Phyl~enetic Patterns and the Evolutionary Process. Cracraft80... · Phylogenetic patterns and the...

I

I

~ ..

Phyl~eneticPatterns

and theEvolutionary

Process.Method and Theory inComparative Biology

Niles EldredgeJoel Cracraft

Columbia University Press

New York

j

Contents

Library of Congress Cataloging in Publication Data

Eldredge, Niles.

Phylogenetic patterns and the evolutionary process.

Includes bibliographical references and index.. 1'. Phylogeny. 2. Evolution. 3. Biology--Classi-

ficetlon I. Cracraft, Joel, joint author. II. TitleOH367.5.E38 575 80-375 .ISBN 0-23Hl3802-X (cloth)ISBN 0-231-Q8378-5 (paper)

Preface viiChapter 1 Introduction-Pattern and Progress in

Comparative Biology 1Pattern and Process: Logical Interconnections 4; The Study ofPhylogenetic Pattern 5; The Study of the Evolutionary Process 13;Integration of Pattern with Process: the Structure of This Book 16.

Chapter 2 Cladograms: Cladistic Hypotheses andTheir Analysis 19Cladograms: Some Introductory Principles 21; Cladistic Analysis:Similarity, Synapomorphy, and Homology 29; Cladistic Analysis:Taxa and Characters 41; Cladistic Analysis: Hypotheses and TheirEvaluation 50; Some Case Studies 74.

Chapter 3 Species: Their Nature and Recognition 87Species Defined 87; Other Definitions of Species 92;Recognition of Species 94; Species Recognition in Practice: SomeExamples 108.

Chapter 4 Modes of Speciation and the Analysis ofPhylogenetic Trees 113Patterns of Speciation 114;Construction and Testing of PhylogeneticTrees 127.

Chapter 5 Biological Classification 147Darwin, Natural Groups, and Classification 149; Classification andthe Dichotomy of A and Not-A Groups 158; Structure of the LinnaeanHierarchy 165; Information Content and Branching Diagrams 171,Branching Diagrams and Their Role in Classification 175; TheQuestion of Ranking: A Critique of Classical EvolutionaryClassification 191;Cladograms, Trees, and Classification 211,Conclusions 238.

Chapter 6 Systematics and the Evolutionary Process 241Relationship Between Evolutionary History and EvolutionaryProcess 242; Contemporary Evolutionary Theory: A Basic

Columbia University PressNew York GUildford, Surrey

Cop~right © 1980 Columbia University PressAll fights reserved:rinte~ in the United States of America

65432

,.,-.----~vi Contents

Characterization 244' C tIssues and r f ',on emporary Macroevolutionary Theory:Ph rans ormanonai Explanations 248' The

Sy,e,neomm,e,.nor092icallevels of Evolution and Thei; Relation toice 72' Rei r h!of Evolution'D 'r a Ions IpSAm?ng the Phenomenological levels

Phen '~oup Ing 283:RelatIonshipsAmongthe

Momenol.oglcal Levels of Evolution: Interconnections 293·

acroevolutlonary Th A A 'DeterministicCom eery: eslatement ~fthe Problem 296;Theand Inform" . oonent of a Macroevotunonary Theory 301 Theory

a Ion In Macroe I ,. . '~'PO'h vo U tonary AnalysIs 305' Macroevolution'•'I eses and Some B . p , .Evolutionary Proces . A asSIC attems 310; Systematics and Ihe

s. ummary 325.ReferencesIndex

.:at

Preface

331343

WE HAVE sought, in this book, to unite two themes too tre-quently disjoined in comparative biology: systematics. the orderingof the elements of the earth's biota, and evolutionary theory, the ex-planation of how that order arose and continues to arise. As such, it isnot a "cookbook" on how to do systematics, nor does it pretend to bean overview of evolutionary theory. Rather, we have tried to charac-terize the basic ontological and epistemological problems of sys-tematics and that portion of evolutionary theory relevant to system-atics.

We are convinced that there is no body of theory so arcane thatit cannot be discussed in simple terms. We have set out to examinethe principles of comparative biology-systematics and evolutionarytheory-as thoroughly and rigorously as possible. But we have en-deavored to keep the language simple. This book should be acces-sible to both the enquiring beginning student and to the thoughtfulprofessional. It is intended for everyone with a serious interest in thesubject. We hope it is accessible to all.

We are much indebted to many people who generously helpedus in our efforts to produce this book. The work seemed Sisypheanat times. Trying to write such a book at a time when comparativebiology has been in an uproar has been both exhilarating and ex-hausting. We are grateful to Dr. Gareth Nelson, of The AmericanMuseum of Natural History, initially our collaborator, for much of theinitial impetus and stimulus for the book. It was Nelson who firstclarified the distinction between cladograms and trees (in an earlymanuscript prepared for this book). The essence of that concept isretained here, and used, no doubt, for purposes far beyond its origi-nal intent. Dr. Stanley Satthe. of Brooklyn College, has similarly pro-vided fundamental stimulation regarding evolutionary theory, also

-----------,viii Preface

via an unpublished manuscript dealing with hierarchical structure inevolutionary theory, as well as through numerous personal converse-tons. We have also benefitted from countless conversations withfriends and colleagues, especially Drs. Eugene S. Gaffney, NormanI. Platnick, Bobb Schaeffer, Randall T. Schuh, and ran Tattersall, allof The American Museum of Natural History.

We are particularly grateful to those who helped directly withproduction of the manuscript. Incisive, helpful commentary on vari-ous drafts and portions was received from Drs. Stephen Jay Gould(Harvard University), David Hull (University of Wisconsin, Mil-waukee), Norman D. Newell (The American Museum of Natura! His-tory), David M. Raup (Field Museum of Natural History), and StanleySalthe (Brooklyn College). Ms. Nancy A. Neff (The AmericanMuseum of Natural History and City University of New York) spottedinnumerable inconsistencies and illogical formulations. Herthorough and painstaking critique of early drafts resulted in substan-tial improvement in many areas of the book. We also thank Mr. Bruc~Manion, Ms. Sharon Simpson, and Mr. Robert Schmitz, all of the Uni-versity of Illinois, for their Comments on parts of the manuscript.

We are glad to acknowledge as well the patience and guidanceof ~essrs. John Moore and Joe Ingram, the shepherding of the ma~u-scnptthrough reviewers by Dr. Vicki Raeburn, and through productionby Ms. Maria Caliandro, all of Columbia University Press. We aregrateful to Ms. E. Penny Pounder and Ms. Majorie She patin for pro-ducing most of the illustrations. Finally, we thank our secretaries, Ms.Cristina Ordonez (The American Museum of Natural History) and Ms.~Iaudette Lake (University of Illinois at the Medical Center), for ~~-Ing some of the drafts, and also Mr. Sidney Horenstein, ScientificA.ssistant at The American Museum, who so ably aided us in mattersbibliographic. To all, and to those unnamed, we are profoundlythankful

Phylogenetic Patternsand the Evolutionary Process

New York and ChicagoJanuary, 1980

1

Chapter

1Introduction-Pattern and Processin Comparative Biology

THIS BOOK is about the twin themes of pattern and process incomparative biology. By pattern we mean aspects of the apparentorderliness of life. By process we mean the mechanisms that gener-ate these patterns. The function of comparative biology is to analyzeand capture biotic patterns and to elaborate a theory of process toexplain pattern.

Comparative biology deals with several of the various sorts ofpatterns to be found in the biotic world, The "orderl iness of life," forsome, might suggest the ecological integration of individuals intopopulations, populations into communities, and on up through a hi-erarchy including provinces, realms, and finally, the entire biosphere(see Valentine 1973). A related pattern is simply the distribution oforganisms in space; analyses of such patterns fall under the dis-cipline of historical biogeography. Similarly, temporal distributionsconstitute the subject matter of biostratigraphy. Ecological and dis-tributional considerations thus lead to different general approachesto the very perception of patterns in the organic world.

The kind of pattern on which we focus our attention here residesin the intrinsic features of individual organisms. Intrinsic featuresrange from the atoms, molecules, and compounds composing an or-ganism, on up through cells, tissues, organs, organ systems, as wellas individual bits of behavior. Intrinsic features contrast with extrin-sic properties of organisms-their distributions in space and time.The type of orderliness exhibited by intrinsic features is expressed interms of relative degrees of similarity: all organisms share at least afew intrinsic properties in common (RNA. for example). Organismstend to share a great deal more properties with certain other crqa-nisms, and share relatively little with others. Thus "orderliness in na-

2 Pattern and Process

ture" in this book refers to the apparent hierarchy of similarity amongthe organisms comprising the biota.

To study this orderliness, we need a method, a set of proceduralrules, Why is a set of rules necessary? Can we not simply map outthis orderliness, which seems so obvious in a general sort of way?We take the position that the history of systematics-that branch ofbiology which seeks to capture the orderliness of nature-recordsprogress from simple perception of the pattern to a conscious under-standing of the nature of the order and the logical structure of theanalysis of that order. The basic problem with simply "perceiving"pattern is that there is no consensus at the outset about what is to becompared, or even what "similar" means. The very fact that orga-nisms can be viewed in a hierarchical fashion (atoms through organsystems) indicates a decided diversity in opinion about the units ofcomparison among organisms. And the question of what, exactly,constitutes "similarity" is perhaps the oldest one in systematics. Aswe shall develop, an explicit contribution to this question has beenmade as recently as a quarter-eentury ago (Hennig 1950). In ourview, this contribution is fundamental and effects any formulation ofa logical structure for pattern analysis in comparative biology.

Thus a fundamental cornerstone of comparative biology is thesimple assertion that there is order in the biological realm of na-ture-an order involving patterns of similarity among the constituentsof the earth's biota. Those who do not perceive such order, or whorefuse to adopt its existence as an axiom, will, of course, have littleinterest in the subject matter of this book.

How did this pattern of similarity of features originate? Are theregeneral mechanisms which operate in the natural world that arecapable of generating this pattern? There are two general sets of ex-planations for this pattern still with us in the twentieth century: evolu-tion and sepa~ate, or special, creation. Evolution asserts that the pat-tern of similarity by which all known organisms may be linked is thenatural outcome of some process of genealogy. In other words allorganisms are related. Just as human children resemble theirparents, organisms which are closely related tend to resemble eachother more closely than do their more remote relatives. The hypoth-esis that such a process of genealogical production of descendantsfrom ancestors has occurred is, of course, called "evolution." It is theonly generally accepted biological mechanism for the production of

Pattern and Process 3

this pattern, ever since the demise of the notion of spontaneous gen-eration long ago .

.Special creation, in its various guises, is not a scientific formula-tion, simply because it asserts that every discrete "type" of organismhas its own unique beginning unconnected with the history of anyother kind of organism. All organisms are created de novo. Undersuch an assumption, there are no generalizations to be found whichwould cover more than a single case. The most common form ofspecial creation today is associated with the religious concept of asupernatural power, or being, which has created the order we see inthe biotic world. Such formulations have always been difficult to in-vestigate experientially. Special creation is, properly speaking, a setof assumptions about the nature of the world that excludes such ex-periential considerations. Yet, if a notion is to be considered withinthe purview of science, it must be capable of evaluation through ourexperiences with the world. Those who adopt the assumptions andaxioms of special creation will not be inclined to adopt the viewspresented in this book.

The notion that life has evolved paves the way for a generaltheory of mechanisms to explain the genesis of life's orderliness.While the concept that life has evolved remains the only viable alter-native to special creation and allows the scientific search for generalmechanisms, evolutionary theory per se is actually in a healthy stateof disrepair. Unlike systematics, where methodologies for recon-structing the pattern of nature form the very basis of theoretical dis-cussion, evolutionary theorists have concentrated almost totally ongenerating and examining statements about how the process works,and have spent almost no effort in evaluating just how the process ofevolution ought to be studied scientifically. We address both epis-temological and ontological questions in evolutionary theory in this

book.Adopting the assumption that life has evolved as our other cor-

nerstone of comparative biology allows us to expand the concept ofpattern, or "order in nature." Putting the two t~gether, it is clear th~tthe pattern is a direct result of the process, r.e., that the pattern IS

historical. Whether the pattern is being "perceived," or "analyzed," itis the contribution to our understanding of the history of life that isactually at stake.


Pattern and Process: Logical Interconnections

Pattern results from process. Beyond this simple assertion, we mightask what, if any, other relationship might exist between these twoareas of systematic biology. The question has many possible an-swers, all lying in the methodological, or epistemological, realm(see Bretsky 1979; Eldredge 1979a; Eldredge and Cracraft 1979). Itis apparent that most systematists do not feel the need to adopt anexplicit set of notions about evolutionary processes in order to pur-sue the task of reconstructing the history of life. Reference to the 27volumes (at this writing) of Systematic Zoology amply demonstrateshow little a concern for evolutionary mechanisms characterizes nor-mal work in systematics. Likewise, reference to such journals asEvolution or American Naturalist reveals of how little concern evolu-tionary historical patterns-or, phylogenetic patterns-are to thosewho are primarily concerned with elaborating ideas of evolutionarymechanisms. Yet, delving a bit more deeply, it becomes clear thatvirtually no one operates in either area without at least some refer-ence to the other. In sum, all systematists working on the analysis ofpatt~rn have som~ set of assumptions at least subconsciously con-cernmg the evolutionary process, and these sets are inferable fromthe nature of the "pattern" they perceive and report in the literature.Conversely, except perhaps for the purely inductive mathematicaltreatments of some population geneticists, all evolutionary theoristshave a general concept of the pattern they are trying to explain and,on a more basic level, usually have some sort of historical data onwhich to base their analyses (see chapter 6). Thus, the connectionsbetween the two areas of comparative biology are deep, if notalways clearly acknowledged.

It is our position that, in the analysis of evolutionary history (i.e.,pattern) in its most general form, we need only adopt the basic no-tion that life has evolved. Only when more detailed statements arerequired are more specific notions of evolutionary processes rele-vant. We believe that the most important connection between the twoareas, an aspect as yet underexplored, involves the comparison ofthe ~atterns of both intrinsic and extrinsic features of organismspredicted from theories of process, with those actually "found" in na-ture. Initially, therefore, the study of pattern must be divorced asmuch as possible from the study of process, to provide an unbiased


baseline for the evaluation of alternative hypotheses about process.In discussing methodological questions in both areas of compara-tive biology, we have adopted the view that the procedures shouldbe hypothetico-deductive in nature: elaborate a hypothesis whichcontains the basis for its own evaluation (i.e., predictions). In thisapproach, nothing is ever "proven;" we speak, rather, in terms ofelimination (vretutation" or "rejection") of manifestly false hypothe-ses, and retention of as yet unfalsified hypotheses. Hypotheses con-sistently resisting rejection relative to alternative hypotheses aresaid to be more highly corroborated. Under this view, facts them-selves are nothing more than highly corroborated hypotheses.

The application of the hypothetico-deductive approach to sys-tematics has been much discussed of late (see Gaffney 1979, for athorough review and guide to the literature). In keeping with the gen-eral disinclination of those interested in evolutionary mechanisms tobe concerned with methodological questions, the subject has beenbarely discussed in this branch of comparative biology. As we shalldiscuss at various stages throughout the ensuing chapters, the rab-bit warren of untestable story-telling ("scenarios") which comprisesmuch of past and contemporary evolutionary theory indicates that fu-ture work in this area could benefit greatly if it were cast more ex-plicitly in hypothetico-deductive terms. It is our position that one waywe might accomplish the task of making evolutionary theory moreexplicitly hypothetico-deductive is to use highly corroborated hy-potheses of phylogenetic pattern to evaluate predictions of patterngenerated from hypotheses about the nature of the evolutionary pro-cess. Thus, in our opinion, improvement in evolutionary theory willdepend, to a great extent on the availability of highly corroboratedhypotheses of evolutionary history. And such hypotheses of .patterndepend upon the adoption of rigorous methods of phylogenetic anal-

ysis.

The Study of Phylogenetic Pattern

The biological discipline known as systematics deals with the theoryand practice of capturing the orderliness in nature that has resultedfrom patterns of phylogenetic ancestry and descent. There are two

,.


closely related general concepts pertaining to phylogenetic pat-terns. First, a notion of evolution implies "descent with modification";new intrinsic features, be they genes, specifiable anatomical struc-tures, or bits of behavior, arise from time to time and are inherited bydescendants. This is the general idea of the origin and maintenanceof morphological diversity among organisms. Because some newfeatures appear earlier than others in evolution (the amniote eggbefore mammalian hair, for example), the expected outcome is anested set of evolutionary resemblances: similarities among orga-nisms are hierarchically ordered as the expected outcome of theevolutionary process itself.

The second concept, a corollary of the first, involves the hierar-chical arrangement of taxa, Taxa are groups, or sets, of organisms,defined and recognized according to a set of criteria. The primarytask of systematics is the recognition and classification (naming setswithin a. hi.era~~hical arrangement) of taxa. The two concepts ofnested Slmllarl~leS and nested taxa are closely related, both epis-tem.ologlcally (I.e .. methodologically) in systematics, and both onto-logically and epistemologically in evolutionary theory. All ap-proache~ to understanding the hierarchical arrangement of taxa insystematics depend on the nested pattern of similarity of intrinsicfeatures as the primary data for analysis: the nested sets of featuresreveal the outline of the nested set of taxa. As far as evolutionarymechanisms are concerned, the relationship between nested sets ofIntrinsic features and nested sets of taxa is a complex issue havingto ~o with such fundamental problems as the very definition of bio-logical evolution itself. For the moment, we shall summarize thespectru~ of opinion regarding this relationship by examining theconventionally recognized "schools" of systematics.

It is generally said that there are three contemporary "schools"of systematics: evO!utionary systematics, phylogenetic systematics(also kno~"n as clad Ism or cladistics), and numerical taxonomy (gen-erally, If Inaccuratel.y, used as a synonym of "phenetics''). None ofthese th.ree s~hoOIS IS monolithic; in terms of the results of their anal-yses of Identical data sets, it might even be claimed that they are notve~ much different at all. But there are general theoretical stancesattnb.utable to each which summarize the spectrum of philosophicalpositions taken by contemporary systematists.

Both numerical taxonomy, the most vigorous modern rnanifeata-


tion of pure "pheneticisrn," and phylogenetic systematics can beviewed as late-coming developments that at least partly representreactions against evolutionary systematics, Cast in the role of the"traditional approach" (largely by virtue of the vocal assaults, first bynumerical taxonomists, then by cladists, and the equally assertivereactions by its generally acknowledged spokesmen), evolutio~arysystematics is actually a multifarious discipline which last receiveda thorough intellectual housecleaning in the 1940s. The "new sys-tematics" is characterized particularly by an attempt to take in-traspecific variation into account, with a concomitant eschewal of"typology" (i.e., the characterization of taxa, especially, but not ex-clusively, species, as if they did not vary). It was widely hailed ~s.agreat advance, in keeping with the spirit and letter of the synthesis In

evolutionary theory, itself wrought in part by the very same bio-logists, It would appear more accurate to view evolutionary system-atics as another, equally vigorous area of systematics, rather thansimply and cavalierly as a body of thought whose time has comeand gone simply because of its imagined advanced age.

Although all three "schools" of systematics are difficult to char-acterize briefly without caricature, evolutionary systematics is .esp~-cially difficult It is tempting to suggest that this state of affairs di-rectly reflects the fact that the much-vaunted "synthetic t.heOry" ofevolution, so closely tied to this approach to systematl.cs, I~ a greatdeal less completely "synthesized" than is popularly Ima~lned (weexpand on this theme later on in this chapter and at length In ch.a.pter6). In any case, there does seem to be a core of basic pr?~OSltlonsprobably agreeable at least to the majority of the practitioners of

evolutionary systematics. .The central tenets of the evolutionary school of systematics

seem to be that life has evolved, that the order we see is a product ofevolution, and that the goal of systematics is to reconstruct th.at e~o-lutionary history as closely as possible. Furthermore, evolution Im-plies ancestry and descent. and therefore the. hierarchic~1 nature ofthis order should be depicted on phylogenetic trees, which specifypatterns of ancestry and descent. Descendants genealogically farremoved from ancestors should resemble each other less than moreclosely related ancestors and descendants, reflecting, at base, de-grees of genetic similarity. Thus in the evolutionary school there al'etwo different kinds of similarity: (a) true evolutionary resemblance,


i.e., resemblance due to inheritance from a common ancestor, and(b) false resemblance (usually termed convergence; "paralellisrn'' isincluded here, but is regarded by most evolutionary systematists asa sort of intermediate case between "true" and "false" resemblance).False resemblance does not reflect inheritance of the similar struc-tures from a common ancestor. Trees derived from such analysescan then be converted into a classification by adopting a set of cor-respondence rules (to use the phrase of Colless 1977).

Not all evolutionary systematists necessarily accept all thesepropositions. Moreover, there is a great deal more to formulations ofthis school than the crude, bare outline above, as readers of Simp-son's (1945, 1961) and Mayr's (1969) books on the subject will attest.But the characterization we have given does at least seem to sum-marize the basic propositions in a fashion probably more or lessagreeable to all.

The late 1950s and the decade of the 1960s saw the develop-ment of numerical taxonomy. With the publications of Principles ofNumerical Taxonomy (Sokal and Sneath 1963; a second book, up-dating the first, appeared as Sneath and Sokal 1973), the theoreticalbasis of numerical taxonomy was established. The main concern ap-peared to be a desire to reformulate the process of delineating life'sorderliness in a more standardized, repeatable, rigorous, and objec-tive fashion (Sakal and Sneath 1963:49). In a sense, this approachtreats the pattern as if it merely needs to be "perceived," as if it werelike any collection of "data." Viewed in this manner, the scientificperception of the pattern is to be formal ized and made objective likeany other precise, but routine data gathering operation in science.This view is far removed from the general notion that any "observa-tion" of nature constitutes an hypothesis.

Numerical taxonomists note the many difficulties in reconstruct-ing phylogenetic history, and therefore tend (e.g., Sakal and Sneath1963) to dissociate their activities from phylogeny reconstruction.Acknowledging that the patterns do reflect phylogenetic history, nu-merical taxonomists nonetheless have claimed that such history isunknowable with certainty, or in fact in any detail. Objecting particu-larly to Simpson's (1961:110) remark that "like many other sciences,taxonomy is really a combination of a science, most strictly speak-ing, and of an art" (a statement which pertains to the construction of


classifications only, and not to phylogeny reconstruction), numericaltaxonomists sought to make systematics more "scie~tific." .

The central methodological principle in numerical tax~nom~ I~phenetics-the clustering of samples ("operational.ta~on.oml~ units,or "OTUs") according to an index of overall stmilartty. Pheno-grams" or "dendrograms," are generated (almost always by comp~~ter) accordinp to an algorithm specifying for one of the many a~ail-

, ' ' QTU Tracing their ideologicalable measures of slrnitartty among s.pedigree back to the botanist Adanson (1763), who advocate.d ex-amination of as many characters as possible and the productlo~ ofclassifications based on overall similarity, numerical .taxono~lstscontrasted their "objective" approach with the "subjective" weiqht-. c. ! rathering (selectivity) of characters practiced, they clalm.e , In avague and arbitrary fashion by evolutionary systematists.

Thus concepts of similarity and its measurement seem t~ con-, .. t d volutlOnary

stitute an important difference between phenetic ISs an ,esystematists. The concept of overall similarity consciously lumps"true" evolutionary resemblance with convergent and parallel resem-blance: the position of most numerical taxonomists seems to ~e that,if enough characters are examined, "real" resemblance will ~hut~weigh "false" resemblance. Evolutionary systematists, on the.o er

"true" (i e evolutionaryhand make explicit the distinction between ' .,or phylogenetic) and "false" (convergent) resemblance. Ho;eve.r,once the "true" set of resemblances is identified, the opera. Ion IS

, " f olutionary systematics, atpurely phenetic' the baste crttenon 0 ev . . .least according 'to Mayr (1969:200) and Bock (1977) IStwhemaxl~I~~

. . d d by the phenotype. e conc u etion of genetic similarity as JU ge . and evolutionarythat the difference between numerical taxonomy b

d I . al differences seem to e,systematics great as the metho OOglC .' fboils down to slightly different views about wha~is attainable l~solaras the scientific analysis of life's orderliness IS concerned. va u-

. k t es with ancestors and descen-tionary systematists strive to rna e re , . f. . I taxonomists are content with statements 0

dants while nurnenca di tl IheI ttve" s'' of samples and thus define taxa tree y onre a Ive nearnes '

basis of clusters found on phenograms generated by some measure

of overall similarity. .' ", d h I "cladistics" or "phylogenetic systematics,The thfr sc 00, ' .' .

Is to occupy an tntarmediate position onseems, in many respec ,

10 Pattern and ProcessPattern and Process 11

cladists adopt a less extreme view than evolutionary systematists, inthat their scientific goals regarding the reconstruction of phylogenyare tempered by the theoretical and methodological difficulty ofdealing with ancestors. On the other hand, the cladists do not sharethe utter despair that numerical taxonomists have expressed con-cerning the impossibility of dealing with the history of life in a ratio-nal manner. We should adopt the position (see also Eldredge 1979a;Wiley 1979) that, logically, the ancestral units of evolution are spe-cles-c-r.e.. that supraspecific taxa do not form ancestral~descendantunits. Therefore any system, be it a branching diagram or a clas-sification, which recognizes supraspecific taxa as ancestors is illogi-cal. In this sense, there is no difference between a cladogram and atree depicting relationships among supraspecific taxa, inasmuch asthe added information of trees (identification of certain taxa as an-cestors) is superfluous. However, species do form ancestral-descen-dant sequences, and thus, logically, their histories are appropriatelydepicted on phylogenetic trees. We discuss the methods by which acladogram can be converted into a phylogenetic tree for species inchapter 4 and further claim, in chapter 6, that phylogenetic trees arethe actual patterns required for the scientific study of speciation.

Again, there seems to be little difference between cladists andevolutionary systematists (and numerical taxonomists, for that mat-ter) except for a d ifterent emphasis on such issues as the nature ofscientific inquiry in systematics and what might accordingly be at-tained in terms of the reconstruction of the history of life. Both Mayr(e.q.. 1974:98) and Simpson (1975:14), as well as other noted evolu-tionary systematists, have acknowledged the validity and importanceof Hennig's explicit statement that the evolutionary process pro-duces, as an expectation, a nested set of evolutionary novelties, Theadditional information that evolutionary systematists wish to incorpo-rate into their trees and classifications seems to be based on certainassumptions about the evolutionary process, particularly the extremeimportance accorded to adaptation as the central theme and prob-lem in evolution (see chapters 5 and 6 for an extended discussion ofthis issue).

But, at the elemental level of recognition of patterns of similarity,we might fairly ask if there are any fundamental differences in addi-tion to the obvious methodological ones among the three schools,Pheneticists talk of overall similarity, evolutionary systematists speak

these various issues. With the evolutionary systematists, ctadists be-lieve that the orderliness of the biotic world derives from evolution,and that the reconstruction of the history of life is the central goal ofsystematics. It was Hennig (1950, 1966) who first made explicit yet athird component of "similarity." He accepted the usual dichotomybetween "true" evolutionary and "false" similarity, but further pointedout that, for any monophyletic taxon (i.e., a taxon composed of two ormore species consisting of an ancestral species and all its knowndescendants), evolutionary similarities shared by a group are of twosorts: those held over from some remote common ancestor (e.q. twopairs of limbs in mammals) vs. those held only by members of thatgroup (e.c.. three inner ear bones in mammals). Hennig pointed outthat older evolutionary novelties can be retained in a sporadic man-ner; consequently their utility for defining and recognizing clusters oforganisms is appropriate only to the hierarchical level at which theyr~present ~ru: e~?lutionary novelties. Thus cladista amplify the set of~Inds of slm.llantles recognized by evolutionary systematists, seem-Inqly departing even further from the position of the pheneticists.

Howeve~,because ancestors never possess a set of evolutionarynovelties unl~ue to themselves, their definition and recognition is,logically, difficult. For this reason, cladists concentrate on a more el-ementa~I:~el: the prime goal of systematics, according to cladists, isthe definition and recognition of monophyletic groups. This is ac-complished by the search for nested sets of evolutionary noveltiesdepicted on. branching diagrams called "cladoqrams.' Cladogramsorder organisms according to nested sets of these novelties' con-sequently, the organisms are ordered as well into nested ~ets-(hypothesized) monophyletic taxa (see chapter 2). The procedurehas the added advantage of being easily converted into classifica-tions Witha minimum of required conventions (see chapter 5).

As diagrams of the history of taxa, cladograms can be in-ter~reted in terms of relative recency of common ancestry. As firstpointed out by Nelson (personal communication, 1976), and furthercharacterized by Platnick (1977b), Tattersall and Eldredge (1977),Cracraft (1979), and Eldredge, (1979a) (see Wiley 1979 for a counterOpinion), a clad~gram subsumes the logical structure of a set oftrees. Phyloqenet!c trees, in specifying actual series of ancestral anddescendant taxa, are more detailed and precise sorts of hypothesesthan are cladograms. Thus, as Platnick (1977b:441) has discussed,


of overall (genetic) similarity in the context of "true" (as opposed to"false") evolutionary similarity, while cladists additionally recognizelevels of "true" similarity. Cladists avoid the confusing issue of"weighting" by recognizing that all nonconvergent characters arerelevant to defining monophyletic groups at some level. The problemis the recognition of the correct level for any character (see chapter2). But congruence of patterns of similarity can be expected, in manycases, to lead to identical groups, whether the analysis is periormedby a numerical taxonomist, a cladist, or an evolutionary systematist(Nelson 1979).

There is, however, one fundamental difference between thesethree approaches to pattern analysis in systematics, The differencepertains to the definition of taxa, and not to the analysis of similarityper se. Without an explicit attempt to evaluate similarities at theirproper hierarchical level, clusters of organisms whose similaritiesare primitive retentions rather than evolutionary novelties typicallyresult. Thus the problem with phenetics in general is not parallelismor convergence (a problem usually resolvable, especially with refer-ence to a parsimony criterion-see chapter 2) but. as in evolutionarysystematics, the failure to evaluate evolutionary novelties at theirproper level. The effect this has had on analyses in both evolutionarysystematics and numerical taxonomy is fundamental: some taxa sodefined are non-monophyletic. Both numerical taxonomists and evo-I~tionary systematists regularly admit groups based on shared reten-tion of primitive features into their systems. Such groups are inevita-bly of dubious cohesion. as the evolutionary systematists themselvesseem to a?m~twhen they endorse the use of evolutionary noveltiesfor the def.lnltlon of taxa. As Farris (1977) and Platnick (1978b) haverec~ntly discussed, monophyletic taxa maximize comparative infor-mation ~boul organisms. Moreover, and this argument is crucial 10all consioerattons of evolutionary process in this book evolution isfirst and foremost a ge , ' , "nea oqsce process. It produces genealogiesof ancestors and descendants, If we are to compare pattern withtneortes of process-as we m t d ' ,us 0 to Improve our very notions ofproce~s--we must have at the very feast an accurate concept ofe~ofutlonary.genealogy. As will be developed in chapter 6, it is amistake to Invent theories of process to explain the origin of non-m~nophyletic groups, It follows. then, that any procedure in system-atics whereby nonmonophyletic groups are routinely recognized dis-

,, t"


tortsnot only the information content of the classificatory system. butalso our very notions of evolutionary processes. This suggests thatthe system best suited to the recognition of monophyletic taxa is thebestsystem for comparative biology as a whole.

The Study of the Evolutionary Process

As we stated earlier in this chapter, most evolutionary theory is on~o-logical, rather than epistemological; there is far more concern Withhowthe process works in nature, than with how we know about thatprocess. Many of the problems in evolutionary theory ~t~m from aninattentiveness to the "how we know" component. Specifically, thereis a very strong tendency amounting to a tradition, to (a) reconstructphylogenetic trees, (b) elaborate some theory of evolutionary mecha-nisms, and (c) subsequently explain the phylogenetic patterns of (a)in terms of the theory of (b). (See, for example, the statement to thiSeffect by Bock and Von Wahlert 1963:140.) Not only do theories tendto be invented which can explain all patterns (a frequent, and oftenjustified complaint about the use of the concept of adaptation by nat-ural selection, for example), but, perhaps more insidiously, the .~at-

, ' h f trion as to fit prevaIlingterns themselves are perceived In suc a as I, to th popularity of recoq-notions of process. As an example, consi er e, , , th I t t9505 and early 1960s. Ap-ruzmq polyphyletic groups In e a e

patently stemming from Huxley's (1958) paper discussing gradesand clades (patterns) as well as anagenesis, cladogenesis, and sta,~, ,) " olyphyly

stqenesis (modes of process prodUCing the patterns, P .. . .. I t' process In our View,enjoyed WIde discussion as an evo u ronary '., f

this flurry of theoretical work represents a sort of hypertrophlcat~on ofthe general view that evolution is fundamentally the transformation 0. . . 't to adaptation via natu-mtrtns!c features, best explained by re erenee .

'zed the eXistence ofral selection, Investigators cheerfully recoqru ., f d t-e-e 9 Mammalia,taxa acknowledged not to share unity 0 escen '.' _

. n nine different times,once claimed by Simpson (1959). to have arise, t were based on con-a view no longer in favor. Rather these axa 't

I xtent [oint retention 0ve.rgences, parallelisms, and, to a. ~sser e '. . endentpnmitive features. (The theme definitely emphaSized Indep

----


acquisition of evolutionary novelties designed to perform the samefunction.) This celebration of adaptation came directly from evctu-tionary theory and, inasmuch as it dwelt on the "explanation" of theevolution of phylogenetically non-existent groups, is best regardedas a bizarre conclusion resulting from the application of a theory farbeyond its appropriate limits (see chapter 6).

It is our firm conviction, in contrast, that the conventional prcce-dure of explaining pattern in terms of notions of process (step 3above) is grossly in error. We agree that phylogenetic patterns mustbe analyzed (as wholly independently from notions of process ashumanly possible) and theory invented to explain pattern-I.e. thefirst two steps. But the crucial third step should be the direct com-parison of predictions of pattern drawn from ideas of process di-rectly with the analyzed phylogenetic patterns themselves, with theaim of critical evaluation of the notions of process themselves. Amajor problem of contemporary evolutionary theory from an epis-temological standpoint is not so much that its propositions are untes-table, but that its main practitioners use theory to explain away pat-tern (see Grene 1959, for a clearly analyzed example). Rather, thetheory should be tested by comparing it with best estimates of actualevolutionary results-I.e., carefully reconstructed patterns of phy-logenetic relationship.

From an ontological standpoint, there is also much to criticize incontemporary evolutionary theory. As systematists, our main concernis with patterns of relationship among taxa. Systematists naturallyhave been most conversant with, and have themselves helped tocreate, that part of evolutionary theory specifically addressed toproblems of the origin, persistence, and extinction of taxa. Mostrecently, this work has focused on species (e.g., Mayr 1942, 1963),and for good reason: species are unique as taxonomic entities (seechapter 3). In discussing among-taxa evolutionary problems, a verysubtle but crucial issue is immediately raised. Just as cladogramsdepict nested sets of characters, and thereby also depict nestedsets of monophyletic taxa, so does the issue of the evolution of taxabecome intertwined and confused with the modification of intrinsicfeatures in evolutionary history. We can define taxa only in terms ofthese intrinsic features. It was probably inevitable that the predomi-nant view of evolution would stress the transformation of intrinsicproperties as the central issue. There is no doubt that this has hap-

Pattern and Process 15pened (chapter 6). That transformation of intrinsic properties is thecentral theme of evolution to most biologists is evident in the generaldefinition of evolution historically acceptable even to prominent sys-tematists (e.g., "evolution is any change of gene content andfrequency within populations"). Because we characterize taxa by in-trinsic properties, we tend to assume that their evolution is a mattersolely of the transformation of these features rather than a matter ofthe origin of species as well. This reductionist viewpoint has beenparamount and underl ies the lack of true synthesis in evolutionarytheory.

In contrast. we have noted (especially in chapters 3, 4, and 6)a historical thread of interest in problems of the evolution of taxawhich is to be sure, related to, and intermixed with, but not the sameas, the problem of the transformation of intrinsic features. Taxonomicdiversity, in our opinion, is not a synonym of morphologic diversity.They are related, and how they are related is an interesting problem.But they are not the same. The bulk of contemporary evolutionarytheory focuses on morphological diversity, to the point. at times, ofpresuming that the two sorts of diversity are isomorphic (see chapter6 for an extended discussion of these issues). In keeping with ourinitial perspective as systematists, we have adopted and defendthroughout this book the position that species are real entities exist-ing in nature, whose origin, persistence, and extinction require ex~planation. All species (except those reproducing strictly asexually,with no exchange whatsoever of genetic materials among lndivid-uals-a distinct minority of organisms) are held together by a patternof parental ancestry and descent that is disrupted when new speciesarise from old. Adaptation and natural selection are hypotheses toexplain changes of intrinsic features within populations from onegeneration to the next. Speciation disrupts these lineages of parentalancestry and descent. The foregoing implies that among-speciesdifferences do not flow directly as a simple, reductionist eareoote-tion of within-population generational change. The central role ofspecies as real units in nature-these units being the ancestors anddescendants of the phylogenetic process-implies a view of distinctphenomenological levels in the evolutionary process. In such a view,rnicroevolution is, perhaps, best defined as change in gene contentand frequency within populations, and macroevolution is defined aschange in species composition within a monophyletic group in

J


space and time, best thought of, perhaps, as a proce~s of differentialspecies origination and survival within mo~o~hyletlc taxa, We ex-plore these possibilities in far greater detail In chapters 3, 4, andespecially 6,

Integration of Pattern with Process:The Structure of This Book

The structure of the remainder of this book can be summarize.d, asfollows: First. in chapter 2, we discuss method and theory. pert~lnln.gto the analysis of evolutionary novelties: cladistic analysl.s. It IS t~ISnested set of intrinsic features which gives us the main signal of in-terrelationships among taxa. It is through cladistic analysis t~at wederive our hypotheses of (a) the composition of monophyletic :a~aand (b) the distributions of nested sets of characters In the bioticworld. Both types of sets have further uses. .

Next, in chapter 3, we discuss the special case of species ..Wedefine species in such a way as to stress their internal cO,heslon,their identity as discrete, real entities, and their unique position asphylogenetic units, No taxon other than ~peci~s serves as a~cestorsand descendants (! e.. as phylogenetic units) In evolution,' Wediscuss two sets of criteria that form the basis for the evaluation ofhypotheses of species composition. , .

Then, in chapter 4, we briefly discuss baSIC hypothesizedmodes of speciation (both those consistent with our own character-ization of the nature of species, as well as some consistent with otherspecies definitions). This departs from our overall plan to use patternto evaluate predictions derived from theories of process. We followthis procedure to expose the theoretical patterns of possible ph~-logenetic trees. We then discuss trees themselves, finding them logi-cal structures only when species are considered, and adduce somemethodological rules for converting cladograms into trees.

Chapters 5 and 6 constitute our effort to characterize howbranching diagrams of interrelationships among taxa may be put tofurther use. In chapter 5, we detail the methods and procedures, aswell as history of ideas and general purpose, of the construction of

Pattern and Process 17biological classifications, We conclude that the history of system-atics might best be characterized as the progressive elimination offalse ("unnatural") sets, especially those sets "defined" by the ab-sence of features used to define a coord inate set; we refer to thesesets as "not-A" (e.g., Invertebrata) and "A" (e.g., Vertebrata) sets, re-spectively. We conclude that monophyletic groups maximize the in-formation content of a classification and, because ancestral taxacannot be monophyletic (by any acceptable definition of mono-phyly), they therefore cannot be arranged in an hierarchic fashionwithout special conventions, We conclude, therefore, that clade-grams alone are necessary and sufficient for the construction of clas-sifications. We present some correspondence rules for the conver-sion of cladograms into classifications.

Finally, in chapter 6, we consider the use to which highly cor-roborated hypotheses of patterns of relationship among taxa mightbe put in the evaluation of statements in evolutionary theory, We takethe position that the majority of evolutionists (whether neeDarwinists, syntheticists, or saltationists), have focused on the expla-nation of change of intrinsic features, primarily on morphologicchange. The most common mode 01 explanation is couched in termsof the paradigm of adaptation via natural selection. We point out thatthe extrapolation of within-population dynamics of variation and se-lection to explain morphological and adaptational differencesamong taxa of higher categorical rank results in a theory of the evo-lutionary process which purports to discuss taxa but in reality onlydeals with the intrinsic properties of organisms. We conclude that ifspecies are discrete reproductive units, microevolutionary pro-cesses cannot logically be extrapolated in a reductionist manner, toexplain macroevolutionary patterns.

We then examine evolutionary theory from the standpoint of phe-nomenological levels. We conclude that rnicroevolution (within-POpulation, within-species transformation of intrinsic features) mayb~ hypothesized to result from adaptation via selection For its scien-title study, microevolution requires direct historical information, on a~en~rational basis, which can be obtained experimentally. Specia-tion Itself requires detailed phylogenetic trees for its scientific study.For the analysis of macroevolutionary patterns-i.e., patterns of ori-g~nof, and fluctuations of diversity within, monophyletic taxa of rankhigher than species-only cladograms are required, which supply j

tI


(a) the nature of the composition of the monophyletic groups, and (b)relative distributions of evolutionary novelties, Each component (i.e.,rnicroevolution. speciation, macroevolution) of evolutionary theoryrelies upon use of the kinds of historical data (hypotheses) apprcpri-ate to it, for its eventual improvement. Such improvement is to be ex-pected particularly when predictions of pattern, based on theories ofprocess, are directly compared with independently derived hypothe-ses of pattern, in a hypothetico-deductive manner. We close thechapter, and the book, with a brief characterization of a testable, tax-fealty oriented theory of macroevolution.

p,.

Chapter

2Cladograms: Cladistic Hypothesesand Their Analysis

THE SUBJECT of this chapter is a fundamental one in compar-ative evolutionary biology: how do we reconstruct the history of life?We will claim that hypotheses about phylogenetic history are neces-sary standing points for virtually all comparative evolutionary studies.Unfortunately, many evolutionary biologists either choose to ignore theproblem of phylogenetic history altogether, or it is dismissed on thegrounds that "direct" evidence is lacking-the implication being thatthere is no fossil record for the group in question. In this chapter wewill suggest that an estimation of phylogenetic history can be ob-tained with or without a fossil record (although paleontological dataare surely welcome in any study), and subsequent chapters willelucidate the importance these estimations have for investigatingevolutionary questions. To many biologists the claim that phylogene-tic history is capable of analysis without paleontological evidencemay sound a dissonant chord, but the problem can be resolved byunderstanding not only the general philosophical approach adoptedin this book but also the specific kinds of phylogenetic hypotheseswe believe are subject to scientific investigation.

In our view, any methodology designed to reconstruct phyloge-netic history should be hypothetico-deductive in structure, that is, al-ternative hypotheses must imply different empirical consequencesand these in turn serve as the basis for evaluating those hypotheses.Thus, each hypothesis is expressed so that it can be criticized andrejected if the evidence so warrants. We are interested, therefore, ina methodology capable of producing and evaluating hypotheses ofthis kind. We reject outright any predilection for narrative-type sce-narios of phylogenetic history if they are not expressed in arigorOusly testable form. The literature is replete with these narra-

"""-z _

20 Cladograms

tives, many of which may sound very reasonable, but all toofrequently there is little or no way to subject them to critical analysis.In such cases, the descriptions of historical events essentially lieoutside the realm of scientific inquiry,

In this book, we are concerned primarily with erecting hypothe-ses about the pattern of life's history, including the supposedsequence of phylogenetic branching events and the distribution ofsimilarities and differences among the organisms being studied,Thus, the initial, primary question is exactly what has happened, notwhy or how. These hypotheses, then, should attempt to explain, assimply as possible, all the relevant empirical data bearing on thequestion of what has been the history of life.

Cladograms, as defined and discussed here, are specific kindsof hypotheses about the history of life. They are hypotheses aboutpattern. The concept of evolution implies change or modification inthe intrinsic properties of organisms from some state or condition, a,to a derivative state or condition, a'. More significantly, however,evolution means that somehow one kind of organism may change orevolve into a different kind of organism. In other words, there is an-cestry and descent. It follows that similarities in intrinsic propertiesamong organisms are, in some way, a manifestation of the processof evolution. From Darwin to the present the evolution of Iite hasbeen viewed as a process of branching and diversification. Such aprocess produces a pattern, and that pattern is hierarchical, or, putanother way, it is a sequence of nested sets of organisms (the"groups within groups" of Darwin), and it is generally agreed that thesimilarities among these organisms must mirror that hierarchy insome manner. Hence, systematists look to the analysis of similaritiesto give some understanding of that hierarchy.

The question arises as to the kind of similarity that might be ex-pected to evidence this nested pattern. After all. organisms may besimilar to one another in various ways. Does this nested patternreflect a general measure of the overall resemblance of one orga-nism to another? Or is this similarity of a more special. more specifickind? The answer is related to the process-evolution-that has pro-duced the pattern of branching and divergence. The assumption ofevolution implies that while there is a continuity in the characteristicsof ancestor and descendant, there is also change. During the evolU-tionary process, newly evolved organisms come to be characterized

by novelties, their desce d' Cladograms 21. n ants retain th

pectatlon of nested sets f ' esa, and thus there isCI d 0 evolutionary nove'f an ex~a ocrams, therefore are res.

neste~ evolutionary novelti~s post~i~~~~ses about the pattern ofo:ganlsms, As SUCh,cladograms are 0 o~cur among a group ofn.,smsor groups of organisms Sam branching diagrams of orga-view the hierarchical structure' of lewhat more formally, one can(set~)and subgroups (SUbsets) th c ~dograms in terms of groupssharrng a branch point com '. e e ements (groups of organismsthemselves defined by higher brlse : Subset within other SUbsets)cladograms are representative r~nc points of the hierarchy. Hence'that as the hierarchy is aseend:d °t~~nowledge about organisms i~POintsare inclUded, statements abou~t~~'as more and more branchbec?me more general. Clado ram ISknowledge can be said toerallzation, and thus specify d~ff s, th~refore,.denote levels of qen-about attributes pertaining strict~r~nt hierarchical levels, Statementseral (more specific) than those y 0 ~a~s,for instance, are less gen-::~ert in turn, are more restriet~e~:~nl~~ to all mammals, and thera es. nose pertaining to all vet-

, The question arises' to whecereo actual rep . at extent are cladograms to be cI resentatlons f h on-e adogram has been sop ylogeny? The concept of theprior literature. The Vi:~o::ous wit~ "phylogeny" in some of the~elves,are not phylogenies, bU~ here IS that cladograms, in them-1n~sted eVOlutionary novelti ~~ther hypotheses about the patternganrsms into sets and subs es. ISpattern allows us to arrange or-a~d descent, Whereas eladoets, Phylogenetic trees specify ancestry~ ylogenetic trees requires grams do not. Thus, the construction 01, emselves are not neces certain evolutionary assumptions whichtl~ctlon between cladog sary for cladogram construction, The dis-c aote- 4, rams and trees will be examined further in

Cladogram . Ss. orne Introductory PrinciplesA Simple ExampleInorder foClad r the reader to qai .. - .

ograms express 0 n some InItIal Insight into the notion thatur general k J dnowe ge about organisms and

p -------------:---------.,I

22 Cladoqrarns . . a be. f nested evolutionary novelties, It my.

are constructed In terms a . organisms familiar to most, Ifirst t example usmqhelpful to turn first 0 an . t d in part to demonstrate' pie IS presen e I

not all biologists, This exam anisms can be expressed' , , d differences among org

that similarities an I I it elf will be used to torrnu-, h d the exarnp e I Sby a nested hierarc y, an b t cladograms. Naturally, be-

, I statements a au itlate a sense of genera . it nnot convey the complex I yI ' a simple one I ca . .cause the examp e IS stemati t undertaking a cladisticf t"ng a system a ISof problems often con ron I , h Id help to formalize the

I the exercise s au ftanalysis. Neverthe ess, _ th ds that until recently have a enthode of many systematists, me 0me ,

remained essentially intuitive. b t s Only five well-known. k f the verte ra e .

The example ISta en rom d small restricted sample ofkinds of organisms are considered, an a (At 'this point in the dis-

'bl h s been chosen.the characters POSSI e a ddi organisms or characters:cussion little would be gained by a Ing licated situations willhow the investigator might handle more camp

be outlined later in the chapter.) I dog ram for the follow-The problem at hand is to construct ahea d lamprey. Suppose

lizard perc , aning organisms: cat. mouse, I , . s encompassing all. ted by specimeneach organism IS represen d specimens, ob-

Phasesof its life cycle. Because these are prel~e~vgematerial, for in-. h b en made on IVln .servations that might ave. e havi are not a matter of rrn-

stance physiology, biochemistry, or be aVlor,. be restricted tomediate concern. For the moment, the anaivsis can

. h ilable specimens.the morphological features In t e avar t my of these or-k of the ana aThe first task is to underta e a survey . This morpho-

, 'I it! and differences. dganisms and to note the Simi an res . 'exhibit a broalogical comparison indicates that the five o~ganlshms considerable

. h organism sowsstructural diversity. Moreover, eac ther Because. rt cle to ana 'variability in form from one stage of ItS ne cy are available,

a number of specimens of each life history stage pecimens of, I b erved among s .variability in some features IS a so a s ming the dts-, , t ount and assuthe same stage. Taking all this In 0 ace , d s! uanues among

, th observe simitinctness of each organism, how are e may be con-the five organisms to be evaluated so that a cladogram

structed? .. ider one of two ap-To answer this last question, we might cons I f e organismS;

preaches First the analysis might be confined to the IV ps by nest-' , I ested grouin this case an attempt must be made to orm n

Cladograms 23ing statements about evolutionary novelties postulated from the simi-larities observed among the five organisms. In the second approach,a comparison with other groups of organisms might be used tosuggest which similarities appear to be evolutionary novelties andthus what the pattern of groups and subgroups might be. As will bedeveloped later in the chapter, there is reason to believe both ap-proaches will lead to similar results, namely, a cladogram that de-picts our knowledge of the organisms in terms of nested sets ofevolutionary novelties.

Upon dissection and study of the specimens a number of possi-ble groups are suggested. Obviously, each organism is unique anddifferent from the others and thus one can recognize five basic setsin the analysis, each containing one kind of organism. Each of thesesets can be defined by the attributes shared by all specimens as-signed to that kind of organism. But the goal of cladistic analysis isto compare similarities and then to group these individual organismsinto larger groups, nested one among the other. The first question tobe asked, then, is: What are the general similarities shared amongthe organisms that could be used to define a group containing allfive (termed the universal set of the comparison)? There happen tobe a number of these similarities; semicircular canals in the head, achambered heart, visceral (gill) arches at one or more stages of thelife cycle, a dorsal nerve cord, a notochord. and appendages ofsome kind. All these similarities help define the universal set, al-though only one is sufficient and necessary. All other observed simi-larities among these organisms, unless they contribute to the defini-tion of the entire group (universal set), must define SUbgroups.

What are these subgroups? A comparison of similarities beginsto yield conflicting results. For example, similarities such as an am-niote egg indicate a set comprised of lizard, cat, and mouse. On theother hand, the perch resembles the lamprey in having a non-amniote egg, and thus might be grouped with the latter; and the liz-ard, perch, and lamprey have neither hair nor mammary glands,Implying they might comprise a subgroup,

How are these conflicts to be resolved when the comparison isConfined to just these five organisms? It will be recalled that the as-SUmptionof evolution provides us with an expectation that groupsand Subgroups will be characterized by evolutionary novelties. Analternative way of looking at this is to consider evolutionary novelties

�I,

..

24 Cladograms

as changes or modifications of preexistng features (say, from a toa'). Because these preexisting features (a) define a group that alsoincludes the subgroup characterized by the modified feature (a'),these preexisting features are more widely distributed within the uni-versal set. We need, then, to search for more restricted features(novelties) that are interpreted as changed or modified conditions ofmore widely distributed features. One approach is to compare themorphology of each organism at various stages of its life cycle inorder to see if some features change during development, that is,from features that are shared generally with other organisms to thoseshared by only a few. Perhaps these changes can be used sub-sequently to hypothesize which features are more widely distributedand which are less so.

At one or more stages of their life cycle all five organisms havemuch of their skeleton consisting of cartilaginous elements. In theperch, lizard, cat, and mouse, bone is seen to replace much of thiscartilage in later developmental stages. These observations suggestthe hypothesis that cartilage is a more widely distributed character-istic of skeletons than bone, and that bone, therefore, can be used tocharacterize a subgroup consisting of the perch, lizard, cat, andmouse, Are there other similarities among these four organisms tocorroborate this grouping? All four have jaws, three semicircularcanals, paired appendages, and a well-developed vertebral column,to name only a few similarities. If early developmental stages are ex-amined, there are indications that some, or all, of these similaritiesare modifications of more widespread conditions present in the earlystages of all five organisms, For example, the early head skeleton ofall five shows many similarities, including the absence of definitivejaws, but the perch, lizard, cat, and mouse go on to develop a l0v.:erjaw. The arrangement of three semicircular canals in the perch, uz-ard, cat, and mouse is a modification of the condition found in earlydevelopmental stages of all five organisms (the adult lamprey hastwo semicircular canals). This evidence seems to support the hy-pothesis that the perch, lizard, cat, and mouse form a subgroupwithin a larger group that includes the lamprey.

Within the subqroup just defined, are there further subgroupS?The lizard, cat, and mouse have an amniote egg, and this featurecan be hypothesized to be a modification of the generalized ver-tebrate egg, more or less typified by those of the lamprey and perch.

Cladograms 2S

This interpretation is suggested by the fact that all five organisms~ho~ similarities in the early stages of development following fertil-izatron, but the lizard, cat, and mouse depart from the general pat-tern and develop an amniotic membrane.

Finally, among the remaining three organisms, the cat andmouse can be grouped together because of their possession of threeear ossicles, hair, and mammary glands. The three ear ossicles aremodified from skeletal elements in the region of the jaw articulation,a pattern common to jawed vertebrates in general; hair and mam-mary glands can be interpreted, developmentally, as modificationsof more general structural characteristics of the vertebrate in-tegument

By nesting these statements about similarity, the cladogram offigure 2.1 can be constructed. Hypotheses about change in features

CAT MOUSE

LIZARD3 ear ossicleshairmammary glands

LAMPREY

PERCH

amniote egg

JOws3 semicircular canalspaired appendagesvertebral column

semicircular canalschambered heartvisceral archesdorsal nerve cordnotOchordappendages

Figure 2.1 A cladogram for five kinds of vertebrates. Each level 01the hier-~rChY(denoted by branch points) is defined by one or more similarities in-erpreted as evolutionary novelties. (See text.)

•

,

I

26 Cladograms

(character transformation), that is, hypotheses about the hierarchicallevels of distribution of the features, were proposed from compara-tive observations of the developmental stages of the five organisms.This comparative method of constructing nested sets of organismsbased on the determination of a feature's distribution, as revealed bydevelopmental sequences, is essentially the method proposed andused by the German anatomist von Baer over 125 years ago. Morewill be said about the validity and limitations of this method later inthe chapter.

An obvious question stemming from an analysis such as theabove is: how can a cladogram be constructed in the absence of de-velopmental data? Suppose we have only adults of our five orga-nisms. The key problem is to determine which similarities are evolu-tionary novelties, and perhaps unexpectedly, the problem isresolved just as in the preced ing example by asking how wide-spread each similarity is. To answer this question other organismsmust be taken into consideration.

Our analysis is predicated on constructing a cladogram for thefive organisms. Thus, the group as a whole can be characterized bymany of the same similarities noted earlier: semicircular canals, dor-sal nerve cord, and chambered heart. The problem, as before, is todetermine which of the similarities within the group are evolutionarynov~lti~~. This can be accomplished by determining whether thesimilarities observed within the group are also found in other orga-nisms outside the group. If they are, then those similarities can bePostulated as being too widespread to define subgroups, that is,those similarities are not novelties within the set of five organisms.

Potentially, a vast array of organisms could be compared withthe five orqanisms, so it makes practical sense to place some limitst~ the kinds that will be studied. Intuitively, it may seem that orga-nisms most like those being investigated would be most appropriate;~fter.al.l,morphological featu~eswithin the group must be co~paredo Similar structures In outstda organisms. A reasonable first ap-P~O~Ch.'.then, is to compare those organisms sharing one or moreslmila~ltles and see how far the within-group comparison can pro-c~ed, t.e., see if a cladogram can be constructed. Thus, several out-side groups are similar to the five organisms in having a dorsalhollow nerve cord and a notochord the lance let amphioxus and thet . t 'umca es (urOChordates),for example. Let these groups serve as a

Cladograms 27basis for postulating novelties within the group of five organisms.

It is observed that the perch, lizard, cat, and mouse are similarin possessing jaws, bone, paired appendages, and three semicircu-lar canals. None of these features is present in the two outsidegroups, nor do they appear in other organisms that are not at thesame time suspected to be part of a larger group defined by thesimilarities characterizing the five organisms as a whole (chamberedheart, semicircular canals). Therefore, we can adopt the hypothesisthat the four organisms noted above form a subgroup.

These four organisms now constitute their own group for sub-sequent analysis, so the outgroups now consist of the lamprey, am-phioxus. and tunicates. Within the group of four. what might be asubgroup? The lizard, cat, and mouse have an amniote egg,whereas the perch does not. Upon comparison with the outgroups,the anamniote egg of the perch is seen to be fairly similar to theeggs found in the outgroups. On this basis, it can be postulated thatthe amniote egg is a novelty defining the lizard, cat, and mouse as asubgroup.

Finally, within these three organisms, is there a subgroup oftwo? The cat and mouse have three ear ossicles, hair, and mammaryglands. None of the outgroups (including now the perch) sharesthese similarities. Therefore, it can be postulated that the three simi-larities shared by the cat and mouse are evolutionary novelties defin-ing them as a subgroup.

This approach, called outgroup comparison, is similar to theapproach using developmental data in that both are methods usedto determine the relative, comparative distribution of observed simi-larities. These different levels of similarity form nested patternswhich in turn define nested sets of organisms. This comparativemethod is the general method of cladistic analysis, and the re-mainder of this chapter will explore the procedure in detail.

Some General Statements about Cladograms

The preceding example can be used to formulate some generalstatements about cladograms, all of which will be discussed in de-tail as the chapter unfolds.

First, cladograms are hierarchical in structure because the re-sults of the evolutionary process can be expressed in terms of a

28 Cladograms

branching diagram (figure 2.1). The aspect being expressed in sucha diagram is the nested pattern of evolutionary novelties, whichthemselves define nested sets of groups within groups. This hierar-chical structure can also be denoted by nested I ists of groups andsubgroups as follows (this constitutes a classification; see chapter 5),

GROUP: Lamprey, perch, lizard, cat, mouseSUBGROUP 1a: LampreySUBGROUP 1b: Perch, lizard, cat. mouseSUBGROUP 2a: PerchSUBGROUP 2b: Lizard, cat, mouseSUBGROUP 3a: LizardSUBGROUP 3b: Cat, mouseSUBGROUP 4a: CatSUBGROUP 4b: Mouse

Second, similarities, not differences, define the sets, theirnested configuration, and, by extension, the hierarchical levels of thecladogram. Furthermore, from assumptions about the evolutionaryprocess, the only kind of similarity defining groups of organisms isevolutionary novelty. It is the nested pattern of evolutionary noveltiesthat is being expressed by the cladogram and gives it its hierarchi-cal structure.

In practice, if a similarity is perceived, it can be assumed, as aworking hypothesis, to be an evolutionary novelty that defines agroup of organisms at some hierarchical level. Following the erec-tion of this hypothesis, various kinds of comparative evidence canbe used to refute it (e.p., developmental data, outgroup comparison).Thus, in the example given, the possession of jaws was postulated tobe an evolutionary novelty defining a group including the perch, liz-ard, cat, and mouse, whereas the possession of hair was hypothe-sized to be a novelty defining a subgroup at a lower hierarchicallevel (cat + mouse).

Given any assemblage of organisms, it is possible to define auniversal set and two or more subsets. The existence of a universalset may be taken as axiomatic, given that a worker wishes to con-struct a cladogram for a specific collection of organisms. On theother hand, most analyses are more open-ended in that the kinds oforganisms to be included are not necessarily predetermined. Never-theless, although not all biologists will necessarily reach agreement

Cladograms 29about the organisms to be included in any group or subgroup, theremust be agreement that these groups within groups can, in principle,be resolved.

Third, for a given universal set of N organisms, it takes at leastN - 1 similarities (novelties) to define the hierarchical structure ofany cladogram that might be constructed for those organisms (as-suming the cladogram to be fully resolved dichotomously). Thus,one similarity (novelty) is sufficient and necessary to define a groupof two or more organisms. In most situations, as in the example,groups will be defined by two or more coincident or congruent simi-larities, but one is sufficient and necessary (e.g., the amniote eggdefining the group lizard + cat + mouse).

The fourth principle is a corollary of the third, namely, a noveltydefining a group at one hierarchical level cannot define subgroupsat lower levels in the cladogram. The novelty, possession of jaws,defines a group comprising perch + lizard + cat + mouse, but thatsame similarity cannot define subgroups among these four orga-nisms.

This last principle also implies that the sharing of a postulatednovelty by two or more organisms is sufficient to include those orga-nisms within a group defined by that novelty, but the membership ofthis group cannot be exhaustively determined until the entire mem-bership of the universal set has been compared. After all, the postu-lated novelty may in fact define a group at a higher level, includingperhaps even the universal set. Thus, for any cladogram, it is neces-sary to specify the highest hierarchical level defined by each hy-pothesized evolutionary novelty.

Cladistic Analysis: Similarity, Synapomorphy,and Homology

Similarity, Synapomorphy, and Group-forming Procedures

Upon initiating a comparison of organisms, a systematist utilizesperceived similarities to choose those attributes (characters) of theorganisms that will then be used for the more detailed comparisonsleading to construction of a cladogram. Thus, the choice of charac-

30 CladogramsCladograms 31

dXV"'o

ters involves a perception of similarity, i.e.. a perception of compara-ble form and spatial relationships relative to other features of otherorganisms. In fact, more fundamental, subconscious perceptionsprobably precede even this elementary level of comparison, andperhaps the basic perceptions are those of "top," "bottom," "an-terior," "posterior," and so on. Eventually, at some stage in thethought process, the characters are sufficiently similar to be ac-cepted as the "same" character and are thus worthy of further, moredetailed comparison. There is, seemingly, an infinite regress in-volved in our perception of similarity, and thus perhaps it can beclaimed that biologists will compare that which is comparable andwill not compare that which is so different-again, in form or spatialrelationships to other structures-as to be termed "not the same, andnot worthy of further comparison." To a certain extent, it is a contra-diction in terms to say that one can compare things that are different;at some level at least. things must be sufficiently similar to invitecomparison even though they may eventually be designated "dif-ferent" and subsequently ignored.

As was previously noted, a major assumption of systematics isthat similarities and differences are a consequence of the evolu-tionary process. Some initial implications of that assumption will nowbe explored.

Consider two groups of organisms, X and Y, to constitute a uni-versal set and that in both some similarity has been recognized-acharacter is thus identified-but that the similarity is not one of iden-tity. Hence, in X this character can be denoted as a, in Y as a' (inconventional systematic terminology, a and a' would be character-states).' The two ways of specifying relationship-by common an-cestry and direct ancestor-descendant (see Chapter 1)-establishthe possibility of three evolutionary trees for these two groups (figure2.2a-c).

How do the Character-states of X and Y relate to these evolu-

tionary trees, and what can we infer about the evolutionary history ofthese character-states? The answer is diagrammed in figure 2.2d-g,Given that there are two groups, X and Y, and two types of rela-tionship, four hypotheses about the evolutionary history of thecharacter-states, a and a', can be specified. These four character-state trees can be generalized by converting them to a single clado-gram (figure 2.2h), which is the cladogram describing the informa-tion about set-membership contained in the three possible evolu-tionary trees. (a-c).

Because X and Yare products of some pattern of ancestry anddescent, it follows that so too must the character-states a and a' be aproduct of that ancestry and descent. The four Character-state trees(d-g) indicate only two possibilities, either a was transformed into a'(d, g), or a' was transformed into a (e,f). Thus, given two character-states, one of them can be considered primitive or plesiomor-phous-"close to" form (Hennig 1966), and the other derived orapomorphous-"away from" form. If a were primitive and a' derived,then only trees 2.2d and 2.2g could be admitted; if a' were prim-itive and a derived, then only trees 2,2e and 2.2f would be possible.

Comparisons of two groups such as this are of little significancein systematics because we can make no important generalizations:only one cladogram is possible. However, once the comparison isextended to three groups, important concepts can be introduced,

If three groups are being compared, four cladograms can be

b. X

ry

c. Y

1e. xe Ya'

Vf r

1. The terms "characte '" d" h diff t hi-. r an c aracter-state'" merely refer to similarities at two I rerenerarchlCal levels. Thus the character "feathers" is a common similarity of all birds, althoughspecific chal1lCter.states of the character "feathers" Ie.g., variation in color, texture, and pat-tern) would be . "I .. " t

SImI antics common to various groups of birds. At the same time, it IS apparenthat even. the character "feathers" could be considered a character-state, say within the ver-tebrates, If the systematist wen: considering the "character" to be the vertebrate integument(see also P1atnick 197811).

Thus, in.this diSCUSSion, the words "character" and "character-state" should be construedto mean relative levels of similarity within a given hierarchy.

32 Cladograms Cladograms 33

four possible trees involving X, Y, and Z upon which the character-state distributions can be mapped (figure 2.3f). The nested structureof all four trees (i.e., the subset, YZ, within the universal set, XYZ)can be represented by a single cladogram (figure 2.3c).

What is the aspect of these trees that enables Y and Z to beplaced together in a subset within the cladogram? It is that the an-cestor of Z-whether Y or an unspecified common ancestor of both Yand Z-can be postulated to possess a character-state, a', that isderived relative to the condition. a, present in X or its immediate an-cestor (whether unspecified or not). Groups Y and Z form a subsetbecause they share a derived (apomorphous) condition-in thiscase a' and its modification a"-relative to the primitive (plesiomor-phous) condition in X. The condition of sharing a derived character-state or a later stage in the transformation sequence of a derivedcharacter-state is termed synapomorphy. One of the most importantconcepts in comparative biology, we shall see that synapomorphyprovides the theoretical basis for constructing and testing clade-grams.

It should be apparent that within a particular cladogram of threegroups perhaps only one of the groups will possess a derivedcharacter-state. For example, in figure 2.4 if it were postulated thatcondition b had transformed to b', then for this cladogram thederived condition, b', would not define a subset. In this examplegroups X and Y share the postulated primitive condition b; a sharedprimitve Character-state is termed a symplesiomorphy. Group Z, onthe other hand, is characterized by a derived condition, b', notshared with X or Y and such a possession is termed an au-tapomorphy (a" in group Z would also be an autapomorphy, but a'in Y would not, because it is the shared condition, a', which isviewed as uniting Z and Y in a subset).

It can be appreciated that neither a symplesiomorphy nor an au-tapomorphy can be used to combine groups into subsets. Take, forexample, the simplest case, in which two groups, X and Y, share aCharacter-state, a', whereas a third group, Z, possesses character-state a (figure 2.5). If it is postulated that a has been transformed toa', then X and Y would comprise a subgroup, and the cladogram offigure 2.5a would be indicated. However, if the Character-state trans-formation were in the direction of a' to a, then this transformation iscompatible with anyone of the three cladograms (figure 2.5a-e)

c.

e.O_O'_ON

f.ZOU

IYo'

IXa

ZOU Yo'

)('\ Y(

Figure 2.3 (a-d) Fourpossiblecladogramsfor threegroups. (e)A postu-latedcharacter-statetransformationsequence.(f) Fourcharacter-statetreesfor X, Y, andZ assumingthe transformationsequenceof e. (Seetext)

constructed to describe the nested pattern of similarities (figure2.3a-d). Thecase in which no resolution is possible (figure 2.3d) willbe ignored for the present. Following the example of figure 2.2, let usassume all three groups have a similar, but nonidentical, charact~~such that it can be denoted as character-state a in X, a' in Y, and ainZ. .

On the basis of this single character it would not appear pOSSI-ble to form a subset within these three groups and thus resolve theuniversal set (XYZ) into one of the three cladograms (figure 2.3a-c).However, one of the fundamental insights of Hennig (1950, 1966)was to specify a method for forming subsets in situations such asthis. Hennig recognized that given some evolutionary change fromone Character-state, a, to another, a', the investigator may also beable to postulate a further change, from a' to a" (figure 2.3e). Somesimple, well-known examples are the sequences: five toes to fourtoes to three toes, fully developed eyes to eyes greatly reduced toeyes absent, and undifferentiated epidermis to scales to feathers.

If one has some basis to postulate a transformation fromCharacter-state a to a' and subsequently to a", then there are only

Cladograms 35

,

34 Cladograms

Zo"b' Yc'b Xob Homology and the Concept of Synapomorphy

The concept of homology has been cited frequently as the most im-portant idea in comparative biology. Simpson (1959c:287), for ex-ample, calls it "the first and greatest generalization of anatomy." Thehistory of homology is complex (Simpson 1959c), but prior to Dar-win, the term, when used, referred basically to similarities amongdifferent organisms. After Darwin, it became readily accepted thatfeatures which were called homologous obtained their similarity as aconsequence of common ancestry. In recent years, both conceptionsof homology have been recommended, although for most contempo-rary systematists the concept of homology impl ies the idea of com-mon ancestry. For example, Simpson (1961:78) defines homology as"resemblance due to inheritance from a common ancestry." Simp-son's view is that homology is similarity, but of an inherited kind.Bock (1977:881 and included references) goes further and assertsthat similarity should have nothing to do with the definition of ho-mology, rather the definition should be strictly phylogenetic: "Fea-tures (or conditions of a feature) in two or more organisms are homo-logous if they stem phylogenetically from the same feature (or thesame condition of the feature) in the immediate common ancestor ofthese organisms." Bock (1977:882) considers similarity to be the"test" of homology, and by this he presumably means that similarityis the "recognizing criterion" of homology. In claiming that "sharedsimilarity is the only valid empirical test of homology" (p. 882, italicsin original), Bock virtually adopts the logical position that homologyitself is an empirical concept, "knowable" only by observed similar-ity and not by any phylogenetic criteria. (This assumption will be dis-cussed shortly.)

A second group of systematists, particularly numerical tax-onomists, finds the phylogenetic definition of homology to be un-satisfactory. These definitions are, they believe, self-defeating be-cause "it is the primary purpose of the phylogenetic school ofsystematics [here they are really referring to the school of evolu-tionary systematics] to work out phylogenies, and for this they needhomologies that are not defined in terms of the conclusions theywish to reach" (Sneath and Sakal 1973:75). Sneath and Sakal haveidentified aspects of this definition that bother systematists con-cerned with methods to reconstruct phylogeny, They propose to re-

Figure 2.4 In thiscladogram,thederivedconditionb' does notdefine asubset(i.e.,Z + Y)but doesdefineZ alone;henceb' is termedanau-tapomorphy. Thesharingof a primitiveconditionb by Yand X is termedasymplesiomorphy.

c. b. c.xe: Yo' ze Yo· Zo Xo' Xo' Zo Yc'

YYYFigure 2.5 A character-statetransformationpostulatedto befroma' toa wouldbecompatiblewithany of the cladograms.Theconclusionis that symplesiomorphycannoldefine sub-sets.

because character-state a would be interpretable as an au-tapomorphy of Z. In this case, the character defined by thecharacter-states a and a' contains no information relevant for forminga hypothesis about a particular cladogram, because no synapo-morphy is postulated.

