sign in sign up
<<
Home , Balnibarbi (trilobite), Deme (biology), Niles Eldredge, Stephen Jay Gould

Punctuated Equilibria: The Tempo and Mode of Reconsidered

Stephen Jay Gould;

Paleobiology, Vol. 3, No. 2. (Spring, 1977), pp. 115-151.

Stable URL: http://links.jstor.org/sici?sici=0094-8373%28197721%293%3A2%3C115%3APETTAM%3E2.0.CO%3B2-H

Paleobiology is currently published by Paleontological Society.

Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at http://www.jstor.org/about/terms.html. JSTOR's Terms and Conditions of Use provides, in part, that unless you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you may use content in the JSTOR archive only for your personal, non-commercial use.

Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at http://www.jstor.org/journals/paleo.html.

Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed page of such transmission.

The JSTOR Archive is a trusted digital repository providing for long-term preservation and access to leading academic journals and scholarly literature from around the world. The Archive is supported by libraries, scholarly societies, publishers, and foundations. It is an initiative of JSTOR, a not-for-profit organization with a mission to help the scholarly community take advantage of advances in technology. For more information regarding JSTOR, please contact [email protected].

http://www.jstor.org Sun Aug 19 19:30:53 2007 Paleobiology. 1977. vol. 3, pp. 115-151.

Punctuated equilibria: the tempo and mode of evolution reconsidered

Stephen Jay Gould and Niles Eldredge

Abstract.-We believe that punctuational change dominates the : evolution is concentrated in very rapid events of (geologically instantaneous, even if tolerably continuous in ecological time). Most , during their geological history, either do not change in any appreciable way, or else they fluctuate mildly in morphology, with no apparent direction. Phyletic is very rare and too slow, in any case, to produce the major events of evolution. Evolutionary trends are not the product of slow, directional transforma- tion within lineages; they represent the differential success of certain species within a - speciation may be random with respect to the direction of a trend (Wright's rule). As an a priori bias, has precluded any fair assessment of evolutionary tempos and modes. It could not be refuted by empirical catalogues constructed in its light because it excluded contrary information as the artificial result of an imperfect record. With the model of punctuated equilibria, an unbiased distribution of evolutionary tempos can be established by treating stasis as data and by recording the pattern of change for all species in an assemblage. This distribution of tempos can lead to strong inferences about modes. If, as we predict, the punctuational tempo is prevalent, then speciation-not phyletic evolution-must be the dominant mode of evolution. We argue that virtually none of the examples brought forward to refute our model can stand as support for phyletic gradualism; many are so weak and ambiguous that they only reflect the persistent bias for gradualism still deeply embedded in paleontological thought. Of the few stronger cases, we concentrate on Gingerich's data for Hyopsodus and argue that it provides an excellent example of species selection under our model. We then review the data of several studies that have supported our model since we published it five years ago. The record of human evolution seems to provide a particularly good example: no gradualism has been detected within any hominid , and many are long-rangirig; the trend to larger brains arises from differential success of essentially static taxa. The data of molecular genetics support our assumption that large genetic changes often accompany the process of speciation. Phyletic gradualism was an a priori assertion from the start-it was never "seen" in the rocks; it expressed the cultural and political biases of 19th century liberalism. Huxley advised Darwin to eschew it as an "unnecessary difficulty." We think that it has now become an empirical fallacy. A punctuational view of change may have wide validity at all levels of evolutionary processes. At the very least, it deserves consideration as an alternate way of interpreting the history of life.

Stephen Jay Gould. Museum of Comnparative Zoology, , Cambridge, Mass. 02138 Niles Eldredge. Department of Fossil Inuertebrates. American Museum of Natz~ralHistory, Central Park West at 79th Street, , N.Y. 10024 Accepted: January 5, 1977

You have loaded yourself with an un- I. Gradualism and Stasis necessary difficulty in adopting Natura In 1944, G. G. Simpson published a book non facit saltum so unreservedly. that brought within the modern Huxley to Darwin, Nov. 23,1859, the synthesis of evolutionary theory. He used his duy before publication of the Origin. title to identify the principaI topics that pale- I see you are inclined to advocate the pos- ontology might pursue to enlighten evolution- sibility of considerable 'saltus' on the part ary theory-tempo and mode. But tempo and of Dame Nature in her variations. I altoays mode do not share an equivalent status as took the same view, much to Mr. Darwin's subjects for study in the fossil record. Tempos disgust. can be observed and measured: modes must Huxley to Bateson, Feb. 20, 1894 be inferred, usually from empirical distribu-

Copyr. @ 1977 The Paleontological Society All rights reserved US ISSN 0094-8373 116 COULD & ELDREDGE

tions of tempos. Such inferences, if they are (Eldredge and Gould 1972). Two other to be made properly, require a random sam- classes of information were explained away ple of tempos-or at least a sample not hope- or simply ignored: 1) morphological gaps in lessly biased by a priori assumptions about stratigraphic sequences-which might have evolutionary rates. Paleontologists have never suggested a punctuational view of evolution- been able even to approach such a random were attributed to imperfections of the fossil sample. Our model of punctuated equilibria record; 2) evolutionary stasis, though recog- is a hypothesis about mode. We claim that nized by all and used by stratigraphers in the speciation is orders of magnitude more im- practical work of our profession, was ignored portant than phyletic evolution as a mode of by evolutionists as "no data." Thus, Trueman evolutionary change. An unbiased distribu- rejoiced in Gryphaea (1922) but never men- tion of tempos must be achieved in order to tioned the hundreds of Liassic species that test this hypothesis rigorously. show no temporal change. Rowe (1899) We may illustrate this dilemma with an monographed Micraster but spoke not a word analogy from genetics. Population geneticists about its legion of static colleagues in the recognized from the outset that a primary English chalk. In fact, the situation in pale- datum of their profession would be a measure ontology is far worse than that confronting of the amount of genetic variability in natural genetics a decade ago. At least the geneticists populations. This basic issue could not be re- were frustrated by an absent technology: they solved simply because no one knew how to knew what data they needed. Paleontologists take a random sample of in order to es- allowed a potent, historical bias to direct their tablish the relative frequency of variable vs. inquiry along a single path, though they could fixed loci. In the absence of direct evidence, have accumulated other data at any time. two opposing schools (advocates of very What's more, paleontologists accumulated limited vs, copious variation) argued for half hardly any good examples: the gradualistic a century, and their debate set the tone and idols that were established had feet of clay concern of an entire profession (Lewontin and rarely survived an intensive restudy. The 1974). Impressive catalogues of variable genes tale of Gryphaea is dead in Trueman's for- had been compiled even before the debate mulation (Hallam 1968; Gould 1972). Mi- began: geneticists suffered no dearth of evi- craster will soon follow. (Rowe's data identi- dence for multiple alleles at loci. The problem fied three successive species, but he had no lay only in the nature of sampling: ge- stratigraphic control for samples within taxa. netic variation had to exist before a trait Even if his gradualistic tale were true-which could be located-invariant loci could not be it is not-his own limited data could not have identified; hence, no random sample could established it.) The collapse of classic after be drawn. With electrophoretic techniques classic should have brought these gradualistic applied during the last decade (Lewontin biases into question. The alienation of practi- and Hubby 1966 et seq. through hundreds cal stratigraphy from an evolutionary science of papers), loci can be identified without prior that required gradualism should have sug- knowledge about their variability. Random gested trouble (see Eldredge and Gould, in samples were established and the central di- press) : always trust the practitioners. lemma of was resolved: This sorry situation led us to postulate our variation is copious. alternative model of punctuated equilibria We believe that paleontology has labored (Eldredge 1971; Eldredge and Gould 1972). under a strikingly similar dilemma. The em- We wanted to expand the scope of relevant pirical distribution of evolutionary tempos is data by arguing that morphological breaks in as fundamental a datum to our profession as the stratigraphic record may be real, and that amounts of variability are to geneticists. Yet, stasis is data-that each case of stasis has as just as geneticists could only identify variable much meaning for evolutionary theory as each traits, paleontologists have worn blinders that example of change. We did this by recogniz- permit them to accumulate cases in one cate- ing that the model of speciation preferred by gory only: they have sought evidence of slow, most evolutionary biologists did not yield a steady and gradual change as the only true prediction of gradual change in large popu- representation of evolution in the fossil record lations. Most evolutionary change, we argued, PUNCTUATED EQUILIBRIA RECONSIDERED 117 is concentrated in rapid (often geologically gested two more radical changes in theory instantaneous) events of speciation in small, pe- and practice : ripherally isolated populations (the theory of 1) We realized that the extrapolation of ). (Our model of punc- punctuated equilibria to macroevolution sug- tuated equilibria works equally well for sym- gested a new explanation for the fundamental patric speciation when two conditions are met phenomenon of evolutionary trends (Eldredge [see Bush et al, in press, for their defense]: and Gould 1972, pp. 111-112 and fig. 10, p. daughter species must arise from a small sub- 113; Stanley 1975a). group of the parental population, and they 2) We knew that we were proposing a re- must do so in no more time than it takes for orientation of empirical work in evolutionary a peripheral isolate to speciate in the allopatric paleontology-away from the search for grad- model.) The norm for a species during the ualism in selected species within local sections, heyday of its existence as a large population towards the quantitative study of evolutionary is morphological stasis, minor non-directional pattern in all members of a fauna. fluctuation in form, or minor directional An a priori bias toward gradualism as the change bearing no relationship to pathways only "true" evolutionary event continues to of alteration in subsequent daughter species. preclude any fair test for relative frequency In local stratigraphic sections, we expect no among the possible tempos of evolution. At slow and steadv transition. but a break with worst, it dictates an erroneous interpretation essentially sudden replacement of ancestors of major evolutionary events. Thus, Durham by descendants: this break may record the tried to estimate the age of conunon ancestry or emigration of a parental species for deuterostomes by stacking species end to and the immigration of a successful descen- end in lineages of phyletic gradualism. He dant rapidly evolved elsewhere in a small, specifies 6 m.y. as an average "species dura- peripherally isolated population. ( Small num- tion" and estimates 100-600 durations strung bers and rapid evolution virtually preclude on a line to reach the common ancestor of the preservation of speciation events in the Early and Middle echinoderms. fossil record; in any case, speciation does not Running further down the string, he places occur in local sections inhabited by abundant the common ancestor of deuterostomes ancestors.) "slightly over a billion years before the be- For all the hubbub it engendered, the model ginning of the Cambrian7' (1969, p. 1128)- of punctuated equilibria is scarcely a revolu- an age considerably earlier than the most gen- tionary proposal. As Simpson (1976, p. S), erous estimate now being offered for the with his unfailing insight, recognized in three origin of the eukaryotic cell (Schopf and lines (where others have misunderstood in Oehler 1976)! Yet Durham states that his entire papers), our model tries to "clarify and estimate can be too great only if he over- emphasize ideas nascent in previous studies estimated the number or length of species of the synthetic theory." We merely urged durations-it never occurs to him that the our colleagues to consider seriously the impli- postulate of gradualism and linear stacking cations for the fossil record of a theory of spe- might be fallacious. In fact, he argues that a ciation upheld by nearly all of us," and to belief in evolution itself requires his mode of recognize the search for phyletic gradualism estimate: "Acceptance of the doctrine of evo- as a bad historical habit not consistent with lution and adherence to a biologic species con- modern evolutionary ideas. Nonetheless, we cept and their corollaries necessitates elapse did recognize that our modest proposal sug- of a considerable interval of time prior to the Cambrian in order for the necessary evolu- * Since the arguments of this paper rely upon some basic tionary events to have taken place." tenets of the 6'biological" species concept, we wish to em- At an intermediate level, gradualistic biases phasize explicitly that we treat species as "real" units in nature, not as arbitrary elements in a continuum ranging from inspire misleading tales of continuity in more individual to kingdom. This "reality" may arise from the con- ventional dynamic argument of direct interaction by restricted, empirical studies. Hurst ( 1975), flow (with reproductive isolation from other species); it may also reflect the historical argument that unique, highly' for example, presents two outstanding figures homeostatic systems arise during the genetic reorganization that accompanies speciation (thus preserving the basic char- of gradualism in the Resserella sabrinae lin- acter of a species even in the absence of among its demes). Species, Mayr writes (1963, p. 621), are "the eage (Silurian from Wales and the real units of evolution, as the tem orary incarnation of har- monious, well-integrated gene compfexes." Welsh Borderland). These figures (Hurst 118 GOULD & ELDREDGE

1975, pp. 320-321) display more than 10 sam- Ozawa (1975) measures nine characters in his ples arranged, so the caption declares, "in exemplary account of the Permian foraminifer chronological order." Hurst depicts only one Lepiclolina multiseptata. Four exhibit gradual reversal in an otherwise continuous trend for trends; five do not. The first four are either both characters-impressive examples of grad- graphed or listed in tables of mean values; the ualism. Or so it would seem, until we read remaining five are discussed in qualitative further in the caption and discover that terms only. "chronological order" is established by "pale- But stasis will not go away; and the punc- ontological and stratigraphic evidence." Some tuations that mark the fossil record do not samples are ordered by superposition, and smooth out as stratigraphic resolution im- these are indicated on the figures (the one proves. Even Darwin had to retreat from the morphological reversal occurs between two lovely tree of pure gradualism that he sketched superposed samples). We then discover that in the first edition of the Origin. For he added gradualism itself is the criterion for "paleon- to the fourth edition the following proviso: tological" ordering-hence no reversals, by "it is far more probable that each form re- definition. Moreover, in no fewer than four mains for long periods unaltered, and then cases (and for four independent pairs of sam- again undergoes modification" (quoted in ples), the "chronological order" of the first Gingerich, 1976). figure is exactly opposite to that of the second figure. One may believe in gradualism with 11. What Eldredge and Gould all one's soul, but how can temporal order run in different directions for different char- Did Not (And Did) Say acters? (We thank G. Klapper for pointing Our model of punctuated equilibria has been this out to us; we missed it the first time widely discussed and applied; we also detect around. ) a good deal of disagreement about what we At best, a gradualistic bias leads to no ob- actually said. We write this paper as a col- vious error but only to the neglect of im- lective response and amplification. It is not portant evidence for stasis. English-speaking a comprehensive review of the debate between paleontologists have rarely approached the gratlualism and punctuation-for this would zeal of some continental workers who regard compel us to chronicle the entire history of deflections from linear trends as "irregularities evolutionary paleontology. We confine our in the course of development" ("Unregelmas- comments strictly to those who have tried sigkeiten im Entwicklungsablauf," Betten- explicitly to test our model and to some staedt 1962, p. 407). But they have bypassed earlier papers cited by them, sometimes to overwhelming evidence for stasis to concen- support us, but usually to refute us. trate on rare cases of gradualism. As Scott Beyond the major substantive criticism that (1974, p. 136) admits for Foraminifera: "Be- gradualism remains a more adequate model cause of their stratigraphical value, unidirec- for assessing the history of life (see subse- tional trends in shell morphology have at- quent sections), the two most frequent criti- tracted most comment, although it is not cisms of our paper (Eldredge and Gould established that they are even a principal fea- 1972) are methodological. In both cases, we ture of foraminifera1 evolution." feel that we have been misunderstood: The bias that regards stasis as "no data" 1. Some critics (e.g Harper 1975) have seen persists even within the few satisfactory ex- our work as restrictive in scope-as an attempt amples of gradualism known to us. One might rigidly to exclude gradualism by establish- have expected that such studies would value ing a new dogma for evolutionary tempos. all characters equally by attempting to de- Lespbrance and Bertrand (1976, p. 610) limit the scope of gradualism among traits charge that we have, "in effect, denied the within a chosen taxon. Ziegler's celebrated existence of phyletic gradualism in specia- study of the Silurian Eocoelia tion." We have never understood punctuated (1966), for example, presents data for five equilibria in this light. We see it as funda- graphable characters. Three display a fluctu- mentally expansive-as a more adequate pic- ating pattern, one no trend at all, and the last ture that should extend the range of paleon- a directional pattern. Only the last is graphed. tological activity by valuing types of data PUNCTUATED EQUILIBRIA RECONSIDERED 119 previously neglected. We never claimed either Using the gradualistic model to the full, that gradualism could not occur in theory, it would be possible, given the enormity or did not occur in fact (Eldredge 1971; of geological time, to derive practically Eldredge and Gould 1974, p. 307). Nature any modern species from practically any is far too varied and complex for such ab- fossil one. Using the punctuated equilib- solutes; Captain Corcoran's "hardly ever" is rium model we should be able to observe the strongest statement that a natural historian the stasis in each recorded species and de- can hope to make. Issues like this are decided velop our phylogenetic schemes accord- by relative frequency. The of axolotl ingly (Walker 1975, p. 7). was proven experimentally in the 1860's, but no one viewed the case as a serious challenge 2. In a misinterpretation inspired largely to Haeckel's recapitulation. A few examples by our own ambiguity, many critics have at- of do not refute tributed to us a notion of testability that we as the directing force of evolution. The funda- wish decisively to reject. Some have claimed mental question is not "whether at all" but that we abjure testability altogether in assert- "how often." The potential neutrality of many ing the necessary truth of punctuated equi- mutations, arising from the redundancy of the libria. Scott ( 1976), for example, charges that genetic code, is another matter. Here we we dismiss gradualism "without trial of the have a challenge based on relative frequency stratigraphic evidence." We made no such -a claim that most genetic substitutions may claim; we merely supported the general state- not be guided by selection. ment about science-defended by all its astute Our unhappiness with gradualism arose historians and philosophers (Kuhn 1962; Han- from its status as restrictive dogma. For it has son 1969, for example)-that observation can- the unhappy property of excluding a priori not be "objective," but must be made under the very data that might refute it. Stasis is the aegis of some theory. This is no denial of ignored as "no data," while breaks are treated testability; it only asserts that data to refute as imperfect data. Relative frequency cannot a theory must usually be gathered in the light be assessed because contrary data are not pre- of an alternate theory. We proposed punc- sented. All facts are gathered in the light of tuated equilibria in order to permit the test, some theory. The explicit formulation of punc- not to avoid it! The empirics of the case tuated equilibria should lead to the casting of should prove our adherence to the cardinal a wider net for data to test the relative fre- principle of testability. If we thought that no quency of evolutionary tempos; for we know stratigraphic evidence were needed, we \vould no other way to make reasonable inferences not have presented quite a bit of it ourselves about evolutionary modes-specifically, in (Eldredge and Gould 1972, pp. 98-108). If this case, the relative importance of specia- most paleontologists viewed our model as un- tion vs. phyletic evolution. Of course, we do testable, we would not be writing this paper not champion punctuated equilibria as liberal -for there \vould be no extensive literature on explicit, putative tests to inspire this com- pluralists with no suspicion about the final mentary (e.g. Johnson 1975, p. 648 on the outcome. We do regard punctuated equilib- deduction of three predictions from our model rium as by far the most common tempo of evo- and their test in the stratigraphic record). lution-and we do assert that gradualism is Others have recognized our adherence to both rare and unable in any case-given its the principle of testability, but have inter- characteristic rate-to serve as the source for preted us as stating that fossil evidence can major evolutionary events (pp. 133-134). decide nothing of importance in evolutionary We are gratified that many paleontologists theory-that all decisions must be made by have used our model to expand a range of evolutionary theorists working with living or- concepts and data. Sprinkle (1976), for ex- ganisms (Roughgarden 1973, p. 225; Kitts ample, realizes that it comfortably permits the 1974, p. 471; Stanley 1975, p. 646). The of- designation of small, but very distinct Paleo- fending statement in our paper follo\vs; if time zoic echinoderm as classes. Walker could move backward and if Omar Khayyam (1975) and Pilbeam (1975) have supported had not written so truly about the moving the utility of stasis in tracing phylogeny: finger, we would value the opportunity to re- 120 GOULD & ELDHEDGE

