The Structure of Evolutionary Theory

by Stephen Jay Gould

  Nonetheless, compendia of such studies do provide a “feel” for generalities of data in admittedly non-randomized samples, and they do establish ar­chives of intriguing and well-documented cases both for pedagogical illustra­tion, and simply for the general delight that all naturalists take in cases well treated and conclusively resolved. I shall therefore discuss this mode of docu­mentation as practiced for two categories central to punctuated equilibrium: patterns of gradualism or stasis within unbranched taxa (part B of this sec­tion), and tempos and modes of branching events in the fossil record (part C). Part D will then treat the more decisive theme of relative frequencies.

  THE EQUILIBRIUM IN PUNCTUATED EQUILIBRIUM: QUANTITATIVELY DOCUMENTED PATTERNS OF STASIS IN UNBRANCHED SEGMENTS OF LINEAGES

  As previously discussed (see pp. 758–765), the main contribution of punctu­ated equilibrium to this topic lies in constructing the theoretical space that made such research a valid and recognized subject at all. When paleontolo­gists equated evolution with gradual change, the well-known stasis of most lineages only flaunted a supposed absence of desired information, and could not be conceptualized as a positive topic for test and study. By representing stasis as an active, interesting, and predictable feature of most lineages most of the time, punctuated equilibrium converted an unconceptualized negative to an intriguing, and highly charged positive, thereby forging a field of study.

  Nonetheless, we cannot argue that a proven predominance of stasis within lineages can establish the theory of punctuated equilibrium by itself. Punctu­ated equilibrium implies and requires such stasis, but remains, primarily, a theory about characteristic tempos and modes of branching events, and the primary patterning of phyletic change by differential birth and death of species.

  Stasis has emerged from the closet of disappointment and consequent non-recording. At the very least, paleontologists now write, and editors of jour­nals now accept, papers dedicated to the rigorous documentation of stasis in particular cases — so skeptics, and scientists unfamiliar with the fossil record, need not accept on faith the assurances of experienced paleontologists about predominant stasis in fossil morphospecies (see pp. 752–755). Moreover, sta­sis has also become a subject of substantial theoretical interest (see pp. 874–885), if only as a formerly unexpected result now documented at far too high a frequency for resolution as an anticipated outcome within random systems (Paul, 1985); stasis must therefore be actively maintained. In any case, pale­ontologists are now free to publish papers with such titles as: “Cosomys pri­mus: a case for stasis” (Lich, 1990), and “Apparent prolonged evolutionary stasis in the middle Eocene hoofed mammal Hyopsodus” (West, 1979).

  The study of McKinney and Jones (1983) may be taken as a standard and [Page 825] symbol for hundreds of similar cases representing a characteristic mixture of satisfaction and frustration. These authors documented a sequence of three successional species of oligopygoid echinoids from the Upper Eocene Ocala Limestone of Florida. The two stratigraphic transitions are abrupt, and there­fore literally punctuational. But available evidence cannot distinguish among the mutually contradictory explanations for such passages: gradualism, with transitions representing stratigraphic gaps; rapid anagenesis for a variety of plausible reasons including population bottlenecks or substantial environ­mental change; punctuated equilibrium based on allopatric speciation else­where (or unresolvably in situ, given coarse stratigraphic preservation), and migration of new species to the ancestral range. Hence, frustration. (More­over, as this pattern represents the most frequent situation in most ordinary sequences of fossils, we can readily understand why the testing of punctua­tional claims within the theory of punctuated equilibrium requires selection of cases — fortunately numerous enough in toto, however modest in relative frequency — with unusual richness in both spatial and temporal resolution.)

  At the same time, however, we gain satisfaction in eminent testability for the set of claims representing the second key concept of stasis. Any species, if well represented throughout a considerable vertical span marking the hun­dreds of thousands to millions of years for an average duration, can be reli­ably assessed for stasis vs. anagenetic gradualism by criteria outlined pre­viously (pp. 765–774). McKinney and Jones (1983) compiled excellent evidence for stasis in each of their three species — the basis, after all, for using these taxa in establishing biozones for this section. (As argued on pp. 751–752, biostratigraphers have always used criteria of stasis and overlapping range zones in their practical work on the relative dating of strata.) McKinney and Jones conclude (1983, p. 21): “These observations suggest there is little chance of species misidentification due to ontogenetic or phylogenetic effects when using this lineage for biostratigraphic purposes.”