,,

L

36 Cladograms

turn to an "operational definition": "When we say that two charactersare operationally homologous, we imply that they are very muchalike in general and in particular" (p. 79). This definition, it wouldseem, is remarkably similar to Bock's "test" for homology.

The diversity of opinion noted above is testimony to the fact thathomology continues to occupy an important position in discussionsabout comparative theory and methodology. However, most of thisdiscourse has contributed only marginally to the important method-ological question of comparative biology: how do we reconstruct thehistory of life? On the one hand, the definitions of Simpson and Bockdo not contain intrinsically any implications for methodological gen-eralizations, and as such they create a paradox, noted by Sneathand Sokal. Are "homologies" to be used to reconstruct phylogeny ormust we, as the definitions of Simpson and Bock imply, have an es-tablished phylogeny before identifying homologies? If the former,then their definition would seem superfluous; if the latter, then thehomology concept would seem to be of little consequence for sys-tematics. On the other hand, the operational approach of Sneath andSokal is also unsuitable. Almost universally, homology has meantsomething more to most systematists than mere similarity; further-more, the Sokal and Sneath definition implies little about the goal ofcomparative biology, the reconstruction of the history of life.

These opposing viewpoints constitute what might be called the"problem of homology," and neither the schools of evolutionary sys-tematics nor numerical taxonomy have offered a conceptual solutionto this problem. The solution does not lie in proposing a concise,easily remembered definition of the word "homology," but rather inunderstanding the fairly simple relationship between inherited simi-larity on the one hand and the use of comparative analysis to dis-cover group-membership on the other: the solution to the "homologyproblem" is the concept of synapomorphy.

Homologous similarities are inferred inherited similarities thatdefine subsets of organisms at some hierarchical level within a uni-versal set of organisms. Viewed in this way, homology can be con-ceptualized simply as synapomorphy (including symplesiomorphy:see below). Indeed, there can be little doubtthat many systematists,particularly those who have formed the post-Darwinian tradition incomparative biology, have sensed the connection between ho-mology and synapomorphy, although it is only recently that this con-

I,

:1,

Cladograms 37

cept has been made explicit (see, for example, Hennig 1966; Nelson1970; Wiley 1975; Cracraft 1978; Gaffney 1979).

Synapomorphy subsumes the concept of symplesiomorphy.Synapomorphies are shared similarities (homologies) inherited froman immediate common ancestor; symplesiomorphies are sharedsimilarities (homologies) inherited from ancestors more remote thanthe immediate common ancestor. But a symplesiomorphy can beviewed as a similarity whose level of synapomorphy is unresolved,as more groups are added to the comparison, what was once con-sidered a symplesiomorphy becomes a synapomorphy defining aset of organisms at a higher hierarchical level. The question of im-portance is: at what level of the hierarchy does a shared similaritydefine a set? For example, the similarity of dense body hair cannotdefine a set including a mouse and a lion but excluding a human,because the shared possession of dense body hair between themouse and lion is, at that hierarchical level, a symplesiomorphy, thatis, there are other organisms with dense body hair excluded from theset. As the hierarchical level is increased by adding other groupswith hair, this similarity is eventually seen to define a set biologistscall Mammalia, and at this level hair is a shared derived similarity, asynapomorphy.

The concepts of synapomorphy and homology can be viewed inanother way, one which pertains to defining group membership. Aswill be developed in more detail later, the "test" for synapomorphy isnot empirically observed similarity (as was Bock's test for homology,noted above), but the congruence of other hypothesized synapomor-phies in defining sets of a cfadogram and, ultimately, the systema-tist's preference for that cladogram in contrast to alternatives thatmight have been chosen. For example, given the cladogram shownin figure 2.6a, the postulated synapomorphy a' defines a subset ASwithin the set ABC. However, if from other evidence the investigatorhad reason to prefer the cladogram of figure 2.6b, then the postu-lated "synapomorphy" does not define a set within the cladogram. Isthe similarity a', therefore. a synapomorphy? No, not in the sensethat a' defines a subset within the hierarchical level ABC. Then, howcan we explain the distribution of the similarity a' in A and B if thereis reason to prefer the cladogram of figure 2.6b? There are two ex-planations. First, the similarity a' may be a synapomorphy, but onedefining set ABC or some still larger set. If this were the case, condi-


A Bb.

C AA B C oo.

C B

a' ..._---ji a a'lo-~ 0'

Figure 2.6 (a) Character-statea' seemsto defineasetA + Bandthusmight be postulatedto be asynapomorphy.(b) If thereis sufficientevidencetoaccept thiscladogram,thena' can no longerbe asynapomorphyat this hierarchicallevel but mustei-ther bea primitivesimilarity(symplesiomorphy)or aconvergence.

Figure 2.7 A cladogramof four groups.A is thesister-groupof B,A + Bthe sister-groupof C, andA + B + Cthesister-group of D.Black rectan-gles signifydefining syna-pomorphtes.

Two additional concepts are associated with the idea that syna-pomorphies define nested sets within a cladogram. First, the coordi-nate subsets of any set defined by a synapomorphy are termedsister-groups. Therefore, in figure 2.7, group A is the sister-group ofB, group A + B the sister-group of C, and group A+ 8 +C the sister-group of D. The concept of sister-groups is relative and not absoluteand, as we shall see, is itself also a hypothesis to be tested. Itsapplication in any specific instance depends upon the number oftaxa included in the comparison (universal set) and the perceivedpattern of synapomorphies. In figure 2.7, if 8 had been omitted forsome reason, then groups A and C would be sister-groups united bysynapomorphy 2.

A second concept is to some extent one of classification(chapter 5) but it has its basis in set-definition and synapomorphy. Aset of organisms is said to be monophyletic if it includes the stemspecies (if known) and alf those kinds of organisms hypothesized tohave descended from it. It follows that the concept of monophyly isapplicable only to groups of two or more species (see Chapter 3). Animportant corollary of this definition is that monophyletic groups aredefined on the basis of one or more synapomorphies. The sets AB,ABC, and ABCD of figure 2.7 are monophyletic, but the set AC isnon monophyletic because B has been excluded even though a syn-apomorphy indicates its inclusion. Sets of organisms may have

tion a' in A and B would be a shared primitive similarity, and thecondition a in C would be derived, an autapomorphy. In order to de-termine this situation methodologically, the analysis must be ex-tended to include groups of organisms other than A, B, and C, andalso to include other features. Secondly, if for cladogram 2.Gb evi-dence could be presented that condition a is indeed primitive and a'derived (for example, through comparison with other groups of orga-nisms), then similarity a' would have to be interpreted as havingbeen derived independently. In this case, a' would be a convergent,or nonhomologous, similarity.

This example, then, implies that similarities may be consideredhomologous (synapomorphous, symplesiomorphous) or nonhomo-logous (convergent) depending on their ability to define subsetswithin a particular cladogram. If the point of reference (a cladogram)were to change, our view about what is or is not a homology wouldbe similarly altered. The notion that the concepts of "homology" and"nonhomology" are not empirically determined and not factual de-scriptions of observed similarity, but rather a consequence of set-definition, clarifies much of the confusion in the literature on ho-mology. Even if evolutionary connotations are imparted to the idea ofhomology, properties of set-definition continue to be essential cri-teria for applying the concept. This should not be too surprising,since the resolution of nested subsets is the primary concern ofcladistic analysis, and of comparative biology in general.

j

I,,,40 Cladograms Cladograms 41names and be part of the linnaean hierarchy, in which case mono-phyly comes within the purview of classification theory. (This will bediscussed more fully in chapter 5.)

Cladistic Analysis: Taxa and Characters

The Search for Nested Sets

The key methodological problem in comparative biology might beviewed as a search for synapomorphy. More specifically, the searchis for nested patterns of synapomorphy, and cladograms are hypoth-eses about those nested patterns.

Cladograms, and thus nested synapomorphy statements, have aproperty in common with all scientific hypotheses: that of prediction(see Platnick 1978b). The prediction is that synapomorphies will becongruent with one another, and this prediction is based on our ex-pectation that the evolutionary process has produced a single,unique history of life, and that changes in organisms and their fea-tures will conform to the pattern of this history more often than not.The essence of cladistic analysis, therefore, is the formulation of hy-potheses of synapomorphy and their nested pattern. Subsequentevaluation of hypotheses of synapomorphy and of the cladogram it-self rests with the prediction that future synapomorphy statementswill continue to be congruent with those of the cladogram. As weshall see, this implies a preference for the cladogram which max-imizes the congruence of individual synapomorphy statements. Thisview of cladistic analysis was impl lcit in the example of figure 2.1:

Lamprey, perch, lizard, cat. mouse: dorsal nerve cordPerch, lizard, cat, mouse: jawsLizard, cat. mouse: amniote eggCat, mouse: hair

There is the expectation that future synapomorphy statements(notochord, semicircular canals, paired appendages, three ear os-sicles. and so on) will conform to this nested pattern. If they do, thenour confidence grows in this c!adogram as a reliable representationof the pattern of life's history. If they do not, then perhaps an alterna-tive cladogram will represent that pattern better.

;,In this section, the practical problem of constructing cladograms willbe considered. Three broad topics are discussed: the kinds ofgroups of organisms that are subjected to a cladistic analysis, thekinds of characters that provide the basic data for the analysis, and,finally, the procedures that might be followed in formulating and test-ing hypotheses of synapomorphy patterns (the cladogram itself).

The remainder of this chapter is a "manual" on cladistic analy-sis only in a very restricted sense. No one description of methodscan lead investigators step by step through the countless problemsand decisions which arise in a real biological example. Our purpose,rather, is to provide the conceptual foundation for the investigator'sown efforts, and, by using examples, we trust the conceptual founda-tion will be brought closer to the real world situations confronted bypractical experience. Our own experiences, and those of our stu-dents, have convinced us the only way to fully understand the con-ceptual aspects of cladistic analysis is to apply the methodologicaltheory to a group of organisms. It is a process of trial and error, andit has rewards beyond the success of a finished cladogram: applyingthe methodology can also commit one to further understanding andclarification of the underlying concepts.

I1ji!,

The Elements of Cladograms: Kinds of Taxonomic Units

In the earlier biological example (figure 2.1) the groups of the setswere loosely termed "organisms." It seems doubtful this usage wasconfusing, since to most, if not all. biologists the five "organisms"would be recognized to be of a distinct kind, what are typicallycalled "species." Before proceeding to the details of cladistic analy-sis, the exact nature of these biological groups should be discussed.

The fundamental units of comparison in systematic studies aretaxa (singular, taxon). Taxa are defined as collections (sets) of indi-vidual organisms that are sufficiently distinct from other sets to begiven formal names and placed in the Linnaean hierarchy (see Mayr1969:4-5; Simpson 1961 :19). For example, the taxon name Columbarefers to a genus (a category of the Linnaean hierarchy) of pigeons,the taxon name Columba Iivia to a species of pigeon, and the taxon

)ijl'

42 Cladograms

name Columbidae to the family including all pigeons. The lowest hi-erarchical category within which formal taxa are sometimes clas-sified is the subspecies, thus Columba livia Jivia (see chapter 5).

If systematists were asked to name the primary taxonomic levelat which systematic analyses are undertaken, most would answerthat of the species. How one might identify a collection of individualsas belonging to a species or whether a precise delimitation of spe-cies is in fact necessary for systematic work to proceed are ques-tions that have been vigorously debated (Mayr 1963, 1969; Sokaland Crovello 1970; Sneath and Soka11973; Wiley 1978). We discussthese problems in chapter 3. It is important to note that, althoughspecies are commonly defined in terms of reproductive criteria, rela-tively few practicing systematists directly use these criteria when ap-plying the species concept in systematic studies (see chapter 3).

Our immediate task at this point is to specify the kinds of taxasystematists might want to investigate using cladistic analysis. Theone basic assumption is that each included taxon composed of twoor more species is strictly monophyletic." In practical terms thismeans the systematist has some reason to accept a corroboratedhypothesis of their monophyly, and the reason, as we have notedpreviously, is evidence of congruent synapomorphies. A hypothesisof monophyly for any supraspecific taxon must be based on synapo-morphous similarity. In general, most species-level taxa will also bedefined in terms of synapomorphy, but, as will be discussed inchapter 3 there may be cases in which this need not be a require-ment. Cladistic analysis can be performed, therefore, usingtaxa of species-level rank or higher.

2. 1be concept of monophyly is applicable only 10 groups of two or ~ species. Obviously,when speaking of single species, the concept of monophyly is inappropriate since stem (ances-tral) species and descendant species are necessarily excluded (rom the discussion.

This observation raises a more general point relevant al this juncture: the ancestral specieswithin any monophyletic group, if present in a sample and correctly analyzed, will be dia-grammed as a monotypic serer-group of the remainder of the species contained in that group(see chapter 5). If the ancestral species is raised to some higher rank (see chapter 5) for etas-sificatory purposes, e.g., placed in a separate family, Ihen that family will not be monophyletic.In general, taxa of any rank which consist solely of a single species ancestral to other specieswithin a monophyletic group cannot themselves be monophyletic. Every monophyletic group,under Hennig's definition, will contain, as a minimum, one such example {i.e., a stem species).As a corollary, such nonmonophyletic taxa will not possess any Wliquely derived characters(autapomorphies).

Cladograms 43

Similarity and Characters

Comparison among taxa implies the search for similarity, and clado-grams are general expressions of the patterns of those similarities.This statement forms the conceptual basis for understanding the pro-cess of constructing and testing cladograms: that process is merelythe systematist's method of arriving at general statements about sim-ilarity. In one sense, the characters of organisms are the carriers ofsimilarity. This view emphasizes an important point it is not the char-acters themselves that are of significance, but rather the similarities.Something more than a semantic issue is involved here. There hasbeen considerable discourse within systematics about charactersand character-states, reliable versus unreliable characters, the"weight" of characters, and so on, but little of this literature mentionsor emphasizes similarity. These discussions of characters have dis-tanced themselves from the central concept of comparison: similarity.The ensuing pages, therefore, focus on similarity and its role in com-parison and the construction of cladograms.

Some general statements about similarity were introduced ear-lier in the chapter. All comparison-biological or not-involves thecorrespondence of similarities. Arms could be compared withheads, yet the basis for that comparison still remains a search forsimilarity; thus, both arms and heads might be seen to be compara-ble in having similar tissues, cells, chemical characteristics, and soforth. This comparison would resolve itself to comparing not thecharacters called "arms" and "heads," but rather the characterscalled "tissues," "cells," and "chemical characteristics." However,similarities must be named, and it is these names that we call "attri-butes" or "characters." If the process of comparison has some gen-eral significance, it is that as soon as a character is designated, thenso too is a similarity: in comparative biology the words are concep-tual. if not entirely operational, synonyms-a character is a name fora similarity. Thus, "paired appendages" may refer to a similarity thatis common to a set of vertebrates, "differentiated paired appen-dages" to a similarity of a more restrictive subset, and a "pentadac-tyl foot" to a similarity of a still more restrictive subset. This suggeststhat discussions in the systematic literature about characters andcharacter-states are really about similarity. A "pentadactyl appen-dage" as a character, and "four-toed" or "three-toed" conditions as

Cladograms 4544 Cladograms

character-states would seem to have little intrinsic usefulness incomparison if they did not refer to the similarities implied by thosenames. A further implication is that the similarities are defining setsand sub.sets.The implication itself is fascinating-name an attribute,an~ ~ttrlbute, and the working hypothesis is that it is a similaritydefining a set. The set may not be "natural" in an evolutionary sense("eyes." of vertebrates grouped with "eyes" of arthropods) but, as willbe pointed out in what follows, problems of this sort are resolved bycomparing "characters" defining higher hierarchical levels-the es-sence of testing cladograms. In any case, more restrictive "charac-ters" (read "similarities") can be seen to define nested sets., .What kinds of similarities observed in taxa can be used in cla-dlS.tlCanalysis? The ~nswer is that all observed similarities may beof In.ter~st.and potentially useful. Logically, those similarities whichare m~rmslc .to the species comprising the taxa would seem to bemor~ I~medlately useful for cladistic analysis than those which areextrinsic to the species. By intrinsic, we mean similarities of anat-omy, . pnysioloqy, biochemistry, behavior, genetics, and develop-m~nt, by extnnsfo, we mean the similarities among species thatmight be,o~served in their distribution in space and time. (Intrinsicand ~xtrrnslc "properties" are also treated in the next chapter onspecies.}

. ~~ato.mic~1similarities are clearly the most commonly used si-rruladties 10 biological comparison. Some of these were illustrated inthe cladogra~ at the beginning of the chapter (figure 2.1); more will.be inClude? In later examples. Anatomical similarities predominateIn systematic analysis primarily because they are the most easily ob-ser~ed, a~d.m~~t can be studied in preserved organisms. Nenana-tomtcal similarities, while important in specific cases, have beenless ~Idely adopted in systematics owing generally to a lack of com-~a.ratlve data. Many nonanatomical similarities must be observed inliVing organisms or under experimental conditions requiring veryelaborat.e and expensive techniques. Nevertheless, through theyears biologists have learned a great deal about the comparativeaspect~ of these kinds of similarities. An example of one of the moreextensive attempts at using nonanatomical similarities to constructand t~st cladograms is Lavtrup's (1977) analysis of the origin andevolution of the vertebrates.

Physiological similarities have seldom been adopted in system-

atic studies, owing mainly to a lack of comparative information. Per-haps the best summary of physiological similarities within animalscan be found in textbooks of comparative physiology (e.g., Prosserand Brown 1965). A few examples will suffice to illustrate the use ofphysiological similarities in defining groups of organisms.

Almost all animals (and some plants) possess respiratory pig-ments that bind oxygen and transport it from a respiratory surface toother body tissues. In most organisms this pigment is hemoglobin,but other kinds of pigments also exist. One of these, chlorocruorln, isfound only in the plasma of species in four families of polychaeteworms, Sabellidae, Serpulidae, Chlorhaemidae, and Ampharetidae(Prosser and Brown 1965). Although some species of Serpul idaealso have hemoglobin, shared possession of chlorocruorin defines agroup including the four families.

Several interesting similarities defining groups of organisms areassociated with the physiology of digestion. Virtually all animalsstudied to date possess amylases for the digestion of starches, andthe possession of this physiological property defines a group includ-ing all animal life. Shared similarities of other digestive enzymesalso define groups: pepsin, for example, is found only in the ver-tebrates (Prosser and Brown 1965).

Physiological similarities will undoubtedly be of great value tosystematists in the future. Several difficulties will always accompanytheir use, but by no means are these difficulties restricted to physio-logical similarities. First, compared to anatomical similarities, rela-tively little is known about the occurrence of any given similaritythroughout the taxon under study. Second, statements that a certainphysiological process is or is not present or whether it is similaramong taxa are often made only on the basis of functional observa-tions, and detailed studies of the precise composition of chemicalproducts or the biochemical pathways involved in their elaborationfrequently have not been undertaken. Thus it is often problematicalthat the "same" or "similar" physiological processes are being ob-served or compared.

Biochemical similarities, to the extent that they can be distin-guished from physiological similarities, have proven helpful to sys-tematists in defining the groups of cladograms. Lavtrup (1977), forexample, lists a large number of biochemical similarities of molluscsand vertebrates said to define a group distinct from other animals:

..

46 Cladograms

these similarities include the distribution of specific qlycosaminoqly-cans (hyal.uronat~, chondroitin, and keratin sulfate), epidermal struc-tural proteins (epidermin), and cartilage (also found in protochordategroups), among others.

Wi~hthe ~x~ep~ionof the findings of comparative anatomy, noother kind of :Imllanty has been utilized by systematists as much as:~:tof b~havlor. Most behavioral similarities have been observed at

specte~ level, and the emphasis most often has been on thedemonstration of how one ..... . . specIes IS distinct from another in behav-torat characterlstlcs Expressed somewhat differently and perhaps~ore correctly,. th~~e investigations use behavioral similarity to de-me ~rohup~of m~lv~du~1organisms which are then called species.

e ~~,oral slmllanties can also be used to form groups of su-praspectttc taxa In his ext .havior Kahl 1971 .enslve comparative studies of stork be-ties def . ( , 1972) discovered a series of behavioral aimilan-

e mlng a sub t ithi, I d se WI 10 the stork family Ciconiidae thatme u es the four species of d t ' ,ibis, I. Jeucocephalus a wo~ s arks (Mycteria americana, Ibissimilarities in sev I' nd I. cmereus). These four species showplay preening baIera. stereotyped behavioral patterns termed dis-unable to find' beh:~.cln~ ~os.tur~"and gaping (figure 2.8). Kahl wasspecies of wood storl~;~ SImilarities defining subsets within the four

Gorman (1968) has em I db' .play behaviorofth . p.oye ehavicra! SImilarities in the dis-

e spectas In the Iizard A'sets within the gen Th genus notis to define sub-us. e roquet spec! . . .species, is defined by th . es-group, conststtnq of SIX

e possession of a shared similarity, a dis-

Cladograms 47

• •

Flgure 2.9 A behavioralsimilarity in dewlapdisplay inAno/is lizardsused to define variousspecies-groups.(FromGorman1968.)

play in which the dewlap (a fold of skin trom the throat) is extendedand then held in this position for the entire display; other species ofAnoUs rapidly pump the dewlap in and out during the display (figure2,9),

An increasing number of systematists are using "genetic" or"molecular" data to form sets of taxa (see Ayala 1976, and Dobz-hansky et al. 1977, for an introduction to the types of work being un-dertaken). These investigations include the analysis of karyotypes(chromosome number and structure), indirect techniques of estimat-ing "genetic distances" among organisms (DNA hybridization, elec-trophoresis, and immunological similarity), or the "direct" compari-son of the amino acid sequences of a given protein among varioustaxa (even this may be considered an "indirect" genetic technique,in the sense that the results only provide an estimate of the similarityin nucleotide sequences).

Most of the studies on karyotypes, chromosome structure, andelectrophoretic mobility are directed at forming groups of organismsat the species level. For example, Greenbaum and Baker (1976)analyzed the chromosome number in populations of two mainlandspecies of bats of the genus Macrotus. One species, M. californicus,has a diploid chromosome number of 40, whereas the other species,M. waterhousii, has a diploid number of 46. Individuals in popula-tions on the islands of Jamaica and Haiti were found to have a

A B,Flgure 2.8 Behavioralsimilaritiesi d' C.storks:(a)gaping; (b) balancin o~u:~lay patterns usedto definewood1972, by permissionof theBrit?"'P0 '!he,(c)~lsplay preening.(FromKahl

rru oloqtsts' Union.)

48 Cladograms

diploid number of 46 and therefore were included in the species M.waterhousii.3 This is only one example of many that could be givento illustrate the use of karyotypes to define species-level taxa.

Techniques such as DNA hybridization, starch-gel elec-trophoresis, or immunological reactions can be used to estimate thedegree of genetic similarity among species. The methods of analysisadopted in these studies, although routinely used to producebranching diagrams depicting concepts of relationship, are qualita-tively different from the methods discussed in this book, in that theymake no attempt to distinguish between primitive and derived simi-larities. Instead, similarity (or dissimilarity) matrices are produced;for example, a matrix might show the degree to which species crossreact immunologically with one another. These numerical data ma-trices are then analyzed by various clustering techniques to yieldbranching diagrams. This approach is analogous methodologicallyto the con~ept of general "overall similarity" adopted by numericaltaxonomy In that similarity is expressed as a numerical value in adata matrix, and then the latter is used to produce a branching dia-gram of some kind. We further discuss numerical taxonomic tech-niq~es in chapter 5; we merely note here that cladistic analysis asde!lned a~d. described in this book is theoretically and methodol-ogically distinct from the techniques mentioned above.

In recent years, biochemists have determined amino acidsequences for homologous proteins in different taxa, and compari-son of these s.equences has permitted cladistic hypotheses to beproposed. Unlike the numerical "similarity" measures yielded byDNA hybridization or im I' '. .muno oqical reaction techniques compan-so.ns. arnonq amino acid sequences can be evaluated in terms ofpr.lmltlve and derived similarities. For example, in cytochrome c pro-teins humans and rhesu k iff. . s mon eys d! er by only one amino acid: atposition 66 the human h . I '.' as ISO eucrne and the rhesus monkey hasthreonine .. Examination of nonprimate mammals indicates thatth~eonlne IS widespread. Therefore, it can be postulated that withinprimates, at position 66 threonine is primitive and isoleucine deri-ved. Crowson (1972) ha di .. s rscussed In more detail a procedure forcompanng amino acid sequences by this approach.

3. For this to be a valid peoced f .number of 46 is ectuall cieri lIre,O course, It would have to be shown that a chromosometion might not necessJat beloved(Synapomorphou~)condition, otherwise the Jamaican pcpula-

y ng to M. WDl~rhousll.

Cladograms 49

Table 2,1 DevelopmentalSimilarities Used to Define the Two Currently Ac-cepted Animal Superphyla"

Annelid superphylum Echinoderm superphylum

Spiral cleavageBlastopore= mouthscneccoeuc coelomDeterminatecleavageDelaminatesnervoussystemEctodermalskeletonTrochophorelarva

Radial cleavageBlastopore= anusEnterocoeliccoelomIndeterminatecleavageInvaginatesnervoussystemMesodermalskeletonPjuteus-type larva

"From tevtrup (1977:60). after Kerkut(1960)

The final kind of intrinsic similarity to be illustrated is that of de-velopment. Similarities in development have played an importantrole, historically and scientifically, in the definition of the majorgroups of animals. Kerkut (1960), for example, discusses many de-velopmental similarities indicating a major division of the animalkingdom into the annelid superphylum and the echinoderm super-phylum; the similarities defining these two postulated sets are shownin table 2.1. Although Levtrup (1977) has questioned the interpreta-tion of these observed similarities, and thus the definition of thegroups themselves, the similarities exemplify the kinds of evidenceto be found in developmental patterns. The appl ication of develop-mental similarities is discussed in more detail later in the chapter.

To many practicing systematists, intrinsic similarities are viewedas a direct manifestation of an organism's genotype. To be sure, theexpression of most of these similarities, perhaps to be equated withthe concept of phenotype, may well be under genetic control. Thepoint is, however, that with very few exceptions systematists do nothave knowledge about the genetics (or developmental expression ofthese gene products) of the similarities they are comparing. nor doesthis information appear to be crucial. Cladistic analysis functions todiscriminate pattern-presumably the consequence of an evolu-tionary process-not the underlying causal mechanisms of that pat-tern.

Extrinsic features of taxa are those associated with their distnbu-tion in space and time. Although extrinsic similarities among taxaare certainly of interest to systematists, the extent to which they bearupon cladistic analysis is slight. Cladograms are hypotheses about

i--

so Cfadograms

the distribution of intrinsic . '.correlation ith Properties: extnnslc features lack a 1'1

WI nested sets of tax It .extrinsic s;milaritie b B., would appear, therefore, thatOnly intrinsic I I s are y nature different from intrinsic similarities.

ea ures are discuss d f rth .struction01 cl d e u er WIthrespect to the ceo-a ograms.

Cladograms 51

o. ~ c.

\{yyFigure 2.10 Three alternative three-taxon statement clade-grams. (See text)

Cladistic Analy" • HIS. ypotheses and Their Evaluation cladistic methodology, it has meant: "taxon A and taxon B shared amore recent common ancestor than either did with taxon C," and inthis sense the statement is one of phylogenetic relationship sensuHennig (1966). The ensuing discussion will admit this interpretation,but something more specific will also be implied: taxa A and B sharea pattern of synapomorphy unique to themselves and nested within amore encompassing synapomorphy pattern defining taxa A, B, andC. Hence in this sense relationship ultimately refers to a hypothe-sis of nested synapomorphy.

The hypothesis stated above, that taxon A is more closely re-lated to taxon B than either is to taxon C (figure 2.10a), can beproposed in the absence of any supporting evidence, for it is im-plied in the formulation of such a cladogram that this hypothesisstands in competition with alternative hypotheses. In the case ofthree taxa, there are two completely dichotomous alternatives (figure2.1Gb,c), and it is the purpose of cladistic analysis to provide amethod to chose among the three hypotheses.

How might a cladistic (cladogram) hypothesis be formulated?What taxa are to be included? It can be assumed at this point that asystematist undertaking a cladistic analysis will have a particularproblem in mind with respect to a specific group of organisms. Ini-tially, the taxa will be those delineated in prior systematic treatmentsof the group. It is important that all taxa included in a cladistic analy-sis be monophyletic, i.e.. be definable by synapomorphous featuresof their own. Thus recognition of these taxa is equivalent to formulat-ing a series of hypotheses of monophyly. In most instances, system-atic tradition will have accepted the monophyly of the basic taxa, inwhich case the next concern is to develop a hypothesis of rela-tionships for them. If, on the other hand, this systematic tradition hasindicated uncertainty about the monophyly of one or more of these

There are two major com . .formulation of a h th .ponents to cladistic analysis. First is the

ypo esrs about t d .the evaluation of this h . nes ~ sets of taxa, and second ISthat might be c 'd ypotheeis relative to alternative hypotheses. ensI ered These treal science in g I . wo components are those of empir-. enera ' one beg· ith .which may sugge I' InSWI a question or a problem,

s an array of pas ·bl I·must in turn be e I Sl e so utlons (hypotheses) thatTh . va uated (tested).e primary question of '.

tory of life? The . systematics IS:what has been the his-. Viewdeveloped' thi hare hypotheses ab thin ISc apter is that cladograms

about synapomorp~u t e. pattern of that history, and statementsCladograms.Inwhat~oir:~de th~ basis for evaluating alternativecprrent of Cladistic h we Will focus on three issues: the devel-th ypotheses the I·e evaluationof clad' I' ' ana YSISof synapomorphy, and

ISIChypotheses.

Development of CladisticA cl d' . HyPOtheses (Cladograms)

a IStlChyp th . .r. a eSISISa clad .'cnsntos (nested set) ogram specifying a pattern of rela-nested Pattern of s s among taxa that is a consequence of aha k ynapomorphy II .ve nowledge of . IS not necessary of course topr! synapomorphie'"lor to proposing I s or their possible nested patternits a c edocram bown and is eventuau .' ecause the hypothesis stands onanotherway, a cladistic YhSubjectto rejection or corroboration. Put~onoPhY,I~,(see Gaffney l~i~thesl.s (c'ecocram) is .a.hypothesis ofat says. taxonA is ). In Its Simplest form It ISa statement

taxon C" (figure 2 1O;)orTeh~'oseIYrelated to taxon B than either is toment ca b . . IS stateme t II' n e variously . tn, ca ed a three-taxon state-

rnerpreteo. In much of the literature on

-

... ------------- 2 -q!lll-- n

52 CJadograms

taxa, it will be necessary to investigate this problem further beforeatte.mpting to group these taxa into larger groups. In either case,thebasic taxa of a cladistic hypothesis can be viewed as a seriesof!Ower-Ievel.hypotheses, themselves subject to testing.

Once It has been accepted that certain basic taxa will formthe:Ubject matt.er for a subsequent cladistic analysis, the next step ishe formulation of alternative hypotheses that can be tested. Al-though there are no established formal rules for erecting thesehy-potheses, systematists have traditionally followed only a few meth-ods,

Cladograms S3

3. Hypotheses suggested by general similarity Many clusters of taxaare suggested by aspects of general similarity. Historically, generalsimilarity has been the principle means used by systematists togroup taxa in terms of relationships and to form classifications. Be-cause general similarity includes a component of symplesiomorphy,it cannot be used, in itself, to define monophyletic groups Nonethe-less, general similarity sometimes can lead to the formation of hy-potheses that can later be tested by synapomorphy. If fact, as isotten the case, an observed similarity in two or more taxa can serveas a preliminary reason for hypothesizing monophyly, and then thesimilarity can be compared more broadly in other taxa in order toassess whether it is a primitive or derived condition (a method thatdoes not differ materially from the next approach).

1. Hypotheses suggested hy prior systematic work To a greater orlesser extent, all systematic analyses rely on the hypotheses of priors~stematic work. Indeed, in those groups previously receiving con-siderable attention, most, if not all, of the possible hypotheses mayalready have been proposed and discussed. Although many of theseInvestigations may not have been undertaken from a cladistic view-point, the hypotheses generated can often serve as a starting pointfor a cladistic analysis nonetheless. For example, in his cladisticstudy, of t~e interrelationships of several major groups of actrrcp-t~rygl~n fishes. Wiley (1976) identified four hypotheses of rela-tionships that had been previously suggested, and then proceededto evaluate each (his study wi/! be discussed further later in thechapter).

Unless one is investigating a poorly studied group or attemptingto deCipher the relationships of newly described taxa, prior system-atic work will probably be a major source of hypotheses to be test~d.Even so, there are at least three alternative methods for generatmgthese hypotheses.

4. Hypotheses of synapomorphy In many cladistic analyses, hypothe-ses of synapomorphy are used directly to formulate hypotheses ofmonophyly. Those latter hypotheses are in turn tested by additionalsynapomorphies that might be postulated. In following this proce-dure, the line between hypothesis formulation and hypothesis testingbecomes blurred. There is usually a prior acceptance 01 basic taxa,similarities are compared, and, on the basis of one or more criteria.synapomorphies are postulated These synapomorphies lead di-rectly to one or more phylogenetic hypotheses Examples of thisapproach will be presented shortly.

2. Hypotheses generated by enumeration Given a specific number oftaxa .to be analyzed, the systematist might I ist all the cladogr~msPOSSible for those taxa and then subject each hypothesis to testing.If one were concerned with a very small number of taxa (say, three orfour), then this procedure might prove feasible. But for 5 taxa 32 fullyresolved cladograms are possible and the number increases mar-kedly thereafter (e.g., there are 10395 possible cladograms for 7taxa). Clearly, enumeration will not be an important method for gen~erating hypotheses unless more complex situations are resolved tostatements of the three-taxon type (figure 2.10).

Analysis of Synapomcrphy

Without doubt. the most important part of any cladistic study IS theanalysis of synapomorphy Not only is the oetirmtanon of the basictaxa dependent upon an assessment of synapomorphy but so too arethe eventual interrelationships oeterrnmec for those taxa Hence.there is ample reason for so many systematists to have expendedconsiderable time and thought to ttus subject Most of ttus literaturefalls under the rubric of character enetvsss. and many workers haveprovided "criteria," "rules." and so forth to recognize pnrninve-derived conditions of similarities. But if the rationale for cladisticanalysis is accepted, that is, if it is agreed that cladograms areexpressions of nested patterns of synapomorphy, then the problemof character analysis is one of resolving character distributions. Nomatter what kinds of similarities are used-anatomical, physiolog-

;;i

t -.i.-:,,,

Cladograms SS

may sometimes be difficult to guard against a preconceived notionof polarity (see also Gaffney 1979). It seems best to avoid unneces-sary assumptions by simply identifying character-states and then fol-lowing procedures to postulate which are relatively primitive andwhich relatively derived. To repeat: "characters" and "character-sta-tes" merely signify different hierarchical levels of synapomorphoussimilarity.

The concept of nested sets of synapomorphy makes sense onlyif it is assumed that change has taken place, that somehow a groupof organisms, itself a subset of a larger set of organisms, is charac-terized by a change (a transformation) in one or more features thatcan be said to characterize the group as a whole. Thus, in analyzingsynapomorphy, the problem is to develop methods to postulate thesequence of change. Basically, systematists have recognized threeways to identify a transformation sequence: (a) the application ofpaleontology, (b) the appl ication of ontogeny, and (c) the applicationof outgroup comparison. We will now discuss each of these in order.

54 Cladograms

ical, behavioral, biochemical, developmental, or genetic-synapomorphy patterns are resolved by nesting more restrictedstatements of similarity within more general ones.

The analysis of synapomorphy begins with the recognition of asimilarity-a postulated homology-and its methodological designa-tion as a character. Concomitant with this is the recognition that sucha character exhibits some variability in terms of discrete character-states distributed in some pattern among the basic taxa. Somewriters have suggested that these character-states can be arrangedin a morphocline, a sequence presumed to reflect the probablepathway (not direction) of change among the character-states (Mas-lin 1952: Schaeffer, Hecht, and Eldredge 1972; Ross 1974). The rele-vant problem thus becomes the establishment of the morphocline'spolarity, i.e., the determination of the primitive and derived ends ofthe morphocline. For example, given the observation that fossil andRecent horses have two toes, three toes, and one toe, the morpho-cline one-toed--two-toed-three-toed might be established asmost reasonable. Subsequent analysis might suggest the polarity tobe three-toed - two-toed _ one-toed, although if one ignores theidea of a morphocline other possibilities clearly can be considered(figure 2.11). (It also could be suggested that the three character-states might have been derived from some other, unknowncharacter-state, but such an assumption would be ad hoc.)

Perhaps once character-states are delimited, the perception of amorphocline may be inevitable in the thought process. However, theestablishment of a morphocline may be more heuristic than neces-sary, and if all possible polarity sequences are not considered, it

a. 1-2-3 b.1-3-2

1. The App6cation of Paleontology A traditional assumptionof paleontological practice has been that fossil taxa are the keyto the history of life, and that, in their absence, all knowledgeabout relationships is speculative at best. Not only has this view-point concerned itself with the identification of possible ancestral-descendant sequences of taxa but also with the determination ofprimitive-derived sequences of character transformations. Indeed,this paleontological viewpoint can be characterized quite simply:knowledge of phylogeny is obtainable only through direct empiricalobservation of the historical record. Among the recent pronounce-ments on this subject is the following passage from Gingerich andSchoeninger (1977:488), who speak for a tradition extending backwell into the nineteenth century:

d. e. These basic ideas about phyletic evolution, like the basic principles ofmechanics, could not be predicted from theory alone-they are essen-tially empirical, and it is only with an adequate fossil record that anyreal understanding can be gained of the importance of size increase,irreversibility, or parallelism. . There is, in the absence of an ade-quate fossil record, no way to be certain which characteristics of an or-ganism are primitive and which are derived, or which evolved in-dependently in different lineages. Knowledge of both is essential toreconstruct a phylogeny on the basis of living forms alone.

g.I. 3-2

-1Figure 2.11 Possible character-Slate-transformation sequences for the threecharacter-states (1) one-toed, (2) two-toed,and (3) three-toed.


Other paleontologists, to be sure, do not necessari Iy expressthemselves in as positive a manner as Gingerich and Schoeninger.Nevertheless, within paleontology, data from fossils are generallyconsidered to be useful in determining primitive-derived sequences.Typically, this view is expressed as follows:

Dabcdef'gh

Cabcd'e'fghFor groups with an extensive fossil record, the character state first ap-

pearing in the record is likely the ancestral state. (Harper 1976:185)

One should have sound biological reasons for hypothesizing a charac-ter state to be primitive when it appears late in the fossil record, andwhen at the same time taxa with what are identified as advancedstates abound in earlier strata. It is therefore expected that theprimitive, rather than advanced, character states should occur withgreater frequency at earlier dates. (Szalay 1977:17)

Ao'b'c'd'e fq'h

Bn'b'c'd'efqh'

The general notion that the fossil record can tell us whichcharacter-states ,ar: primitive and which derived is appealing. How-e~er, most sophisticated paleontologists who basically support thisvlew.(e:g", Harper, quoted above) recommend using the fossil recordonly If It IS relatively dense. Other paleontologists, on the other hand,are .altogether scepttcal: "The fossil record for most groups of or-~anlsms IS too mcomplete to allow the assumption that relative stra-tigraphic position is necessarily indicative of morphocline polarity"(Schaeffer, Hecht, and Eldredge 1972:37).

Perhaps a central problem within the trad itional paleontolog icatmethod has been the tendency to equate a chronocl ine-thesequential distribution of character-states through time-with an an-cestral-descendant sequence of taxa possessing those character-sta.tes. Moreover, allegiance to the fossil record for determining po-larity of character-states necessari Iy requires acceptance of the fol-lowmq assumptions: (a) that the record being used is sufficientlycomplete to make a determination of polarity (i.e., additional fossilsshould not mate~ially influence the determination), and (b) that moreor, I~s.s aXiomatically, those character-states occurring earlier arep~I~I.tlve, The acceptance of both assumptions seems to be a mattero aith, and perhaps the general "experience" of the investigator.But, clearly, both assumptions will always be open to doubt, andhow are we to say Whether they are valid for any particular case?

That there can be no prior reason against expecting primitiveCharacter-states later in the record and derived states earlier is illus-

Figure 2.12 Four taxa (A-D) distributed in three different stratigraphic timeintervals. Each taxon has eight characters, derived character-states beingindicated by prime marks. This hypothetical example, which may not be un-common, illustrates the danger of assuming that character-states occurringearlier in the fossil record are primitive. (See text.)

trated in figure 2.12. In this example, four species of a monophyleticgroup (A-D) are distributed through time (the reader is invited tosupply any time scale desired) and, on the basis of synapomorphypatterns, are related as shown, The lineage A + B is characterized bythe possession of a number of derived character-states (a', b', c'),whereas species C and 0, which are sampled from a later time, aremarkedly more primitive in morphology. For some character-states(e-e' and f_f'), the fossil record would give a true picture of thepolarity, but for most others that picture would be misleading.

Although this example is didactic in nature, it is neverthelesseasy to see that each lineage must be characterized by one or morederived character-states, and thus whenever sister-species have un-equal survivorship in the fossil record, it is an expectation of thephylogenetic process that some taxa occurring earlier in the fossilrecord will possess derived character-states relative to those occur-

;'i,

.........


ring later in the record." Because unequal survivorship of sister-species is a common phenomenon, the early occurrence of derivedcharacter-states should also be common, It is also easy to see thatthe more incomplete the fossil record, the less confident we can bein using that record for determining polarity.

Does this argue against using paleontological data in the analy-sis of synapomorphy? No. not entirely. Rather, it is an argumentagainst the axiomatic acceptance of the traditional paleontologicalapproach. If other methods of analysis, for example outgroup com-parison, are unable to provide a sufficiently clear indication of polar-ity for one or more characters, then in such cases, paleontolog icaldata might be used, with caution, to hypothesize a primitive-derivedsequence (see also Delscn 1977). This hypothesis itself will be sub-ject to testing when alternative cladograms are compared andevaluated (see below). Even so, in most studies involving fossil taxa,outgroup compa~isonwill yield a more reliable estimation of polaritysequences and IS, therefore, to be preferred over a strict adherenceto the occurrence of character-states in stratigraphic sequences. Ifboth outgroup comparison and stratigraphic occurrence indicate thesame hypothesis of synapomorphy, one's confidence in the analysisIS understandably increased.

Historically ontogenetic observations have been considereduseful for determining the course of phylogeny in two ways. First.there is the straightforward Haeckelian viewpoint that "ontogeny re-capitulates phylogeny," a statement known as the biogenetic law. Inthis view, the ontogenetic stages of a descendant are said to tracethe sequence of adult ancestors (the reader is referred to Gould1977a, for a historical review and detailed characterization). Thus,by "reading" ontogeny, it is said, phylogeny can be reconstructed.This strict interpretation of recapitulation has long since been re-jected by modern biology, yet the notion that each individual some-how has its phylogeny "locked up" in its development has continuedto intrigue systematists.

The second way in which ontogenetic data have been applied tophylogenetic reconstruction is actually much older than the Haecke-lian formulation. In the late 1820s the German comparative embry-ologist K. von Baer generalized the results of his detailed studies ina set of rules or laws." These have been expressed by deBeer(1948:3) as follows:

1 In development from the egg the general characters appear beforethe special characters.

2. From the more general characters the less general and finally thespecial characters are developed.

3. During its development an animal departs more and more from theform of other animals.

4. The young stages in the development of an animal are not like theadult stages of other animals lower down on the scale, but are likethe young stages of those animals.

It is doubtful that anyone since von Baer has made as importanta contribution conceptually to the question of the parallel betweenontogeny and phylogeny-"the most important words in the history ofembryology" (Gould 1977a:56). Indeed, it is surprising that so littleattention has been paid to ontogenetic data in twentieth-century sys-tematic writings. Von Baer's fourth rule is a direct contradiction ofHaeckel's biogenetic law. The first three rules pertain to the problemof analyzing synapomorphy, and our subsequent discussion focuseson them.

2. The ap.plication of ontogeny Early in the chapter, we appliedontogenetic d~ta to ~h~problem of identifying evolutionary novelties(synapornorphies) within a group of five vertebrate taxa. The discus-sl.on~as presented in terms of deciphering levels of character dis-tribution (synapomorphy) by examining the distribution of differentcharacter~statesboth in the adult and in early developmental stages.Thus cartilage was conside d id . .. ,re more WI espread (primitive) relativeto bone because the forme . ..r IScommon to all five taxa In early stagesof deve~opmenta~d is transformed to (replaced by) bone in later on-togenetic stages In the perch, lizard, cat. and mouse. In the discus-sion that foll?wS, we expand on the use of ontogenetic data in char-acter analysts and discuss its strengths and possible limitations.

4. This statement is true if a peleom 10" '. '.

"

If . . 0 gist IS restncnng the analysis to a given monophyleticoup. outgroup comparison IS used hod .

the comparison it be' as a met to determine polarity, then at some point inher in time But tmay ~slble to show that the primitive character-stale in fact occurred ear-expect ~ deriv: ~y given collection of t~a, and using l:he paleontological approach, we

c aracter-states to OCcur1lIearlier strata (see text).

5. We do nor mean to imply, of course, that von Baer was an evolutionist or that he interpretedhis laws within an evolutionary context. The parallel between orderly ontogeny and phylogene-tic inference was an interpretation of these laws initiated after 1859.

60 Cladograms

The question we shall examine here is twofold: how can on-togenetic data be used to determine the level of distribution ofcharacter-states of adult taxa? And how do we interpret ontogeneticcharacter transformations in terms of their level of synapomorphy?Consider a monophyletic group of four species, A through D, in whichA and B have an endoskeleton composed of cartilage and C and 0an endoskeleton composed of bone. Which condition of the adult en-doskeleton is primitive, cartilage or bone? That is, which is morewidely distributed? Examination of the adult stage of these four spe-cies does not provide an answer. Assume that the ontogenies of thefour species are examined and all are found to have a cartilaginousendoskeleton in the early stages of development: in species A and Bthat condition is retained in the adult, whereas in species C and Dthe cartilaginous skeleton is transformed into bone. Based upon acomparison of just these four species, we would conclude that acartilaginous skeleton is a more general, more widely distributedfeature (it is possessed by all four species). A bony skeleton is in-terpreted as a more restricted similarity, a synapomorphy. Theseconclusions support the hypothesis that species C and D comprise asubgroup within the larger group, A-D. Nothing further can be saidabout the existence of other subgroups-(C +D)+A, (C + D)+ B,(C+ 0) + (A+ B)-because the cartilaginous endoskeleton of A andB is interpretable as a symplesiomorphy.

Viewed in this sense, then, ontogenetic data can be examineddirectly for information about the relative distributions of adultcharacter-states. A procedure such as this seems equivalent to anapplication of von Baer's first two rules and also of the biogeneticlaw as restated by Nelson (1978:327): "Given an ontogenetic charac-ter transformation, from a character observed to be more general to ach,ar~~terobserved to be less general, the more general character isprimitive and the less general derived."

The question arises as to how we are to interpret the level ofcharacter transformations themselves, and, in particular, how we aret~ analyze the problem of neoteny (Nelson 1978}.6 It is easily appre-elated that neoteny invalidates the application of von Baer's lawsand the biogenetic law (figure 2.13). In figure 2.13a, the ancestor

6. "The term neoteny, or, more completely, "phylogenetic neoteny." as used here is equiva-lent 10Gould's 0977a'221-28)" "- hosi . " .. ierm peeoomorp OSIS;features that appear only In the juvenilesof ancestors later appear as juvenile and adult characters in descendants.

o.~Q)

8'-c;o

b.

Species f

cartilage

Species 2bane

phylogeny•

cartilage corti loge

Species 2bone

Species f

cartilage

phylogeny•

cartilage cartilage

cartilage bone cartilage

>- neotenycQ) • •c. tn0-c:0

cartilage cartilage cartilageFigure 2.13 Therelationshipbetweenontogenyand phyloqeny.(a)Thede-scendantspecies2 hasmodified its ontogenyin termsof changefromageneralcharacter(cartilage)to a specific character(bone);this transforma-tion canbe intepretedwith referenceto the biogenetic law. (b)Thedescen-dant species 1 hasmodified its ontogenyin termsof changefromthe spe-cific to thegeneral,this transformationrepresentsa caseof neotenyandnegatestheuseof the biogenetic law. (c) Neotenyrepresentsa characterreversaland thusconfusesanalysisof characterphylogenyfor that charac-ter. (Seetext)

\1 ,

62 Cladograms

(species 1) has a cartilaginous skeleton in both juvenile and adultstages: the descendant (species 2) has evolved a transformation to abony skeleton in the adult. The developmental histories of both spe-cies, interpreted within the framework of von Baer's laws, would leadto the phylogenetic inferencethat the condition in species 2 isderived relative to the condition in species 1, The occurrence of neo-teny, however, would falsify the use of von Baer's laws (figure 2.13b):ontogeny cannot be interpreted as going from the general to the spe-cific.

Suppose, in some third species descended from species 2, thatthe ontogenetic transformation of cartilage to bone is lost--cartilageis retained throughout the developmental sequence and becomessecondarily an adult character (figure 2. 13c). Phylogenetic ally, then,a ca~i1age-cartilage transition is primitive for the group containings~ecles 1+ 2 + 3, but is also advanced (an autapomorphy) for spe-cies 3. Neoteny produces character reversal, which creates the con-fusing effect of the "same" character with two different distributions(species 1 + 2+ 3, and species 3 only),

Thus for any comparative sequence of ontogenetic transforma-tions, that .sequence thought to be more general, or primitive alongthe usual line of ~rgument for ontogenetic data may, in any particularexample, :eflect Instead an instance of neoteny. Neoteny can only bedetected If the analysis of other features suggests acceptance of acladopram with an arrangement of taxa such that an adult characterIn fact has two ~eparate distributions. A corroborated hypothesis oftwo separate distributions of what was initially taken as a singlecharacter establishes the presence of neoteny and further impliesthat the "ch~racter" cannot be homologous between the two distribu-tl~ns retention of cartilage in the adult as a derived character (Le.,W:lth bone su~preSSed) is not the same, in the evolutionary sense, assimple retention of the primitive state, Since the characters cannotb~ told apart on inspection, only demonstration of conflicts of dis-tribution can be used to detect their presence as two discrete char-acters. Thus neoteny is .. . a convergence (or parallelism) and is ra-solved In precisely the same fashion as in all such instances: twocharacters taken to be the "same" (homologous) can be shown to beconvergent only if the analysis of further characters demonstratesthat the taxa sharing th bl .e resem ance are In fact more closely re-

•

Cladograms 63

lated to some other taxa which do not share the resemblance (seepage 70 for further discussion of the analysis of parallelism andconvergence).

3. The application of outgroup comparison The most commonly ap-plied method of determining the relative distributions of two or morecharacter states (or determining their "polarity") is the procedure ofoutgroup comparison. It is based upon a rather simple method-ological principle: if we are attempting to resolve the scheme ofsynapomorphy within competing three-taxon statement cladograms,those character states occurring in other taxa within a larger hypoth-esized monophyletic group that includes the three-taxon statementas a subset can be hypothesized to be primitive (plesiomorphous)and those character-states restricted to the three-taxon statement it-self can be hypothesized to be derived (apomorphous). In short, theprocedure is one of mapping the distributions of characters-stateswithin and without the group under consideration. This principle ofoutgroup comparison can be illustrated with a simple example. Infigure 2.14a and b are two alternative three-taxon statements. In fig-ure 2.14a taxa A and 8 share a similar character-state a', whereas infigure 2.14b taxa A and C share a character-state b'. If taxa A, 8, andC are themselves members of a still larger monophyletic group A-E,then outgroups, that is, taxa outside the three-taxon statement, canbe compared for the possible occurrence of alternative character-states, in this case a or a' and b or b'. Those character-states foundin the outgroups, then, are considered primitive within the three-taxon statement. If, for example, taxa 0 and E possessed character-states a and b' (see figure 2.14c), this would suggest the hypothesisthat character-states a' and b are derived, and this in tum wouldlead us to prefer the cladogram of figure 2.14a, because the postu-lated synapomorphies are consistent with its branching pattern. Inan evolutionary context, we would postulate the evolution of a -ve'between ancestors Y and X, and b' ---,lob within the lineage leading totaxon 8.

This procedure, then, can be viewed as a search for levels ofcharacter distribution in the observed similarities. In this sense, simi-larity a' would have a distribution (define a set) at the level of taxonA+ 8, and b at a still lower level, taxon 8. Character-states a and b',

64 Cladograms

o. b.Aa' Ba' Ca Ab' Cb'

c.Aa'b' Ba'b Cab' Dab' Eab'

Figure 2.14 Anexampleof outgroupc .incompatiblecharacterdistr"b r b ompanson.(a and b) TwoapparentlyComparisonofoutgroups0 ~n~I~ns asedonsh~r~dsimilarities (a', b'). (c)b' a symplesiomorphy.(Seetext.)suggeststhata ISa synapomorphyand

on the other hand wo Id h ..t A

.wou ave a distribution at least at the level ofaxon -E and perhaps hi h (h·. .. Ig er t !Swould be subject to independentInvestigation).

The use of outgroup com . .level of distribution f panson ISa search for the hierarchicalmeth d

. 0 each character-state. Put another way theo IS a search forth I I· '

character-state is a s e eve of the hierarchy at which eachtaxa At all I I ynapomorphy and therefore defines a set of

. Ower evels that ch .So h aracter-state ISa symplesiomorphy.

ow were levels of . ·1 .termined for th h simi anty, and thus synapomorphies. de-D and E h e c eracter-states of the present example? Why were

c osen for the 0 t .number of 18 h . u groups In contrast to the potentially largexa t at might have been chosen? Examination of the

•

Bb

Cladograms 6S

three-taxon statement shows that character-state a' defines thegroup A + B, and a character-state b' the group A + C. Clearly, toresolve the conflict in the level of synapomorphy of these two simi-larities, additional taxa must be examined. If more As, Bs. or Cs wereadded to the analysis, the distribution of similarities would probablyremain pretty much the same and we would consequently be deal inginstead only with basic taxa of higher taxonomic rank than was doneinitially. It is necessary to investigate taxa that do not belong to thesethree groups, because the strategy is to discover whether a or a' andb or b' are found in other taxa at a higher hierarchical level. But howwas the level A + B + C determined in the first place? What are theirpossible synapomorphies? Thus, if A B, and C were the cat, mouse,and lizard, using our example at the beginning of the chapter, wemight make a preliminary definition of this hierarchical level by say-ing they are vertebrates with an amniote egg. This immediately leadsus to ask what are the distributions of character-states a, a', b, andb' in vertebrates lacking an amniote egg. We might observe, for ex-ample, that frogs and fishes exhibit character-states a and b' andthus we conclude that a' and b are derived within amniotes.

To summarize: in order to determine the level of synapomorphyfor a character-state within an initial three-taxon statement, it is nec-essary to compare taxa at a higher level. This is accomplished, ul-timately, by using other similarities to erect a preliminary hypothesisabout the group-membership of that higher leve\. It would seem thatit is not possible to decipher the level of synapomorphy of one simi-larity without tentative acceptance of a higher level of synapomorphydefined by a different similarity. In systematic work of this kind, oneis always accepting, as working hypotheses, higher level hypothe-ses of synapomorphy (and thus relationship) as a means of evaluat-ing lower level hypotheses. All levels of hypotheses are continuallysubject to testing and reevaluation (see below).

Within the systematic literature, outgroups are traditionally de-scribed as being the "closest relatives" of the taxa within the three-taxon statement. In many systematic studies, a hypothesis aboutthose "close relatives" frequently has been corroborated, or at leastthe range of possibilities has been narrowed to a small number oftaxa. Choosing outgroups to be compared is usually not as difficult atask as some critics of cladistic analysis would have us believe. For,after all, it is rare that a given three-taxon statement cannot be con-

66 Cladograms

sidered a subset of some higher level hypothesis of relationship, andtherefore there will always be some basis for evaluating hypothesesof synapomorphy. Contrary to these criticisms, it is not necessary tospecify the close-relative exactly but only a higher level of synapo-morphy. This does not mean, of course, that the determination ofsynapomorphywill be easy, or even possible in some cases. But thisis not a limitation of outgroup comparison, or of cladistic analysis ingeneral. Rather, it is a limitation set by the available data, a limita-tion of our knowledge, as it were, and such a situation would seem-ingly jeopardize all methods of systematic analysis equally. Someexamples and problems of outgroup comparison will be presentedlater in the chapter.

Character weigbting It is our view that each hypothesis of synapomor-phy, if arrived at by careful comparative analysis, can contribute tothe. :valuation of alternative cladograms. Recently, however, theopinion has been expressed that some hypotheses of synapomorphyare less useful and should have less weight than other hypotheses(Hecht ~976; Hecht and Edwards 1976, 1977). This viewpoint hasalso gal~ed some approval from evolutionary systematists (Szalay1.977).It IS thoroughly consistent with the prevalent tradition in evolu-tionary systematics to set up a system of weighting whereby somesorts of characters are deemed axiomatically to be more (or less)~seful In phylogenetic analysis than others. We merely reiterate the~ndamental observation that all characters are evolutionary novel-ties (synapomorphies) at some level. The problem is to find that levelfor each character, and not to assume that "conservative" featuresare "more or less valuable than "variable" ones, or that "ncn-adap-~lve features are better or worse than obviously "adaptive" features,~ustto name two different sorts of characters occasionally deemedI~portant as criteria for the purposes of phylogenetic inference. It isstili commonly said t . t ., " ' or ins ance, that obviously functional ("adap-tive ) features are m Hk I. ore ley to represent convergences or parallel-IS~Sthan are structures whose functions are less obvious. But paral-lelisms and convergences can only be hypothesized by arguing thatone or more of the tax ith th. a WI at structure are more closely related tosome org,anlsm without that structure. Because the function of astructure IS assumed t b "und " .. 0 e un erstood IS no reason to assume thatIt evolved more th San once. uch an assumption reflects the preoc-

• •

Cladograms 67

cupation with adaptation predominant in contemporary evolutionarytheory (chapter 6). We return to the analysis of parallelism and con-vergence later in this chapter.

Synapomorphy as a Test of Cladistic Hypotheses

In previous sections of this chapter, we presented the notion thatcladograms are hypotheses about synapomorphy patterns, and that,given a unique pattern to the history of life, there is an expectationthat a cladistic hypothesis which closely approximates, if not pre-cisely parallels, the unique historical pattern should exhibit maximalcongruence of the nested synapomorphies. If one views the scienceof systematics as being subject to the same rules of inference asother branches of hypothetico-deductive science, then these as-sumptions and expectations take on the nature of an axiomatic meth-odological principle: because we cannot empirically have knowl-edge of the true historical pattern, science must formulate a criterionby which to judge the relative merits of our close approximations(hypotheses), That criterion, in effect, is parsimony, and it specifiesthe most preferred hypothesis to be the one exhibiting the mostcongruence in the synapomorphy pattern, This is not to suggest thatthis hypothesis is necessarily true or that it precisely mirrors the truehistorical pattern, but only that it appears to be a better hypothesisthan the alternatives. As Wiley (1975:236) so aptly remarks, the appli-cation of parsimony must be accepted, "not because nature is parsi-monious, but because only parsimonious hypotheses can be de-fended by the investigator without resorting to authoritarianism orapriorism. "

Synapomorphies are tests of cladistic hypotheses (cladograms),but reciprocally these cladograms, with their expected pattern ofcongruent synapomorphies, are also tests of the synapomorphies.Such reciprocity is an integral part of hypothetico-deductivescience, particularly if one views all scientific statements-eventhose considered to be "facts"-as being theory-laden (see Popper1959, and many other philosophers of science). Detailed discus-sions about the application of synapomorphy to test cladistic hypoth-eses can be found in Wiley (1975), Gaffney (1979), and papers citedtherein. We will describe some basic principles of this procedure.

Consider the three alternative (competing) three-taxon state-

«

lD

lD

o

«

<.>

lD

«

o

c

o

o

o

o

o

o

c

«

lD

o

lD -<.>

Q)

«

<.>

lD

«

Ctadograms 69

ments in figure 2.15. If there is a postulated synapomorphy, a',present in taxa A and S, whereas the primitive condition, a, is in C,then we can say that cladogram 2.15a is corroborated and clado-grams 2, 15b and 2.15c are rejected." A decision about corroborationor rejection is dependent upon the ability of the synapomorphy todefine a set within each cladogram.

If similarity a' is truly a synapomorphy within the three-taxonstatement A + B + C, then the foregoing argument is clearly accept-able. But, what if a' is not a true synapomorphy within A + B +C?How do we account for this observed similarity in A and B? Thereseem to be three, and only three, possibilities and the first one reallydoes not count: (1) The investigator has not actually perceived a sim-ilarity a' at all; a mistake has been made. This is the least likely ofthe possibilities, for after all, if a' can be recognized and named,then some similarity must be apparent. This leaves us with two alter-natives: (2) a' is in fact a true synapomorphy but it defines a set at ahigher level of the hierarchy, say A+ B+C, or even higher. If this isthe case, then the shared similarity a' in A and B is a symplesio-morphy and, a in taxon C is almost certainly to be interpreted as anautapomorphy. (3) The similarity a' in A and B is not a synapo-morphy at any level of the hierarchy, therefore it must be interpretedas a convergence, in other words the similarity was independentlyacquired during the evolution of A and B. If this is true, the conditiona in C cannot be analyzed given present information. It should beclear at this point that a convergence is the antithesis of homology; itis neither a synapomorphy nor a symplesiomorphy.

The preceding discussion inevitably leads to the question: if wehave perceived a similarity distributed in two taxa of a three-taxonstatement, how do we decide whether a postulated synapomorphy isa "true" (corroborated) synapomorphy or a convergence? The ques-tion can be restated: how do we corroborate or reject a hypothesis ofsynapomorphy?

In order to answer this question, let us return to our example offigure 2.15. If a second postulated synapomorphy also defines the

7. In the systematic literature, one will also find the word "falsified" to refer to hypothese,such as those represented by cladograms b and c. "Falsified" implies that the hypotheses areproven false, bUI this is not the meaning we (or other phylogenetic systematists) wish to con-vey. It may be thai, eventually, the preferred hypothesis will itself be "rejected" by somesynapomorphies. Another word that is occasionally used is "refute." but this has essentially thesame implications as "falsify."


subset A + 8, the cladogram of a is again corroborated whereas band c are rejected. It is clear, then, that continued testing of thecladograms will tend to increase the degree of corroboration or re-jection of each hypothesis. Let us assume, for example, that afterrepeated testing we arrive at a situation similar to that shown in fig-ure. 2.15d--f. There are six postulated synapomorphies (a'-t'), ofwhich four (a', b', d', and t') define the set A + B (figure 2.15d) andtwo (c': e') define the set A +C (figure 2.15e), No postulated synapo-morphies appear to define the set C + 8 (figure 2.15f). On this basis,th:n, an investigator would conclude that hypothesis f is the mostrejected.hypothesis, e the next most rejected hypothesis, and d theleast rejected hypothesis, The criterion of parsimony specifies ourac.cePtanceof the least rejected hypothesis. If our comparative anal-ysrs of.outgroups has clearly indicated that the postulated synapo-morphies a' f' co Id t b .. . - .u no e Interpreted as being synapomorphies ata higher hierarchical level, and thus primitive within A + B + C, then

1 it th I d' . ... e ca IStlChypotheSISof 2.151were true, it would necessitateSIXcasesof convergence(a'-f');

2. If the cladistic hypomest '215f SIS0 . e were true, it would necessitate.ourcasesot convergence(a', b'. d', and I'); or

3. II thecladistic hypothesis012.15dwere true, it would necessitatetwocasesot convergence(c' ande').

In evolutionary terms th f. . '" ' ere ore, we prefer the cladistic nypoth-ests which minimizes coh . nverqence. Once a preferred cladistic hy-

pot eSlshas been chos " .(t r en, I In turn serves as a basis for evaluatinge~ Ing) the synapomorphies. If, for example the hypothesis 2 15d ispre erred, then postUlated synapomorphies~' b' d' and t ' ~re ac-ceptsedas synapomorphies at the given hiera:chi;al level

uch seems to be th If·era eo synapomorphies in testing clade-grams, and cladograms in t tialso leads t es Ing synapomorphies. The discussionrnorph tte 0 a more detailed consideration of conflicts in svnapo-y pa ern.

vergence refers to similarities independently acquired and notderived from a common ancestor. 8y definition, then, one must havea hypothesis of relationships prior to applying the concept of con-vergence. Thus, in figure 2.15, character-state a' was recognized asa homology (synapomorphy) because cladogram d was eventuallyaccepted, If either cladograms e or f had been accepted, then a'would have been considered a convergence, As we said, this con-cept of convergence has been standard in the systematic literature,although phylogenetic systematists have only recently made its for-mulation more explicit

There is a belief among some systematists that it is somehowpossible to assess the probability that certain similarities are theresult of convergence or parallelism. It is said, for example, thatthose similarities with a high probability of being convergent or par-allel can be eliminated from the analysis, or at least given "lowweight." Such is the rationale behind the weighting scheme of Hechtand Edwards (1977). We will now comment specifically on the prob-lem of analyzing parallelism and convergence, because this bearsdirectly on the problem of testing cladistic hypotheses.

To many systematists (e.g., Mayr 1969:226--28) convergence offeatures does not present a major analytical problem because studyof a sufficient number of characters usually reveals that the two taxapossessing the similarity are in fact not related to each other but tosome other group. We agree and further assert that the same appliesto the concept of parallelism, Other systematists, on the other hand,believe that parallelism is different in important ways. What do theseworkers mean by parallelism? Some definitions follow,

Parallelismis the development01similar charactersseparately in twoor more lineages 01commonancestryand on the basis 01,or chan-neled by, characteristicsof that ancestry.(Simpson1961'78){Parallelismsare] similarities resulting 01joint possessionof indepen-dently acquired phenotypiccharacteristicsproducedby a sharedgen-otype inherited from a common ancestor (similarity through parallelevolution).(Mayr1969:202)_[In] parauel evolution ... the character is present in the ancestralform but a commonderived characterstate has been independentlyevolved in eachdescendantform. (Hechtand Edwards1976:654)

Conflicts in Syna hpomorp y Pattern and Their AnalysisThe preceding section t I' .shared char t s rong y Implies that the identification of a

ac er-state as a .Whetherthe st >' ' convergence IS strictly a matter of

Iml anty can be d t d . . . . .cladogram, Indeed ~~e 0 efine a set within a specified

, the traditionally accepted definition of con-

The concept of parallelism is most easily compared and con-trasted with convergence by reference to branching diagrams {figure

L

72 Cladograms

a. b.

~ ~ Ca Da ~ Ba Ca ~

XaYa

Zac.

Aa' Ba' Co Do

Xo'

Figure 2.16 The concepts of parallelism and convergence.(a) An example of parallelism. (b) An example of conver-gence. (c) An alter~':ltive hypothesis to parallelism in whichthe ancestral condition In X is postulated 10 be a' be--cause the sisler~groupsA and B both have a' (See text).

2.16). The condition of the similarity in the common ancestor is thecentral tion i .ques ron In applying the two concepts. In parallel ism (figure2.~6~), the common ancestor X is assumed to have a relatively moreprJmltl~econdition of the character than its descendants, and a sfrni-lar derived co diti .. n I Ion IS assumed to have evolved in each descen-dant Hneaqe. In convergence (figure 2.16b), on the other hand, thecommon ancestor Z is 1a so assumed to have the primitive condition,but the descendant t havi .axa avmp the derived condition are not sister-groups of one another Thi 1disti . . ISwou d seem to be the only meaningful1~s;~ct~onbetween the two concepts. Hecht and Edwards (1976,

, or example, consider the atlas-axis complex of birds andmammals to be a par II I· ba e Ism, ut because the two taxa are so dis-tantly related their Itlon b tw ' examp e would seem to vitiate all useful distinc-

e e~n parallelism and convergence.There IS however .

plication of' ar II .' a senes of assumptions surround ing the ap-p a euem that are seldom if ever applied to con-

vergence In parau I" '. "cies hav· ·.1 e Ism, It IS argued, because closely related spe-

e Simi ar genetic b k . .ac grounds and therefore similar

Cladograms 73

developmental potentials, they are likely to respond to similar selec-tion forces in a like manner; hence parallelism is to be expected inclose relatives living under similar environmental conditions. Wemight. for the moment, accept this line of argumentation and ask:how are we to analyze such a situation in systematic studies?

In order to apply this concept of parallelism (say, as in figure2.16a), a minimum of four sets of information is required:

1. It must be known or assumed that the common ancestor X infact had the primitive character-state a.

2. It must be known or assumed that the taxa (or individual or-ganisms) of the lineages leading to A and B had genetic and devel-opmental potentials which were sufficiently similar to some specifieddegree.

3. It must be known or assumed not only that the taxa of A and Blive in similar enviornments, but that they were subjected to similarselection forces.

4. It must be known or assumed that if they were subjected tosimilar selection forces, these taxa would in fact respond to them insimilar ways.

Given the above four assumptions, one could conclude that par-allelism as defined by evolutionary systematists had occurred.e

It is our opinion that this conception of parallelism is epis-temologically impossible to evaluate and therefore unscientific. Toapply this concept-and, accordingly, a weighting scheme-to theanalysis of synapomorphy necessitates the inclusion of numerous adhoc assumptions. Thus, one cannot test any hypothesis of parallel-ism without resorting to knowledge about ancestral conditions, ge-netic and developmental potentials, the selection forces present inthe past, and the expected response to those selection forces.

The concept of parallelism is also open to criticism on thegrounds of parsimony. An argument of parallelism requires that twoevolutionary lineages evolved a derived character-state indepen-dently. A properly expressed cladistic hypothesis, on the other hand,would necessitate the evolution of the derived condition only once,

8. In actuality, all that would be required 10 invoke parallelism would be knowledge of rela-tionships and knowledge that the common ancestor X has the primitive condition. Most work-ers, however, especially evolutionary systematists, also include assumptions about develop-mental-genetic potential and natural selection.

r - • i

74 Cladograms

between common ancestors Y and X (figure 2.16c); the unspecified o. W b. Xcommon ancestor X would be inferred to have the derived condition A B C W C Da'.

In conclusion, we recommend that the concept of parallelism be ..omitted from systematic studies. We suggest that the term con-vergence be applied to all cases of nonhomologous character simi-

!larities, identified through character conflicts arising from normalanalysis of the patterns of synapomorphy.