write it: "We can apply and test, but we can- a sad commentary indeed on the status of not generate new mechanisms. If discrepan- paleontology if so basic an item as the prob- cies are found between paleontological data ability distribution of evolutionary tempos and the expected patterns, we may be able could not be reasonably inferred from the to identify those aspects of a general theory fossil record, and once we learn something that need improvement. But we cannot fomu- about the distribution of tempos, we will be late these improvements ourselves" (1972, pp. able to make strong inferences about modes. 93-94). Our critics have taken this as a gen- As Sylvester-Bradley writes (in press, p. 63) : eral statement about paleontology. Sylvester- "If the punctuation is real, so is the grammar Bradley (in press, p. 42), though he likes our and syntax." model, charges us with "a defeatist attitude, Two general strategies are available: 1) which suggests that paleontology can never Build towards a distribution inductively by decide which theory is wrong." Roughgarden examining individual cases with optimal fea- writes (1973, p. 225): "On its face the com- tures: For satisfactory tests, we must work mitment is false, for it is possible that paleon- with species-level lineages well preserved over tologists could be the first to discover causal the full span of an extensive geographic and mechanisms with long time constants." Kitts temporal range (see Eldredge 1974, p. 479). replies (1974, p. 471): "Why can't paleonto- (Imperfect cases also provide useful data: we logists formulate theoretical improvements would love to know whether most species themselves? . . . Paleontological events may sporadically preserved in local sections dis- legitimately be used to test the long range his- play stasis [as we suspect], significant fluc- torical consequences of evolutionary theories." tuating change, or directional evolution.) We agree with Kitts and Roughgarden. We, Most commentary on our model has involved with our primary interests in evolutionary its application to individual cases; we will theory, would never have become paleonto- record our opinion of these works in the fol- logists had we regarded the data of our field lowing subsections. 2) Devise more general as fundamentally unavailable for evolutionary tests based on quantifiable features of entire insight. Of course the data of paleontology clades or communities: In supporting our will inspire macroevolutioilary theory and model, Stanley (1975) has recently proposed play a major role in testing it. Where else four such tests. His last "test of generation would we go? time" rejects the gradualistic inference that Our statement, as cited above out of con- a correlation should exist between length of text, seems to be a general indictment of pale- generations and macroevolutionary rate. If ontological potential. In the context of its frequency of speciation controls the rate of paragraph, however, we can only read it as we evolution, then no such correlation should be intended it-as a limited statement treating expected. We do not regard this test as con- one issue only: microevolutionary theories clusive because gradualists can claim that a about the mechanisms of speciation. We will number of effects-variation in the intensity stick by this limited intent: alone will of natural selection in particular-over.cvheIm neither decide the issue of how speciation oc- the real correlation between generation and curs, nor will they, by themselves, supply the rate, and swamp it out completely in actual information needed to infer new mechanisms. data. We could not be more optimistic about Stanley's first three tests (adaptive radia- paleontology and its potential, yet unrealized, tion, Pontian cockles and living fossils) ad- role in evolutionary theory. In exhorting pale- dress the same issue: major morphological ontologists to leave microevolutionary theory evolution must occur by repeated, rapid speci- largely to neontologists, we merely tried to ation since too little time is available for assert the inviolability of our own, extensive change by standard, gradualistic rates. (Liv- turf-time. ing fossils represent the flip side of argu- ments from adaptive radiation-morphological change is slow because lingulae and their 111. Testing Puilctuated Equilibria allies have never been diverse.) We were first A) Introduction.-The model of punctuated inclined to judge these tests as inconclusively equilibria is eminently testable. It would be directed against a straw man of extreme grad- PUNCTUATED EQUILIBRIA RECONSIDERED 121 ualism-surely, most gradualists would not Hecht writes (1974, p. 302): "The morpho- trv to build an ada~tiveradiation without an logical trends within U. maritimus affected uiusually high rateLof speciation; they would the entire population. . . . The changes within insist only upon the V-shaped pattern of con- the polar bear are classical phyletic trans- tinual divergenceu between ~arental and formational changes." But we could scarcely daughter species. We are no longer so cer- ask for a better case of rapid, allopatric speci- tain that Stanley's gradualist is an extremist. ation. The "entire population" is a small, pe- Durham (1971), as cited on p. 117, depicted ripheral isolate (Hecht estimates the breed- a monumental adaptive radiation by end-to- ing population of females at never more than end stacking of species, without any increase 6,000), and it speciates rapidly. in rate of speciation or phyletic evolution. Hayami and Ozawa (1975) present a similar And Gingerich (in press) has denied any in- case, not as a counterexample, but as com- crease in rate of meciation for the ugreat plementary to our larger aim of legitimizing Paleocene-Eocene mammalian transition that punctuational events in Darwinian terms. They yielded rodents, of modem aspect, trace the introduction and gradual increase, bats, primitive Carnivora, artiodactyls and beginning in the middle Pleistocene, of a dis- perissodactyls. He postulates a late Paleocene crete phenotype within populations of the climatic deterioration. drivingu the ancestors scallop Cryptopecten vesiczcloszcs. The pattern of these groups into limited Central American of increase in frequency is gradual, but it does spaces where reduced populations subject to not confute our model for several : severe competition evolved under such intense i) The change was rapid compared to the selection that phyletic change alone yielded duration of most species; in Recent popula- the required transitions. tions, the new phenotype seems to have stabi- B) Invalid claims of gradzcalism made at the lized at a frequency near 40 percent. torong scale.-The model of punctuated equi- ii) The change has nothing to do with speci- libria does not maintain that nothing occurs ation: it is a minor event (an allelic substitu- gradually at any level of evolution. It is a tion in Hayami and Ozawa's view), occurring theory about speciation and its deployment in in thousands of years, within a lineage des- the fossil record. It claims that an important tined-like all lineages-for extinction unless pattern, continuous at higher levels-the it produces daughter species. "classic" macroevolutionary trend-is a con- iii) The geological pattern of change is not sequence of punctuation in the evolution of morphological intermediacy, but the sudden species. It does not deny that allopatric speci- introduction and subsequent increase of a dis- ation occurs gradually in ecological time crete phenotype. Traditional paleontologists ( though it might not-see Carson, 1975), but would recognize two species with overlapping only asserts that this scale is a geological range zones. Thus, as Hayami and Ozawa microsecond. Our model must be tested at ( 1975) argue (see also Hayami 1973), allelic the appropriate scale-by considering tempos substitution can act as a Darwinian mechanism of change in species and in the process of to produce geological patterns of punctuation. speciation during geological time. 2. Scales too macroscopic: Sequential spe- 1. Scales too microscopic: Against our cies in a larger clade often display continual model, Hecht (1974) offers all the evidence of transition for some traits of their mean mor- Darwinian, neontological gradualism: "Phy- phologies. How could it be otherwise? In a letic transformation can be seen in the Dro- sequence of three, taking the initial point as sophila population cage and in the develop- given, unidirectional trends occur with a prob- ment of domestic types of and plants" ability of 1 in 4 for any character. (One in 2 (Hecht 1974, p. 300). Of course-and at rates might be a better figure since we test for uni- that would propel a peripheral isolate to full directionality by asking whether the third spe- speciation in a geological instant. cies continues a tendency set by the first two Hecht then sics the polar bear upon us. as given). For a sequence of four species, the Ursus maritimus evolved from an isolated sub- chance of unidirectionality is 1in 8, or 1 in 4 population of Asiatic brown bears that in- by the second argument. Since most sequences vaded tundra and ice-pack areas in Cromer- are short, and since have an abun- Mindel times. Against the punctuational view, dance of characters, we expect numerous uni- 122 GOULD & ELDREDGE directional "trends" on purely stochastic order (p. 336, the others are drawn as bi- grounds. variate "growth" diagrams); even the most These cases are irrelevant to the issue of ardent gradualist will find this figure, to say punctuated equilibria. Yet the majority of the least, inconclusive. Finally, as a lovely claims for gradualism are based on such dis- illustration of a priori bias, we cite Drooger's continuous, unidirectional series of discrete way of telling us that he has not detected steps. These sequences test nothing, for the gradualism (1963, p. 319): "It might be pos- crucial data lie unreported in the blank spaces sible that the trend towards reduction of the between successive species (Eldredge 1971). total number of spiral chambers in this oldest Our model predicts that these trends reflect group was less rigorous." the differential success of phenotypic subsets Johmon, though a strong supporter of our within an essentially random pool of rapid model ( 1975), cites among rare counter-cases (see Eldredge and Gould 1972, two of his own works (Johnson et al. 1969; pp. 111-113; Stanley 1975; and part IV of Johnson and Norris 1972). But both involve this work). Gradualism, on the other hand, only a shift in mean morphology among three requires insensibly even transition both within sequential species of Devonian brachiopods; and between taxa in the sequence. The re- neither includes any evidence for bed-by-bed markable fact that such blatantly inadequate gradualism within a species. Tecnocyrtina data have been so widely accepted as con- missowiensis, for example, split from Cyrtina vincing proof of gradualism only reinforces in the late Middle Devonian, presumably by our claim that gradualism has always rested allopatric speciation (Johnson and Norris on prior prejudice rather than paleontological 1972). In so doing, it developed plications on data. its fold and sulcus. The claim for gradualism The Micraster sequence of the English rests only upon the observation that a sub- Chalk has resisted the fate of other famous sequent, non-overlapping species, Tecnocyr- gradualistic tales. Rowe's story (1899) re- tina billingsi, evolved even more plications. mains unchallenged as the exemplar of grad- Given a gradualistic assumption that morphol- ualism in the English literature. Yet an ex- ogies will alter from bed to bed, we remind amination of Rowe's data reveals that it tests readers that it had a 50-50 chance of so do- nothing of the sort. Rowe had no stratigraphic ing in the simplest, coin-flip model of random resolution within these species of heart ur- processes. chins; his celebrated story is a sequence of C) Invalid claims of gradualism based on in- tluee discrete points-named, with a good adequate data.-We have been struck by the deal of taxonomic panache, the bull's heart, extreme ambiguity (to our admittedly biased the turtle's heart and the eel's heart (M. cor- eyes) of several cases advanced as contrary to bovis, M. cortestudinarium, and M. corangu- our model. The data are consistent with a inum). Some traits are unidirectional through variety of evolutionary schemes, and exclu- the series of three, and almost all of these re- sively gradualistic interpretations can only flect the same basic feature: increasing com- arise from fervent desire. Makurath and An- plexity of pustulation and surface ornament. derson (1973, see also Eldredge 1974 and Others, equally important, show no trend at Makurath 1974), for example, search valiantly all (the middle species, M. cortestudinurium, for gradualism in three successive samples of is elongate and rectangular, while both its the Devonian brachiopod Gypidula. Their "ancestor" and "descendant" are shorter and oldest sample, G, prognostica from the Keyser heart-shaped) . Formation, is an agglomeration of 52 speci- Many cases, cited explicitly against our mens from Hyndman, Pa. and 11 from Warm model, are equally irrelevant to its test. Scott Springs, Va. The middle sample contains 79 ( 1976), for example, presents Drooger ( 1963) specimens of G, coeymanensis from the Lower on the evolution of myogypsinid forams as Coeymans Formation near New Salem, New a primary example of gradualism. But York. Ninety specimens, also of G. coey- Drooger's work treats evolutionary trends in manensis, define the upper sample (Upper a large clade of 29 species and cites no evi- Coeymans Formation, Cherry Valley, New dence for bed-by-bed gradualism within taxa. York). Eighty feet of section separate middle Drooger presents only one plot in stratigraphic and upper samples; at least as much (by in- PUNCTUATED EQUILIBRIA RECONSIDERED

ference) separate lower and middle. This is scarcely a bed-by-bed compilation. Makurath and Anderson make only two closely-allied measurements: spondylium width (y) and beak length (x) as "an esti- mator of spondylium length" (1973, p. 305). They begin by computing allometric regres- sions, obtaining slopes of .98 for the lower sample, 1.31 for the middle, and 1.35 for the upper. No gradualism; the samples seem to separate into their two designated species. Then they compute a distance matrix, and cal- culate 48.89 between lower and middle, 35.46 between lower and upper, and 11.48 between middle and upper. Again, no gradualism; in fact, the lower sample is closer to the upper than to the middle! FIGURE1. Three samples of Gypidula plotted on Still searching for gradualism, they perform two canonical axes for two original variables. From Makurath and Anderson 1973. Numbers are centroids a canonical analysis on their two variables (Fig. of the three samples in stratigraphic order; inner 1). Once again, they find no gradualism in circles are 95% confidence intervals; outer circles projections on either axis or in the general represent one standard deviation around the mean distances. Only the nlean shape ratio (Fig. 1) canonical values. Dotted lines are contoured values of the simple shape ratio, y/x, plotted with respect to changes in a unidirectional manner. (One the axes. Although we see gradual change in the scarcely needs the multivariate apparatus of mean shape ratio, we find no gradualism for projec- callollical analysis to find this out; the simple tions of centroids on either axis, or for general dis- calculatioll of y/x would suffice.) They in- tances between centroids. terpret these difference~in shape as "real," even though they are no larger than the dif- ferences in size between middle and upper differences in shape between middle and samples "interpreted as functions of sampling" upper samples may represent two points in (1973, p. 308), presumably because they con- the normal spectrum of geographic variation fute the gradualistic interpretation. Makurath existing at all times in G. coeymanensis, and and Anderson conclude (1973, p. 309): "The utterly unexamined in this study. shape change trend persists between species Klapper and Johnson (1975) present their and between successive populations of the study of the Lower Devonian conodont Poly- later species. Speciation in gypidulid bra- gnathus partly as a test of our model and an chiopods thus provides an example of phyletic affirmation of phyletic gradualism. They rely, gradualism." in small measure, on traditional arguments But what are we to make of such limited for intermediacy in supposed phyletic transi- and ambiguous information? The authors pre- tions, but more importantly on their own cri- sent only three samples with no control of terion for speciation by gradual separation of geographic variation (beyond a dubious ag- sympatric populations. This criterion involves glomeration of distant collections into a single the search for "y-branched" phyletic patterns sample). Given the limited set of alternatives, (1975, p. 66), defined as "lineage splitting it is not particularly impressive that a third characterized by an interval of stratigraphic sample continues a "trend" set by the first overlap of the ancestral and descendant spe- two. The "trend" reflects only one ratio mea- cies, and by accompanying intermediate sure of shape; parameters of ontogenetic forms" (p. 66). We are dubious enough about growth display no gradualism, and general this criterion, since the presence of two spe- distances refute it. If the major force of cies accompanied by intermediates is, among Makurath and Anderson's argument lies in living organisms, more often a sign of hy- their claim that an intraspecific trend (middle bridization than of incipient divergence. More- to upper sample) follows the same direction over, Klapper and Johnson claim to find the set by the species' origin, then we reply that same y-branched event occurring in many far GOULD & ELDREDGE