  Smith and Paul (1985) studied vertical variation of the irregular echinoid Discoides subucula in a remarkably complete and well-resolved sequence of Upper Cretaceous sands. The species occurred throughout 8.6 m of section, apparently representing continuous sedimentation within one ammonite zone spanning less than 2 million years. The authors were able to sample meter by meter through a section with an interesting inferred environmental his­tory: “The sediment that was then being deposited changed from clean, well-washed sand to a very muddy sand, and so one might expect to find evidence of phyletic gradualism in response to these changes” (1985, p. 36).

  Smith and Paul did measure a steady change in shape towards a more coni­cal form, a common response of irregular echinoids to muddy environments. But such an alteration can be ecophenotypically induced during ontogeny, and the authors see no reason to attribute this single modification to geneti­cally based evolution (while not, of course, disproving the possibility of such genuine gradualism). Otherwise, stasis prevails throughout the section: “In other, more important characters, D. subucula remains morphologically static and shows no evidence of phyletic gradualism” (1985, p. 29). [Page 826]

  This case becomes particularly interesting, and merits consideration here, as a demonstration of how far reliable inference can extend, even when the tempo and mode of origin for a descendant species cannot be directly resolved (the usual situation in paleontology). The potential descendant, D. favrina, enters the section near the top and overlaps in range with D. subucula, thus implying cladogenetic origin rather than anagenesis. The descendant's larger size and hypermorphic morphology suggest a simple heterochronic mechanism for the production of all major differences, hence in­creasing our confidence in (although clearly not proving) a hypothesis of di­rect evolutionary filiation. Finally, the fact that no morphological differentias of the species undergo any phyletic transformation within the lifetime of the putative ancestor further underscores the punctuational character of the tran­sition, whatever the mode followed. The one character that does change dur­ing the tenure of D. subucula (perhaps only ecophenotypically, as discussed above) does not move towards the morphology of descendant D. favrina. The authors conclude (1985, pp. 36-37):

  Clearly the sedimentary record is complete enough and represents a sufficiently long period of time to be able to detect phyletic gradualism. Yet throughout this period D. subucula remains otherwise morphologically static. Characters that have been modified in closely related species show no evidence of undergoing gradual transformation within the duration of the species ... The overlapping ranges of the two species and the total absence of phyletic gradualism in the characters that serve to distinguish the species suggests that punctuated equilibrium is a better model for speciation in this particular case.

  In a later section (pp. 854–874), I shall discuss the generality of stasis within taxa or times under the more appropriate heading of empirical work on relative frequencies. But I shall also note this broader argument here, and in passing, if only to underscore the strong psychological bias that still per­vades the field, thereby conveying a widespread impression that gradualism maintains a roughly equal relative frequency with punctuated equilibrium, whereas I would argue that, in most faunas,
only a small minority of cases (surely a good deal less than 10 percent in my judgment) show evidence of gradualism. Under this largely unconscious bias, most researchers still single out rare cases of apparent gradualism for explicit study, while bypassing ap­parently static lineages as less interesting.

  Johnson (1985), for example, studied 34 European Jurassic scallop species, and concluded (p. 91): “One case . . . was discovered where . . . the sudden appearance of a descendant form could fairly be ascribed to rapid evolution (within no more than one million years). Inconclusive evidence of gradual change over some 25 million years was discovered in one of the other lineages studied . . . but in the remaining 32 lineages morphology appears often to have been static.” Yet Johnson virtually confines his biometrical study to the two cases of putative change, presenting only a single figure for just one of the [Page 827] 32 species in stasis. Johnson's title for his excellent article also records this bias in degrees of relative interest — for he sets the unmentioned but over­whelmingly predominant theme of stasis in opposition to his label for the en­tire work: “The rate of evolutionary change in European Jurassic scallops.”