Internesting Statements of Monophyly

One can view the problem of cladogram construction as a problemd. Z c. yof intemesting two or more three-taxon-statement hypotheses of mon-

ophyly. Any monophyletic group can be resolved to a basic tax- y E F X D Eanomie unit of a cladogram at a higher hierarchical level (see Wiley1975). For example, in figure 2.17 the monophyletic group A + B canbe represented by the taxon-name W at a higher level of the hierar- .-chy (figure 2.17a), W + C by X at a still higher level (figure 2.17b),and so on (see also Gaffney 1979).

All levels of the hierarchy are susceptible to testing by postu-

\lated synapomorphies. Each identified synapomorphy can test atonly one level of the hierarchy, Furthermore. to test at anyone level,

Zthat is, to make a determination of synapomorphy, requires com pari-yson of taxa at higher levels. It can be appreciated, therefore, that the

Xchoice of an outgroup at anyone level of comparison is itself a e.hypothesis of monophyly to be tested at still higher levels.W

A B C D E F

Some Case Studies

In this chapter, we have attempted to provide systematists with amethodological foundation for formulating their own cladistic hypoth-es~s. Most of the examples were hypothetical, and while such theo-retical ta~a as A, B, and C have their value, real-world examples alsohave their own advantages. Thus, we close this chapter with somee~~mples of cladistic analysis. The presentation of examples is adifficult task in a general book of this kind because the readership

Figure 2.17 The resolution of lower-level hypotheses of monophyly tohigher-level hypotheses (see text).

76 Cladoqrarns

will have diverse interests and experience. To properly evaluate anygiven study requires not only famil iarity with the comparative biol~gyof the organisms themselves but also a little experience in applymgthe theory we have discussed up to this point. Nevertheless, wehope the examples will be of value, and that they will give a generalidea of the procedures and kinds of evidence used in clad istic anal-yses. Most of the studies begin with a comparison of similarity, fol-lowed by an analysis of synapomorphy, and postulated nested setsof taxa are then constructed on the basis of the postulated synapo-morphias.

Cladograms 77

c G A G AT c T

Example 1. Actinopterygian fishes Wiley (1976) recently performed acladistic analysis on some major groups of actinopteryg ian fishes.His main concern was the relationships of the species within thefamily of gars (Lepisosteidae) and the relationships of that family toother groups of actinopterygians. Our discussion here is restricted tothe latter portion of his study.

The relationship of gars to other actinopterygians has consti-tuted a systematic problem for many years. Wiley identified fourcladistic hypotheses that have been suggested frequently in the lit-erature (figure 2.18), and then proceeded to evaluate each usingpostulated synapomorphies. The similarities studied were primarilythose of the cranial and postcrania! osteology, but some myologicalsimilarities were noted. The determination of the level of synapo-morphy for each similarity is seldom easy within the higher taxa offishes, and in formulating his hypotheses Wiley compared a broadrange of taxa, both actinopterygian and nonactinopterygian.

The results of Wiley's analysis are shown in figure 2.19. A largenumber of the postulated synapomorphies indicate that ami ids (bow-fins) can be placed in a group along with the teleosts (synapomor-phies 1-13) and that this group, in turn, is the sister-group of thegars (synapomorphies 14-20), Not shown in this figure are the syna-pomorphles linking gars, amiids, and teleosts to the chondrosteans:the reader should consult Wiley (1976:14-37) for his detailed ratio-nale r~garding determination of synapomorphies.. Wiley's study is noteworthy because it is a cladistic analysis in

which a set of specific hypotheses was first formulated on the basisof prior systematic work, and then each hypothesis was tested by theresults of an analysis of synapomorphy. In this case, the synapomor-

sG A Tc A TG

d.e.

Figure 2.18 Thefour cladogramssubjectedto testinqbyWiley in evaluat-ing the interrelationshipsof somemajorgroupsof ~ctlnopteryglanfishes.A: amiids; C: chondrosteans:G: gars; S: semtonottos:T: tereosts.(FromWiley 1976:14.)

phies themselves were not used directly to construct inlernestingstatements of monophyly.

E I 2 Side-neckedturtles Gaffney (1977) has analyzed the syna-xamp e . f h I'dpomorphy patterns of cranial similarities for seven genera,o c .e I.

turtles, The family itself is well-defined by synap~morphles withinturtles as a whole, thus the study was initially restricted o~ly to thisf ·1 Gaffney used all other turtles as a basis for evaluating syna-ami y. . d . t tedpomorphies within the chelids: in addition, this stu y was In erpre

c G A

Figure 2.19 The prefe dte.rygianfish jnlerrelatj~~~hicladogramof actlnop,?Icated by shaded rectan IPS. S~nap?morphjes are in-eans. G: gars; T: teleosls 9(eFs.A.Wa~lIdS; C: cnoncros-. rom Hey 1976:38.)

TCladograms 79

•• 2

• 3

• •• ,• 6

• 7

• •• •• "• "• "• ta

"ts

"" Figure 2.20 A cladistic hypothesis for the genera of turtles in the family

Chelidae. (From Gaffney 1977:17, figure 10.)"" within the framework of a prior cladistic analysis of all major groups

of turtles (Gaffney 1975). For example, Chelodina and Hydromedusaare the only two genera of chelids to have four claws on their forefeet(all other genera have five), Moreover, the sister-group of the Chell-dae, the Pelomedusidae. also have five claws. Therefore, the condi-tion in Chefodina and Hydromedusa was postulated to be derived.

Gaffney's results are summarized in figure 2.20. Six sets ofsynapornorphies were found to define six sets of nested taxa. In thisexample, because of a paucity of previous work on creuos. well-defined alternative hypotheses were not recognized prior to the anal-ysis, as was done in the preceding example. Thus Gaffney proposedhypotheses of synapomorphy and then used these to internestvarious three-taxon hypotheses of monophyly.

20


Example3. Ratite birds The large. flightless ratite birds have excitedconsiderable comparative work for over a century, but almost nostudies postulated interrelationships because, until very recently,few workersconsidered them to be a monophyletic group. Followingprior work arguing for their monophyly (Bock 1963; Parkes and Clark1966),Cracraft (1974a) constructed a cladogram of nested synapo-morphles. This study concentrated on cranial and postcraniai simi-larities, but the results were congruent with synapomorphiespostulated for similarities in behavior and breeding biology (Meise1963),

Because one or more ratite taxa have frequently been consid-ered the possible sister-group of all other birds, their study pre-sented some special problems for determining the level of synapo-morphy of the different character-states. If ratites are the sister-groupof all other birds, this would necessitate extending the comparisonbeyond birds to other vertebrates, particularly diapsid reptiles. Butthe attributes of these nonavian groups are not comparable to thesimilarities of ratites and other birds in sufficient detail to make sucha comparison useful (the detailed similarities of ratltes and otherbirds do not, generally speaking, extend to diapsids). Hence, it wasdecided that those similarities within ratites having widespread dis-tribution in all other birds, particularly nonpasseritorms. would behypothesized to be synapomorphous at a hierarchical level higherthan ratites (t.e.. to be symplesiomorphous within ratites). Such aprocedure seemed to work, for it was discovered that seven sets ofpostulated synapomorphies defined nested subsets for the eight in-cluded taxa (figure 2.21), and no other cladogram was suggested bythe data. As an example of the reasoning used to postulate synapo-~~rphies. we can examine the morphological trends seen in thetibiotarsus, one of the hindlimb bones (figure 2.22). In primitive ra-tires (e.g., the tinamouCryPturellus), and in nonpasseriform birds ingeneral, the cnemial crests of the tibiotarsus are bladelike structuresand not greatly enlarged. Not only is there a trend within ratites forthe crests to become enlarged anteriorly, but their base becomesstrongly constricted (figure 2.22F) and the outer cnemial crest be-comes k~oblike. This latter feature is unique, and clearly derived,and was Interpretedas a synapomorphy of Rhea and Strutbio. Othersynapomorphies were postulated in a similar manner,

f~ra-~ec~~~~aft1c~~~~:~~,h6~~~~:I~;i~nt~~tf~:~;~~s~f~~;~~Ii~~~ts'Union.)

, ld rimates In recent years cladistic analysis hasExample 4. HomlDOI p .' h' d rI' d t the question of primate Interrelations IpS, an ou

been app re a . orne studies of the great apes andexample here pertains mostly to s It Delson and Andrewsman. In particular, the reader can consu

10

9

"15

"13

-

82: Ctadograms

- ,/- -D F

CowrJ,ius

Fi,gu.re 2.23 A cladistic hypothesis for the great apes and man. [With per-rrusston from Delson 1977:445, figure 1. Copyright by Academic Press Inc.(London) Ltd.l• •Example 5. Spiders Ptatnick and Gertsch (1976) undertook a cladisticanalysis of the major groups of spiders, and their paper dealt almostexclusively with the three-taxon statement including the taxa Liphis-tiidae, Mygalomorphae, and Araneomorphae. In order to determinewhich similarities are primitive and which derived, they presentedevidence for the arachnid order Amblypygi being the sister-group ofthe spiders as a whole. They then hypothesized "that any character-state found in some but not all spiders and also in amblypygids isplesiomorphic, and its homo logs apornorphic" (p. 2). In addition tooutgroup comparison, Platnick and Gertsch employed ontogeneticdata, effectively following von Baer's principles to define the polar-ity. As an example of their approach to character analysis, theynoted that IiphisWd spiders have the third abdominal segment as adistinct sclerite ventrally. The Mygalomorphae and Araneomorphae,on the other hand, have lost atl ventral indication of segmentation.Because the amblypygids have a distinct ventral sctertte. the condt-

ABr C

JgUl'e2.22 Hypothesized primitive-ct . .ties of the proximal tibiotarsus f . enved sequences In some similari-CrYPturellus is similar to that fO;atlte birds. Th~ condition of the tinamout~lated to be primitive (see te~) (~er noroasserttorm birds and is thus pas-sion of the British Ornithologist~' ~~:;~acraft 1974:501, figure 7, by permis-

(1975), Eldredge and TattersEld:edge, and Tattersall 197all (1975), Delscn (1977), and person,a discussion of the t ( 7). All of these papers not only include

axa and the sy .but each also present f napomorphlas used to unite them,

s use ul sum .theory and methodo!o f manes and statements about the

gy 0 cladistic I·good examples of I ana YSIS. These papers are alsorecent taxa' in hypothes ~~es combining data from both fossil and

, eSlzlng I'used outgroup com' po anty of character-states the authorst! paneon (to other "I " .Igraphlc distribution. A re r .' ower primates) and stra-mates is shown in fig p esentatlvs cladogram of the higher pri-

ure 2.23 (after Delson 1977).

r--~---------""-!!!!!~!!'!!!_""--"'----------'"• •• •~ ~• 0- e-• •

"0 0

~E E0 0

:c " •0> e0- ~ •::; " •-e

14

Cladoqrarns 85

tton in mygalomorphs and araneomorphs was postulated to be de-rived. The results of a similar analysis on 14 characters are shown infigure 2.24. For another application of cladistic analysis to thestudy of spider relationships, the reader is also referred to Platnick(1977c)

Mesotherae

Opisthothelae

hFiguhre 2.24 A cladistic hypothesis for the19 er taxa of S "dd I Pi ers. Shaded squares in-rca e synapomorphies' e 9syna ' - ". pomorphous character-states 3b-9

unite the Myg Ih a omorphae and Araneomor-f~ ae. (From Platnick and Gertsch 1976'9Igure 6.) ..

Chapter

3Species: Their Nature and Recognition

THE ENGLISH word "species" derives from the identical Latinword "species." In Latin, the word means "kind," The primary pur-pose of this chapter is to clarify the meaning of the English word asused in modern comparative biology.

Species Defined

Linnaeus (1758) applied the term "species" in an explicitly Latinsense: species are particular kinds of generic entities. In other words,"genera" were more general entities and species were more specifickinds of those entities. This original biological usage remains a partof OUf general concept of the nature of species today.

The notion of organic evolution, defined for the moment as "de-scent with modification," leads directly to an expectation of nestedsets of evolutionary novelties. These are the synapomorphies dis-cussed in the previous chapter. Nested sets of synapomorphous re-semblances are in turn used to define and recognize nested sets ofmonophyletic groups, or taxa. The basic aim of cladistic analysis isto define nested sets of monophyletic taxa on the basis of nestedsets of evolutionary novelties. Consonant with Lirmaeus' original useof the term "species," we would expect there to be a base level,where patterns of synapomorphous resemblance linked individualspecimens into clusters which are not further subdivisible. These in-divisible clusters would form the base of the hierarchy of nestedtaxa. In modern garb, they would conform exactly to the Linnaeanconcept of species as the basal units of the taxonomic hierarchy.

r-~---------""'----------88 Species

10tilThis .essentially analytic view of the nature of species conformsIs e notlbonof or~anic evolution as descent with modification There

no pro lem with the co t .strictly as the t f . neap as long as evolution is regardedhierarchicall rans ormation of characters, and taxa are viewed asdetermined ~ya~~angedclusters of individuals. the hierarchy beingnovelties. Two SimSlappar.e~t. pattern of ~ested sets of evolutionarycept follow' "sp .p e defln.'tl?nS of species consistent with this con-

. ectes are minimal rna hi'are taxa of the lowest cateoor! nap ~ e~lcgroups," or "specieschy." qcrical rank within the Linnaean hierar-

There is a distinct advanta ...merely the lowest-ranked cat qe to such ,definitions of species. Asthis rank are recogni d i egory of the ttrmaeen hierarchy, taxa of. ze In precisely the s f h'Inclusive (higher-rank d) arne as Ion as are morechapter 2 would SUff"e taxa. The .general principles adduced inspecial problem Wer~c~han? s~ecles recognition would pose noimmediately to s~me pra et'sltluatlon that simple, we could proceed

. c Ica examples of' , .associated taxonomic species recognition andpie. procedures. But the situation is not that sim-

Factors complicating th .from several consideration \~ery notion of what species are arisespecies are somehow"reals:,. ere has ~~ng.~xisted the notion thatvtdua's'' actually ex' t! .' t.e., actual entities," "things" or "indi-

ISIng In nat The! 'ferent from their status as "ki ~,re. err status as "entities" is dif-and 1978 for the de I rnds. (See Ghiselin 1974, and Hull 1976

. ve opment of this thPOintof view.) erne from a philosophicalThe t"no Ion that species are" " ' _

From ancient times a d real units In nature is an old one.t .. ,nasanon'oday, lt ISnoted that W'th' gOing cross-cuiturat experienceof different kinds of o I I.na local area, there seem to be a numberu I I rganlsms These k' da s a I more or less sim'l .:": In S are clusters of individ-arate kinds are all f 'I I ar I.n~ppearance and behavior. These sep-. . an y easily Ide tifIndividuals COmprising .. n I led (i.e.. they are distinct). Thereproduction among the each dIstinct kind exhibit some pattern oflike k' d mselves produ ' .In ,In other Words . ' clng still more individuals oforganisms organized int' ,nd.anyone area, there are different kinds ofOver a rscreta reprcd ., each of these units uctive communities. More-are either identical or Sli~:~yb~.~ound in adjacent areas, where theythe common experience of rna ki erent in appearance. In any case,

n md includ! ,, Ing our perception of our-

Species 89

selves, is that there are discrete reproductive units in nature. Nearlyall contemporary definitions of species stress their status as discretereproductive entities. Species, moreover, are expected to have a fi-nite distribution in space and through time.

Returning to the view that species are those taxa of the lowestrank in the Linnaean hierarchy, or, in a cladistic sense, minimalmonophyletic groups, the resultant picture can be neatly dia-grammed in a cladogram. But in a cladogram all taxa are terminal,whereas the general concept of evolution implies a pattern of an-cestry and descent. What kinds of units are ancestors and descen-dants? A purely transformational view of evolution as "descent withmodification" does not specify what the units are. The literature ofevolutionary biology is not very helpful on this point. as there is noclear consensus: some authors think that codons are the units ofevolution, and there are all shades of opinion ranging from thesegenetic loci to structures, individuals, populations, species, and onup the Linnaean hierarchy. It is not uncommon, for example, to hearof one phylum being ancestral to another. The Procaryota arefrequently alleged to be the ancestor of the Eucaryota. It would ap-pear that a clarification of the nature of the units of evolution wouldenable an improvement in existing evolutionary theory; we pursuethis theme in detail in chapter 6, But the identification of the units ofevolution-c-i.e.. what kinds of units are the ancestors and descen-dants implicit in the very concept of organic evolution?-also bearsdirectly on the nature of species.

Consideration of the ecological reality of populations and spe-cies and the nature of the reproductive plexus binding these unitstogether, implies a functional hierarchical organization of molecules,codona, cells, tissues, organs, individuals, populations, and spe-cies, A codon can change (t.e.. undergo mutation) within an individ-ual. Somatic mutations rest with the individual; mutations in germcells may be transmitted to other individuals reproductively. This istrue for all genetic changes and their phenotypic manifestations,Such changes can have import in an evolutionary context only to theextent that they are, in fact. inherited. The conclusion is obvious:Genetic and phenotypic heritable change assumes evolutionary im-portance (persistence in time) only to the extent that the reproductivecommunity and its descendants persist. For organisms among whichthere is at least occasional sexual reproduction, this unit would con-

j

"., ..~

90 Species

form to the reproductive concept of species. We are led to the ine-luctable conclusion that species, when conceived of as reproductiveunits, are the units of evolution.

Returning to the notion of phylogenetic ancestry and descent, jfspecies are evolutionary units, it follows that some must be ancestralto others. In other words, not all species can be viewed as terminaltaxa In terms of the distribution of characters, a descendant unitmust by definition possess at least one derived attribute not sharedwith its ancestor. Logically a species cannot have any synapomor-phies unique to itself (t.e., autapomorphies) if it is ancestral to anyother species. To the extent that each species in a collection of orqa-nisms appears to possess one or more autapomorphies, we mustjudge that either (a) all species sampled are terminal taxa or (b) theirdescendants are not included in the sample.

We conclude from the foregoing that strict application of cladis-tic methodology to the problem of species recognition is not entirelyconsistent with the concept of species as individuals and evolu-tionary units. Use of patterns of synapornorphy to delineate mono.phyletic groups requires that all taxa be terminal, whereas the verynotion of evolution requires some unit-and it must be the species-serve as ancestors and descendants. Strict adherence to cladisticmethodology may tend to underestimate the true number of speciessampled. A corollary, and somewhat ironic, observation is that spe-cies are not always monophyletic units in a cladistic sense.

Thus far we have developed a composite view of the nature ofspecies which includes at least two dual aspects. (1) Species are si-multaneously the lowest ranked taxa of the Linnaean hierarchy (Le.,they are placed in the "species" category) and real entities in nature,representing the highest rung of the hierarchy of molecules, throughin?ividuals and populations. As we have developed elsewhere inthis book (chapter 6), taxa of higher rank do not share with speciesthe quality of being individuals. Rather, taxa of higher categoricalrank are collections of species (see note 2 to chapter 2 for an impor-tant exception). In our view, it is important that they be monophyletic.(2) ~ .second duality sees species as integrated reproductive com.mUMles-i.e., they exhibit a pattern of parental ancestry and des-cent-which gives them internal cohesion and separates them fromoth~r .~UCh groups. Simultaneously, species are evolutionary unitse~hlbl.tmg a p~ttern of phylogenetic ancestry and descent-i.e., theygive rise to units of like kind (other reproductive communities).

Figure 3.1 The view of the nature o! spec~sT~fr::X~;~~i~:-areproducing organisms adopted In thts bO? within roupdepicted as discrete clusters demonstr~tl~~eathree s~eciespattern of parental ancestry and desc~n h lagenetic ancestryalso exhibit an among-group cettem a p ~n ancestor for theand descent, with o~e species servl~~ a: species are indivld-other two in this pa~lcular example. u ive rise to de-uals, with their own Internal ~oheslon, a~dl~dra buds offscendant individuals rather like a paren ayoung individuals.

92 Species

. Figure 3.1 ,illustrates our basic concept of species. They may bepictured as being somewhat like individuals of Hydra spp. (treshwa-~er h~drozoa.ncoelenterates). Each individual possesses its ownidentity and lO,tegrity. Occasionally new descendant individuals ap-pear ~y?uddmg, eventually detaching from the parent Hydra andestablishing themsel.ves as tully discrete descendant individuals.i ~e can ~um.marlZ~the foregoing into a simple definition of spe-

~ es. ~ species IS a diagnosable cluster of individuals within whichtthere ~sa parental pattern of ancestry and descent beyond which

ere IS not and which exh 'b"r 'd d' I I S a pattern of phylogenetic ancestryan escent among units of like kind.

Species 93

other such groups." Later, Mayr emended the definition (1969:26):"Species are groups of interbreeding natural populations reproduc-tively isolated from other such groups." In removing the phrase "ac-tually or potentially," Mayr avoided the pragmatic difficulties inevaluating "potential" interbreeding.

Simpson (1951) proposed an "evolutionary" species definition,subsequently modified slightly to read: "An evolutionary species is alineage (an ancestral-descendant sequence of populations) evolvingseparately from others and with its own unitary evolutionary role andtendencies" (Simpson 1961:153). This lineage concept of speciesemphasizes reproductive continuity through time. Wiley (1978) hasrecently reviewed various species concepts, concluding that Simp-son's seems to agree better with patterns in nature. The definitions ofboth Mayr and Simpson cited above are perfectly consistent with thedefinition we have adopted. All center around species as natural anddiscrete reproductive units.

It is to be stressed at this juncture that these and other, similardefinitions of species are concepts, pictures of patterns in nature.Mayr's deletion of "potentially" from his earlier definition points outan ironic, if not amusing, pseudo-problem with the species concept:it was repeatedly pointed out that it is realistically impossible toassess whether or not individuals living in the remote corners of afar-flung species' range could in fact mate successfully with oneanother. Similarly, paleontologists have spent a great deal of timeworrying if, within a hypothetical lineage persisting, say, 1 millionyears, individuals within that lineage a million years apart couldhave successfully interbred. How far back in time within the lineageHomo sapiens could modern individuals mate with their predeces-sors? Such questions seem ludicrous. As pragmatic problems inspecies recognition (see Sakal and Crovello 1970, for a critical ap-praisal of the "biological species concept" from a pragmatic stand-point), such problems are not laughable. But the concept of repro-ductive continuity is another matter, related to but not the same asthat of species recognition. Thus allopatric (geographically sepa-rated) individuals need not be able to mate in order to belong to thesame species. The important issue, insofar as the concept is concer-ned, is a plexus of reproduction across the entire distribution of aspecies. Similarly, the concept demands merely an unbroken plexus

Other Definitions of Species

Species have long been r .an evotuticna ~cog~.zed as coherent units in nature. Inequate evolut~nC~~t~~~bioloqists famil~ar with species tended to1859 e onqm of species. For example Darwin's

monograph on organic evol 1" .. 'on adaptation and I' u Ion, which IS basically a treatisevery notion of evol :.e ectlon as an argument for the validity of theSpecies" even th U 10hn,wa~ nonetheless entitled "On the Origin of

, aug relatively little lie!modes of origin of th exp tort attention was paid torole species play in es~ r~productive groups. Appreciation of thetimes,1 nonetheless has r utIO~,however muddled it has seemed at"biological" or "evol t! remaJnedsufficiently clear that all so-called

u ronary'' d f ..upon their nature as . e mtttons of species have centered(1940' 1942'120) def.r~t~rOductlvecommunities. For instance, Mayr's

, . 10I Ion states that scec! "or potentially interbreedin . pectes are groups of actuallyg populations reproductively isolated from

l . ~ major theme of this book as dearchitects of the "synthetic theory" ( veloped at length in chapter 6, is that many of tlte keythe significance of the eXistence of mostly ~eneticists and paleontologists) have overlookedthenlfore ~ueed a rather CuriouS:C:s as ducre/e entities in time as well as space and have(reproductIve communities) in the I'Y of theory. Nonetheless, the importance of speciesmoSl ardent " evo Utlonary proces h al. fransform~lionist" th all s as ways been sufficiently dear to the~nt of view have stressed ,"_ at . modem definitions of species from an evolutionaryllOQof our . UJI;: reproduclive communh . .VIeWSon COntemporary I' I Y aspect of species. For a fuller exphca-

evo unonary theory, see chapter 6.

94 Species

of parental ancestry and descent through time, and not that the endmembers, separated by our hypothetical 1 million years, could havesuccessfully interbred given the opportunity.

Discontent with the pragmatic problems of recognizing repro-ductive units in nature (discussed more extensively below), has re-sulted in the invention of other sorts of species concepts. Thesevarious sorts of species have been ably reviewed by Cain (1954).The concepts are unified, seemingly, solely by the desire of the ex-pon~nts to find an easy way of delineating basal taxonomic groups.Their eschewal of an evolutionary connotation robs them of any theo-retical biological interest. Consequently, we shall not review themfurther here.

Species 95

species' properties arising from the definition, with the patterns of in-trinsic and extrinsic properties displayed in the material at hand.

What are the general predictions arising from the definition ofthespecies concept we have adopted? There are two components toour definition: (a) within-species reproductive cohesion (parental an-cestry and descent), and (b) among-species phylogenetic ancestryand descent. The first of these components gives rise to the generalcriterion (direct demonstration of reproductive cohesion) and threeanciuary predictions: Species should exhibit (a) generally continuousdistributions of intrinsic properties within species (figure 3.2), (b)continuous distributions in space (figure 3.3), and (c) continuousdistributions in time (figure 3.3). All three of these latter predictionsare relatively weak; exceptions to the first two occur frequ~ntly, andthe third relies on notoriously spotty information from fossils. ~or.e-over, each has a circular aspect; simple demonstration of oontiruntyof any or all three of these properties is no guarantee that represen-tativesof more than one biological species have not been inclu.d~~.

The second (phylogenetic) component of the species defmlt,?nleads to a different set of predictions concerning the pattern.of ~IS-tribution of characters. The full strategy for recognizing species In ahypothetico-deductive framework includes both sets of procedures.First we discuss the general reproductive criterion and its ancillarythree predictions.

Recognition of Species

~e recognition of reproductive units is sufficiently difficult in prac-tice as"to.hav~ led many professional taxonomists to ignore the so-called bl?'?glcal species concept." Our position is that, whereas aconc~pt difficult to apply in practice is not thereby ipso facto a poordescriptor of actual pattema ! ,. ems In nature, a specrss concept shouldnonetheless contain sufficient elements to allow predictions aboutthe nature of species useful in practical species recognition.

The essence of the problem of species recognition is the simul-taneous () 1". .. a. e munation of all sampled organisms beyond the limitsof a biological species, and (b) inclusion of all those sampled ele-ments properly belonging to that species. We want neither to over-nor to underestimate th t Ie ac ua number of species represented in asample of organisms' we I ' h ",. . .we a so WIS to minimize the misclassification,or rrusallocation, of organisms to particular species

beFurthermore,delineation of species can-and properly should-

regarded as a hypoth t' ....,,1 • ,ti f . e 'co-deductive operation. The compost-Ion a a species and d t ' ,

th . e errrunation of its limits constitutes a hy-po esis. The systemaf t' b 'likely h thesi IS S pro lem ISto reject all but the least un-

ypo eats of species" ,b " ccrnpcsltion among the orgamsms

elOg consIdered Eval I' f "compos't' d· ua Ion 0 conflIcting hypotheses of species

I Ion ecencs upon 'a companson of the predictions about

Evaluation of Hypotheses of Species Identity: TheReproductive CriterionThe first step in the analysis of species composition is the seg~ega·tt . . "'1" f almilar or nearly iden-Ion of the organisms sampled into pi es 0 I .tical organisms such that no further subdivisions (clusters) are eVI~

, d'l'kltobedent. Thus the number of species actually represente IS leyoverestimated. For example there may be two piles, one of malesand one of females, which 'may eventually be shown to be sex~aldi , I th number of species'rnorphs of the same species. Alternative y, e .may be underestimated; sibling species (see Mayr 1963, chapter 3,Dobzhanakv 1951:267 ft.) are closely similar and closely related true

. , "j) hich are nonetheless,species (t.e. reproductively Isolated urn s WI,. .. . . .' 'hable The point here ISso similar as to be practically Indtslln9UlS· .

that species must be diagnosable, If the systematist fails to notice

a.1I6 Species

b.

c.

Flgure 3.2 Variation of a sin Imodal distribution wW .9 e character. (a) A uni-dal va~iation (dimor~hi~n ) 81.ng,lespecies. (b) Sima-(c) UnimOdal distribulio~ w~~t!llt~single species,c~e~.Patternsof vanatl !n 0.separatespe-dlff!c~lt, if not jmpossib~ depl:::t~d In band careof limited use as eVidenc~to tstln~UiShand thus areth . 0 species composition. e slight differences betw .IScarried along throughOU~~~two closely similar species the errorto detect sibling species 'f th e analysis. There is no automatic wayoutset. Only additional dat e~ are not diagnosed as different at thereveal this error. a In subsequent analyses are likely to

Species 97

c )=a.

=b

SPACE

Figure 3.3 Problems in use of distributional data (extrinsic properties) todelimit range of a biological species. (a) Geographic distribution. The rangeof the hypothesized species may actually represent the combined ranges oftwo separate species either partially sympatrtc or wholly parapatric or sym-pattie. (b) Temporal distribution. The range of a putative species may actu-ally represent the combined ranges of two separate species which eitheroverlap or are mutually exclusive (l.e., time-successive).

Hypotheses of species composition are therefore various combi-nations of the initial discrete diagnosable clusters discriminated atthe outset of the analysis. The reproductive criterion-direct demon-stration of reproductive cohesion-comes into play first. For in-stance, there may have been two separate samples, not initially con-sidered the "same," which may turn out to be males and females.Sexual dimorphism is a common characteristic of many differentgroups of sexually reproducing organisms. Morton (1958:139) dis-cusses an extreme example. In deep-sea cephalopods of the genus

98 Species

Argonauta, sexual dimorphism is pronounced. The large female car-ries a "paper" shell as an egg case, while the tiny male (only aboutan inch long) is reduced essentially to a reproductive organ, andthus not immediately recognizable as an argonaut in terms of its su.perticial morphological features. To confuse the issue even fU~h~r,the male reproductive organ detaches and moves about freely Withinthe mantle cavity of the female. The detached organ was longregarded as a parasite, even Jj,yCuvier who, according to Morton,dubbed it Hectocotylus octop.6b'is-which is dOUbly bizarre, as thegeneric name is the anatomical term for the male reproductive organin cephalopods! Only when the full anatomical details of the malesbecame known were there ample grounds for the hypothesis thatthese two quite dissimilar groups should be united into a single, at-belt highly dimorphic, species. Other, less extreme cases, wherereproduction has been observed (as in birds of paradise) have re.suited in me synonymizing of two "species" into one-in otherwords, the recognition of two or more clusters as being the "same"and so naming them. In general, direct demonstration of in.terbreeding-preferably in the wild, but also in the laboratory-asconfirmation of the general reproductive criterion of the species con-cept we have adopted, is the most direct means of uniting two ormore previously separated taxa into a firm hypothesis of the identityof a biological species.

In the absence of direct information on reproductive behavior,some patterns of bimOdality will nevertheless suggest the presenceof sexual dimorphism. In recent years, some systematists workingwith ammonites (externally-shelled fossil cephalopods) have as-serted that, in cases where two closely related and similar "spe-cies," which are found together in the same rock units, differ only in(a) relative size, (b) size of the initial chamber, and (c) presence orabsence of a lappet (an armlike projection of the shell margin), theywere dealing with a single dimorphic biological species. Thesmaller shells, bearing lappets, are hypothesized to be the males,by.analogy with the known morphological characteristics of dimor-phism in JiVingcephalopods. However persuasive these cases maybe, the formUlation of a hypothesis of sexual dimorphism based onthese kinds of data is actually based solely on inference from mor-ph?logy and can only be evaluated further with the phylogenetic err-tenon, as we develop below.

Species 99

D. --..... - or hosisof the acornbaf!18cle,Figure 3.4 Larvaldevelop~ent andf~~:~~g~ naupuus.(Fr~ Ba~I~~_aleBalanus balanoides. (a'lb9)7F31r~~-g~~~117.)(e-e) Metamorpth.?SI~r~~~aestner(1936),fromAnderson, . I dult stage.(FromRunnsromtachedcyprid stageto (e) early~ of GustavFischerVerlag.)1970 figure 9-15, with perrmsston .

' Iso comes directly into. .' of parentage a es-The reproductive cnten~n . c cle of individuals areplay when different stages I~ the ~eex~eriment. The life cycle oftablished by direct observation an r I discrete larval stages".Earlybarnacles (figure 3.4) includes sev.ea . and thus more typically

free-SWimming, Much re-in ontogeny the larvae are nted adult stage. .tdorceme, . ups.uscrustacean like than the ro~ eo, . th various parasite gro '.

search in systematics, partlcularl.y Inha: larvae found either free-liv-devoted to the direct demonstratl~n t the Ijuveniles of an adult

h orqamsrn areing or parasitizing anot er

100 Species

form parasitizing another, wholly different kind of organism. This isanother way in which demonstration of parentage can reduce thenumber of clusters of discrete taxa into hypotheses of the existenceof a single species.

Again, inference of continuity in the dissimilar ontogeneticstages of fossil organisms, or those elements of the Recent biotastudied only from preserved materials, becomes a matter of the eval-uation of intrinsic properties, and is not the same as the direct obser-vation of such ontogenetic transformation. However, cases abound inwhich larval stages have been plausibly linked to adult stages onthe basis of (a) associational (distributional) evidence, or (b) themore compelling demonstration of a complete series of morphologi-cal intermediates. Such demonstration is only possible within groupswhere progression between ontogenetic stages is gradual (e.g., inechinoderms such as edrioasteroids; see figure 3.5) and does not in-volve metamorphosis from one discrete stage into another, as inholometabolous insects and many parasites.

The reprOductive criterion also Comes into play directly in in-stances of hybridization. For present purposes, we can recognizetwo distinct aspects of hybridization: (a) occasional hybrids betweentwo species, in Which case the hybrids do not represent a taxon inand of themselves, and (b) situations where a new species has beenformed by hybridization between two pre-existing species. Only theformer consideration is relevant in terms of the general problem ofspecies recognition. Again, direct field and controlled laboratory ex-perimental eVidence is frequently adduced, demonstrating the oc-casional hybridization between two species otherwise considereddistinct. For example, hybrids between the American toad (Butoamericanus) and Fowler's toad (Buto tow/eri) are occasionally en-countered. and can be raised in the laboratory. But again, putativehybrids, (Le., specimens which appear to be intermediate betweentwo taxa otherwise considered discrete biolog ical species) aremerely suggestive. What is the significance of such observations?How rampant must hybridization be before two clusters, formerlyCOnSi?ere~discrete reproductive communities, are merged as a sin-gle b.lotoglCalspecies? We enter a gray area here: If the hybrids arefUI!y Interfertile with both parental "species," we reject the hypoth-~IS .th~t t~~se "Species" are in fact discrete. Any sign of reproduc-tive InViability Within the hybrids, Whether among themselves or with

II •I 1

l'llq~ J 1

DDDtJD

102 Species

either of the two "species," is presumptive evidence that the twoclusters indeed constitute two separate reproductive communities,coherent within themselves but imperfectly separated from eachother. Cases where only morphological intermediates are observed,with no concomitant breeding data, are incapable of decisive resolu-tion, and further evaluation can only come from analysis based onthe phylogenetic criterion.

In organisms reproducing wholly asexually, direct observationalso confirms patterns of parentage, but in any case would appear tobe of tittle use in reducing the number of putative diagnosable taxain the manner described above for sexually reproducing organisms.There are grounds for doubting whether any organisms are strictlyasexual, however, in which case the application of the parentage crt-tenon from the species concept we have adopted is generally valid.As long as occasional sexual reproduction occurs (as in, for ex-ample, Foraminiferida), the parental pattern of ancestry and descentcomponent of the species concept applies.

Thus the reproductive criterion, to the extent that parentage canb~ actually observed, is useful in the direct merging of separate,dlapnosable clusters of organisms into single taxa (biological spe-cres) Incases of (a) sexual dimorphism, (b) disparate morphologicalstages in ontogeny, and (c) in some instances of hybridization. In theabsence of direct evidence, the reproductive criterion may still besuggestive of such hypotheses in those instances where both extrin-sic and intrinsic features conform to known patterns in relatedgroups, but only as an inference much in need of further test.

~p.art from direct observations of parentage, the three ancillaryp~ed.lctlonsfrom the species concept (continuity in distribution of in-~nn~,~properties among individuals and continuity of distribution ofindividuals within the species in space and in time) offer furthermeans of uniting previously sorted population samples in order toevaluate various hypotheses of species identity and composition.

. Use of intrinsic properties-observable traits of individuals-atthl~ s~agedepends upon the nature of the subdivision of the basicumts In the first place. If the clusters of specimens were originallysorted because they seemed to be recognizably different, diaqncs-ab~et~a, then recognition of a continuous distribution of a variabletra.'t (figure 3.2), if such exists, would conflict with the already per-carved pattern of dis t· . . dcan mucus vanation, and further analysis waul

Species 103

have to include the search for synapomorphtes. However, if the clus-ters to be united were originally sorted on some other criterion--e.g.,different localities in space and time-then such general continuityof intrinsic properties, including identity among two or more sam-ples, arising as a prediction from the notion of parentage, may beused to unite clusters as a hypothesis of species identity.

Further preliminary merging of clusters can be effected by unit-ing clusters with contiguous distributions. If clusters were origi~~lIydelineated primarily on the basis of their provenance from localitiesseparated in space and time, and provided that the general aspectof the intrinsic properties suggests continuity among samples, thendistributional contiguity (Le., conformance to either of the two pre-dictions of continuity in extrinsic properties) may be us~d to mer.geclusters as a hypothesis of the existence of a biolcqical species(figure 3.3). Samples of organisms which appear to be nearly the"same"-Le., the within-sample variation is nearly as great as t~eamong-sample variation-and which, when mapped, show a co~tln.uous or quasi-continuous spatial distribution, may be hypothesrzedto be drawn from a single species. (see figure 4.12 for an example.)

This sort of approach, while arising as a prediction from the def-inition of biological species, and by conforming in a general way tothe perception of intrinsic properties, nevertheless s.uff~rsfrom twobasic flaws; (a) the negative use of the parentage cntertcn ~romthedefinition, and (b) the inability to specify limits, at least In m~recomplicated cases to the distribution. In other words, there re~alnsthe problem of either including too many or too few samples In anidentified species. . d. .

. " di lay a geographiC istn-Allopatnc ("of another country) taxa ISP. trt (" f the same coun-bution wholly exclusive of one another. Sympa nc a" I ·ng ranges such thattry ) taxa occupy the same, or at least over appr, . '

individuals belonging to these taxa come into direct ~o~tact.Parapatric distributions involve close approximations of the limits ofranges of essentially allopatric taxa. Although species ~ay ~CCUpyth . f ct occur In differente same geographic area they may In ahabitats within that area and thus not be truly sympatric, a~d there-f . b OO·ng Lions and tigers areore not have the opportunity for Inter re I. .f II h do not occur to the sameu Y capable of interbreeding, yet t ey t .habitats in India and therefore do not interbreed in natur~. Allopa ncdistributions-cases where there are disjunct distributions of taxa

104 Species

which mayor may not belong to the same species-are numerous. Insuch instances, obviously, there cannot be any positive evidence forinterbreeding (parentage), and the point remains moot, on distribu-tional criteria, as to whether or not one or two species are actuallyrepresented by the samples. In sympatric distributions, a pattern ofparentage may be observed within each of two units hypothesized torepresent distinct species, but the absence of evidence for paren-tage between the two "species" is negative evidence; the hypothesisthat the two putative species are in fact distinct, true species standsuntil rejected by evidence of interbreeding.

The third and final ancillary prediction of the parentage compo-nent of the species definition is continuity in temporal distribution.By definition, such parental continuity must remain unbroken throughtime. Temporal distributions, naturally, can only be studied em-pirically with fossils. There are few circumstances in the fossil recordwhere distributions are thoroughly continuous for any hypothesizedspecies except for relatively short intervals in the total duration of ahypothesized species. The only exceptions are for species whosetemporal durations are extremely short; a species may be knownfrom only a single occurrence, in which case, for all practical pur-poses, it is not known to have a non-trivial temporal distribution. Inall other instances, vertical gaps occur in ranges of taxa, and the in-ference is made that (1) the species was living there but individualswere no.tprese~ed, or (2) the species was living elsewhere duringthat pen?d of time, when the environment was perhaps inimical tothe species. In other words, the species remained in existence, butoccurred elsewhere. But what limits can be drawn to such infer-ences? A paleontologist might be willing to "allow" a gap of a fewthousan.d'. or even perhaps a few million, years, but would probablybe un~lIllng to hypothesize that two samples, one from the MiddleJurassic and the other from the Upper Cretaceous, with no known in-tervening samples and otherwise appearing to be exactly the same,would actually represent the existence of a single species. Such atem~oral gap (some 100 million years) would seem to be too long-yet It must be admitted that it might be possible for a species to existthat .Iong.Again, negative evidence of this sort is simply not a con-clusive means of test" h .

109 a vpothests of species composition.We conclude that the basic proposition of the biolog leal species

Species 105

concept-eontinuity of parental ancestry and descent within but notwithout the species' limits-serves as positive evidence for themerging of diagnosable clusters into species. A systematist mayreject the hypothesis that two samples are drawn from two discretespecies, if it can be shown that they are males and females regularlyinterbreeding in nature, But failure to demonstrate such a pattern ofreproductive cohesion does not reject the hypothesis that the twosamples constitute a single species, as there are too many sourcesof experimental and observational error, as well as problems of allo-pattie and allochronic distribution, which could account for such afailure. Likewise the three ancillary predictions arising from theparentalcomponent of the species definition offer only weak tests ofa hypothesis that any two samples may actually represent a singlebiological species. Negative evidence, whereby samples show nei-ther unimodal distribution of intrinsic properties, nor continuous dis-tributions in space and time cannot constitute a particularly strongrejection of the hypothesis that any two samples represent a singlebiological species. Clearly, there is a need for a stronger means oftesting hypotheses of species compositions. Such a test must beconsistent with the parental criterion and the evidence adduced forthe three predictions arising from this criterion. But it must be in-dependent. As such it must arise from a wholly different predictionfrom the definition, The only way for this to be effected is to look tothe other component of the species definition we have adopted ear-lier.

EValuation of Hypotheses of Species Composition:The Phylogenetic Criterion

Withthe exception of sibling species, the procedure for forming andevaluating hypotheses of species composition under the reproduc-tive criterion initially overestimates the number of species present ina sample of organisms. In contrast, as we showed earlier in thischapter, hypotheses of species composition based on distributionsof ~ynapomorphies (one or more autapomarphies for a minimallydefined monophyletic group) underestimate the number of species~resent.This discrepancy arises from the status of species as evolu-tionaryunits. Some species, perhaps including some represented by

- ..

106 Species

the samples at hand, are ancestral to others. We do not knowwhether ancestors are present in the sample or not.e The possibilitythat some of the species within a sample may be ancestral to otherspecies (whether or not the direct descendants themselves are in-cluded in the sample) implies the possibility that not all species in asample will be characterized by its own set of synapomorphies. Thusa straightforward application of the procedures of cladistic analysis,which evaluates hypotheses of synapomorphy and hence hypothe-ses of composition of monophyletic taxa, may not be sufficient torecognize all species represented in a sample. It would tend to un-derestimate the number of species actually present.

There is one case in which the phylogenetic procedures-search for patterns of synapomorphy-may overestimate the actualnumber of species represented in a given sample of organisms. Thiscase involves a species morphologically well differentiated over itsgeographic range. A feature of most models of allopatric speciationis that morphological differentiation (Le., development of evolu-tionary novelties) may occur within the geographic range of a spe-cies prior to the onset of fult geographic (and hence reproductive)isolation among the populations. Thus, by the reproductive criterion,there is one species. But. by phylogenetic analysis, specifiable clus-ters, complete with autapomorphies, would indicate the presence oftwo (or more) species.

Such cases commonly arise in the analysis of atlopatric or para-pattie distributions of distinctive populations which themselves dis-play some morphological homogeneity. In the hypothetical example,It was assumed that reproductive isolation was known not to have~curred. In reality, the number of species present in such a situa-tion must remain moot, as earlier noted.

Nonetheless, terminal species, as well as species without anyd~scenda~ts (no matter how remote) within the sample being con-stdered, Will be diagnosable in terms of uniquely derived characters,or autapomorphies. This consideration suggests a general proce-dure: patterns of similarity (synapomcrphy) should be evaluated ac-cordmg to th~ criteria developed in the preceding chapter on clade-gram analYSIS.Those organisms which do not cluster into piles

2. As will be devel"""d . .., ' . 1.~• -y- In our UlSCUSSlOnof panerns of speciation, ancestral Species can I

on, ccevat wlIh and perhaps even outliving, their descendants. Thus fossils are not requited forancestors 10 be present in a sample.

I

Species 107

basedon synapomorphy will, therefore, appear relatively plestomor-phousin all features considered. According to the criteria developedin the preceding chapter, it is impossible to postulate the existenceof a (monophyletic) taxon based solely on plesiomorphous similari-ties.Yet, as we have seen, species may not conform to the definitionof monophyly.The residual organisms are diagnosable, presumably,only in terms of the absence of apomorphles. or, put in a more posi-tivefashion, the unique retention of plesiomorphies. It is to be notedthatwe are not here considering whether or not a species is an an-cestor(see chapter 4). We are concerned, rather, with the problem ofrecognizing species, which might happen to be ancestors, as spe-cies in the first place.

Further evaluation of the identity of collections of such plesio-morphousorganisms as species can only come from a comparisonof the clusters resulting from phylogenetic analysis with those sug-gested by the reproductive criterion and its three ancillary predic-tions.The procedure in the latter case initially overestimates the ac-tual number of species, whereas in the former phylogeneticprOCedure,the number is generally underestimated. Should the ple-siomorphousclusters resulting from the phylogenetic procedure ap-pear cohesive under the reproductive criterion as outlined above,the hYpOthesisthat the plesiomorphous cluster constitutes a singleCoherentspecies should be provisionally accepted. The hypothesis?analways be rejected by showing that a portion of the species soIdentified is more closely related to (shares synapomorphies with)SOmeother group or is itself autapomorphous and best regardedas a distinct species.

In a situation in which one sample is plesiomorphous with re-spect to a second-and thus possibly its ancestor-we are meth-odologically forced to recognize two discrete species where, in fact,there might only be one. If the two species are contemporaneous,thereis no confusion. We recognize two species and then consider,a?Cordingto the procedures set forth in chapter 4, whether the ple-StOmorphousspecies was in fact ancestral. But if the first, plesiomor-phcuaspecies is exclusively known from an earl ier period of time:an that of the relatively apomorphic species, the possibility arisesn at the two "species" are samples from a single reproductive con-Inuum. Should this be the case, under the species definitionadopted here, the two samples would be ccnspecitlc. Method-

108 Species

oloqlcally, however, there would be no way to establish the (former)existence of the continuum. We would be forced to continue to rec-ognize two distinct species.

Analysis of Species Composition: A Summary

We have developed two approaches to formulating and testing hy-pothes~s.~f species composition. Each stems from a component ofthe .deflnltlon of species adopted in this book. The reproductive co-~estonaspect or component of the definition initially tends to overes-timate the actual number of species. The number of diagnosableclusters is reduced primarily by direct demonstration of patterns ofparental ancestry and descent; hypotheses are further evaluated byref~rence to ~redictions of continuity in intrinsic and extrinsic prop-erties of species. The latter predictions are, however, weak.

The second approach " , I .. . .s a vanant a cladistic analysts. Speciesancestral to other species in the sample cannot by definition,possess a uni I d ' 'rquey enved set of characters. We have advocatedthat clusters of plesi h '". romorp ous individual organisms can be hypoth-esized to constitute a b! I' ,'th h 10 oqica! species. In such a case agreementWI t e r~productive criterion is particularly crucial. The' hypothesescadn.~ rejected with additional information showing that some of theIn IVlduals allocated to th h .'Ih e ypctheslzed plesiomorphous speciesare er er related to h. some at er cluster, or are members of anotherunique group,

Species Recognition in Practice: Some Examples

~e~ ~ew m~erial Platruck and Shadab (1976) have recently re-scribed S:m:P~7er genus ~imiro.mus. In so doing, they have de-viously b new species (Le.. species which had never pre-

een recognized' th .procedur In e taxonomfn literature). Their generale was as follows' II 'lected field . co ectiona were procured (newly col-

samples of m .amined in the I be useum specimens) and prepared and ex-a ratory Spec' . .

synapomorphies with' lIl~ens were recognized as shO~lOgother Species of the genus Zimiromus, r.e.,

Species 109

they conform to the synapomorphies in the diagnosis of the genus.Such allocation permitted comparison with other species of thegenus, which was essential for the evaluation of the distribution ofcharacters within the samples at hand (because a higher-levelhypothesis must be accepted prior to analysis to permit outgroupcomparison; see chapter 2).

Subsets of the specimens were further recognized as being, insomeways, unique, i.e.. they were different in some specifiable wayfrom atl other known species within the genus. They were thus diag-nosable, and the characters utilized for the diagnosis were unique tothe group, but not necessarily apomorphous. For instance, the twospecies Z. penai and Z. brachet (both from Ecuador) were distin-guished solely by the relative width of the median epigynal ducts (afeature of the reproductive tract in females). One state may well beplesiomorphous, the other apomorphous. What matters is that a char-acter be unique to a (putative) species; a uniquely retained ptesio-morphy is as valid a criterion of a species as a uniquely derived fea-ture.

In instances in which both males and females are known (e.g.,the description of Z. jamaicensis), there were two initial clusters. In-asmuch as spiders in general are known to reproduce sexually andmales and females of Z. jamaicensis share a large number of intrin-sic features, the procedure was simply to include both males andfemales in the species, even in the absence of any direct informationon mating behavior.

Platnick and Shadab (1976) also described Z. bimini from SouthBimini, in the Bahama Islands, The males of this species share anapomorphous condition with males of Z. jamaicensis (the presenceof single, long retrolateral apophyses). Thus the authors recognizeda nested set of synapomorphies between the level of the genus andt~atof its discrete, included species. Why then were these two spe-ctes not simply recognized as comprising a single (by virtue of thesynapomorphy) albeit variable (by virtue of the diagnosable differ-ences) species? In point of fact, were the pattern of parental an-cestry and descent known in detail for both groups, it is conceivablethat the pattern would be shared by the two (despite the rather largegeographic distance between the two known samples) and that thust~ey would constitute a single species. But absence of any suchdirect information left the systematists no choice: they clustered the

---"~

no Species

diagnosable units within the spiders sampled (Le., males and fema-les, etc.) on presumed parental criteria, consistent with the modifiedversion of cladistic analysis that we have presented.

Case 2. Revision of species Much of the work of the practicing sys-tematist lies in the revision of hypotheses of species composition al-ready present in the literature. Rearrangement of hypotheses of spe-cies composition has two aspects. One is the removal of somesamples previously allocated to a certain species. In such a case,the samples so removed are referred to one or more other species,either already described, or described as "new," In so doing, theconcept of variation of the intrinsic or extrinsic (or both) properties ofthe species is decreased. The second aspect is the inclusion of ad-ditional material (newly found, in which case the range of variabilityof intrinsic and extrinsic features of a species is expanded), or twospecies are found not to be separate, diagnosable groups, as pre-viously hypothesized in the literature. In the latter case, the speciesare synonymized, i.e.. are formally merged, and the name chosenfollows the Rules of Zoological Nomenclature. A typical species revi-sion may encompass both aspects: inclusion and elimination, in therefinement of a hypothesis of species composition.

As an example of simultaneous exclusion from one notion ofspecies composition, with concomitant inclusion in another, in revi-sionary work, Eldredge (1972) studied a large sample of availablespecimens of the trilobite genus Phacops in (Upper) Middle Devon-ian rocks of eastern and central North America (figure 3.6). Severalspecies had been previously described. All but two had been synon-ymized by earlier workers. Eldredge examined the type material forall described species and concluded that, on the basis of the exami-nation of the morphology (i.e., solely on the basis of intrinsic proper-ties), the previous synonomies appeared correct: some of the pre-viously described species appeared indistinguishable from eachother as diagnosable taxa. There seemed to be only two basicallyseparate, diagnosable clusters of specimens, conforming to the de-scribed species Phacops rana and Phacops iowensis. One of these,Phacops rana, exhibited obvious subclusters, whereas P. iowensisshowed far less variability.

Examination of samples, particularly from the Michigan Basin,revealed that many of the Phacops specimens previously identified

Species 111

o. b.Figure 3.6 Comparison 01 (a) Phacops rana rana and(b) Phacops iowensis. Though similar In general appearance,the two species differ in a number of specifiable attributes.including the number of columns of lenses in the eye.

as P. rana shared instead those general features of P. iowensis mostcritical in distinguishing that species from P. rana (e.g, number ofcolumns of lenses in the eye, and mode of development of the tuber-cles on the external side of the dorsal cuticle). Again, whether thesedifferent characters were primitive or derived was irrelevant to theanalysis of species composition. In fact, the sister-species of P. ranaappears to be a slightly older European and African species,whereas P. iowensis apparently shares a pattern of synapomorphywith older species in North America and elsewhere. The reclassifiedspecimens simply shared the attributes (intrinsic properties) of P.iowensis, rather than those of P. rana. As formerly constituted, thesamples did not fit evenly into discrete, diagnosable groups.

The diagnosable subclusters of P. rana presented the familiarproblem of assignment of such clusters to the appropriate rank. Inone instance, two subclusters were nearly identical except in termsof the total number of lenses in the eye; earl ier systematists had sug-gested that these two diagnosable clusters represented a case ofdimorphism, presumably sexual. The early ontogeny of these twogroups is identical; in addition, sexual dimorphism in lens number isknown in other groups of aquatic arthropods (e.q., Umulus polyphe-mus the horseshoe crab). But the two taxa seldom cooccur in thesame bedding planes (which perhaps would suggest sympatry) andoccur alone in some regions. For this reason, Eldredge (1972) re-jected the hypothesis of sexual dimorphism and did not merge thetwo subclusters.

Each of the five diagnosable subclusters within Phacops rana

.'" ==-=--~---

.

112 Species

was defined on the basis of unique intrinsic characters. Whenmapped, the distributions of each proved continuous (as nearly soas fossils are likely to be) in space and time (see figure 4.12). El-dredge chose to label each of these five subclusters as a "subspe-cies," but the point is clearly moot. Indeed, the procedures outlinedand illustrated in this chapter suggest that each of the five "subspe-cies" should have been accorded status as actual discrete species.However, it makes little difference as the point necessarily will re-main moot in the absence of direct information on reproductive pat-terns within and among the diagnosable taxa.

I

Chapter

4Modes of Speciation and the Analysisof Phylogenetic Trees

TIE GENERAl. rnothodoIogy lor pooducong el~_-oped in chap .... 2 __ only one _Ill 8UUI11Plian _ !helMllulianaly proc_: Ihol -.aoon hoi occurred end hoi ~.- ..... 01 evolutianaly ~ (~ eMilie_a).The considenlhtw'l of the nature of spec .. ...-d their ideOOficabon__.e_~0I!he-.a....., ~. ThoBCluBlLWlits of 8'\IOlution are the reproducttye corm'U'libes called spec ...That some spectes are ancestral to 0Chefs necessrtales 8 modIfica-tion of normal cladistic procedures 1Mthe analystS of spec .. com-posrtoon IchaplBf 3) Evolutoon 01 __ spec .... end !he d.. ,.,,_,of evolutionary novelties are nor one and the same theng. Though re-lated. nove"ies can be attaIned without new spec.teS betng tonnedand the converse is lnJe as well We now dtseuss the topIC otphylogenet:iIC tJeea, their nature and methOd of conSlruchon And mconsidering trees we shall need 10 examine more dosety the ectualevotution.-y petlems ot the onglfl of new spec ..

We define a phyk)geneIlC tree as a br8flChlOg di8Qfam that de-POClSactual -.... 01 ~ and ~ among • _ .... 01 ....Thus trees are far more specifIC sorts of SUdef1leOtS than cladogramsCfadograms depICt nesled sets of ~ees.. thereby defining:monophyletiC groups and 5lmuttaneously presentIng a hypolhes.lsof the relatlOnatups among the uua Ciadogtams oeocr rete-leonshtpl ,n a relalrve way laJlOn A and taxon B are more ctosefyrelated 10 one .-other tr.l erthef is 10 taJa:)n C For Instance. the_ement IhBl C8II and dOgs lFelodae end c.nodae' ..... moreclOl8ty "'818(1 to -=tl otheI than ellhef II 10. say, seals (PIn-nipedia) iB an -..pIe 01 • elBdootoe hypoIhes<a Such _

,,

•

114 Speciation and Phylogenetic Trees

can be made about any presumed monophyletic cluster, of whatevercategorical rank.

Trees, in contrast. depict specific hypotheses as to the mannerin which taxa are related. We have already concluded that, logically,only species can serve as evolutionary units. Taxa of higher rank aremerely monophyletic aggregates of one or more species, and thusdo not exist in the same sense as do species and cannot serve asancestral or descendant units (see also Wiley 1979; for a contrastingview, see Bretsky 1979). Nonetheless, the literature is replete withphylogenetic trees depicting ancestor-descendant relationshipsamong genera and taxa of even higher rank. Thus our conception ofphylogenetic trees is more restrictive than the usual notion. Phy-logenetic trees as understood here are branching diagrams showingpatterns of ancestry and descent among species. In chapter 6, weshall discuss the use of highly corroborated phylogenetic trees inevolutionary theories of species formation. At this juncture, we brieflyreview generalizations on patterns of speciation to expose the possi-ble sorts of trees that can be produced.

Speciation and Phylogenetic Trees 115

Consideration of the modes of origin of new species is compli-cated by the general model of phyletic evolution. We briefly men-tioned phyletic evolution in chapter 3 in discussing the problem ofhandling samples a million years apart within a hypothesized sin-gle lineage, unbroken in terms of parental ancestry and descent. Thepoint here is that phyletic transformation of one or more characterswithin such a lineage is frequently alleged to have produced de-scendant morphologies so different that the systematist has nochoice but to recognize two discrete species. From a pragmaticstandpoint, an ancestral species evolves directly into a descendant(figure 4.1). Procedures for tree construction outlined later in thischapter offer a general means of distinguishing situations in whichnew species arise by splitting or "budding" from those representingphyletic change within a single species, a microevolutionary pro-cess culminating in "pseudospeciatfon." For classificatory purposes(see chapter 5), all diagnosable species, whatever their mode of ori-gin, are treated the same,

Patterns of Speciation

I,,/ GAP,,,

F

bQ. c.

Species 2 Sampled Species 2

Accepting the concept of species as discrete reproductive units, itfollows that speciation is the origin of new reproductive communi-ties. Such an event can only result from a pattern of splitting, orbudding off from, an ancestral species. Descendant species arise,by definition, from a portion of the ancestral species. The only alter-native, the wholesale, complete phyletic transformation of one entirespecies into another, is inadmissable because there is no postulateddisruption of the continuum of parental ancestry and descent in sucha situation Separately diagnosed species resulting from such phyle-tic transformation are rather to be regarded as end members of anunbroken continuum: as we have earlier remarked, such endmember "species" might be diagnosed as two, but constitute a sin-gle reproductive lineage and thus constitute a single evolutionaryentity. Our review of speciation will therefore focus on modes of ori-gin of new, discrete reproductive communities from ancestral spe-cies.

•E..:

Species 1 Sampled Species 1

.. Morphological Change ..

Figure 4.1 Aspects of the neo-Darwiruan concept of the derivation of newspecies by transformation. (a) Transition from species 1 to species 2. (b) Thetransition is viewed as even and gradual, such that specifiable differencesbetween species emerge only if samples widely separated in time are con-sidered. (c) The incompleteness of the lossil record provides the gaps be-tween samples, obviating the necessity for the systematist to divide the con-tinuum in an arbitrary fashion.

•

116 Speciation and Phylogenetic Trees Speciation and Phylogenetic Trees 117

cized as an unjustified extrapolation into an inappropriately muchlarger time dimension (Eldredge and Gould 1972, 1974; Gould andEldredge 1977; also chapter 6). Of greater importance at this point isthe empirical evidence of pattern" what is the evidence for (a) within-species gradual change through time, and (b) what is the evidencefor successional occurrence of ancestors and descendants? The firstkind of evidence is crucial to the neo-Darwinian theory of transforma-tional speciation (but not to saltation ism), and the second kind of ev-idence is crucial to the entire concept of transformational speciation., T~e dearth of examples of continuous change within lineages,Including change within segments designated as nominal species,in the fossil record has been recognized since Darwin's day (seeespecially Gould and Eldredge 1977, for a review of putative exam-pies of phyletic evolution). Adherents to the neo-Darwinian theory oftransformational speciation resort to the ad hoc hypothesis that thefossil record is too poor to reflect adequately this mode of evolution.And there is little reason to doubt that the fossil record is indeedspotty. Nonetheless, there are many cases where species, recog-nized according to techniques developed in chapter 3, have quasi-continuous ranges; in the case of marine invertebrates, these rangesmay be 5-10 million years in duration, and in some instances evenmore. Within species delineated in this fashion, there is generally nonet accumulated change in those intrinsic properties used by sys-tematists to discriminate alleged ancestors from their supposed de-scendants. Most of the examples purporting to document such grad-ual transformation have since been rejected by subsequentworkers. '

On the other hand, the more general proposition, i.e., that spe-cies succeed one another and do not display overlap in time (whichis the more general prediction of the transformational hypothesis)remains a more open question. Indeed, many apparent ancestors

Phyletic Transformation: Neo-Darwinism and Saltationism

As characterized in greater detail in chapter 6, models of phyleticevolution stem directly from notions of the transformation of intrinsicproperties of organisms. They are not, at base, theories of speciationat all. The most common version of phyletic evolution in the English-speaking world is nee-Darwinian in nature. The essence of the neo-Darwinian model of phyletic evolution is that, over time, the genetic(hence biochemical, physiological, anatomical, and behavioral) at-tributes of species become modified such that, given the elapse ofsufficient time, were the lineage to be sampled, it would be found todiffer so much in its intrinsic properties that a systematist would rec-ognize diagnosably different "species" (figure 4.1 b). The theory ern-phasizes a completely continuous, evenly gradational and usuallyrather slow transformation of the intrinsic properties of the ancestralspecies into those of the descendant. Paleontologists and those theyhave influenced (see Cain 1954:111 for a lucid exposition of thismodel) have repeatedly stated that, were the fossil record perfect,systematists would be faced (even plagued) by the continuousseries of forms that could only be broken up in the most arbitraryfashion and pigeonholed into the taxonomic hierarchy. Thus, fromthis point of view, we are told that it is fortunate indeed that the fossilrecord is so incomplete (figure 4.1c), such that the isolated glimpsesafforded by sporadic preservation of samples of these otherwisecontinuous, gradational lineages allows systematists to recognizediscrete taxa in a non-arbitrary fashion: the gaps in the fossil recordprovide the necessary arbitration.

We discuss the putative mechanisms underlying phyleticchange in detail in chapter 6. For the moment it is only necessary toconsider the evidence appropriate for the evaluation of this model asa general hypothesis of morphological change in the evolutionaryprocess. Two independent bodies of evidence have been cited insupport of the general notion of phyletic transformation as a majormode of evolutionary change: (a) extrapolations from experimentaland theoretical (mathematical) genetics and (b) empirical evidencefrom the fossil record Only the latter is directly concerned with pat-tern The former is concerned with a hypothetical mechanism andhas been ably characterized and defended by Simpson; we discussit in greater detail in chapter 6. The model has been severely crtti-

1 Many of the classic examples of phyletic gradualism in the paleontological literature arestudies performed in Great Britain between 1899 and 1919. These include Rowcs stud .. of Cre-taceous echinoids (1899), Carruthers' (1910) ,tudy of the Carboniferous coral "Za;hremis"delanouei, Trueman's (1922) paper on the Jurassic oyster Gryphaea. and Brinkmann's (1929)examination of the Jurassic ammonite Kosmoceras. Restudy of these and other cases has consu.rently failed to verify the gradualistic patterns of morphological change claimed by the originalauth~r. (Some cases have yet to be thoroughly restudied; see Gould and Eldredge 1977. for dis.cussion and .Cltat.,on of subsequent reevaluations of these and other putative examples of phy-lenc gradualism inthe paleontological lherature.}


Figure 4.2 Stenzel's example oftime-successive specieshypothesized to form anancestor-descendant sequence,and thus an example of derivationof diagnosable species by trans~formation. The upper three speciesare of the Cubitostrea selliformisslock, the lower the putative ances-lor, Stenzel's explanation follows:"These are outlines of the leftvalves of large specimens of eachspecies, They show the develop-

I men! of auricles and increase In

t shell size from species to species,The species are arranged with the

~ earliest at the bottom and the latestat the top [N,S: species are _

---:~ confined to the geologic formationsi named at the right side of theII' ctaqrarn-caus.]. Horizontal barsI are the extensions of the hinge9 axes, Auricles are shaded differ-

'----- ently than the main shell body.~~ Some growth lines are indica,ted by

dashed lines," (From Slenze1949:44, figure 8.)

seem to disappear from the record, succeeded by apparent descen-dants. These successively occurring taxa are recognized and char-acterized in the manner outlined In chapter 3. For example, Stenzel(1949; see figure 4,2) described a case in which four species, whichhe hypothesized formed an ancestor-descendant sequence, ap-peared to succeed one another in time, and in no instance was t~eancestral species known to have survived as a contemporary of Itsdescendant. Thus the general hypothesis of transformational specia-tion cannot be ruled out.

The most striking aspect of the nee-Darwinian (see chapter 6, p.246, and footnote 1 for a discussion of the meanings of the terms"neo-Darwinian" and "synthetic") theory of transformational specia-tion is that it is not really a theory of speciation at all. Rather, it is atheory of change of intrinsic properties, effected primarily by natu.ralselection working on a groundmass of genetic variation. Species

•


emerge, at the end of an analysis, as semi-arbitrarily defined clus-ters of convenience. There is no qualitative difference alleged to dif-ferentiate species from genera and other higher taxa. Evolution pro-ceeds "at the species level" but produces successive "species" thatare all part of the same evolving lineage. Occasional splits in thelineage produce diversity, but genera and taxa of higher categoricalrank are ex post facto summations of species-l ineages that go on in-definitely. Species have no discrete limits in a phylogenetic sense;this view of the evolutionary process is, at base, incompatible withthe view of species developed in the previous chapter and is notconsidered further until chapter 6.

"Saltation ism" is less coherent than the nee-Darwinian view as atheory of transformational speciation. There have been several dif-ferent mechanisms proposed for saltation. Moreover, neo-Darwlnistshave nearly unanimously viewed saltation as the only competingparadigm and have caused its nearly total eclipse in contemporarybiology. Neo-Darwinists claim to have a mechanism (natural selec-tion through geological time, as an analogue of experimental and ar-tificial selection; see chapter 6), whereas most saltation theorieshave invoked mechanisms even more difficult to establish on an em-pirical (including experimental) basis,

However, saltationism-literally, the view that evolution pro-ceeds by sudden jumps from one state to the next-never whollyabandoned the notion that taxa (species) do the jumping (figure 4.3).Thus at least most theories of saltation are theories of speciation.eAlthough most discussions of saltation by proponents and oppo-nents alike seem to focus on transformational problems (e.g., ex-plaining the origin of major structural differences among taxa withinmonophyletic groups), the creation of a "hopeful monster" directlyimplies the creation of a new species. Indeed, the only cogent argu-ments against particular theories of saltation appear to be thosewhich see grave difficulties in creating a new, sexually reproducingspecies through the appearance of a single, individual hopeful mon-ster. It is interesting, in this connection, that the term "mutation" was

2, In discussing sahationism as a general alternative to neo-Darwinian processes in the explana-tion of patterns of phyletic evolution, we do not wish to imply that saltational models have notalso been concerned with splitting phenomena-speciation in the strict sense. Carson (1975l,for example. writes of "saltatinnal speciation." But much. and pernap, the bulk. of the oldersaltaticnal literature deals with ancestral and descendant species occurring succe,sively, in aphyletic sense.


Species 3

Species 1

.. Toxic discontinuity •

Figure 4.3 Speciation by transforma-tional saltation.Eachspecies is derivedthroughthe transformationof the ances-tral species into its descendant.Trans-formationsoccur as sudden and discreteevents(horizontaldashed lines), how-ever,ratherthanby gradual, progressivechange,the usualversion of the nee-Darwinianmodel.

coined, not by a geneticist. but rather by the paleontologist waacen(1869). And as Simpson (1944:48; 1953:81) has correctly pointed out,Waagen's mutations were taxa which, in a taxonomic context hecalled "varieties," but which seem to conform to the criteria of spe-cies recognition discussed in chapter 3. Waagen's mutations in-volved a concept of time-successive (i.e., nonoverlapping temporalranges) monophyletic taxa arranged in a direct ancestor-descendantsequence. As previously acknowledged, the question as to the oc-currence and relative frequency of such patterns remains open, in-sofar as the analysis of the fossil record is concerned. There is atleast no evidence that clearly and decisively refutes the generalproposition of saltation as a pattern of morphologically discrete,clearly differentiated successive species related in an ancestral-descendant fashion. Any particular model of saltation proposing a~pec~fic set of mechanisms should be experimentally testable, and itIS stnctly on this basis that saltation ism would appear eventually tos~andor fall, unless the general pattern predicted to occur in the fos-s!l record is itself ultimately shown not to exist.


Speciation by Splitting

If phyletic transformation, whether guided by natural selection orthrough some other agency, does not suffice as a model for theproduction of new reproductive communities, it follows that the onlyrea! mode of speciation is that of splitting, the budding off of a por-tion of a reproductive community to form a new, descendant unit. Incontemporary biology the words "speciation" is actually taken tomean "formation of new species by splitting." The definition of spe-cies adopted in the preceding chapter virtually requires such a pro-cess for the formation of new species. In his succinct review oftheories of speciation, Bush (1975) refers to three basic types ofspeciation: allopatric (with two basic variants), parapatrtc. and sym-patrie. All three involve the derivation of a descendant species bythe splitting off of a portion of the ancestral species.

Use of the terms allopatric, parapatr!c, and sympatric as de-scriptive modifiers emphasizes a central unifying factor in contem-porary speciation theory: geography plays a key role in speciation.Speciation, the fragmentation of an ancestral species into two ormore different species, can happen in another place (allopatric), andadjacent places (parapatric), or in the same place (sympatric). Thusstages in the differentiation and development of new species can, atleast theoretically, be mapped. In conjunction with the represen-tation of phylogenetic histories by branching diagrams, the "map-ability" aspect of the speciation process leads directly into the re-lated field of historical biography, the study of the distributionalhistory of the earth's biota. Nelson and Ptatruck (1980) have dis-cussed the relationship between systematics and biogeography atlength, and the diversity of methodological approaches to this fieldis well treated in a recent symposium volume on vicariance bioge-ography (Nelson and Rosen 1980).