P. costatus I 7

I I F! laticostatus --

basal -- inverted %avitv

FIGURE2. The phylogeny of Polygnathus as presented by Klapper and Johnson 1975. Numbers 14 indi- cate Klapper and Johnson's assessment of the quality of evidence for their conclusions. We confine our com- ments to all events in their categories 1 and 2-i.e., all events depicted by them with solid lines (not dashes or question marks). flung areas. We know no model of sympatric communication] that he regards only the "y- speciation that postulates such in situ diver- branched patterns, i and iv, as exceptions to gence on a global scale. But we let these our model ) . theoretical objections pass, and analyze only i) the y-branched pattern, as P, gronbergi the fit of Klapper and Johnson's own evidence departs gradually from its ancestor P, de- to their phyletic conclusions. hiscens. Klapper and Johnson epitomize their con- ii) the gradual phyletic transition from P. clusions in an evolutionary tree, unambig- gronbergi to P. laticostatus. uously presented (Fig. 2). These are diagrams iii) the phyletic transition from P. perbonz~s that work their way into textbooks, there to to P. inversus through an intermediate form, convince the uninitiated that paleontologists P, aff. P, perbonus. can specify with assurance the (gradualistic) iv) the y-branched pattern, as P, sp. nov. history of life. But let us examine all of D diverges gradually from P, inversus. Klapper and Johnson's evidence for all four i) Klapper and Johnson cite only three sec- evolutionary events depicted without question tions to support y-branched gradualism (p. in Fig. 2: (Klapper has written [personal 6s). In the first, ancestral P. dehiscens is fol- PUNCTUATED EQUILIBRIA RECONSIDERED 125 lowed by intermediates. In the second, beds dence supposedly arises from three areas so that should contain intermediates bear no widely dispersed that local origin by allo- Polygnathus at all. In the third, both species patric speciation becomes impossible: Nevada, are found together, apparently without inter- Yukon Territory, and Australia. But the Aus- mediates. No other evidence is cited. From tralian section only yields an overlap in range such scrappy data it is hard to see how any- with no intermediates. The Yukon section con- one could derive with confidence the grad- tains both species in with inter- ualistic interpretation of Fig. 2-unless one mediates (no biometrical confirmations are were predisposed to gradualism from the start. presented). These intermediates could be hy- Klapper and Johnson conclude (1975, p. 68): brids or unusual variants of one or the other "This stratigraphic overlap suggests a 'Y- taxon. A third Nevadan section again con- branched' evolutionary pattern. P. gronbergi tains both species in sympatry with no inter- represents a relatively minor, but discernible mediates. In the fourth Nevadan section a morphologic modification from P. dehiscens, large sample of ancestral P. inuersus includes chiefly involving the beginning of inversion at some individuals tending towards P. sp. nou. the posterior end of the platform." We do D; Klapper and Johnson present no evidence not wish to invade the taxonomic prerogatives to test the likely possibility that these variants of specialists in groups foreign to us, but we lie within the normal range of P. inuersus. cannot help wondering whether P. gronbergi Higher in the section, P, sp, nou. D occurs is really a good taxon at all. alone. Again, we do not understand how ii) We quote verbatim the only cited evi- anything other than prior preference could dence for gradual transition between P. gron- dictate the exclusive choice of gradualism as bergi and P. laticostatus (p. 68): "P. lati- an interpretation for such limited and ambig- costatus succeeds P. gronbergi at Lone Moun- uous data. tain. A specimen [our emphasis] occurs in In summary, we cited the evidence of LM 18, which has a lower platform surface Gypidula and Polygnathus in detail not pri- like that of P. laticostatus and an upper sur- marily to reveal the fragility of stories built face like that of P, gronbergi, therefore sug- upon it; for most "phylogenies" based on fos- gesting a transition between the two species." sils rely on flimsy data. Rather, we wish to No further comment. demonstrate that most cases presented as falsi- iii) Some of Klapper and Johnson's argu- fications of punctuated equilibria are circular ments are circular-from gradualistic presup- because they rely, for their gradualistic inter- positions to gradualistic conclusions (based pretations, not upon clear evidence, but upon on the presuppositions, not on fossil evidence). the gradualistic presuppositions they claim to P. perbonus is only known from Australia. test. Yet, Klapper and Johnson choose it rather than D) Potentially valid (but mostly unproued) the local Nevadan P. gronbergi as an an- cases of gradualism.-Among the few poten- cestor for P. inuersus for the following reasons tially valid cases urged against our model, we (p. 68) : "We do not favor an alternate origin find only one (Ozawa's forams) that meets of P, aff. P. perbonus (and consequently P. all criteria for an adequate test-good geo- inversus from P. gronbergi, because at Lone graphic coverage, long sequence of closely Mountain there is no evidence of transitional spaced samples, unambiguous definition of forms between P. gronbergi and P, aff. P, per- taxa, and adequate biometrical testing on suf- bonus. Thus, P. aff. P, perbonus appears ficiently large samples. Two others (Kellogg's abruptly at Lone Mountain." But our model radiolarians and Gingerich's mammals) are of of punctuated equilibria predicts that no particular interest because they include a long transitional forms will be found between an- suite of samples treated biometrically. cestors and descendants in local sections. Our A few additional cases fit a gradualistic in- model cannot be falsified by citing the very terpretation better than any other, but do not evidence it predicts, and then choosing grad- contain enough data to convince. Ziegler's ualistic explanation based on hypothetical an- (1966) celebrated study of the brachiopod cestors half a world away. Eocoelia from Upper Llandoverian strata of iv) Klapper and Johnson champion this the Welsh borderland documents a "progres- case of Y-branched gradualism because its evi- sive suppression of ribs" (1966, p. 523) among GOULD & ELDREDGE

FIGURE3. The phyletic pattern of change in the radiolarian Pseudocubus vema from core E14-8; from Kellogg 1975. Vertical lines through points are 95% confidence intervals for means; numbers above lines represent the sample size. four successive species in 14 samples. Of four Hurst (1975) studied the delthyrial-width traits treated quantitatively, only the height/ length-of-delthyrial-chamber ratio in 11 sam- width ratio of ribs exhibits a consistent trend ples of the Resserellu lineage from Wenlockian among all species (modal rib number increases and Ludlovian strata of Wales and the Welsh then decreases, size does the same, while the borderland. The first four samples show no angle enclosed by six ribs first decreases and trend. Only one sample contains more than then increases). For the height/width ratio, 10 specimens (18), while 8 of 11 include 6 or 6 of 14 samples have 6 or fewer specimens, fewer, and six (i.e., more than half the sam- while only three have more than 15 (20, 25, ples) contain four specimens or fewer. and 39). Coefficients of variation range up to 1. Kellogg's radiolarians. Kellogg ( 1975, p. 51 within samples. 359) bases her claim for "a long-term phyletic PUNCTUATED EQUILIBRIA RFCONSIDERED 127

TABLE1. Seven selected samples chosen non-randomly to illustrate the evolution of thoracic width in Pseudocubus uema.

Chosen by Kellogg Chosen by us Depth in interval between mean width Depth in interval between mean width core (cm) samples (cm) (microns) core (cm) samples ( cm) ( microns ) 1750 150 89.81 1710 110 93.79 1600 40 89.92 1600 GO 89.92 1560 220 105.67 1540 100 106.84 1340 60 108.28 1440 80 105.71 1280 135 117.41 1360 120 104.23 1145 60 119.33 1240 117 119.16 1085 134.67 1123 118.58 trend in a continuous fossil sequence" on 2 the single sample (9th from the left on Fig. n1.y. of the radiolarian Pseuclocubus vema 3) between plateaux 1and 2. We work against from a single antarctic deep sea core (Fig. our own hope of zero slopes by recognizing 3). We applaud this excellent study, with its no intermediate samples between plateaux 2 careful collection and presentation of copious and 3. Samples 1-8 from the first platea~r, data from what are probably the most con- 10-22 the second, and 23-32 the third. We tinuous sedimentary sequences in the geolog- compute the following three relationships, all ical record. Though we cheerfully admit our with positive slope, but none anywhere near prejudices, we find it hard to view Kellogg's a statistically significant difference from zero; pattern (Fig. 3) as anything but a series of three Y = 31.19 .00129X for the first plateau plateaux, periods of stasis interrupted by very + Y = 36.35 .00286X for the second rapid rates of change, all admittedly in the + and same direction. (We argued on pp. 121-122 Y = 39.46 .00434X for the third that unidirectionality of mean points in short + sequences is no test of gradualism). Indeed, (width in micrometer units at 2.9 to the mi- Kellogg admits that her trend is "'stepped' cron, depth in cm). A hypothesis that all in- rather than straight" (1975, p. 367). None- crease in width occurred only during very theless, she writes (1975, p. 366) : "The single brief periods of rapid change between pla- feature which dominates both is a strong trend teaux is fully consistent with the data. Since toward increased size running the entire the punctuations occur within a plexus of pre- length of the stratigraphic range of P. vema dominantlv asexual clones rather than between within the core." In her tabular calculation reproducthely isolated taxa, we cannot fit of evolutionary rates (1975, p. 365), Kellogg them strictly within our model (but see pp. selects 7 of her 34 san~plesto display a con- 141-142). We restricted punctuated equilib- tinuous increase in thoracic width (Table 1). ria to conventional speciation in sexually re- If we select a different seven, spaced even producing Metazoa (Eldridge and Gould more regularly through the core, we note four 1972, p. 94), but we believe that the general decreases with only two increases (Table 1). phyletic geometry of long stasis (or mild, Our samples are, of course, selected a priori directionless fluctuation) followed by rapid to fit our biases-decreases within plateaux change is more widely applicable. and increases only between them. But hers Other intrinsic limitations of Kellogg's study are equally selected to suit her interpretation preclude any proof of her preference for phy- -and this is the only point we try to make in letic gradualism directed by natural selection. Table 1. First of all, she has no control on geographic From data kindly supplied by Dr. Kellogg, variation. Gradualism cannot be verified in we have computed least squares regressions local sections of widely-dispersed species. As of mean width vs. depth in the core for each Newell argued long ago (1956), spurious of the three apparent plateaux. We omit 3 "phyletic change" may arise in local sections of 34 samples as representing intermediate by successive immigration of normal geo- periods of rapid transition-the last two, and graphic variants responding to changing local GOULD & ELDREDGE

z C0 Diameter of Proloculur in mtcronr m -290 , 400 690 , 890 FIGURE4. Geographic and temporal variation in prolocular diameter for the Permian foraminifer Lepi- dolina multiseptata in East Asia; from Ozawa 1975. Black rectangles are 95% confidence limits for the mean; white rectangles span two standard deviations. Samples grouped together between wider intervals are, in Ozawa's words, "not of successional occurrence but of almost the same horizonn-hence, they represent the rather narrow range of geographic variation for any one time over wide areas. The mean changes by more than 100% during this phyletic sequence. environments. Take, for example, a mammal mammals can be explained in this manner. distributed through its range according to Kellogg proposes that P, vema increased in Bergmann's rule, with larger animals in colder width as waters became colder ( 1975, p. 361); climates. As a local area cools gradually her "trend could reflect migration rather through time, increasing size up section may than phyletic change. And if the increase is record nothing more than the shift to lower phyletic (affecting the entire species, though latitudes of a static range of geographic vari- only sampled in one spot), why must we in- ation. Many local "trends" in Pleistocene voke genetic change mediated by natural se- PUNCTUATED EQUILIBRIA RECONSIDERED 129 lection-as Kellogg does (p. 368). For basic be merely migrational since geographic vari- dinlensions of simple creatures, a purely ation has been measured and eliminated. It phenotypic response of an unaltered genotype is too long, extensive and persistent to be ran- to changing environments seems just as likely. dom. The correlation with altered habitat and We must not make up stories about the power the complexity of morphological change (in- of natural selection, just because modern theory ternal as well as prolocular size) points to favors it as an evolutionary agent. In so doing, some genetic alteration: shifting habitat sup- we do not strengthen the Darwinian cause, but plies a reasonable selective pressure. only display our biases. Finally, Kellogg states We are delighted with these results, and that she chose P. vema (1975, p. 362) because believe that they reflect well upon our model. she had noted considerable change of form We expect counter cases, especially among between its first and last appearance. Un- predominantly asexual forms (see p. 142). We biased tests of gradualism must study all are, like all honest men, anxious to vindicate adequately preserved species in faunas, not the substantive predictions of our model: but only the ones that appear to change. Stasis we also hope that punctuated equilibria will is data. serve as an organizing device for reorienting 2. Ozawa's forams. Ozawa's superb study the central study of evolutionary tempo- of the Permian verbeekinoid foraminifer Lepi- away from a biased consideration of rare cases dolina multiseptata should stand as a model towards a fair assessment that can resolve for the testing of evolutionary tempos and the some important issues in evolutionary theory. inference of modes. It represents the only In this case, Ozawa used our model to recog- case of gradualism that we find fully satis- nize that adequate tests must include an ex- factory. Ozawa studied 9 characters. Five plicit study of geographic variability. do not display gradualism; of the four that do, 3. Gingerich's mammals. Gingerich ( 1974, only the change in prolocular diameter is im- 1976, in press) has been our most forceful and pressive-but it is very impressive (Fig. 4), cogent critic. He has attempted to trace the for a suite of reasons: phylogeny of several early Eocene mammals i) Samples are sufficiently numerous to by followiiig their stratigraphic history in preclude any claim of random change. Raup Wasatchian strata of the Big Horn Basin in (in press) has shown that this sequence of 34 Northern Wyoming. His recent study (1976) samples could not, at any reasonable level of includes the condylarth Haplomylzls and the probability, arise by a random walk. Pelycodus, but Gingerich bases most ii) Sample sizes are large enough to estab- of his claims on the more complex phylogeny lish means and variances with sufficient con- of the condylarth Hyopsodus (Fig. 5)". Gin- fidence. Sample size ranges from 8 to 292; gerich uses the logarithm of length x width only 6 of 34 samples contain fewer than 30 of the first lower molar as his only criterion specimens, and only 1has fewer than 10. for a quantitative phylogeny and a defense of iii) The stratigraphic range is large, span- gradualism. ning Middle to Upper Permian times. In advocating gradualism as the primary iv) Most importantly (for it successfully tempo of life's history, Gingerich lays proper overcomes the Achilles Heel of most other stress upon a claim that rates of nlorphological claims for gradualism), Ozawa has studied separation in speciation are basically the same geographic variation explicitly. His samples as rates of phyletic transformation within come from a wide geographic area of South- lineages. This is a central point often neglected east Asia, Southern China and Japan. He has by defenders of gradualism. No gradualist demonstrated that geographic variation is slight compared with the magnitude of mean * Gineerich has ~ublishedtwo other fieureq with-... more...-.- -wm--... ples than the sect^ion reproduced SgG 5-1974 and stratigraphic change. 1976, p. 13. But the test of punctuated equilibria, as' Gin- gerich agrees, must be based on Fig. 5 (his Fig. 4 of 1976. v) The change in prolocular diameter can p. 12)-for only this section orders-samples by Tield evidence of stratigraphic superposition alone. The more detailed sec- be correlated with a shift in habitat from shal- tions include samples interpolated by assuming gradualism and ordering samples of Hyopsodus accordin ly. Gingerich low carbonates to deeper elastics. Here states ( 1976 pp. 10-11 ) that "the stratigrapsic framework must be set' up completely independently of the fossils of enough- information is available to defend a interest": he then admits (1976, p. 11) that the sections with interpolations "cannot be cited as evidence for gradual phyletic selectionist interpretation. The change cannot evolu~tion." 130 GOULD & ELDREDGE