  The most brilliantly persuasive, and most meticulously documented, exam­ple ever presented for predominant (in this case, exclusive) punctuated equi­librium in a full lineage — Cheetham's work on the bryozoan Metrarabdotos, more fully treated on pp. 843–845 and 868–870 — began as an attempt to il­lustrate apparent gradualism. Cheetham wrote (in litt. to Ken McKinney, and quoted with permission from both colleagues): “The chronocline I thought was represented ... is perhaps the most conspicuous casualty of the restudy, which shows that the supposed cline members largely overlap each other in time. Eldredge and Gould were certainly right about the danger of stringing a series of chronologically isolated populations together with a gradualist's ex­pectations.” Cheetham's biometry led him to the opposite conclusion of ex­clusive stasis: “In nine comparisons of ancestor-descendant species pairs, all show within-species rates of morphological change that do not vary sig­nificantly from zero, hence accounting for none of the across-species differ­ence” (Cheetham, 1986, p. 190).

  The establishment of stasis as an operational and quantifiable subject behooves us to develop methods and standards of depiction and characteriza­tion. Several studies have simply presented mean values for single characters in a vertical succession, but such minimalism scarcely seems adequate. At the very least, variances should be calculated (and included in published dia­grams in the form of error bars, histograms, etc.) — if only to permit statistical assessment of significances for mean differences between levels, and for corre­lations of mean values with time.

  Smith and Paul (1985), for example, presented both ontogenetic regres­sions and histograms for samples from each meter of sediment to illustrate stasis in relative size of the peristome in Discoides subucula (Fig. 9-14). Cronin (1985) also used both central tendency and variation to illustrate sta­sis throughout 200,000 years of intense climatic fluctuation (during Pleisto­cene ice cycles) for the ostracode Furiana mesacostalis. Cronin (Fig. 9-15) en­circled all specimens of the species at three expanding levels of time in a multivariate plot of the first two canonical axes (encompassing 92 percent of total information): (1) variation in a single sample spanning 100 to 1000 years; (2) in one formation encompassing 20,000 to 50,000 years; and (3) across two formations, representing 100,000 to 200,000 years. Two features of this pattern provide insight into the anatomy of stasis: first, relatively small increase in the full range of variation over such marked extensions in lengths of time; second, the concentric nature of the enlarging ellipses, indicating no preferred direction in added variation, but merely the regular expansion an­ticipated in any random system with increasing sample size. As Cronin notes, this lack of directionality seems all the more surprising when we recognize that this lineage persisted in stasis through several ice-age cycles. Stasis must [Page 828] be construed as a genuine phenomenon, actively maintained — and not as an absence of anything. Cronin writes (1985, pp. 60-61): “Total within-sample variability representing 102 to 103 years is only slightly less than variability over 105 to 2 × 105 years. Puriana mesacostalis shows no secular trends in its morphology over this time interval that might be evident from a lack of concentricity of the ovals — stasis is directionless. Yet high-amplitude environmental fluctuations occurred during this time that could have catalyzed speciation or caused extinction.”

  Once we construe stasis as an interesting evolutionary phenomenon, ac­tively promoted within species, we then become eager to know more about its fine-scale anatomy and potential causes. A remarkable series of studies by Michael A. Bell on the Miocene stickleback fish Gasterosteus doryssus (Bell and Haglund, 1982; Bell, Baumgartner and Olson, 1985; Bell and Legendre, 1987) provide evidence at a maximal level of paleontological resolution, for these fossils occur in abundance in varved sediments with yearly bands — surely a summum bonum for attainable temporal precision! Bell and colleagues

  9-14. An impressive demonstration of stasis in the peristome size of the echinoid Discoides subucula. From Smith and Paul (1985). All specimens are shown for each narrow collecting interval, spaced one meter apart.