The core of any viable speciation theory is a mechanism for thedisruption of the within-species pattern of parental ancestry and de-scent. A phylogenetically descendant species becomes reproduc-tively isolated from its ancestor. Allopatric distributions, particularlyinvolving species of low vag iI ity, offer a de facto disruption of thepattern of parentage: disjunct populations simply cannot share a co-hesive pattern of parental ancestry and descent unless there are dis-persal mechanisms that bring them into contact from time to time.However, such disjunct distributions offer no prima facie evidence


that two (or more) species are involved (see chapter 3). Shear (1975)has described the occurrence of the phalangids (harvestmen)Caddo agilis and Caddo pepperella from Japan, These species areotherwise known only from eastern North America. But sampleswithin each species from eastern North America and Japan are in-distinguishable (in terms of the characters examined thus far) andtherefore, according to the criteria developed in the preceding chap-ter, there is no basis for recognizing these disjunct allopatric occur-rences as representing two separate species, inasmuch as they arenot separately diagnosable. The problem of the lack of an immediateand tight correlation between (apparent) reproductive isolation onthe one hand, and morpholog ical divergence (attainment of novel-ties, diagnosability) on the other is raised. Geographic isolation isan important (some would say "necessary") cause of disruption ofpatterns of parental ancestry and descent. Such disruption is thesine qua non of speciation But such disruption may not be reflectedin intrinsic properties needed to determine the presence of two spe-cies rather than one (see Ehrlich and Raven 1969).

Nonetheless, from the earl iest days of analysis of modes of spe-ciation, it has been recognized that geographic isolation affords thesimplest. most direct means of effecting reproductive isolation.Wagner (1869), Romanes (1886), and a number of other nineteenth-century biologists noted that apparently closely related distinct spe-cies tended to "replace" each other geographically as ecological"vicars" or "vicari ants," suggesting that the formation of new speciesis primarily a matter of geographic differentiation followed by repro-ductive isolation. Mayr (see especially 1942, 1963, 1970) has mar-shaled a large amount of data from the primary literature and hasconcluded that allopatr!c speciation is the predominant mode of for-mation of new species, at least among sexually reproducing animalspecies, This view has come to prevail in recent years in contempo-rary biology.

Bush (1975) has made a heuristically useful distinction betweentwo basic variants of the theory of allopatric speciation (figure 4.4).In Bush's (1975) type a allopatric speciation, an ancestral species ofrather broad geographic distribution becomes subdivided into twoor more daughter-species. In such cases, the geographic range isgenerally thought to become disrupted through some change inphysical geography within the ancestor's range. The two sections

<2E,•o•

".- <••, Eo.DO

-•- "- .O~<

< 02 -..

o >•.~e.-E ~ In< --".c:: :::iD·0<-~ :;a:w

z-••~N

.,"&,>•

•- .:::i:>=""cig~co c,0: EQ ,c: ..EE.c-->~iQ• •~o- ..e a."00<

- 0.-~>c •-~~=• •- .c,

,0"0_o •Co••o.-i'Do __ 0

£

,;; ".-,;.0.0

o.=_ 0

"0e-•

.·-·.• 0o.00~=00• •o


simply gradually diverge-gene flow (i.e., mutual pattern of parent-age) is automatically cut off, and each subunit gradually adapts (vianatural selection) such that, should sympatry reoccur at some laterpoint in time, the two subunits would be unable to reestablish a mu-tual pattern of parental ancestry and descent. They will have becometotally reproductively isolated as an accidental byproduct of theirphysical separation. In such a modeJ, natural selection is not en-visioned as establishing reproductive isolation "deliberately" itself;rather, through selection and perhaps genetic drift, enough geneticdifferences accumulate to prevent the formation of viable hybridsshould the opportunity subsequently arise. Or the reproductive be-havior of the two units could have become sufficiently different inisolation so that hybrids will not be formed (see Dobzhansky 1951 fora thorough review of isolating mechanisms). The point to be stressedhere is that, in isolation, the two units simply go their own way to apoint where they cannot merge on the eventual onset of sympatry.Thus the model itself is a hybrid between the transformational viewof phyletic change or "speciation" and splitting per se: intrinsic prop-erties are gradually and slowly modified through time in two or morediscrete but large populations that started as fragments of a widelydistributed ancestral species. In cases where both fragments di-verge perceptibly, such that each becomes diagnosably differentwith respect to the common ancestor (some situations in the fossilrecord might, at least theoretically, offer the possibility of recogniz-ing this pattern), the pattern shown in figure 4.5a results. If one of thetwo fragments remains indistinguishable as a diagnosable taxonfrom the ancestral species, then the pattern of figure 4.5b emerges,which is indistinguishable from the basic pattern of splitting whichresults from the other model of allopatric speciation (type b of Bush1975), as well as parapatric and syrnpatr!c speciation.

To the extent that the gradual divergence model of allopatricspeciation can be corroborated {i.e., if it seems to fit the data insome cases better than alternative models), it affords perhaps theonly means of preserving the transformational aspect of speciationwithout abandoning the concept that species are discrete evolu-ti?n~ry ~nits. According to Bush (1975), this mode of allopatric spe-ciation IS a long-term process and is fairly common, especiallyamong terrestrial vertebrates.

The second mode of allcpatrtc speciation, the development of


a. b.

Species2

Species3

Species1

Species2

Species1

Species1

Figure 4.5 Expected consequences of various speciation mod-els. (a) The expected outcome of type a auooatrtc speciation(Bush 1975). The two descendant species, derived from thesplitting of the ancestrai species, each develop autapomorphies.The ancestral species becomes "extinct by transformation." (b)The expected outcome of type b allopatric, parapatric, and sym-pattie speciation. The descendant species develops au-tapomorphies and the ancestral species persists unchanged.

peripherally isolated populations much smaller in size than the en-tire ancestral species, is, again according to Bush (1975:346) muchmore common than type a allopatric speciation. As in type a, selec-tion in type b allopatric speciation is envisioned as effecting changein the adaptations of the isolated population, but smaller populationsize makes more rapid change possible. The expected result (figure4.4) is a vicariant initial distribution of the ancestral species and itsdaughter species, which has become diagnosable by virtue of de-veloping specifiable autapomorphles-c-evotutionary novelties whichin this case are hypothesized to relate to adaptation, the occupationand exploitation of niche space. Thus the resultant pattern is asdrawn in figure 4.5b, where a new, discrete species arises from anancestor (which continues on unchanged, hence not diagnosable asa "new" species itself) in a relatively short span of time (geologicallyinstantaneous). Moreover, the initial geographic pattern is vicar! ant,though, as Bush (1975) notes, there is ample opportunity for broadneosympatry, inasmuch as parent- and daughter-species are al-leged to have somewhat different sets of adaptations, i.e., they oc-cupy somewhat disparate niches.

As illustrated in figure 4.4., there are two fundamental forms ofallopatric speciation on the one hand, and parapatric and sympatricspeciation on the other hand. These latter two modes involve (a) aninitial distribution, in which the daughter-species is directly adjacent


cestral species as a contemporary of its descendant). Therein liesthe only formal distinction, in terms of pattern only, between classicsaltation ism and speciation in the strict sense: in alfopatr!c (type b),parapatric, and sympatric speciation, the expected pattern is syn-chronous overlap of ancestor and descendant. whereas the oppositeis expected from classic saltation. In situations involving fossils, sal-tation can be rejected only by demonstrating temporal overlap be-tween ancestor and descendant. Failure to demonstrate such con-temporaneity is obviously negative evidence and thus not a strongtest of successional or "classic" saltation. But, aside from this con-sideration, no amount of data pertaining to fossils is likely to allow asystematist to distinguish unequivocally among the possibilities ofsaltation, on the one hand, or speciation (whether type b allopatr!c.parapatric or sympatric), on the other. And there is the even morefundamental problem of the formulation and testing of hypotheses ofphylogenetic ancestry and descent-phylogenetic trees-the sub-ject to which we now turn.


to (parapatric) or wholly within (sympatric) the distribution of theparental species. Thus (b) selection for reproductive isolation is hy-pothesized to occur directly; i.e., reproductive isolation does not de-velop merely as an accidental by-product of divergence. The prob-lem with such models has always been the difficulty of concocting aplausible explanation of how reproductive isolation can developwithout the aid of extrinsic (specifically, geographic or physical en-vironmental) barriers (see Mayr 1963 for an extended discussion).Bush (1975) has reviewed the possibilities and presented evidencethat such .processes can occur. Of direct interest to a systematist,how.ever, IS.th.e fa.ct that the main effect of both parapatric and sym-patnc specla.tlon IS identical to that of allopatr!c speciation involvingperipherally Isolated populations (i e., type b allopatric speciation)'the end result is a new species diagnosably different from the un-modifie~ ances.tr~1 species. Thus type b allopatric, parapatrtc. andsyrnpatr!c speciation all are predicted to result in the development ofthe patter.n of fi~ure 4.5b. Moreover, all three are held to producenew specres rapidly, Thus. from the point of view of systematics theonly distinction to be drawn among these three models is the natureof the initial distributions of the ancestral and descendant species atthe o.n~~t of reproductive isolation. If sympatric speciation is a realpossibility (as Bush argues), then demonstration of sympatry be-t~een two closely related species does not automatically imply atime I.ap~e s~fficient to allow a change from an allopatr!c to a sympa-tr~c distribution. These considerations have implications primarily forhi t . I bis once roqeoqraphy. In terms of representing phylogeneticevents on branching diagrams, there are no meaningful distinctionsamong these latter three modes of species formation by splitting.in C.arson (.1975) ha~ recently discussed rapid speciation events

volvinq peripherally Isolated populations, stressing the possibilitythat such events can be so rapid as to appear "sattational" even inec I . I tihO oglC~ nne .. ~hen. compared with the empirically established~ enotypic st~bliity eVld.enced by many minimally diagnosable taxa(,.e.. hypothesized species) over truly long expanses of geologicalt~me, such rapid speciation events might indeed appear to be salta-ttonal. In most classic cases where saltation has been hypothesizedIe wh tt . '". ere en Ire species were judged to be transmogrified into de-scendant species by sudd . . ..svnchronat en Jumps, by definition there can be noynchronalty of ancestor and descendant (r.e.. no survival of the an-

Construction and Testing of Phylogenetic Trees

In this section we present a general strategy for the production andevaluation of phylogenetic trees. We shall develop the basic themethat trees have a relationship with cladograms such that, for anygiven problem, there are a number of possible trees consistent withanyone cladogram (see also chapters 2 and 5 for further discussionof the concept that cladograms are sets of trees). The central prob-lem of tree construction is to reject all but the least unlikely tree con-sistent with any given cladogram. As more specific statements. treesare in some ways more easily tested (i.e., rejected) than cladograms.Construction of trees, however, entails an additional set of assump-tions nol needed for cladogram analysis. Incorporating more infor-mation (e.g., on extrinsic properties of species), a tree is also furtherremoved from the "data base" of pure character distributions amongorganisms than the background cladogram on which it is based.Thus trees probably can never be as highly corroborated as ctado-grams.


As diagrams purporting to show actual evolutionary events,trees portray hypothesized sequences of ancestry and descentGiven the corollary of any general notion of evolution, i.e., that suchpatterns of ancestry and descent must exist, the systematist canhardly be faulted for seeking them. And, as we shall see in chapter6, highly corroborated trees are absolutely essential to the testing ofcompeting theories of speciation. And it would be well to reempha-size at this juncture that only trees involving species as ancestorsand descendants have any meaning beyond the cladogram level ofanalysis. There is no formal difference-just semantic confusion andthe retention of non-monophyletic groups-between cladogramsand trees involving taxa of rank higher than species.


a, b.A oB C oA

CB

xd,c,

IA 10

B(/ ('" ',,? ,/

XFigure 4.6 Comparison of cladograms and trees. (a) Acladogram. Species are represented by dots; lines con-nect dots to indicate nested patterns of synapomorphy. (b)A phylogenetic tree, one of several consistent Withthe cia-dog ram of a. Notational symbols are exactly the same asin the cledoqram: species are represented by dots, andthe lines, in addition to indicating nested sets of synapo-rnorphfes. connect the species in ancestor-descendantfashion. The hypothesized ancestral species ISrepre-sented by X. (c) A more conventionally rendered phyloge-netic tree. In this diagram, species A, S, C, and 0 are de-picted by solid lines indicating known temporal ,distributions. Dashed lines depict hypothesized lines ofdescent. The only "dots" used are for X, ~h~unknown, hy-pothesized ancestral species. (d) The original cladogramredrawn using the notational conventions of the phyloge-netic tree of c.

The Topology of Trees

There are two different ways of drawing phylogenetic trees, Clarifica-tion of these topologies is essential to avoid ambiguity. In figure 4.6we compare, purely from the standpoint of conventional notation, thevarious uses of "dots" and "lines" in cladograms and trees. In figure4.6a, a typical cladogram is shown. Taxa are represented by dots;lines merely connect taxa to depict the hypothesized nested sets ofsynapomorphies. In figure 4,6b, we show a phylogenetic tree whichis drawn with the same symbolic conventions as in the preced ingcladogram: taxa (species) are dots, and lines depict a pattern of rela-tionships, in this case one of ancestry and descent. In figure 4.6, weshow the same tree, redrawn this time to show the known duration intime of each of the species. In this instance, solid lines stand for theentire known range of the species; hypothesized patterns of rela-tionship are shown as dashed lines. The only "dot" is the unknown,hypothetical species X, the inferred common ancestor.

For ease of comparison, it is convenient to utilize the same nota-tion in both the cladogram and its derivative trees. However, the no-tation of figure 4.6c, where taxa are represented by lines rather thanby dots, is the style adopted in the overwhelming majority of treespublished in the literature. Conceptually, it makes little difference,inasmuch as either system agrees with the fundamental notion ofspecies as discrete individual entities in nature. For the remainder ofthis chapter, we shall adopt conventional symbolization, not onlybecause it agrees with the bulk of past usage, but also because it

, series of ancestorsallows depiction of overlapping ranges among a, t Ily and conceptuallyand descendants. Such overlap IS both prac rca . .

, , I' 0' speciation predictImportant. Pragmatically, different concep Ionsdi tnb ticn among ancestorsvarious different patterns of temporal IS fl U I

, nges in a tree thereforeand descendants' incorporation of known ra, , I species as an ancestoraids in evaluating the status of any particu ar


vs. a collateral descendant, and contributes as well to the evaluationof the relative merits of competing speciation models in any giveninstance. And conceptually, inclusion of time-range information on abranching diagram, be it a tree or a cladogram, is directly tied to no-tions of differential species survival, discussed in detail in chapter 6.As an example, in figure 4.6d, we redraw the cladogram of figure4.6a using the conventions of the tree of figure 4.6c.

Construction of Trees: Additional Assumptions in the System

Construction of both cladograms and trees assumes a pattern of an-cestry and descent connecting the species being analyzed. Jt fol-lows as a natural goal of systematics to reconstruct that pattern in asf~ne a d~tail ~s possible~hence the zeal with which early evolu-tionary biologists converted the Scala Naturae into trees purportingto show actual evolutionary lines of descent. However, constructionof trees, with ancestors and descendants specified among a seriesof spec~es at the outset requires two additional assumptions notmad: With cladograms. The first assumption is that the species actu-ally Involved in connecting speciation events are at least potentiallypresent In the sample, Another way to put it is that some of thespecies present in the sample may be the direct ancestors of otherspecies under consideration. Cladograms, of course, require nosuch assumption. Cladograms depict nested sets of synapomor-phies,' In terms of relatedness of the taxa, a cladogram merely orderstaxa In term,s of nearness of common ancestry. Cladograms requireno assumptions about the presence of ancestors in the sample (fig-ure 4.7), Inasmuch as it is impossible to assume with certainty thatdirect, ancestors are Included in the sample, it appears that the con-structl?n of phylogenetic trees has an intrinsic flaw, requiring an as-sumption perhaps many systematists would not be willing to make,This accounts .for the predilection for drawing trees among taxa ofqenenc and higher rank (otherwise a logical absurdity, as alreadynoted), which loosens the restrictions of this assumption.

Non~theless, in certain circumstances, the assumption that ac-tual spectatton events are at least potentially included within a sam-ple o.f s.pecies may appear reasonable. According to the majority ofspecrat.on models reviewed earlier in this chapter, ancestral speciesfrequent!y survive as contemporaries of their descendants, and areapt to display characteristic geographical distribution patterns with


o. b.

Species 1 Species 2 Species 3Species 2 Species 3

Species 1

F1gure4.7 Diagram illustrating the more general nature of c1adograms andphylogenetic trees from the point of view of specification of number of spe-ciation events represented in the system. (a) A cladogram in which species1 is the sister-group of species 2 + 3, Any number of speciation events be-tween species 1 and species 2 + 3 may in fact have occurred and areallowed by the c1adogram. (b) One phylogenetic tree derivable from thecladogram. This tree depicts the hypothesis that species 1 gave rise to anunknown descendant species ("X"), which divided to produce species 2 and3, Thus the tree specifies exactly two actual speciation events. Pattern a (thecladogram) is thus a more general statement than pattern b (the phylo-genetic tree).

respect to one another. Thus, in the Recent biota, in a region wheretwo species are found, if these two species have been demonstratedto be more closely related to each other than either is to any otherknown taxon, the possibility exists that one of the species is ances-tral to the other, or that both evolved from a single, immediate com-mon ancestor, For example, in a case discussed in chapter 3, Plat-nick and Shadab (1976) described the spider species Zimiromuspenai and Z. brachet from Ecuador, alleging their closer relationshipto each other than to any other known species. The possibility is im-mediately raised that a speciation event connects the two species.Further, in the case of the four nominal subspecies of Phacops ranadiscussed by Eldredge (1972; see chapter 3), the pattern of synapo-morphies (see figure 4.8), the absence of close relatives within thedepositional basin, the relative continuity of presence of the mono-phyletic unit (i.e., the whole group) over a period of some 8-10million years, and the pattern of occurrence (with respect to eachother) of the constituent taxa in space and time raise the distinctpossibility that they are linked by direct patterns of ancestry and de-Scent. Indeed, Eldredge (1971) and Eldredge and Gould (1972) as-sumed such a tree to exist (see figure 4.9) to serve heuristically as ameans of discussing patterns of speciation in geological time (see


(18) (18)

(18)Figure 4.8 Acladogram of relationships amongfour subspecies of Phacops rana. Numbers in pa-rentheses refer to the number of vertical columnsof lenses in the eye typical of each SUbspecies orthat hypothesized as characteristic of the com-mon ancestor. Other patterns of synapomorphysupport the cladogram. 18: Phacops ranacrassituberculata and P. rana milleri; 17: P_ r.rene; 15: P. r. norwoodensis.

below for further discussion of this example). All theories pertainingto speciation mechanisms hinge on field observations of populationsand species which. at least implicitly, have been analyzed in pre-cisely these terms, i.e., as ancestors and descendants. In spite of theadditional assumptions required before phylogenetic trees can beconstructed, phylogenetic trees appear to be crucial for further re-finement of theories of the nature of the evolutionary process, specifi-cally that part of theory devoted to the origin of species (see chapter6; Wiley 1979). Thus these assumptions. in the long run, appear tobe worthwhile, unreasonable as they might appear in any particularcase.

The second assumption. that character reversal does not occurin phylogenesis. is vital for the evaluation of conflicting trees. Let usassume that a comparative analysis of distributions of synapomor-pnies has been performed. Emerging from that analysis are severalputative species, plesiomorphic or equally synapomorphic in all re-spects to their hypothesized sister species. Such species are good

EPEIRIC SEA MARGINAL SEA

+ TTAGHANIC

-----J--J+STAGE

---------- .......+

TIQUGHNIOGASTAGE

------ 1-------+ -r-+

III

CAZENOVIA 1STAGE I

I1I

-J:.....,

16-11ia ie I ", "rs• . th ized pattern of ancestryFigure 4.9 Phylogenetic tree showing the hypo eS~~ivedfrom the claoo-

and descent among subspecies of Phacops r.ana. d rinsic and intrinsicgram of figure 4.8 and consistent with all available e1 ure 4 8. Dotted linesdata. Numbers at the base of the diagram are as In ~g of c~lumns ofrepresent origin of a taxon with ~ n~w {red~ced} numhe~ lines representlenses by type b auopatr!c speciation; .horlzonta~daSe~raltaxon in a portionmigration; vertical lines represent persl~tence °d anCendanttaxon occurs.of the marginal sea other than that In which the ~c 197t figure 5. andCrosses denote final disappearance. (From Eldre ge ,Eldredge and Gould 1972, figure 5-8,)

--~~-~-~-----------------~

106 yr r"';-;-.-"'o;;:;;;;;;;-----,---------.=;;-w-.;;::=,--0H.IiI.$. \., Neands H.$.s. I I~eand

H·1 Ii J.I 1 I I....

I IH.h. ~ A.r. A.b. A.b. A.r. ., H.h.

- 21-----JI-:--I--I--____j____j~-I--_I___1____j----j


candidates for being ancestors to their more derived sister species.Though there are, as we shall discuss, some relatively weak tests ofsuch hypotheses available based on extrinsic data, it is desirable toevaluate the hypothesis on the basis of intrinsic properties.

If a taxon is plesiomorphic (or at least synapomorphic) in all(analyzed) respects when compared with a series of closely relatedtaxa (as determined by the presence of at least one synapomorphycommon to them all), there is no way to reject any specific hypoth-esis of relationship of that species beyond specifying its status asthe plesiomorphic sister-species of the remaining taxa in the series.For example, if species 1 of figure 4.7a is plesiomorphic in most re-spects when compared with species 2 and 3, there is the possibilitythat species 1 is the ancestor of the other two species (as in figure4.7b). But there is no formal means of rejecting this hypothesis, atleast with sole reference to the evaluation of intrinsic properties (seeEngelmann and Wiley 1977; Platnick 1977a). Eldredge and Tattersall(1975) claimed that, at least in terms of cranial morphology, the earlyPleistocene hominid Australopithecus africanus is so completely ple-siomorphic that it is difficult to assess its relationships among l-lo-minidae. There is certainly no way of formally rejecting the hypoth-esis that A. africanus is the ancestral species from which allsubsequent species of Hominidae evolved (figure 4.10).

• I •• A.8. •• I I• •0 3 • I;;• • "u • I0 • •~ • • ;•, • ;

• ,• ;,";•"• I

I• I0 I•u

Ie 8,I• Is I

10

• I •" I A.8. :II I ••, I • •

'I • •• •, " •• I, •," I-,: /

" ;

H.s.s. N. He.", ', ,, ,, ,

R.p. R.p.H.h. A.b. A.r. A.a. R.p. • 120•u

0,.e 14-e

~

_16 A B

;IIII,'1;

b.Figure 4.10 (b) Two phylogenetic trees consistent with the cladoqram. Rp.,Ramapithecus punjabicus; Aa. Australopithecus africanus; A.r., .A.ustralo-pithecus robustus; A.b., Austra/opithecus boisei; H.h., Homo rebitis; n.e.,Homo erectus; N, neandertaloids; H.s.s., Homo sapiens sapiens. Note that,according to the cladogram, the genus Australopithecus is non~onophy~lenc, and alternative views of the relationships of the naandertaloids are ex-pressed. (From Tattersall and Eldredge 1977, figure 7.)

o.Figure 4.10 (a) A claocqram of relationships among species of Hominidae~ased solely on cranial characteristics. (After Eldredge and Tatlersal11975,figure 4, and Tattersall and Eldredge 1977. figure 5.)

---.---------------------------------- - -


ancestry and descent are to have- any degree of testability what-ever.


Thus there is a profound, if formal, difficulty in the testing of hy-potheses of ancestry and descent. This difficulty stems from the log j-cal necessity of postulating that an ancestor must be plesiomorphicor equally synapomorphic with respect to its descendant-whichmust. by definition, be true, If a species fits these requirements, thehypothesis of ancestry is formulated. Testing of the hypothesiscomes when more characters are sought. For example, demon-stration that a species hypothesized as being an ancestor of anotherspecies in fact possesses an autapomorphy rejects the hypothesis ofits ancestral status without rejecting the underlying cladogram. Ofcourse, demonstration that a hypothesized ancestor shares apo-rnorphies with a third taxon (I.e. some taxon other than the putativedescendant), results in the rejection of the entire cladogram and notjust the tree. Thus there is an element of testability to hypotheses ofancestry and descent among species, but this element comes at theexpense of our second assumption, that character reversal does notoccur in phylogenesis (see also the discussion of phylogenetic neo-teny, chapter 2). If autapomorphies are sufficient to reject the hypoth-esis of ancestry, there is an implicit assumption that a particular au-tapomorphy cannot revert to its "primitive" condition. This is nothingmore than the criterion of parsimony and, viewed in this light, per-haps not an overly steep price to pay for testability of phylogenetictrees. However, paleontologists in particular are fond of tracing char-acter changes through vertical sequences of rocks. If a verticalsequence "documents" a character reversal, it is taken at face value.The second assumption, the invocation of parsimony in the form of amethodological rule ("character reversal does not occur"), becomesunacceptable when viewed in this context, e.g., the "stratophenetic"approach of Gingerich (1976, 1979. See figure 3 in Gingerich, 1979,for an example of character reversal within an hypothesized continu-ous lineage through a sequence of layered rocks). Even though ple-stornorpny and apomorphy are qualities themselves hypothesized bythe systematist, it is nonetheless parsimonious to assume that an ap-parently more apomorphous state will not, in fact. become furthermodified into an apparently more plesiomorphous state. Paleonto-logists have for years been rejecting trees on the basis that a certaintaxon is "too specialized" to have given rise to another. The assump-tion of Irreversibility is methodologically essential if hypotheses of

Cladograms to Trees: Formulation and Testing

Figure 4.11 shows the simplest form of phylogenetic tree and itsmost general relation to a cladoqrarn. The case is trivial, but none-theless epitomizes the most fundamental, formal distinction betweena cladogram and a tree: the presence of taxa specified at nodes(branching points) on a branching diagram. Cladograms do not ,havetaxa at branching points, trees do. This most general tree (figure4,11b) might be termed an Ai tree ("A" for ancestor). Species 1 and 2are implicitly linked on the cladogram by a pattern of synapom~rphyderived from some common ancestor of unknown remoteness(figure4.11a). In figure 4.11b, the Al tree, the ancestor is merely ~ade ex~plicit. There are three possible values for i: i= 1 (species 1 IS anc~s-

·s1)'orl~tral to species 2); i= 2 (species 2 is ancestral to specie. '1,2 (neither species is ancestral to the other), If i= 1 or 1=2, there

Q. b. c.

\ /Species 1 Species 2

\(Species 1

\A;

f. g." /s,e'y'" 2

Species X

I

d. eI \ I I7'"2 8"''''\}'''' 2 7"1

Species 1 Species 2 species 2

/ / /. .' eena (a)two-taxonctadcqram

Figure 4.11 Diagramtuustratmqthe relattonsfupbe,twt ) Alsoshownarelreescorte-and(b) an A; tree (thesimplestformof a p~Ylogene~l)ci~~.' {gl i :..1,2,i.e,,' '" somespondingtothefollowingvaluesofi:(c,d)/-:= 1; (e. -,othertaxon (X).


are two possibilities in each case (figure 4.11c, d. e, f). If i~1,2 (Hq-ure 4.11 g)there is a total of five potential forms of an Ai tree in anycase involving two known species. The problem of converting a cla-dog ram into a tree, then, resolves down to a matter of specifying thevalue of i.3

The process of identifying a value of i consists of consideringeach possible specific form of a hypothesis of ancestry and de-scent at each branching point (node) between sister-species on acladogram. The simplest case involves a dichotomous node, whereonly two species are involved at anyone branch point. Making thefirst assumption, that it is possible that the two species are directlyrelated by a speciation event. the first hypothesis to be considered isthat the relatively more plesiomorphous of the pair of sister-speciesis ancestral to the other. In other words, on the basis of the secondassumption, the hypothesis that the relatively more apomorphoussister-species is the ancestor of the other is automatically tested andrejected.

In testing the hypothesis that the relatively more plesiomorphousspecies is ancestral to the other species, there are two possible out-comes: the hypothesis is rejected (in the general manner suggestedabove, i.e.. by discovery of one or more autapomorphies or by rejec-tion of the underlying cladogram itself) or the hypothesis is provi-sionally accepted. Should the plesiomorphous sister-species turnout to be relatively plesiomorphous or equally synapomorphous withrespect to all available intrinsic data, the hypothesis that it repre-sents the ancestor of the second species might further be tested byconsideration of configurational (extrinsic) data: distributions of thetwo species in space and, if data are available, in time.

The simplest case occurs when the two species are I iving con-temporaneously and are known from only a single point in time. thebest example being most elements of the Recent biota. Two speciesknown as fossils from only a very short interval of time conformequally well. In such circumstances, type b allopatric. parapatric,and syrnpatnc speciation are the three major models of speciation

3. In this example. the distinction between notational conventions in trees is clearly seen.When species are depicted as lines, there are indeed two forms of trees for each case (i.e., fori = I and i = 2), a total of four possibilities (i ~ 1,2 gives the fifth possibility), conforming toconflicting theories of speciation reviewed earlier in the chapter. If, however, species arerepresented as dots, there are only three trees possible, one each for j = I, i = 2, and i ..1.2.


which are theorized to produce a new species living alongside (as acontemporary of) its ancestor. Referring to figure 4.4, distributionaldata may be suggestive of the actual mode of speciation, and serveas a relatively weak test of the hypothesis of ancestry. All three mod-els of speciation predict that. initially, i.e.. during and right after thespeciation "event" (which is hypothesized to be rapid but which mayin fact take as long as several thousand years), the size of the de-scendant species (measured in numbers of individuals) will be smallwith respect to that of the ancestral species. Thus should the apo-rnorphous sister species, hypothesized to be the descendant, have arestricted distribution with respect to its putative ancestor, andshould the two species display disjunct distributions, the data areconsistent with an interpretation of type b allopatric speciation, andthe hypothesis of ancestry and descent may be considered to havebeen corroborated somewhat further. However, size and distributionof descendant species relative to their ancestors emerge as pre~ic-tions from theory only in the early history of the descendant species.Distributions of both descendants and ancestors mayor may notchange subsequent to speciation. No element of speciation theorycan predict subsequent patterns of distribution in detail (see chapter6 for further discussion of the prediction of distributional attributes ofspecies). An initially allopatnc descendant may come to ran~e aswidely as, and even become fully sympatric with, its ancestor (figure4.4) Moreover, there are no criteria whereby relative recency of spe-ciation-which would allow more confident predictions about thesignificance of distributional data--can be assessed .. Spatial dis-tributional data for two contemporary species offer no firm means oftesting hypotheses of ancestry and descent. .

The only direct prediction that arises from the nonon of ancestryand descent is the rather obvious one that the ancestor must havebeen older than its descendant There must have been som~ seg-ment of time during which the ancestor was in existence, and Its de-scendant was not. Fossils offer the only direct evidenc: of the exts-tence of taxa in pre-Recent times. Shaw (1964) has pointed o.ut that

fossil . must be consideredthe actual temporal range of any OSSI species .'unknowable. The observed stratigraphic range, especlatly In ~nylocal area, cannot be assumed to be the total life-s~an of a ~peclesfWhat is sampled is actually a portion, somewhere In t~e middle, ~the total life-span. Similarly, total geographic distributions of fossil


taxa are virtually unknowable. In many instances, sediments werenot even deposited in all areas where a species might have been Iiv-ing. This is especially true for terrestrial organisms, but pertains toaquatic organisms as well And where deposition did occur, there isno guarantee that individuals of a given species are fossilized evenif they had been living there. Moreover, erosion has destroyed atleast portions of many sedimentary rock units, particularly aroundthe margins of sedimentary basins (where speciation might be pre-dicted to occur most frequently according to the model of type ballopatric speciation). Or sediments may remain buried under milesof younger rock, wholly inaccessible to the systematist. Subduction,metamorphism, and diagenesis (geochemical alteration of sedi-ments and fossils) are additional factors that limit the credibility oftemporal and spatial distributional data of fossil species, In an hy-pothesis as specific and detailed as those postulating derivation ofone species from another, the fossil record can rarely if ever betaken literally,

Nevertheless, if a plesiomorphic sister species is always foundin younger (independently dated) rocks than its hypothesized de-scendant, then, even though the hypothesis is not decisively rejec-ted, there would seem to be little reason to retain the hypothesis thatthe younger species is ancestral to the older one. Such data, how-ever, have no bearing on the cladogram: the younger but more ple-siomorphous species remains the plesiomorphous sister of the older,more apomorphous species.

The phylogenetic tree (figure 4.9) for the Phacops {ana ranaspecies-group depicts two putative instances of type b allopatricspeciation For example, P. {ana {ana (number 17 in figures 4.8 and4,9) is derived with respect 10P. {ana ctessitubercutete (number 18on the diagrams) P, r. crassituberculata is known from slightly olderrocks in eastern North America. P. {ana {ana is first known to occur inslightly younger rocks representing near-shore environments in cen-tral New York State. In one quarry, a few specimens have been col-lected which are morphologically intermediate between the two taxa.Higher in the same quarry, undoubted specimens of P. rana {anaoccur alone. In still younger sediments, P. {ana {ana is foundthroughout an approximately 2 million-year interval of time, in thenear-shore environments preserved in the present-day AppalachianMountains. During that same temporal interval, P. r. crass-


itutsercutete became restricted to the more offshore waters ofwhat is now the American Midwest. Geological evidence suggeststhat the seaway disappeared in the Midwest at the end of this inter-val. When the sea reappeared, P. rana {ana was found to be ubiquit-ous and P. r. crsssitubercutete was no longer found. Thus distribu-tional histories of fossil taxa can be sequentially mapped (figure4.12).

Eldredge (1971) interpreted the sequence as an instance of typeb allopatric speciation, where P. r. crassituberculata gave rise to P.rana {ana as a peripheral isolate. The two taxa remained allopatricwhile P. {ana {ana expanded its range (along the present-day Ap-palachians), The distributional data of this example are consistentwith this interpretation. Further tests are possible by examining addi-tional data: expansion of range data (both temporal and spatial)would serve to reject the specific form of the hypothesis. Were P.rana {ana to be shown to occur as early as the earliest known oc-currence of P. r. creeettubercutsie. or should its inferred initial geog-raphic range be shown to have been broader, interpretation of thedata as a case of type b atopatnc speciation would be falsif.ied. .

In short, distributional data can be interpreted as consistent witha specific hypothesis of ancestry and descent, and one of the sev-eral available models of the speciation process will usually appearto fit the data rather well But lack of strict predictability of necessarypostspeciational patterns of distribution, coupled with necessarilyfaulty distributional data, severely restricts the use of extrinsic dataas a critical test of hypotheses of ancestry and descent-phylogene-tic trees. There are no strict criteria available to allow judgment as tohow much anomalous distributional data can be "allowed" before ahypothesis of ancestry and descent is disallowed. Species with dis-junct (whether spatial or temporal) distributions may nonethelesshave a direct ancestral-descendant relationship, and there ISno wayof specifying the limits in distributional data beyond which such hy-potheses are to be rejected.

If the hypothesis that either of two species is ancestral to eachother is rejected, a fifth possibility remains to be consld~red. :ype aallopatric speciation produces two descendant species Simulta-neously from a single ancestor. Thus, in a situation involving ~otaxa, the fifth possibility, that both species were derived from an Im-mediate common ancestor, results directly in a tree where an ances-

}s }s Z<t>

IS 0z'"N<t<J

}s Is 18,17

117

17 17


tor is postulated to have existed, but is unknown. Such a hypothesisfurther states that, whatever that ancestral species was, it could nothave been either species 1 or species 2. Such a tree is a real tree,not a cladogram, simply because it does postulate a direct, albeitunknown, ancestor.

A three-taxon statement (a cladogram) may be investigated as asource of the identity of the unknown ancestor. There are severalpossibilities. In the cladogram of figure 4.13a, the possibility is con-sidered that species 3 is ancestral to both species 1 and 2. But, ac-cording to the cladogram, species 1 and 2 share a synapomorphynot present in species 3. Thus their common ancestor must havepossessed that synapomorphy, and species 3 is thereby automati-cally rejected as their immediate common ancestor, The only kind ofcladogram involving more than two species that could resolve into aphylogenetic tree representing type a allopatric speciation is that infigure 4.13b, Species 3 is discovered or considered after species 1and 2 are depicted as sharing an unknown common ancestor. Thereare no synapomorphies linking any two species as a subset of thethree, Trichotomous cladograms may be considered unresolveddichotomous cladograms (i.e.. as if the cladogram of figure 4.13bmay ultimately be shown to have a structure like that of the clade-gram of figure 4.13a). This position is especially pertinent to theoriesof relationship among three monophyletic taxa of generic or higherrank. Insofar as species are concerned, there is a further possibility:a trichotomous cladogram might represent the closest approximationto the actual evolutionary relationships among the three species. Theexpected outcome of type a allopatric speciation is precisely thisform of cladogram. All that is required is that the three species beseparately diagnosable, i.e. that species 1 and 2 (the putative de-scendants) each have at least one autapomorphy. In such a situa-tion, the hypothesis is tested that the newly discovered or con aid-

Ze;

'"oZI

'"=>o>-

17 17

Figure 4.12 Five sequential maps showing distribution of seaways in east-ern North America in the Middle Devonian Period. Numbers on the map referto five representative localities. Vertical lines show the stratigraphic occur-rence of different subspecies of Phacops rana, represented by numbers ofcolumns of lenses in the eye (as in figures 4.8 and 4.9), Blank spaces in-dicate intervals when a sea was absent from a locality, Major subdivisions ofthe Middle Devonian of eastern North America are shown; a time span of ap-proximately 10 million years is represented. (From Eldredge and Eldredge1972:55.)


Species3 Species2 Species1 '" /S"""y""'" f

Species3

Ia

Species3 Species2 Species1 "" /S'OC'"y''' 1

Species3

Ib. =

Figure 4.13 Diagram illustrating relationships between cladograms andphylogenetic trees lor cases involving three species. (a) Cladogram inwhich species 1 and 2 are the sister-group of species 3. Such a diagramimplies possession of at least one synapomorphy held by species 1 and 2,but not species 3, Since the synapomorphy must have occurred in the com-mon ancestor, it is impossible to construct a phylogenetic tree with species3 directly ancestral to species 1 and 2. (b) A trichotomous cladogramamong three species. Such a situation could represent type a allopatricspeciation, in which a single species (e.q., species 3) splits into descendantspecies 1 and 2. All that is required is that the three species be separatelydiagnosable and that one be plesiomorphic or equally synapomorphic withrespect to each of the other two species, Species 1 and 2 do not uniquelyshare any synapomorphies in this instance.

ered species 3 is the ancestor of both species 1 and 2, Theprocedure is exactly the same as that outlined above for the hypoth-esis that i= 1 or i= 2, Given this possibility, further speciation couldproduce taxa of higher rank ultimately linked as a genuine polytomy.Thus not all potytomous cladograms need to be mere expressions ofIgnorance.

Actual instances of simultaneous polytomous splitting, in whichtwo or more species are differentiated (by type b allopatric. parapa-tnc, or sympatric speciation) at the same time, have the result that notwo descendant species will share a synapomorphy; thus, again, thec!adogram will perforce be polytomous, and the situation will again

,'~~•• - •••• -... : L,.P2S 0l0) I x........ .....' ..... 1I,IUlui,o,o ~~. •

,o' " I. '~ .,o" \ ~ pzs (xV) I', r

,l ! ~~ 11111111 1P(Sb(XY) C".

"~I ~ -,",~ Ilbul' "I: : '\4" : PZ4(XO) \

" ':"~..,:\ ,.... .... IlillllOI :"" .,P(Sc. \"" .:A • ',o '.

O , ,', ',._.,o' "~"'" .' ,"'"\1"0""" , - r.... -. ,.. ,o' • " '.'...... \ 1IIIIh I' \ /• , J.....~ " ,. <,. '"~ .. : ViaticafJ ' ',. '.,

\''''''''''', (~ 1i;:I~::~:::)'"...~_ __~"."·X........................ 1111111I.' Vlaticl,.1 100km

, h Australia, showing distribution ofFigure 4.14 Map of a portion of soul re~ine rasshcopers. Species arecertain taxa of the viatica group of (0 a P4S2). haploid karyotypes are alsoidentified by chromosome numb~r e,1,'\Othe a~cestral "viatica" type, (Fromshown, P45c appears to be t.heC,Ogs:b the American Association for theWhite 1968, figure 1. Copyright YAdvancement of Science.)


struction of the history of life, then specification of ancestral-descen-dant relationships among the phylogenetic units of evolution-species-remains at least ideally desirable. Trees express interpre-tations of actual patterns of descent and are thus vital to the furtheranalysis of the nature of the evolutionary process. It is at this levelthat specific theories of the nature of the evolutionary process have arelationship with systematics: phylogenetic trees must be consistentwith both cladograms and concepts of evolution. Cladograms, on theother hand, merely require the hypothesis that life, somehow, hasevolved.

ff

-

Chapter

5Biological Classification

IN its most elemental form, classification is the grouping of ob-jects into a larger more inclusive set; alternatively, classification canbe viewed as the subdivision of a larger set into smaller subsets. Inclassifying there is always a reason for grouping some objectstogether white at the same time excluding others. and that reasonmust be based, ultimately, on similarity. Words, or, perhaps morestrictly, the meanings of words, are themselves classifications. Theprocess of classifying is thus a very ancient activity within humanhistory: "From earliest times the human mind sought out the elementsof order in the world, and the first step in this direction consisted inthe noting of similarities between things. Such noting of similaritiesbetween things constitutes an implicit, if not an explicit, classifica-tion of them" (Wolf 1930:139).

The Greek philosophers provided the foundation of ideas fromwhich subsequent views of classification are derived. The philoso-phy of essentialism, originating with Plato and Aristotle, held that acollection of objects can be defined as a set (called by them a"species") when each member of the set shared the same essence:"In Aristotle's view three things can be known about any entity-itsessence, its definition, and its name. The name names the essence.The definition gives a complete and exhaustive description of the es-sence" (Hull 1965:6). The definition of the essence, then, was a list ofjointly held properties-similarities-that justified applying a nameto the set of objects. Essences were considered real entities in na-ture, not artifacts of human thought; consequently, groups character-ized by such essences were natural. The problem of "natural clas-sification," Le., the discovery of "natural" groups, has captured theinterest of biologists since Aristotle.

Carl von Linne constructed his now famous system of botanical

148 Biological Classification

classification. from the standpoint of Aristotelian philosophy (see Lar-son 1968). HIs "essences" were extracted from the varieties in formof the reproductive system of plants. Soon thereafter the FrenchZ~IO.9i,~tG~orge~Cuvlertooka similar approach uSin~"functionalcriteria to Identify the most important characters (essences) thatcould ..se~e to delineate groups (Coleman 1964). Belief in naturalclasSlflcatlo~ was widespread throughout seventeenth-, eighteenth-,~nd early nineteenth-century biology, and most biologists of thesetimes followed a simple rationale: natural groups are those createdby God and it Is the task of the pious scientist to discover and makethem manifest 1

The ~bservation that natural historians have always been con-cerned with creating "natural" classifications is important. for itsuggests a commonality in purpose that transcends the intellectualdogma hegemonic at any particular period of time. In a real sense,the problem of classff ,.. . .'. rca ron, In particular, IS the problem of compar-ative biolopy in gene I· thra. e search for natural groups. Indeed, it hasbeen,. and always will remain, one of the fundamentai questions andpur~UI~s of biology, for the simple reason that it reflects mankind'scurros~ty a.bout reconstructing the history of life.

Hlstorrcally, the problem of classification-the search for naturalgroups~preceded evolutionary theory and thus may appear to belargely Independent of it. This, of course, is not the conclusion mostcontemporary systematists might draw were the question put tothem. But consider The h f. . . searc or natural groups is a search for pat-tern. Darwinian evofutio th .nary eory IS a theory about process, notpattern analysis, and the question of mechanism can be shown to~ave a tenuous link to the methods used to discover natural groups,I.:.,.to reconstruct the pattern of life's history (see chapter 2 on cla-distic analysis), This is not to assert, of course that evolution has notproduced the pattern-natural groups. But ha~ it been demonstratedthat the imrooucnon of evoiutionary thought altered the method ofsystematics to any great extent? Louis Agassiz for example con-structeda" I'''' ". . genea oqical diaqram (Figure 5.1) 15 years prior to TheOf/gm of Species a d!, raqram that does not differ materially in its

I. During the eighteenth ceetu F h .sentialist criterion f ry: ~nc botanists sought to discover "natural order" by an es-characteristics" w 0 gmu: slm~lllllty: those plants most like one anOlher in their "essential"stem of L·'""~Q,~.recom 1Il~~nto "natural" groups (Burkhardt 1977:49). The classificatory..,,,,,, was consluo::red"anifi '81" . .the sexual system alene. Cl In thltl It was based upon the characterics of

Biological Classification 149

content from those drawn after the introduction of evolution. Only theinterpretation was to chanqe.s

By the beginning of the nineteenth century, the search for natu-ral groups had become more intense: "The belief in natural clas-sification rose to dominance with remarkable swiftness and lack ofdebate. Perhaps the accumulated experience of taxonomists hadbuilt up until they knew that they must be on the right track. Theysensed that their efforts were revealing nature herself, and by'1800they felt this with new confidence" (Winsor 1976:3).

That nature might be hierarchical in arrangement, rather than alinear chain, began to be considered by nineteenth-century bio-logists, The famous Swiss botanist A P. de Cando lie, for example,stated in 1813 that the pattern of nature is characterized by groupswithin groups (Winsor 1976). This belief gained further acceptanceprior to Darwin, and while most biologists of the period relied onScripture to provide the ultimate causal explanation of these "pat-terns in nature," the actual recognition of natural groups was ex-pressed through classification. Ultimately, it was left to Darwin togive a new meaning to the concept of natural groups.

Darwin, Natural Groups, and Classification

In chapter 13 at The Origin of Species, Charles Darwin presented adiscussion of the principles of classification and articulated a cost-

2. Patterson (1977:580) notes of Agassiz's diagram: "Agassiz's example shows clearly thatbelief in evolution is not necessary for the production of such diagrams. . The informationcontained in these diagrams is therefore not necessarily concerned with evolution or phy-logeny...

Several systematic botanists have also had similar pen:eptions about the introduction ofevolutionary thought: "It seems to be too readily assumed that the doctrine of evolution is thebasis on which classification builds, whereas in practice it is rather the reverse. . Evenwithout any theory [of evolution}, taxonomy would proceed as it always has done, with the aimof classifying organisms in the most convenient manner. which is to place together obviouslyrelated genera, species, and other groupings" (Ramsbottom. cited in Wilmott 1950:44). Wil-mott (1950:44) also notes: "The fact remains that systematists had produced a rather surpris-ingly natural system before Darwin. "

Zoologists, too, have commented on the fact that actual taxonomic practice was little af-fected by the introduction of evolutionary thought. The Origin of Species. says Hopwood(1950b:59), has had "little effect on the principles of taxonomy. which remain much as theywere in the days of Linnaeus. of Lamarck, of Cover, and of S. P. Woodward."

• •

ORDRE DES GANOIDES ORDREDES PLACOIDES.ORDREDES CYCLOIDES ORDREDES CTENOIDES

~r;; • '" '" 1• '"• z ~ ." " "- e :L; • • '" •

~• •~

•, , 0 ] H i ~ §~ ~

.=, • -e~ ·• .= 'i ... •

~,- ,~ •• , • 0 0 E •

-e '000: ~ ;; • 'l! S ~ • JJ 'c e .i' ~.~-e -$ <J ' - e • 0

.a~

" g . " • , ,~., t , E .. ~.~ ;, -= i:. 5 "ell .E .- ~-• E ee •j; J;~c.:! • J <5 <IJ :.It;; J"S -e ~ ;E

'"v. '" ~

j js~

;;,

" j},

I~-l

l .~.,"e-, ~

~~

j1"::; il ~:i

, l0• {'"

Biological Classification 153152 Biological Classification

tion that, despite some controversy through the years, has persisteduntil the present. Recently, there has been considerable debate as toexactly what Darwin meant, particularly with respect to how closelyhis views can be reconciled with one or the other of the systematicphilosophies currently in conflict Thus, evolutionary systematists(Simpson 1959, 1961; Mayr 1969, 1974) have interpreted Darwin'swritings as indicating classificatory principles similar to their own.On the other hand, Nelson (1971a, 1974) has taken issue with this in-terpretation and suggests that Darwin was, in principle, closer inphilosophy to the school of phylogenetic classification. It mightseem unimportant, except from a historical point of view, whetherDarwin's position on classification was closer to this or that modernschool of thought. From the standpoint of justifying a particularmodern-day philosophy of classification, such a debate does indeedseem unimportant, since modern conflicts in theory and methodmust be decided on merit, not historical precedent. However, aclose examination of Darwin's views is important here because theyhave had a significant impact on subsequent classification theory,and can help us understand a fundamental dilemma that has charac-terized systematic practice since the time of Aristotle (this is the di-chotomy of Aznot-A groups, to be discussed shortly). There is one ad-ditional reason: the principles enunciated in this chapter can beseen to have their germination with Darwin, and our discussion willacknowledge this debt. We also hope that this discussion will clarifyaspects of the current debate over Darwin's philosophy of classifica-tion.

Darwin perceived the pattern of organic diversity to be hierar-chically arranged (1859:411): "All organic beings are found to resem-ble each other in descending degrees, so that they can be classedin groups under groups. This classification is evidently not arbitrarylike the grouping of the stars in constellations." Referring to the onlydiagram in The Origin of Species (our figure 5,2), he makes this no-tion of hierarchical arrangement more explicit, and for the first timerelates it to classification. Darwin's detailed interpretation is at thecenter of the contemporary debate and hence is quoted here in itsentirety:

the modified descendants proceeding from one progenitor becomebroken up into groups subordinate to groups. In the diagram each let-ter on the uppermost line may represent a genus including severalspecies; and all the genera on this line form together one class, for allhave descended from one ancient but unseen parent, and con-sequently, have inherited something in common. But the three generaon the left hand have, on this same principle. much in common, andform a sub-family, distinct from that including the next two genera onthe right hand, which diverged from a common parent at the fifth stageof descent. These five genera have also much, though less, in com-mon; and they form a family distinct from that including the threegenera still further to the right hand, which diverged at a still earlierperiod, And all these genera, descended from (A). form an order dis-tinct from the genera descended from (I), So that we here have manyspecies descended from a single progenitor grouped into genera; andthe genera are included in, or subordinate to, sub-families, families,and orders, all united into one class. Thus the grand fact in natural his-tory of the subordination of group under group, which, from its familiar-ity, does not always sufficiently strike us, is in my judgment fully ex-plameo. (Pp. 412-13; italics added)

Three points are worth stressing. Darwin clearly views the evolu-tionary process as producing a hierarchical pattern of nested sets oftaxa (his "subordination of group under group"). Moreover, speciesare grouped into sets (genera, families, and so forth) because theyall have descended from a single ancestral species. Finally, heimplies, and quite strongly, that the species descended from a com-mon ancestor are united by features inherited from that ances-tor. Later in the chapter he emphasizes this last point, again referringto his illustration: "All the modified descendants from A will haveinherited something in common from their common parent, as will allthe descendants from I; so it will be with each subordinate branch ofdescendants, at each successive period" (p. 421).

From these statements it would seem Darwin was outlining aviewpoint foreshadowing the more explicit formalization of Hennig(1966)' that the pattern of the phylogenetic process is one of nestedsets of taxa defined in terms of nested sets of shared derived char-acters (synapomorphies). And, like Hennig, Darwin was quick to em-phasize the distinction between grouping on the basis of a generalmeasure of similarity and the special kind of similarity that is syn-apomorphy. Darwin also saw the genealogical pattern, manifestedby shared derived similarity, as providing the long sought key to a

I reque~t the reader to turn to the diagram lour figure 5.2] illustratingt~e action, as formally explained, of these several principles [i.e. ofhis theory of evolution]; and he will see that the inevitable result is that

• u'__ T*EH=-_f".::u: .~6"~ _!_r~H~.----~ ...~---,.,.~__I--,!: \"- J

" I I : i \,! f• i ;' 1': r :T ...-: i -----------\\,:----"i,---·------"\"---~i'f---f-

\ : t .' ", " , 'I'" \ I--.---\ -__,_+_ ---- -\'__ --._L____ \. J i\ : " \ : " : .~i:' :1 'I::

---- " ..l..J.. 'L.J.'~' . ~Lli...J~' _I 1 \ , I

.." r"""' ..,._)....;:'.i/ rl

../z'

-'-" .....1'.....,'-y>

.\·,·t/~·

i'

-, ,.; I'-,~\:.'_.'

ABC D E... \, 1 i\, j ",/. \ ,• \ I.. ~ : :", ' , ;', ,:... .

\ \ ! I

Figure S.2 Darwin's phylogenetic diagram on which he based his discussions of the

."

-,1# .1_':'- ,.,.. a~, -'--T" ----rI \, : :

! \ r !i-i ~~~ ....,~!---i--: ' !: \, !I ---------+--4---:---"

" ., ", ~' .! ._LLL _

•PL" _ .__ .... _---~.

~--~..'~--~.:_XI.V\ : !\ : :" : --1-.- XIII, ,, ' ,, ' '

• ': XII-----~-+-.-:--_.., , ,, , I~' ,, ' ,_-------.l.1 : Xl

._~ z" X

- .. .;... -""-'"-_ ~~ __ :_,<,-,,-:......_ ... __ IX

__________ -"u' ".l-.::'-- ~ -_._.------ ,...,. -'.::~.,z WI

... .;... ~u~·:.~'_'J\.;:~·:.'--------'.•';c· ·~xcs~ V

-.. '-' f- ·_-:.;::.•'fJi-2~c:.' w

-L .:.. tf'',- C''!--'!-C· ,xC''- ~ m

_----..;..--'~-'t'':.··''-··+-v-- -,-~.,,/--'-"i:,-' ~_

--\/1;1

D

_~ I

F G I L

hierarchical structure of evolutionary history and classification. (From Darwin 1859,)

H K\,•,\\\

i,,,,,,,i

!/

./,,,,:,.

,,,,:,:,!;,


order for lineage A with three families, an order for lineage I with twofamilies, and he suggests that the Recent genus FI4 should "rankwith the parent-genus F." Although he does not state so explicitly,Darwin would presumably place F and Fl4 as separate genera intheir own order. This arrangement, as Darwin notes, is genealogical,and the differences in ranking between lineages A, F, and I areminimal and reflect the frequency of branching more than morpho-logical divergence. That Darwin argued against using morphologicaldivergence to delimit classificatory groups was illustrated by the fol-lowing example: "If it could be proved that the Hottentot had des-cended from the Negro, I think he would be classified under theNegro group, however much he might differ in colour and other im-portant characters from negroes" (p. 424; italics added)."

Darwin's major efforts at classification involved barnacles, and itwas with this group that he had to confront the realities of the prac-ticing systematist. Genealogies within the group were not easy todiscern, and this ambiguity found its way into his classification at-tempts. Ghiselin and Jaffe (1973) have shown that not all of Darwin'sgroups were based strictly on genealogy and that he sometimesgave greater consideration to differences between groups than totheir similarities: "Darwin seems to have erected a system whichcompromised between evolutionary principles and utility. But he wasformulating a system for his times, and the times were not yet ripe fora strictly genealogical arrangement" (Ghiselin and Jaffe 1973:138)Nevertheless, in principle, Darwin clearly advocated genealogicalsystems and. in view of this, it would be historically inaccurate toclaim him the founding father of the school of evolutionary classifica-tion as advocated by Mayr and Simpson (see below). In his theoreti-cal writings he certainly can be placed near the proponents of phy-logenetic classification, despite the fact he was unable to achievehis theoretical ideals in practice (Ghiselin and Jaffe 1973:139).

Although Darwin unquestionably advanced the meaning of natu-ral classification, perhaps his failure to obtain such a system in prac-tice stems in part from retention of a conceptual tradition that has its


"Natural System." Indeed, Darwin was the first to propose groupingby descent rather than by general similarity (pp. 413-14):

But what is meant by this system? Some authors look at it merely as ascheme for arranging together those living objects which are mostalike, and for separating those which are most unlike . But manynaturalists think that something more is meant by the Natural System:they believe that it reveals the plan of the Creator. . Such expres-sions as that famous one of linnaeus that the characters do notmake the genus, but that the genus gives the characters, seem toimply that something more is included in our classification than mereresemblance. I believe that something more is included: and ttiat pro-pinquity of descent-the only known cause of the similarity of organicbeings-is the bond, hidden as it is by various degrees of modifica-tion, which is partially revealed to us by our classifications. (Pp.413-14)

For Darwin, and many naturalists before him, the concept of nat-ural classification meant that the perceived pattern of groups withingroups must be translated into a Linnaean hierarchy as faithfully aspossible. Darwin had some specific thoughts on how this was to beaccomplished, and his discussion has been variously interpreted bymodern systematists:

I believe that the arrangement of the groups within each class, in duesubordination and relation to the other groups, must be strictly genea-logical in order to be natural; but that the amount of difference in theseveral branches or groups, though allied in the same degree in bloodto their common progenitor, may differ greatly, being due to the dif-ferent degrees of modification which they have undergone: and this isexpressed by the forms being ranked under different genera, families,sections, or orders. (P. 420; italics in original)

What exactly did Darwin mean here? He referred the reader tohis diagram. He supposed that A through L were related genera liv-ing in the Silurian, all were descended from a single species in theremote past. Only three genera (A, F, and I) are considered further,and the question of interest is how to classify the Recent species a 14

to Z14 (the superscript refers to the fourteenth time level since theirorigin from A. F, and I). The derivative species of ancestors A and Iare supposed by Darwin to have diverged morphologically from theirrespective ancestral species, whereas the species of genus F14 havenot diverged significantly. Lineage F exhibited no branching andlittle change. Darwin suggests that there should be recognized an

3. Darwin (1859:426) seems 10emphasize this point in another. later passage: "We can under-stand why a species or a group of species may depart, in several of its most imponant character-istics, from its allies, and yet be safely classed with them. This may be safely done, and is oftendooe. as long as a sufficient number of characters, let them be ever so unimportant, betrays thehidden bond of community of descent."


origins in Aristotelian philosophy and continues to the present. Thattradition is characterized by attempts to dichotomize groups of or-ganisms into mutually exclusive sets, A and not-A sets. Historically,a distinction has not generally been made between constructing setsof taxa on the one hand and classifying them on the other; the proce-dures were identical in the minds of most workers during their at-tempts to obtain a "natural" system. A major theme of this chapter istha~ failure to appreciate the conceptual and methodological impli-cations of the Alnot-A dichotomy is a primary reason for the slowprogress that has been made in achieving natural classifications.


ticn. and organized it in terms of A and not-A groups (Lamarck 1914;

Hopwood 1950a):Animals with blood (the A group)Viviparous quadrupedsOviparous quadrupedsFishesBirds

Animals without blood (the not-A group)MollusksCrustaceansTestaceansInsects

Aristotle's A and not-A classification dominated biology until thework of John Ray (1627-1705), who extended the classification toinclude other organisms and increased its complexity but did notchange its fundamental logical structure (Hopwood 1950a:28-29).(See the cladogram on page 160.)

The classification of Linnaeus did not alter Ray's scheme in anyfundamental way with respect to the defining properties of thevarious groups. The class Vermes of Ljnnaeus, e.g., was reservedfor those animals lacking both skeletons and articulated legs.

The early nineteenth century was a time when classificationsbegan to take a truly modern form. Primarily through the efforts of theFrench zoologists J. B. Lamarck and G. Ouvter. major groups of or-ganisms were delineated and defined. However, both these scien-tists followed earlier tradition and admitted not-A groups into theirsystems (see Hopwood 1950b). The most famous example, ofcourse, is Lamarck's division of the Animalia into two major groups,the Vertebrata and Invertebrata. Lamarck (for a summary, see 1914)arranged his 16 classes of animals in an ascending series of com-plexity, and a class was often defined in terms of the absence ofcharacters that defined succeeding, "higher" classes. Thus, wormslack legs, cirri peds lack eyes, mollusks lack a spinal cord, fishesand reptiles lack hair or feathers, and birds lack mammary glands.

The realization that groups which are defined in terms of "nega-tive" characters might not be natural was seldom stated explicitly.Lamarck realized that Linnaean groups such as Vermes were"wastebaskets" and contained a heterogeneous assemblage of taxa.Nevertheless, he too continued this tradition and erected classes onthe same basis. Indeed, as a general theme in classification, the rec-

Classification and the Dichotomy of Aand not-A Groups

"A qroups'' are characterized as those groups defined by the pos-session of shared derived character-states (synapomorphies),whereas "not-A groups" are those formed either on the basis of lackof c~aracters defining A groups or by sharing character-states con-trasting. to those defining A groups. In either case, not-A groups arerecognized most generally because of possession of the primitivecondition. (It is not implied, of course, that all groups defined in thisway are necessarily non-monophyletic, only that they generally are.)

The history of A and not-A groups is as old as classification it-self. There never has been a time when efforts to classify organismsaccording to the AJnot-A dichotomy have not been commonplace.Such a conception of natural order formed the basis for the first bio-logical classifications and even today remains dominant-indeed,t~e contemporary school of evolutionary classification, as will bediscussed, extolls the virtues of the Alnot-A conceptualization... The presence of A and not-A groups in biological classification

originated with man's earliest attempts to develop a logical ap-proach to thought processes. Specifically, such methods of groupingwere a natural outgrowth of the "Laws of Thought" including the"Law of Contradiction" (an object cannot both be A and not be A)and the "Law of the Excluded Middle" (an object must either be A ornot be A). Aristotle made the first significant attempt at a classitica-

160 Biological Classification Biological Classification 161

better show just how pervasive the influence of the AJnot-Adichotomyhas been, several examples can be considered.

Animals

/Blooded Not-blooded

l \./ \1~.\TW\'.\

~1~~~~~:dS serpent,' Vi.ipo '\ "'i .. rous

Aqotic Terrestrialognition of not-A groups has been accepted systematic practice,even to the present. However, there has always existed an aware-ness that classificatory groups should not be founded on negativecharacters, i.e., that not-A groups are somehow not "natural." Darwin~a.s.explici~ in advocating that groups be defined by inherited simi-anties, which would, in effect, make those groups A groups. Al-though there has not always been a conscious recognition of theAJnot-Adichotomy, the history of classification seems to describe atend~ncy to eliminate not-A groups; intuitively, biologists have re-cognized that not-A groups are not natural in the same way that Agroups are. Nevertheless, not-A groups are still prevalent. In order to

Example 1. The Kingdoms of Organisms

Since ancient times, organisms have been divided into animals(sensible, motile) and plants (insensible, nonmotile)-a perfect ex-ample of A and not-A groups. This division has continued essentiallywithout change until recent decades when technical advances per-mitted more detailed analyses of morphology, biochemistry, andgenetics. Now, considerable discussion is ensuing as to what con-stitutes a "natural" classification of living organisms (see review inMargulis 1974). Despite considerable attention having been directedtoward eliminating some not-A groups (e.g., placement of all"plants" together, or recognition of the dichotomy within animals ofprotozoans and metazoans), it would seem that current attempts at"natural" systems continue to admit not-A groups.

Most modern specialists, for example, accept a dichotomy be-tween organisms with prokaryotic cells and those with eukaryoticcells. The prokaryotes are interpretable as a classic not-A group, ascan be seen when comparing their defining properties with those ofeukaryotes (from Margulis 1974, table 3):

Prokaryotes

Small cellsNucleoid notmembrane-boundNo centrioles,mitotic spindle,or mlcrotubules

SexualsystemabsentNo mitochondriaNo intracellularmovement

Eukaryotes

LargecellsNucleusmembrane-boundCentriolesgenerallypresent,mitoticspindles,microtubulespresent

SexualsystemgenerallypresentMitochondriaIntracellularmovement

In general, then, prokaryotes (e.g., bacteria) are defined in termsof negative characters, and this raises the question of whether theyshould be considered a natural group on that basis. The implicationof the above definition is that they are a group of organisms, none ofwhich can be classified as a eukaryote.

The second major division of organisms, the superkingdom Eu-karyota, contains "protists" (protozoans, some algae, flagellatefungi), amastigote fungi, plants (bryophytes, tracheophytes), and an-imals (multicellular animals). Here, too, the concept of A and not-A


Cuvierhimself oftendescribed the purposeof a naturalclassification,and its advantageover an artificial arrangement,in terms of beingable to make positive general statementsthat would be true for allmembersof the group.Thiswould not be thecase for a groupconsist-ing merelyof remainders,animals left over after the formationof an-other group, having only negative characters in common. It seemsclear that the Radiatawas sucha collectionof left-overs,andmuchoftheworkdone on its classificationin subsequentyearscan be viewedas a searchfor positivecharacterswith which to define true radiates,while animals assignedthat place by defaultwere being removedtotheir proper home,(Winsor1976:15),

The Radiata of Cuvter was viewed by most subsequent workersas a simplified collection of basically dissimilar forms. Lamarck(1914), for example, split the Radiata into at least four different inver-tebrate classes: infusorians, polyps (rotifers, hydrozoans, antho-zoans, sponges), radtanans (various coelenterates, echinoderms),and worms (of various kinds). Many of the subdivisions within thesefour classes were based on negative characters and thus constitutednot-A groups (Hopwood, 1950b:54-55).

During the nineteenth century, many biologists turned to theproblem of describing and comparing the vast numbers of inver-tebrates being discovered, and eventually a more nearly naturalclassification came into existence. Of particular importance were thetedious analyses of life histories of various coelenterate groups, andthe discovery of alteration of generations united the polyps on theone hand and acaleph medusa ids on the other. Likewise, life historyand detailed anatomical investigations began to clarify the positionof the echinoderms relative to other "radiate" groups such as thecoelenterates. Soon it was realized that echinoderms were distinctand could be defined by their own positive characters, includingtheir internal organization, separate gastrointestinal track, and pecu-liar water-vascular system. As Winsor (1976), Hopwood (19S0b), andothers have noted, all these advances led to the replacement of ataxonomic arrangement of large, ill-defined groups (not-A groups) byone emphasizing smaller, more "naturally" defined taxa (A groups).

groups pervades the classification, for the protists include "eukar-votes. primarily microorganisms, which are neither metazoan ani-mals, embryophyte green plants, nor amastigote fungi" (Margulis1974:56).

Within the kingdom Animatla. subdivisions are also organized interms of not-A and A groups. For example, the Parazoa versus Eume-tazoa; Radiata versus Bilaterta; the Acoelomata versus Pseudocoelo-mata, versus Coelomata; and possibly even the Protostomia versusDeutero.stomia.. It remains to be seen whether the not-A groups ofthese dichotomies are natural groups in a genealogical sense.

Example 2. Classification of Invertebrate Animals

The "Invertebrata," as noted above, is a classic example of a not-Ag~oup, but at the time of its introduction near the beginning of then~~eteenthcentury, it represented an advance over previous etas-siffcatory systems. Linnaeus, for example, recognized only insectsand worms. as the major groups of invertebrates, an arrangement~arcety different from that of Aristotle. But by the end of theeighteenth and beginning of the nineteenth centuries there was ad~.ter~ination on the part of zoologists to produce a' natural etas-slfication of these animals. In these early attempts at classification,n?~-A.groups were prevalent, and the history of invertebrate clas-aiftcation through the first half of the nineteenth century appears tod~c~ment an unconscious attempt on the part of systematists toeliminate such groups (see especially Winsor 1976 from which thisaccount is largely taken). '

In 1812 Cuvier recognized four major divisions, or embranche-men~s,of the animal kingdom: Vertebrata, Mollusca, Articulata, andRadiata. The fi~st three contained bilaterally symmetrical organismsand were considered easily definable in terms of the construction oftheir nervous and circulatory systems. The fourth group, the Radiata,was not so easily distinguishable and was characterized by the ab-sence of a heart, a brain, and a well-developed nervous system. Fur-thermore.' they were "rayed animals" with their internal organs dis-played In a radially symmetrical pattern. The Radiata includedgrou~s as disparate as sipunculid worms, echinoderms, polyps, me-dusol? fo~ms,and even "infusorians" (protozoans, rotifers). The tax-ononuc history of the Radiata seems a microcosm of the aspirationsof systematics in general:

Example 3. Classification of the VertebrataUnlike the concept of the "Invertebrata," few systematists have ques-tioned the monophyly and hence the "naturalness" of the Vertebrata.Most of the major groups of vertebrates-mammals, birds, reptiles,


fish----have their origins as taxonomic concepts back in antiquity,and their naturalness has seldom been disputed, even in moderntimes. But it has become increasingly apparent that much of the gen-erally accepted vertebrate classification is in need of re-examina-tion, and it can be suggested that in large part this has been neces-sitated because of a historical reliance on the AJnot-A tradition. Thistradition can be illustrated by the following classification, abstractedfrom modern general zoology, comparative anatomy, and vertebratepaleontology texts (see particularly Kent, 1965; Romer, 1962, 1966;Storer and Usinger, 1965):

A. Invertebrata(non-Chordata)AA. Chordata(notochord,dorsal nervecord, etc.)8. Acrania(absenceof cranium,vertebrae,brain)88. Craniata(= Vertebrata)(cranium,vertebrae,brain)C. Agnatha(absenceof jaws and paired appendages)CC. Gnathostoma(jaws,paired appendages)D. Pisces(nontetrapod,"aquatic" gnathostomes)DO. Tetrapoda(tetrapod,"terrestrial" gnathostomes)E. Amphibia(anamnioteegg)EE. Amniota(amnioteegg)F. Reptilia(absenceof advancedavianor mammalian

characters)G. Anapsida(no temporalopening)GG. Otherreptiles (temporalopening)FF. AvesFFF.Mammalia

Plan of the ChapterThe remainder of this chapter will be devoted to a detailed discus-sion of the principles of biological classification, current controver-sies within classification theory, as reflected by differences in theprocedures adopted by alternative systematic "schools," and a con-sideration of methods of classification. The approach of the chapteris therefore tripartite. In the first part, the logical structure of the Lin-naean hierarchy itself will be considered, with an emphasis on thelimitations placed on biological classification by that structure. Thesecond part will present a general discussion of branching diagramsand their relevance for classification. Moreover, here will be in-cluded a critical analysis of several contemporary systematicschools. Finally, in the third part, the various procedures needed toproduce phylogenetic classifications will be outlined and examined.The principles discussed there should provide the reader with abasis for investigating the systematic literature in more detail andproducing satisfactory classifications of his or her own.

Throughout the chapter, in the expositions and critiques, runs acommon theme: that biological classification, to be natural, shouldbe based on the concept of recognizing A groups (monophyletictaxa, as defined in the chapter on cladogram analysis) and eliminat-ing not-A groups (nonmonophyletic taxa). It is recommended thatthis be accomplished by classification based on the precise repre-sentation of the nested sets of taxa uncovered by cladogram analy-sis. Since it has been amply demonstrated in previous chapters thatcladograms are more general concepts than are alternative kinds ofbranching diagrams, the position is adopted here that classificationsbased on cladograms will result in a general reference system thatnot only is natural (as conceived from a historical and Darwiniantradition) and of great practical utility, but also one that reflects t.heunderlying scientific (hypothetico-deductive) structure of systematics

in general.