1--1L I ,80 .90 1.00 1.10 1,20 1.30 1.40 LOG (LxW) OF MI FIGURE5. The phylogeny of Hyopsodus in the Big Horn Basin as given by Gingerich 1976, and based only on samples in measured sections. Horizontal bar is standard error; horizontal line represents total range of variation; sample size indicated at the right of each distribution; points are single specimens.

can deny that species often arise by splitting- ii) the pattern of Fig. 5 does not, in fact, how else could diversitv ever increase? It is establish the prevalence of such unidirectional not enough simply to demonstrate a few cases change; of gradualism in phyletic transformation; iii) even if the pattern of Fig. 5 did indi- rates of separation between splitting lineages cate gradualism, the inferred rates are too must also be no greater than the average slow to account for anything important in tempo of phyletic change. Gingerich writes evolution. (1976, p. 2): "If rates of divergence are rela- i) Gingerich presents no study of geo- tively slow, that is, as slow as documented graphic variation. Gradualism throughout a rates of phyletic evolution, the diversification species of broad range cannot be inferred from of life is properly viewed as a gradual process, local sections. Gingerich determines only species are dynamic links in a great chain, and what happened in one place-the Big Horn the traditional branching tree is an adequate Basin of Wyoming-though he admits (1976, representation." p. 5) that there is "no to think that the We welcome Gingerich's approach to test- Paleocene and Eocene faunas of the Big Horn ing by meticulous stratigraphic work; his data Basin were geographically isolated from their are among the most important ever collected counterparts preserved in other basins in the by paleontologists for the study of evolution- Rocky Mountains." Gingerich's most impres- ary rates. Nonetheless, we cannot agree that sive evidence for sustained unidirectionality his information, as presented, confirms a wide- is particularly subject to an alternate inter- ranging claim for gradualism; his data on the pretation based on migration and an unaltered splitting of lineages seem to fit our model of pattern of geographic variation with no shift punctuated equilibria better. We proceed of mean values through time. In Lower Gray by three levels of argument: Bull strata, all three lineages ( Hyopsodus, i) the data, as given, are inadequate in Haplomylus, and Pelycodus) display a sus- principle to validate gradualism even when tained, simultaneous increase in size. Such the pattern of Fig. 5 seems to indicate uni- impressive coincidence leads us to wonder directional change; whether Gingerich's notion of similar response PUNCTUATED EQUILIBRIA RECONSIDERED to independent regimes of selection provides the most likely explanation. An alternate hy- pothesis based on geographic variability is at least as likely: all species vary as Bergmann's rule with larger animals in more northerly latitudes; climate became colder in the Big Horn basin during Graybullian times and larger bodied, northern populations, moved d 50 100 150 200 into Wyoming. We do not assert this alter- STRATIGRAPHIC LEVEL (LEVEL 1140 OF GlNGERlCH=O) native as intrinsically preferable; we merely FIGURE6. Plots of mean tooth size versus strati- point out that Gingerich does not provide the graphic level for a segment of Gingerich's Hyopsodus essential data to test it. Any judgment about lineage, as discussed and defined in the text. The gradualism requires a knowledge of how Hy- regression indicates stasis during a period of supposed opsodus changed in other basins. size decrease in the H. miticulus to H. lysitensis lineage. Dotted line is the least squares regression ii) We do not see the same unambiguous calculated from Gingerich's data. evidence for gradualism that Gingerich af- firms in Fig. 5--for several reasons: a) The dotted guidelines that Gingerich latidens-H. minor) is not a unidirectional draws about his data are gradualistic inter- trend, but a zigzag of intermediate increase pretations, not literal renderings of the evi- ending up just about were it started (see dence. We find long segments of apparent Fig. 5). stasis within supposedly gradualistic se- c) We see no evidence for Gingerich's most quences. Consider, for example, levels 160- important contention: that speciation is a 500, a period of supposed size increase lead- process of gradual divergence at rates similar ing from H. loomisi to H. htidens. If we take to those in phyletic sequences. Consider all all nine samples (excluding only the single cladogenetic events in Fig. 5: H. simplex is a specimen at level 500 because it is morpho- single specimen and we can say nothing about logically separate from another sample of six it. H. miticulus (again omitting the single specimens from the same level), and plot them specimen at level 920) arises disjunctly at a against stratigraphic level, we compute a maximal distance from its ancestor, and then slope insignificantly different from zero. We evolves in parallel with it. The ratio of means are especially grateful to Dr. Gingerich for for the two lineages in 1.21 (1.15/0.95) at the supplying us with his raw data, knowing full inception of H. miticulus (level 1140) and 1.25 well that we would attempt to use it against (1.09/0.87) and 1.19 (1.06/0.S9) at their last him. two sympatric occurrences. H. mentalis has only one sample in the Big Horn Basin. Finally, H. powellianus includes only three Secondly, the H. miticulus to H. lysitensis samples, displaying no trend (the first and lineage supposedly displays gradual size de- last samples have the same mean value, though crease. Indeed, it does to our satisfaction at Gingerich's guide lines indicate size increase). upper levels; but the first 9 of its 15 samples Cladogenetic events in Pelycodus tell the same (spanning 50% of its stratigraphic range) tale (there are no such events in the third, yield a slope insignificantly different from Haplomylus lineage). Gingerich places spe- zero, and a decrease (calculated from the cial emphasis upon the supposed gradual di- curve) from 1.15 to 1.14 in log (1 x w) over vergence of "N." nunienus from "N." uenti- the entire sequence (Fig. 6). colis (see Fig. 7; these species are now placed in Notharctus, but will be revised by Gin- gerich) : "The two species, once established (We omit from this calculation only the sin- sympatrically, diverge in a regular way from gle specimen at level 920, disjunctly situated each other" (1976, p. 25). We admit the 220 ft below the initiation of a lineage that gradual size decrease of "N." nunienus, but only endured for 340 ft over its 15 samples.) see no evidence for the key claim of gradual b) The major episode of apparent gradu- departure in sympatry from "N." uenticolis. alism within Hyopsodus (H. loomisi-H. Gingerich has only three samples of "N." venti- GOULD & ELDREDGE

. *----*------. ,------0 -----4 , . --..1/ I . "N." nunienus "N. " vanticolis , '. I -3 .. 3 1/

-. . 14) ,' -6 3 13 ,' P jorrovii

4 ,,I / r / STRATIGRAPHIC RECORD OF 6

I ,-2 /' EARLY EOCENE PELYCODUS ,I ,:-7:-*- P rokloni ,, :...... \ --

[ 0 ;f," /' -;.;,:..,:: := ,"- I I' I I I I' I I I, 12 I I 1 I L I 1.00 1.10 1.20 1.30 1.40 LOG (LxW) OF MI FIGURE7. Supposedly gradual in the Pelycodus lineage, according to Gingerich 1976. Symbols as in Fig. 5. We see no evidence for gradual divergence of "N." venticolis from its ancestor. colis; the first two are based on single speci- a random pool of speciation events produced mens, and the literal pattern among the three by punctuated equilibria. Species selection is a zig-zag of increase followed by decrease. depends upon the validity of "Wright's rule" In short, Gingerich presents no evidence for (Gould, in press)-the claim that speciation his most important claim that "the paired de- is essentially random with respect to the direc- scendant species appearing after speciation tion of a macroevoIutionary trend (Wright events . . . were not initially distinctive, but 1967). Wright's rule must be tested exten- only became so after a significant period of sively and affirmed if species selection is as time" (1976, p. 25). A literal interpretation fundamental a process as we believe. Gin- of his cladogenetic events would support our gerich provides the first adequate test. The model of punctuated equilibria. general trend in the entire Hy'opsodus clade d) On a positive note, Gingerich's data for is towards larger size (all three final species Hyopsodus offer the finest confirmation now are larger than ancestraI H. loomisi). Gin- available for the most important implication gerich writes (1974, p. 108): "The Hyopsodus of punctuated equilibria (see section 1V)- radiation clearly conforms to Cope's rule." Yet the explanation for trends that we chose not of the nine species descended from H. loomisi to christen explicitly (Eldredge and Gould on Fig. 5, five evolve towards smaller size, 1972, pp. 111-113), and that Stanley (1975) and only four (H. latidens, H. miticulus, H. has since called "species selectionJ'-i.e., mac- powellianus, and H. walcottianus) become roevolutionary trends are not a result of grad- larger. We cannot attribute size increase to ualistic orthoselection, but arise from a "higher gradualism within lineages, for not a single level selection" of certain morphologies from lineage displays it (contrary to Gingerich's PUNCTUATED EQUILIBRIA RECONSIDERED 133

TABLE2. Rates of evolution (assuming gradualism) for Gingerich's most rapid evolutionary events.

% change (absolute difference between beginning % and end change Stratigraphic Inferred Log (LX W) L)c\y \/wdivided by interval time beginning beginning beglnnlng beginning m%&n Event (ft) (rn.~.) and end and end and end value) years A-decrease H. latidens-H. minor 640 1.60 1.08-0.87 12.02-7.41 3.467-2.723 21.5 13.4

C-increase in Haplomylus 660 1.65 0.75-0.83 5.62-6.76 2.371-2.600 9.5 5.8 D-increase in Pelycodus 1140 2.85 1.11-1.33 12.88-21.83 3.5894.624 28.8 10.1 guidelines, there is no increase within H. of length, rather than in his calculation of log loomisi-see p. 131). The increase must oc- length2 [length x width of molars]; thus, we cur in cladogenesis, and we have already work with the square root of the antilog of argued that cladistic events conform to punc- mean values presented in Gingerich's papers). tuated equilibria. The phylogeny of Hyop- a) Size decrease in H. lutidens to H. minor sodus therefore affirms Wright's rule. Size (from the first latidens at level 680 to the last increase in the entire clade arises from the good sample of minor at level 1320-neglect- differential success of larger species in a ran- ing for Gingerich's sake the single specimen dom subset of cladistic events. (In fact, new of larger size at level 1340). In 640 ft of sec- species are more often smaller than larger.) tion (1.6 m.y.), length decreases by 21.5%, iii) Quite apart from the bickering that will for a rate of 13.4% per million years. continue for years over whether this or that b) Size decrease from H. miticulus to H. case really exhibits gradualism, we must con- lysitensis (neglecting, for Gingerich's sake, sider the characteristic rates of supposed grad- the segment of stasis within the miticulus ualistic events. When this is done, one cardi- lineage (see p. 131). In 180 ft of section nal fact emerges: they are too slow to account (450,000 yr), size decreases by 11.9%, for a for most important evolutionary phenomena, rate of 26.4% per million years. particularly for adaptive radiations and the c) Size increase within Haplomylus (Gin- origin of new morphological designs. We re- gerich 1976, p. 15, neglecting the Clarkforkian gard gradualism as unimportant in evolution specimens collected elsewhere and the single not only because it occurs rarely, but also specimen at the top of the sequence). In 660 because its rates are only sufficient to cast ft of section ( 1.65 m.y. ), we calculate a 9.5% a superficial molding upon the pattern of evo- increase in length, for a rate of 5.8% per mil- lutionary change. As Stanley writes (1975): lion years. "Phyletic evolution is much more sluggish and d) Since increase within Pelycodus (Gin- less significant than has generally been recog- gerich 1976, p. 16, again neglecting the Clark- nized." forkian sample collected elsewhere and run- Let us give Gingerich the benefit of all ning to the sample at level 1300, since decrease doubt and choose the most rapid rates accord- in size begins immediately thereafter). In ing to his phyletic interpretations (Table 2). 1140 ft of section (2.85 m.y.), length increased (Gingerich gives three slightly differing esti- by 28.8%, for a rate of 10.1% per million years. mates of sedimentary rates-20 ft in 6,000, Rates for simple increase in size (without 50,000 or 60,000 yr. We use the intermediate any complex alteration in shape) of 6 to 26% value of 50,000 though any of the three would per million years will not encompass the early support our interpretation. We also present Tertiary radiation of mammals. We again give Gingerich's rates in the traditional dimension Gingerich the benefit of all doubt by using a 134 COULD & ELDREDGE

"compound interest" rather than an additive verged to minimize competition with each model; still, at the most rapid rate of 26% other." Not at a rate of 10% change in length (twice that of any other calculation), a per million years! For such a rate is invisible doubling of length still requires 3 m.y. At the ecological time. Character displacement is average rate of 13.9%, it requires 5.3 m.y. In a geologically instantaneous process (see this particular case, all of Gingerich's lineages Eldredge 1975). originated abruptly in basal Clarkforkian If such tiny, sustained rates of change times before the deposition of his Big Horn actually exist in paleontological time, what do sequence. These originations fit our model they mean? Traditionally, they have been of punctuated equilibria by rapid events of interpreted as the primary defenders of con- speciation-and they involve fundamental al- ventional selection theory. But we demur, terations in morphology, not simple change in for the rates are far too slow and far too con- size. Even if we claim that these lineages tinuous for selection in such a dynamic world. arose phyletically in areas not yet explored We wish cautiously to float a radical proposal: or devoid of appropriate strata (as Gingerich, perhaps these rates do not provide a com- in press, speculates), rates must have been fortable confirmation of traditional panselec- many times higher than those calculated in tionism; perhaps they constitute a fundamental the Big Horn Basin. Gingerich's "trends" mystery worthy of our serious thought and seem to be minor fluctuations in the history of attention. lineages that arose abruptly not much earlikr. E) Cases that confirm punctuated equilibria. How can we view a steady progression -By emphasizing cases presented as contrary yielding a 10% increase in a million years as to our model, we do not wish to imply that anything but a meaningless abstraction? Can we are without supporters. In fact, most pub- this varied world of ours possibly impose such lished commentary on punctuated equilibria minute selection pressures so uninterruptedly has been favorable. We are especially pleased for so long? The 's surface is not like that several paleontologists now state with the calculating pad of a population geneticist. pride and biological confidence a conclusion Lande (1976) has used three of Gingerich's that had previously been simply embarrassing Hyopsodus sequences to calcuate selective ("all these years of work and I haven't found rates in paleontological events. The minimum any evolution"). amount of selection required to explain these Several lineages have been cited in support rates is only 1selective death per 100,000 in- of our model. Kellogg and Hays (1975) argue dividuals per generation, corresponding to a that about 1.9 m.y. ago, a small population of truncation point 4.3 standard deviations from the radiolarian Eucyrtidium calvertense in- the mean. In a burst of understatement, Lande vaded subarctic waters north of the Arctic writes: "this is very weak selection." In fact, Convergence. There it evolved rapidly and it is so weak that the change could easily be allopatrically to E. matuyamai, a larger and accomplished by genetic drift, even in large more robust species. E. matuyamai then in- populations (the smallest population size for vaded the range of its ancestor. For the pre- which such change could occur by drift at least 5% of the time is only 10,000). vious two m.y., E. calvertense had been stable The scale of these events is simply all wrong in size, but it decreased rapidly after contact for the usual interpretations extrapolated up with its larger descendant. When E. matu- from observations made by neontologists in yamai became extinct a short while later, the ecological time. Thus, Gingerich (1976, p. evolution of E. calvertense to smaller size 26) would attribute the gradual divergence ceased. of two species in sympatry to competition for Reyment (1975) presents an interesting case resources (though in another place, 1974, p. of "pseudo-gradualism," resolved to abrupt 108, he seems to be aware that his rates are disjunction by the study of covariance rather too slow): "Both the differing deme and its than static adult morphology. In upper Turo- homogeneous neighbors might survive, each nian strata on Hokkaido, the "generic" transi- creating a strong (sic) directional field of se- tion from Subprionocyclus normalis to Reed- lection. . . . A parent species divided into two sites minimus is clearly continuous in basic new sister species which subsequently di- dimensions of adult morphology. Means and PUNCTUATED EQUILIBRIA RECONSIDERED 135