  [Page 829]

  9-15. Two hundred thousand years of stasis, during intense climatic fluctuation of Pleistocene ice cycles, in the ostracod Puriana mesacostalis. From Cronin (1985). The three circles around the specimens of this species show increased variation with expanding amounts of time: a single sample representing 100 to 1000 years, one formation encompassing 20,000 to 50,000 years, and two formations representing 100,000 to 200,000 years. The range of variation ex­pands but the modal values do not change at all — as a hypothesis of stasis would predict.

  have documented extensive and complex temporal variability, both for single characters and correlated complexes, over tens of thousands of years (the 1985 study, for example, included several sampling pits covering about 1/3 of the total sedimentary record in a full sequence of approximately 110,000 years). They found some gradual trends in parts of the sequence, a great deal of fluctuation, and a few levels of abrupt alteration, often com­pletely reversing a gradual change built through most of preceding time. In sum, this extensive and multifarious variation includes no sustained or ac­complished directionality, and means for most single characters end up about where they began, whatever the internal wanderings between endpoints. Bell et al. (1985, p. 264) conclude: “Despite the temporal trends and heterogene­ity of all characters through time, the end members [oldest and youngest sam­ple] of only two of the time series (i.e. dorsal spine and dorsal fin ray numbers) [Page 830] are significantly different from each other; most characters return to their original states.”

  This complex anatomy of stasis again illustrates an active process of main­tenance. Perhaps Futuyma's insight (see pp. 796–802) about linkage between speciation and achieved, stable morphological change can also help in this case. I suspect that much of the fluctuation, especially the occasional abrupt changes, represents a complex mosaic of shifting geographic borders for tran­sient local populations. Vertical sequences probably record a mixture of tem­porary change within local populations and successive migrations of distinct local populations in and out over the same geographic spot (sample pit in this case). But both of these sources — short-term changes within a population and the mosaic of differences between demes — will be transient and fluctuating unless a set of differentia can be “locked up” by reproductive isolation within a newly formed species. Since Bell's sequence includes no events of speciation, sustained changes do not accrue. The unusual extent of directionless fluctua­tion then records the especially high degree of temporal resolution.

  Bell and colleagues (1985, p. 258) conclude correctly “the irregular patterns and great magnitude of phenotypic changes that are observed indi­cate that conventional paleontological samples may miss important evolu­tionary phenomena and are not comparable to shorter-term evolution in ex­tant popu
lations.” Fair enough; I asserted the same argument earlier (see p. 801) by metaphorical comparison to our current cliché for illustrating dif­ferent scales of fractal self-similarity: one cannot measure around every head­land of every sea-cove (transient changes over years in local populations) when calculating the coastline of Maine (macroevolutionary trends over mil­lions of years) at the scale of a single page in an atlas. But one must firmly re­ject the tempting implication that either scale can be judged “better” or more complete. Yes, the paleontological scale misses “important evolutionary phe­nomena” of transient fluctuation in local populations. But measurement of details at this local scale cannot be extrapolated to encompass or explain a macroevolutionary trend either — and such local details therefore miss “im­portant evolutionary phenomena” as well. Rather than accusing any level of insufficiency for its inevitable inability to resolve events at other unrecorded scales, we should simply acknowledge that any full understanding of evolu­tion requires direct study and integration of the fascinating uniquenesses (as well as the common features) of all hierarchical levels in time and structure.

  I have emphasized that one cannot achieve a proper “feel” for the relative frequency of punctuated equilibrium merely by tabulating published cases for individual lineages (whereas the study of relative frequency in a well-bounded, taxon, time, or environment — provided that researchers do not preselect their circumstances based on well-documented subjective appear­ance in favor of one side or the other — has yielded valuable data, as discussed on pp. 854–874). Claims for gradualism attain their highest frequencies at “opposite” ends of the conventional chain of being — that is, for foraminifers and for mammals. In my partisan way, I suspect that the former case may be valid, but attributable to biological differences that predict gradualism within [Page 831] asexual protist “species” as the expected consequence of punctuational clone selection, and therefore a proper analog of gradual trends in metazoan lin­eages that arise from cumulated punctuational speciation (see pp. 807–810). For mammals, I suspect that published reports follow traditions of work and expectation more closely than they record actual relative frequencies of na­ture.