A classification closely approximating this form has dominatedvert~br~te biology for the last half-century, if not longer. The per-vasrve Influence of the Arnot-A dichotomy is readily apparent. In re-cent years, however, systematists have come to question the natural-ness of the not-A groups of this classification, and alternativegroupings are being proposed. It is apparent, for example, that theAcra~~a,Agnatha, Pisces, perhaps the Amphibia, and certainly the~ePtllla are groups of unrelated taxa combined on the basis of nega-tive characters (a general review of this evidence can be found inMay-Thomas and Miles 1971; Greenwood, Miles, and Patterson1973; and Levtrup. 1977). Clearly, the history of vertebrate classifica-tion shows a tendency toward the elimination of not-A sets, and thesame historical trends are evident even within the vertebrates [for ex-ample, teleost fishes (Patterson 1977) and mammals (McKenna1975)].

Structure of the Linnaean Hierarchy

As a general concept, classification is a way of ordering or arrang-ing knowledge. Biological classification is an attempt to order or ar-


a different hierarchical level, or rank. Linnaeus himself recognizedonly five categories:

KINGDOM

CLASS

ORDER

GENUS

SPECIES


range knowledge about biological phenomena. Such a statementleads one to ask: What kinds of phenomena are to be classified? Fur-thermore, what are to be the principles or methods to be adoptedwhen producing a classification and how is that classification to beexpressed?

Within biology, the last question has had a clear, almost univer-sally accepted answer for over 200 years: biological classificationsare expressed in terms of the Linnaean hierarchy. The fact that Lin-naean classification has gained such wide acceptance is surprisingin itself, considering the remarkably diverse systematic philosophiesthat have existed and continue to do so. Because of this, one mightbe tempted to wonder whether the structure of Linnaean classifica-tion reveals an underlying unity within systematic biology, regard-less of the methods and philosophical leanings of the individual sys-tematist. At the same time, the philosophical diversity seen withinsystematic biology seems to suggest that many systematists mayhave lost sight of the logical and biological implications of the Lin-naean system itself, and consequently have not perceived the con-flicts that exist between some of the proposed kinds of knowledgethat might be expressed in classification and the capability of theLinnaean system to organize and express that knowledge. Actually,both observations seem to be true, and together they may explainthe general allegiance of the systematic community to the Linnaeansystem and the incredible diversity of opinion-and controversy-concerning the purposes, methods, and uses of classification.

What is it about the Unnaean system that seems to unify suchdiverse approaches to systematics? What kind of basic knowledgeis being expressed within Unnaean classifications? To answer thesetwo fundamental questions, the structure of the Unnaean system it-self must be considered.

Through the ages, biologists have perceived a hierarchy of taxa,and the Linnaean classificatory system came into being as a way tocommunicate that taxic hierarchy by means of an ordering of ca-tegories. The Linnaean system was created, not as an attempt toorder all kinds of knowledge about organisms, but rather to order thekinds of organisms themselves. Consequently, Linnaean classifica-tions are lists of taxa. The order of these taxa is specified by a seriesof categories arranged hierarchically: certain categories are includedwithin (or are subordinate to) others, so that each category assumes

Through the years, many additional categories have come into use.Simpson (1961:17), for example, lists 21 levels, commonly appliedin classifications:

KINGDOM

PHYLUM

SUBPHYLUM

SUPERCLASS

CLASS

SUBCLASS

INFRACLASS

COHORT

SUPERORDER

ORDER

SUBORDER

INFRAORDER

SUPERFAMILY

FAMILY

SUBFAMILY

TRIBE

SUBTRIBE

GENUS

SUBGENUS

SPECIES

SUBSPECIES

Not all of these categories are necessarily used in any given clas-sification, but then again, all these and more have been adoptedwhen some systematists have attempted to classify complex groups.As a matter of practicality, the number of categories to be employedwill depend upon the manner in which the systematist wishes to .rep-resent the hierarchy of taxa. This point will be dealt with in consider-able detail as this chapter progresses.

Earlier chapters on cladograms, species, and trees showed howtaxa can be defined and how knowledge concerning taxa can berepresented in a hierarchical pattern. Taxa, at all hierarchical levels,

..


Such systems consist of sets within sets (see Hull 1964, and Buckand Hull, 1966, for an extended discussion of the logical structure ofthe Linnaean hierarchy). Information is a result of objects being clas-sified together into sets. The hierarchical system conveys a patternof nested sets, defined first by the subordination of subsets at a levelor rank within a higher, or more inclusive, level (or rank), and secondby the sequencing of sets at the same level or rank. Hence, the fol-lowing hierarchical classification conveys information by subordina-tion (the placing of orders A and B, within Class 1, families a and bwithin order A, and so forth) and by sequencing (families c, d, and ewithin order B, for example).


consist of one or more species, and, as we have emphasized, spe-cies are ontologically "real" units in nature. Categories, on the otherhand, cannot be said to "be real" in the same sense as taxa.Categories are conceptual-linguistic repositories of named taxa. Fur-thermore, categories are independent of the included taxa in thesense that named taxa may be assigned to any category and do notinherently belong in one as compared to another (this is not to saythat an arbitrary placement of taxa within categories would necessar-ily make biological or heuristic sense). One of the keys to under-standing many of the current conflicts among alternative methods ofclassification is to recognize this dichotomy between reconstructingthe hierarchy of taxa and then using that hierarchy as the basis ofclassification. Systematic methodology, viewed as a whole, mustproceed in one direction, from discovering the taxic hierarchy toclassifying it. By classification, one simply means the assignment ofthe various taxa to the different categories (i.e., assignment of rank).By definition, then, taxa logically precede categories.

To a biologist with some systematic background, the above maysound trivial and self-evident, but it can be argued that such is notthe case if one views a classification as a repository of knowledgeabout a group of organisms. If classification is a repository or a rep-resentation of our knowledge, it can be so only in the sense that itrepresents a hierarchy of taxa. A Linnaean classification is nothingmore than a system of names hierarchically arranged.

There is, of course, ample historical precedent within the sys-tematic literature for the view that classifications are devices to storeand retrieve "information" about organisms:

A classification is a communication system, and the best one is thatwhich combines greatest information content with greatest ease of in-formation retrieval (Mayr 1969:98)In practice, a classification is a descriptive arrangement not only lor?onversing about its included objects, but also for storing and retriev-Ing information concerning them. (Ross 1974:258)

If the elements of Linnaean classifications are lists of names,then the "information content" of that classification must be thedirect consequence of the structure (i.e., the topology) of the clas-sificatory hierarchy itself. What aspects of the hierarchy provide thebasis for conveying information? Clearly, that information can onlyreside in the logical structure of hierarchical systems in general.

CLASS 1ORDER AFAMILY aFAMILY b

OROER BFAMILY C

FAMILY dFAMILye

CLASS 2ORDER CFAMILY IFAMILY g

At this point of the discussion it does not matter what .the basiswas for forming the above classification. What is important ~sth~t theinformation content of the classification is manifested by Its hierar-chical topology. Indeed, that topology, and thus the information, canbe translated into a branching diagram (see p. 170).

What is the information contained in these "classifications," onea Linnaean hierarchy, the other a branching diagram? QUit~ sirr:Ply,it is that the taxa (or objects, if the classification is nonblolog

lcal)

form nested sets and the elements of these .sets must share someset-defining properties. It is not the properties themselves tha.t .arespecified by the classification, only the sets. In the above classlf.l~a-. . ' , 1 2 d rs A + B and familiesnon, five sets are specified: classes + ,or e ..a + b, c + d + e, and f + g. This classification implies that familIes a and

.... th f mcd and e and thatb share some property distingUishing em ro " ' .families a.b.c.o. and e share some defining property not shared Withf and g. What are these properties? The classification cannot tell us.

170 Btoloqical ClassificationBiological Classification 171

branching diagrams are to be used as the basis for classification?As will be discussed below, there are several kinds of branchingdiagrams, each of which is translatable into a classification. The rel-evant question would seem to be: are some kinds of branching dia-grams more useful than others?

Given these two points, it can be concluded that biologists whoadhere to the Linnaean system in biological classification must becareful to respect the logical structure of that system. Branchingdiagrams representing knowledge about taxa become the focus ofcontroversies in classification, rather than the classifications them-selves. Hence, to a considerable extent, the scientific analysis of thepurpose, expression, and formulation of those branching diagrams is

the key issue of systematics.

Family a b c d e f 9

Order B C

Closs 2

Information Content and Branching Diagrams

Thus the information content of a classification is only the nestedsets; nothing more can be retrieved. 4

The foregoing statements have two possible implications forresolving some of the arguments about methods of classification soprevalent in the systematic literature. First, it would seem that infor-mation or knowledge about organisms can be used for classificationonly if it can be expressed in terms of a branching diagram of thetaxa under study. In other words, such information must have a logi-cal structure equivalent to that of the Linnaean hierarchy. This maynot be an important restriction, since it should be possible to repre-sent shared properties of taxa in a hierarchical arrangement regard-less of the properties or the ways in which similarities among taxaare evaluated. Second, the question of what kinds of classificationsare useful within biology perhaps should be restated: what kinds of

The preceding section stated, but did not pursue in detail, two s~~m-ingly important principles of classification. First, Linnaean clasaifica-tions only express a hierarchical arrangement of taxa. They cann~tdirectly convey anything about the properties of taxa, or abo.ut theirsimilarities and differences. If such information is expressed In clas-sification, it is implied only through the mind of an individu~1 r thetemattst and mayor may not be shared with others contem.platlng thsame classification. Strictly speaking, then, the information conte.ntof a classification is a hierarchy of taxa, nothing more. As Will be d~s-cussed however there may be some general agreement (or dis-

, , " th b! logical implicationsagreement. even) among biOlogists as to e 10

of that hierarchy. .Second in order for Linnaean hierarchies to convey informa~lon

precisely, there must be a one-to-one correspondence-an ISO-metry-between the branching diagram of the taxa th~t was used asa basis for the classification and the classification Itself. In ot~er

. . f mation-as most bicl-words, If classifications are to convey In or dt t te that a clas-ogists agree they should-then common sense IC as.'. , .' t tion (i e the taxrcsification should represent precisely the In orma .. ,hierarchy) on which it is based. This one-to-one correspondence

4. Farris (1977:84]) also raises this point in his response to phenencists arguments that phy-logenetic classifications do net express any phenetic information: "No classification, even aphenetic one, is by itself any more than a collection of groups of organisms. No grouping of or-ganisms. however arrived at, can directly express anything more about the organisms than themembership of the groups. Classifications may possess 'infonnation content' only in the indi-rect sense that a classification may be used as a reference system," (See also arguments ofevolutionary systematists. e.g., Johnson, 1970:233.) It can be added, in the context of thepresent discussion, that "reference system" can only refer to the property of set-membership.

..


provides a basis for comparative evaluation of alternative classifica-tions: that classification which most nearly represents the original hi-erarchy of taxa is best Naturally, such an evaluation depends in parton the amount of agreement regarding the taxic hierarchy, and thisin turn is amenable to scientific analysis, as discussed in chapter 2.

Although it may seem persuasive to most readers that Linnaeanclassifications can express information only about the structure ofthe taxic hierarchy, many systematists have been explicit in theirbelief that classifications can convey more than just this topologicalinformation. It has been stated repeatedly that classificationsvariously express phylogenetic relationships (genealogy), phenetic(overall) similarity, genetic similarity, evolutionary history, and soforth. Thus, to some systematists, the topological structure of thetaxic hierarchy is a direct translation of statements about phyloge-netic relationships or phenetic similarity, To others, classificationsare to be viewed as "heuristic" systems and as schemes that canexpress an assemblage of evolutionary information, albeit only im-precisely.

What is to be said of these viewpoints? What kinds of informa-tion can be expressed by (or contained in) Linnaean classifications?Should not that information be readily accessible to all interestedbiologists, even those lacking special knowledge of the taxa inquestion? One cannot help but agree with Bock (1977:861), "Judg-ment of the usefulness of classifications is not in terms of the sys-tematists who construct them, but with respect to the other biologistswho are dependent upon classifications as the foundation for theircomparative studies and the formation of their generalizations."

Consider the following hypothetical Linnaean classification:

FAMILY ASUBFAMILY a

SPECIES 1SUBFAMILYa'TRIBE b

SPECIES 2TRIBE b'GENUS CSUBGENUS dSPECIES 3SPECIES 4

SUBGENUS d'SPECIES 5

GENUS c 'SPECIES 6SPECIES 7

If there are generalizations or principles about information storageand retrieval appl icable from one classification to the next, it shouldbe possible to identify and discuss them relative to the above hypo-thetical classification, If, on the other hand, there are no such gener-alizations or principles, then it would seem impossible that any spe-cial classification of real taxa could be rationally defended in termsof its ability to store and retrieve information.

What information is expressed by this classification? Clearly,the classification is expressing something about the set-relations ofseven species, and these relations are defined by the hierarchicalarrangement of the categories in which those species are placed,The classification can be translated directly into a taxic hierarchy orbranching diagram:

Species1 Species 2 Species3 Species4 Species5 Species6 Species7

1.

But why this specific branching diagram? What is inherent inour conception of the Linnaean hierarchy that might prevent us fromconsidering the following diagram (p. 174) or anyone of the manyalternatives that could be constructed?

. ti ontent of hierar-The answer of course is that the Informa Ion c ," I T tonschical arrangements, whether in the form of Linnaean c assI rca I

or branching diagrams, is expressed in terms of neste? s~ts.Branching diagram 1 is a direct translation of the nested set.sImpliedby the classification' branching diagram 2 is not. As a Simple ex-ample, diagram 2 i;"Plies a nested set of species 5, 6, .and 7,Whereasthe classification indicates that species 5 is nested Instead


derivetwo alternative branching diagrams. In actual taxonomi~ prac-tice, of course, some conception of a branching diagram will pre-cedethe classification. The example is also utustrative of the secondprinciple of classification referred to earl ier: the informat~on contentof a classification is conveyed precisely only when there IS a one-to-onecorrespondencewith a branching diagram. If branching diagram 1wereused as the basis for constructing the classification, then theclassification itself could be used subsequently by any biologist-howeverunfamiliar he or she might be with the group-to reconstructthebranching diagram. Obviously, the classification would not yieldbranchingdiagram 2.


Species1 Species2 Species3 Species4 Species 5 Species 6 Species7

2.

with species 3 and 4. So it is evident that a Linnaean classificationdoes contain information, in the sense that a topography of nestedsets can be retrieved. What other kinds of information might be con-tained within that classification? The inference might be made thatthose taxa contained within the same set share some similarity notshared with those taxa excluded from the set. Thus, it might be sup-posed that species 5 shares some attribute or attributes with species3 and 4 not shared with species 6 or 7. Consequently, some may saythat the hierarchy implies the sharing of set-defining properties. Ifthis is the case, the classification cannot tell us how many sharedsimilarities are involved, what kinds of similarities they might be, oreven that species contained within a set are actually more similar toone another than to a species excluded from the set. The only con-clusion implied by the hierarchical arrangement is the possiblepresence of set-defining properties. However, all of the foregoingrests on the assumption, adopted prior to the investigation itself, thatthe hierarchy is to be characterized by set-defining properties pos-sessed by the taxa, On the other hand, hierarchies, including per-haps the classification above, can be formed without the assumptionof set-defining properties, and indeed, one branch of systematics(numerical taxonomy) advocates just that. So there would seem to benothing inherent in an hierarchical classification (branching dia-gram), other than the arrangement of nested sets itself, conveying in-formation of a special kind about the included taxa.

In the above example, a Linnaean classification was used to

Branching Diagrams and Their Role in Classification

Themyriad of controversies, which for years have festered within theliteratureof biological classification, can be attributed, for the mostpart, to a single, critical issue, i.e., the kind of branching diagramthat should be used as the basis for classification, Naturally, what-ever branching diagram might be chosen by a systematist will con-vey that individual's conception about the purpose of classification.

Mostmodern systematists have advocated one of two systems ofbranching diagrams: phenograms or evolutionary trees. It will be amajor thesis of this chapter, however, that a third type of branchingdiagram-the cladogram, as defined and discussed in previouschapters-constitutes a more general class of branching diagramand is, consequently, preferable as a basis for classification Beforepursuing that line of argument, the salient characteristics of pheno-grams and evolutionary trees, and their implications for classifica-tion, should be discussed,

Pbenograms

~henograms are branching diagrams depicting the phenetic rela-tIonships of the included taxa. By "phenetic relationships" is meantthe degree of phenetic similarity, a measure of "overall similarity,basedon all available characters without any weighting" (Cain and

iIII

p ----------- ••----------- .. ------ •••••• II!!!!!!!!!!!!_!!!!!_!!_ •••• ~~~.~:~~:~-~:::r:=~.~...: ••••••••• ~.


Harrison 1960;3). Phenograms are the creation of numerical tax-onomy (Sakal and Sneath 1963; Sneath and Sakal 1973) and tradi-tionally are defended by numerical taxonomists as providing clas-sifications that are "general," "natural," and preferable scientifically.Numerical taxonomy expanded rapidly in the early 1960s, recruitingmany young practitioners, but since the end of the decade, and cer-tainly through the 19705, its influence in bioloqical classification haswaned.

To summarize their methods briefly: In order to produce pheno-grams a large suite of characters is chosen, and the alternativecharacter-statesof each character are coded for each taxonomic unit(data may be qualitative or quantitative). There follow two computeroperations manipulating these basic data: first. the estimation of sim-ilarity among taxonomic units by calculating a similarity coefficientusing one of many algorithms designed for this purpose; and sec-and, the clustering of taxonomic units into a hierarchical arrange-ment (a phenogram; see figure 5.3) by application of a clusteringalgorithm to the similarity matrix produced by the first computeroperation, [Sneathand Sokal (1973:114-308) discuss in considerabledetail those algorithms cOmmonly used to produce similarity matri-ces and phenograms; Farris (1977) presents a concise review ofphenetic principles and techniques.]

How successful are phenograms in affording a basis for clas-sification? To jUdge from their own statements, numerical tax-onomists have been less than satisfied with their results:

A number of guides have been proposed for evaluating and determin-ing what is the "best" phenetic classification, given several potentialcandidates. However, it is clear that no complete hard and fast set ofrules for finding the best Possible classification is possible at present(and such a presumably ideal approach will probably not be availablefor some time to come). (Schnell 1970:294)

The reason for this pessimistic outlook, and it has been commonknowledge Within systematics for some time is that minor alterationsin numerical taxonomic procedures may radically alter the hierarchi-cal structure of the included taxonomic units (see Minkoff 1965;Schnell 1970. and other citations in Johnson 1970:223). Changes inthe number of taxa being studied, the choice of characters beinganalyzed, the type of similarity coefficient being computed (Sakaland Camln 1965), and the clustering technique that is adopted, all

TAXONOMICo 2 •

-f---

'--,o 2 4

TAXONOMIC

RANK6

6RANK

8

, C...TH.... C' ... """.

" '''''''''"''''''J' POfWll>O.JS, S,,"C,,"""!,," p .... "'"ncus• STE""""",",," LOfoM;IC",-",-,>, C","AN.J5 "",Incus

6 G... IAN.J5 """""sen, e'G""H!L. ,."'''"'.• LAAUS FU..!CI>;osu:;, L""us ""'-"XSt,,"

, 0 L"'US "i"'",""",",I" L'""us L£U<;OP~"",,- ......12 L""lS ........ ~IOi!I" L US .£Lo-<:"",. L llS "''''''''r.OST.I., L""US >LOov:rr<uI. L... lS ""L .......""""." L"'US '''''--IS" L"'US "'CENUTlJ5'" L"'C-S T",,¥<U>0 LAA'"" C"SOJs

", L"'U' '",--:tF""""CU5"' LMus OCCIOCNT "'--ISi" L"''''' DO<'UNICN<JSe. L""US SOoIsnSlious" LMUS .... Iru;e. LAAus "'--'<.crSCEP<S" L"'US ~YP£I<"""<LIS,. LMu" LCLIC(lPt""",'" LMUS r<:HTHY""rus" LMUS "Rlcru...'T L"'US .~LN<IC&_L15« L""US CIJ<ROC!;PHALus" LAOlJ5 "".~AN.J5" L US """D<eN<" ._ .J; NOV'oc'"""-LMrlI,OC,. L"'U; ~~~_us" LMus ""--,-",,rlo L"'US "'=rPE .... !5'" '--""US .ID'OlHll.l5'0 L"'US """"I., LAAL<8 prcr..."OCLP~Lo.'2 L"'US ~I""TL<8.. L""US S"'""""""'•• RHOOOSTn><Lo, .0"""•• U's>. ,"W.cTYLA•• ~I5S. "~Evr.OStFns., a>""""us FU<<;ATl"•• """, """INI." C,",-IOO'<IAS ><Y".IOAsa Oi...roo,<r .... L',""OPTOl>51 0i...1OONIAS NI"".." P>W:T""", S_LCx" CEL<Jc>.<:LIW< "'Lonc.,. ~,"""OPR"""" TSCH<""~".SS STeRN. "-"N<n ...SO STeRNA HIIlL.l'OINA""." St"" .... """'--"'"SO STE"NA "AA""IS""~OS STERNA VIT T~.0 Sf"' ,,""c ....., ST£I< ~"".T£I<I6" ST<"N. "'Lct:"'-JIS! "'£RNA Da.K;"-L!!6 •• T"" ST.IAT..5 5T< ~P.ESSA•• STERNA SU'1"'TR"""6' STD> 'U-""""~STe"6. Sre ,,-runc.5' STe"NA LLIU.T'"" .Te N<AE_TUS" STER ".'SC.T>" ST£"N. "'"ErS

,. STE~"'" "--"ERCILI ... IS" ST"'N'" B",--"""""",'" STERNA LOIlAT." 5T1:...... "-BInlONS18 T_ ""SElJ5 "O>QII" TH"--"SS£lS tw<I"-'S80 T>"'-ASSEus BENC"-ENSIS., T_.""SEus ZII+E"""""",", TH"'--ASSCUS £U",,,,,.T,,,," THALAS"""S £L'C""",•• T_""",,us S"I'<lvIC<NSISe. LAAoSTD> ..... INCA.s eROC£LST~R"'" ~<.LEA

"' """'-'S S'OLIDJSOR """"-JS TEN.JIl<OSTllISOS """"-'s ~IN.ftFJS'0 cyeIs ,,-0.., "YNCHCf'S N!""~9' "YNCHCf's ",-"VIJ<OSTRISOJ "YNCHCf's "--"ICOLLIS

Figure 5.3 An example ofa phenoqram, a branchingdiagram showing thehierarchical pattern ofgeneral overall similarityexhibited among sometaxa of birds (see text),(From Schnell 1970:36.)

,

....178 Biological Classification

have an influence on the nested pattern of taxa that is on the resul-tant phenoqram. As a consequence, and seeminqfy as a way toelude t~IS problem, numerical taxonomists often suggest the use ofmany dlff~rent kinds of classifications, each to be employed for theirown special purpose. But even this proposal appears questionableto some phenel",c" t " "Wh"l ". IS S. J e numerical taxonomy can supply morerefined special classifications it is doubtful if the effort of creatingthem would be worthwhile in most cases" (Ehrlich and Ehrlich1967316)"

The guides 0 iter! ", r en ena, for choosing among alternative pheno-grams are mathematical in nature (e.g., see Schnell 1970:294).Sneath and Sokal (1973:278), for instance, recommend acceptingt~at phenogram which is the "best fit" to the original similarity rna-tnx. To measure that fit th '.' .(S k I, ey use a copnenetlc correlation coefficiento al and Rohlf 1962), which is a measure of the relative closeness

among taxa implied bv a ot . ... .t th y a gIVen phenetic classification with respectOf the actual values of overall similarity calculated for the attributeso e taxa. While su hd f c a measure may serve as an adequate proce-ure or evaluating h "acce t" h P enograms, ItS adoption creates problems for

as Fp I~g P enograms as a general classificatory scheme because,arns has recent! h

h",gh h ' Y sown, cladograms consistently produceer cop enetlc cor I t! . . .h re a Ion coefficients than do the corresponding

p enograms for the same set of taxa (1977:838):

In every case the IT'soecrar " '" . c ass, rcanon constructed by clustering according tosimi artty [i eperior to th I ' -,synapomorphous similarity] was found to be su-mea" "a' a ,produced through clustering by overall similarity. The

".ueothesquad h' . "sification d re cop enenc correlation coetttcrenr for eras-063 anothe r uceo by clustering according to overall similarity was

", e <crrescondln I " ""larity w . "g mean or analyses based on special sum-as conSiderably h h 8 " "larity doe Ig er at O. 9,. clustering by overall sum-

to the pri:c7°: seem to be an optimal classificatory method accordingpes of phenetics, (Italics added)

These mathematic Ioverlook' a approaches to evaluating phenograms also

an Important point"· .natural ph : SCience ISnot the process of measuring

enomena' that !1970'229" u I' .' _ IS merely a technique of science" (Johnson

. , I a ICS In ongin I) Phmeasurements f . . . a, enoqrams summarize and representfundamentall fa similarity among taxa and, as such seem to differ

y rom other bran hi di "that the latter d c mg raqrams (trees, cladograms) mo not manifest measurements but rather constitute tty-


pothetical constructs capable of testing. Phenograms, trees, andcladograms all depict a nested pattern of taxa. With phenograms,the nested pattern is generated by a measurement of similarity, butnot directly by a search for a nested pattern of similarity, since thedifferent synapomorphies are not being partitioned hierarchically.With cladograms, and to a certain extent with trees, the nested setsof taxa are a direct consequence of hypothesizing a pattern aboutobserved similarities (synapomorphies). And that pattern can be sci-entifically analyzed or tested: alternative patterns are evaluated withrespect to the same observed similarities and "best" patterns arethose with fewest conflicts in forming nested sets of taxa (seechapter 2). Choice among phenograms derives not so much from dif-ferences in the capability of those phenograms to organize patternwithin a given collection of observed similarities, but rather from acalculated measure of how well a particular clustering algorithm inprinciple reflects the underlying structure of the similarity matrix ofthe taxa (itself dependent upon method). Differences in computa-tional methodology in effect account for variability in phenogramsand, ultimately, in preferring a single phenogram as "best." In phy-logenetic (cladistic) analysis, on the other hand, the same methodcan produce competing cladograms which are evaluated by a crite-rion (parsimony) not itself a consequence of the method but of hypo-thetico-deductive science in general. In contrast, in numerical t~x-onomy different methods of analysis produce alternativephenograms which are evaluated by a calculated value (the cophen-etic correlation coefficient) that is linked closely to the methodsthemselves.

Trees

Trees comprise another group of branching diagrams, typically spe-cifying, by one form or another, phylogenetic relationships among aset of taxa. Through the years different kinds of trees have beenproposed; unfortunately, workers have seldom indicated exactly howtheir trees should be interpreted, hence confusion has been preva-lent.

Trees commonly are constructed using three topographical f~a-tures: dots, lines, and branch points (figure 5.4). Dots, when used, in-variably represent identified taxa. lines, on the other hand, mayor

180 Biological Classification 5 4 3 2

8' Is 6\ f/5a. 7' {4

b.

3~ f21\7 6 5 4


named (say, with a dot), indicating an identified taxon ancestral tothe two or more descendant lineages, or unnamed, indicating a hy-pothetical speciation event with the ancestral species unspecified.In all trees figured in this chapter, taxa are represented by dots andtheir relationships by lines.

As many paleontologists have stressed for years, continuouslysampled stratigraphic sequences are a rarity; thus, in most cases,trees consist of lines purporting to show the hypothesized ancestral-descendant relationships among taxa perceived to be discrete. Withrespect to classification, the important questions are, first: how arenamed taxa of the tree to be subdivided and allocated to largergroups? And second: what categorical rank shall be given to thesegroups? We will first discuss the question of subdividing the tree.

Paleontologists since the time of Darwin have noted the arbitrarynature of subdividing a continuous lineage of evolving populationsfor classificatory purposes. The problem was important to Darwinand has remained so for succeeding generations of paleontologistsbecause it focuses on the origin and definition of species. How cansomething continuous-for example, a gradually evolving Iineage-be subdivided objectively? Of course, it cannot

Supposing Band C to be two species, and a third, A. to be found in anunderlying bed; even if A were strictly intermediate between Band C,it would simply be ranked as a third and distinct species, unless at thesame time it could be most closely connected with either one or bothforms by intermediate varieties unless we obtained numeroustransitional gradations, we should not recognize their relationship, andshould consequently be compelled to rank them all as distinct spe-cies. (Darwin 1859:297)

Darwin is suggesting here that a continuous lineage could notbe subd ivided. that, given continuity, only a single taxon is recogniz-able, and that gaps are necessary for subdivision. This belief in anarbitrary subdivision of evolutionary-phylogenetic trees persists.even though most systematists deal not with "continuous" lineages.but with discrete taxa connected by lines, lines that also branchAnd because trees branch two kinds of relationship are envisioned.

3

c.

F"Igure.S.4 !hree examples of evolutionary trees,~ranchlng dlagr~ms ouroornno to show the ancestor-descendant relationships am~mgspecies. In these treeraqrams.species are sometimes represented by linesas in Ia). or as dots as in (bl and (c). Trees in which '~aec~esare represented by dots (specified named taxa). y epicthypofhesized ancestors, as in (c). Trees arediscussed In detail in chapter 4.

may not represent named taxa (see chapter 4); in either case, linesare always used to signify a relationship of some kind (depending onhow relationship is defined by the systematist) among the identifiedtaxa. But lines per se may be used to symbolize a continuous ances-tral-descendant lineage of one or more identified taxa, continuous in~esense~at. be. a systematist believes a sequence of change candiscerned am . e or. ong a senes of specimens sampled from a marless uninterrupted stratigraphic record. Lines imply continuity; it maybe.a conceptual continuity between dots (named identified taxa) inwhich '" . 'f, intermediate" taxa are unknown or the line may signify can l-nUity (a lineage) among specimens assumed to be assignable to thesame taxon. Branch points represent a speciation event; they may be

among successive taxa in an ancestral-descendanllineage, andamong contemporaneous taxa of more or less distant common origin.In accordance with the usual coordinates of tree representation. theformer relationships are called vertical and the latter honzontal In


dealing with recent animals or with contemporaneous faunas of fossils,only horizontal relationships are directly involved. In temporalsequences of fossils vertical relationships are directly presented.(Simpson 1961:129; italics in original)

The dilemma for the paleontologist is resolving these two kindsof relationship within the framework of the Linnaean hierarchy:

Examination of any extensive tree or dendrogram at once reveals thatclassification by either vertical or horizontal relationships alone is ab-solutely impossible. . In translating the phylogeny into taxa a com-promise must somewhere be effected; some divisions among taxamust be horizontal and some vertical. Choice as to just how to effectthe compromise is part of the art of taxonomy, a matter of taste and in-genuity. (Simpson 1961:130)

How is the "art" of classification to be appl ied to trees so thatnested sets of taxa, and thus a Linnaean hierarchy, can be formed?Various criteria have been suggested (Simpson 1961; Mayr 1969): (1)Temporal sequences of taxa are rarely continuous-gaps exist (thatis, morphologically and temporally intermediate taxa are unknown),and these can serve as convenient, arbitrary boundaries. (2) Taxacan be clustered so that the sizes of groups of the same rank are notespecially disparate. (3) Taxa can be clustered so that morphologi-cal diversity within groups of the same rank is more or less the same,And (4) taxa that are morphologically distinct or are thought to haveevolved into a way of life significantly distinct from their sister-taxoncan be classified apart from that sister-taxon (this may pertain pri-marily to the issue of ranking, but frequently involves the procedureof Clustering). These are the main criteria used to subdivide trees assuggested by some well-known systematists. The first two apply totopographical aspects of trees, whereas the latter two call for addi-tional information not capable of expression within the logical struc-ture of a tree (e.q .. Simpson 1961 :113). The resolution of this lastproblem will be discussed in more detail shortly (see also the earlierdiscussion, pp. 171-75).

The question of how a tree is to be subdivided is thus answered:the procedure is unspecifiable and dependent on precedent andthe "arbitrary" decision of the investigator. Given a tree, it can besubdivided in numerous ways, resulting in many classifications. Ex-amples of some trees and their possible subdivision, at least at the


~-----------Ih .,,g ,, ,, ', ", ,~~'---------~...~! 0 b cd", ', ', ", ,\ ------ --""- --

- - --- .....-----:-----/r Q po, I J k I \/s n \ h }, , r\ ml I, , r-, ...!---- /

-s, -> c de - ''''"/ b \, f ', ,,0 ,, ,, ,

-, ->------- ---a b

,....----------- -----', ...h i j klmno\- ,-- ,,,,,

g~--../,,,,,,,- ,

,------- -,,---- ---- ------,'bcdefgh i)k I m n 0/, ,, ,, ,-:0 ........, ---I 1----...., ," ,, ,, ,

, t, ,, ,,~_....

'''-b-Z~-:~~f, 0,,,,----

c dFigure 5.5 The ways in which evolutionary trees are often subdivided forclassificatory purposes by evolutionary systematists (see text). (From Simp-son 1961:119 and 206.)

higher taxonomic levels, are given in figure 5.5 (all from Simpson1961). In these trees, internodes are not named, that is, ancestorsare not specified. One obvious consequence of following the criteriaof Simpson (1961) and Mayr (1969) for subdividing trees is the con-struction of not-A groups-sets of taxa clustered characteristically bythe possession of shared primitive characters (symplesiomorphies).The trees of figures 5.5a and b can be assumed to exemplify this sit-uation. In figure 5.5a taxa a-d are grouped together, the implicationbeing that they lack those characters defining the lineage of taxae-h. The subdivisions of figure 5.5b, illustrated by Simpson, imply asimiliar situation: taxa a-f are presumably morphologically primitiveto more advanced and well-defined groups g-l and m-s.

As noted in the introduction to this chapter, the history of clas-sification has been, to a great extent, the history of eliminating not-Asets, Although many not-A sets have been rejected and are no

h


longer in use, e.q.. Radiata or Vermes, others, such as Invertebrata,Pisces, or Reptilia, can still be found in some modern classifica-tions.

To some systematists the recognition of not-A sets is a matter oflittle or no concern, However, in accepting not-A sets the power, util-ity, and logical structure of the Linnaean hierarchy is sacrificed, ifnot destroyed altogether. As developed earlier, the Linnaean systemof classification can represent only a nested pattern of taxa. The rec-ognition of not-A sets assures that the nested sets of taxa within theclassification cannot be translated back to the original tree. The treeand classification are independent of one another, and the clas-sification therefore loses or misrepresents the information content ofthe tree.

What information content contained in trees might be useful forclassification purposes? First, trees (as defined in this book: chapter4) reflect the nested synapomorphy patterns of the taxa. Second,trees contain statements (hypotheses) about the identification of an-cestors.' In this regard, trees are appropriate only for the depictionof hypothesized ancestral-descendant relationships among species.There do not seem to be additional kinds of information that can beexpressed by the topographical structure of the tree itself (Nelson1973c). Some workers (e.g., Mayr 1969) have suggested that thelengths of the lines and the angles of the branches can express rela-tive degrees of divergence, but for present purposes it can be as-serted that these kinds of information are not organized hierarchi-cally within the context of a tree, and therefore are not applicable tothe issue of constructing Linnaean hierarchies from trees (the "solu-tion" of Mayr and other evolutionary systematists will be taken upshortly)

So, like cladoqrarns. the structure of trees manifests nested syn-apomorphy patterns, but unlike cladograms trees propose to make


statements about ancestors. How are these two aspects of trees tobe contained in Linnaean classifications?

Leaving aside for the moment the substantial difficulties, bothscientific and philosophical, of recognizing ancestral species(chapter 4), the conversion of trees to classifications can be consid-ered, Figure 5.6a depicts a tree in which ancestors are specified butnot at internodes, hence common ancestors are not identified. Thesynapomorphous relationships of this tree are easily expressed inthe form of a classification:

h d9 e 9

f c f

b

Q

bagfhbecd 9 f h

5 It is important to point out that the paleontological concept of trees just discussed is seldomas rigorously defined as in chapter 4. First, in traditional paleontological (or neontological)tree" it cannot always be assumed that construction is based on a nested pattern of synapomor-phies_ On the contrary. the tree itself may be a vague schematic representation of stratigraphicposition and general overall similarity _ Second, traditional trees often are unclear as to whetherJines are meant to represent actual species continua (i.e., the line itself is named) or somesymbolic conception of relationship. Here, we specifically consider taxa to be discreteunits, either terminal (descendants) or nonterminal (ancestral), and lines to be representationsof hypothesized phylogenetic relationships.

e d

c

b

Q

a e bed

Figure 5.6 The lop two diagrams are evolutionary trees in whichancestors are specified (dots). (In the left-hand tree, ancestors arenot at internodes, whereas in the right-hand tree they may be.) Thelower two diagrams are cladograms expressing the synapomor-phic information of the trees. (See text for details.)


TAXON: a.o.c.o.e.t.o.hTAXON: aTAXON. b.c.d.e.f.q.hTAXON: b.c.d.eTAXON: bTAXON: c.o.eTAXON: eTAXON: c,d

TAXON.I,g,hTAXON: I,gTAXON: h


mation content of the tree, However, this requires the introduction ofa convention (in this case, the use of parentheses) not traditionallyconsidered a component of Linnaean classification systems.

What of a tree in which common ancestors are recognized, as infigure 5.6b? The traditional classification might be as follows:

TAXON: a,b,c,d,e,l,g,hTAXON: aTAXON: o.c.c,eTAXON bTAXON: eTAXON: c.o

TAXON: I,g,hTAXON hTAXON: (g

Once again, ancestors are concealed within the traditional Lin-naean system, and the most such a classification can do is reflectthe structure of the cladogram of the tree (figure 5.6b, bottom dia-gram). However, as in the previous example a convention might beadopted to identify ancestors within the Linnaean hierarchy. In thiscase, let a parenthesis which encloses a taxon sharing equal rankwith two other taxa identify the common ancestor of those taxa, but ifthe taxon in parenthesis shares equal rank with only one other taxon,then the former is to be considered the ancestor of the latter:

TAXON: a,b,c,d,e,l,g,hTAXON: (a)TAXON b.c.o.eTAXON (b)TAXON e

TAXON: c.cTAXON: (c)TAXON: d

TAXON: I,g,hTAXON: hTAXON: I,gTAXON: (I)TAXON: g

However, if it is desired to convert this classification back intothe tree fromwhich it is derived, i.e.. if we want to express the infor-mation content of the classification, the classification-tree rela-tionship is seen to be asymmetrical.What is obtained in this processis the cladogram of the original tree (also shown in figure 5.6a, bot-tom). In this classification, species c is classified with species dbecause the two shareone or more synapomorphies not shared withspecies e; but within the context of a traditional Linnaean hierarchy itis not possible to discern whether c is the ancestor of d, d of c, orwhether neither is an ancestor. The important point, of course, is thatthe specified ancestor, species c. is classified with its descendant,species d, because both possess an evolutionary novelty settingapart their lineage from all others of the tree. .

At this point, a convention might be agreed upon and In-troduced whereby ancestors could be identified within the frame-work of the Linnaean hierarchy. For example, let namesof ancestorsbe placed in parentheses:

TAXON. a.b.c.o.s.to.nTAXON: (a)TAXON: b.c.d.e.f.q.hTAXON: b.c.o.eTAXON: (b)TAXON: c,d.e.TAXON eTAXON (c).o

TAXON. t.q.hTAXON: h

TAXON {f),g

The original tree, figure 5.6a, can be faithfully reproduced from thisclassification; consequently the latter accurately represents the infer-

There are several generalizations contained in these two ex-amples. For any tree, those aspects of its structure that can be ex-pressed by a Linnaean classification are precisely those aspectsexpressed in the corresponding cladogram, i.e., the nested patternof synapomorphy exhibited by the taxa. Within traditional Linnaean

•188 Biological Classification

classification schemes, concepts of ancestry and descent cannot bestored or retrieved. Such concepts cannot by themselves be treatedhierarchically, but with the addition of a variety of conventions, an-cestors can be notated within a Linnaean classification. One mightconclude, then, that a cladogram is a more basic, more generalbranching diagram than a tree. The hierarchical structure of a clado-gram is identical to that of a Linnaean classification system; the twoare isomorphic.

All of the above assumes, of course, that ancestors can be iden-tified or specified. for without this assumption it would seem prefera-ble to base classification on a cladogram. We have already noted(chapter 4) the difficulty inherent in testing hypotheses of ancestryand descent

Up to this point the concept of trees being discussed is thatmost often identified with paleontological systematists (perhaps bestcharacterized by the writings of G. G. Simpson) with their concernfor "horizontal" and "vertical" relationships. The trees themselvesgenerally attempt to depict an accurate representation of thesequence of phylogenetic branching The classifications derivedfrom these trees do not necessarily, as Simpson admits, reflect thisbranching sequence in a precise manner. Brundin (1966) and Nel-son (1972b) have correctly noted that "horizontal" relationships donot exist as a separate category of kinship relationships and thatwhat Simpson refers to are instead "similarity" relationships. Simp-son ian classifications tend to emphasize "vertical" relationships-that is, genealogical or cladistic-but deviate from them when sub-jective assessments of "similarity" relationships are desired.

Another group of systematists-neontological evolutionary sys-tematists-differ in that their concept of relationship emphasizes"similarity relationships" rather than kinship relationships (see Nel-son 1972b, for an extended discussion)

It is no longer legitimate to express relationship in terms of genealogyThe amount of genetic similarity now becomes the dominant consider-ation for the biologist

When an evolutionary taxonomist speaks of the relationship ofvarious taxa. he is quite right in thinking in terms of genetic similarity,rather than in terms of genealogy, (Mayr 1969:79)

This concept of "genetic similarity" has never been defined pre-cisely by the evolutionary systematists and for obvious reasons-the


genetic systems of the taxa are never the subject of investigationRather, "genetic" similarity is estimated. and the process of estima-tion is based on "weighted" morphological similarity: "weighting,then, can be defined as a method for determining the phyletic infor-mation content of a character" (Mayr 1969:218; italics in original).Mayr (1969:220-21) lists a number of criteria used to determine thetaxonomic "weight" of characters: complexity, joint possession ofderived characters, constancy through large groups, characters notassociated with an "ad hoc" special izaticn. characters not affectedby "ecological shifts," and correlated suites of characters, Most im-portantly for this discussion, one is never told precisely how thesecriteria are used to construct a tree; as with the statement by Simp-son noted earlier, subjective "artistic" interpretations are considereda normal aspect of taxonomic procedure (some of the problems ofweighting characters in cladistic studies were discussed earlier; seepp. 66-£7.

This methodology produces an ambiguous concept of a tree: "Inan orthodox phyletic diagram [see figure 5.7] three kinds of informa-t~on are conve~ed, degree of difference in the abcissa, geologicaltime In t~~ ordinate, and degree of divergence by the angle of di-vergence (Mayr 1969:255-56). Actually, some other kind of informa-

• , c o r r

,

<.~~,~. of 4;(I .. on<o )

~igure 5.7 Mayr's conception 01phylogeny (an evolu-ttonary tree With terminal taxa) as expressed in a branch-~~gdiagram (see text). (From Mayr 1969:256, Copyright ©69, and used by permission of McGraw-Hili Book Compaoy.) -

---------


tion must also be contained in the tree, for what, after all, is meant bythe branching sequence? Because the methods of evolutionary sys-tematists are designed to estimate "genetic" (= "weighted morpho-logical") similarity and not cladistic relationships, it cannot be as-sumed the branching sequence conveys information aboutphylogenetic relationships, In this regard, "Mayrtan'' trees (Nelson1972b) can be viewed as intrinsically similar to phenograms; thebranching sequences attempt to reflect relative degrees of similarity.

All of this creates fundamental problems for converting"Mayrian" trees into classifications, for although it would be possibleto mirror the branching sequence of such a tree in a Unnaean clas-sification, only nested sets of taxa comprise the information contentof those classifications, not a measure of similarity. Furthermore,Mayrian classifications do not mirror Mayrian trees because othercriteria, such as those advocated by Simpson, are employed to sub-divide the tree; consequently, not-A sets are formed.

In summary, the important aspect of trees that has trad itionallyattracted biologists is the specification of ancestral taxa. But if thisinformation is to be contained within Linnaean classifications, thensome ad hoc conventions must be adopted in order to identify an-cestral taxa within the hierarchy. Only then is it possible to specifyfully the information content of the classification and reconstruct,precisely, the original branching diagram. Only through this conven-tion can the classification and branching diagram be isomorphic.

If this were the only problem with using evolutionary trees forclassification, then the major issue would be whether the tree con-cept is amenable to scientific analysis. But neontological and pa-leontological evolutionary systematists go far beyond the question ofaccommodating the isomorphy between trees and classificationsand propose a set of procedures for subdividing trees that in fact de-stroys the isomorphy itself. What these workers perceive as thestrength of their approach-greater flexibility, more artistic and sub-jective control Over the process of classification-is actually itsgreatest weakness. for by destroying the isomorphic relationship be-tween trees and classifications. they destroy the information contentof both, that information is, of course, the nested sets of taxa. More-over, this approach also counteracts 200 years of systematic aspira-tions: the elimination of not-A sets.


The Question of Ranking: A Critique of ClassicalEvolutionary Classification

Linnaean hierarchies are so structured that named taxa are nestedby means of categorical ranks such as order, family, genus, andspecies, to name only a few. Whereas all species must have a genusand species rank (thus the species taxon name is a binomial), it isnot a requirement of the International Code of Zoologica! Nomencla-ture that a species be included in a suprageneric category; it is con-ceivable, for example, that a species may be considered of uncer-tain status and therefore not be classified as to family or order, Whatthis means, of course, is that the relationships of the species are un-certain, its set-membership within a cladogram not having been re-solved satisfactorily, But most species are classified within a family,order, cress. etc., and it is the question of how sets of species are tobe assigned a rank that has engendered much of the vehementargumentation in the systematic literature. The crux of the problem isthe extent to which ranking, and thus the Linnaean hierarchy itself, isto be used to depict accurately the set-membership of the taxa in abranching diagram.

In an earlier section, it was seen how ranking can be utilized todepict set-membership, and it has been argued throughout thischapter that the logical structure of both Linnaean hierarchies andcladograms are compatible with one another only when the set-membership between the two is isomorphic. Within systematics,then, the question of ranking has been debated primarily from twodifferent viewpoints, First, should we require that the ranking systemof Linnaean classifications be used to depict precisely the set-mem-bership of a cladogram, or should that set-membership bedisregarded in the name of "equally important biological consider-ations"? Second, if in fact the Linnaean ranking system should de-pict faithfUlly the set-membership of cladograms, is it actually possi-ble to classify the known diversity of organisms in this manner? Thissection will be organized around the debate over the first 01 th~setwo questions; the second question will be examined in the sectionimmediately following. This chapter has been an extended discus-sion supporting the notion that Linnaean classifications should bestrictly isomorphic with the branching diagrams from which they are

-

r ---


derived, If this is not done, then the information content of the class-ification is sacrificed to a greater or lesser extent. In contrast to thisview, there are systematists who advocate procedures of rankingthat do not lead to isomorphic set-membership of taxa betweenthe classification and branching diagram. It seems necessary,therefore, to consider their position in some detail

The process of classification, as Mayr (1974:113) has noted,"consists essentially of two steps, 1. the grouping of lower taxa(usually species) into higher taxa, and 2. the assignment of thesetaxa to the proper categories in the taxonomic hierarchy (ranking)"(italics in original).

The methods for grouping taxa were discussed in the chapter oncladogram analysis and are not of primary concern here; indeed, inrecent years evolutionary systematists have shown a tendency to ac-cept the use of synapomorphy to cluster sets of species (see chapter1). However, philosophical differences over acceptable approachesto ranking have consistently led to conflict.

Evolutionary systematists reject the idea of an isomorphic rela-tionship between the set-membership of branching diagrams (phy-logenetic hypotheses, to them) and classifications. This viewpointfollows from their belief that classifications should be constructed soas to reflect a variety of information about the included organisms,and not just the set-membership of the taxa as expressed in thecladogram or phylogenetic diagram For example,

Organisms are classified and ranked, according to this theory [of theevolutionary systematists], on the basis of two sets of factors, 1. phy-logenetic branching ("recency of common descent," retrospectivelydefined), and 2. amount and nature of evolutionary change betweenbranching points. The latter factor, in turn, depends on the evolu-tionary history of the respective branch. e.o.. whether or not it has en-tered a new adaptive zone and to what extent it has experienced ~major radiation, The evolutionary taxonomist attempts to maximize SI-

muttaneousfv in his classification the information content of both typesof variables (1 and 2 above), (Mayr 1974:95)

.,. tonIf classification is based on evolutionary theory, a natural ctassr rcamust be in agreement with evolutionary theory and with the whole ofevolutionary theory, including all laws mechanisms of change andsubfactors thereof, By the whole of evotctionarv theory, I mean all f~c-tors and mechanisms. In particular, I mean all studies of function, bio-logical role, behavior and environmental factors required to under-. msstand the evolutionary mechanisms. And I mean all mechame


including that 01phyletic evolution, especially antactors of the forma-tion of genetical mechanisms and of natural selections [sic] under-lying evolution in single phyletic lineages, Speciation is not the onlyevolutionary mechanism of importance to classification contrary tostatements of many phylogenetic systematists. (Bock 1977:864)

Evolutionary classification is a system 01taxa arranged in a Linnaeanhierarchy. The taxa must be monophyletic in the sense of Simpson andtheir rank reflects the attempt to maximize simultaneously the twosemi-independent variables of amount of phyletic change as reflectedin the degree of similarity and the phylogenetic sequence of events asreflected in the pattern of phylogenetic branching and ultimately thepattern of speciations. No one-to-ore relationship or conetetton existsbetween the eVolutionary classification and the phylogenetic diagramof the groups contained in the classification (Bock 1977:869: italicsadded)

Evolutionary classification takes into account: degrees of homologousresemblance in all available aspects: the most probable phylogeneticinferences from all data (including the foregoing resemblance plusevolutionary analysis and weighting 01 the various characteristics):and also the practical needs 01discussion and communication (Simp-son 1963:25: italics in original)

The criteria used by evolutionary systematists to determine theranks of taxa are not absolute, nor can they necessari ly be appl iedobjectively in any specific case. Classifications, in their view, mustbe approached as "an art with canons of taste, of moderation. and ofusefulness" (Simpson 1961:227) But criteria for assigning rank havebeen suggested, Mayr (1969:233). for example. lists five: (1) d istinct-ness, or size of gap between two or more sets of species, (2) evolu-tionary role, or uniqueness of the adaptive zone occupied by the setof species under consideration, particularly compared to that ofclose genealogical relatives: (3) degree of difference between setsof species, (4) the size of the taxon, and (5) the equivalence of rank-ing in genealogically related taxa. To this list might be added Simp-son's (1963:27) call for "linguistic convenience" as a basis for es-tablishing rank,

The scientific merit of this approach to the question of rankingmust be judged in terms of their views of the functions of biologicalclassification. Merit cannot be decided in terms of the isomorphicrelationship between a branching diagram (phylogenetic hypothe-sis) and the classification, for evolutionary systematists haveadopted, by definition, the position that this relationship in fact


frequently obscures functions of classification more important thanconveying information about genealogy. An analysis of their viewsmust focus on their own claims regarding first, the purpose of biolog-ical classification; second, the ways in which their procedures ofranking allow them to fulfill that purpose; and third, whether or notthese methods accomplish the purpose for which they are intended.

Mayr (1969:229), like most evolutionary systematists, views clas-sification as a system for the storage and retrieval of information, andin their system information is to be evolutionary in content: "The com-plex relationship of species with each other and the varying rates ofbranching and divergence must be translated into a system of taxa,conveniently ranked in the appropriate categories, in such a way thatit forms as efficient as possible a system of information storage."Thus, to Mayr, a classification must meet two objectives: groupingclose relatives (determined on the basis of weighted overall similar-ity and the five criteria fisted above) and facilitating information re-trteval.s

Evolutionary systematists assert that preferred classificationsare those which are the most useful, have higher empirical content,have greater capability for predictions (or lead to broader general-izations), and which best form the basis for comparative studies. Asa general view of classification, one could hardly take issue withsuch idealism.

The goal of evolutionary systematics is to construct classifica-tions that go beyond the mere expression of set-membership ex-plicit in the branching diagram. An evaluation of the evolutionaryhistory must be included (see Bock's comments above), and this isaccompl ished by criteria such as those of Mayr, Iisted earl ier. Whatare the characteristics of such classifications and how might the6. Michener (1978) also views classifications as information storage-retrieval systems. Bock(1973) talks about maximizing the information content in classifications of two "semiindepen-dent variables": degree of genetical similarity, which is said to be correlated with phenetic sim-ilarity ("greater phenotypical similarity implies greater genetical similarity," p- 377), and thebranching sequence of a phylogenetic diagram. Recently, Bock {I 977:865) deperted somewhatfrom this more traditional viewpoint: "Many workers believe that the information is actuallystored in the classification. This is not true, nor is the classification, itself, an informationstorage-retrieval system. Rather the classification serves as the foundation on which an efficientinformation storage and retrieval system can be constructed, be it a book, the arrangement ofspecimens in a museum, a computer system or Whatever." The view that classifications do norstore information, and that one cannot retrieve something from {hem, will be rejected in sub-sequent discussions. Indeed, throughout history, classifications-whether biological or nor-,have universally been thought to store and express a special kind of information, namely, set-membership.

ilI,!J

1,I

'j

Pan Homo./

Australopithecus-:Ramapithecus


methodological criteria of evolutionary systematists be expresse.d inthe actual process of constructing a classification? These questionswill be examined by some examples,

PangoHylobates

Pliopithecus

Oreopithecus

.>Apidium

Figure 5.8 Simpson's view of hominoid. phylogeny as of 1963. (After Simp-son 1963:27, figure 6. Viking Fund Publications tn Anthropology, n~, 37.Copyright © 1963 by the wenner-Oren Foundation for Anthropoloatcat Re-search, Inc., New York).

I. Tbe bominoid classification of Simpson (1963) Figure 5.8 attemp~st~summarize Simpson's 1963 view of the phylogeny of the Hommo.l-dea. The purpose here is to examine the consequences of hi:method of ranking, and not evaluate the validity of his taxa or theirrelationships. Some genera (e.q. Kenyapithecus, Proconsul) areomitted because it is uncertain, either from the figure or the accom-panying text, precisely how Simpson would relate .themto ~t~er ~e~-era. For nine genera Simpson provided the followmg classtfication:

SUPERFAMILYHominoideaFAMILYPongidaeSUBFAMILYHylobatinae

PliopithecusHylobates

r - .-~-- ...


SUBFAMILY DryopithecinaeRamapithecus

SUBFAMILY Pong inaePongoPan

FAMILYHominidaeAustralopithecusHomo

FAMILYOreopilhecidaeApidiumOreopithecus

The Apidium and Oreopithecus lineage was ranked as a family"because of Its ancient separation plus its marked divergence fromany other group now usually given family rank" (1963:26). Hylobatesand Pffopithecus are classified with Pongo and Pan because all ofthem

almost certainly had a common hominoid ancestry. . and its [Hylo-bates lineage] evolutionary divergence from [Pan and Pongo] is de-cidedly less than that of either Homo or Oreopithecus. Thai would jus-tify placing the Hylobates group as a subfamily of a family alsocontaining Pongo, Pan, and some, at least, of the dryopithecine com-plex. Both arrangements are consistent with reasonable interpretationsof the available data, and choice [between placing Hylobates in aseparate family or with Pan and Pongo] becomes a matter of personaljudgment and convenience. I continue to prefer the second alternative,partly as a matter of linguistic convenience. (pp. 26-27)

Although Simpson clearly accepts the fact that Pan and Homoare more closely related genealogically than either is to Pongo, heclassifies Pan with Pongo and elevates Homo to separate familialrank because (a) "Pan is the terminus of a conservative lineage" and"Homo is both anatomically and adaptively the most radically dis-tinctive of all hominoids," (b) "Pan is obviously not ancestral toHomo," and (c) Homo has diverged until it deserves family status,and if Pan were placed in the Hominidae, "carrying this processdown will eventually require inclusion of all descendants of earl iersplittings also in the latest family----eventually the whole animal king-dom would be in the Hominidae on this principle."?

7. It seems useful at this point to reject this last argument. Following the logic of Simpson, wecould pick ally tenninal group te.g., any Recent species) and then classify all the animal king-dom in the. f~i1y. of that species. More important, of course, Simpson is in error (see also1961.165) In VIewIng phylogeny, and therefore classification, as proceeding from top to bot-

I,I

• • ace S U·£4-&\lGWJ


Simpson's approach to hominoid classification has been ac-cepted by other evolutionary systematists, notably Mayr (1969:70):"To rank taxa according to branching points is nearly always mis-leading. It might necessitate, for instance, the inclusion of the Afri-can apes (Pan) in the family Hominidae and their exclusion from thefamily Pongidae."

2. The galliform classification of Mayr (1974:119) Basing his discus-sion on figure 5.9, Mayr argues: "There are three major families of liv-

Phasianidae Cracidae Megapodiidae

Figure 5.9 A hypothesis of relationshipsfor three families of gallifarm birds.

tom, rather than vice versa {see Wiley 1978, 1979, for an extended discussion of this problem}.Such views stem from the notion that it is species that evolve, not genera or taxa of higher cate-gorical rank-a concept with which we heartily agree. But it does 1101 follow that, for anymonophyletic group, an ancestral species appears first, followed at some later time by genera,then families, and so forth-the view that Simpson and Mayr (quoted below) seem to support.As Wiley (1978, 1979) has pointed OUl, the age of a taxon depends upon the time of first ap-pearance of the synapomorphy defining the set. The first member of the group will indeed be anancestral species but, to be included in the set, must have at least one synapomorphy linking itto its descendants. That one {or more) synapomorphy will likewise define the higher, moreinclusive set. Thus any taxon of higher categorical rank, because it is a set defined by one ormore svnapomorpmes, "originates" at exactly the same time as its earliest included species.The confusion arises from the mistaken belief that supraspecinc taxa are real entities in naturethat somehow "evolve. '" As Wiley (1978:21) has noted: "There is no doubt that one can runfrom man 10 protist in one classificatory taxon, but. in my opinion, that taxon would beEucaryora. not species Homo sapiens, There was a genus HomQ before there was a speciesHomo sapiens, just as there was a class Vertebrata before any of the Recenr vertebratesevolved. Thus, we tie together increasingly ever larger taxa on the basis of the continuum theyare hypothesized to have shared in the past, and if we adopt a truly natural classification, thisclassification will document past continua, not bury their reality or existence ."

We note also the statement of Mayr (1974: 105) that phylogenetic systematists are guilty ofa "fatal flaw" in reverting to Aristotle's "downward" classification rather than the "upward"classification supposedly characteristic of modem Linnaean systems. From the foregoing, it isreadily apparent that this statement is incorrect: the ancestral vertebrate species, after all, ex-isted before the ancestral primate species, and it in tum existed before Homo sapiens.


ing gallinaceous birds. Among these the MegapodHdae have thegreatest number of primitive characters while the Phasianidae havethe greatest number of derived characters. The South American Cra-cidae are intermediate. They share a few derived characters with theadvanced Phasianldae but a far greater number of primitive charac-ters with the Megapodiidae." On this basis, then, Mayr supports aclassificatory arrangement such as the following:

ORDERGalliformesSUBORDERCracoideaFAMILYMegapodiidaeFAMILYCracidae

SUBORDERPhasianoideaFAMILYPhasianidae


"0--~ 00 1<" 0 20 -c .~ 0 ·0~ .20 E .2 0

~ g - 0- .~ 0 0 .!0 .0 ~ '0 ·c ! ·00 ~ - 0,; • 0 - - ~E " 1;; 0 ·0 2 • 0 -c 0 s 2 ~0 c 0 c e .0. 0; -c • 0 ~ 0- - .i ~~ 00 0;

~ .ll 3 ~" w ~ is " " ~ UJ« o :I:

In this classification Mayr would unite cracids and megapodiidson simple overall similarity, without applying, it would seem,"weighted" phenetic similarity.

3. The vertebrate classification of Romer (1962, 1966), and its advocacyby Bock (1977). Based on his studies of gill arches and other data,Nelson (1969) proposed the following genealogical classification ofthe Recent vertebrates using a cladogram similar to that depicted infigure 5.10 (Nelson's classification includes many other groups, in-cluding fossils, which are omitted here for simplicity):

SUBPHYLUMVertebrataSUPERCLASSCyclostomataSUPERCLASSGnathostomataCLASSElasmobranchiomorphiCLASSTeleostomiSUBCLASSActinopterygiiINFRACLASSObonorostetINFRACLASSNeopterygiiDIVISiONHolosteiDIVISIONTeteostei

SUBCLASSSarcopterygiiINFRACLASSDipnoiINFRACLASSChoanataDIVISIONBatrachomorphaDIVISIONReptilomorphaCOHORTSauropsidaSUPEAORDERCheloniaSUPERORDERArchosauria

Figure 5.10 A cladistic hypothesis of the vertebrates. (After Nelson1969.)

SERIESCrocodiliaSERIESAves

COHORTMammaliaSUPERORDERProtother!aSUPEROROERTheriaSERIESMetatheriaSERIESEutheria

Nelson's classification was critized by Bock (1977:868-69; see belowfor comments), who advocated adoption of the evolutionary clas-sification of A. S. Romer (see, for example, 1962, 1966). This clas-


sification is as follows (where h .taxon, the name in parenth . t .erehare different names for the same

ears IS t at of Nelson):SUBPHYLUM VertebrataSUPERCLASS PiscesCLASS Agnatha (= Cyclostomata)CLASS Chondrichthyes {> EI 'CLASSOsteichthyes(_ T- I asmobranchlomorphaj

S - e eostorru)UBCLASSActinopterygiiINFRAClASS ChondrosleiINFRACLASSHolosteiINFRACLASSTeleostet

SUBCLASS SarcopterygijORDER Dipnoi

Su~ERCLASSTetrapOda (= Choanata)c~::~;~~:~ia (= Batrachomorpha)

SuBCLASS AnapsidaORDER Chelonia

SUBCLASS LepidosauriaSUBCLASSArchosauria

ORDER CroCOdiliaCLASS AvesCLASS MammaliaSUBCLASS ProtothensSUBCLASS TheriaINFRACLASSMetatheriaJNFRACLASS Eutheria

:rhe general patternof bran .details depicted' N chinq, and even probably most of the

In elson's cl dmostvertebratezoolog. t. a ogram, have been accepted by1966) ISs, mciuo: R.." The classification of ng .omer (see especially, Romer,manly in the etevatton r Romer differs from that of Nelson pti-e d n m rank ofre to have diverged . '. a number of taxa that are consid-sumedancestral condif significantly in morphology from the pre-a d I ron, e gn c ass Mammalia . " superclass Tetrapoda, class Aves,

Oncea' .gam,we see this latt .systematists' evaluatio f er classification reflects evolutionaryamong taxa and the Sin ?f' degrees of morphological differencePre qm rcanca of th d .sumed to haveoc e a aptive changes that are. hi curred wittu ,WIt m the classificatory hi I In certain lineages, to decide rank

hav b rerarcny Manv aoct .e een presented b t . any a ditional examples couldcons~~uently,these thre~ c:one would have differed in principle;classificatory methods d n be used to evaluate the qoals and

a vacated by I·eva uncnary systematists.


In analyzing the scientific merit of evolutionary classification, itis well to keep in mind several claims of its advocates. First, theybelieve classifications are designed to contain informationof severaldiverse kinds, and that this information is manifested in the structureof classification; in other words, such information can be retrieved,Certainly they believe information should, in principle, be retrievablefrom classifications. Second, they believe evolutionary classifica-tions are the best general reference system for biology because theytacit itate a greater number of testable predictions and lead to moregeneralizations than alternative classificatory systems. The thirdclaim is that the information being stored and retrieved includes ge-nealogy, morphological-adaptive divergence, genetic similarity, andvarious other evolutionary parameters,and that such information is tobe classified by judicious decisions in assigning rank,

It can be argued that the above three beliefs lead to a paradox,incapable of solution within the framework of their ideas. On the onehand, they assert that a diversity of evolutionary information must beused to construct classification, and their writings imply they believesuch information is "stored" in those classifications. On the otherhand, they fail to provide any rules or criteria for retrieving that infor-mation from their classifications, and they obviously cannot. be-cause it would seem to be impossible. Earlier sections of thischapter demonstrated that information, to the extent it can be storedand retrieved from Linnaean hierarchies, must be expressed in termsof nested set-membership. Evolutionary systematists dream aboutthe virtues of all possible evolutionary information (see remarks byBock, 1977:864, cited above) being included within the structure ofeach evolutionary classification. It simply cannot be done. To besure, set-membership can depict general notions of similarity rela-tionships among species-as numerical taxonomy has demon-strated-but it has not yet been shown how the property of set-mem-bership can convey, simultaneously, as evolutionary systematistsdemand, knowledge of genealogy, genetic similarity, adaptive di-vergence, and whatever else it is they may want to include in theirclassifications.

Evolutionary systematists place considerable importance on thecapability of a classification for prediction and for leading to test-able generalizations. Not unexpectedly, evolutionary systematistsbelieve their classifications are superior to other approaches in thisregard:

2 2 2


I claim that using the criterion f h'tent or of restabut . 0 a, .Igher degree of empirical con-dation for all biOIO~'i;Vlolutlonar~ctasstticaton provides the best faun-best classification' th compansons. Second. that this criterion for theevolutionaryClaSSi/i~ali~n:os~hessentialdistinction be~een classicaltics on the other. (Bock 197;:38~)one hand and ohenettcs and cladis-

Howare we to evaluate h la!themselvessetd .s~c a c aim? Evolutionary systematists

om seem willing to eng . dipredictive powersof cl T' age In a latoque about theto state their claim Th ass I rcatton Rather, they are content merelyterms of precisely ~h ~ .co~c:Pt of prediction can be examined insifications in general a ISb,elng predicted and what renders etas-

. capa e of predictionEvolutionary systematists . .

tion in terms of g I ' apparently universally, view predic-absolute numbers e~er~ .~v~r.a/l (phenetic) similarity, that is, in the

o smu antres that might be expected:Thecommongenetic pro rural taxon guarantees .~ am characteristic for the members of a nat-this taxon sharecertajnW~ha hrgh p.robability that all the members ofthrush 1 can make . aractenstlcs, If I Identify an individual as ah " preCisestatements "

P YSlology,and reproduct· 1M concerning Its skeleton, heart,Ion. ayr 1969:79-80)

Thebest classification is basedacters in organismswh' h h on the greatest concurrence of char-at unknowncharacters I'CB'ko

Uld permit the most accurate prediction. oc 1973:379J

Thus, the evolutiona s .and Pongo in one fa '1 ryh ystem~tlst would argue that placing Panily, the Hominidae rru y, / e Ponqidaa, and Homo in a separate fam-characters than ,.,p'a wou d allow for more predictions of unknown

n were cl T d .it is m~st closely related. Pa:

SSIIe with Homo, th.e genus to ~hich

more srmilar to Pongo th ' they would argue, IS phenotypicallytures should show th Poan to Homo, hence newly discovered tea-concept of predictio e, ngo,-Pan pattern more often than not. ThisIndeed, the method~1 rs pr~clselY that of the numerical taxonomists.evolutionary systemat~~: i~ the .'a.t1erseems superior to that of theIronical, then that I' Providing for this type of prediction. It isand methods 'that c eV,odutlon~ry systematists reject the philosophyd ou provide the ki d fa VOcate.The solution t tho n, 0 prediction they seem toclassifications should ~e IS p.rob/em IS to modify the concept thatRather, classifications . bdeslqned .to predict overall similarity.nrra, should be cap b! In emq consistent with their logical struc-

a e of predicting the nested pattern of strrulan-


ties observed among taxa. Viewed in this way, classifications pro-duced by numerical taxonomy and evolutionary systematics are notmaximally efficient predictors of unknown characters (Platnick1978b).

The information content of classifications is in their nested set-membership. Classifications state that the universal set of all spe-cies under consideration can be subdivided and allocated to sub-sets in particular ways. Whereas one can utilize a variety of informa-tion to form those subsets, including random draw if one were soinclined, the only information to be retrieved directly from a clas-sification is the pattern of nested sets and their membership.

How, then, can classifications predict? The pattern of nestedsets provides the answer, because there is the expectation that pat-terns of similarities are themselves nested. This expectation seemsan integral part of all taxonomic philosophies, but only the phyloge-netic systematists explicitly attempt to partition the kinds of similarityinto nested sets. The classifications of evolutionary systematists donot consistently partition sets of taxa according to the nested pat-terns of similarities, and indeed their procedure of ranking is theprimary reason why such classifications are not efficient predictors.To illustrate this point, the three examples can be considered.

In order to better understand the nested pattern of taxa ex-pressed in the hominoid classification of Simpson, the galliformclassification of Mayr, and the vertebrate classification of Romer(and Bock), the set-membership of these classifications can be "re-trieved" and expressed in the form of branching diagrams (figure5.11). None of these diagrams should be interpreted as a phylogeny,but they are equivalent to the classifications in set-membership, andso are equivalent to the latter in terms of their predictive capability,

Upon comparing Simpson's phylogeny of the hominoids (figure5.8) with the branching-diagram-equivalent of his classification(figure 5,11a), some striking differences are apparent; namely, theclassification is incapable of predicting: (1) that Apidium and Oreo-pithecus lack certain similarities shared by all other genera; (2) thatRamapithecus shares certain similarities with Australopithecus andHomo; (3) that Ramapithecus, Austra/opithecus, and Homo sharecertain similarities with Pan; and (4) that Ramapithecus, Australo-pithecus, Homo, and Pan share certain similarities with Pongo.

This would appear to be a rather extensive list of similarities

�u j0

� ~ ~ ~'- E- .g~ I ~e 0 - .g

~ ~ - ::: .!! '- " .2 6 ~.2

.~ .~ 0 g~

"8- .~ Il " e i ~ ·2 0 .., ~ .g ~ ~.g

:l! :~ 1 - "~

:l! j'§ ~ '0 -e - :c 0~ -

0- d ~ ~ ~ S0 s 0 g- O Ji! ~

'" it t ~-2 .!! 0 0 0- ! ~ -!a ~ ;! .2- " '" e 0 s s ~ .;ja o " " " "

o. b.

c

FIgure 5.11 Branching diagrams de tct!p 109 the set-membership expressed by the evolutionary classifications proposed by (a) Simpson, (b) Mayr, and (e) Romer and Bock (see text for details).


that the evolutionary classification of Simpson does not predict. In-terestingly, these are quite special similarities. They are derivedsimilarities that define four lineages within the phylogeny. If the setmembership of the original branching diagram (phylogeny) wereisomorphic with a classification, then the existence of these four setsof similarities would have been predicted. For example, the follow-ing classification preserves the predictions of those four sets of simi-larities (taxon names and absolute rank should be ignored; it is theset-membership which is important):

SUPERFAMilYHominoideaFAMilYOreopithecidae

ApidiumOreopithecus

FAMilYHominidaeSUBFAMILYHylobatinae

PliopithecusHylobates

SUBFAMilYHomininaeINFRAFAMllYPongiPengo

INFRAFAMllYHominiTRIBEPaniiPan

TRIBEHominiiSUBTRIBERamapithecininii

RamapithecusSUBTRIBEAustralopithecininii

AuslralopithecusHomo


tion. The difference is that the evolutionary classification makessome of its predictions at the incorrect level of the hierarchy, andthis error is derived directly from the ranking procedures of evolu-tionary systematics. Some errors involving actual morphologicalsimilarities will be pointed out below in conjunction with the discus-sion of Romer's vertebrate classification.