covariances are so similar that the first four species seem to have real meaning among these eigenvectors of principal components for each olenids rather than being arbitrary points on a species taken separately are virtually identical. continuous morphological spectrum changing But the pattern of covariance between these with time." variables and measures of ornamentation We are particularly pleased that our model changes abruptly. Correlation coefficients be- of punctuated equilibria has been used to re- tween ornamentational and dimensional vari- assess the evolution of the clade most near to ables range +.39 to f.55 in R. minimus and all our hearts-the peculiar pongid offshoot from -.32 to -.SO in S. normalis. Reylnent con- whose only living representative is H. sapiens cludes (1975, p. 675) : (see Walker 1975; and Pilbeam 1975). In no other group, has the a priori assumption of The major morphological difference be- gradualism been so pervasive. It supported tween the two genera did not take place racial classifications to justify imperialism in gradually. . . . The change was not the nineteenth century, and it still regulates heralded by a gradual change in the fre- our exalted view of our own estate. The bias quencies of the ornamentational charac- is so strong that Brace (1967) once labelled ters. . . . The concept of 'punctuated equi- as "hominid catastrophism"-a rearguard ac- libria' of Eldredge and Gould seems to fit tion of those who still do not wish to accept these observations well. our brutish ancestry-all attempts to deny di- Other paleontologists have presented pat- rect ancestral-descendant relationships among terns in larger clades as affirmations of our known hominid fossils, and to attribute ob- model. Henry and Clarkson (1975) discerned served patterns to extinction and subsequent a clear trend to greater complexity of enroll- migration. Human evolution has been viewed ment structures in the as a ladder of progress. Placoparia. But they detect no gradual Recent discoveries have discredited the na- change within any species, while patterns of ive notion of a single lineage, Australopithecus geographic variation point to the prevalence africanus-Homo erectus-Homo sapiens, with of allopatric speciation. The trend seems to gradual increase in brain size within each be a result of species selection. taxon. All new evidence points to a branching In his study of Arenigian olenid bush with rapid origination and subsequent in Spitsbergen, Fortey has followed both our stasis within taxa (Eldredge and Tattersall primary recommendations for a reformulation 1975; Gould 1976). On mechanical and bio- of method in the study of evolutionary tempos metric grounds, Oxnard (1975) has argued -study all taxa, and consider stasis as data. that the australopithecines, although a sister Fortey (1974) collected bed by bed through group to us, were not directly ancestral to any an "exceptionally complete record (1974, p. subsequent hominid. (Several paleoanthropol- 4) and established the phylogeny of an ex- ogists who generally support our model do not tensive adaptive radiation in the Balnibarbi- accept Oxnard's specific conclusion-E. Delson inae. He concludes that "the period of deriva- and A. Walker, for example). In any case, there tion of a new species from its ancestor is short is no direct evidence for gradualism within compared with its subsequent duration, and any hominid taxon-A. africanus, A. robustus, that this speciation pattern is consistent with A. boisei, H. habilk, H. erectus, and even H. the allopatric model" (1974, p. 4). Fortey sapiens. Each species disappears looking rejects the attribution of morphological gaps much as it did at its origin; admittedly "pro- to breaks in the stratigraphic record because gressive" trends result from the differential times of origin do not correspond in different survival of discrete taxa. lineages and because he finds no lithological Richard Leakey's discovery of hominid E.R. indication of non-sequence or even of reduced 1470 has shattered the conventional view that sedimentation at points of origination for new Homo evolved gradually from A. africanus; species. Fortey concludes, in confirmation of for this member of our genus, with its cranial traditional stratigraphic practice (1974, p. 20) : capacity of nearly 800 cc, lived in sympatry "Evolution in this case defines discrete mor- with australopithecines, perhaps as long as 3 phological groups with a particular strati- m.y. ago. The more recent discovery of a graphic range-that is, that paleontological remarkable H. erectus from the Koobi Fora 136 GOULD & ELDREDGE

Formation, east of Lake Turkana in East Ager (1973, 1976) shares the same view for Africa has discredited the strongest traditional Mesozoic brachiopods: "In twenty years tale of hominid gradualism-a progressive in- work on the Mesozoic Brachiopods, I have crease in brain size from primitive demes in found plenty of relationships, but few if any Java to the advanced population at Choukou- evolving lineages. . . . What it seems to tien (Peking Man). This specimen, older mean is that evolution did not normally pro- than any non-African H. erectus, has a cranial ceed by a process of gradual change of one capacity "well within the range of the Peking species into another over long periods of time. specimens" (Leakey and Walker 1976, p. 572). I have long criticized the notion that evolu- (See Howells, in press, for a defense of H. tion can be studied by chasing fossil oysters erectus as a stable taxon, not a grade in a tale up a single cliff. . . . One must clearly study of continuous improvement. ) the variation of a species throughout its geo- Finally, several paleontologists have sup- graphical range, at one moment in geological ported our model for broader patterns in the time, before one can claim that it has changed major groups of their expertise. These qualita- into something else" (1973, p. 20). tive assessments follow our criterion of rela- Reyment (1975) has supported stasis as a tive frequency. After a major study of Middle predominant tempo, even for microorganisms Cambrian agnostid trilobites in western North from continuous sections-the usual arena of America, Robison ( 1975) concludes : "I have greatest hope for abundant gradualism: "The found a conspicuous lack of intergradation in occurrences of long sequences within species species-specific characters, and I have also are common in boreholes and it is possible to found little or no change in these cl~aracters exploit the statistical properties of such se- throughout the observed stratigraphic ranges quences in detailed biostratigraphy. It is note- of most species" (1975, p. 220). "Most species worthy that gradual, directed transitions from of the suborder Agnostina are non-intergrad- one species to another do not seem to exist in ing and best fit an allopatric model of specia- borehole samples of microorganisms" ( 1975, tion" (1975, p. 219). p. 665). MacGillavry (1968, p. 70) was also Johnson (1975) has assessed the relative forced by his own observations to abandon frequency of evolution by allopatric specia- a previous commitment to gradualism for mi- tion in Devonian brachiopods and has come croorganisms in local sections: "During my down strongly on our side, primarily because work as an oil paleontologist I had the op- most variation within species is geographically portunity to study sections meeting these rigid rather than temporally distributed. Johnson requirements. As an ardent student of evolu- reached this conclusion only after years of tion, moreover, I was continually on the watch search for preferred examples of phyletic grad- for evidence of evolutionary change. . . . The ualism. Recalling the hope engendered by great majority of species do not show any ap- gradualism in Eocoelia ( Ziegler 1966), he preciable evolutionary change at all. These writes (1975, p. 657) : species appear in the section (first occur- After completion of Ziegler's paper we rence) without obvious ancestors in under- talked a number of times about the possi- lying beds, are stable once established, and bilities for duplicating his efforts with disappear higher up without leaving any de- other fossils and in other times. It was scendants." We do not cite these two state- a heady prospect. . . . In subsequent ments about local sections as documentation years many workers have attempted to of our model, for we have emphasized that seek out and define lineages of bra- the study of geographic variation is an indis- chiopod species and other megafossils in pensable component of any decision. These the lower and middle Paleozoic with statements do, however, confute the basic little success. My conclusion, subjective tenet of gradualism that even transition pre- in many ways, is that speciation of bra- vades entire populations. chiopods in the mid-Paleozoic via a phy- In summary, our model has survived its letic mode has been rare. Rather, it is first five years in excellent shape. It has re- probable that most new brachiopod spe- ceived much empirical support and has fit cies of this age originated by allopatric the impression of many specialists who spent speciation. a professional lifetime searching for gradual- PUNCTUATED EQUILIBRIA RECONSIDERED 137

ism \vithin their groups. It is very hard to showing incipient reproductive isolation, often find probable cases of gradualism, even in in the form of sterility, exhibit signif- geologically optimal situations; most reported icantly greater genetic distances, involving ma- cases resolve to little more than wishful think- jor allelic changes at up to 20% or more of ing. Moreover, all these tests are based on a structural genes" ( 1976, p. 113). Again, a ma- biased selection of cases kno\vn to exhibit jor genetic alteration characterizes the early some evolutionary change. What would hap- stages of speciation, following spatial isola- pen if paleontologists carried out large-scale, tion and the establishment of (even im- unbiased studies that admitted stasis as data perfect) reproductive isolation. Avise con- and considered all taxa in a fauna? We cannot cludes (1976, p. 120): "Arguments that avoid the prediction that punctuated equilib- speciation is normally accompanied by little ria \vould assume even greater importance. genetic change are clearly refuted." The fact F ) A note on indirect tests from the genetics that strongly differentiated species exhibit of living organisms.-Punctuated equilibria far less genetic similarity speaks neither for needs one "enabling criterion" from the ge- nor against our model; for we do not know netics of speciation in living organisms. We whether increasing genetic distance results must be able to assert that a major amount of from gradual separation or from the interposi- genetic change often accumulates in the event tion of additional speciation events between of speciation itself. If nearly every case of re- the common ancestor and surviving taxa. cent speciation separated a daughter popu- A few studies in biochemical genetics speak lation only minutely from its parent, and if more directly to our hypothesis. They also greater genetic divergence were primarily a speak very ambiguously, some for and some function of time, then gradualism would re- against. In a work that provides evidence ceive some support. We must be able to state against our model, Avise and Ayala (1976) that Mayr's "genetic revolution" (Mayr 1963, studied average structural-gene differences p. 538) is a common component of specia- between species in two clades of fishes, the tion. (As Stanley [I9751 emphasizes, it need centrarchid Lepomis with 11 living species, not be a ubiquitous property. Indeed, since all North American, and the North American speciation is defined only as the acquisition of minnows, with 250 species. Scrappy fossil reproductive isolation, it must often occur evidence indicates that both clades are equally with only minor genetic input-as when, for ancient (the first Lepomis appears at the example, small genetic changes produce a Miocene-Pliocene border, while the first fos- shift to a new host plant in host-specific in- sil minnows are Miocene). If both clades are sects-Bush 1969.) the same age, then their major difference lies Mayr supported his notion with indirect in the number of speciation events. By our evidence of external morphology; recent re- model, random pairs of minnow species should search in biochemical genetics has confirmed be more genetically dissimilar than random his suspicions. Ayala et al. (1974) studied pairs of Lepomis since, on average, they will the genetics of speciation in the Drosophila be separated by more speciation events. Aver- toillZstoni group. They found that a substan- age distances are similar for pairs of species tial amount of genetic differentiation (aver- within the two clades. aging .23 electrophoretically detectable allelic We applaud this approach to the genetic substitutions per locus) occurred during what study of evolutionary divergence, but we do they called the "first stage" of speciation-i.e., not consider this case as strong evidence in allopatric, isolated populations that have against our model-and neither do Avise and already acquired at least partial reproductive Ayala (personal communication). First of all, isolation. Sister species in the "second stage" too many uncertainties envelop the study (as (now in sympatry and completing reproduc- Avise 1976, p. 118, admits). Are the clades tive isolation) differed no more than pairs really of equal age? Are the selected min- still in the first stage. Avise (1976) has gen- nows, all from California, truly a random sub- eralized these conclusions. He finds that indi- set of their clade, or might they represent a viduals in interacting populations generally subgroup with rather recent times of diver- share up to 99% of the structural genes sur- gence from each other? veyed by electrophoresis, while "populations Our second reservation is far more impor- 138 GOULD & ELDREDGE

