equilibrium cannot

  [Page 819]

  9-12B. Redrawing and simplification of these data in the excellent paleontological textbook of Raup and Stanley (1971). The bottom icon, showing an early mode and a right skew, has become canonical in textbooks. The data are firm and fascinating, but the interpretation has general been faulty as a result of gradualistic and anagenetic assumptions. Lineages did not stagnate in any anagenetic sense; rather, species diversity became so dramatically lowered (and has always stayed so — only three genera of lungfishes remain extant today) that speciational processes have never again had enough fuel to power further exten­sive phyletic change.

  be affirmed merely by showing that realized evolutionary patterns must record speciation and cannot be rendered by anagenetic, end-to-end stacking. Even the most committed anagenetic gradualists never denied the importance and prevalence of speciation. They hold, rather, that speciation generally oc­curs in the gradualistic mode — as two cases of divergence at characteristic rates for unbranching lineages — and not, as supporters of punctuated equilib­rium maintain, as geologically momentary bursts representing the budding of descendant populations from unchanged, and usually persisting, ancestral species in stasis. Thus, the best possible test for punctuated equilibrium must distinguish between the expectations of bifurcating vs. cladogenetic models of speciation.

  I am embarrassed to say that neither I nor my colleagues working on the validation of punctuated equilibrium ever conceptualized the simple and obvious [Page 820] best test for distinguishing the bifurcating model of speciational gradu­alism from the cladogenetic model of punctuated equilibrium. In this case, the impediment may be clear, but I can offer no legitimate excuse for my opac­ity — and I congratulate Wagner and Erwin on their formulation.

  The solution lies in the distribution and frequency of “hard” polytomies in cladogenetic topologies. I failed to appreciate the following point: under punctuated equilibrium, new species branch off from unchanged and persist­ing ancestors. The successful ancestor remains in stasis and may live for a long time. Therefore, these “
stem” species may generate numerous descen­dants during their geological tenure, while remaining unchanged themselves. Now what cladistic pattern must emerge from such a situation? A group of species branching at different times from an unchanged ancestor must yield a cladistic polytomy. Cladograms cannot distinguish different times of ori­gin from an unaltered ancestor, and can therefore only record the phenetic constancy of the common and unchanging ancestor as a polytomy, for all branches emerge from an invariant source. Bifurcation, on the other hand, can produce a range of cladistic topologies (Wagner and Erwin, 1995, p. 92), but not domination of the overall pattern by polytomies. Thus, gradualistic vs. punctuational models of speciation should be distinguishable by distribu­tions of polytomies in the resulting cladogram.

  I suspect that many of us never recognized this point because we have been trained to view polytomies negatively as an expression of insufficient data to resolve a true set of ordered dichotomies. (Shades of our profession's former failure to conceptualize punctuated equilibrium because we had been trained to view geologically rapid appearances as artifacts of an imperfect fossil re­cord!) Thus, we never recognized that polytomies might also be denoting a positive and resolvable pattern — multiple branching through time of several species from an unaltered ancestral source. Of course — and, again, just as with punctuated equilibrium itself — polytomy can also result from imperfec­tion, and we need criteria to separate “real” polytomies representing a signal from the history of life from polytomies that only record artifact of an im­perfect record. Wagner and Erwin (1995) develop such a criterion by distin­guishing between “hard” polytomies that include the persisting ancestor and “soft” polytomies that arise from an inability to resolve true sets of ordered dichotomies.

  Wagner and Erwin's modelling demonstrates the translation of punctua­tional speciation to a cladistic pattern of predominant polytomies. (Wagner and Erwin used my own model of punctuational phylogenies, done with D. M. Raup in the 1970's (Raup and Gould, 1974), to show this mapping of punctuational phylogeny to a polytomous cladogram — see Figure 9-13 — but I had never made the connection myself.)

  Wagner and Erwin then applied their modelled differences to cladograms for two well resolved, but maximally different (in taxon and time) species-level phylogenies in the fossil record: two Neogene clades of planktonic foraminifers (Globigerinidae and Globorotaliidae), and Ordovician representa­tives of the gastropod family Lophospiridae. In both cases, the cladograms indicated [Page 821] an overwhelming predominance of speciation by cladogenesis as a cause of phylogenetic patterning — thus affirming the predictions of punctu­ated equilibrium. For globigerinids, the cladistic topology revealed 40 specia­tion events, 5 probably anagenetic, 35 cladogenetic, and none bifurcating. Wagner and Erwin did not present full tabulations for the globorotaliids, but stated (p. 105): “The results are not presented here, but they were similar to those found for globigerinids: cladogenesis is significantly more common than anagenesis, a positive association exists between having long temporal ranges and leaving cladogenetic descendants, and no such association exists for anagenetic ancestors.”

  For the lophospirid gastropods, they write (p. 106): “Our preferred cladogram for lophospirids is rife with polytomies. Of the eleven polytomies, only

  9-13. To my embarrassment, Wagner and Erwin (1995) — for I had not seen the obvious implication that would have enormously helped my argument — showed how phylogenies based upon iteration of several species from an unchanged par­ent stock (as Raup and Gould, 1974, had generated, and Wagner and Erwin reproduced, at the top of this figure) must yield, in cladistic representation, a polytomy. Thus, polytomies may provide evidence for punctuated equilibrium and do not necessarily represent the “signature” of missing data needed to re­solve the system into dichotomies. If the ancestral form doesn't change through­out its geological range, all descendants must in principle arise at a polytomous junction of a cladogram.