The above arguments also apply to Mayr's classification of thegalliform birds. His classification, and its derived branching dia-gram (figure 5.11b), does not predict the expectation of a set ofsimilarities common to the Cracidae and Phasianidae and notshared with the Megapodiidae. However, a phylogenetic classifica-tion such as the following:

OROERGall iformesSUBORDERMegapodiFAMilYMegapodiidae

SUBORDERPhasianiFAMilYCracidaeFAMILYPhasianidae

would make this prediction in addition to predicting that certain fea-tures will be shared at the ordinal level, and these are precisely thefeatures that cracids and megapodiids do share. The phylogeneticclassification cannot predict how different the Phasianidae might befrom the other families, but neither can the evolutionary classifica-tion.

Finally, a comparison can be made between the vertebrate clas-sification of Nelson, on the one hand, and that of Romer (along withBock's advocacy) on the other. In this example, the predictive capa-bilities of each approach become particularly clear.

Without providing any support for his assertions, Bock(1977:869) severely criticized Nelson's classification:

Although Nelson's classification contains at one rank or another, thegroups of Romer's classification, a quick comparison of the two clas-sifications shows that the classical system advocated by Romer IS farsuperior to that produced by Nelson in the number of possible gener-alizations and hypotheses that may be generated from each. Indeedsome of the groups in Nelson's system are almost devoid of usefulgeneralizations.

It can be suggested that what is needed to evaluate the merits ofthese two classifications is not a quick comparison, devoid of anyempirical support, but a more temperate consideration of some po-

An evolutionary systematist might well respond: "But such aclassification as this could not predict that Pan and Pongo have~any simila.rities not shared with Homo. The evolutionary classifica-tion does this, and is superior in this respect."

In f~ct, this counterargument is not entirely true. What the phy-logenetic classification shown above cannot do is predict how dif-f~rent Homo may be from Pan and Pongo. But neither can the evolu-~Ionary classification. Furthermore, the phylogenetic classificationIndeed predicts that Pan and Pongo share a set of similarities thisprediction being made for the set of taxa included under the' sub-family rank, tile Homininae. In fact, probably all of the similaritiest~~t c~n be expected for Pan and Pongo under the evolutionary clas-aitication can also be expected under the phylogenetic classifice-


tential empirical conflicts in the predictions of the two classifica-tions. For example, keeping in mind the phylogenetic hypothesis onwhich both classifications have their foundation (figure 5.10), we findthat the evolutionary classification of Romer and Bock lacks the ca-pability to make the following predictions:


predicted for Romer's Pisces are included in the similarities pre-dicted by Nelson's Vertebrata. Romer's Pisces is an artificial grabbag set of taxa formed because they share some primitive features(and probably few at that; most likely it is their "fishiness" thatprompted Romer to separate them from the land vertebrates); thosesame features, if they can be recognized, would be derived at thelevel of the Vertebrata. The second point can be answered similarly:Romer's Osteichythyes is another grab bag, and those featuresshared by the dipnoans and actinopterygians are primitive and pre-dicted by the taxon Teleostorni in Nelson's classification. The thirdpoint can be rejected entirely: whereas the phylogenetic classifica-tion cannot express information about divergence or adaptation,there is nothing inherent in the logical structure of Romer's clas-sification that makes that information any more apparent. We arecompelled to interpret Romer's classification only in terms of set-membership.

The second response of the evolutionary systematist probablywould take the following form:

Because the approach of evolutionary classification does not insist ona one-to-one correspondence of the classification and the phylogeny,separate sets of hypotheses about groups are needed to cover bothaspects of reratlonstilos. And it is necessary at the completion of anevolutionary classificatory analysis to present the conclusions in a for-mal classification and in a phylogeny. Many workers omit the phy-logeny; this omission causes problems for others who may need theexact phylogeny for their studies and yet cannot obtain this informationfrom the classification. (Bock 1977:872)

But this response places the evolutionary systematist in a pre-carious position logically. Their classifications are supposed to con-tain phylogeny as one of their "semiindependent" variables, yet theyadmit that it cannot be retrieved from their classifications, and there-fore a phylogenetic diagram must be published along with the clas-sification. Furthermore-and Bock seems to recognize the threads ofthe problem in the above citation- neither can the second "semiin-dependent" variable-morphological and adaptive divergence-beextracted from evolutionary classifications without an accompanyingphylogenetic diagram. One cannot interpret their system of ranking,as a means to convey information about morphological similarity ordissimilarity, without reference to a phylogenetic hypothesis. Thispoint is also emphasized by Farris (1977:847):

1. That there exists a set of similarities common to the Gnatho-stomata. Under the evolutionary classification, for example, the exis-tence of a set of species having jaws and the associated branchialapparatus cannot be predicted, nor, apparently, is it considered a"useful generalization."

2. That there exists a set of similarities common to all ver-tebrates excluding agnathans and elasmobranchs.

3. That there exists a set of similarities common to the holosteanand teleostean fishes.

4. That there exists a set of similarities common to the Dipno!and the choanate vertebrates.

5. That there exists a set of similarities among the "reptiles,"birds, and mammals. Thus, the evolutionary classification cannotmake the prediction that there is a set of vertebrate species sharingan amniote egg.

6. That there exists a set of similarities common to birds andcertain taxa of "reptiles."

This is a substantial list of predictions, involving suites of char-acters used by systematists for over a hundred years to define setsof species. Yet, this particular evolutionary classification does notexpress, in its logical structure, the capability to make these predic-tions.

An evolutionary systematist might make the following two-foldresponse to this criticism. Their first rejoinder might be predicatedon the major reason for forming their classifications in the first place,namely that the phylogenetic classification of Nelson itself could notpredict: (1) that there exists a set of similarities common to all fishes(Romer's Pisces); (2) that there exists a set of simi larities common tothe dipnoans and actinopterygian fishes; and (3) that Chondrich-thyes, Osteichthyes, Aves, and Mammalia all exhibit significant mor-phological divergence from their presumed ancestors and have en-tered strikingly new adaptive zones.

The first two points can be answered very simply. All similarities

210 Biological ClassificationBiological Classification 211

Evolutionary systematists, it should be clear, subscribe to an as-sortment of methodological rules having the overall effect of revers-ing, or at least resisting, this historical tendency. Their procedures ofranking are specifically designed to produce nonmonophyletic taxa,i.e., not-A sets. The classifications discussed above amply demon-strate this (Pisces, Reptilia, Pongidae, and so on). It seems that littlecan be done, at this point to convince established evolutionary sys-tematists of the need to reevaluate their premises more carefully: "Ido not care whether taxa are paraphytetlc (reptiles) or halo phyletic(birds). I readily accept both" (Michener 1978:114). On the otherhand, the tendency to eliminate not-A sets is vigorous and gains in-creasing support from the systematic community in particular, and

biology in general.

The difficulty with attempting to use categorical rank to convey infor-mation about degree of distinctiveness of taxa is that in order for it towork the categorical level of separation between the two taxa ob-viously must generally be interpretable in terms of degree of dif-ference. But Homo [in Simpson's classification, discussed earlier] isseparated from Hylobates at the same categorical level as that atwhich Homo is separated from Pan, conveying the "information" thatHomo is equally distinct from Pan and Hylobates. Homo, however, isadmitted to be much more similar to Pan than it is to Hylobates. Bas-ing categorical rank on degree of difference is thus seen to distort orconceal the very kind of information it was intended to express (Italicsadded)

Thus, it seems that neither of the two kinds of information pur-ported to be contained in evolutionary classifications can in fact beretrieved. If a method of classification, in itself, without reference toall potentially accompanying diagrams, text discussion, and so forth,cannot stand alone as a general reference system, what is there torecommend it? The poverty of evolutionary systematics is not that itis a "sloppy system" because of "sloppy material" (Bock 1977:867);rather the method and theory lack logic and conceptual clarity,qualities so necessary for placing the system in order.

Cladograms, Trees, and Classification

Cladograms and Linnaean classification schemes share a similarlogical structure: they consist of sets within sets. In cladograms, thenested sets of taxa reflect nested sets of synapomorphies; in Lin-naean classifications, the nested sets of taxa are the primary infor-mation, for the classification itself could have been based on abranching diagram derived from one of a variety of approaches.Why, then, are cladograms, as branching diagrams, considered tobe of value in classifying organisms?

So far, this book has developed the viewpoint that cladogramscan be used to make general statements about the history of orga-nisms. Given the acceptance of the evolutionary paradigm, ctado-

Conclusion

It was shown at the beginning of this chapter that a historical ten-dency has existed in classification since the time of Linnaeus. Thattendency has been toward "natural classification" and has beenmanifested by a general desire of systematists to eliminate not-Asets. Thus, systematists through the years have strived to recognizestrictly monophyletic taxa and to eliminate those which are non-monophyletic."

8. The literature on the possible kinds of norunonophyly and their definitions has been tor-tuous to say the least (the reader is referred to Ashlock 1971, 1972; Farris 1974; Hennig 1965,1966, 1975; Nelson 1971b, 1973b; and Platnick 1977a). Whereas most workers now accept thegeneral concept of monophyly as formulated by Hennig' 'A monophyletic group is a group ofspecies descended from a single ('stem ') species, and which includes all species descendedfrom this stem species" (Hennig 1966:73), there are still diverse opinions on how to view non-monophyly. The term holophyletic has been used by Ashlock and other evolutionary systema-tists to refer to the concept of monophyly sensu Hennig and other phylogenetic systematists;because holopbyletic is a synonym of monophyletic, only the latter term will be used here.There have been two aspects of nonmonophyly, peraphyty and polyphyly. Aside from the in-trades of how these terms are to be formally defined and applied (we suggest those of Farris

1974' Plamick 1977aj, the conception of many workers seems to be that paraph)'l}' refers totaxa classified together primarily on the basis of shared primitive characters, with some othertaxon genealogically related to taxa within that group ranked as a separate. higher taxon (thehominoid classification of Simpson. in which Australopltbecus and Homo are not c1asslfledWIthPall, is illustrative). Polyphyly/ie, on the other hand, has been viewed as referring to taxa clas-sified together on the basis of sharing some presumably derived character later shown 10 be in-dependently acquired (convergent). Perhaps the only consequence of this semannr distincuon IS

that evolutionary systematists admit both monophyletic and paraphyletic groups into their clas-sificatory system, whereas phylogenetic systematists accept only monophyletic groups. Con-sequently, a distinction more refined than that between monophyly and nonmonophyly hardlyseems important, and only these two terms will be used in this book.


grams are hypotheses about the structure of that history, that is, not A B C D E A B C D Especifically about the history itself, but about the structure of therelationships of the organisms as expressed in their patterns ofshared evolutionary novelties. This underlying structure is a morefundamental, more general aspect of evolutionary history than the o.numerous hypotheses that might be constructed about specific evo-lutionary events and, consequently, is the reason why a number ofevolutionary trees may have identical structures, although not neces-

Tree Clodogramsarily identical topologies. Classifications based on cladograms areimmune to the various problems and special requirements enqen-dered by alternative evolutionary trees, yet contain what is the com-mon denominator to all alternative trees, the nested sets of taxa. It isappropriate at this point to examine the structural characteristics of A B C D E A B C D Etrees and cladograms in some more detail and to demonstrate therelationship of that structure to classificatory hierarchies.

b.The Structure or Trees and Cladograms

With regard to configurations of trees, two separate topologies canbe identified: (1) trees in which none of the included taxa is specifiedas being ancestral to any other taxon or group of taxa within the tree(i.e., all taxa are terminal), or (2) trees in which one or more of the in-cluded taxa are specified as being ancestral to some other taxon orgroup of taxa within the tree.

Furthermore, trees in which ancestral taxa are specified may becategorized as follows: (2a) trees in which the specified ancestor orancestors are directly ancestral to only one other taxon; (2b) trees inwhich a specified ancestor or ancestors are ancestral to two or moretaxa (common ancestors are specified); (2c) some combination of 2aand 2b,9

Given a tree of type 1 in which no ancestral taxon is specified,the tree is identical in configuration to the corresponding cladogram(figure 5.12 top). In the tree, branch points signify hypothesized spe-ciation events; in the cladogram. branch points signify the joint pos-session of synapomorphy. Neither the cladogram nor the tree has to

Cladogram

, 9. In this discussion, unlike tile formalization of Nelson (l973c), hypothetical ancestorswill nOlbe considered; the question of interest is the structure of known taxa not of unknown,bypotbetical entities. Despite this difference, Nelson's treatment and the one 'given here exhibitmany similarities.

Figure 5.12 Diagrams depicting the similarities in the structures (set-mem-bership) of trees and cladograms {see text}.

be dichotomized fully in order for this isometric correspondence instructure to hold (figure 5,12 bottom). Thus, in cases in which ances-tors are not specified, the structures of the cladogram and its corre-sponding tree are identical and can be expressed hierarchically in aclassification. The classifications corresponding to the diagrams inthe top and bottom of figure 5.12 are as follows:

(top) (bollom)TAXON ABCDE TAXON ABCDETAXON ABCD TAXON ABeDTAXON AB TAXON ABCTAXON A TAXON ATAXON B TAXON B

TAXON CD TAXON CTAXON C TAXON 0TAXON 0 TAXON E

TAXON E


Any alternative classification for these two examples thatchanged the set-membership of the taxa would destroy the topologi-cal information implied by the cladogram and the tree.

In the case of Type 2a trees, none is isomorphic with the clade-gram (figure 5.13); nevertheless, the underlying structure (Le.. theset-membership) of all the trees is identical to that of the cladogram.Thus, a single classificatory statement can summarize the basicstructural information contained in a diversity of evolutionary trees:

c oA B

Cladogram:

TAXON ABCD A ATAXON ABC ,TAXON AB

TAXON ATAXON B a. b.

TAXON CTAXON 0

c 0 A

BC

c.

B C B

A

f.

o

c oA o ABOnce again, that information is the nested pattern of synapomorphyexpressed by the cladogram.

Type 2b trees, in which common ancestors are specified, corre-spond in structure to cladograms containing one or more tricho-tomies (figure 5.14). In the example, taxa A, S, and C share one ormore synapomorphies, but no two of the taxa share synapomorphiesexcluding the third. The set-membership implied by the trees isidentical to that of the cladogram and, again, can be expressed inhierarchical form:

TAXON ABCDETAXON ABCDTAXON ABCTAXON ATAXON BTAXON C

TAXON DTAXON E

d. e.o

o A C B C

h.~ X9

B

AC

Figure 5.13 (a-i) Some possible evolutionary trees for the four species.A-D. The structure (set-membership) of all trees can be expressed by thesame cladogram. (See text.)

tion unique to trees, i.e., ancestry. The information of cladograms, onthe other hand, consists entirely of set-membership and this is easilytranslated into a scheme of Linnaean hierarchies. The remainder ofthis chapter is devoted to discussing the procedures necessary toproduce phylogenetic classifications based on cladograms.

It can be concluded on the basis of these examples that clade-grams are more general hypotheses about the history of life than aretrees, and that, as such, there is much to recommend them as abasis for classification, Unless one wants to introduce some conven-tion to identify ancestors within the Linnaean hierarchy, as was dis-cussed earlier in this chapter, the logical structure of a Linnaeanclassification is incapable of expressing the specific kind of informa-

The Nature of Phylogenetic Classifications

Throughout this book, a distinction has been maintained betweencladograms and trees. This distinction has arisen in recent years as


A B c Eo

Cladogram:

a.

A Bc

c. o

A

€. g BoE

B CB C o

Ak.I. J.

E

CA B o A

mE

o.

Figure 5.14 (a-c) Evolutionary trees in which common ancestors are specified.The structure (set-membership) of all trees can be expressed by the same clade-gram. (See text.)

a means of separating the problems associated with analyzing pat-tern (e.g., the distributions of similarities and differences amongtaxa) from those involving a more direct consideration of evolu-tionary events and processes. Such a distinction seems desirable for


E ABO E

Cd.

E A C 0 E

h. B

E BCD E

I.

E ABO E

p.

it tends to require greater analytical precision on the part of a sys-tematist investigating phylogenetic and evolutionary ~uestions. T.hisis particularly true if the primary goal of the research IS constructionof hypotheses about evolutionary trees. How, after all, can one inves-tigate evolutionary trees without first dealing with hypotheses aboutcladograms, i.e., about synapomorphy pattern?

The reader will realize, however, that phylogenetic systematists

I\i I


have not always recognized a conceptual difference between clado-grams and trees, and indeed some may still prefer not to do so. As aconsequence, since the mid-1960s, many of these workers haveviewed cladograms as trees in which ancestors are not being speci-fied: cladograms defined in this way are taken to be hypothesesabout phylogenetic branching patterns and discussions about taxaare in terms of phylogenetic (genealogical) relationships. Thisstance poses no special problems for our discussion of classifica-tion, because both the more traditional and the newer concept ofcladograms share a similar basis: both depict synapomorphy pattern,and it is that pattern, expressed by nested sets of taxa, which canserve as a basis for classification.

Given a cladogram or hypotheses of phylogenetic relationships,how are these to be expressed in a classification? The answer thatfollows will summarize the conventional views outlined by phylogen-etic systematists from Hennig (1966) to the present (see Brundin1966; Nelson 1972a, 1973; Cracraft 1974b; Bonde 1977).

Both cladograms and Linnaean classifications depict hierarchi-cally arranged sets of taxa. The problem, then, is to effect a symmet-rical conversion of the cladogram to a classification in order to max-imize the information content of the latter. It has been pointed outthat taxa are classified according to the two aspects of Linnaeanhierarchies-subordination and sequencing-and it is these aspectswhich provide the basis for constructing any kind of classification,including one that is phylogenetic (genealogical). The aspect ofsubordination has preoccupied taxonomic thinking since the time ofLinnaeus, and only recently has sequencing been proposed as amethod to convey phylogenetic relationships (see below).

Given the cladogram. or phylogenetic hypothesis. shown in fig-ure 5.15, what might be a corresponding classification? Ignoring forthe moment the absolute rank of the included taxa or their propernames, one possible classification might be as follows.


Figure 5.15 A cladistic hypothesisfor somemajorgroupsof vertebrates.

CLASSMammaliaSUBCLASSMonotremataSUBCLASSTheriaINFRACLASSMetatheriaINFRACLASSEutheria

This classification accurately depicts the nested sets of taxashown in the cladogram and, consequently, also precisely ex-presses the implied phylogenetic relationships. Phylogenetic clas-sifications, including this one, contain only monophyletic taxa (to theextent that we have corroborated hypotheses of relationships); thatis, in a phylogenetic sense, all taxa, regardless of rank, contain allknown species hypothesized to be descended from a common an-cestral species. Moreover, in those classifications in which subordi-nation is used to indicate cladistic relationships, it is implied thatcoordinate sister-groups stem from a common ancestral species;hence these groups are of the same absolute age. Within any phy-logenetic classification of this kind, therefore, the categorical rank ofa taxon is an indication of relative age. But to repeat, the most impor-tant consideration in constructing phylogenetic classifications is thecorrect depiction of the set-membership exhibited by the branching

SUBPHYLUMVertebrataSUPERClASSBatrachomorphaCLASSAmphibia

SuPERClASSAmniotaCLASSArchosauriaSUBCLASSCrocodiliaSuBCLASSAves


diagram (cladogram). Thus, in this example, the information contentwill be completely specified in the classification as long as the set-membership takes the following form:

GROUP:Amphibia, Crocodilia, Aves, Monotremata, Metatherta. EutheriaSUBGROUP1: AmphibiaSUBGROUPta: Crocodilia, Aves, Monotremata, Metatheria, EutheriaSUBGROUP2: Crocodilia, AvesSUBGROUPza: Monotremata, Metatheria, EutheriaSUBGROUP3: MonotremataSUBGROUPSa: Metathena, Eutheria

compromise possible. each specialist can erect a consistent phy-logenetic system for his group without any necessity for corre-spondence on the basis of equivalent age between the absolute rankorder of his categories and the absolute rank order of other groups ofanimals.

Hennig (1966) did not discuss in detail the limitations of usingsubordination of ranks, by itself, to convey the hierarchical informa-tion of a branching diagram. But it did not take long for systematists,includ ing proponents of phylogenetic classification, to call attentionto the problems of relying entirely on subordination. Those systema-tists sharing Hennig's goal of conveying phylogenetic information inclassification have offered some imaginative solutions to these prob-lems; those not sharing his goal, on the other hand, have used theseproblems as a reason for rejecting outright his system of classifica-tion and for simultaneously ignoring alternative solutions proposedby phylogenetic systematists,

Perhaps three major problems of Hennig's subordinationscheme have been identified.

The names of the taxa or their absolute ranks are of somewhat lessconsequence.

In laying the foundations of phylogenetic classification, Hennig(1950, 1966) was attempting to provide principles whereby thehierarchical taxic relationships of phylogenetic hypotheses could betranslated unambiguously into a classification, and vice versa. Hisphylogenetic hypotheses represented relationships dichotomously(to the extent that analysis so permitted); consequently a dicho-tom?us hierarchical treatment of subordinate ranks, as just sum-mafl~ed and illustrated, permitted the precise expression of thoserelatlon.shi~s. Hennig's approach to classification, then, logically re-sulted In sister-group taxa being coordinate and of the same abso-lute rank-he even viewed this as "the fundamental constructionprinciple of the phylogenetic system" (1966:156).

. Much of ~~nnig's discussion centered around the goal of havinga single classificatory system for all groups of animals. Under such asystem, there. would be a direct correlation between age of origin ofa group and ItS absolute rank. Hennig also suggested that if such asystem could be constructed, taxa of the same rank might be morecomparable than is now possible with current classifications. AI-thoug.h his suggestion has been ridiculed by various systematists,Hennig himself actually took a somewhat more realistic view of cur-rent systematic practice (1966:191):

The r~~uirement that rank designations must express the com-parability of [taxa assigned to the same] categories-however re-m?tely related.the qroups-c-is not a fundamental principle of phylogen-etic systematics to the same degree as the requirement that the~stem must c~ntain only monophyletic groups and that sister groupsust be coordinate and be given the same rank. This fact makes a

Problem 1 Because each branch of a branching diagram (phylogeny)necessitates subordination, it becomes apparent that such a schemecannot easily accommodate the known taxonomic diversity of mostgroups of organisms (assuming that their phylogenetic relationshipsare generally accepted). Such a classification would require (a) thecreation of new categorical ranks and (b) the creation of new namesfor taxa at each of those ranks.

Some typical comments in the literature follow:

If phylogenetic classification proceeds usually by dichotomous divi-sions, and very unequal ones at that, it will necessitate the use ofmany more categories than were needed for older, "formal" systems.(Crowson 1970:260)Since in any dichotomous dendrogram there is one less branchingpoint than there are terminal points, the number of names needed for acomplete description of the cladogram is one less than the number ofspecies contained in the group. .' I don't believe that the cladistswant to burden systematics with the number of names needed to makeformal cladistic classifications completely describe cladograms. (Ash-lock 1974:96)It appears very doubtful that the increased expression permitted by aphylogenetic classification will compensate for the confusion and

--

m Biological Classification Biological Classification 223

orderproblems that will accompany the proliferation of categorical levelsand taxa names. (Bock 1977:862)

The "problem" of subordinated phylogenetic classifications is il-lustrated in McKenna's (1975) preliminary attempt to classify phy-logenetically the higher taxa of mammals, It was necessary, in thisinstance, for McKenna to create new categorical rank names (e.g.,Parvorder, Superlegion, Legion, Sublegion, Supercohort. Magnorder,Grandorder, Mirorder), along with many new taxon names. Neverthe-less, in so doing, he presented a classification reflecting phylogene-tic relationships, to the extent that he discussed them.

order

suborder suborder

superfamily

Csuperfamily

A B o Et

"II/I

//

/III,

Problem 2 The notion that coordinate sister-taxa are to have the samerank in a subordinated phylogenetic system is said to lead to objec-tionable consequences when classifying fossils:

The greatest possible difficulties for phylogenetic classification whichcould arise from these cases would be if Aecent species of Nautiluswere traced to separate Mesozoic ancestors. . We might then findourselves obliged, by the logic of phylogenetic systematics, to placein separate tribes or even subtarmues species which systematistshave regarded as only barely distinguishable from each other, or tomake separate families for what have been considered as poorly sepa-rated genera. These problems, however, have not yet arisen, and maynever arise. (Crowson 1970:248)The kind of systematic presentation Hennig then had in mind would bevery much open to a charge of conceptuai redundancy because of aproliferation of monobasic group names. All species which originatedand became extinct in the Triassic would, if orders are defined asgroups originating in the Triassic, be considered to represent monoba-sic orders, families, and genera. (Griffiths 1973:339)The problem raised particularly by the inclusion of older fossils is thatthey are likely to engender a procession of higher and higher ranks toexpress the relationships of increasingly plesfornorph sister-taxa. (Pat-terson and Aosen 1977:155)

Hennig desired a classification encompassing all known orga-nisms, and proposed that ranking within such a system be deter-mined not only by relative time of origin, which would be a logicalresult of any hierarchical, subordinated, phylogenetic system, butalso by absolute time of origin. The latter would only be possible, itwould seem, if in fact biology had a complete branching system forall organisms, otherwise ranking by absolute age would be mean-

Figure 5.16 A cladogram for five taxa, four of them Recentand one fossil (see text).

ingless from group to group. As this situation is moot within biology,the question of ranking by absolute age can be safely ab~ndon~duntil some future date. But the Question of relative rankmg stilicauses some concern with regard to fossil taxa. Consider, for ~x·ample (figure 5.16), four Recent taxa, A-D, each, let us say, with.many subgroups and species, and traditionally classified as follows.

OROER A-DSUBORDER DSUBORDER A-CSUPERFAMilY CSUPERFAMilY A-B

d h t be placed in itsA newly discovered fossil, species E, woul ave 0own order since it would be the coordinate sister-taxon of order A-~.It is conceivable therefore that the discovery of additional fosS.1s

, , , b 'ng created for sln-could result in a number of higher categories er .

. As Henn Ig notesgle fossil species or for only a few specIes. .'b idered obJectlOna•(1966:192) however this mayor may not e cons

ble "only i~sofar as 'it contradicts ideas associated wit~ o~~ moore.Ote'I h! h r categories espress typological way of thinking of the Ig e . I' t toh! , Ih part of systema IS st IS, there has been a general desire on eavoid classifications of this sort.


Problem 3 If ranking within phylogenetic classifications is deter-mined by the branching sequence of a cladogram, then newly dis-covered taxa (Recent or extinct), or changes in our understanding ofrelationships, or uncertainty about those relationships in the firstplace, all might necessitate major alterations in the ranks of namedtaxa, thus creating the potential for an inherently unstable classifica-tion.

It can be argued, of course, that change in our understanding ofrelationships is a worthwhile reason for a change in classification,and surely few such changes tend to wreak havoc with a system ofranking. Nonetheless, it is desirable to incorporate newly discoveredtaxa and reallocations of systematic status into classifications with aminimum of change in rank and categorical names, One of the func-tions of ranking, and classification in general, is to convey informa-tion about set-membership, and in order for this to be carried out ef-fectively, categorical and rank names should be as stable aspossible.

It should be noted, then, that these problems do not directlychallenge the logical structure or the capability of phylogenetic clas-sifications to convey precise statements about set-membership.Rather, these are problems more of convenience and communicationresulting from the difficulty of remembering categorical and taxicnames. Moreover, they are not necessarily problems restricted tophylogenetic classifications but are shared by other classificatorysystems to one degree or the other. And they are problems to the ex-tent that systematists view current systems and their conventions asmonolithic and resistant to change.

Phylogenetic systematists have proposed a number of clas-sificatory conventions to circumvent the problems just noted. Hennig(1969) himself suggested replacing categorical names with a num-bering system (table 5.1): the resultant classification is hierarchicalbut not Linnaean. His classification maintains a dichotomous ar-rangement of coordinate sister-groups, but the numbering systemhas several disadvantages. First. the numbers are functioning likecategorical rank names and, as long as the classification is dicho-tomous, any changes that might create problems for these Linnaeanranks would appear to do the same for number ranks. Second,placed within the context of our systematic tradition, numbers are notas effective for verbal or written communication as the pre-existing

Biological Classification 22S

Table 5.1 Hennig's Systematization of the Insecta"

1, Entognatha 2.2.2.2..3.2..2. Condylognatha1.1. Diplura 2.2.2.2..3.2..2.1. Thysanoptera1.2. El!ipura 2.2.2.2..3.2..2.2. Hemiptera1.2,1. Proarra 2,2.2.2..3.2..2.2.1. Heteropteroidea1.2.2.Couembora 2.2.2.2..3.2..2.2.1.1. Coleorrhyncha

2, Ectognatha 2.2.2.2..3,2..2.2.1.2. Heteroptera2.1. Archaeognatha (Microcoryphia) 2.2,2.2..3.2..2,2.2. Sternorrf1yncha2.2. Dicondylia 2.2.2.2.,3.2..2.2.2,1. Aphidomorpha2.2.t. Zygentoma 2.2.2.2.,3.2..2.2.2.1.1. Aphidina2.2.2, Pterygota 2.2.2.2..3.2..2.2.2.1,2. ccccma2.2.2.1. Palaeoptera 2.2.2,2..3.2..2.2.2.2. Psyllomorpha2.2.2.1..1. Ephemeroptera 2.2.2.2..3.2..2.2.2.2.1. Aleyrodina2,2.2.1..2. Odonata 2,2.2.2..3.2..2.2.2.2.2. Psyllina2.2.2.2. Neoptera 2.2.2.2..3.2..2.2.3. Auchenortryncha2 2 2 2 1 PI , 2.2.2.2.,3.2..2.2.3.1. Fulgoriformes. . . ... ecop era2.2.2.2..2. Paurometabola 2.2.2.2..3.2..2.2.3.2. Cicadiformes2.2.2.2..2,1. Embioptera 2.2.2.2..4. Holometabola2.2,2.2..2.2. Orthopteromorpha 2.2.2.2..4.1. Neuropteroidea2.2.2.2..2.2..1. Blallopteriformia 2.2,2.2..4.1..1 Megaloptera2.2.2.2..2,2..1.1 Notoptera (Grylloblallodea) 2.2.2.2..4.1..2. Raphidioptera2 2 2 2 2 2 1 2 0 , 2.2.2.2..4.1..3. Planipennia. . . .. . .. ,. ermap era2.2.2.2..2.2..1.3. Blallopteroidea 2.2.2.2..4.2. Coleoptera2.2.2.2..2.2..1.3.1. Mantodea 2.2.2.2..4.3. Strepsiptera2.2.2.2..2.2..1.3.2. Blattodea 2.2.2,2..4.4. Hymenoptera2.2.2.2..2.2..2. Orthopteroidea 2.2.2.2..4.5. Siphonaptera2.2.2.2..2.2..2.1. Ensifera 2.2.2,2..4.6. Mecopteroidea2.2.2.2..2.2..2.2. caeutera 2,2.2.2..4.6..1. Amphiesmenoptera2.2.2.2..2.2..2.3. Phasmatodea 2.2.2.2..4.6..1.1. rncncctera2.2.2.2..3. Paraneoptera 2.2.2.2..4.6..1.2. Lepidoptera2.2.2.2..3.1. Zoraptera 2.2.2.2..4.6..2. Antliophora2.2.2.2..3.2. Acercaria 2.2.2.2..4.6..2.1 Mecoptera2.2.2.2..3.2..1. Psocodea 2.2.2.2..4.6..2.2. Diptera

"From Hennig, 1969.

categorical rank names (although the taxon names are the ~ame inboth systems), Finally, given a sufficiently large etassiflcaticn. thenumber ranks preceding the taxon name may be inordinately lo.n?,

. . teat! Hennig sthus further reducing its effectiveness 10 communlca Ion..replacement of rank names with numbers was undertaken 10 the ap-parent belief that dichotomous subordination was the only mean~ ofCOnveyingphylogenetic relationships. 1t therefore seems appropfl~teat this point to consider some of the alternative methods for formingphylogenetic classifications within the Unnaean hierarchical system,methods that do not rely necessarily only on subordination.

226 Biological ClassificationBiological Classification 227

sequencing convention this branching diagram can be specifiedprecisely by the following classification:

GROUP1: AgnathaGROUP2: ActinopterygiiGROUP3: AmphibiaGROUP4: AvesGROUP5: MonotremataGROUP6: Metatheria (or Eutheria)GROUP7: Eutheria (or Metatheria)

This classification signifies that the Agnatha are the sister-groupof the six taxa listed below it, the Actinopterygii the sister-group ofthe five taxa below it, and so on Rank was not specified; its choicewould have no effect on the information content of the classification.Givenonly these taxa, one rank is sufficient to specify relationships;compare this to a conventional, subordinated phylogenetic clas-sification in which, minimally, six ranks are required:

SUPERCLASSAgnathaSUPERCLASSGnathostomataCLASSActinopterygiiCLASSChoanataSUBCLASSAmphibiaSUBCLASSReptilomorphaINFRACLASSAvesINFRACLASSMammaliaDIVISIONMonotremataDIVISIONTheriaCoHORTMetatheriaCOHORTEutheria

In a classification in which taxa are sequenced only, absoluterankWould be determined by (a) the rank given to the group as aWholeand (b) an arbitrary choice of rank for the included taxa aslong as that rank is lower than that applied to the group as a whole.Thus,if the seven taxa as a group were classified within the subphy-I~mVertebrata, then, in a sequenced classification, each could begiven a rank as high as superclass or as low as, say, family; andrankWould have no influence on information content.. Phyletic sequencing by itself solves one of the major problemsIdentified for purely subordinated classifications, namely, the prolif-erationof categorical and taxon names.

If SUbordination and phyletic sequencing are combined, there

Phyletic Sequencing

The topology of Linnaean hierarchies has two aspects, subordinationof ranks and the linear listing of taxon names of the same rank. Untilrecently this second aspect was not used to convey informationabout set-rnernberstnp but Nelson (1972b) showed that it can be soutilized. Phyletic sequencing, as this can be called in order to distin-guish it from conventional listing in which information about rela-tionships is not implied, has been discussed and applied by anumber of workers (e.g., Cracraft 1974b; Wiley 1976; Patterson andRosen 1977; Bonde 1977). Phylogenetic relationships (or set-mem-bership in its most general form) are conveyed by sequencing usinga simple convention: the first taxon in the sequence is considered thesister-taxon of all taxa listed below it, the second taxon the sister-taxon of a/I below it, and so on. As a simple example, consider thebranching diagram or phylogeny of figure 5.17. Using the phyletic

~ .~.S> r Jf>.S> ~ .~ t

~~ .~ R~ <; s p l .,? <;.>' ~

'" sIii $ ~ ". "" of' ""

FIgure 5.17 A cladistic hypothesis for some majorgroups of vertebrates.


exists the potential for many different classifications, all of whichspecify the same set-membership. For example, two of these clas-sifications might be as follows:

ApidiumOreopithecus

FAMilY HylobatidaePliopithecusHylobates

FAMILY PongidaePongo

FAMilY HominidaeSUBFAMILY PantneePan

SUBFAMILY RamapithecinaeRamapithecus

SUBFAMILY HomininaeAustralopithecusHomo

Alternatively, the Paninae, Ramapithecinae, and/or Homininae couldbe raised to family rank.

An evolutionary systematist thus may believe that such a clas-sification has an advantage over the phylogenetic classification pre-sented earlier (p. 206) in that Austra/opithecus and Homo are nowclassified at a higher rank (subfamily, or even family) than be.fore(subtribe); classifying Pan with Homo instead of Pongo emphaslz~stheir genealogical relationships and, of course, their. clo.ser geneticsimilarity. On the other hand, even though some subjective elementmight be used to express degree of difference in such a ~equenc.edclassification, it should be obvious by now that the only Informat~onbeing retrieved from that classification is set-membership, whichwould be identical to that expressed in the cladogram.

SUPERCLASS AgnathaSUPERCLASS Gnathostomata

CLASS ActinopterygiiCLASS AmphibiaCLASS AvesCLASS Mammalia

SUBCLASS MonotremataSUBCLASS Metatherta (Eutheria)SUBCLASS Eutheria (Metatheria)

SUPERCLASS AgnathaSUPERCLASS Gnathostomata

CLASS ActinopterygiiCLASS Choanate

SUBCLASS AmphibiaSUBCLASS AvesSUBCLASS MonotremataSUBCLASS Metatheria (Eutheria)SUBCLASS Eutheria (Metatheria)

When phyletic sequencing and subordination are used in thisway, considerable subjectivity enters into the choice of ranks, thenumber of subordinations, and the number of taxa to be sequenced.Any nonterminal dichotomy will necessitate subordination; for in-stance, if Crocodilia had been included as the sister-group of birds,then Aves and Crocodilia would need to be subordinated belowsome other taxon name (e.q.. Archosauria). Thus. using subordina-tion and phyletic sequencing, rank will be determined by (a) the ini-tial rank given the group as a whole, and (b) an arbitrary choice ofranks, lower than the initial rank. and depending on the scheme ofsubordination and sequencing.

It has been suggested (Cracraft 1974b) that a combination ofsequencing and subordination might serve as a bridge between thegoals of the phylogenetic systematists on the one hand, and evolu-tionary systematists on the other. Such classifications are strictlyphylogenetic, i.e .. they admit only monophyletic taxa and the phy-logeny can be specified precisely; this is certainly the goal of phy-logenetic classification. Moreover, the freedom given to the systema-tist regarding choice of rank and the pattern of sequencing andsubordination would permit an evolutionary systematist to express ageneral statement about the degree of divergence of certain taxaand still preserve the phylogenetic relationships. To illustrate thispoint. ~impson's (1963) phylogeny of the hominoid primates (figure5.8) might be classified as follows using subordination and phyleticsequencing:

SUPERFAMILY HominoideaFAMILY Oreopithecidae

Indented-List Classifications (Farris 1976)

The incorporation into classification of certain unique taxa-DC-casionally Recent but most often fossil-frequently has the effect ofcreating monotyotc taxa of high rank. For example, if a fossil orRecent species were newly discovered and determined .to be thesister-taxon of some living higher taxon, under the conventIOnal rules

. I Tdttheof phylogenetic classification this species must be c asst Ie asame rank as its coordinate sister-taxon. There also has been a ten-dency to accept the principle of "exhaustive subsidiary taxa": "If aSUbsidiary level is used within any group it should, as far .as pos-sible, be used for all organisms in that group. For example, If a sub-


family is used within a family, then the whole family should be di-vided into subfamilies" (Simpson 1961:18). As Farris (1976:272)points out, this and other rules often have the result that a singlespecies may be placed in its own order, family, subfamily, andgenus, thus increasing the number of categorical and taxon namesfor anyone classification. Moreover, Farris (1976:272~73) notes,"several taxa must be constructed, all of which have the same mem-bership-that is the 'several' taxa are nothing more than differentnames applied to one and the same group."

Whereas most systematists have, in the past, been resigned toaccept such a system, its difficulties nonetheless have been gener-ally deplored, and evolutionary systematists have pictured phy-logenetic classifications as being particularly vulnerable to thesedifficulties:

A o E FtB

Ct

Figure 5.18 A claooqram for six taxa. Dag-gers indicate fossil taxa. (See text.)

FAMILYC-ESpecies C't

SUBFAMILYD-ESpecies DSpecies E

In this example the fossil species F is placed in the class A~F todenote its inclusion relationships with species A-E, but otherwise Fis classified only to genus and species, as required by the rules.ofnomenclature. Its indentation, in juxtaposition to the order A-E, sig-nifies it as the sister-taxon of the remaining species in the ~Iass,however there is no need to classify species F to order or family asthis would be redundant. Order A-E is subdivided conventl~nally e.x-cept for fossil species C, which is not classified to sUbfa~lly but Itsjuxtaposition next to subfamily D-E indicates the two are slster~taxa.

if fan the vertebrate re-As another example of list-form class! rca I ,lationships depicted in figure 5.17 might be represented as follows:

SUBPHYLUMVertebrataAgnatha

SUPERCLASSGnathostornataActinopterygii

SUBCLASSChoanateAmphibia

INFRACLASSAmniotaAves

For the dubious advantage of being able to read off 'phylogeny' from aformal listing of taxa, craotsts are willing to pay, as Hull has said, aprice too high for many biologists. Species that split off in the Precam-brian but give rise to no other species would have to be classed asphyla. Such classifications would be highly monotypic and highlyasymmetrical. (Ashlock 1974:97)

Farris has proposed the adoption of indented I ist classificationsas one possible mechanism to classify fossil species. He suggeststhat monotypic taxa be abandoned except in those circumstances inwhich a generic name is needed for a species that cannot be as-signed to any known genus. According to Farris the only require-ments that should be adhered to are "that each monophyletic groupmust be a taxon, each taxon must be a monophyletic group, and thenatural inclusion relationships of the monophyletic groups must beretained by the taxa" (1976:275).

In indented list classifications, ranks are abandoned when in-dentation within the hierarchy conveys the same information aboutinclusion relationships. Thus, given the relationships shown in figure5,18, the list-form classification might be as follows:

CLASSA-FSpecies Ft DIVISION

ORDER A-EFAMILYA-B Monotremata

COHORTMetatheriaEutheria

Species ASpecies B


sitication of Recent taxa without disturbing the traditional ranking ofthe latter. Such classifications may have an important disadvantagein that, until one is thoroughly familiar with how to read them, their in-terpretation may be confusing, and this could be of some signifi-cance when trying to communicate with non-systematists.

A B c o E F

The "Plesion" Concept (patterson and Rosen 1977)

In one of the more extensive discussions about the classification offossils Patterson anti Rosen have recommended that fossil taxa notbe assigned the customary categorical ranks coordinate with theirsister-groups. Instead, they suggest that fossil groups be designated"plesions" (in reference to "plesiornorphtc sister-group") and in-serted into thecclasslflcation using the convention of phyleticsequencing to specify that each ptes ion taxon is the sister-group ofall taxa listed below it in the classification, In this manner neithersubordination nor new categorical and taxon names would be nec-essary; conventional subordination of ranks would continue to beapplied to Recent taxa,

Either single fossil species or fossil groups containing manyspecies may be designated plesions. Furthermore, a given pies ionmight be coordinate with a Recent taxon assigned to any categoricallevel. As Patterson and Rosen point out, the use of pleaions wouldbe appropriate in general classification of major groups or in morerestricted classifications of paleontologically rich groups; plesionswould not be needed if the classification were of strictly extant taxa.te

The application of the pies ion concept can be illustrated with anexample taken from Patterson and Rosen's analysis of the rela-tionships and classification of fossil and Recent teleost fish (for anadditional example see Wiley 1976:93). Figure 5.20 depicts the hy-pothesized relationships of a sample of fossil and Recent taxa; Pat-terson and Rosen (1977:163) propose the following classification'

SUBDIVISIONTeleosteipies ion tPholidophorus becheipleaion tPholidolepis dorsetensis

Figure 5.19 A cladogram for six taxa (see text).(After Farris 1976:276, figure 1.)

In this classification redundant rank and taxon names are elimi-nated. For example, the Agnatha are not classified as to superclass,subclass, infraclass. and so on. Nevertheless, the classification con-veys a precise statement about relationships,

List-form classifications can also be used to depict relationshipsthat are not dichotomous. For the cladogram of figure 5.19, the clas-sification might take the following form:

ORDER A-FSpecies A

SUBFAMILYB-eSpecies BSpecies C

FAMILY D-F

Species DSUBFAMILYE-FSpecies ESpecies F

. That there is a trichotomy involved is signified by placing spe-cies A at a lower rank than species Be-C. and species 8 and C at alower rank than the taxon, D-F. Thus, these three groups (A, B-C,and D-F) are equal subtaxa of order A-F, none being a subtaxon ofthe other.

One of the advantages, then, of tndented-ust classifications isthat redundancy of categorical rank and taxon names is minimized.Also, fossil taxa can be intercalated into or removed from a clas-

, 10. There is no reason, OfCOUfSe,why a convention such as "plesicn' could not be usedor an extant taxon comprised of one f . . .

group. ' or a ew species. that 15the sister-taxon of a much larger

l

·0 i) ~~ ••1l E

I •• ~ j ••! u -e C • •-e • ] c

l • 1 :§ C ~

1l ~ • I E

J •• • ~~ " " ~ " ~c ! e g §" ." ~&

c il' :E~ " a :§ E c

~ " ~ " 8- • ~ " •3' c, E

~ "- c .s ,0 z w U • <5 ,

0 - w-


pies ion tLep/olepis coryphaenoidesplesion t/chthyodec/iformesolesron tThars;s dubiusSUPERCOHORT Osteoglossomorpha

ORDEROsleoglossiformesSUBORDER Osteoglossoidei

SUBORDER Notopleroidei

SUPERFAMILYHiodonloideaplesion tLycopleridae

FAMILYHiodonlidaeSUPERFAMILYNotopleroidea

SUPERCOHQRT Elopocephala

COHORT Elopomorpha

COHORTCfupeocepba!aSUBCOHORT Clupeomorpha

Order Clupeiformespies ion tOrnategulumpies ion tDiplomystidaeSUBORDER Oenticipitoidei

SUBORDER Clupeoidei

SUBCOHORTEuteleostei

The plesion taxa of this classification include those of a singlespecies and those containing many (e.q.. Ichthyodectiformes): theyare also seen to be coordinate with supercohort, suborder, and

family rank taxa.. The concept of pies ions appears to be a useful tool in construct~Ing phylogenetic classifications of large groupS of organisms havinga rich fossil record. At first thought, the use of phyletic sequencinga!one might seem to have the same advantages as the use of ple-sian, but in order to effect such a classification redundant highertaxon names must be invented. In this example, each of the first fivefossil taxa in the classification would have to be placed in a taxon of

supercohort rank:

SUBDIVISION TeteosteiSUPERCOHORT Pholidophorei

Pholidophorus becheiSUPERCOHORT Pholidolepei

Pholidolepis dorsetensis

Fi( gore 5.20 A cladistic hypothesis for ssee text). (After Patterson and Rosen 19c;n7e,5R3ecent and fossil teleost fish

, , figure 54.)

II

236 Biological Classification Biological Classification 237SUPERCOHORT Leptolepe!

Leptolepis coryphaenoidesSUPERCOHORTIchlhyodecteiOrder Ichthyodectiformes

SUPERCOHQRT TharsetItuusis dubius

SUPERCOHOATOsteoglossomorphaSUPERCOHORTE'opocephala

In this case, the creation of new supercohort names is redundantand serves little useful purpose. The plesian concept avoids thesedifficulties.

A B CtD'C-\y--------- ./\<~,

\? -

'I.)

A B C o

0_ b_Figure 5.21 A cladistic hypothesis for three Recent taxa and one f~ssiltaxon of uncertain relationships (see text). (After Nelson 1972b:229, figure 2.)

GENUS CGENUS 0

If the classification as a whole is not expressing relationships byphyletic sequencing procedures, then the above method of denotinga trichotomy (and the implied uncertainties in relationships) withinthe classification will be unambiguous, If, however, portions of theclassification include sequencing, then the above notation might betaken to imply that B is the sister-qroup of C-D. In this case, unre-solved trichotomies will have to be tagged for recognition by someconvention, perhaps by placing all three groups incertae sedis:

FAMILY A-D

SUBFAMILY ASUBFAMILY B-DGENUS B (incertae sedis)GENUS C (incertae sedis)GENUS 0 (incertae sedis)

Another possibility is that trichotomies might be bracketedwithin classifications to distinguish them from sequencmg:

FAMILY A-DSUBFAMILY ASUBFAMILY B-D

[

GENUS BGENUS CGENUS 0

Classifying Taxa of Uncertain Status

In routine practice a systematist frequently confronts situations inwhich the systematic status of taxa is uncertain. This uncertainty mayarise from (a) our inability to resolve relationships despite the pres-ence of adequate comparative material, (b) the fact that the taxa maynot have been studied in sufficient detail by previous workers, or (c)our inability to resolve relationships because of the absence of ade-quately preserved material (this is of common occurrence in fossilstudies). For the systematist, the problem is how to incorporate suchkinds of taxa into a phylogenetic classification,

In cases a and b. the problem almost always involves an unre-solved trichotomy in a cladogram analysis. Although a single taxonmay be the focus of attention, for example, when an attempt is madeto interpolate a fossil taxon into a scheme of relationships based onRecent groups, the uncertainty of the situation extends to some ofthese Recent taxa as well. Consider the example of Recent taxa,s-C, related as shown in figure 5.21a, and the fossil taxon 0, of un-certain relationships. Taxa Band C are determined to have a closerrelationship to each other than either has to A, but the precise rela-tionships of 0 to 8 and/or C are ambiguous (obviously B, C, and 0share some uniting synapomorphy). How is this situation to be ex-pressed in a classification? The simplest answer would be to basethe classification on the cladogram (figure 5.21b):

FAMILY A-DSUBFAMILY ASUB FAMIL Y B--DGENUS B

- 1- ith the uncertaintiesPatterson and Rosen (1977), m dea mg WI .- - ti ns: (1) fossil taxadiscussed here, adopted the tonowma conven 10 .

- - tu re then expressedwere tagged pies ions and their relations IpS we .- _ _ e subordinated In theusmq phyletic sequencing; (2) Recent taxa wer


traditional manner; (3) unresolved trichotomies within Recent taxawere expressed by a simple listing (at the same rank) of the threetaxa; and (4) the use of incertae sedis was restricted to fossil groupswhose relationships could not be resolved. Most classifications ofmajor groups will sometimes include taxa that are known to be non-monophyletic, Such taxa may not have been the focus of investiga-tion but need to be Included in the classification for one reason oranother, or for various reasons they have not been the subject of de-tailed analysis by previous workers. Patterson and Rosen suggestthat such taxa be placed in quotation marks within classifications todistinguish them from taxa assumed to be monophyletic.


because they lack an obvious connection with the expected hierar-chical structure of nature produced by the evolutionary process. Evo-lutionary trees and cladograms, on the other hand, have such a con-nection.

5. The information content of evolutionary trees cannot be ex-pressed readily in Linnaean classifications, because specified an-cestors cannot be easily incorporated into classifications, Supraspe-cific ancestors, in any case, do not exist in nature.

A cladogram can express the underlying set-membership of anevolutionary tree and thus can serve as a basis for classification. Thelogical structure of cladograms and Linnaean hierarchies is iden-tical, and thus the set-membership of the two can be made isomor-phic.

6. Evolutionary classifications, as advocated by Mayr, Simpson,and others, cannot be recommended because the logical structureof the Linnaean hierarchy makes it impossible to express simulta-neously both genealogical relationships and a measure of overallsimilarity. Consequently, the classificatory procedures of evolu-tionary systematists make it difficult to retrieve from their classifica-tions any of the information putatively used to construct the clas-sification.

7. We recommend that phylogenetic classifications based oncladograms be adopted as general reference systems in biology.Such phylogenetic classifications explicitly avoid acceptance of not-A sets. They have been shown to predict character distributions,including general phenetic resemblance, more efficiently thanphenograms or "evolutionary" classifications. They are also isomor-phic with the hypothesized genealogical relationships of the in-cluded taxa. Genealogy determines set-membership, thus genea-logical information (hypotheses) can be incorporated directly into aclassification. Using conventions such as subordination, phyleticsequencing, the indented-list procedure of Farris (1976), and thepreston concept of Patterson and Rosen (1977), it is possible to con-struct phylogenetic classifications for any group of organisms, nomatter how taxonomically diverse it might be. Phylogenetic clas-sifications, because they attempt to express the underlying pattern ofnature produced by evolution, yield natural classifications and thusprovide the soundest foundation for future stud ies in comparativesystematic biology.

Conclusions

The important conclusions of this chapter on classification are asfollows:

1. The history of systematic biology documents repeated ex-amp!es of attempts to classify organisms into A and not-A groups,that IS, those groups defined by derived attributes and those definedby primitive characteristics. That history also documents a tendencyon the part of ~ystematists to recognize only A groups as being natu-ral and to eliminate not-A groups from classifications. A theme of thechapter is that phylogenetic classification provides a solution to theproblem of obtaining natural classification, in contrast to the pro-posals of other systematic philosophies, because the method of phy-logenetl~ systematics is designed to discover and classify A groups.

2. t.innaean classification systems consist of lists 01taxa hierar-chically arranged. The only information inherent in a Linnaean hier-archy is set-membership.

3. Branching diagrams and Linnaean classifications have thesame .Iogical structure, that is, of nested sets of taxa. Because oftnls, Llnnaean classifications must inevttaoty be based on some typeof branching diagram.

4. There are three main kinds of branching diagrams: pheno-grams, evolutionary trees, and cladograms. Phenograms have notgenerally been favored as a basis for general classification systems

(

,,,.0-

1•{

r

Chapter

6Systematics and the Evolutionary Process

THE FIRST part of this book has been devoted to biologicalsystematics: method and theory pertaining to the elucidation of pat-terns of phylogenetic relationship among organisms. In so doing, wehave adopted the view that for basic phylogenetic research, all thatis required is the assumption that life has evolved, i.e., that all lifeconstitutes a monophyletic assemblage. The formulation and testing(including eventual corroboration) of such hypotheses of phylogene-tic relationship are of great intrinsic scientific interest. Highly corrob-orated hypotheses of phylogenetic relationship are a fundamentalgoal of research in biological systematics and need no further ratio-nale.

Yet there clearly are further uses for such hypotheses. We havealready considered one--classification-in the preceding chapter.Classifications express the hierarchical arrangement of taxa andgive names to monophyletic groups, thereby recognizing the statusof such groups as entities and allowing biologists to talk about themconveniently. The phylogenetic system expressed in cladogramsand classifications is the reservoir of corroborated hypotheses on thehistory of life, which forms the basis for all further predictions andgeneralizations about the nature of life. The importance of this under-taking, and the complexity and enormity of the task, is sufficient tooccupy the lifetime attention of a professional systematist working ononly a fraction of the earth's past and present organic diversity.

Yet, clearly, there is even more to the implications of the phy-logenetic system than recognizing and naming monophyleticgroups. The phylogenetic system codifies and epitomizes all that wethink we know about the genealogical history of life on earth. And ofcourse, in this context, "genealogy" is just another expression for"evolution." If we need know nothing specific about how life has

242 Systematics and the Evolutionary Process

evolved to formulate and test genealogical hypotheses, there is stillthe. opposite side of the coin to consider: can corroborated genea-logical hypotheses be used in any way to study the process of evolu-tion? And if so, how? In the remainder of this book, we shall developthe theme that such corroborated phylogenetic hypotheses can in-deed ~ used to investigate the evolutionary process, and that phy-logenetic hypotheses of the kind treated in the first half of this bookare consistent with, and suggest further, lines of inquiry and kinds ofhypotheses of evolutionary process somewhat different from those ofthe standard so-called "synthetic" paradigm.

Relationship between Evolutionary Historyand Evolutionary Process

The general form ?f w~at we take to be the correct relationship be-~een phylogenetic history and the scientific study of the evolu-tl?nary pro~ess is simply stated: hypotheses of evolutionary mecha-nlsm~ (derived de novo or taken in whole or in part from otherthe:ones) generate predictions about results (historical patterns)which are ~hen compared with highly corroborated hypotheses ofph~l?genetlc and distributional patterns. The predictions are eitherverified or found not to agree with the corroborated hypotheses ofactual e~ents. T~us, in its simplest form, the study of the evolutionaryprocess ISno dl~erent from any other subject of scientific enquiry.

The conclusion that evolution can be studied scientificaffy willof course, come as no surprise to evolutionary biologists. But it isn~ces~a~ to ~ake this simple point because evolutionary theos~~ce~s Inception as a subject worthy of serious scientific consid~~a~on,baSh~een.plagued by an assumption-ridden inductive processw er~ y IStO~IC~1patterns are examined, narratives dreamt u toexplain them (In Itself a valid way of formulating new ideas) andfur-:hn~~~~j~~:~~e to fit, ra~er than used to assess, prior ass~mPtions

" process. uch a procedure constitutes a misuse ofhistorical patterns and has impeded prcqress l t I

t evcluti s In a east some areaso eve utfonary theory. We develop a more detailed critique of someaspects of contemporary evolutionary theory b IF'e ow. or now, Just to

Systematics and the Evolutionary Process 243

provide a single example of the misuse of historical patterns in theanalysis of evolutionary mechanisms, consider the way in which theconcept of "natural selection" is used to explain why an organismhas a certain morphological configuration. There mayor may not bescientific grounds for preferring a neo-Larnarckian. neo-Darwinian,or even a neo-Kiplingian version of "why the giraffe got its longneck," but the fact remains that the results of history (i.e., existenceof a monophyletic assemblage of artiodactyla, Family Giraffidae,with, among other things, long necks) are often explained as "justso" stories, rather than being used to provide the critical evidencefor the evaluation of the relative merits of conflicting views on the na-ture of the evolutionary process itself.

On the other hand, it is equally obvious that historical patterns-or some concept of historical events-have been absolutely essen-tial to the development and progress of evolutionary theory, and thisholds true for each of the several levels of evolutionary phenomenadiscussed in this chapter. Microevolutionary phenomena (changesin gene content and frequency, and concomitant effects within popu-lations and species) are perhaps the best understood of all levels ofthe evolutionary process. This is true simply because, at this level,the phenomena are directly susceptible to experimental analysis inthe laboratory, as well as to independent theoretical mathematicalanalysis and field investigation. Thus hypotheses are tested directly,empirically. Less well appreciated is the fact that the data derivedin, say, a routine Drosophila population-cage experiment are highlycorroborated hypotheses (to the point of being fact, since all individ-uals are known to be descendants of at least some of the individualsfrom the initial population) concerning the historical relationships ofgenerations. These corroborated hypotheses provide the historicalbackground against which the nature of the observed changes ingene frequency (if any) are evaluated. This is, of course, a trivial ex-ample, as these historical hypotheses are the natural outcome ofrigidly controlled laboratory conditions. Yet it is the historical natureof the data that renders such experiments of any relevance whateverto evolutionary theory. At higher phenomenological levels, the histor-ical data are more explicitly acknowledged as important, but be-come more susceptible to doubt, because they are hypotheses' ofvarying degrees of corroboration. We explore the kinds of historicalhypotheses appropriate to the various levels of the evolutionary pro-

--------- ,..- -- 2


cess in depth later in this en thypotheses as criteria f ap ~r.In essence, the use of historicalcess (and thus the th or. evaluatlnq predictions from theories of pro-evolutionary analysis. eones themselves) is the same at all levels of

Systematics and the Evolutionary Process 24S

In contrast to the transformational concept, our adopted defini-tion of species requires that they have a definite origin (from an an-cestral species), and, logically, a definite termination. This is to saythat species are discrete entities. All theories of speciation involvingsplitting mechanisms (briefly reviewed in chapter 4) implicitly takethe position that species are ontologically "real" entities in nature.This view has been further developed from a philosophical point ofview (for slightly different reasons) by both Ghiselin (e.g., 1974) andHull (e.g., 1976). We reiterate these points at this juncture because itis evident that these two contrasting points of view concerning spe-cies and their origins can be generalized even further to character-ize two prevailing and conflicting approaches to evolutionary theory(especially, but certainly not exclusively, within the synthetic para-digm). Furthermore there is a general geometric scheme for the re-sults of the evolutionary process that conforms to each of theseviews. Distinguishing between these two approaches clarifies thebasic issues in evolutionary biology and points the way towardssome alternative approaches to some of the older problems in evolu-tionary theory.

This dichotomy in evolutionary theory is simply stated. It canbest be characterized by the two contrasting definitions of evolutionprevalent today, which highlight a duality in what evolutionists taketo be the central problem in evolution. One view holds that evolutionis primarily a phenomenon of the transformation of genes, theirfrequencies, and their products (reflected in the phenotype, behav-ior, and physiology). All modern definitions of evolution include thistransformation aspect, and most restrict themselves to it:

Contemporary Evolutionary Theory:A Basic Characterization

In the earlier chapters deal in . '. .systematics we cha I . g With speciation and Its relationship to

, rae ertzed the esssnn Itlcular theories of sp . . ra aspects of various par-geometric concepts ~c~~tlon, and related these theories to definitetion theory leads to It e results of these processes. Each specie-with its own g'eomel~ own s,etof expected historical outcomes, each

no configuration re t dgrams-specifically hI' presen e on branching dia-lar modes of specia~'P ~ o~en~tlc trees. We contrasted two particu-ing in the denv t! Ion. p ytetic transformation of a lineage result-

a Ion of a descend t· 'ancestor, as opposed t .an species directly from itsIsolation OCCursto eep. concepts of lineaqe splitting, where genetictors, We briefly noted thatm descendant species from their ances-lineage) new spec, a. 10 the former case (transformation within a

, ctes anse as a by p dthe genetic propert! t - ro uct of the transformation of. les 0 a lineage a d th 'arbItrarily delineated ,n us are viewed largely as

View of species is ac~~~~e.nts of t~at li~eage. This transformationaladopted earlier in the bay I~ contflct With the definition of speciescestry and descent d f ~Ok, Inasmuch as the pattern of parental an-

o. e In109 a species'nrtron, throughout the n ISvery much present, by deti-our adopted def 'I' Ie-span of the entire lineage. Therefore under

InJIon of "spec! " 'cogently pointed oUI] . cres, [and as WHey (1978) has

a contmuous lin . .event Wheresome fra 1" . eage, unbroken by a splittingnew species, must bc Ion o~the lineage is split off and becomes along it persists or hoe considered a single species no matter howtime span We a W m~ch phenotypic change occurs during this

. gree With Wiley th t thiagrees exactly with S' a IS concept of "species"species, while remajni~mpson'~ (1961.) concept of an evolutionaryearlier (chapter 3, p. 93~ consistent With Mayr's (1942), as we noted

Organic evolution is a series of partial or complete and irreversibletransformations of the genetic composition of populations, based prin-cipally upon altered interactions with their environment It consistschiefly of adaptive radiations into new environments. adjustments toenvironmental changes that take place in a particular habitat, and theorigin of new ways for exploiting existing habitats. These adaptivechanges occasionally give rise to greater complexity of developmentalpattern. of physiological reactions, and of interactions between popu-lations and their environment. (Dobzhansky et at. 1977)

One of the many implications of this approach, as codified and sym-bolized by definitions of this sort, is the particular point developedearlier: that species emerge as a logical consequence of such trans-

-Systematics and the Evolutionary Process 247

lion has Jed to the development of a set of propositions which seem,at base, rather difficult to test. In short, the within-population phe-nomena of mutation, variation, and selection have been extrapolatedacross the board to embrace all aspects of evolution, including 50-called macroevolution. In so doing, an important ingredient isusually missing, an ingredient that makes all the difference in the na-ture and composition of theories of the evolutionary process.

That missing ingredient is speciation. Alongside the dominanttheories of transformation has been another theme, that of the evolu-tion of taxa. Although one rarely finds in the literature definitions ofevolution conforming to this alternate approach, it is clear that manybiologists (historically, mostly neontological systematists) who havestudied species and theorized on speciation implicitly adopt theview that evolution is quintessentially a matter of the origin of newtaxa-c-i.e.. species. This, the "taxic'' approach to evolution, simplyalleges that. at its core, it is species that evolve-not individuals,and still less, anatomical or genetical properties of individuals. Inthe process of speciation, such genetic change might well occur,even as an overwhelming expectation. In an earl le- characterizationof the differences between these two approaches to evolutionarytheory, Eldredge (197gb) argued that the taxic view is the more gen-eral. embracing concept of the two, in that speciation cannot be un-derstood strictly with reference to accumulated changes in genes,their frequencies, and their products, but that speciation providesthe meaningful context for understanding and interpreting any suchchanges in genotype, whether adaptive or not We do not pursue thisview here. It is clear that, under the concept of species adopted inthis book, that a great deal of genetic and morphological changecan, at least theoretically, occur within the total time of existence of asingle species, without any splitting events occurring. To define evo-lution simply as the origin of new species would be to deny thatphyletic change within a lineage (single species) is also evolu-tionary. Rather, the gist of the problem is that the existence of dis-crete species-and the attendant problems of their origins, sub-sequent histories, and eventual demises-represents a pheno-menologicallevel distinct from that of the population.

One of the logical consequences of the foregoing observationhas been badly neglected: considerations of all macroevolutionaryphenomena (defined for the moment as evolutionary phenomena


formations within linea It' fges. IS, 0 course, possible to retain thisgeneral concept of I f ,. . evo U Ion and to diSCUSS the origin of new dis-crete species via ge ti , I ' ''. ne IC ISOation. But the particular notion thatspecies artse as a cowith! I' . nsequence solely of accumulated changes

I m meaqes IS the vi .m t I lew most consistent with the general transfer-

a lona concept and' . .the thinkin of m ' more Im~ortantly, has historically dominated

g ost paleontologists and not a few geneticists.For most contempo bi I '

proach is generall d rary I? OgIStS, the transformational ap-natural selection I: .eve.loped m terms of adaptation, usually viatransformation otth thfs VIew, the ~entral problem in evolution is theresponse to I / genotype which occurs as organisms adapt ind'etre for h se ec. Ion pressures. Adaptation is the ultimate raisontive agent

CMantget.In gen~ frequency, Natural selection is the effec-

. u a Ions ultlmatel d .recombinat',onj' y, an venous mechanisms (such as

proximally suppl th" ,upon which sel t! ' y e rsqursrta panoply of variationec Ion acts Th! . th

sion of the transfo .' IS IS e core of the nee-Darwinian ver-the neo-Darwinian ~m:fl~~al ~octrin~. "Our purpose is not to attackhighly corroborated a ec Ism Itself-Indeed, we find it is probably asin biologY-but rath aS

tany other equally complex set of propositions

to the basic transtoer

°t.Show that an exclusive and rigid adherencerma ionar view h '. .

application of these as resulted In the Inappropriateena. Thus adapt tt co~cepts to all aspects of evolutionary phenom-

a Ion Via select' derffa explain the results of th Ion .un erltes most previous attemptscommented that such ada e ~volutlonary process. We have alreadypropriate to the scienn ctanonar scenarios are themselves not ap-reSUlts), Considered I IC study of evolution (either the process or itsrnational approach t as a. research strategy, an exclusively transfer-

o major problems of rates and modes of evolu-

I We use the term "tio"- neo-Darwinism" t f, n as tUl" Ultimate source of ' ..0 re er 10 Ihe body of theory that identifies muta-SIZes r' geneuc vanatlOn withi .are Imlted and that there' th ~ I In populatIons and states that populationral selection) within po,ulatio JS hie! ore a resultant pattern of differential reproduction (natu-of th " ns w ich sysreman alle vanalJon present in one ' IC Y alters gene frequencies since only someloose! def generation can be .Y <I<: ned body of work ' d present In the nex!. "Syntheticism" refers to aby came out from th 192a tendency to extra .....late D" e Os to the early 19508, marked especiallyters .......ific . ,,- neo- arwmlan " .,,,~J evolUtIOnary ,1<00 T nncJp es to encompass speciational and in-recentl . mena. hus the teh Y ~Inted oUI. To the elllent tha nns are not synonyms, as Simpson (1978) hasypo!hesls accOunting for ch . t we accept seleclion as the best available deterministic

remain nO' " ange In gene content d f '" eo- arwJnlaJ!s. The failure f an requency Within populations, weInto a testable theory of rna 01.0 the syntheticislS to inCOlporale speciation theorypresented here. croev UIJon leads us to avoid the term in describing the views


pertaining to genera and taxa of higher categorical rank) cannot beaddressed without explicit reference to species-their origins, his-tories, and patterns and modes of extinction. As we shall now show,nearly all attempts by syntheticists to deal with macroevolution, atleast until recently, fail even to mention species, much less to tncor-porate hypotheses pertaining to species into the analysis. In short,the transformational approach has given us a view of evolutionbased on the assumption that the entire tree of life is one huge,smoothly continuous gene pool. In this view, evolution is merely amatter of progressive change of intrinsic features and its productsare appropriately to be understood almost strictly in terms of adapta-tion and selection (although genetic drift enters the picture as a ran-dom component in some formulations). Discrete species become anembarassment instead of the very key (as we believe) to derivingmore accurate and more testable hypotheses pertaining to evolu-tionary phenomena at and above the species level.

Contemporary Macroevolutionary Theory:Issues and Transformational Explanations

Any summary and critical evaluation of a body of theory must havesome limits, however loosely defined. In this case, we construe "con-temporary" to refer to work performed during the past fifty years orso, especially the so-called synthetic theory. It is our bel ief that themajority of Western biologists interested in macroevolution still fun-damentally adhere to many of the basic tenets outlined in the 1930s,1940s, and 1950s. This is not to say that theory developed even ear-lier is uniformly uninteresting or without relevance, but merely thatthe more common notions held today were formulated and codifiedduring the past half-century or so, and were themselves partly basedon, and partly a reaction against, earl ier work. Nor do we suggestthat authors not explicitly numbering themselves among the synthe-ticists (or, indeed, explicitly dissociating themselves from the synth-eticists. or being disavowed by the latter) have had no importantcontributions to make. But there is no discrete "school" (even amongthe "sattationists'') that can be easily labeled and characterized. The


isolated writings of Willis (1940), Goldschmidt (1940), Schindewolf(e.g., 1950), Cuenot (1932), Rosa (1931), Grasse (1973), and Levtrup(1977), to name a few biologists who have addressed evolutionarytheory in more or less nonsyntheticist terms, are a most heteroge-neous array, united only by the fact that each deviates in one or moreways from orthodox synthetic ism, and au display a general tendencyto adopt a transformational viewpoint. In focusing on the syntheticschool (itself rather heterogeneous when examined closely), not onlydo we emphasize the most widely held views, but we are also ableto discern trends and styles of thought which are also present in thewritings of most of the "nonsyntheticists" as well. The point of thisreview of contemporary macroevolutionary theory is to pick out suchbasic modes of thought in addition to exposing the core of the theoryitself, in the belief that analysis of the epistemological structure ofsuch existing theory is essential to the development of a protocol forfuture scientific research in evolutionary theory.

Simpson (1944) spoke of "tempo" and "mode" as a useful di-chotomy of evolutionary phenomena. Tempos, of course, refer toevolutionary rates: genetic, morphological, and taxonomic, whereasmodes refer to styles of change (for Simpson 1944, there were threedistinct modes: speciation, phyletic evolution, and "quantum evolu-tion"). Specific topics usually considered as macroevolutionary in-clude all types of rates viewed as a problem among species andtaxa of higher categorical rank, and, especially, the very nature andmode of origin of taxa of higher categorical rank.