tant: We do not know what these changes in TI,CE A B structural genes represent in terms of adaptive evolution. After years of squabbling and out- of-hand denial, evolutionary biologists must now take seriously the proposition that many, if not most, changes in structural genes drift I I,, ,',,\,:,/, , ,' , ,,,, I,,, I to fixation in the neutral mode, do not affect II;] [,I phenotypes, and are therefore both irrelevant and invisible to Darwinian processes (e.g. Nei 1975). Neutral substitutions will accumu- ILI \,,,, 1',,,1 late in clock-like fashion and lead to the con- ,I clusions affirmed by Avise and Ayala in their P, genetic studies of fishes. But neutral substitu- J ,, OPmH^.LOGY tions are irrelevant both to gradualism and to 1 punctuated equilibria, for both deal with + I/, adaptive, phenotypic evolution. (A gradualist can take no comfort in the prospect of ran- / dom genetic divergence, even among genes with ~henot~~iceffects, for this process will FIGURE8. The origin of an evolutionary trend by yield directional trends in morphology only at species selection, rather than directed gradualism. the low frequency that stochastic processes From Eldredge and Gould 1972. Stasis prevails allow-see Raup, in press.) within species, and speciation is random with respect We wish to propose a potential use of our to the direction of the evolutionary trend. model in molecular -genetics. (We cannot advocate it he'' because we are arguing for extent than on collstancy of genic structure" the validity of punctuated equilibria, not for (Zuckerkandl 196S, p. 270). schemes that work by its JWe applaud the burgeoning emphasis on truth. ) If a cornensus judges our change in regulatory genes as the stuffof mc,r- favorably, then it can serve as a test for neu- phological evolutioll (Wilson 1976; King and tralism, and the information provided by min- Wilson 1975; valentine and canlpbe1l 1975 1, and L'pomis suppo* the hy- if only because one of us has written a book pothesis that most changes in structural genes to argue that the classical, and ignored, are neutral. The argument is similar to that data on evolution by should be advanced Zuckerkandl ( 1''' ). Assuming exhumed and valued as a primary demollstra- the validity of von Baer's laws, Zuckerkandl tion of regulatory ( Gould 1977) We predicted that fetal proteins would be less dif- do not see h,, point mutations in structural ferent than corresponding adult proteins in genes can lead, eveIl by gradual accumula- the same species. But he measured equal tion, to new morphological designs. Regula- levels of divergence. This discovery leads in tory changes in the timing of complex onto- two opposite directions: either von Baer's genetic programs seem far more promising- laws are wrong and we should not expect and potentially rapid, in conformity with our greater evolutionary conservatism in fetal than punctuational predilections. The near iden- adult structures, or "on Baer's laws are valid tity of humans and chimps for structural genes and the changes are neutral with respect to (King and Wilson 1975), and the evidence of . Since von Baer's laws have re- major regulatory change indicated by human neoteny (Gould 1977) provides an important ceived overwhelming support for 160 years, confirmation. Zuckerkandl chooses the second option. He we are pleased that some recent molecular concludes that adaptive evolution occurs pri- evidence, based on regulatory rather than lnarily by changes in genetic regulation; these structural gene changes, supports our model. are not examined in conventional studies of Ferris and Whitt (1975) have studied the genetic distance based upon structural genes evolution of diploidization in catostomid alone: "Reproducible morphogenesis depends fishes. They regard the ancestor of their clade on constancy of genic regulation to a larger as tetraploid. More specialized modern genera PUNCTUATED EQUILIBRIA RECONSIDERED 139 owe their status to a greater number of in- herited genetic constitution in the an- tervening speciations between common an- cestral components of a lineage, lead to cestor and living fish (Ferris, talk at A.I.B.S., new and improved efficiency. Improve- New Orleans, 1976). Unspecialized genera ment would be consistently greater within are equally old, but are separated by fewer this hypothetical subset of local conditions speciations from the common ancestor. If the that a population might invade. The over- Avise-Ayala conclusion for structural genes all effect would then be one of net, ap- were to apply, we would expect an equal parently directional change: but, as with amount of diploidization in specialized and the case of selection upon mutations, the unspecialized genera. But Ferris and Whitt initial variations would be stochastic with (1975, p. 30) find a positive correlation be- respect to this change. We postulate no tween this change in regulation and the num- 'new' type of selection. ber of speciation events: "We conclude that a slow return to a diploid level of gene expres- We declined to designate" with a new name sion is occurring in all genera, but that this this phenomenon of macroevolution by dif- diploidization occurs more rapidly in the more ferential success of speciation events because specialized genera. This may be attributed we regard it as so fundamentally consistent to the progressive loss or inactivation of some with basic . It represents no de- duplicated loci." parture from Darwinian mechanisms, but only the previously unrecognized mode of IV. Punctuated Equilibria as the Basis operation for natural selection at hierarchical for a Theory of Macroevolution: levels higher than the local population. Surely, The Speciation Theory the extinction and persistence of species is as Darwinian an event as the spreah of genes When we were writing our initial paper, no through populations. The differing geometry conclusion excited us more than the insight of change is simply an "allometric" phenom- offered by punctuated equilibria for a new enon of scale: the same process works in dif- interpretation of evolutionary trends. We fering ways at different ievels of complexity argued that evolutionary trends did not usually and organization. Stanley (1975a) has since arise by orthoselection in gradually-changing designated this process as "species selection." lineages, but that they represented the dif- We were inclined, at first, to reject this special ferential success of subsets from a potentially name (Gould, in press), since it carries the random pool of speciation events (Fig. 8, re- unfortunate implication of a novel mechanism produced from our 1972 paper). In this per- inconsistent with natural selection-while we spective, speciation is the raw material of believe that "species selection" represents no macroevolution, and genetic substitution more than the operation of natural selection within populations cannot be simply extrap- at higher levels. Nonetheless, recognizing a olated to encompass all events in the history peculiarity of human psychology-that the of life. We therefore challenged the central importance of a phenomenon is not recognized assumption that secured the admission of pale- unless it has a special name-we drop our ontology into the modern synthesis of evolu- objection, while reiterating our hope that the tionary theory (Simpson 1944 and 1953): relationship of species selection to Darwinian change in gene frequency within populations theorv will not be misconstrued. is the building block of major evolutionary Species selection follows directly from the events. We wrote (1972, p. 112): validity of two premises; these must first be A reconciliation of allopatric speciation tested and affirmed if species selection is to with long-term trends can be foimulated fornl the basis for a theory of macroevolution: along thi following lines: we envision i) the model of punctuated equilibria itself; multiple 'explorations' or 'experimenta- ii) the proposition that a set of morphologies tions' (see Schaeffer 1965)-i.e. invasions, produced by speciation events is essentially on a stochastic basis. of new environ- random with respect to the direction of evolu- ments by peripheral isolates. There is tionary trends within a clade. nothing inherently directional about these (1967) made this suggestion in explicit anal- invasions. However, a subset of these new ogy with the lower-level phenomena of ran- environments might, in the context of in- dom mutation and natural selection within 140 GOULD & ELDREDGE populations. We suggest that this proposition place of mutation. Whereas, natural se- be termed "Wright's rule,"* and that its test- lection operates upon individuals within ing be an item of high priority in paleobiol- populations, a process that can be termed ogy. We have already showed (pp. 132-133) species selection operates upon species that Gingerich's phylogeny of Hyopsodus pro- within higher taxa, determining statistical vides a first, impressive confirmation. trends. Wright's analogy represents the key to a claim that a new theory of macroevolution lies If the speciation theory of macroevolution is of general interest, it should perform as all in the expression: punctuated equilibria $ Wright's rule = species selection. Let us good theories to resolve paradoxes, reinterpret simply call it the "speciation theoiy" of macro- old observations, and synthesize under a com- evolution. Previously, mutation and natural mon rubric phenomena previously uncoordi- selection within populations were regarded nated. To cite just four examples of its po- as fully sufficient to render macroevolution: tential application: one had only to extrapolate their action di- 1. Stanley (1975b) has developed an rectly to longer times and higher taxa in elegant proposal to resolve, at least in part, larger clades. But if we (1972) and Stanley the classicial paradox of why so inefficient a (1975a) are right, then speciation interposes system as appears so com- itself as an intermediate level between macro- monly in nature. Traditional arguments focus evolutionary trends and evolutionary events on immediate adaptive benefits to the sexual organisms themselves: sex accelerates evolu- within populations. Species become the raw tion by providing efficient genetic recombina- material of macroevolution: they play the tion or by permitting the rapid spread of use- same role, at their level, as mutations do in ful mutations through populations. But Stanley local populations. All movement from micro proposes an equally forceful argument based to macroevolution must be translated through retrospectively on evolutionary history-an the level of species by Wright's grand anal- eminently paleobiological input to a tradi- ogy, not merely extrapolated up in continuity. tional neontological dilemma. He points out As Stanley puts it so well (1975a, p. 648) : that asexual clones do not speciate easily, Macroevolution is decoupled from micro- while sexual clades more readily divide them- evolution, and we must envision the pro- selves into separate species because interact- cess governing its course as being anal- ing individuals form interbreeding populations ogous to natural selection but operating that often split into geographically isolated at a higher level of organization. [We subgroups. Thus, sexual species are not more would say that it is natural selection, numerous because sex itself provides strong working at a level higher than the local adaptive advantages. Asexual species are just population.] In this higher-level process as successful and abundant by number of in- species become analogous to individuals, dividuals. Sexual species predominate simply and speciation replaces reproduction. The because they maintain a high capacity for random aspects of speciation take the speciation, while asexual clones do not. 2. Gradualism is not the only prior prej- udice constraining paleontological thought. A * We choose Wright's name for our designation because he (1967, p. 121) explicitly suggested that speciation might prominent place must also be awarded to our by truly stochastic with respect to the direction of evolu- tionani trends. But we wish to record our debt to Ernst Mavr propensity for explaining all questions of di- who has so forcefully and consistently supported the idea versity and success in terms of morphological that speciation is the stuff of evolutionary change. In fact Mayr ( 1963, p. 621) also analog~zed speciation with inuta: adaptation (Eldredge, in press). Consider, tion in the following remarkable passagel for example, the allied opposites of "over- I feel that it is the very process of creating so many spe- cies which leads to evolutionary progress. Species, in specialization" and "Cope's law of the unspe- the sense of evolution are quite com arable to mu- tations. . . . It appears) then that a prozigious multipli- cialized." Extinction, it is said, inevitably cation of species -l's a prerequisite f6 evdutionary pro- overtakes overspecialized taxa because they gress. . . . Speciation, the production of new gene complexes capable of ecological shifts is the method by are narrowly committed, through elaborate which evolution advances. withoui speciation there would be no diversification of the organic world, no and complex morphology, to a very small adaptive radiation, and very little evolutionary progress. The species, then. is the keystone of evolution. range- of environments. These taxa are almost We also note that Mayr, with his notion of the "genetic invariably large in body size, and owe their revolution" (1963, pp. 538-541), is the architect of a punc- iuational view of speciation within modern Neo-Darwinism, mol~hO1ogical at least in part, PUNCTUATED EQULLIBRIA RECONSIDERED 141 to allometric intensification (Rensch 1960). duces to the single phenomenon of ordinary Small bodied, morphologically unspecialized speciation, if general arise by spe- taxa. on the other hand. serve as the source cies selection. The Lmecondition for "general adaptation need not be many individuals in for almost all major groups because they re- A tain the morphological flexibility to change a large population; many events of speciation in many directions. The basic observations within a clade should serve just as well. Evi- are probably true: large, complex forms are dence is, to say the least, spotty: but nothing ultimately doomed, while small, simple species in the fossil record can enCOUraBe Darling- often proliferate. But we suspect that relative ton's belief that major morphologcal innovi- ability to speciate, not morphological flexibil- tions occur in large, stable, and widespread ity, provides an interpretive key, heretofore populations. unrecognized. Large animals live in small Darlington's paper illustrates a belief al- populations and do not fraction easily into most universally held among adherents to the isolated subgroups. (Some exceptions to this modern synthesis (Rensch 1960; Huxley 1958, generality appear among large mammals with for example)-that evolution proceeds in two complex social systems involving fractionation basic modes: anagenesis, or progressive evo- into family and kin groupings. Bush et al. lution ("improvements," or general adapta- [in press] explain high rates of speciation tions by phyletic transformation); and clado- among horses in this manner.) Small animals genesis, or diversification by splitting. Ayala maintain populations large enough to weather (1976, p. la), for example, writes: severe density-independent mortality, while Anagenesis, or phyletic evolution, con- their limited mobility and coarse-grained per- sists of changes occurring within a given ception of the environment permit an easier separation into isolated subgroups. All spe- phyletic lineage as time proceeds. The stupendous changes from a primitive form cies, the phenomenally successful as well as of life some 3 billion years ago to man, or the narrowly committed, will succumb in due some other modern form of life, are ana- time unless they leave descendants via speci- genetic evolution. Cladogenesis occurs ation. In wright7sgrand analogy, speciation = when a phylogenetic lineage splits into variability; clades and populations are equally two or more independently evolving line- doomed without variability. ages. The great diversity of the living 3. Darlington (1976) would separate mac- world is the result of cladogenetic evolu- roevolution into small-population events for tion. the rapid production of many species and large-population events for the development We find intellectually satisfying (though not of broad and general adaptations. Adaptive necessarily true on that account) the reduction radiations beginu with the evolution of key of these supposed processes to the single phe- adaptations in large populations, followed b$ nomenon of speciation-for, in our model, their general deployment through speciation anagenesis is only accumulated cladogenesis in small ones. Darlington writes: "The two filtered through the directing force of species parts of evolutionary explosions-the primary selection. adaptations and the secondary radiations- - 4. The speciation theory may help us to should be clearly distinguished. The first part predict differences in evolutionary tempos is probably mainly a large-population process among groups. A higher frequency of gradu- (p. 1363). Large groups evolve more effec- alism seems to characterize the predominantly tively than small ones at the same level" asexual protists (though still not very high- (p. 1362-because "they can put together se- see statements of Reyment and MacGillavry, lectively advantageous events that occur sepa- pp. 136). Gradualists would attribute this to rately, -and they can pay relatively heavy, a better stratigraphic record and would antici- complex costs of adaptation"). All this makes pate its occurrence across all groups. But the excellent sense under Darlington's "inarticu- best examples of protistan gradualism (e.g., lated major premise" (hidden assumption) of Ozawa 1975) do not come from continuous phyletic gradualism. If evolutionary events oceanic cores, but from traditional continental must occur within populations, then we cannot sequences containing, in abundance, other taxa deny his arguments. But his dual scheme re- that do not display gradualism. GOULD & ELDREDGE

FIGURE9. The resolution of apparent gradualism in asexual lineages to punctuational events at the appropri- ate level of clones. ( 1.)A trend produced by species selection in a sexually reproducing species, a is an an- cestor; d, a descendant. The trend arises only because species with favored traits live longer and branch more often than others. Note stasis within each species and equal number of branching events in each di- rection. (2.)Apparent gradualism in the branching of a descendant species from its ancestor in asexual or- ganisms; asexual "species" are assemblages of independent clones; these clones, rather than the taxonomist's artificial "species," should be the unit of macroevolution in asexual forms. (3.) Magnification of the ap- parently gradualistic segment of (2) to show individual clones. These clones follow the punctuational pat- tern of stasis within and rapid (in this case truly sudden, by definition) change between units. The trend is a result of clone selection; clones with favored traits live longer and branch more frequently. Note stasis within each clone, and equal number of branching events in each direction.

We predict more gradualism in asexual trend produced by punctuated equilibria and forms on biological grounds. Their history species selection. should be, in terms of their own unit, as punc- We suggest that the "speciation theory of tuational as the history of sexual Metazoa. But macroevolution" be explored by working their unit is a clone, not a species. Their evo- through the details of Wright's grand analogy lutionary mode is probably intermediate be- and considering the consequences. Stanley tween natural selection in populations and (1975, p. 649) has begun this process, and we species selection in clades: variability arises extend it here (Tables 3 and 4). No strategy via new clones produced rapidly (in this case, is as slippery and dangerous as analogy. (It truly suddenly) by mutation. The phenotypic may also be true that none is more rewarding; distribution of these new clones may be ran- Darwin, after all, constructed the theory of dom with respect to selection within an asex- natural selection as an analog to artificial se- ual lineage (usually termed a "species," but lection by breeders.) Many phenomena at one not truly analogous with sexual species com- level have no legitimate correspondent at posed of interacting individuals). Evolution another; in these respects, the levels are funda- proceeds by selecting subsets within the group mentally different. Recombination, the inter- of competing clones. If we could enter the action of individuals within populations, has protists' world, we would view this process of no common analog at other levels; clones may "clone selection" as punctuational. But we interact through occasional sexuality and spe- study their evolution from our own biased per- cies (at least in some clades ) by hybridization, spective of species, and see their gradualism but these are exceptional rather than charac- as truly phyletic (Fig. 9)-while it is really teristic processes. Similarly, the programmed the clonal analog of a gradual evolutionary course of individual has no analog PUNCTUATED EQULLLBRIA RECONSIDERED 143

TABLE3. Three distinct levels of evolution in the light of Wright's analogy. Starred items do not translate well across levels. level in sexual species asexual "species" clade organization ( individuals sharing ( clones sharing (species sharing a basic genome and dis- basic genome and pro- basic adaptive plan) tributing variation gram of regulation; by recombination) variation not well distributed by inter- action among units ) individual unit individual clone species of the level *individuals recombination (limited recombina- ( limited interact by tion in occasional hybridization ) sexual events ) *life course programmed ontogeny stasis by defini- no inherent of an individual tion till extinction direction; stasis or (a new mutation fluctuation probable; forms a new clade) gradual change very rare source of new mutation formation of new speciation variability clones by mutation relation of random by chemical random by chemical random if Wright's variability to the nature of mutation nature of mutation rule holds direction of evolution mechanism for individual selection clone selection species differential ( conventional selection increase of natural selection) favorable variation in the history of clones and species; in fact, ary analog of speciation rate be included in the overthrow of this analogy was a major ac- our study of successful clades (Table 3). A complishment of evolutionary paleontology- macroevolutionary analog to an r-strategist discrediting the theory of racial life cycles. might be a clade that consistently produces We believe, however, that the insight far many species-let us call these the increasers; outweighs the danger; the analogy is, at least, an analog to a K-strategist might be a clade a most fertile field for speculation, ten percent that produces species resistant to extinction, of which might be fruitful. Try, for example, either because they compete well against to translate the theory of r and K selection others, or because they survive diversity-in- from ecological to evolutionary time. Many dependent mass -let us call these paleontologists have tried to apply this theory, the survivors. We recognize that the "sur- without great success, especially to periods vivors" include species that are both r and K of mass extinction in the history of life; but strategists in ecological time. We are not they have not translated it properly through bothered that a macroevolutionary analog of the analogy. K selection might be a clade that contains r- Consider successful clades that are both di- selected species, for we have emphasized the verse and long-lived. Traditional thought fallacy of direct extrapolation between levels. would attribute their abundance and persis- tence to good morphological design, fashioned In this case, the key notion to translate through and tested in competition against species of the analogy is rapid production of individuals other clades. But just as life history param- (species) vs. production of good survivors. eters of maturation time and reproductive ef- If this analogy to r and K selection fails, it fort have been used to explain "success" in will fall because increaser and survivor clades ecological time, so must the macroevolution- may not exhibit the trade-offs so essential to 144 GOULD 8.z ELDREDGE

TABLE4. Determinants of evolutionary success in species and clades.