  [Page 822]

  two do not include plesiomorphic species. Thus, nine may represent hard polytomies.” Of 42 implied speciation events, a maximum of six may have been anagenetic, while only one may represent a bifurcation. Again, cladogenetic speciation, the expectation of punctuated equilibrium, dominates the phylogenetic pattern.

  Wagner and Erwin's overall conclusion accords fully with patterns ex­pected in phylogenies built primarily — one might say overwhelmingly — by punctuated equilibrium (Wagner and Erwin, p. 110):

  • Cladogenesis is significantly more common than anagenesis.

  • Species with longer temporal and geographic ranges are more likely to leave descendants via cladogenesis or the factors contributing to wider temporal and geographic ranges also contribute to the likelihood of cladogenetic evolution.

  • If anagenesis occurs, it only applies to species with restricted temporal and geographic ranges.

  • Bifurcation accounts for a negligible amount of speciation.

  We cannot often obtain well-resolved species level phylogenies from paleontological data, and inferences from higher taxa will probably remain too murky and insecure to permit general use of such models for testing hypothe­ses explicitly based on the evolutionary behavior of species. Still, other data sets do exist in fair absolute abundance (while representing a low percentage of the total number of potential lineages in life's history). Studies like Wagner and Erwin's can be replicated and extended for many taxa — and such a strat­egy can provide powerful tests for the relative importance of punctuated equilibrium in the history of life and the generation of phylogenetic patterns. The first tests have been highly favorable, but we have scarcely any idea what an extended effort might teach us about the basic modalities of macroevolution.

  Sources of Data for Testing Punctuated Equilibrium

  PREAMBLE

  Punctuated equilibrium has generated a fruitful and far ranging, if sometimes acrimonious, debate within evolutionary theory (see appendix to this chap­ter). While we feel much pride (mixed with occasional frustration) for the role that punctuated equilibrium has played in instigating such extensive rethinking about the definitions and causes of macroevolution, we take even more pleasure in the volume of empirical study provoked by the theory of punctuated equilibrium, and pursued by paleontologists throughout the world. These carefully documented case studies (both pro and con) build a framework of proof for the value of punctuated equilibrium, as illustrated by the most important of all scientific criteria — operational utility. Such cases have been featured in numerous symposia and books dedicated to the empirical [Page 823] basis of punctuated equilibrium. This literature includes: the 1982 symposium in Dijon, France, entitled Modalites, rythmes, mechanismes de devolution biologique: gradualisme phyletique ou equilibres ponctues and published as Chaline, 1983; the 1983 Swansea symposium of the Paleontological Association (United Kingdom) on “Evolutionary case histo­ries from the fossil record” and published as Cope and Skelton (1985); the book The Dynamics of Evolution: The Punctuated Equilibrium Debate in the Natural and Social Sciences (Somit and Peterson, 1992) that began as a sym­posium for the annual meeting of the American Association for the Advance­ment of Science, and then appeared as a special issue (1989) of the Journal of Biological and Social Structures; the 1992 symposium of the Geological Soci­ety of America on “Speciation in the Fossil Record,” held to celebrate the 20th anniversary of punctuated equilibrium, and published in book form as Erwin and Anstey (1995); and the 1994 Geological Society of America sym­posium on coordinated stasis, published in a special issue of the journal Palaeogeography, Palaeoclimatology, Palaeoecology in 1996 (volume 120, with Ivany and Schopf as editors). Several other unpublished symposia, in­cluding the notorious Chicago macroevolution meeting of 1980 (see pages 981–984), focused upon the topic of punctuated equilibrium. Finally, several books have treated punctuated equilibrium as an exclusive or major topic, including the favorable accounts of Stanley (1979), Eldredge (1985, 1995), and Vrba (1985a), and the strongly nega
tive reactions of Dawkins (1986), Dennett (1995), Hoffman (1989), and Levinton (1988).

  As emphasized throughout this book, most general hypotheses in natural history, with punctuated equilibrium as a typical example, cannot be tested with any single “crucial experiment” (that is, by saying “yea” or “nay” to a generality after resolving a case with impeccable documentation), but must stand or fall by an assessment of relative frequency. Moreover, we can't estab­lish a decisive relative frequency by simple enumerative induction (as in clas­sical “beans in a bag” tests of probability) — for individual species cannot be treated as random samples drawn from a totality with a normal (or any other kind of simple) distribution, but represent unique items built by long, com­plex and contingent histories. Time, taxon, environment and many other fac­tors strongly “matter,” and no global evaluation can be made by counting all cases equally. We may, however, be able to reach robust solutions for full pop­ulations within each factor — for planktonic forams, terrestrial mammals, De­vonian brachiopods, or species of the Cambrian explosion, for example. Part C of this section reports several such studies, nearly all finding a predominant relative frequency for punctuated equilibrium.

  Nonetheless, hundreds of individual cases have been documented since we proposed punctuated equilibrium in 1972. I do not think that most authors pursue such work under any illusion that they might thus resolve the general debate, but rather for the usual, and excellent reasons of ordinary scientific practice. Researchers pursue such studies in order to apply promising general concepts to cases of special interest that draw upon their unique skills and expertise. [Page 824] Such studies are pursued, in other words, to resolve patterns within Australopithecus afarensis, or among species in the genus Miohippus, not to adjudicate general issues in evolutionary theory.