That species are real, i.e.. that species have ontological exis-tence in nature, and display beginnings, continued history, and dis-crete ends, is one of the more important premises of this book. Thattaxa of categorical rank higher than species do not exist in preciselythe same sense as do species is crucial both to an appraisal of pastand current macroevolutionary theory, and to our own elaboration ofsuch theory. What all taxa, from species up through kingdoms, doshare is presumed descent from a single ancestral species. Whatthey do not share are similar reproductive patterns. During the entiretime interval of their existence, species display patterns of parentalancestry and descent within themselves, but not without, as an over-whelming generalization. It is these units (species) that evolve,largely by producing isolated descendant units of like kind (morespecies). Taxa of higher categorical rank cannot be defined with ref-


outer-simulated taxa. Their work examines the ex~ent to which pat-terns of change in diversity within monophyletic groups reflectdeterministic (non-random) processes versus "random" processes.(See p. 298 for a more complete discussion of the meaning of theseterms and the significance of this research.) There appears to be asignificant nonrandom component to such changes in ?iversity pat-terns which alone suffices to show that a macroevolutionary theorymust' have a strong deterministic component. Although we will becritical of contemporary macroevolutionary theory in its insistence onreductionist transformational explanations, nevertheless many of theissues raised retain their importance and can, we believe, be ap-proached if properly restated.


erence to reproductive patterns (a within-species phenomenon) ex-cept for the trivial observation that the same disruption of parentalancestry and descent among species by definition applies to gen-era, families, and so on up the Linnaean hierarchy. Thus the defini-tion of taxa above the rank of species is identical for all ranks: amonophyletic group comprised of one or more species. Such taxa,ideally, at least, reflect our knowledge of life's genealogical history.As discussed in chapter 2, these groups of species can be recog-nized in a fairly objective and scientific manner. As hypotheses ofmonophyletic pathways of descent, they surely have relevance to ourelaboration, and testing, of hypotheses bearing on the evolutionaryprocess. But they do not evolve in the same sense as species, aim-ply because they are not the same kinds of biological entities. Theyare all collections of species. The evolution of a genus, an order, ora kingdom is fundamentally the same thing, amounting to varyingpatterns of origin and survival of component species.

The distinction between the nature of species, on the one hand,and taxa of higher categorical rank on the other, is crucial, simplybecause much has been written on the subject of the origin of taxa ofhigher categorical rank. And much of this literature is written underthe tacit assumption that such taxa possess an ontological statusover and above monophyly of descent. For example, Simpson (e.q.,1953:340 ft. and elsewhere) clearly recognizes the fact that taxa ofhigher rank are recognized ex post facto and thus are constructs ofthe systematist's mind, not existing in nature in any real sense; yethe has recently written (Simpson 1978:77) that" 'Species selection'is simply natural selection at a particular taxonomic level, but Stan-ley has emphasized well that major changes are best studied at this( or higher) levels" (italics added). This statement directly imputesan ontological equivalence in status to species and taxa of higherrank-the very confusion we believe has beclouded much of mac-roevolutionary theory to date.

If taxa of rank higher than species do not exist. or evolve, exceptas their component species do, we must seriously ask: what is theproblem? Is there any sense to speak of macroevolution at all? In aseries of recent papers, Gould, Raup, Schopf, and other colleagueshave investigated shapes of "clades" -i.e., patterns of relative andabsolute species diversity within both "real" (supposedly) monophy-letic taxa (using data drawn from the systematic literature) and com-

The Adaptive Landscape

In 1932 Sewall Wright introduced the concept of the adaptive land-scape. 'In its original formulation, the topographic.g~id of hills andvalleys served as a graphic, two-dimensional deplctl?n ~f th~ rela-tive "adaptive value" of various "harmonious combinations of anumber of loci, each with a number of alleles. Dobzhansky (1937and later editions; see also 1970 and Dobzhan~ky et al: 1977) has,perhaps more than any other author, utilized ~hls gra~hlc rep~esen-tat ion to pinpoint some of the more crucial Issues In ~volutlonarytheory. Dobzhansky and his coauthors point out that, wlt.h the two-dimensional landscape, the two axes "symbolize the alteltc variantsof two gene loci. The Darwinian fitness (adaptive valu~) of the com-binations of these alleles is symbolized by contours, like a topogra-phic map" (1977.168). Hence, they retain the force and gist ofWright's original metaphor, although the equatio~ of fitnes~ (d.e:lnedelsewhere in the book as differential reproduction) With individualvigor or survival is problematical. However, Wright hi~self (1~32)explicitly wrote in terms of relative survival value of atlelic combina-tions. At its core, the concept of the adaptive landscape allows one(but not necessarily Wright himself-see below) to pictur~ n~tura,~selection as directly favoring the "more harmonious combinations(peaks) over the less favorable allelic combinations, given the fullrange of such possibilities. .

The problem immediately arises: how does one (metapborlcallyspeaking) get from one peak to another? Or, as Wright (1932:358) put

.,


it: "The problem of evolution as I see it is that of a mechanism bywhich the species may continually find its way from lower to higherpeaks in such a field." Wright went on to elaborate a hypothesiswhereby "the course of evolution through the general field is not con-trolled by direction of mutation and not directly by selection, exceptas conditions change, but by a trial and error mechanism consistingof a largely nonadaptive differentiation of local races (due to in-breeding balanced by occasional crossbreeding) and a determina-tion of long term trend by intergroup selection" (1932:365). The im-portant point here is that Wright explicitly intended the "adaptivelandscape" to apply strictly and solely to various allelic combina-tions within species (especially within semi-isolated subpopulationsof species; see Wright 1931, 1945:415 and many of his other works).Above that level, some other process of intergroup selection not in-volving an adaptive-landscape motif was envisioned to be in opera-tion. Further, it is of interest that Wright emphasized a "trial anderror" mechanism, rather than either mutation or selection, as themajor means of travel between peaks. Later workers, most notablyDobzhansky (1937) and Simpson (especially 1953, after abandoningthe "inadaptive phase" of quantum evolution as conceived in 1944)agreed about the (non)role of mutation in this matter, but came downfirmly in favor of a model of selection as a means, not just of climb-ing peaks, but of traversing the valleys as well. We shall not pursuethe issue here, since it is a within-population and within-speciesproblem.

Or, at least, the issue was at the within-species level when origi-nally formulated and, in part, proselytized in Dobzhansky's severalbooks. Through the three editions of his Genetics and the Origin ofSpecies, Dobzhansky (1937, 1941, 1951) progressively augmentshis discussion and utilization of Wright's imagery. By 1951, Dobz-hansky had explicitly changed Wright's original context. applyingthe concept of the adaptive landscape to gene combinations amongspecies, families and on up the Linnaean hierarchy:

Theenormousdiversityat organismsmay be envisagedas correlatedwith the immensevariety of environmentsand of ecological nicheswhich exist on earth. But the varietyof ecological niches is not onlyimmense,it is also discontinuous.One speciesof insectmay feed on,for example.oak leaves,and anotheron pine needles;an insect thatwould requirefood intermediatebetweenoak and pine would proba-


bly starveto death. Hence,the living world is not a tormless~~ss ofrandomlycombining genes and traits, but a great arrayof famlll~sofrelated gene combinations,which are clustered on a large but finitenumberof adaptive peaks. Each living species may be thoughtof asoccupying one of the available peaks in the field of gene combina-tions. The adaptive valleys are deserted and empty. (Dobzhansky1951:9 fl.)

Dobzhansky then goes on, in the following paragraph, to extendthis generalization up the Linnaean hierarchy, concluding that: ":rhehierarchic nature of the biologic classification reflects the objec-tively ascertainable discontinuity of adaptive niche~, in other.word~the discontinuity of ways and means by which organisms that Inhabitthe world derive their livelihood from the environment." Thus is thewithin-species topographic conception of Wright translated into amodel for the diversity of life. The interesting point about these ratherbeguiling quotations from Dobzhansky is that. just as Wright saw forhis conception, they imply an immediate problem: the ce~tral ~rob-lem in evolution is, how does a taxon (i.e.. not a population With atrial and error means of arriving at relatively more harmonious allelecombinations) get from one peak to another? How are valleys tra-versed and hills climbed? In one neat and logical stroke, Dobz-hansky equated a within-species problem with the entire problem ofevolution and, more subtly, suggested that the answer to the one wasthe answer to the other. The key word, in any case, remains the same:"adaptive." Wright's "trial and error" of aile Iic combinations wastacitly dropped, and the way was cleared for a true synthesis of thedata of systematics (including paleontology) with the theory andresults of experimental and mathematical population genetics. If theconcept of the adaptive landscape neither completely accounts for,nor covers all topics addressed by "syntheticlst" writings on macro-evolution it at least clearly symbolizes the route (and it is a logicalone) thatwriters have overwhelmingly taken to bridge the perceivedgap between microevolution within populations and species, on ~heone hand, and the phenomena of evolution "above the specieslevel," as Rensch (1960) puts it, on the other hand..The c~nclusionthat all evolutionary phenomena were not only consistent With gene-tic theory, but could actually be explained by reference to such "fir~tprinciples," was obviously an attractive one. Our purpose in thischapter is not to say that such findings (e.g., in genetics) are in fact


inconsistent with the data of systematics, but rather that the data andtheory of genetics and systematics have been integrated in an inap-propriate fashion. And the reason for this is that a model, especiallybut not solely Wright's adaptive landscape, which was developed fora multiple loci, multiple allelic circumstance within species, is inap-propriate for wholesale translation to among-species levels-letalone to higher taxonomic levels.

Nearly all major works pertaining to modern, nee-Darwinian for-mulations of evolutionary theory address the problems of macroevo-lution to some extent. However, we agree with Mayr (1963:586) thatthe major works on what he terms "transepecific" evolution are thoseof Simpson (see especially, but by no means exclusively, 1944 and1953) and Rensch (1960, an English version of a work first publishedin German in 1947). An examination of their works is therefore inorder, along with ancillary writings by these and other authors.

Rensch (e.g., 1960:1, 57, and 97) repeatedly asks the rhetoricalquestion: are the processes accounting for formation of racesand species (mutation, selection, etc.) sufficient to explain trans-specific evolution "leading to the emergence of new genera,families, orders, classes etc., and hence to the formation of neworgans and new types of organizations" (p. 57)? He concludes that,whereas there are rules which govern trans specific evolution (mostlyenvironmental and "somatic" constraints on the possibil ities for ana-tomical modification), we basically must say "yes." In each of thesummations of his lengthy considerations of rates (p. 96) and cla-dogenesis (p. 279) occur statements such as the following (p. 279):"Summing up, we may state that the evolution of new structural typesand of new organs needs no other explanation than specific and ge-neric differentiation, i.e.. the combined effect of mutation and selec-tion." Thus it is quite clear that Rensch (1) views evolution fun-damentally as a problem of the transformation of anatomical featuresand (2) denies the existence of different phenomenological levels inevolution. The processes which Rensch discusses (e.g., alteration ofdevelopmental pathways) are (as in the case of Goldschmidt-seefootnote 4 to this chapter) undoubtedly relevant to a complete theoryof macroevolution, but only if considered in the proper context, i.e.with regard to the phenomenological levels developed later in thischapter.

Simpson (1944) set out with the avowed purpose of achieving a


"synthesis of paleontology and genetics." In a sense, he succeededadmirably, and no longer was there any rational doubt that the prin-ciples of genetics were somehow connected to, and consistent with,the results of the evolutionary process, bits and pieces of which areto be found in the fossil record. He also achieved his goal ofsuggesting "new ways of looking at facts and new sorts of facts tolook for" (1944:xviii) in the work.

The intellectual tradition in vertebrate comparative anatomy andpaleontology vis a vis notions of organic evolution began amid astrong revulsion which set the tone for nearly all subsequent discus-sions. Cuvier and Agassiz, both comparative anatomists and paleon-tologists, were outspoken in their opposition to any notion of evolu-tion. The stumbling block that anatomists have cited over and overagain is the problem of transforming one complex structure into an-other. They have been, in short, overawed by the complexity andsupposed magnitude of differences between fully developed struc-tures. They have worried about obviously inviable, but supposedlynecessary, "intermediate" structures, such as useless rudimentaryeyes, in statu neeceno;» Proevolutiontst paleontological writing ac-cordingly has been geared to counteract these objections, andSimpson's works (see especially 1944, 1953) are certainly no excep-tion. Simpson (1944:89) adopted Wright's adaptive landscape andextended it to embrace among-species phenomena before Dobz-hansky did, alleging that it portrays "the relationship between selec-tion, structure, and adaptation." Inasmuch as Wright (1932) (a) delib-erately downplayed the direct role of selection, (b) wrote ofmultiallefic loci, not anatomical structures, and (c) used the expres-sion "adaptive value" (i.e., survival value of particular loci) and not"adaptation," it is immediately evident that Simpson at the very out-set modified Wright's model for some larger purpose. The remainderof Simpson's characterization of this landscape is quoted in full:

The field of possible structural variation is pictured as a landscapewith hills and valleys, and the extent and directions of variation in apopulation can be represented by outlining an area and a shape onthe field, Each elevation represents some particular adaptive optimumfor the characters and groups under consideration, sharper and higheror broader and lower, according as the adaptation is more or less spe-

2. W. J. Bryan, as recently as 1925, raised similar objections during the Scopes trial, andthe theme persists in some creationist literature to the present day.


HittLinea,.selectionCenfrifu941 seJ.

Asymmt'fnc41centrl~tal

fi<!lcfio~fin9

Flgure 6.1 Simpson's depiction of the adaptive landscape. Hisoriginal caption follows: "Selection landscapes. Contours analogousto those of topographic maps, with hachures placed on downhillside. Direction of selection is uphill, and intensity is proportional toslope." (From Simpson 1944:90.)

clftc. The direction of positive selection is uphill, of negative selectiondownhill, and its intensity is proportional to the gradient. The surlacemay be represented in two dimensions by using contour lines as intopographic maps (figures 11, 12) [reproduced here as figures 6.1 and6.2]. The model of centripetal selection is a symmetrical, pointed peakand of centrifugal selection, a complementary negative feature, abasin. Positions on uniform slopes or dip-surfaces have purely linearselection. The whole landscape is a complex of the three elements,none in entirely pure form. To complete the representation of nature,all these elements must be pictured as in almost constant motion-ris-ing, falling, merging, separating, and moving laterally, at times morelike a choppy sea than like a static landscape-but the motion is slowand might, after all, be compared with a landscape that is beingeroded, rejuvenated, and so forth, rather than with a fluid surface.(Simpson 1944:89-90; repeated nearly verbatim 1953:155-56)

There follow immediately (Simpson 1944:90 ft.; 1953:157 ft.) sev-eral pages wherein this version of the landscape metaphor is ap-

-+Diredionof evolution

Figure 6.2 Simpson's depiction of.splitti~g phen~mena interms of the adaptive landscape. HIs anginal caption fol-lows: "Two patterns of phyletic dichotomy; shown on selec-tion contours like those of {figure 6.1]. Shaded areas repre-sent evolving populations. A, dichotomywith populationadvancing and splitting to occupy two different ad~ptlvepeaks, both branches progressive; B, dichotomy wl~hmarginal, preadaptive variants of ancestral pcoutatton mov-ing away to occupy adjacent adaptive peak, ancestralgroup conservative, continuing on same peak, descendantbranch progressive." (From Simpson 1944:91.)

I1


plied directly to equid evolution. It is a superb example of what wehave earlier termed a "scenario" (see also Tattersall and Eldredge1977). We here reproduce Simpson's original illustration presentedin conjunction with his discussion (figure 6.3). In this figure there arethree subfamilies (two of which are non monophyletic; see MacFad-

TAXONOMY: E9ulnae ~fJ/:::h- Hy,acoth~nin8e


den 1976, for a recent discussion of equid phylogeny), whose phy-logenetic trees and adaptations are simultaneously depicted as aseries of four successive adaptive topographies. The point to beemphasized here is that the discussion focuses on the shift from onepeak to another, or the old dilemma of the transformation of one

morphological configuration into another. The illustration (our figure6.3) is particularly concerned with the shift from browsing to grazing(browsers later becoming extinct) which, as Simpson notes, involvesmorphological features used to characterize and define the threesubfamilies listed on the diagram. The critical part of Simpson's dis-

cussion is quoted in full, as evidence of our characterization.

id1e MioceneIn the Eocene browsing and grazing represented lor the Equidae twowell-separated peaks, but only the browsing peak was occupied bymembers 01this family, That peak had moderate centripetal selection,which was asymmetrical, because one kind of variation, on one side,away from the direction of the grazing peak (teeth lower than optimum,and so forth) was more strongly selected against than on the otherside, in the direction of the grazing peak.

As the animals became larger-throughout the Oligocene, espe-cially-the browsing peak moved towards the grazing peak, becausesome of the secondary adaptations to large size (such as highercrowns, as previously discussed) were incidentally in the direction ofgrazing adaptation, Although continuously well adapted in modallype, the population varied farther toward grazing than away from itbecause of the asymmetry of the browsing peak. In about the lateOligocene and early Miocene the two peaks were close enough andthis asymmetrical variation was great enough so that some of thevariant animals were on the saddle between the two peaks. Theseanimals were relatively ill-adapted and subject to centrifugal selectionin two directions. Those that gained the slope leading to grazing were.with relative suddenness, SUbjected to strong selection away frombrowsing, This slope is steeper than those of the browsing peak, andthe grazing peak is higher (involves greater and more specific. lesseasily reversible or branching specialization to a particular mode ofHte). A segment of the population broke away. structurally, under thisselection pressure, climbed the grazing peak with relative rapidityduring the Miocene, and by the end of the Miocene occupied its sum-mit. Variants on the browsing slope tended by slight. but in the longrun effective, selection pressure to be forced back onto that peak, andthe competition on both sides from the two well-adapted groupscaused the intermediate, relatively inadaptive animals on the saddleto become extinct. Thereafter browsing and grazing populations werequite distinct, each differentiated in minor ways. The browsing types

PHYLOGENY

Eocene

Figure 6.3 Simpson's depiction of "major features of equid phylogeny andtaxonomy repre~e~ted as the movement of populations on a dynamic selec-lion landscape .. Simpson referred the reader 10the seclion of text quoted infull here for additional explication. (From Simpson 1944:92.)


eventually, at about the end of the Tertiary, failed to become adaptedto other shifts in environment and became extinct, while the grazingtypes persist today, (Simpson 1944:91-93; 1953:158-59)


nomena and he expressed the hope that it would serve as "one validclassification of basic descriptive evolutionary phenomena"(1944:197). Labeling the "practical study of changes in adaptation"as "the most essential single phenomenon of modes of evolution"(1944:189), Simpson explicitly relates his three modes to the "adap-tive grid." Each mode, moreover, is assigned a "level" (in terms ofthe Linnaean hierarchy, and apparently phenomenologically) whereit is most typical, though Simpson claims that each mode is not nec-essarily restricted to its most general level. Using the concept of"adaptive zones" (vconsideratfon of the environment as composed ofa finite, and a more or less clearly delimited set of zones or areas";1944:189), Simpson's tripartite classification of evolutionary modesfollows [see figure 6.4, a reproduction of Simpson's (1944) figure 31,p. 198].

The essence of this passage is clear: the evolution of taxa of rankhigher than species is to be explained by reference to the modifica-tion of those anatomical structures on which the recognition and def-inition of those taxa are based. Such changes are understood to bepurely the result of the action of various modes of natural selectionacting on a groundmass of phenotypic variation (itself a reflection ofunderlying genetic variation). Apart from the manifest difficulty onewould have in testing any of the specific statements about selectionvis-a-vis equid fossils, the statement is a logical extrapolation ofwhat was then, and remains today, generally understood to be thenature of the dynamics of within-population genetic change. Wemerely wish to point out two things about this passage: (1) it repre-sents the very essence of Simpson's approach to macroevolution-an approach which indeed does blend paleontology with geneticsand is also consistent with at least 100 years of previous paleonto-logical thought on the central problem of evolution (transformation ofstructure), and, more importantly, (2) nowhere in the discussion arespecies mentioned. Unlike Wright's admittedly terse comments onthe subject (see page 273), there is no recognition of the possibilitythat species exist and are to be regarded as units of evolution andas taxa distinct in kind from taxa of higher rank (elsewhere and indifferent contexts Simpson acknowledges much of this to be thecase; see also page 196). Simpson also fails to point out that at theamong-species level, the actual geometry of evolution might be radi-cally different, such that the bald extrapolation of Wright's within-species allelic imagery might be consistent with, but inappropriateas a simple depiction of, among-species evolutionary phenomena.Put another way, the nature (direction and intensity) of selectionwithin populations and species produces a pattern of within-speciestemporal variation which may well be different from, or even op-posed to, a net pattern of change resulting from differential speciessurvival. The assumptions in Simpson's approach ignore this posai-bility entirely.

As already mentioned, Simpson (1944, especially p. 197 ff.;1953) recognized three modes, or styles, of evolution. He pointed outthat his classification simplified the complexity of evolutionary phe-

1. Speciation: differentiation, usually a within-species phenome-non. In terms of the adaptive grid, it involves either differentiation ofa population into subzones of a single zone, or the elaboration ofnew adaptations, allowing later invasions into new subzones.

2. Phyletic evolution: the "sustained, directional (but not neces-sarily rectilinear) shift of the average characters of populations," amode "typically related to middle taxonomic levels, usually genera,subfamilies, and families. In relation to the adaptive grid, phyleticevolution is usually or most clearly seen as a progression of singleor multiple lines within the confines of one rather broad zone" (Simp-son 1944:203).

3, Quantum evolution: "the relatively rapid shift of a biotic popu-lation in disequilibrium to an equilibrium distinctly unlike an ances-tral condition" (Simpson 1944:206). Like other modes, it can give riseto taxa of any rank, but Simpson proposed quantum evolution as "thedominant and most essential process in the origin of taxonomic unitsof relatively high rank, such as families, orders, and classes," Interms of the adaptive grid, quantum evolution pertains to interzonalshifts."

3. Crucial to our argument here is the demonstration or a direct relationship betweenSimpson's version of Wright's topographic landscape and his own concept of the "adaptivegrid," which he characterized as follows: "The course of adaptive history may be pictorializedas a mobile series of ecological zones with time as one dimension. Within the limits of the flatpages. the basic picture resembles a grid. with its major bands made up of discrete smallerbands and these ultimately divided into a multitude of contiguous tracts" (Simpson 1944: 191).That the two sets of imagery were related emerges particularly from Simpson's discussion ofquantum evolution. as we shall discuss.

I

262 Systematics and the Evolutionary Process Systematics and the Evolutionary Process 263

tion, and shall further discuss the relative importance of these modesto the evolutionary process. The mode that Simpson proposed as re-ally critical in macroevolution----..quantum evolution-is a rather bril-liant conception designed to explain just how the distance betweenSimpson's version of adaptive peaks is in fact traversed. As stated. itis purely a transformational postulate. The equation of the pictorialconcept of the "adaptive peaks" and Simpson's own concept of the"adaptive grid" is neatly shown by Simpson's (1944) second discus-sion of the evolution of equid hypsodonty (see figure 6.5, a repro-duction of Simpson 1944, figure 34). Simpson says (1944:209): "Indiscussing equid hypsodonty it was shown that there are two mainadaptive zones, each embracing a shifting adaptive peak-one cor-responding with browsing habits and one with grazing habits." Therefollows a long restatement, in terms of the quantum evolution mode,of the equid scenario quoted above. He then states (p. 210): "On arelatively small scale, the distance from the browsing to the grazingpeak in this example is a quantum, a step that must be made com-pletely or not at all, although in the example and, I think, in generalthe genetic processes involved do not permit making the step with asingle leap," and the selective processes do not make the unstableintermediate forms inviable."

Simpson later (1953, 1959) modified details, but did not alter hisclassification of the three modes. And thus has the foremost Ameri-can student of macroevolution firmly equated the origin of taxa ofhigher categorical rank, in essence, with the problem of new andchanging adaptations-a problem of getting from one peak to an-other.e

Adaptive- ZoneSuhzone SuhlJJne

,,

A.spec:id lion Phylelic Evolution

4. This remark represents an evident side-step from Goldschmidt's (1940) concept of"systemic mutations .. , Goldschmidt saw the problem of applying concepts of within-populationvariation and selection as a direct explanation of macroevolutionary phenomena (to the point ofviewing speciation as a pure epiphenomenon. as did Willb 1940), but also saw macroevolutionpurely as a problem of morphological and genetic change-hence his ad hoc hypothesis of sys-temic mutations, Gould (l977a) and others have recently resurrected Goldschmidt's concept>(via regulatory genes, for the most part). an interesting and possibly fruitful action. providedthe mechanism is located within the correct phenomenological level. obviously that of develop-ing individuals within a population, 11is not our purpose to deny that evolutionary theory mustoffer an explanation of how morphological and genetic change occurs. but rather to establi,hthat such explanation is frequently not made within the proper context.

5. As is clear in the quote in the last pantgraph (pertaining to the quantum evolution ofequid hypsodontyj, although Simpson repeatedly referred to it as especially concerned with"higher categorical levels," quantum evolution is a phenomenon thai can only logically takeplace at the population and species level and can involve at most a series of species. Thus quan-

Quanfum £VDltrh"on T/m~----1.Sf"uc'l'u'e-

Figure 6.4 Simpson's visual characterization of the three majormodes of evolution, In which "broken lines represent phylogeny andthe frequency curves represent the populations in successive stages.(From Simpson 1944:198.)

t Exti,u''/"

The point of this brief recapitulation of Simpson's three modes isto s~ow their total dependency on notions of adaptation, with a con-C?mltant, underlying selection argument. Simpson apparentlyviewed speciation per se as an epiphenomenon; in any case, wehave already presented our views on speciation and phyletic evolu-

.. Interval of instabi/it>jBrowsintj • Grazin9zone zone


Simpson (1959) returned to the topic of macroevolution, repeat-ing his "conviction that the basic processes are the same at alllevels of evolution, from local populations to phyla, although the cir-cumstances leading to higher levels are special and the cumulativeresults of the basic processes are characteristically different at dif-ferent levels" (Simpson 1959:255). The italics are ours, emphasizingagain the confusion of phenomenological levels; the mechanismsSimpson cites can only be proposed to operate within populations.Quantum evolution is not mentioned in this later paper, but Huxley's(1958) concept of the "grade" is added to his considerations, andthe discussion remains firmly rooted and wholly steeped in adapta-tion. There is a greater emphasis on splitting phenomena (producingadaptive radiations, presumably by speciation), but the fundamentalconclusion." that "higher categories generally arise by acquisition ofa basic general adaptive complex" (Simpson 1959:270), continuesthe theme that evolution is a matter of adaptation via selection whichgoes on at the within-population level and on up through that ofphyla.

Huxley (1958:27) wrote of grades as being "just as 'natural' or atleast non-arbitrary as the customary monophyletic levels." They arelevels of anagenetic ecvaoce->!e.. groups of organisms defined bysome organizational (usually structural) improvement. Improvement,to Huxley, "covers detailed adaptation to a restricted niche, speciali-zation for a given way of life, increased efficiency of a given struc-ture or function, greater differentiation of functions, improvement ofstructural and physiological plan, and higher general organization.

AdapfiYephase

Preadapfivephase

Inaddpfivephase

tum evolution is, at base, a special aspect of speciation, an obvious consideration which mayexplain why Simpson (1976:5) and Boucot (l918) have claimed that "quantum evolution" isthe early and exact equivalent of "punctuated equilibria" (Eldredge and Gould 1972): the letterauthors equate punctuated equilibria with speciation. Earlier, Simpson {l953:389l stated thatquantum evolution is "a special, more or less extreme and limiting case of phyletic evolulion,"i.e. not speciation. In any case, those portions of the discussions of "punctuated equhbria" per-taining explicitly to modes and rates of morphological change deliberately avoid use of termssuch as "adaptation" or "natural selection" and are certainly oot concerned with ways andmeans of getting from one adaptive peak to another.6. We use the word "conclusion" loosely here. As Van Valen (l978) notes, Simpson's ideashave a way of retaining their plausibility and appeal despite the fact that the "data" upon whichthey are supposedly based have become outmoded or been superseded or outright falsified.Thus the genealogical hypotheses cited by Simpson as partial corroboration of his views arenot, in fact, used 10test the ideas at all. Rather, the ideas exist (perhaps originally suggested bypatterns of data), and examples that appear to tit are selectee to illustrate the concept. This con-trasts with the approach we develop later in this chapter.

Figua;e 6.5 ~impson's "phases of equid history interpreted as~~~~ Up~P~I~lutlo~. 'k: The phase designations refer to the part

zone" IF Sl~n rea mg away and occupying the grazing. rom Impson 1944:208.)


more careful, thorough presentation of the transformation of entirepopulations, forming single lineages or "phyla," be found than inSimpson's (1944, 1953) discussions of phyletic evolution. It is to benoted that Simpson's belief-at least in 1944-is that phyletic evolu-tion is most evident at the "middle levels" of genus. subfamily andfamily, a sort of "low macro." Directionality, including, but not re-stricted to, linearity, is generally considered the direct consequenceof directional selection (e.g., Simpson 1953: Hecht 1965). Thustrends.' especially in morphological features of fossils collectedfrom successive horizons, are most commonly considered as an ex-ample in geologic (or true evolutionary) time of exactly the samephenomenon on the generic and familial level as seen in directionalselection experiments within Drosophila population cages. Eldredgeand Gould (1972, 1974) and Gould and Eldredge (1977) have char-acterized this particular set of ideas at length. In addition to the theo-retical reasons why directional selection would not as a rule be ex-pected to persist over millions of years, it is apparent (see p. 283 ff.)that data pertaining to within-species variation in most examples oftrends in the fossil record in effect falsify the hypothesis that among-species phyletic phenomena, including trends, are a direct reflec-tion (and hence outgrowth) of within-species trends that might plau-sibly be attributed to natural selection. (The reader is referred tothese papers, and references cited therein, for further discussion ofthese points.) An alternative view, that among-species (and, ofcourse, taxa of higher rank) trend phenomena result from a processof differential species survival and are not directly attributable towithin-species patterns of temporal variation (whether directional ornot) has been available in the literature for many years, but until


At all levels it is the direct consequence of natural selection" (Huxley1958:19).This paper had a great influence on subsequent thought onmacroevolution. Many authors were quick to point out that such andsuch a taxon was not, in fact, a natural monophyletic group (i.e., aclade), but was, rather, a grade-a non-monophyletic group definedon the basis of convergences or on non-homologous "pseudosyna-pomorphies." Simpson (1959) himself wrote a paper claiming thatthe Mammalia were patently polyphyletic and that the defining char-acter (in this case, the anatomy of the jaw articulation) of Mammaliawas not homologous in all mammals. The multipl icity of papers ongrades in the late 1950s and early 1960s points up two things:(1) Many of the groups in conventionally accepted taxonomy of orga-nisms were not monophyletic; elimination of not-A sets (chapter 5)begins with their recognition, so the concept of grades marked ahealthy turn of events for systematics, and (2) The concept of adap-tation via selection, a view central to most previous discussions ofmacroevolution, was so completely dominant that the ultimate ex-treme was reached: generally thoughtful biologists could seriouslyconsider the supposed evolutionary processes lead ing to non-monophyletic (hence nonexistent in a genealogical, or evolutionarysense) groups.

The ad~ptational argument extends in detail to other aspects ofmacro.evolutlonary theory, involving directions of (morphological)evolutionary change as well as evolutionary rates. Perhaps the mostclas~lc, long-standing extrapolation of microevolutionary (within-species) .phenomena to among-species (macroevolutionary) phe-nomena Involves directionality. In no other aspect of evolutionarytheory is the confusion of within- and among-species phenomena~or~ clearly seen. The literature on directional phenomena in evolu-tion IS enormous, but there is little to be gained from a detailed con-sideration of particular examples. Suffice it to say that there is a fun-damental assumption, pervasive in neo-Darwinian theory, if not heldby all neo-Darwinists, that the same phenomenon-fitness dif-fer~nces among individuals within a population, or natural selection,whl~h by definitio~ alters the complexion of the gene pool from gen-eration to generatlo.n-can, when viewed over evolutionary time, beseen as a ~umulatlve process with rather large net effect. Amongpaleomoloqlsts, Gingerich (e.g" 1976, 1979) has been most elo-quent In recent years in defense of this view. And nowhere can a

7. All patterns, including trends. discussed in this chapter are considered to have at least threecomponents: (a) a deterministic component. We are interested specifically in the determmlSIlCelement of trends at this juncture; (b) a random component. We discuss stochastic processes andtheir role in the production of patterns, especially on page 298: (c) an error component. Thereare many sources of error in the perception of pattern. In this connection, we note earlier claims(e.g., Eldredge and Gould 1972:111) to the effect that "many, if not most, trends involvinghigher taxa may simply reflect a selective rendering of elements in the fossil record" (see alsop. 323). Salthe (1975:302) has gone further, suggesting that trends tvseres ,oj have no ontologi-cal status whatever, being in all cases solely creations of evolutionary biologists. We here as-sume that at least some major component of the palterns we call "trends" are de~nninist.ic ~norigin, i.e., not only that the patterns reflect an actual shift in value of some specified mtnnsicfeature during the genealogical history of a group, but also that at least some such shifts are the

result of deterministic processes.

Systematics and the Evolutionary Process 269morphological rates, which are in turn a function of genetic rates. Aswill be developed below, the converse possibility-that genetic andmorphologic rates might, in some sense at least. reasonably be hy-pothesized to be a function of speciation rates-emerges from therecognition of the dichotomy between within- and among-speciesevolutionary phenomena.

But the main point concerning rates is that, at base, it can beentirely appropriate to compare rates of genetic change a~on.gmonophyletic groups of any rank, regardless of magnitude. This IS

because hypotheses of evolutionary mechanisms are funda~~ntallya matter of mode, i.e., styles of change, including recognition ofunits of change and biotic and physical parameters moderating suchchange. To assert that species evolve, not anatomical parts, or alle-lic loci, of individuals, is not to deny that changes in gene contentand frequency result from the evolutionary process. As long as con-siderations of rate imply no necessary mode of change (e.g., phy-letic transformation vs. speciation), relative rates of genetic changeamong phyla may be studied. The problem of mixed phenome-nological levels arises only when comparative studies of morpho-logical and genetic rates are used to deduce generalizations aboutthe actual nature of the evolutionary process.


recently has been by far subordinate to the conventional view ofphyletic modification of entire populations via the direct action ofnatural setecttoo.e

The transformational view of macroevolution has also dominateddiscussion of evolutionary rates. Under the conventional definition ofevolution ("changes in gene content and frequency") it is natural thatevolutionary rates have been taken to mean "rates of change of genecontent and frequency." Taxonomic rates, a topic first fully devel-oped by Simpson (see 1944 and, especially, 1953) have frequentlybeen used as a means of estimating rates of genetic change. Ratesof diversification of taxa of higher categorical rank are common inthe literature of paleontology; frequency curves showing the numberof new taxa first appearing, or disappearing, or taxa persisting (the"standing crop") during an increment of geologic time are handydevices to summarize fluctuations in diversity of a group throughtime from the point of view of a given taxonomic level. It is unclearwhat else such rates of origin of higher taxa might mean, except assuccessively poorer approximations to actual speciation rates, thehigher up the Linnaean hierarchy one goes.

If taxonomic rates are viewed as a means of estimating actualgenetic rates, it is perhaps, but not necessarily, true that there is ahidden assumption of adaptation/selection working equally at all tax-onomic levels, i.e., that taxonomic rates are a function, ultimately of Transformational Macroevolutionary Theory:

Summary and CritiqueThe concept of natural selection (fitness differences, or differentialreproduction of individuals within populations) appea~s to be a cor-roborated, within-population phenomenon, and constltutes. the bestavailable explanation for the origin, maintenance, and p.osslble mod-ification of adaptations. Its corroboration results primarily fro.m labo-ratory and (mathematical) theoretical studies of populations. Inbridging the gap from within-species phenOmen? to patterns of tax-onomic and morphological diversity of the earth s past and presentbiota these principles have been applied wholesale across the tax-cnorn!c hierarchy, as though there were no difference between apopulation, on the one hand, and a kingdom, on the .o~her.There a~etwo fundamental objections to this approach, both aristnqtrom a fail-ure to integrate notions of the nature of species an.d their m?d~s. oforigin fully into the paradigm. First, if species are viewed as lndivid-

8. In his discussion of SlufenreiM, or "sjeplike" evolution, Simpson (1944:194, figure29; 195):220) notes that many, and perhaps most, epperent trencs reflect "relatively abundant,relatively static populations of successively occupied adaptive zones" (1944: 195). As shown inhis figure 29, the change within each "structural stage'" is irrelevant to the overall direction ofthe trend The phenomenon is not discussed at length, and structural stages (i.e., grades) areconsidered. rather than clades, but there nonetheless is a deviation from wholesale extrapolationof within-population and within-species patterns of (temporal) variation as a direct correlate of,and underlying mechanism for, directional among-taxon patterns (trends). Van Valen(1978:211) claims. without citation, that the notion of lineage selection has a long history,belllg used explicitly by Lyell, Darwin, Simpson and himself. To which it may be remarkedthat, in the works of authors as prolific as these four, there is usually to be found a little bit ofvirtually anything at all relevant to the topic. This is as it should be. But cenatmy. tbe main gistof Simpson's writings, to take one example, is far removed from any general notion of "lineageselection. '" Van Valen (1978:211) further states that such notions somehow imply that evolu-non within spl'Cies is unimportant, whereas it is just the other way around: the only phenome-nological level at Which adaptation (via selection) actually can occur is within species. It is onething to postulate that once a species has developed its own set of adaptations, selection tends toconserve it and quite another to state that within-species evolutionary phenomena are not impor-tant'


uals {i.e., discrete entities with origins, subsequent histories, anddefinite terminations), their evolution must be explained. Only a viewthat species are transitory, arbitrarily defined segments of an evolu-tionary continuum permits the notion that within-population phenom-ena may be extrapolated directly to higher levels. Recognition of theexistence of species as discrete entities in effect contradicts thevision of change in gene content and frequency-whether or not ef-fected by natural selection-as a continuous process from the popu-lation on up through the phylum.

The problem is particularly acute in terms of the concept of ad-aptation. We accept, for purposes of discussion, the basic notion ofadaptation, the adjustment of intrinsic features in response to naturalselection. As developed, it is a within-population, generational phe-nomenon. The problem arises from speciation theory: there is no nec-essary relationship between natural selection and speciation, atleast in terms of allopatric speciation (see Eldredge 1974a:542;"Though selection invariably plays an important role during any spe-ciation event, it has never been shown to be the effective "cause" ofspeciation in the sense that the selective regime originates and ismaintained for the "purpose" of developing a new species"). Asreviewed in chapter 4, Bush has pointed out that selection for repro-ductive isolation is a necessary component of sympatr!c and para-patric speciation. But nowhere in contemporary works on speciationtheory is the notion developed that speciation is fundamentally aprocess of adaptation. New adaptations, or the perfection of oldones, might be acquired, particularly in situations involving rela-tively small, peripherally isolated populations. But such adaptations,especially in allopatr!c situations, are incidental to the major phe-~ome~on of the establishment of reproductive isolation. Any changetnvolvinq behavior, morphology, or cytogenetics may be sufficient toeffect reproductive isolation. Such change may be adaptive as well,b~t to .conclude that speciation is a phenomenon of adaptation is adistortion of contemporary speciation theory. Speciation is a matterof establishment of reproductive isolation; adaptation via selection~ay or m~y not be involved incidentally. And, at least in the allopat-nc case, It cannot be construed that selection itself acted to createtwo s~ecies from a single ancestor, l.e.. that the prime "object" ofselection was the actual creation of two species from a single ances-tor.


From this point of view, the direct extrapolation of the concept ofadaptation via selection to explain differences in intrinsic featuresamong taxa of higher categorical rank cannot be appropriate. Theuse of the adaptive landscape beyond the limits of a single species,with its fundamental underlying premise that the within-species phe-nomena of selection and adaptation can be extrapolated in direct,unbroken fashion, violates the notion that the origin of new reproduc-tive communities (speciation) is not logically to be considered aphenomenon of adaptation. Rather, speciation breaks the smooth,within-population generational process of adaptation via selection.Thus the entire landscape metaphor is inappropriate for purposesother than the one for which Wright originally conceived it: within-species patterns of distribution of relatively more "harmonious" com-binations of alleles.

The second objection is that taxa of rank higher than species donot exist in the same sense as species exist. Species are reproduc-tive communities. Taxa-at least properly defined, monophyletictaxa--consist of one or more species connected by unity of descent.That is the definition of the word "taxon," holding for orders, classes,and taxa of all other categories. This consideration suggests thatgenera, orders, and so forth, do not evolve except as their co~p~-nent species do-that the patterns of fluctuation of diversity wlth.1ntaxa of higher categorical rank are a reflection of patterns of Origin,survival, and extinction of their constituent species.

Thus the notion of species as discrete units simultaneo~slydisrupts the ebb and flow of population continua (based on continu-ous, generational change) and collapses the study of macroevol~-tion down to a single phenomenological level When nature IS

viewed as a dynamic, functional system, the next step above thelevel of species is the ecosystem, not the genus, to parap,~ra~eSalthe (1975 and in press), who uses the terms "population a d"community" for "species" and "ecosystem," respectively. (See Val-entine 1969. for a similar view of natural hierarchies.) .

'. .. f t ary transformationalThis brief characterlzanon a con empor .'I t and was written pn-macroevolutionary theory has been ec ec IC .

marily to document the evident confusion in phenomenological, T d t 'he wholesale ex-levels that has plagued the subject. 0 ocurnen .

, . cross to higher {Lin-trapolanon of microevolutionary phenomena a .. nt pt of the adaptivenaean) levels, we have chosen Wng t sconce


landscape and some of the uses to which it has been put. We chosethe landscape imagery partly because it has played a crucial role inboth the analysis and the pictorial representation of these diversephenomena in terms of adaptation and selection, and partly becauseWright himself avoided the pitfall of such bald extrapolation. Thelandscape metaphor continues to provide the point of departure formost discussions of transspecific phenomena right up to the presentday (see, for instance, the chapter on transscectttc evolution inDobzhansky et al. [1977] and the somewhat equivocal usage of theimagery by Lewontin 1978). We must agree with Waddington(1967:14) who said: 'The whole real guts of evolution-which is, howdo you come to have horses, and tigers, and things-is outside themathematical theory [i.e., within-populational phenomena of changein gene content and frequency]." This is not because that theory iswrong, but because there are indeed discrete phenomenologicallevels glossed over by such across-the-board applications of popu-lation-genetics theory to among-species phenomena.

We have also parenthetically noted that most of the deviationfrom orthodox syntheticism has been equally plagued by the notionthat the essential problem of macroevolution is the explanation ofmorphological differences, and that the majority of such efforts havealso ignored the central importance that species have in the prob-lem. Other biologists, as we shall soon see, have perceived thesame set of problems. Moreover, resolution of the problem involvesclarification of the phenomenological levels themselves, The actualingredients of a revised theory of macroevolution consist of long-and well-understood biological principles, and require no inventionof new mechanisms. Our next task. after reviewing earlier work, is toset forth an alternative approach to the problem.


and Wallace in 1857. Most criticisms have thrown the baby out withthe bath water; of greater interest are statements and hypotheseswhich explicitly or implicitly recognize the phenomenological levelsof evolution, particularly those which have promulgated a concept ofintergroup selection.

Darwin's (1859) subtitle to his Origin of Species, "Or the Preser-vation of Favoured Races in the Struggle for Lite," implies a conceptof group selection (differential group survival) which was not. in fact,strongly developed within the book itself. Nonetheless. over the in-tervening years, hypotheses of intergroup selection have continuedto crop up (rather like hypotheses of sympatric speciation) and haveperiodically been attacked by biologists asserting the purity andprimacy of natural selection sensu stricto-differential reproductionamong individuals. Within the last 50 years or so, the time periodwith which we have been most concerned in terms of the historicaldevelopment of evolutionary theory, concepts of group selectionhave pertained mostly to interdemal phenomena. Interspecific com-petition and differential persistence-"species selection"-havefared poorly as general concepts, warranting, for example, a singlesentence in an otherwise generally excellent recent text on evolu-tlon.s

Largely ignored have been the occasional remarks of Wrightover the years. For instance, Wright (1931) clearly recognized thatspecies are the units of evolution ("the evolutionary process is con-cerned, not with individuals, but with the species" 1931:98) andwrote briefly (p. 154) of intergroup selection (interdemal) as beingimportant "in the origin of peculiar adaptations and the attainmen~ofextreme perfection." In the paper in which he elaborated the notionof the adaptive landscape (Wright 1932) containing the sentencequoted earlier in this chapter (p. 252), intergroup selection is rep~at-edly brought up, but again with emphasis on interdemal selectionwithin species. And in his review of Simpson's (1944) Tempo andMode in Evolution, Wright (1945:416) expressed support for ~heviewthat intergroup selection is "creative," illustrating his point with refer-ence to the elimination and "compensatory adaptive radiation" of"families and orders of vertebrates," clearly expanding his view ofgroup selection to the species level and far beyond-though he9. "Related species compete for resources that both are in need of, and one species mayoutbreed and crowd out another'" (Dobzhansky et ai., 1977:125).

The Phenomenological Levels of Evolutionand their Relation to Systematics

Previous Work

Dissatisf~ction with the concept of natural selection as the processof evolution has been evident since its initial espousal by Darwin


stressed that his main point remained inter-demic selection. He thenwent on (1945:416 ff.) to discuss interdernic selection based on altru-ism, the arena wherein much contemporary discussion (starting withHaldane 1932, according to Williams 1966:92) of group selectionpersists. Wright (1956:21 ff.) further discussed selection "among non-interbreeding species and higher categories," pointing out that(p. 22): "The raw material for such selection is even less random inan absolute sense than in the case of Interdeme selection since itconsists in the differentiation of these categories that has comeabout as a result of all of the preceding evolutionary processes. Thecompeting groups may, however, be looked upon as random trialsfrom the standpoint of the course of evolution of life as a whole thatcomes out of their competition." He further writes: "if there were noselection between such categories, we would expect to find all ofthose of an early geologic period persisting and all branching tosimilar extents." Inother words, the very nature of the historical recordindicates that differential species survival is a self-evident fact. Heconcludes his paper (Wright 1956:23) with the following remark: "Thecourse of evolution of vertebrate life and of life in general has beenguided throughout by a hierarchy of processes of selection rangingfrom selection between genes to selection between orders, classes,and even phyla." We would, of course, demur only with the claimthat differential survival of clades of rank higher than species repre-sents anything more than differential species survival.

Finally, as already quoted and extensively discussed by El-dredge and Gould (1972:111-12), Wright (1967:120) wrote: "With re-spect to the long term aspects of the evolution of higher categories,the stochastic process is speciation. This was treated as directedabove but may be essentially random with respect to the subsequentcourse of macro-evolution. The directing process here is selectionbetween competing species often belonging to different highercategories." Eldredge and Gould (1972:112, figure 5-10) utilizedthis notion as the basis of their reconciliation of the existence oflong-term trends (i.e.. net, directional changes in one or more mor-~hologi~al features within a monophyletic group through geologictlm~),. Wlt~ their rejection of the hypothesis that net, within-speciesvariation IS usually, or even commonly, directional through time.Stanley (1975) clarified and expanded this line of argument, refer-ring to "species selection," and concluded that inter-specific


evolutionary phenomena are in fact "decoupled," from intra-specificphenomena. Gould (1977b) and Gould and Eldredge (1977) usedthe expression "Wright's Rule" to refer to this aspect of the decoupl-ing of phenomenological levels. ""

Similarly, Grant (1963:397; 1977) has used the term In-terspecific selection" as an alternative to the more conventional hy-pothesis of "crthoselection'' as an explanation of directional, in-terspecific trends within monophyletic groups. Mayr (1963:586),addressing the topic of "transspectttc evolution," began his chapterby writing: "The proponents of the synthetic theory maintain that allevolution is due to the accumulation of small genetic changes,guided by natural selection, and that transspecific evolution (Rensch1947) is nothing but an extrapolation and magnification of the eventsthat take place within populations and species" (italics added).However, in the last three paragraphs {p. 621) of his chapt~r .ontransspeciflc evolution, Mayr clearly states the notion of specla~lon

, I ti . "Speciessupplying the raw material of variation for macroevo u Ion,. " . 'in the sense of evolution, are quite comparable to mutations'. Thisview is identical to Wright's (see especially 1967) and at variance,for reasons we have already developed at length, with the statement(quoted above) with which Mayr opened his chapter. .'

Hull (1976) has discussed hierarchical levels in conjunctionwith the notion that species are individuals but leaves open the

, tch t "possess suf-question (p. 184) whether monophyletic hlg er axa .. '. . ' I ti "His oonclusfon on theticient unity" to function as "units of eva u Ion. I .

, ronarv ch 'more than a summationtOPiCfollows: "If macroevolutionary c ange IS ., I more tnctusrve thanof rmcroevofutlonery events, then cornp exes . . "" 'I t! d count as tndlviduals.species might also form units of eva u Ion an

I· "more than aAlthough we disagree that a view of macroevo unon as .. . "I d to the conclUSion thatsummation of rnicroevolutlonary events ea s .

, his indiViduals andtaxa of rank higher than species are t emse ve . '" ' th t tus of species as In-evolutionary units" Hull's arguments on e s a

, " t f different levels ofdividuals have advanced the general concep 0

evolutionary phenomena, h thereApart from such theoretical state~ents and hypot. eses. rob-

has been relatively little explicit analysts of macroevolutionary IP I' h nomena are c ear ylerna in which within- and among-species p e . ,.

, d of such discnrm-differentiated One example of the use an power 'ad ti.. , I'ofthe'aaplvenation can be seen in Bock s (1970, 1972) ana ysts

Systematics and the Evolutionary Process 277276 Systematics and the Evolutionary Process

radiation" of the Hawaiian honeycreepers (Aves, Drepanididae). Incontrast to his earlier approach (Bock 1965)10 to the origin of birds,in which the emphasis was actually on the adaptation of avian flightand a gradual model of successive stages was developed, Bock'sanalysis of honeycreeper evolution (see figure 6.6) involved: (a) atheory of relationships among known (i.e., extant) species of Hawai-ian honeycreepers. Although no cladogram was presented, therewas clearly developed a hypothesis of degrees of relatednessamong the species, coupled with a hypothesis of primitive andderived morphologies, and generalized vs. specialized feedingtypes, The analysis proceeded despite the acknowledged possibleabsence of extinct (unknown and unknowable) species from thesample. The analysis further involved (b) explicit invocation (espe-cially in Bock 1972) of the concept of interspecific competition asthe source of the "selection force" underlying the origin of newmorphologies. The study is an excellent example, at least at the hy-pothetical level, of the recognition of separate phenomenologicallevels and discussion of the production of the hypothesized patternof evolutionary relationships among species of Drepanididae explic-itly in terms of the component species sampled from that family.Testability of such hypotheses is considered in more detail below.

There are, undoubtedly, still further examples wherein the notionof distinct levels, with species and speciation forming the crux of thematter, has been developed. However, it is manifestly clear that thetheme is not a dominant one in contemporary evolutionary theory,We believe it should be.

PseudonnlOl"~Q,,'IIQphfYI

H.I. hanoplp'

. fI nd hypothe-Flgure 6.6 Diagram illustrating head morpholOQYIn pro I e a a of Drepanidi-Sizedphylogenetic relationships of species Within three generdae. (From Bock 1970:714,)

Methodological Considerations, tence of species di-We have emphasized our view that the very exts .

, ' t ature The propertiesrectly implies a hierarchical organization 0 n . ", I "decouples (Stanleyof species as discrete entities, effective y

, , I ti nary phenomena.1975) within-species from among-species evo u 10 . .'. I t evolution' within-Thus there are two phenomenological leve so· I" ")

, cy ("microevo U Ionspecies change in gene content and frequen . ., . .' nophyletlc group Inand change in species composition Within a mo

space and time ("macroevolution"). h nisrnIati ) is the mec aOrigin of new species from old (specra Ion . d

' I f microevolutlon anwhich effectively decouples the two leve s 0macroevolution. It should further be noted that other

AP, ,hhenol mvee~

, 'I tlonary theory, e enologlcal levels are also pertinent to eva u I .. . f. .' . that the modification 0of the Individual for example, It IS obVIOUS t,'

, t i trinsic ea ures 10developmental pathways underlies all change 0 10

10, As this book was going to press, Bock published 3 review of macroevolutionary theory(Bock, 1979). His position in this paper is thoroughly and exclusively transformational. AI.though he initially defines macroevolution (p. 20) as "the appearance and subsequent speciali-zation of distinctive new features and taxa, ,. neatly combining and confounding the transforma-tional and tallic elements, the remainder of the paper views macroevolution as "simply a largeamount of phyletic evolutionary change." (Bock 1979:36, under the heading "Speciation andMacroevolunon '.J. To accomplish his avowedly reductionist task of explaining macroevolutionstrictly within the terms of microevolutionary processes, Bock (p. 28 ff.) concludes that the bi-ological species concept [e.g., one like Mayr's (941), cited here on p. 92] has no timedimension: phyletic lineages exist, but have no properties as discrete units or entities in time.Denying the ontological status of species as real, discrete units in time and space is, of course,a logical necessity for all neo-Darwinian or symheucist versions of transformationalism (seeEldredge 1979b:8, for a discussion of this point). Bock's conclusion is therefore inevitable(1979:39): "The key to all explanatory models of macroevolution is the concept of biologicaladaptation." We conclude, in contrast, that Bock's paper (1979), although citing his work ontbe Depanididae, fails to appreciate ils full significance and instead reverts to the more clas-sically conceived transformational approach seen in his earlier work (e.g., Bock: 1965),

!

-


phylogeny. At the cellular level, the physicochemical constraints andinducers of codon changes become relevant And so on. Any com-plete theory of evolution must deal with all definable levels, as wellas with their independence and interdependence. In this book, weconfine our attention to micro- and macroevolution, for the simplereason that the data of systematics are pertinent to these levels. In-deed, as we shall soon see, in terms of the analysis of processesacting within levers, the data of systematics pertain solely to dif-ferential species origins and extinctions-the dynamics of macro-evolution.


{i.e., level) other than that for which it is appropriate-fails utterly.Applications of the concept of natural selection to among-speciesphenomena are therefore inappropriate on methodological groundsbecause the data are themselves not appropriate to evaluate thespecific hypothesis at hand.

Thus it follows that there are also methodological grounds forrejecting adaptation as a fruitful way of conceiving of and address-ing issues in macroevolutionary theory. This statement holds, ofcourse, only if adaptation is strictly and exclusively tied to naturalselection, a conventional view which we find highly corroborated inpast considerations of the subject. Accepting this relationship, itwould appear that adaptation-adjustment of behavioral and ana-tomical traits to perform explicit functions with respect to biotic andabiotic parameters of the environment-is effected at the within-species level. Inasmuch as it is populations which are integratedinto ecological communities, the conclusion seems ineluctable thatstudy of the process of adaptation (true, evolutionary adapta-tion-not just physiological adaptation which goes on at the level ofthe individual) is fundamentally an ecological and experimentalproblem. For its scientific study, genetic data and well-corroboratedhypotheses of functional morphology on successive generationsover a number of years are required. The requirements are, therefore,stringent, but perhaps not impossibly so. The process of adaptationsimply cannot be applied to, or studied from the vantage point of,the higher phenomenological level of the evolutionary process thatwe call "macroevolution." For these reasons, in addition to thosecited above, we should therefore drop adaptation as the focus of ourresearch into macroevolution. Adaptation (via natural selection) is amost important process in evolution-but it is a within-populationgenerational phenomenon that requires data unavailable to all pa-leontologists and to most systematists as well. As Salthe (un-Published manuscript) has remarked, the dynamics of within-popula-tion phenomena are the best understood of any of the hierarchicallevels; they are best left to population biologists, especially ge-neticists and ecologists, and will not be discussed further here.

Microevolution We have already commented at length that the twinneo-Darwinian concepts of adaptation and selection, which arisefrom and are appropriate to the population level (Le., within-specieslevel) have been inappropriately extrapolated to the higher level ofmacroevolution. We have also expressed our agreement that naturalselection, despite persistent criticism, is an actual dynamic processin nature. Within populations there is inevitably a sampling of theavailable genetic variability of a given generation represented in thesucceeding generation. Some of this sampling, at least, may be non-random-a reflection of differential reproductive success within theparental generation, linked in some way to relative survival value ofa given behavioral, physiological or morphological trait.

But natural selection is strictly a within-population phenomenonby definition (see Salthe 1975). Nearly all criticisms of natural selec-tion are based, in the final analysis, on the perception that explana-tions of evolutionary phenomena frequently invoke natural selectionat the wrong phenomenological level. Natural selection should onlybe hypothesized under conditions in which it can be tested directly.Data required are comparative gene frequencies in a parental andone. or more descendant populations. By definition, the concept isdesigned for, and limited to, within-population situations. It shouldnot be surpri~in~, therefore, that the concept can only be applied bythe evolutionist In such situations. At higher levels, the use of a con-c~pt such as natural selection (e.q., as in Simpson's equid scenariocl.ted and quoted above) is inappropriate, both conceptually as justdls.cusse.d,. and also epistemologically: in an investigative protocolw.hlch minimally demands that hypotheses be susceptible to cnti-cram. the concept of natural selection-if applied to any situation

Speciation The origin of the reproductive units we have been callingspecies has already been discussed from the standpoint of system-atics earlier in this book. We developed our views on the nature of


speciation at that juncture because of the evident link between no-tions of speciation, on the one hand, and the construction of phy-logenetic trees on the other. To convert a cladogram into a tree, onemust specify ancestors at branching points (chapter 4) and a discus-sion of modes of speciation is of heuristic value, relevant to an anal-ysis of the numbers and nature of phylogenetic trees that may be ob-tained from any given cladogram.

Here, we briefly stress what is, in effect, the converse: that thestudy of speciation requires, among other things in its data base, ahighly corroborated phylogenetic tree linking all taxa (species, "sub-species," and perhaps even populations) in some definite ancestor-descendant pattern. In other words, a detailed reconstruction of thehistorical relationships of two or more taxa at or below the specificlevel is the sine qua non of all analyses of speciation. Temporal andgeographic distributional data-c-or hypotheses-are also requisites.

Although most studies of speciation list many examples (pat-terns of relationship and distribution which appear to confirm or il-lustrate particular aspects of speciation theory under discussion), inpoint of fact, the only means whereby speciation theory can betested adequately is by comparison of predictions arising fromtheory with patterns (trees plus distributional information) worked outindependently from that theory. That this is a tall order quite difficultto fulfill in most instances arises both from the intrinsic difficulty oftesting phylogenetic trees themselves (chapter 4) and from the near-impossibility of constructing these trees without reference (con-scious or unconscious) to some theory of speciation or another. Aswe have already discussed extensively in chapters 3 and 4, one'svery concept of species limits the choice of speciation theories toone subset or another of all those available and makes it almost cer-tain that such a theory will be held a priori in the mind of one who istrying, nonetheless, to study speciation.

Thus, study of speciation offers some interesting contrasts to thelevels of microevolution below it and macroevolution above it Su-perficially, the testing of speciation hypotheses requires "data"simpler in detail than those required for microevolutionary study: forthe former, we need a highly corroborated tree plus distributionaldata, whereas for the latter, we need detailed, generation by genera-tion data pertaining to changes in gene content and frequency. Butgiven the highly dubious nature of phylogenetic trees, as hypotheses


difficult to test (i.e., of relatively low corroborability), the data basefor testing hypotheses of speciation is in reality a great deal morecomplex than that required for testing microevolutionary hypotheses.

Another contrast between the phenomenological levels wouldappear to be the order of magnitude of time required for the pro-cesses to work. At first glance, the most simple generalization wouldbe that rnicroevolution is a generational phenomenon, and specia-tion involves origins of new units from old, each made up, minimallyof one, but realistically of hundreds and usually thousands of gen~r-ations. Macroevolution, invoking differential survival of specieswithin monophyletic groups as its dynamic process, involves all pos-sible lengths of time, bounded only by the age of appearance of thefirst species of that clade. But intergenerational wlttun-specres (rni-croevolutionary) phenomena, technically at least, go on and.can ac-cumulate for the entire length of time an individual species per-sists-many millions of years not, apparently, being uncommon,although we have noted (chapter 4) that, to judge from patterns ?fphenotypic modification, relatively little change tends to accrue msuch instances. In contrast, speciation may take only a few years,and perhaps normally requires only a few thousand years, to takeplace. Thus there is no simple relationship between the phenome-nological level and amount of evolutionary time required for e.vents

. . I te Accumulation ofto occur, or for the additive effects to accumu a . .macroevolutionary change takes longer than accumulation of rtu-

croevolutionary change simply because the cumulative existence.of. . . II ater than the durationtwo or more species IS almost autornatrca Y gre

of any single species. .We conclude that the scientific study of the process of specia-

tion demands, as its data base, the existence of well-corroboratedtheories of relationship among populations, or closely related spe-

. . "serm-cies (or allopatric taxa whose precise status IS moot, e.p.. .. . th Ives add an unavoid-species"). Problems in testing the trees emse _

. h t di s a methodologicalable element of uncertainty to all suc s u re , .. . . . . ti hypotheses. For thislimitation on our ability to assess compe mg

. b II understood as arereason, speciation probably Will never e as we ... . I I of microevolutlon.the processes at the lower, within-species eve

tit t the nextMacroevolution Among-species phenomena cons I '' ehigher phenomenological level above the within-speCies level. The

-----~------------------------- --_.


dynamic process acting at this level is taken to be differential spe-cies survival within monophyletic taxa. Data required to test specifichypotheses of macroevolutionary phenomena, under this conceptionof the level, would appear to be temporal and spatial distributionaldata of component species, plus well-corroborated cladogramsshowing these species as members of a monophyletic group. Formost hypotheses, detailed trees appear not to be required,

Inasmuch as there is only a relatively small number of publishedhypotheses of this nature available, it seems relevant at this point toconsider two possible courses: (a) generalization {"laws") arising di-rectly from patterns of differential species survival, or (b) invention ofhypotheses of process which can generate predictions about pat-tern. against which such independently analyzed patterns can becompared. The first would be strictly an inductive process, while thesecond contains a deductive element. Whether the difference be-tween the two is of any ultimate significance, the latter course, offer-ing direct means of criticizing hypotheses, seems preferable, albeitmore difficult.

That the testing of macroevolutionary hypotheses requires only acladoqrarn whereas testing of those pertaining to speciation per serequires trees, may at first seem surprising, inasmuch as speciationwould therefore require the seemingly more complex hypothesis ofpattern. But taxa of higher rank than species do not exist in the sameway as do species (p. 249), and, as a corollary, such "higher" taxacannot logically be ancestral to one another, Thus, even on this levelof analysis, it is a false issue to speak of the "evolution" of theEquinae from the Anchitheriinae. Thus we require branching dia-grams involving species themselves, Since trees can only give usdetails of precisely how any two species may be hypothesized to berelated, it would be superfluous, at least in most instances to con-sider whether a pattern 01 differential species survival involves an-cestors vs. descendants, rather than, say, "plesiomorphous" vs."apomorphous" species. A further methodological consideration isthat to formulate macroevolutionary hypotheses requiring treeswould be, in effect. to require the assumption that all species thatever existed within the monophyletic group were represented in theavailab!e sample under study. We conclude that there are very realconstraints on macroevolutionary hypotheses: to be testable at all in-sofar as phylogenetic patterns are concerned, they must deal only


with branching diagrams of corroborated hypotheses of monophy-letic descent (cladograms) and they must deal with species.

There are some hypotheses at the macroevolutionary levelwhich can be most conveniently tested with data pertaining to taxaof rank higher than species. Such data (e.p., those pertaining to ap-pearance of new, disappearance of old, and "standing crop" of gen-era, families, orders, etc" usually given in some graphical form) areonly directed to such hypotheses as here conceived when it is as-sumed that such data are successive approximations away from (thehigher up the hierarchy one goes) actual species diversity data.Much has been written in recent years about the use of generic andfamilial data, which are thought to allow us to estimate true speciesdiversity in the fossil record, Use of generic and familial data, i~ isargued, helps avoid the loss of direct sampling of many rare specieswith restricted temporal and geographic ranges, but is of lowenough rank to minimize the distortion of true species diversity (seefigure 6.7). (This consideration, of course, applies only to compara-tive diversity among groups with an equal and reasonable chance offossilization in the first place.) When such data are presented as anestimate of species diversity data, they are legitimate to use In theevaluation of macroevolutionary hypotheses. As we have alreadynoted, use of such data to evaluate inappropriate hypothese~

. .. "higher-level taxaregarding putative evolutionary processes amongis not legitimate.

Relationships among the Phenomenological Le~elsof Evolution: Decoupling

. . th t the phenome-In this section, we shall reiterate the conclusion a ""decoupled atnoloqical levels of the evolutionary process are '

. . terence from oneleast to the extent that direct extrapolation or In ..... that by deflnlllOn, nat-level to another IS not possible, We are aware ,

., .' n and thus cannotural selection IS a within-population phenomena,. . I At' sue at the moment,be applied appropriately to higher leve s. IS . .

Id I cepted reduction 1Sthowever, is the counter position, the WI e Y ac. h d that patterns of rna-thesis that the levels do not exist as sue an


'00 r---;;';-'~---'---'--.1--'--'-_-l.-,,.,.,:.,'1'I' :'

I :',. f

i fi :iI


ton within and across taxa (a possible example will be cited below,however), the hypothesis that the levels exist and are in fact decou-pled might be restated as follows: patterns of within-species variationare not the same as those among species within a monophyleticgroup. This concept is illustrated in figure 6.8. If time-averaged pat-terns of variation within the span of existence of an ancestral speciesare of the same nature (have the same direction through time) asthose of its descendant (figure 6.8a), the observation is consistentwith the hypothesis that among-species differences arise smoothlyand continuously from patterns of within-species variation. We has-ten to add that identity or strong similarity of such directions doesnot utterly reject the existence of discrete levels. As Gould and El-dredge (1977) have argued, coincidence of within- and among-species patterns of variation within a hypothesized series of ances-

,,,,,,._-..;, I

A ,-\ /i i ,'- • • r

-""'.' ! II '

Classes<,

I~Orders

l · _ »:" .r -.f"'" ..J. \'. • II! families _\ I

/ ! , ,/ ,! i ;

, II ; ".....

/ iI,

.,.,.,/ •.................•.....

...............•

,,,,, -,,,,,//,//•:

Species --;I,

"

,1,1

~"~".. ,'''\.......- .... ...Q - •• _-_ ...... - --- ' .. _-_ .....

,!!!

!! a. b. c d

Species I~"'" 3

Species :3 ~,o~

• \ Species 2 \~,."2E Species 2>=

Species 2 /Sp,,,,, I \ Species , \""" ,

,·······•·•·····•.• Genera ./

'\ r.... ./'.............

,,'."

Camb. Old. Sil. Dtv. Carb. Perm.Tri. Jur. Cret" TertFigure 6.~ Comparison of apparent Phanerozoic diversity curves as mea-sured at different taxonomic levels. Only well-skeletonized marine inver-tebrates were tabulated. (From Raup 1972:1070. Copyright 1972 by theAmerican Association for the Advancement of Sctence.)

croevolution are a direct reflection of within-population dynamics ofchange

Stanley (1975) has discussed at length the testing of these rivalhypothese~ ("?eCOupling" of discrete phenomenological levels vs. adlre.ct continuity between within- and among-species patterns of se-lection),. The best example of the latter form of hypothesis, phyleticgradual,l~m, asserts. inter alia, that long-term (including in-t~rspe.clf.IC, intergeneric, etc.) trends are the direct product of selec-tion WIthin species (true natural selection). Inasmuch as there are nodata sets known to us that can definitely be shown to reflect setec-

-----------..<II Morphological Change

Figure.6.8 Relation between within- and among-species tim~-avera~ed .patterns of variation. (a) A hypothetical situation in which wlthm-specles.di-~ectionof change through time is oonttouous and unbroken across species~oundaries." This type of change is consistent Withthe hypotheSISthat

Within-species variation and direction of change are responsible for amono-Species patterns. (b) A situation in which the fossil record of t~ree speciesshows within-species patterns similar to the total arnonq-specreetrend, .llke-WiseConsistent with the above-stated hypothesis, but also consistent Withthe hypothesis that the levels are decoupled. (c.d} Directed, net change (atrend) among ancestors and descendants where the net change amongspecies is inconsistent with the hypothesis that such trends are the directproduct of Within-species patterns of directional change.


tors and descendants may still reflect a "decoupl ing" phenomenonat speciation, the resumption of the trend within the descendantspecies reflecting pure chance (the stochastic element, discussedparticularly on p. 298 ff.}, or the resumption of a pattern actuallydisrupted by the wholly unrelated event of speciation (figure 6.8b). Ifpatterns of time-aggregated variation, especially in those charactersused to distinguish the related species, are "neutral" (i.e., there is nonet change in mean value, figure 6.8c) or in the opposite direction(figure 6.8d) of the pattern of net directed change among species,then the hypothesis that among-species differences arise as a mere,and direct, extrapolation and accumulation of within-species pat-terns is effectively rejected.

Such tests are inherently weakened by the basic assumptionthat directed patterns of net aggregated variation through time in factreflect selection. They are further weakened because they deal withthe critical level of speciation, and therefore ideally require highlycorroborated trees. The proposition can perhaps be generalized tostate that relatively apomorphic species within a monophyletic groupshould not exhibit patterns of within-species variation coincidentwith such patterns within plesiomorphic sister-taxa. If the hypothesisthat with in- and among-species patterns of variation through time aredifferent is valid, then the pattern should hold for monophyleticgroups whose relationships are depicted on cladograrns. We nowpresent three examples from the fossil record. In each case, taxawhich are putative descendants, or simply putative apomorphicsister-species, show a pattern of within-species variation radicallydifferent from that of the putative ancestor or plesiomorphic sister-species.