Species Clades criterion differential reproductive differential survival of success of certain genotypes certain species enhancer of abundant genetic variability many peripheral isolates success (or incipient species by another model of speciation) strategies for success rapid high 1. high speciation rate rates of ( increaser clades ) increase by high fecundity consistently high speciation rate by early maturation preemption of adaptive zone by abundant speciation during early history of a clade differential differential survival of resistance to extinction persistence favored genotypes ( survivor clades ) abundance: high frequency large populations to resist of a gene-to resist diversity independent density independent mortality extinction flexibility : "all purpose" large environmental range genes conferring pl~ysiological (law of the unspecialized) or molphological plasticity competitive superiority: triumph over other success in direct, intra- species in direct specific competition competition ( survivors ( K selected genotypes ) by competition ) the r-K distinction; i.e, the species of increaser newly-risen land, or a relatively empty world clades may also be good survivors. decimated by faunal catastrophe. The in- In any case, we make this distinction to creasers may win simply by being first, not em~hasize& that traditional arguments" of mor- by being better in some Newtonian, mechan- phological advantage apply only to one sub- ical sense; having been first, they may be category of survivor clades-survivors by com- able to hold on almost indefinitely, unless petition (Table 4). (We assume that survivors decimated themselves by a major diversity- by escape from diversity-independent mass independent episode of extinction. As R. Bak- extinction do not owe their rescue to generally ker (personal communication) argues, a prin- superior design but to eurytopy or to life in ciple of incumbency seems to hold for both surviving habitats, etc.) Clades that achieve the history of life and Western politics: it is their success through gradual increase in mem- very difficult for a better candidate to dis- bership during geologically untroubled times lodge an entrenched occupant; an aspirant may indeed owe their good fortune to evolu- sits tight and awaits its opponent's demise- tionary trends in characters with clear func- viz, Mesozoic mammals. tional advantage. Increase in average mam- The virtual irrelevancy, in many cases, of malian brain size throughout the Tertiary morphological superiority to a clade's success provides a good example (Jerison 1973). may largely explain the puzzling observation But we do not believe that most successful that so few stories of increasing perfection in clades follow this route: thev are more likelv design can be read from the history of life. 2 , to be the progeny of parental species that man- Our own initial surprise at this conclusion aged to take advantage of rare opportunities matches the insight that E. 0. Wilson had occasionally provided when the earth's general (personal communication) when he set out to equilibria are broken: unfilled ecospace around define biogeographical "success" in the con. PUNCTUATED EQUILIBRIA RECONSIDERED 145 ventional terms of morphology and finally con- ualism-ignoring Huxley's advice that he did cluded that abundant, expanding, and wide- not need it to support the theory of natural se- spread species are simply the r-selected, good lection-he translated Victorian society into colonists-not the "better" s~eciesin anv usual where it need not reside. As his A. sense of the term. The subset of increaser astute biographer W. Irvine remarks (1959, clades among survivors of diversity-indepen- p. 98): dent mass extinction may be the "supertramps" Darwin's matter was as English as his of macroevolution (Diamond 1975). method. Terrestrial history turned out to History, as Hegel said, moves upward in a be strangely like Victorian history writ spiral of negations. We needed Simpson's large. . . . The economic conceptions [of "continuationism" between micro and macro- laissez-faire liberalism] . . . can all be evolution to rid paleontology of accumulated paralleled in the Origin of Species. But so, nonsense and to place it firmly among the alas, can some of the doctrines of English evolutionary sciences. Secure in this status, political conservatism. In revealing the paleontologists now need to emphasize that importance of time and the hereditary the higher-level study of long times and large past, in emphasizing the persistence of clades requires a separate apparatus of evolu- vestigial structure, the minuteness of vari- tionary theory not fully available in the study ations and the slowness of evolution, Dar- of living organisms. We believe that the need win was adding Hooker and Burke [fa- to translate micro to macroevolution through- mous English conservatives] to Bentham the level of speciation guarantees that paleo- and Adam Smith [equally famous lib- biology shall not be a derivative field, but erals]. The constitution of the universe shall provide essential theory to any complete exhibited many of the virtues of the En- science of evolution. glish constitution. , who admired Darwin greatly and V. Towards a General Philosophy once stated that the Origin contained "the of Change basis in for all our views," Punctuated equilibria is a model for discon- made the same point in a famous letter to tinuous tempos of change at one biological Engels ( 1862): level only: the process of speciation and the It is remarkable how Darwin recognizes deployment of species in geological time. among beasts and plants his English so- Nonetheless, we believe that a general theory ciety with its division of labor, competi- of punctuational change is broadly, though tion, opening up of new markets, 'inven- by no means exclusively, valid throughout tion,' and the Malthusian 'struggle for biology. existence.' It is Hobbes' 'bellum omnium The general preference that so many of us contra omnes,' [war of all against all] and hold for gradualism is a metaphysical stance one is reminded of Hegel's Phenomenol- embedded in the modern history of Western ogy, where civil society is described as a cultures: it is not a high-order empirical ob- 'spiritual kingdom,' while in Dar- servation, induced from the objective study win the animal kingdom figures as civil of nature. The famous statement attributed society. to Linnaeus-nutura non facit saltum (nature does not make leaps) may reflect some bio- We mention this not to discredit Darwin in logical knowledge, but it also represents the any way, but merely to point out that even the translation into biology of the order, harmony greatest scientific achievements are rooted in and continuity that European rulers hoped to their cultural contexts-and to argue that gradualism was part of the cultural context, maintain in a society already assaulted by calls not of nature. for fundamental social change (see A. 0. Alternate conceptions of change have re- Lovejoy's [I9361 classic, The Great Chain of spectable pedigrees in philosophy. Hegel's Being, for an extended discussion on gradu- dialectical laws, translated into a materialist alism as a metaphysic traced back to context, have become the official "state phi- and beyond). losophy" of many socialist nations. These laws When Darwin cleaved so strongly to grad- of change are explicitly punctuational, as be- 146 COULD 8~ ELDREDGE fits a theory of revolutionary transformation tive feedback-provides a goldrnine of meta- in human society. One law, particularly em- phor for advocates of punctuationism. phasized by Engels, holds that a new quality We emphatically do not assert the "truth emerges in a leap as the slow accumulation of of this alternate metaphysic of punctuational quantitative changes, long resisted by a stable change. Any attempt to support the exclusive system, finally forces it rapidly from one state validity of such a monistic, a priori, grandiose to another (law of the transformation of quan- notion would verge on the nonsensical. We tity into quality). Heat water slowly and it believe that "gradual change" characterizes eventually transforms to steam; oppress the pro- some hierarchical levels, even though we may 1etaria.t more and more, and guarantee the attribute it to punctuation at a lower level- revolution. The official Soviet handbook of the macroevolutionary trend produced by spe- -Leninism (anonymous, undated) pro- cies selection. for exam~le.We make a simpleA claims: plea for pluralism in guiding philosophies- The transition of a thing, through the ac- and for the basic recognition that such philos- ophies, however hidden and inaTtictllated, do cumulation of auantitativeL modifications. from one qualitative state to a different, constrain all our thought. new state, is a leap in development. . . . Nonetheless, we doUbelieve that the punc- It is the transition to a new quality and tuational metaphysic may prove to map tem- pos of change in our world better and more signalizes a sharp turn, a radical change in development. . . . We often describe often than any of its competitors-if only be- modern Darwinism as a theory of the evo- cause systems in steady state are not only lution of the organic world, implying that common but also so highly resistant to change. We believe that the punctuational view will this evolution covers both aualitativeA and quantitative changes. Leap-like qualita- become important in at tive changes in social life are designated levels both lower and higher than the events by the concept of revolution. . . . The of speciation covered in our model of punc- evolutionarv develo~mentof societv is in- tuated equilibria. Dodson (1975, 1976) has evitably coAsummated by leap-lik; quali- made a promising start, partly in explicit anal- tative transformation, by revolutions ogy with our model, towards the application (anon., pp. 88-89). of Thorn's "catastrophe theory" (a punctua- tional topology) to change within populations. It is easy to see the explicit ideology lurking At the next level of speciation, Carson (1975) behind this general statement about the na- has proposed a model that is punctuational ture of change. May we not also discern the even in ecological time. He argues that sexual, implicit ideology in our Western preference for diploid organisms possess two differing sys- gradualism? tems of genetic variability: 1) An "open" sys- In the light of this official philosophy, it tem of freely-recombining, polymorphic loci is not at all surprising that a punctuational responsible for gradual, adaptive change view of speciation, much like our own, but within populations-e.g. clinal and subspe- devoid (so far as we can tell) of reference cific variability. 2) A "closed" system of co- to synthetic evolutionary theory and the allo- adapted, internally balanced gene complexes patric model, has long been favored by many that cannot vary without drastic effects on the Russian paleontologists (Ruzhentsev 1964; ontogenetic program of a species; thus, the Ovcharenko 1969). It may also not be irrele- "closed" system varies between, but not vant to our personal preferences that one of within, species. Gradualistic models work us learned his Marxism, literally at his daddy's only for the adaptations that arise within pop- knee. ulations by a fine-tuning of organism to en- The punctuational view is also congenial vironment; for these adaptations reflect with some important trends of Western change in the open system. But these gradual thought during the twentieth century. In- changes cannot be extrapolated to a model for formation theory, with its jargon of equilib- speciation, because the origin of new taxa in- rium, steady state, and homeostasis maintained volves rapid, drastic reorganization of the by negative feedback-not to mention the ex- closed system. Speciation, the source of mac- tremely rapid transitions that occur with posi- roevolutionary variation, is qualitatively dif- PUNCTUATED EQUILIBRIA RECONSIDERED 147 ferent from local adaptation within popula- of new Bauplane. We believe that a coherent, tions. It is also punch~ationalin ecological punctuational theory, fully consistent with time. Carson writes (1975, pp. 87-88) : Darwinism (though not with Darwin's own unnecessary preference for gradualism), will The classical view of speciation holds that be forged from a study of the genetics of reg- it is a gradual microevolutionary process. ulation, supported by the resurrection of long- Thus, the genetic events which lead to- neglected data on the relationship between ward speciation are considered to be in- ontogeny and phylogeny (see Gould 1977). dividually trivial or simple. Changes, Ager (1973, p. loo), whose small book ad- such as a shift in gene frequency, may vocates a punctuational view of the strati- accumulate slowly in a population. This graphic record, speaks in simile of the tempo view does not invoke any unusual or that we support as most characteristic of the quantum set of changes for the origin of way our world works: "The history of any interspecific differences. Most theories . . . one part of the earth, like the life of a soldier, of speciation are thus wedded to gradu- consists of long periods of boredom and short alism, using the mode of origin of intra- periods of terror." s~ecificadaptations as a model. . . . I would neveptheless like to propose that the following possibility be seriously con- VI. Suggestions for a Program sidered. Speciational events may be set of Research in motion and important genetic salta- Cosmic philosophical waffling aside, our tions towards species formation accom- highest hope for punctuated equilibria is that plished by a series of catastrophic, stochas- it might guide a change in paleobiological tic genetic events. . . . Speciation is practice. We believe that the agenda of paleo- considered to be initiated when an un- biology should accord a prominent place to usual forced reorganization of the epi- establishing the unbiased empirical distribu- static supergenes of the closed variability tion of evolutionary tempos; our model makes system occurs. . . . I propose that this this possible by tabulating as data several cycle of disorganization and reorganiza- important phenomena previously ignored. We tion be viewed as the essence of the speci- suggest the following as a protocol for the ation process. Under most circumstances, further testing and utility of punctuated equi- it seems unlikely that gradual microevolu- libria: tionary changes could easily accomplish 1. In testing for gradualism vs. punctua- the drastic changes in the closed system tion in individual taxa, study the geographic which seem to be required. variability of species over their entire pre- Carson argues that these reorganizations of served range. Do not confine a study to local the closed system may occur very rapidly sections or single cores. when natural selection is relaxed during a 2. Study the distribution of evolutionary population flush-crash-founder cycle. tempos for all members of an ecosystem or At the higher level of evolutionary transi- community. We hope that the emphasis in tion between basic morphological designs, testing our model will shift away from ab- gradualism has always been in trouble, though stracted individual cases towards the unbiased it remains the "official" position of most West- assessment of entire faunas; for, as we have ern evolutionists. Smooth intermediates be- emphasized throughout this paper, the essen- tween Bauplane are almost impossible to con- tial question is one of rehtiue frequency. We struct, even in thought experiments; there is hope that all students engaged in such work certainly no evidence for them in the fossil will keep our favorite motto before them: record (curious mosaics like Archaeopteryx do stasis is data. As MacGillavry wrote (1968, not count). Even so convinced a gradualist p. 70): "What, to my knowledge, is com- as G. G. Simpson (1944) invoked quantum pletely lacking, is a quantitative study of the evolution and inadaptive phases to explain entire fauna of such successions. A study of these transitions. Recently, Lovtrup (1974) this kind should pay attention to the percent- and Frazzetta (1975) have written books to age of forms which do not show any evolu- support a punctuational theory for the origin tionary change." 148 GOULD ti ELDREDGE

Of our two most important predictions for rections of speciation within a clade random local sections-"sudden" transition from an- with respect to the direction of evolutionary cestor to descendant and stasis within individ- trends? They need not be: it is quite pos- ual taxa-we regard stasis as more amenable sible, for example, that most species will arise to study. Morphological breaks can usually at sizes larger than their ancestors within a be attributed equally well to punctuated equi- clade exhibiting Cope's Rule as a general libria or to an imperfect record. Rut stasis trend. In such a case, species selection would in a great majority of taxa-particularly for play no creative role, just as natural selection the characters used to define species-speaks would not direct evolution within popula- clearly for us, while a high frequency of di- tions if mutations occurred preferentially in rectional change would speak just as strongly adaptive directions. (Natural selection might against us. Moreover, the relative frequency still eliminate the unfit, but the central postu- for stasis can be tabulated with value even in late of Darwinian theory-that natural selec- local sections, for gradualism predicts that in tion creates the fit-would collapse.) If situ change will be directional. Wright's Rule fails in favor of a preferential 3. With a model that allows us to assess tendency for speciation toward the general relative frequency, a basic evolutionary ques- direction of a larger trend, then species selec- tion about phyletic gradualism can finally be tion is not needed because a random removal posed-where and how often? In what ecolog- of species would still yield the trend. ical situations does it occur? Is it unusually Wright's Rule also requires that speciation common in certain taxa? We have already be common in order to provide enough raw suggested that the frequency of gradualism material for species selection. We therefore might be higher in asexual species because predict that the left end of Van Valen's (1973) their "continuous" trends are produced by random extinction curves will actually be con- clone selection, punctuational at a lower level. cave rather than straight as he depicts them- These data are important for evolutionary leading to a marked increase of species with theory, and they could not be gathered when very short durations. (These are the easiest paleontologists regarded cases of gradualism species to miss in the fossil record, and the as the only legitimate illustrations of evolu- straightness of a raw, empirical line for a sam- tion. ple biased by geological constraints might im- 4. Study general patterns in the history of ply concavity in the true universe. This could diversity, whenever possible, at the species be tested by asking whether groups with level. The species, in our view, is a true, basic better records yield greater concavity.) Large and stable evolutionary unit, not merely an numbers of species with short durations would intermediate rank in a hierarchy from individ- supply a sufficient pool of variability for spe- ual to kingdom, defined as an arbitrary seg- cies selection. ment of a continuously changing lineage. It In bringing paleontology within the mod- is ironic that so much paleontological activity ern synthesis, Simpson emphasized evolution- has been devoted to the description of species, ary rates and strongly advocated their further but that virtually all interpretive studies con- study. Yet rather little has come of his plea, centrate on generic and higher rank. The rea- primarily, we believe, because underlying as- son for this has been expediency, not desire. sumptions of gradualism yielded much in the Many evolutionists have lamented the spotti- way of discouragement and rather little in ness and inconsistency of species-level taxon- concrete suggestions for research. Since the omy in the fossil record, and have studied 19407s, paleontology has undertaken several genera and families faute de mieux. We agree long excursions in other directions-partic- that students of-macroevolution cannot merely ularly into the functional morphology of in- tabulate species from the primary literature. dividual taxa and the reconstruction of fossil But Stanley (1975a, 1976) has devised some communities. These studies, at their best, ingenious methods for reliable inferences have been elegant and persuasive; they have about the origination, duration, and extinc- certainly been of great value. But they have tion of species in larger clades. not produced new evolutionary theory; they 5. Test Wright's Rule as a precondition for have not asserted the theoretical independence species selection. Are the morphological di- of paleobiology; they have, in essence, shown I'UNCTUATED EQUILIBRIA RECONSIDERED 149 that fossil organisms and assemblages work effectiveness of evolution in multi-level situations. much as modern ones; they have, in this sense, Proc. Natl. Acad. Sci. 73: 136C-1364. DIAMOND,J. 1975. Assembly of species communi- continued the tradition of paleobiology as a ties. pp. 342444. In: Cody, M. L. and J. M. Dia- derivative science, indebted for all its insights mond, eds. and Evolution of Communities. to the evolutionary biology of modern organ- Harvard Univ. Press; Cambridge, Mass. isms and ready to provide nothing but simple DODSON,M. M. 1975. and the confirmation in return. But whv be a aleo on- fold catastrophe. Evol. Theory. 1:107-118. A DODSON,M. M. 1976. Darwin's law of natural se- tologist if we are condemned only to verify lection and Thom's theory of catastrophes. Math. imperfectly what students of living organisms Biosci. 28:243-274. can propose directly? An evolutionary theorist DROOGER,C. W. 1963. Evolutionary trends in the should pursue paleobiology only if he be- Miogypsinidae. pp. 315-349. In: von Koenigswald, G. H. R., J. D. Emeis, W. L. Buning, and C. W. lieves that the direct study of geological time Wagner, eds. Evolutionary Trends in Foraminifera. can yield new insights to enlarge a theory Elsevier; Amsterdam. constructed for events in ecological time. We DURHAM,J. W. 1971. The fossil record and the believe that paleobiology can do this. Let us origin of the Deuterostomata. North Am. Paleontol. now heed Simgson's recommendation of thirtv Conv., Chicago, 1969, Proc., H: 1104-1132. ELDREDGE,N. 1971. The allopatric model and years past. ~6estudy of evolutionary temp& phylogeny in Paleozoic invertebrates. Evolution. lies in the exclusive domain of paleobiology. 25: 156-167. It is, we believe, our most promising arena for ELDREDGE,N. 1974. Stability, diversity, and speci- the derivation of indegendent macroevolution- ation in Paleozoic epeiric seas. J. Paleontol. 48: 540-548. ary theories. Conceptual tools are now avail- ELDREDGE,N. 1974. Testing evolutionary hypoth- able for a fruitful study of evolutionary tempos eses in paleontology: a comment on Makurath and and modes. Anderson ( 1973). Evolution. 28:478481. ELDREDGE,N. 1975. Character displacement in evolutionaiy time. Am. Zool. 14: 1083-97. Literature Cited ELDREDGE,N. AND S. J. GOULD. 1972. Punctuated AGER,D. V. 1973. The Nature of the Stratigraph- equilibria: an alternative to phyletic gradualism. ical Record. 114 pp. John Wiley; New York. pp. 82-115. In: Schopf, T. J. M., ed. Models in AGER,D. V. 1976. The nature of the fossil record. Paleobiology. Freeman, Cooper and Co.; San Fran- Proc. Geol. Assoc. 87:131-159. cisco, Calif. ANONYMOUS.Fundamentals of Marxism-Leninism ELDREDGE,N. AND S. J. GOULD. 1974. Reply to Manual. 891 pp. Foreign Languages Publishing Hecht ( 1974). Evol. Biol. 7:303-308. House; Moscow. ELDREDGE,N. AND S. J. GOULD. In press. Evolu- AVISE, J. C. 1976. Genetic differentiation during tionary models and biostratigraphic strategies. pp. speciation. pp. 106-122. In: Ayala, F. J., ed. 2440. In: Kauffman, E. G. and J. Hazel, eds. Con- Molecular Evolution. Sinauer Associates; Sunder- cepts and Methods of Biostratigraphy. Dowden, land, Mass. Hutchinson and Ross; Stroudsburg, Pa. ELDREDGE,N. AXD I. TATTERSALL.1975. Evolu- AVISE,J. C. AND F. J. AYALA. 1976. Genetic dif- ferentiation in speciose versus depauperate phylads: tionary models, phylogenetic reconstruction, and evidence from the California minnows. Evolution. another look at hominid phylogeny. In: Szalay, 30:46-58. F. S., ed. Approaches to Primate Paleobiology. AYALA,F. J. 1976. Molecular genetics and evolu- Contrib. Primatol. 5:218-242. tion. pp. 1-20. In: Ayala, F. J., ed. Molecular FERRIS,S. D. LYD G. S. WHITT. 1975. Gene dupli- Evolution. Sinauer Associates; Sunderland, Mass. cation and functional diploidization in the Cata- ) AYALA,F. J., M. L. TRACEY,D. HEDGECOCK,AND stomidae. Genetics. 80(suppl. :30. R. C. RICHMOND. 1974. Genetic differentiation FORTEY,R. A. 1974. The Ordovician trilobites of during the speciation process in D~osophila. Evolu- Spitsbergen I. . Skrift. Norsk Polarinst. tion. 28:576-592. No. 160. 129 pp. BETTENSTAEDT,F. 1962. Evofutionsvorgiinge bei FRAZZETTA,T. H. 1975. Complex Adaptations in fossilen Foraminiferen. Mitt. Geol. Staatsinst. Evolving Populations. 267 pp. Sinauer Associates; Hamburg. 31:385-460. Sunderland, Mass. BUSH,G. L. 1969. Host race foimation and speci- GINGERICH,P. D. 1974. Stratigraphic record of ation in frugivorous flies of the genus Rhagoletis early Eocene Hyopsodus and the geometry of mam- ( Diptera: Tephritidae ). Evolution. 23:237-251. malian phylogeny. Nature. 248: 107-109. BUSH,G. L., S. M. CASE,A. C. WILSON,AND J. L. GINGERICH,P. D. 1976. Paleontology and phylog- PATTON. In press. Rapid speciation and chromo- eny: patterns of evolution at the species level in somal evolution in mammals. Proc. Natl. Acad. early Tertiary mammals. Am. J. Sci. 276:l-28. Sci. GINGERICH,P. D. In press. Patterns of evolution CARSON,H. L. 1975. The genetics of speciation at in the mammalian fossil record. In: Hallam, A., the diploid level. Am. Nat. 109:83-92. ed. Patteins of Evolution. Elsevier; Amster- DARLINGTOX,P. J., JR. 1976. Rates, patterns, and dam. 150 GOULD & ELDREDGE

COULD, S. J. 1972. Allometric fallacies and the at Lone Mountain, Nevada. Geol. et Paleontol. evolution of Gyphaea. Evol. Biol. 6:91-119. 9:65-83. GOULD,S. J. 1976. Ladders, bushes, and human KUHIV,T. S. 1962. The Structure of Scientific Revo- evolution. Nat. Hist. 85(No. 4) :24-31. lutions. 172 pp. Univ. Chicago Press; Chicago, Ill. GOULD, S. J. 1977. Ontogeny and Phylogeny. LANDE, R. 1976. Natural selection and random Harvard Univ. Press; Cambridge, Mass. In Press. genetic drift in phenotypic evolution. Evolution. GOULD,S. J. 1977. Eternal metaphors of paleon- 30:314-334. tology. In: Hallam, A., ed. Patterns of Evolution. LEAKEY,R. E. F. AND A. WALKER.1976. Australopi- Elsevier; Amsterdam. In Press. thecus, Homo erectus, and the single species hy- HANSON,N. R. 1969. Perception and Discovery. pothesis. Nature. 261 3572-574. 435 pp. Freeman, Cooper and Co.; San Francisco, LEWONTIN,R. C. 1974. The Genetic Basis of Evo- Calif. lutionary Change. 346 pp. Columbia Univ. Press; HALLAM, A. 1968. Morphology, palaeoecology, New York. and evolution of the genus Gyphaea in the British LEWONTIN,R. C. AND J. L. HUBBY. 1966. A molec- Lias. Phil. Trans. R. Soc. London B. 254:91-128. ular approach to the study of genic heterozygosity HARPER, C. W., JR. 1975. Origin of species in in natural populations. 11. Amount of variation geologic time: alternatives to the Eldredge-Gould and degree of heterozygosity in natural popula- model. Science. 190:4748. tions of D~osoplzila pseudoobscura. Genetics. 54: HAYAMI,I. 1973. Discontinuous variation in an 595-609. evolutionary species, Cryptopecten vesiculosus, from LESP~RANCE,P. J. AND R. BERTRAND.1976. Pop- Japan. J. Paleontol. 47:401420. ulation systematics of the Middle and Upper Ordo- HAYAMI, I. AND T. OZAWA. 1975. Evolutionary vician trilobite Cyptolithus from the St. Lawrence models of lineage-zones. Lethaia. 8: 1-14. lowlands and adjacent areas of Quebec. J. Pale- HECHT,M. K. 1974. Morphological transformation, ontol. 50:598-613. the fossil record, and the mechanisms of evolution. LOVEJOY,A. 0. 1936. The Great Chain of Being. Evol. Biol. 7:295-303. 376 pp. Harvard Univ. Press; Cambridge, Mass. HENRY,J. L. AND E. N. K. CLARKSON.1975. En- LOVTRUP,S. 1974. Epigenetics. 547 pp. John Wiley; rollment and coaptations in some species of the New York. Ordovician trilobite genus Placoparia. Fossils and MACGILLAVRY,H. J. 1968. Modes of evolution Strata. 4:87-95. mainly among . Bijdragen Tot HOWELLS,W. W. In press. Homo erectus in hu- De Dierkunde. 38: 69-74. man descent: ideas and problems. Can. Assoc. MAKURATH,J. H. 1974. Evolution of Appalachian Phys. Anthrop. Symp. gypidulid brachiopods: A reply to Eldredge HURST,J. M. 1975. Resserelln sabrinae Bassett, in ( 1974). Evolution. 28:481-483. the Wenlock of Wales and the Welsh Borderland. MAKURATH,J. H. AND E. J. ANDERSON.1973. Intra- J. Paleontol. 49:316-328. and interspecies variation in gypidulid brachio- HUXLEY,J. 1958. Evolutionary processes and tax- pods. Evolution. 27:303-310. onomy, with special reference to grades. pp. 21- NEI, M. 1975. Molecular Population Genetics and 38. Uppsala Univ. Arsskr. Evolution. 253 pp. American Elsevier; New York. IRVINE, W. 1959. Apes, Angels, and Victorians. OVCHARENKO,V. N. 1969. Transitional forms and 399 pp. Meridian Books; New York. species differentiation of brachiopods. Paleontol. JERISON,H. J. 1973. The Evolution of the Brain J. 3: 57-63. and Intelligence. 482 pp. Academic Press; New OXNARD,C. E. 1975. Uniqueness and Diversity in York. Human Evolution: Morphometric Studies of Aus- JOHNSON,J. G. 1975. Allopatric speciation in fos- tralopithecine~. 133 pp. Univ. Chicago Press; sil brachiopods. J. Paleontol. 49:646-661. Chicago, Ill. JOHNSON,J. G. AND A. W. NORRIS. 1972. Tecno- OZA~A,T. 1975. Evolution of Lepidolina mzrlti- cyrtina, a new genus of Devonian brachiopods. septata (Permian foraminifer) in East Asia. Mem. J. Paleontol. 46:565-572. Fac. Sci. Kyushu Univ., Ser. D Geol. 23:117-164. JOHNSON,J. G., A. RESO, AND M. STEPHENS. 1969. PILBEAM,D. R. 1975. Middle Pleistocene hominids. Late Upper Devonian brachiopods from the West pp. 809-856. In: Butzer, K. W. and G. L. Isaac, Range Limestone of Nevada. J. Paleontol. 43: eds. After the Australopithecines. Mouton; The 1351-1368. Hague. KELLOGG,D. E. 1975. The role of phyletic change RAUP, D. M. In press. Probabilistic models in in the evolution of Pseudocubus vema (Radiolaria). evolutionary paleobiology. Am. Sci. Paleobiology. 1:359-370. RENSCH, B. 1960, Evolution Above the Species KELLOGG,D. E. AND J. D. HAYS. 1975. Microevolu- Level. 419 pp. Columbia Univ. Press; New York. tionary patterns in Late Cenozoic Radiolaria. RE~ENT,R. A. 1975. Analysis of a generic level Paleobiology. 1:150-160. transition in Cretaceous ammonites. Evolution. KING, M. C. AND A. C. WILSON. 1975. Evolution 28:665-676. at two levels in humans and chimpanzees. Science. ROBISON,R. A. 1975. Species diversity among 188:107-116. agnostoid trilobites. Fossils and Strata. 4:219-226. KI~s,D. B. 1974. Paleontology and evolutionary ROUGHGARDEN,J. 1973. Possibilities for paleontol- theory. Evolution. 28:45&472. ogy. Science. 179:1225. KLAPPER,G. AND D. B. JOHNSON. 1975. Sequence ROWE,A. W. 1899. An analysis of the genus Mi- in conodont genus Polygnathus in Lower Devonian craster, as determined by rigid zonal collecting from PUNCTUATED EQUlLLBRIA RECONSIDERED 151

the zone of Rhynchonella Cuuie~ito that of Mi- ical tests of evolutionary theory. pp. 41-63. In: craster COT-anguinum.Q. J. Geol. Soc. London. 55: Kauffman, E. G. and J. Hazel, eds. Concepts and 494-547. Methods of Biostratigraphy. Dowden, Hutchin- RUZHENTSEV,V. Y. 1964. The problem of transi- son and Ross; Stroudsburg, Pa. tion in paleontology. Int. Geol. Rev. 6:2204-2213. TRUEMEN,A. E. 1922. The use of Gryphaea in the SCHAEFFER,B. 1965. The role of experimentation correlation of the Lower Lias. Geol. Mag. 59:256- in the origin of higher levels of organization. Syst. 268. Zool. 14:318-336. VALENTINE,J. W. AND C. A. CAMPBELL.1975. Ge- SCHOPF,J. W. AND D. Z. OEHLER. 1976. How old netic regulation and the fossil record. Am. Sci. 63: are the eukaryotes. Science. 193:47-49. 673-680. SCOTT,G. H. 1974. Biometry of the foraminiferal VAN VALEN, L. 1973. A new evolutionary law. shell. pp. 55-151. In: Hedley, R. H. and C. G. Evol. Theory. 1:1-30. Adams, eds. Foraminifera. Academic Press; New WALKER, A. 1975. Splitting times among homi- York. noid! deduced from the fossil record. Burg Warten- SCOTT, G. H. 1976. Foraminifera1 biostratigraphy stein Symp. 65. Progress in Molecular Anthrop. and evolutionary models. Syst. Zool. 25:78-80. 23 PP. SI~~PSON,G. G. 1944. Tempo and Mode in Evolu- WILSON,A. C. 1976. Gene regulation in evolution. tion. 237 pp. Columbia Univ. Press; New York. pp. 225-234. In: Ayala, F. J., ed. Molecular Evolu- SIMPSON,G. G. 1953. The Major Features of Evolu- tion. Sinauer Associates; Sunderlan, Mass. tion. 434 pp. Columbia Univ. Press; New York. WRIGHT, S. 1967. Comments on the preliminary SIMPSON,G. G. 1976. The compleat palaeontolo- working papers of Eden and Waddington. In: gist? Annu. Rev. Earth Plan. Sci. 4:l-13. Moorehead, P. S. and M. M. Kaplan, eds. Mathe- SPRINKLE,J. 1976. Classification and phylogeny of matical Challenges to the Neo-Darwinian Theory "pelmatozoan" echinoderms. Syst. Zool. 25:83-91. of Evolution. Wistar Inst. Symp. 5: 117-120. STANLEY,S. M. 1975a. A theory of evolution above ZIEGLER, A. M. 1966. The silurian brachiopod the species level. Proc. Natl. Acad. Sci. 72:646- Eocoelia hemisphae~ica (J. de C. Sowerby) and 650. related species. Palaeontology. 9 :523-543. STANLEY,S. M. 197513. Clades versus clones in evo- ZUCKERKANDL,E. 1968. Hemoglobins, Haeckel's lution: why we have sex. Science. 190:382-383. "Biogenetic Law," and molecular aspects of devel- STANLEY,S. M. 1976. Stability of species in geo- opment. pp. 256-274. In: Rich, A. and N. David- logic time. Science. 192:267-268. son, eds. Structural Chemistry and Molecular Biol- SYLVESTER-BRADLEY,P, C. In press. Biostratigraph- ogy. W. H. Freeman; San Francisco, Calif.