Makurath and Anderson (1973) analyzed the phylogenetic his-tory of the Upper Silurian and Lower Devonian species of the pen-tamerid brachiopod genus Gypidula in the Appalachian region. Bas-ing their conclusions on statistical analyses of two morphometricvariables (beak length and spondylium width), the authors con-cluded that over three successive stratigraphic horizons, evolu-tionary change of the two anatomical features was smoothly continu-ous both within and among the two species. Eldredge (1974b)criticized the experimental design of this study, concluding that thedata were inadequate for a true test of the rival hypotheses of"gradualism" vs. "punctuated equilibria." Of interest here is the


,.. •••.. ..LONG• NAnow

2 ,, ,,,,1 ,,

" ,,, ",

1~3,11131

33DF 2 • , ,

11313333, , 3333 33 ,, r , , , J--.--, " " ,-'-, , , ," ,,,, , ,,-2 ,

SHORTWIDE-. SMALL

•2-. -2 ,-, .DF J

FIgUre 6.9 Within-sample patterns of variation of Gypidula prognostica(group 1) and G. coeymanense (groups 2 and 3). Specimen scores are plot-led against the first and second discriminant functions. Large numerals ~rerespectivegroup centroids; small numerals are individuals. Grou~ 2 individ-uals are omitted. We have drawn lines around all individuals Within agroup.indicating the essentially perpendicular relationship between the major axesof the ellipsoids of dispersion for groups 1 and 3. (After Makurath and An-derson 1973:308.)

point that a bivariate plot of individual specimen scores against thefirst two discriminant functions clearly indicates that the within-sample pattern of covariation of the two variables is in fact radicallydifferent within each of the two species (see figure 6.9). The shape ofthe ellipsoid of dispersion for the sample of Gypidula prognostica(the Putative ancestor of G. coeymanense) indicates a trend to "long.narrow" shape with increase in size, whereas in the one sample of G.Coeymanense plotted in like fashion, specimens become "short.Wide" with increasing size. It is unclear whether the size-correlatedtrends Within popUlations are strictly an ontogenetic phenomenon, ora simple Covariance of a trend with variation in size, or a mixture of


G.(Glabrocingulum) G. (Anonias) G.(newsubgtJnusA) G.(new subgenus B)


the two. Nonetheless, enough analysis is presented to indicate thatthe anatomical features of the brachiopods are put together in ratherdifferent fashion, suggesting that straightforward, linear modifica-tions of the structures within species is not a model for the transfor-mation of the characters among species.

Another example, drawn from unpublished measurements onsome undescribed species of Permian pleurotomariacean gas-tropods, even more vividly illustrates differences in within-samplepatterns of variation between ancestors and descendants. In the fol-lowing example, patterns of pooled within-species variation alongfactor analytic axes are compared with among-species patterns. Thetrends are thus not the same as stratigraphic trends, but nonethelessbear on the issue at hand. In any case, the eight species involvedbroadly overlap in their stratigraphic distribution. We are grateful toDr. Roger L. Batten for supplying us with the data for the analysispresented here. Batten (unpublished manuscript) has recognizedtwo species of the conservative gastropod genus G/abrocingulum(G/abrocingulum) in the Permian of the southwestern United States.In addition, there is one species of the derived subgenus G. (Anan-ias) and two species in each of two new subgenera of G/abroc-ingulum. These latter two subgenera are, insofar as is known, en-demic to the Permian of this region. They, like Ananias, differ fromspecies of G. (Glabrocingulum) in being more highly spired. Thesehigher-spired taxa differ among themselves largely in terms of dif-ferences in whorl shape. In terms of the mathematical parameters ofmolluscan shell growth (see Raup 1966 for a review), the variationwithin and among the taxa under study involves predominantly (a)rate of translation of the generating curve along the axis of growthand (b) changes in shape of the generating curve itself. Batten's rec-ognition of taxa is based on a qualitative assessment of these andother features, such as shell ornamentation.

G. (G/abrocingulum) is virtually cosmopolitan and its includedspecies occur in Mississippian, Pennsylvanian, and Permian rocks;G. (Ananias) is found over broad areas in Pennsylvanian and Per-mian rocks. As stated above, the four remaining species (in the twoas yet undescribed subgenera) appear to be Permian endemics ofwhat is now the southwestern United States. Thus, at most, we canconstruct a cladogram for all species involved, knowing full well thatspecies from other times and places, omitted from the present analy-

Figure 6.10 Cladogramof relationshipsamongthefoursubgeneraofGlabrocingulum discussed in the text.sis are involved in the phylogenetic history of at least a portion of

, , B tt that the two un-the group discussed here. But we concur with a en . f Gdescribed subgenera plus probably the undescribed species 0 .

, . h i t is the apomor-(Ananias), form a monophyletic group whic I~ urn Th I do-phic sister-group of the subgenus G. (G~ab~OCmgulum). e c agram implicit in this statement in shown In figure 6.10.

. B t we may reason-Thus we have not obtained a precise tree. u tf . tion in the relevanably inquire whether or not the patterns 0 vana I .

, 1 h I siomorphlc subgenusvariables within the various species 0 t e p e f aturesG. (Glabrocingulum) agree with variation in homolo~ous e

f pomorphlc subgenera.within and among species in the group 0 a . 01t me 187 specimensWe have subjected linear measuremen s on so rrnalized by

, h the data were nothese taxa to a factor analysis. w.ere . nd subsequently, byvariables (to give each variable unit variance) a., h specimen

, tn ef ct equal slze--eaccases (giving each specimen, In e e , asurementsbeing represented by a vector of unit length). Twe~ty ~~hS of whorl~including spiral angle, were taken on heights .an WI 'dth) on each

( g selemzcne WIand homologous whorl characters e.. , I' I features andspecimen. (For a similar study, where morpho oqrcameasurements are defined, see Eldredge 1~68.). d (i e redun-

d ere mirror-Image .'The first two factors extra?te w. . Ie taxic discrimination.

dant) and produced a parabolic plot with Iitt . t those of tac-lotted aqameWhen specimen scores for factor I were p e neatly and

. d by Batten wertor III, the four subgenera recognize. 6 11) a simplified plot ofclearly discriminated. We here show (fiqure . individual specimensthe clusters of each of the subgenera, with

290 Systematics and the Evolutionary Process Systematics and the Evolutionary Process 291

plex can be examined. Both species of G. (Glabrocingulum) displaya large range of variation, delineating an ellipse of variation whosemain axis denotes a relationship of scores of factors I and III with aslope of approximately (-1), i.e., there is a high negative correlation,within these three species, of scores on these two factors. Thustaller, more narrow shells within this group have, if anything, rela-tively smaller upper whorl faces.

Remaining at the within-species level, it is also apparent thatwithin each of the five species within the three high-spired subgen-era there is also a strong correlation between the two factors, but inall of these cases, the relationship is positive: variation in shell heightand narrowness is positively correlated with relatively large valuesfor the upper whorl height. Each of the eight individual species dis-plays variation in spire height and narrowness of shell with respectto upper whorl face values, but the nature of this within-species vari-ation differs between the plestomorphic and apomorphic groups.

At another level, it is also evident that within-species variation inG. (Glabrocingulum) is not only greater than within-species variationwithin the aporncrphic group, but also greater than the pooled varia-tion within the entire apomorphic group. More significantly, pooledvariation within the derived group is actually in the same direction asthat within the G. (Glabrocingulum) group. Thus, within the apomor-phio group, the subgenera are to be distinguished by differences inwhorl height, total height, and narrowness of spire as determined byvariation in rate of translation of the generating curve along the axis[just as within the G. (G/abrocingulum) group]. It seems to be thecase that the among-species pattern of variation within the apornor-phic group is the same as the within-species patterns within theplesiomorphic qroup-c-an apparent corroboration of the continuityhypothesis. But the within-species pattern within the apomorphicgroup happens to be perpendicular to the among-species pattern.Thus the hypothesis is falsified, and in its stead we have corrobo-ration of the decoupling hypothesis, given the validity of the assump-tions discussed above.

We draw our third example from a paper written in defense ofthe continuity hypothesis. Gingerich (1976) has forcefully arguedthat within-species differences gradually accumulate to produceamong-species differences. Especially with regard to size increase(as reflected in the logarithm of the area of the first lower molar in

I

~---:-,-,--=--__ ----- mFigure 6.11 Ellipsoidsof scoresof specieswithin eachof thefour subgen-era of Glabrocingulum, (Seetext for detailed oiscusston.)

deleted, showing the basic shape of the variation within species(each of the species individually shows the same shape as thatshown for the entire subgeneric cluster). Factor I is an axis of varia-bility contrasting relatively narrow, tall shells with relatively short,broad ones, A high score on factor I implies relatively narrower,taller shells. Factor III is an axis of variation pertaining to the uppersurface of the whorl in relation to the remainder of the measuredmorphology. High scores on factor III imply specimens with rela-tively larger upper whorl surfaces.

With this in mind, and with reference to figure 6.11. the patternsof within- and among-group variation of the Glabrocingulum com-


various mammals), trends of increase in size within a lineage as awhole directly reflect within-species trends. These cases where thespecies are defined on the basis of arbitrarily chosen ranges of val-ues within a plotted continuum, of course, automatically create a situ-ation which confirms the continuity hypothesis, But it is relevant,nonetheless. to note (M. C, McKenna, personal communication,1978) that in several examples, especially in the graphs [Gingerich1976, figure 7 (our figure 6.12)) of the data for Pelycodus retetoni-trigonodus-jarrovii, there are a number of "micro trends" towardssize reduction of mean size within an overall trend of size increase.possibly providing another example of within-taxon patterns differingfrom among-taxon patterns.


Thus we question seriously the generalization that among-species patterns of net variation arise as large-scale versions ofwithin-species patterns which are, presumably, a function of naturalselection. Those who would argue for among-species differencesarising as an accumulation of within-species patterns over time citereversal in direction of the selection vector to cover cases such asthe Pelycodus example. Like all other forms of selectionist argu-ments at this level, such a proposition is utterly immune to criticism;it is, rather, a description in dynamic terms (based on a set of as-sumptions) of data, or refined statistical parameters, plotted onpaper. An accurate test can only be made by comparing patterns ofvariation within and among species that are defined and recognizedon independent criteria,z

~'"00

~"00

~ 000

~000

~_~-----o- _ :------~----_.~.J_ "N.rtMlicQ/if. ",-----. /

, ",,,Relationships among the Phenomenological Le.vels

of Evolution: Interconnections

"N."1lIIli«wI

Once the existence of levels and their mutual "decouplmq," is ac-cepted, we must examine the connection between them. There islittle evidence that when new species arise from old. their morpho-logical or genetic differences at the outset represent major, su~denshifts (but see Gould 1977a, for the resumption of a contrary view).Rather, available data suggest a spectrum of possibilities" b~t theoverall pattern seems clear enough-there seems to be continUity.' ornear continuity, in intrinsic attributes among closely related specle~ .Conspicuous morphological gaps within and among monophyleticgroups appear to be far more a function of differential extinction thansaltatory modifications produced by speciation "events."

This observation in no way contradicts the earlier argument thatdecoupling among phenomenological levels means that the n~-ture of among-species differences within a monophyletic group ISnot a function of the accumulation of within-species change overti d h st commonly repre-nne. When new species arise from 0) , t ey rna .

_, otype-the d If-sent a sam pi ing of the ancestral phenotyptngen . .f f h lation WIthin theerences were already there as a subset 0 t e vanancestral species. Some authors (e.g., Mayr 1963; Eldredge and

,-'P jrnwii-:

-' ----"--" - ',/,'----- ,, -~-. ~",

,p.,,~

=§/~-,.,,/~ :::::'..-":-' . ,/, --.-- ".,,~ _ ,I STRATIGRAPHIC RECORD OF

......--. 's,/ EARLY EOCENE PELYCODUS,

~ 0,' ""'M ' <"~I,••" ~.. . .•

!~--=~;===-:===7"iO--'-'--,;;;:----;-;;;--'---'-j-::~:-,..~'-.---.:roo ~~ f~ ~ ~

LOG {L.wl OF M,

Figure 6.12 Gin.gerich's diagram of the "evolution of the Early Eocene pri-mate Pelycad1.!s, We ~ave Indicated (arrows) the possible "rntcrotrencs" inreverse direction relative to the overall trend mentioned in the text. Log(L x W) of M, refers to t~e logarithm of the surface area (product of the lengtha~d Width of the crown In dorsal view) of the first lower molar. (FromGingerich 1976:16.)

=


thus can feed back and affect phenomena at a lower level. Evolu-tionary theory to date has insisted that evolution at all levels is es-sentially a population phenomenon The view of levels articulatedhere holds that processes at various levels affect adjacent levelsand the vectors can go both ways. The general hypothesis of charac-ter displacement provides an excellent example of how phenomenaat one level can be "worked out" or "resolved" at a lower level (asnoted by Salthe 1975 and personal communication, 1977). We a.rehere concerned with the structure of the hypothesis of character dis-placement and its relation to levels in the evolutionary process,rather than in the general validity of the hypothesis itself. , .

The general hypothesis of character displacement IS Inter-specific in nature. It may affect any two species which are compet-ing for resources, or are reproductively similar (see Grant 1972, for areview). Thus character displacement is theoretically expected to be

d ies are The hypoth-more frequent the more closely relate two spec I .es:s, in its original form (Brown and Wilson 1956), states that twocompeting species will be more different in those asp~cts of the be-havior and/or morphology in which the relevant adaptations are maru-

. . h th e in sympatry thanfest In those portions of their ranges were ey ar .where they are in allopatry. Grant (1972) has generalized. this hy-

. . . f fh two species Will be el-pothesis to read simply that In sympa ry, e _di I ment) or more drt-ther more similar (convergent character ISP ace

. . tl th hen in allopatry. Interent (divergent character dlsplacemen an w. . mpatry are hypothe-any case the similarities and differences In sy

.. . h '~enceofthecom-sized to reflect an adaptive response to t e coexr _. t' And these adaptivepetmq species allowing syrnpatry to con mue.

. . ff ted by natural selec-responses in morphology and behavior are e ec , _· I I ('nterspeclflc compe-lion. Thus we have a phenomenon at one eve I .' .·. . I b proprtate within-tition) resolved (according to the hypothesIs Y ap "

. f adaptation viaspecies (within-population actually) processes 0 ill

. f . t cific cornpetr Ionnatural selection One of the outcomes 0 In erspe h

d non by one, or bot ,(according to the hypothesis) is the accommo a I I.. . f th actual local popu a-

species to each other by the rnoditication 0 e .· . . f lation is another POSSI~tiona involved. The local extinction 0 a popu h .

. f h nomenon at t e In-ble outcome. Here too resolution 0 ape lon-e-fit

" - f selectfon-'- I nessterspecttlc level can be thought of In terms 0 t

t between any wodrops to zero, A third outcome of neosympa ry


Gould 1972) have pointed out that if speciation involves relativelysmall populations isolated on the periphery of the ancestral species'range (where, by definition, edaphic conditions least resemble thenorm for the species), selection might be particularly intense, andchange might accordingly be much more rapid at the onset of ge-netic isolation. But such need not necessarily be the case.

Thus speciation-the origin of new reproductive units-does notautomatically imply great discontinuity in genetic and phenotypicproperties of species. Again we refer the reader to the relationshipbetween selection within populations, on the one hand, and specia-tion on the other (see chapter 4, p. 121f. and p. 270, this chapter).These ideas form the major theoretical argument that microevolutionis decoupled from macroevolution as discrete phenomenologicallevels. In the allopatric case, natural selection mayor may not effectchange within isolated populations such that, upon neosympatry,reproductive isolation (via behavioral. anatomical, physiological, orother changes) mayor may not eventuate. If reproductive isolationhas occurred, it can only be viewed as fortuitous-in no way canselection be said to have acted to create two reproductively isolatedtaxa from a single ancestral taxon. No benefits could accrue to indi-viduals (the essence of natural selection) from such a phenomenon,Should partial reproductive isolation occur, whereby hybrids pro-duced after neosympatry have reduced fitness, it is consistent withthe theory of natural selection to predict that selection will improve re-productive isolation as a necessary consequence of the failure of in-tergroup offspring to survive. In models of parapatric and sympatricspeciation, selection enters earlier to establish. reproductive isola-tion (Bush 1975), Problems with these models are the same as thoseof envisaging intraspecific, sympatric interdemal selection and willnot be discussed further here. The general pattern is that naturalselection, as a population dynamic, accounts for differences amongancestral and descendant species, but is blind with respect to theactual creation of new species,

Salthe (1975 in press; personal communication, 1977) has com-mented on the nature of the connections between the phenome-nological levels of evolution, The general nature of such connectionsis that both lower and higher levels place constraints, or initial andboundary conditions, on any given level. Phenomena at one level


species is a lack of interactive effects at all-which, in a way, canalso legitimately (if needlessly) be thought of in terms of selection-as a selection coefficient of zero,

Phenomena at the interspecific level, such as competition, maybe resolved in part by processes at a lower level. By the same token,the range of possible reactions at the higher level are clearlybounded, or constrained, by the lower level. Within populations,change via selection is clearly limited by (a) the range of variation,with its myriad underlying controls available, and (b) the limits towhich the epigenetic process can be modified at the individuallevel The limitation of possibilities for adaptation via selection re-strict the range and determine the nature of possible outcomes interms of new morphologies and behaviors emergent in speciation.The number, nature, and distributions of species existing at anymoment, in like manner, determine the nature, and delimit the possi-bilities of the composition of clades of higher rank at a later time,

Thus the levels are complexly, but directly, interrelated. Whatgoes on at one level simultaneously affects, and is affected by, pro-cesses at adjacent levels. But this is by no means to say that thephenomena and processes at these levels are the same throughout,as some aspects of contemporary theory, particularly "synthetic"macroevolutionary theory, are prone to have it. Rather, the levels areat one time decoupled and interacting.


of such change (i.e .. evolutionary vectors) within monophyletic taxa, ' d b e 'he preponderance ofof higher rank, As we have reviews a ov ,

, den! phenome-discussion of such macroevolutionary tOPiCS erues any" ' otutionary processesnological distinction between within-species ev

(the mechanics of which are fairly well understood) and am.on.g-d pted the view that within-species processes. In contrast we have a 0 .

, , I election) Indeed effectspecies processes (specifically. natura s .th t oration of consider-change in gene frequency but that smoo ex rap

, h 'thin species is not an ap-ations of rates and directions of c ange WI b" ch as it ignores the pro -propriate model for macroevolution, maemu . W' h'

, W h ve agreed with rigtern of the origin of discrete specl.e~. e a f differentialand other authors that species origins and patt~rns 0 t ratesspecies survival are the relevant factors controillng apparen .

, I change among speciesand directions of genetic and morphologlca. k It now re-

, , 'f hi h r catagoflcal ran.within a monophyletic taxon 0 Ig e I tton in these, I th 0'1 of macroevc u Imains for us to formulate a genera e . . tlt component

terms and, subsequently, to consider the testability 0 I'SrecogniZing, h th ry we are mere yhypotheses. In formulating suc a eo, in the literature, More-

biological principles already well developed , d ed be con-over, our goal is to show that such a theory .can ~: f:rmulating astructed in testable form; we make no pretenslon~ acroevolution.complete or even particularly satisfactory theory 0 ~ macroevolution

On a relatively superficial level, the theoryh0 ral change oc-h I tc and be aVIOstates that: (1) All genetic, morp oioqr , sees at the in-

. .' . t .ned by procecumnq in evolution anses and ISmain ar I" n genetic drift,

t natural se ec 10 ,traspecific level and below (muta ron. ditterencef (in-d as Inter-taxon I

etc.) but. when seen expresse h suit of the origin.. , h h ge is also t e re

terspectttc and higher), sue c an irecti s of such change, , ' f " (2) Net direc Ionand differential survival 0 species" period of timeI stained over a

among species may be neutra or su R' of such change. ' . nd (3) a esand Involve a number of species: a , I of (a) speciation

, .' I on the interp ayamong species depend pnrnan y h! I tter connection, all

, lncti rates In t IS a ,rates and (b) species extinc ron tate>. f relative diversity, (' e patterns 0 ,spindle-diagram phenomena .1.., 0 h letic group), including

through the entire history of a given ~on p Y f taxa of low species"adaptive radiations," long-term persl~tence 0e "bradytely"I. are todiversity (and with little net morphologic c~angf 'the interplay of spe-

" th erspectlve 0be understood primarily from e Pelation and extinction rates.

Macroevolutionary Theory: A Restatement of the Problem

Many of the basic questions of evolution remain, whether addressedfrom a "taxic" or "transformational" point of view, whether phenome-nological levels are explicitly considered, and whether species andtheir origins are included or not. A fundamental task of any evolu-tionary theory is an elaboration of a coherent theory of how change-genetic, morphologic, and behavioral-is effected. Within-popula-tion mechanisms, inc Iuding natural selection acting on agroundmass of variation, are of great importance in this regard, aswe have repeatedly emphasized. However, in the context of macro-evolution, there still remains the large subject of rates and directions

_.J,


In elaborating on this skeletal outline of a theory, it is immedi-ately apparent that, just as in rnlcrcevolutionary theory, there are twocomponents involved. As Raup (1977) has most recently discussedin this connection, Western science continues to seek deterministicexplanations for patterns in the natural world. Yet it is quite clear thatmany of these patterns may arise from pure chance alone. Moreover,what appears to be pure chance at one level may have a determinis-tic base at another level. Individual point mutations, it is still as-sumed, each have a specific (deterministic) cause; yet viewed at thepopulation level, what point on a chromosome and what individualwithin the population will show that mutation is, for all purposes, tobe considered random. In addition, mutation is random with respectto the regime of natural selection obtaining within a population at themoment the mutation occurs. Yet the mutation rate within a populationas a whole is statistically, at least, a constant, and thus deterministicat the level above the individual. Similarly, speciation is determinis-tic (in the sense that a set of proximal causes presumably underl iesany actual speciation event), yet can be considered random with re-spect to long-term trends (as Wright pointed out) and, in terms ofmacroevolution, trends can still be considered deterministic viadirected differential species survival ("Wright's rule").

But the probability is at least potentially equal that any suchspecified patterns are themselves the result of chance. Returning tomicroevolution for an example, change in gene content andfrequency within a population can result from a variety of causes(hence all be "deterministic" in one sense), and yet be either deter-ministic (natural selection) or random (genetic drift, migration, etc.)in a larger sense. The effect produced, in other words, is bestviewed as an accidental product of the process, however determinis-tic that process might be. Examples include "sampling error" in gen-erational reproductive patterns within populations (genetic drift); ori-gin of reproductive isolation between two groups of populations inallopatry as a sheer accidental byproduct (change in morphologyand behavior) effected either by chance (e.g., drift) or selection, butin any case without regard to possibilities of resumption of a patternof reproductive continuity sometime in the future, at the macroevolu-tionary level, interspecific trends may reflect some sort of "speciesselection," but could also arise, one might assume, from chance.The argument in the latter case mirrors that for genetic drift, a con-


I k f pure and exact correlationcept arising from the troublesome ac 0.. ithi populationbetween the quality of an individual flourlshmg ':'1 In a

h·

1'inc-

, . " s" Similarly, w I e exand reproductive behavior and success. ) II caused bytions are clearly (and speciations somewhat les.s ~o a the

. f biotic and abiotic processes,some particular concatenation 0 I . .' I cases of speciesaccidental ("bad luck") component of md~~~dtU~heelaboration of aextinctions is sufficiently apparent to pro.1 I T formulate such apurely deterministic theory of macroevolut.lon. 0

theory as purely deterministic ,:,ould be n~lve'f valuating such ran-Raup (1977) has summarized ~etho s ~t ens Recognition that

dom components of intricate evolutionary pa.ner a' iven pattern hasthere may be "random" components unde:YI t

gte~ (in a statistical

led to some fruitful hypotheses tha.t ca~asere~:ntIY suggested thatsense) in recent years RauP"(1977.63),, Assi ning probability of ex-species may be regarded as partl~les. s I~tting) of new particlestinction of the particle. or production (by P, lth 'he elements of

, I e agreemen wfrom old is conceptually m c os G Id and their colleaguesmacroevolution discussed here Raup, ~ Id 1974' Gould et at(especially Raup et ai. 1973; Raup and o~'n dev'elopment and

d Patterns of the onqr ,1977) have simulate many t of rank higher thanextinction, at the species level, of clades ~~~: shape (summarizedspecies). Gould et al (197~) have com~a;~ulated clades with thoseas spindle diagrams; see figure 6.13) 0 h n that simulated pat-

ti d ta and have s ow Ibased on actual systerna IC a . respects "random y. tions but In many .terns, involving certain assump . "clade shape" seen In

generated," essentially duplicate th.e range m valuation of the "ran-. k ! leading to an e " tsystematic data. This wor IS "patterns and, JUs as

. acroevolutlonary .,'dom" component underlying m It of systematiC blOlogl-significantly, involves the use of actual resu s nts (species) within

, ' . f the com ponecal tnvesttqatrcn-cd'strtbutton 0t cific hypotheses. fmonophyletic taxa-to tes spe t here as the death 0 any

Van Valen (1973) has sugge~ted tnat. w ( ne of many possible),h direct cause 0 f dindividual is assumed to ave a " lation is often so Ixe

frequency of death of individuals Wlt~l~ a PfoP~inction of the popula-. . f robability 0 ex tas to allow the asttrnation 0 P . within a higher axon

h lzed that speCies Ition as a whole. He hypot eSIZ "a statistical genera-lsti te of extlnctlon- d th Smay exhibit a cbaracterts IC ra . ,'ve isotope-an u

h It life of a radloac I Th·ization comparable to the a - I d an be assessed, IS. t 'h entire cia e cthe probability of extinction 0 e

__ A

300Systematics and the Evolutionary Process 301

relative to the random component. The remainder of this chapter isdevote? to .an elaboration of a deterministic component It should be?orne In mind throughout that the patterns it addresses may also beIn any particular case, randomly generated. '

Systematics and the Evolutionary Process

liS,- 88 vi...... 11 U

t-tr The Deterministic Componentof a Macroevolutionary Theory

-C'

} ,0

, ~,

I,,1j

z I

As we have outlined, the basic problem of macroevolution is the ex-planation of genetic, morphological, and behavioral differencesamong species and taxa of higher rank in terms of the origin and dif-ferential survival of species. Rates and directions of such change inparticular are a function of the origin, survival, and extinction ofspecies and the interrelationship of these components. Thus any de-te '. ,rmm.suc theory of macroevolutionary processes must be con-c~rned with factors controlling (a) speciation rates, (b) species sur-vival (and its converse, species extinction) It is also clear thattheoretical, field, and experimental work pertaining to these subjects

ISecological in scope and contentEcology, at least that portion of it germane to our present dis-

cussion, is primarily concerned with the integration of populations ofsom,efinite number of species into a community. The community oc-cU~les some specific geographic area (limits may be imprecise andshift through time) and persists for some period of time. Precisespecies compositions of communities are conventionally theorizedto vary both as a function of some developmental process (i.e., asuccession of communities, or sere, leading ultimately to a "climax"community) and from oscillations in species composition in commu-nities in equilibrium, reflecting discrete episodes of local extinctiono~populations, as well as migration of individuals of different spe-c~es.Thus factors hypothesized to control speciation rates and spe-cies persistence and extinction are analyzed and understood at thelevel of the local ecosystem. At least epistemolOgically, then, anymacroevolutionary theory must consider the ecological controls ofspeciation and extinction. Whether such a resolution will prove onto-logically equally compelling remains to be seen. A further important

t'<, <, ~,,

------

Figure 6.13 Comparisonof Ih 'and realclades.Top,spindle ~i~p:~~le diagramsof randomlygenerated

dProgramat branchingand extinctTocpSgbecberatedby one runof the MBLraqramsfor 9 . . ro a 1III18Sof 0 1 b

vertebrate pale~~~~~o~~~I(~~e~o~fldbCatChioPOdS,fromtheO~r~~;i;~i~~I,~_, e al., 1977·24)

approach IS again consistent with ou ' . .offers a means of understanding th r view of macroevolution, andvotutionary processes e random component of macroe-

There are, however, sufficient aswithin and among taxa of higher c pects of patterns of variationrate and direction of genetic a dategoncal rank (e.g., differences inproduction of new species) t n morphological change and in thelutio I 0 warrant a ge I, n a ong deterministic lines. S nera .th~oryof macroevo-extst as one of two camp uch a deterministic theory would

onents of unsp if deCI te (as yet) importance


implication of me relevance of ecological controls on the origin, per-sistence, and extinction of species is that the concept of adaptationcan be brought back into our consideration of macroevolutionarytheory.

There is a set of relationships which describes in general termsthe role of ecological theory in tying together such disparate strandsof evolutionary phenomena as adaptation, on the one hand, andspecies diversity (encompassig origin, persistence, and extinction)on the other. The relationship, stylized as an equation in figure 6.14,hinges on the fact that both species diversity and the morphology ofindividuals are related to the occupation and exploitation of ecologi-cal niches. Therefore niche theory should act as an interactive cata-lyst in linking these two topics in evolutionary theory.

In the first statement of figure 6.14, morphology is implicatedwith niche theory. All this says is that organisms are adapted to theecological niches they occupy (a tautology) and that such adapta-tions are expressed in the intrinsic properties of individuals compris-ing the local population occupying a given niche, (If this latter pointis not exactly a tautology, it is nearly so.) In other words, traits of or-ganisms are either directly related to the occupation and exploitationof the population's niche, or, at the very least, neutral (irrelevant) withrespect to the occupation of the niche, Traits which appear unrelatedor superfluous to the exploitation of a particular niche abound in allspecies (e.g., the presence of five arms, not four or six, in ophiuroids

MORPHOLOGY adaptation via ECOLOGICAL.. selection NICHE


) Wh'l some of these traits may, inscuttling over the sea bottom, let arising by pure chancefact, represent adaptively neutral charachers the ophiuroids) rspre-, .. t such examples (sue as .10 speolation. mos. . I of distribution (all ophiuroldssent characters With a higher leve. '1' echinoderms known arepossess five arms; all but the most ~nml I~~he evolutionary system,pentaradial). In general, as an ~Iom °h' al (taxoromcl level-

hi t whatever tuerarc IC .synapomorp les-a . volutionary novelties, according torepresent evolutionary novelties, E I tion generally arise as

t d th ry of mfcrcevo u ,the highly corrobora e eo 'I ble variation which max-an outcome of natural selectio~ on a;:~ apart;cular exigencies of aimizea adaptation of a populat~on, tOniche. Thus there is a firm con-local environment-the population s 1 and ecological niches, aceptual connection between morpho ogy .

d ptation via selection-connection based on a a I t diversity to niche theory.

, f tiqure 6 14 re a es I'The second line 0 I, ne of the more ac rve, di tty has become 0 20Regulation of species iversr h during the past

areas of eccloq ical and evolutionary res~ttarcby paleontologists inth t the two books wn en . 'syears-so much so a f publication of Simpson

the United States since 1953 (t~e y~ar °d the word "evolution" inE I ti n) which tnclu e . dMajor Features of vo u /0 . 'gin maintenance, an

ned With the on , '1the title are primarily cancer .' stems not necessafl ydi 'ty (Within ecosy, . ,

degradation of taxic rversr . d here). Both Valentine 5, ps as discusse d Bouwithin monophyletic grou f the Marine Biosphere an -

(1973) Evolutionary Paleoecology a centrals come close to, d Extinction Rate . S h acot's (1975) evotouoo an f di ersity regulatIOn. uc

. . tl as a matter 0 IV . rt thetreating evolution stnc Y . 'work reflects, at least In pa ,shift in emphasis since Slmpso~ s I gy since World War II de-

ff rt l theoretical eco 0 .large amount of e 0 In . of species diversity.voted to the ecological regulation I between rates of appear-

. fl cts an interp ay . andSpecies diverSity re e. d disappearance of speCies, ,

ance (by migration or evolutIOn) an I d to consider these issues .Inecologists have been ineluctably e ufation of diversity have 10-

. tt of the reg . Tbterms of niches. Consldera Ions ra hic area and diverSity (equr I -volved correlations between geog :rthur and Wilson 1967), correla-rium biogeography, started b~ Ma~nd diversity, correlation betweention between latitudinal gr~dlent~ environments, and the effects ofwater depth and diversity In ~aflne diversity. Underlying each of

I leal time on in as a tau-ecological and ge,o og e t of niche number-agal ,these formulations IS the conc p

DIVERSITY regulation1 _ ECOLOGICALNICHES

-;:~m~C~he=:::~MORPHOLOGy.... -theory

Figure 6.14 Three stylized "equations." The upperrelates morphology to ecological niches effectedby adaptation via natural selection; the middle state-ment recognizes the control over species diversityexerted by niches (number and breadth). The twoUpper statements are therefore related, with nichetheory serving as the common denominator.

DIVERSITY

,i


tology, the number of species present is equal to the number ofniches realized (occupied and exploited in a given area). This is trueeven jf the analysis is statistical and merely seeks to correlate spe-cies diversity with, say, geographic area. Thus any consideration ofthe origin and maintenance of diversity must consider problems ofecological niches themselves, including relative niche breadth,changes in niche breadth, the effects of interspecific competition,resource partitioning, and the like.

Utilization of resources and problems of relative niche breadthsimultaneously lie at the heart of the controls of species diversity,on the one hand, and on the other hand, are the bases of the adapta-tiona! (functional and behavioral) characteristics of any given spe-cies. Thus we can connect the first two lines of figure 6,14 into asingle "equation" as given in the third line of the figure. Niche theoryprovides the nexus between considerations of adaptation and con-siderations of the origin, maintenance, and degradation of speciesdiversity.

It is not our purpose, in a general book on phylogeny recon-struction and aspects of macroevolutionary theory, to discuss in de-tail contemporary ecological theory pertaining to niches, either fromthe standpoint of diversity or from the standpoint of functional ana-tomical analysis of adaptations (niche utilization). Such theory in anycase ~annot be tested with reference to branching diagrams in sys-tematics, ~he central concern of this book. The purpose of this sec-tion has Simply been to suggest the nature of the dynamic processinvolved in macroevolution, as well as the nature of the connectionbetwee~ two areas (i e., adaptation and species diversity) concep-tually divorced earlier in this chapter.

HYPo:heses of macroevolutionary processes involve speciationand specres persistence and extinction, and their interrelationships.Subordinate hypotheses pertaining to these processes belong to therealm of ~cOI?gy and ecological theory. Such theory is focused oninterspecific Interactions (especially competition) and resultant ef-fects In terms of niche width. Relative niche width is, in turn, relatedto the concepts of relative eurytopy-stenotopy. Originally defined interms of habitat breadth, the terms eurytopy and stenotopy havebecome more conventionally used to refer to relative breadth of tol-erance to specifiable parameters of the physical and biotic environ-ment. or the relative breadth of ability to exploit specifiable pa-


rameters of the resource space. The terms as used in ecology haverough equivalents in the evolutionary expressions "broadly adapted"and "narrowly adapted" (i.e.. when those expressions are applied tosingle species) and these latter terms are themselves near-synonyms of "generalists" and "specialists," These various pairs ofterms are not strictly synonymous, but are sufficiently close to permita general discussion of the nature of the relationship between thesevariables and others encountered in macroevolution.

Theory and Information in Macroevolutionary Analysis

The systematist attempting to confront the issues of macroevolutionhas the following sorts of data (in the form of well-corroborated hy-potheses) to work with: (a) a theory of relationships among specieswithin a monophyletic taxon of higher categorical rank; (b) the dis-tributions of the component species in space and, where possible, intime (i.e., fossil evidence); and (c) an evaluation of the relative de-gree of apomorphy of each specifiable morphological feature ineach of the species considered. This last system of hypothesesarises from the very analysis upon which the theory of interspecificrelationships is based. When the relative distributions of homologousfeatures are viewed from the standpoint of evolutionary novelties, aconceptual link between adaptive specialization and relative apo-morphy, vs. adaptive generalization and relative plesiomorphyemerges. The relationship holds only for comparisons of homol-ogous traits among species where the one (or more) relatively apo-morphous cond ition may be hypothesized to constitute a functionalspecialization. With respect to that particular anatomical specializa-tion-assumed or hypothesized to relate to the tolerance for, or utili-zation of, some specifiable component of a species' niche-specieslacking the apomorphous condition (i.e., retaining the more plesto-morphous state) are hypothesized to be more generalized with re-spect to the homologous niche parameter.

In plainer language, we simply reiterate the oft-noted correlationbetween anatomical specialization on the one hand, and behavioraland functional, hence adaptive, specialization on the other, and this


with a narrowing of range (with respect to a certain parameter) of aniche. Thus, with respect to comparison of niche width among spe-cies related in some fashion on a cladogram, it may be hypothesizedand tested for individual characters, that relatively more apomor-phous species occupy narrower niches, at least in terms of nicheparameters associated with those particular anatomical and behav-ioral attributes. We suggest that this general sort of correlation be-tween evolutionary (anatomical and behavioral) specialization andniche width among a series of closely related species, is a fairlywell-corroborated generalization already; but we point out that fieldecological analysis coupled with functional anatomical analysis cantest the hypothesis in any given instance. All that is needed at theoutset is a well-corroborated hypothesis of relationships among thespecies involved. The correlation between anatomical and behav-ioral specialization on the one hand, and narrowing of width of a par-ticular niche on the other, and the converse (ecological generalistswith relatively plesiomorphous morphologies) arises as a general-ization from such work as Bock's (1970,1972) on the Hawaiian Dre-panididae, and Lack's (1947) analysis of Darwin's finches on theGalapagos Islands, But we repeat that such a blanket generalizationneed not be accepted for our argument; rather we simply point to theconceptual Iink between morphology and niche width and furthersuggest that in any specific case (involving Recent species, atleast), hypotheses of this form are directly testable.

The original definitions of "eurytopy" and "stenotopy'' referred tohabitat breadth and were used in ecological geography almost assynonyms for "widespread" and "narrowly distributed," These latterterms, of course, further suggest the contrasting set of terms "cosmo-politan" and "endemic," terms generally used as summary descrip-tors of relative geographic spread, devoid of any implication of eco-logical specialization or generality. We do not mean to suggest thateurytopic species cannot be endemic to small areas or that steno-topes cannot be widespread. Nonetheless, as a rule and within amonophyletic group, cosmopolitan species tend to be eurytopes (inthe sense of being relative ecological generalists) whereas relativelymore restricted species tend to be stenotopes. Again, the readerneed not accept this generalization as ironclad; for any given case,the hypothesis can be tested directly, and all that is needed at theoutset, is, once again, a well-corroborated theory of relationshipsamong the species involved,


While on the subject of the correlation between species d istribu-tions and relative niche breadth, we note the generalization-goingback in the literature at least as far as Williams (1910)-that narrowlyadapted species (as judged by lack of within-species variability orpresence of autapomorphies) not only tend to occur in relatively re-stricted geographic regions, but also tend to persist for shorterperiods of time in the stratigraphic record. This suggests, ~f c?urse,that specialized species are relatively more prone to extinction. along-standing suspicion in evolutionary theory. The problem with t~isparticular generalization is that it is extremely difficult to test; relativeduration in the stratigraphic column may be highly corroborated (Inspite of the sampling problems inherent in studying the fossilrecord), but evaluation of relative eurytopy/stenotopy must be basedon ancillary, secondary correlations, which themselves are untesta-ble and possibly false for any particular case. In fact. the general-ization may not hold: relative lack of variability may indeed correlatewell with stenotopy within a group of related species, but then againit may not. Evaluation of relative stenotopy and eurytopy among fos-sils hinges on the assumption that such a relationship pertains, or ,IS

based on geographic distributional data (i.e" range of occurrence In

reconstructed paleoenvironments). In either case, relative eury-topy/stenotopy among fossil species is not susceptible to testingand thus the general ization about the correlation between such eco-logical strategies and temporal duration is itself not susceptible totesting in any particular case.

However, distributions of species in time and space can becompared directly with a matrix of relatively primitive/derived mor-phological features. Thus the hypothesis that relatively plestornor-phous species within a monophyletic group tend to be more Wide-spread geographically or to persist longer (stratigraphically) can beeasily tested. All that is required is a prior, highly corroboratedtheory of relationships among the species, and the relevant distribu-tional data. The only further requirement is that the theory of rela-tionships cannot itself be based, in whole or in part, on the distribu-tional data.

There is thus a conceptual, and for the most part, for specificcases, testable set of three interrelationships between three classesof variables: (a) relative niche breadth, (b) relative degree of apomor-phy, and (c) distribution of species in space and time, all in the con-text of well-corroborated hypotheses of relationship among the spe-

,I


cies involved. If these three sets of variables are arguably, buttestably, interrelated, it still remains to be shown how they might beimplicated in the control of speciation, species persistence, andspecies extinction.

At this juncture, we reiterate our acknowledgment of "random"factors; a new species may appear as the accidental by-product ofchange in the physical geography of the area, or a portion thereof. ofits ancestor. A species might last longer than its sister in the nextvalley for purely accidental reasons, the converse of the observationthat many extinctions are unlucky accidents. Entire ecosystems canbe degraded relatively quickly (less than a million years, to judgefrom such "events" in the Upper Devonian, Upper Permian, UpperTriassic, Upper Cretaceous, and Pleistocene, among others). In suchcases, involving thousands of species in many unrelated clades, anenvironmental event (literally a catastrophe) occurs that is utterly ac-cidental with respect to individual species adaptations and particu-larly random with respect to the interspecific interactions and resul-tant patterns of differential species survival resulting fromdeterministic macroevolutionary processes within monophyleticgroups.

With this in mind, it is now relevant to consider the relationshipthe three sets of variables discussed above might theoretically haveto patterns of differential species survival-macroevolution. If spe-cies diversity is, at base, a function of niche width, a general set ofpredictions about distributions in space and time of relatively ape-morphous/plesiomorphous species emerges. In general, within a sin-gle monophyletic group, highly speciose groups (at anyone giventime, throughout the. entire collective geographic range of the spe-cies-group) should be (in terms of each component species) rela-tively narrow-niched, narrowly dispersed geographically, and rela-tively apomorphcus (including rather autapomorphous) (see Bretskyand Lorenz 1969, for a similar lisl of correlated variables). Individualspecies within Jess speciose sister-groups should prove to be (a)relatively plesiomorphous (b) more widespread geographically, and(c) broader-niched. We cite Fryer and lies (1969) and Greenwood's(1974) analyses of species flocks of cichlid fishes in east Africa, ifnot as a conclusive test of the generality, at least as an example thatfails to falsify it. The cichJid genus Tilapia is relatively nonspeciose(depauperate), and its included species tend to be relatively broad-


niched and geographically widespread (within a given lake or drain-age system) than species of the related, highly speciose genusHaplochromis.

There is one important objection to the correlation between nichewidth and diversity as a means of approaching macroevolutionarytheory: ecological theories of diversity, including niche width and in-terspecific interactions (especially competition), pertain to all spe-cies within a geographic area, not to all component species of amonophyletic group, regardless of their occurrence (i e., in allopatryor sympatry). Thus we conclude that if such theory be relevant tomacroevolutionary phenomena involving monophyletic groups, thenthe discussion is limited to interspecific interactions among syrnpa-tric and parapatric species within the group-a condition alreadyclearly recognized by Bock (1970, 1972) in his analysis of the evolu-tion of Drepanididae.

What, then, are the ultimate conceptual links between nichebreadth on the one hand, and control of species diversity withinmonophyletic groups (speciations, species persistence, and extinc-tions) on the other? Although detailed ecological theory is both inap-propriate for this book and, in any case, outside the bounds of ourexpertise, the basic nature of this relationship does seem fairly clear.and is susceptible, of course, to further examination. We have al-ready noted the healthy disagreement in contemporary ecology overthe issues of the origin, maintenance, and degradation of diversitypatterns. One of us (Eldredge 1979b) has hypothesized that specia-tion rates themselves are, in an important sense, a function of the"niche strategy" of individual species within a monophyletic group.The hypothesis states that, within a given group eurytopic speciesreact differently to interspecific competition than do stenotopes:eurytopes tend to react to interspecific competition by mutual exclu-sion, whereas stenotopes more commonly react to such competitionby further subdivision of resource space (niches are further nar-rowed). Hence, within a monophyletic group, if there is a discerniblespectrum of stenotopy and eurytopy, there are frequently many sten-otopic species and rather fewer eurytopic species. The hypothesis issuggested by actual patterns of distribution in nature: within mono-phyletic groups relatively eurytopic species tend to be allopatricwith respect to one another (or "vicari ant") and far-flung, whereascongeneric stenoptopes are more likely to occur sympatrically.


There are problems with this hypothesis. Why, for instance, do eury-topic species comprising a relatively eurytopic species lineage, re-main eurytopic? In any case, the hypothesis requires a great dealmore testing (and mathematical investigation) than we can give ithere. We also point out that the hypothesis in essence constitutes anargument about dampening controls on speciation rates, though theconverse (that stenotopy actually causes high rates of speciation) isnot held to be true. We include it here as an example of the kind ofconceptual link between rates of speciation and niche widthrequired in macroevolutionary theory.

We have already briefly mentioned the supposed connectionbetween niche width and extinction (species persistence being theopposite of extinction, the two are treated as the same problemhere). According to conventional wisdom, more broadly adapted or-ganisms (eurytopes) are expected a priori to be able to survive un-predictable and large-magnitude environmental disruptions becausethey exhibit (by definition) broader physiological tolerance ranges.The hypothesis is intuitively appealing and seems well corroboratedin mathematical and experimental treatments given it by ecologists(see Bretsky and Lorenz 1970). Insofar as there is a deterministic el-ement to relative rates of extinction within monophyletic groups, pat-terns of relative niche width are at least implicated if not exclusivelythe controlling factors involved.


(evolutionary analysis of patently non-monophyletic, including "gra-dal," groups, under these terms is obviously meaningless).

2. Tabulate the pattern of species diversity.For Recent organisms, the pattern is simply a tabulation of all

known species. For groups known in whole or in part from the fossilrecord, a time-averaged appraisal of the "standing crop" of knownspecies within the shortest recognizable interval of time is used.

3. Depending upon the characteristics of the pattern seen in (2),specific predictions are made about the component species sam-pled.

The predictions deal with the geographic and stratigraphic distribu-tion of each species in conjunction with the distribution of apomor-phous and plesiomorphous characters and, especially with Recentspecies, relative degree of eurytopy and stenotopy. If the predictionsdo not agree with the observed (l.e., hypothesized but corroborated)attributes of the component species, the general set of predictionsabout diversity controls is falsified. Should successive examplesalso be falsified, the basic assumption-that diversity and adapta-tion are a function of modes of ecological niche occupation andexploitation-s-can be seriously questioned. Testing such complexpropositions is always difficult. Potential error in tabulating distribu-tion patterns and in characterizing stenotopy vs. eurytopy is so greatthat we suggest rejection of a macroevolutionary hypothesis cast inthese terms only if 6 percent or more of the putative cases fail to sub-stantiate the predictions. We are aware that, even if our proposedrejection of the conventional extrapolation of microevolutionary pro-cesses to the level of macroevolutionary phenomena is accepted,our proposed reorganization of macroevolutionary theory may not besuccessful. But at least we have identified a means whereby our al-ternative theory can be directly evaluated, criticized, and perhapsrejected. And perhaps it may eventually emerge well-corroborated.

The specific kinds of diversity patterns addressed below dealwith both fossils and Recent organisms. Both are important, for dif-ferent reasons. Patterns of diversity in the fossil record (which can begraphed in a variety of ways, including the familiar spindle diagram)display a sampling of actual patterns in true evolutionary time. Onlywith a fossil record can changes in diversity patterns within a groupbe tabulated. And part of the complex of predictions important in as-sessing controls of the rates of both speciation and extinction

Macroevolution: Hypotheses and Some Basic Patterns

We have adopted the general assumption that macroevolution isbasically a phenomenon of differential species survival. If the line ofargument linking adaptation and species diversity through utilizationof species' niche widths is basically correct, macroevolutionary pat.terns (to the extent that they are deterministic at all) can be inves-tigated by formulating predictions about the nature and distributionsof individual species sampled in the monophyletic group understudy. The general procedure involves three steps:

1. Perform phylogenetic analysis, arriving at a well-cor-roborated cladogram of relationships among all species sampled


require accurate evaluation of at least relative (absolute would evenbe better) duration of species in time. Only the fossil record cansupply such information. On the other hand, evaluating relative eury-topy and stenotopy-a difficult task at best-ean only be done withany degree of credibility with Recent organisms. And the Recentbiota is, after all, a direct product of differential species survival fromthe Pleistocene epoch. Thus analyses, such as Bock's (1970, 1972)on the "adaptive radiation" of Drepanididae, are by definition validand contain much functional anatomical appreciation of niche utili-zation not basically possible to secure with fossils.

Table 6.1 is a classification of macroevolutionary patterns ac-cording to (a) criteria of recognition (diversity pattern, ratio of extinc-tion to speciation, and absolute rate of speciation), (b) hypotheticalnature of pattern of species selection, and (c) predictions about thenature of component species for each pattern. Other patterns couldno doubt be listed; the classification is intended to be illustrativerather than exhauatlvs. More important, other theories of speciationand extinction controls will produce other predictions. We aim in thissection to show how macroevolutionary theory can be hypotheti-co-deductive under the general notion of differential species survivaland not necessarily to establish a specific version of such theory.We shall now discuss each macroevolutionary pattern in the orderpresented in table 6.1.

Adaptive Radiations

The expression "adaptive radiation" is itself a value-laden descrip-tion, hinting at the causes underlying the particular pattern. A moreneutral definition of the pattern is simply a relatively rapid prolifer-ation of species of a monophyletic group-"relatively rapid" con-trasts the rate of proliferation of species within the group both beforeand after the radiation begins. Although such radiations may occurtoward the beginning of the history of a clade, such need not be thecase.

The adjective "adaptive" appears to be appended to the gen-eral term for this pattern because the proliferation is felt to result di-rectly from the SUddenavailability of a new opportunity. The new op-portunity might simply be new habitat space (e.g., an islandunoccupied by ecologically similar organisms), or new "adaptive


s~ace" (e.g., i.nvasj~nof the terrestrial environment by aquatic orga-nisms, or the Invention of a new behavioral and/or anatomical com-plex ~hjCh so improves utilization of the ecological space that theway IS opened up for the existence of a host of variations on atheme-one way of dividing up the new or "redefined" resourcespace). Simpson calls the latter phenomenon "key innovations." 11

. .Whatever the mechanism involved, as pure pattern, adaptive ra-diations consist of many closely related species occupying a dis-c~ete geog.ra~hic region. Sympatry, or near sympatry, tends to behigh. Speciation rate. of course, is high and exceeds extinction rate.A~ a n.on-testable descriptor of the pattern, species selection isdisruptive-many vanat! . .Ions on a morphological theme bemg pro-duced (perhaps as in 8 k' .. .oc s VIsion of divergent character displace-ment resulting from inte if ... rspecr IC competition among closely relatedspecies of Drepanididae).

Fromthis pattern a t f .., se 0 specific hypotheses about the compo-nent species can be add d: (tl Helati ..sh Id uce . 1) Relatively stenotopic species

b Of0. Igre.atlyoutnumber eurytopes; thus (2) Extinction rates shoulde airy high (i e .

(3) E ":" , component species should be fairly short-lived);ach species should b t .th . e res ncted to a relatively small portion of

e geographic range acc . d .(4) Th upre collectively by the entire group' and

ere should be hi h 'n b f

a Ig percentage of species marked with aum er 0 autapomorph .f' res, with synapomorphies among subgroups

o species relatively less commonWe now take a patter tom .di t' n Conorrrunq to the descriptive criteria of ara ra ron, characterize th

The pattern stem e pa~e.rn,and apply the set of predictions.is at this d t s from the anginal research of one of us (N.E.) andMoreove a e only partially publ ished, in the systematic literature.

r, we state at the out t th . . .We present thi . .se at the predictions are confirmed.evolutiona his analysIs :or ItS.heuristic value, to show that macro-can be tesied YbO~heSes.InVOlvlng species origins and extinctionsor all of th . t erstmilarly patterned data may well refute some

e predictions Th "II A . e pcrnt IS that such falsification canrelated COlle""t tha of .. d .

~,~ a~~~"{ .ghrank as the concept of niche ' '. s see p. 261). seems to be to taxa of hi erentities. they can hardl ' sn~~to species. Since such "higher" taxa do not exist as ecological.... ' y mvace or occupy anvthi If h·· I thapll\'e zone" it is ,J' mg. t ere IS anything to the concept 0 me

, . ' as a summanon of all th ' . .within the lar""'r mono hi' e reahzed niches of all the component specIes

0- uvp y etlc taxon' the J' . .unclear, ' re alton of this concept to evolutionary theory IS


occur, not that the hypothesis of adaptive radiations presented hereis correct in any or all of its details,

The test case involves the Calmoniidae, a family of acastid trilo-bites (superfamily Acastacea: see Eldredge, 1979a, for a brief dis-cussion of higher-level systematics of these trilobites), The Cal-moniidae were endemic to the Malvinokaffric Faunal Province of theUpper Silurian and Lower and Middle Devonian. The faunal provinceencompassed the marine environment south of 600 south latitude,and included an Andean component (southern Peru, Bolivia, andnorthern Argentina), the Amazon Basin of Brazil, the Falkland Is-lands, and South Africa, Ghana and Antarctica may also have be-longed to the province. Geophysical evidence places the South Polejust north of Cape Town in the Lower Devonian. All of these continen-tal areas were united into the supercontinent Gondwana during thistime. The total span of time during which this clade persisted was atleast 20 million years. However, the record is brief on all platformareas, and it is only within the Andean sequence that anything like acontinuous stratigraphic record is available, with much more of the20 million years represented in some local sections up to 10,000 feetin thickness. Stratigraphic correlation between areas is still quite ru-dimentary, and the amount of time missing from the record betweenthe Upper Silurian and the lowermost Devonian beds is not as yeteven roughly known. Despite the patchy distribution of rocks andfossils, enough of a pattern remains to indicate a radiation.

The Calmoniidae consist of some 29 known genera and subgen-era, and at least 60 species, according to a recent tabulation (El-dredge and Ormiston 1979). If this seems a moderate diversity oftwenty million years or so, it is to be recalled that (1) Only some frac-tion of existing species have been fossilized, sampled, and de-scribed and (2) Nonetheless, compared with most other known trilo-bite groups over comparable spans of time, diversity is high indeed(there are, in addition, many other noncalmonud trilobites in thefauna), Thus the pace of evolution, at least by tri 10bite standards,was rather high for Calmoniidae.

In figure 6.15 we present a cladogram of relationships among alldescribed genera and subgenera. The cladogram can hardly besaid to be "highly" corroborated; it is defended in outline in El-dredge and Branisa (1980) and more rigorously and completely in

<,__ BAlHEt.LA (1,51

BELENOPS f1,3)

TARklACTINOIDiES m


Eldredge (unpublished manuscript). The range of anatomical varia-tion within the Calmoniidae is unusually great. The pattern of rela-tionship shown on the cladogram indicates that the variability is par-celled out in a regular fashion: there are (a) a few calmonllds soplesiomorphic as to resemble most closely their more primitiveacastid relatives. (Acastids are cosmopolitan in the Silurian and areknown from Australia and the Northern Hemisphere in the Devonian.)The remaining calmoniids are divided into two subfamilies, the (b)Calmoniinae. containing two related subgroups, in which there is anarray of different forms united by a distinctive set of features, and (c)the Metacryphaeinae, members of which, while united in some fea-tures, are greatly different one from another and reminiscent of other,nonacastid groups of trilobites. The Metacryphaeinae thus fulfill thestandard conception of an adaptive radiation. Details of the com-position of these groups, including the anatomical peculiarities ofthe component species and genera, are given in Eldredge (un-published manuscript).

We now compare the set of predictions with properties (ob-served or hypothesized) of the component species. The first crite-rion, that stenotopes ought to outnumber eurytcpes. cannot be exam-ined directly. Indirectly, the restriction of most species to specificsedimentary environments (see Eldredge and Ormiston 1979), aswell as to highly localized areas, agrees with, but does not criticallytest, the prediction. We also predict high rates of extinction (rela-tively short stratigraphic ranges), a high degree of (sub)regional en-demism, and a characteristic preponderant development of au-tapomorohic features as anatomical markers of specific-leveladaptations. Figure 6.16 is a chart of the approxiate stratigraphicranges of all known calmoniid species in Bolivla.t" The pleaicrnor-phic genera Acastoides and Voges ina each have long-ranging spe-cies, and each genus persists through most of the total time span.Only species of Schizosty/us, Metacryphaeus, Ma/vinel/a, and Bou-/eia exhibit comparable ranges. Apart from the two species of the

ANDINACASTE (1,2)

I'tlACOPlNA ((,2)

ACASTOIO£S III

YOGESINA 0)

NEW GENUS (2)

CALMOHIA (2,3)

PENNAIA (2,?3)

n PARACALJ,IONIA (2)•c~

OELTACEPHALASPIS '"

I~PRESTALIA (I)

SCHIZOSTYLUS III

CURUYELLA (1,3)

NEW GENUS 1Il

PR080LOPS rn

NEW GENUS (I)

CRYPHA£OIOES III

I(OZLOWSI(IASPlS (I)

ROMANIELLA OJ

METACRYPHAEUS (1,2,3)

MALYINELLA (I)

NEW GENUS (2)

NEW GENUS (I)

PARA90ULEIA (I)

90ULEIA (I)

TYPHLONISCUS (3)

NEW GENUS ru12. Correlation of the Andean sequence with other regions in the Malvinokaffric Province is in-sufficiently precise to incorporate non-Andean calmoniids on this chan, Inasmuch as 23 of the29 genera and subgenera of Calmonidae (i.e., 79 percenn occur in the Andean region, the chanof figure 6.16 should effectively mirror the gross pattern of temporal variation among species ofthe Calmoniidae. The chart is stylized; e.g., the concordant ranges within the Scaphiocoeliazone reflect isolated presence wilhin the zone, not necessarily concordant appearance and ex-tinction events.


Silurian plesiomorphic taxon Andinacasfe, the beginning of the radi-ation appears in the Scaphiocoelia beds, with 2 of the 14 calmoniidspecies plesiomorphic acastidlike forms, the rest being rather highlyderived. The Ca/monia group presents a rather odd mixture of primi-tive and derived forms: the group as a whole retains the preslornor-phic. typically acastid construction of the central region of the head(glabella), and at the same time, each subgenus is marked with astriking set of autapomorphies involving development of unusualspinescence and other features. This group undergoes its own "radi-ation," and only one genus (Schizosfy/us) has surviving species inyounger horizons. Thus, in some respects, the more primitivemembers of the Calmoniidae, the Cafmonia group of the Cal-moniinae, occur in the earliest horizons-yet each species is, aswell, highly autapomorphic, presumably reflecting specialization.Each species is short-ranged and highly localized. There is no sym-patty among Con(sUb)generic species (according to the taxonomicscheme adopted prior to this analysis), but there is extensive sym-patry among species of these closely related aenera.»

The Prob%ps group is composed of five known species, eachquite autapomorphic and each with a very short known stratigraphicrange. The group is entirely confined to the Andean region. Nonethe-less, the group is present throughout the duration of Malvinokaffricprovincial time. We return to the Prob%ps group, as an example oftrends, below.

The anatomically highly varied Metacryphaeus group is repre-sented in the Scaphiocoefia Zone by two species of Romaniefla, theaoornorpnn, sister-group of the genus Koz/owskiaspis, and by a spe-cies of Paraboufeia. Originally described (Eldredge 1972b) as theprimitive ancestor of the highly apomorphic eootete, additional ma-terial has shown that, in addition to the relatively plesiomorphic gla-bellar features, Parabou/eia Possesses a highly derived, stalkedlensless "eye," whereas its sUpposed descendant, Bouleie, retainsnormal calmoniid eyes. The group dominates the bulk of the middleand upper sediments of the Devonian sequence in Bolivia, andshows a mixture of short- and long-ranging species. The group is

13. Given the fact thaI the Malvinokaffric Province OCCupied a relatively small area, there is aneXtremely high degree of endemism within it. Geographic ranges are given on the c1adogram(figure 6.15); only MetacryPhaeus occurs in all three recognized subregions of the province.The bIogeography of these taxa is discussed e:nensively in Eldredge and Ormiston (1979).

2

,IC"SIC"(OE"A")

SEDS

,~.seLENBEDS .,

gs,~0, ,

; , ,I

I, I

SC'P~IO-cow'aeos III ,

ill

s of known speciesFO 6 16The approximate relative stratigraphic rangeOfl~~~o~iid trilobites of Bolivia. (See text.)

Arrested Evolution

~rrested evolution [8im' "tlonally defined as littl peon s (1944) bradytely"] has been tradi-I I e or no mar holoo!o a monophylel·,c I" P 0 oqlcal change within species. . meaqe Com 1million years and I . man y, cases involve at least 100

, requently mreell.y with a nearly identi . are, where a fossil is compared di-fossils," moreover I d cal living representative. So-called "living

,entobemeboppossums (metathe . m ers of pJesiomorphic clades-{i . nan mammals). .Inarticulate brachio d VS.euthenan mammals lingulid

I po s) VS othe b . 'P acophoran mollusc) . r rachlopods, Neopifina (a mono-that these "plesiom VhS:other Mollusca, and so forth. (We admitth orp rc clad "ough none is blatantl es may not all be monophyletic,(1953:319 ft. and 303 ft r a nonmonophyletic, not-A group). Simpsonerar phenomenon from' t:rese~ts an excellent discussion of this gen-

Another general I IS pomt of view,eature of brad I I· .ye IC Irneages, less frequently

Steady State


highly divergent there bein n d' .derived features' ·Ih· h g 0 rscerntbla spectrum of primitive-

WI m t e cladogram (We m f thiin another connection below.) , en Ion IS group again

There is clearly a gre t d Ibites' the an~IYs·,s I' th a ea more to be said about these trilo-

, ,or e present purpo .Enough of the patt h ses, must be rather skimpy.an adaptive rae! I~rn as bee,n presented, however, to indicate that

ra Ion, when Viewed as a tt I . ..and differential exf I" pa ern 0 species onqms

mo Ions IS a comple ff' , hsubgroups developi then x a air, Wit monophyleticreiterate that '1' ng err own patterns. We should at this point

I IS as yet unclear h d ' ,these patterns may be di " ow eterministic components ofp. 298 for discussion) Nsc~mlnated from random components (seechance alone' still . 0 oubt many of these patterns are due todoes not do 10' b more may be attributable to sampling error. It

e overly opt" . tic ! .stories about such tt nrus IC m framing deterministic just-socomponents wh,·ch p~ ems, But the predictions about the species

ansa from the "ditf 'preach to macroevol t! I erential species survival" ap-tantly, they might n ~ ha, hold true in a general way, More Impor-analyses of these :nd :~~ held true, and may not in subsequentdata pertaining 10 er sets of data, We believe that other

apparent "ad .amined from this p . a apt/Ve radiations" should be ex-erspectlve,


discussed, is their low species diversity.t" Given the possibility thatit is the constant low diversity through time of bradytelic lineages(commonly, but not necessarily, after an initial "burst," or radiation)that has controlled the low observed rate of morphological change,arrested evolution can be characterized (see table 6.1) as low diver-sity for over, say, 80-90 percent of the time of a clade's persistence,a nearly equal ratio of speciation and extinction, and a low rate ofspeciation. These are commonly observed patterns, as for examplein dipnoan and xiphosuran phylogeny, as graphed in Eldredge(1979b).

The mode of species selection would appear to be "neutral"-that is to say, one species is much like another throughout the geo-graphic range and the geologic history of the group. The constituentspecies are, hypothetically at least, nearly interchangeable. Thisconsideration arises from the testable predictions: species withinbradytelic lineages are eurytopic, geographically widespread, dis-play long stratigraphic ranges (rate of extinction is low), and rarelydevelop autapomorphic anatomical specializations, but instead re-tain and pass along plesiomorphic anatomical features, Interspecificmorphologic differences are subtle-there is a monotony about thecollective morphologies of these groups, which of course initiallyinspired the question of arrested evolution. Eldredge (1975 andespecially 1979b) has considered the conceptual link between theseparameters at length and has claimed that the hypothesis ofbradytely stated above, when tested against actual examples, hasyet to be refuted, The reader is referred to these papers and toStanley (1975) for a more detailed discussion of the phenomenonfrom the standpoint of "species selection."

The next macroevolutionary pattern in the list of table 6.1 is the mostimportant-at least if "most common" suffices as a criterion of high

14. Eldredge (1979b) has discussed the chicken-and-egg dilemma inherent in this problem: dobradytelic lineages display low diversity because there is 100 lillie morphological transformationto allow a systematist to "see" change and thus recognize and name more species or are specia-tion rates so low in these lineages thatHttle morphological change can occur? The former possi-bility keeps the issue safely within standard transformational lines of argumentation. wltereasthe latter expands the question into lhe area of interspecific macroevolutionar)' phenomena ashere construed,

1 _

-

I


making relatively precise, testable predictions about componentspecies. But steady state seems to be by far the most common situa-tion in macroevolution, and we include it here to show that it, too,can be understood (at least conceptually) in term of patterns of spe-ciation and differential species survival. As an example, we cite theMetacryphaeus group of the calmoniid trilobites discussed above.The pattern is seen within the speciose genus Metacryphaeus itself(no real change accumulating within this genus through time) andwithin the entire group. The group, to be sure, is anatomically highlyvaried, but the limits of this variation are present from the earliest ap-pearance of the group, suggesting a centripetal mode, with no fur-ther development of major anatomical specialization.


importance. Simply labelled "steady state," we refer to the situationof long periods of equilibrium diversity shown by the bulk of groupsof organisms represented in the fossil record. Major changes in di-versity typically are correlated with similar events in other, unrelatedmonophyletic groups-particularly episodes of widespread extinc-tion, followed in some instances by a rediversfficaticn which mayormay not bring diversity rapidly back to a level even higher than theprevailing equilibrium level prior to extinction. Whatever its specificmanifestation, steady state is "normality." It is also macroevolution(though rarely so considered) as it clearly represents a pattern ofspeciation and differential species survival, an equilibrium interplaywhereby the ratio of speciation to extinction is, on the average, aboutone, and rates of speciation can be low, but are usually moderate torapid.

The mode of species selection is hypothesized to be either neu-tral ?r "centripetal," where radically new morphologies, when ap-pearing, do not lead to new subclades. hence little substantial di-vergence (anatomically speaking) occurs. The reason this generalarea of "normal macroevolution" is little discussed is that the basiccharacterization of its essential pattern (moderate to high speciationrates, neutral or centripetal species selection) is rather messy andimprecise, as are the attendant predictions. These are that thereshould be (a) a mixture of eurytopic and stenotopic ecological stra-tegies; (b) a range from high endemism to virtual cosmopolitanism ingeograp~ic occurre~ces ~f individual species; (c) a wide range inthe duration of stratigraphic occurrences of individual species; and(d) a mixture of specialized and generalized (relatively apomorphicand .plesio~orphic) forms. Furthermore, these variables might be ap-portioned In a nonrandom way among monophyletic subgroups(e.q , some species-groups may be more eurytopic and plesiomor-pbic, etc. than their sister groups), but the variables might also be"normally" d~stribu~ed such that, for instance, there may be a spec-tr~m of .stratlgraphlc or geographic ranges, but that most speciesmight display some intermediate value.

It is th: extreme cases-the radiations, the cases of bradytely,a~d the eXls.ten~e of trends-which traditionally have prompted de-tailed examination as macroevolutionary phenomena rather than"st:ady state." The reason seems to be that, in the steady state, thevan abies are quite "messy," and there is concomitant difficulty in

Trends

The final macroevolutionary topic listed in table 6.1 is trends, Wehave discussed trends at length (see pp. 266, 283) in our attempt toshow that evolutionary phenomena can profitably be viewed as "de-coupled" from microevolutionary processes. Consequently, we shalllimit our discussion of this phenomenon here to a brief character-ization of its properties, and to predictions which can be tested,

Eldredge and Gould (1972:111) and Schaeffer, Hecht, and El-dredge (1972:35) have noted that many trends seem to be selected aposteriori renderings of morphological sequences arranged in stra-tigraphic order. However, trends do exist: the comparative anatomyof anthropoid primates suggests that the present brain size of Homosapiens is an enlargement over the plesiomorphic condition for ho-minids. Discounting a saltationist assumption, it might be predictedthat, were an adequate sequence of hominid crania available, a cor-relation between progressive enlargement of cranial capacity andprogressively younger sediments would be found. And this is what isfound, at least in the overall pattern of the fossil record.

This and other similar trend patterns seem to involve low diver-sity, a speciation/extinction ratio of about 1, and moderate to highrates of speciation. The hypothesized mode of species selection isdirectional (the sense in which the expression "species selection" ismost commonly used, e.g., Grant 1963, 1977; Stanley 1975; Gould1977b; Gould and Eldredge 1977). Predictions include: (1) The spe-cies are stenotopic with respect to the structures and behaviors dis-

I

Systematics and the Evolutionary Process 325of little or no appreciable change in cranial capacity within a spe-cies. The trend is not, manifestly, a function of intraspecific, gradual,progressive, microevolutionary processes. We owe our braininess, itwould seem, to directional species selection.


playing the trend-Le., these features are to be regarded as specia-lizations pertinent to the actual occupation and exploitation ofniches; (2) Stratigraphic ranges should be relatively short; (3) Geo-graphic spread probably limited; and (4) Almost by definition, thereshould be a progression from relatively more plesiomorphic to rela-tively more apomorphic forms, coincident with the stratigraphic occur-rence of the species. Actually, the latter "prediction" is the very crt-terion of recognition of a trend, and the pattern of species diversitycan, in this case, perhaps be regarded as a prediction.

The Prob%ps group of calmoniids offers a test case. Earliestmembers develop a spine on the middle of the glabella (rather like aunicorn) and have thoracic and pygidial spines which emerge abovethe margin. Progressively up through the stratigraphic column, thesespines occur nearer to the margin, so that in the youngest-knownspecies (Cryphaeoides rostratus), the head spine is directly frontal,and the thoracic and pygidial spines are likewise entirely marginal.Whether the component species are stenotopic or not, their geo-graphic and stratigraphic ranges are definitely short, and the struc-tures involved are progressively apomorphous through time.

But perhaps the best test of the general species-selection hy-pothesis of trends is the lack of correlative changes within the stra-tigraphic ranges of component species. We did not list this as aprediction because (see pp. 285) an intraspecific trend may be coin-cident with an interspecific trend (I.e., in the same feature) and yetnot, strictly speaking, falsify the hypothesis (or, of course, falsify theconventional syntheticist hypothesis either). But the converse holds:demonstration that an interspecific trend is not coincident with what-ever pattern pertains within a species falsifies the hypothesis of con-tinuity between microevolution and macroevolution. And. perhapsonly by dint of lack of imagination, these are the only two sets of hy-potheses we have been considering.

In any case. the Probofops group displays insufficient stra-tigraphic evidence to evaluate within-species patterns. But thehominid fossil record (figure 4.10) while agreeing with the generalset of predictions (increase in brain size is here assumed to beadaptive and related to niche occupation and exploitation of ho-minid species), also demonstrates (a) considerable stratigraphicoverlap between species with different cranial capacities and (b)long periods (i.e.. the entire stratigraphic range of a given species)

Systematics and the Evolutionary Process: A Summary

The main points we have argued in this chapter are as follows:1. We have two basic disagreements with contemporary evolu-

tionary theory. The first is methodological, and the second substan-tive. (i) Corroborated theories of relationship can be used as part ofthe data array to test predictions of evolutionary patterns stemmingfrom theories of the evolutionary process. However, much of contem-porary evolutionary theory, especially that pertaining to interspecificevolutionary phenomena, has not been formulated in a manner whichrenders evolutionary hypotheses easily susceptible to criticism. (Ii)The standard syntheticist assumption is that processes consideredimportant in changing gene content and frequency within popula-tions, especially natural selection, account for all such changes inevolution. Within-population mechanisms are directly extrapolated tohigher taxonomic levels in an effort to articulate such a connection.We do not argue here that within-population processes as en-visioned in the neo-Darwinian paradigm are false. We recognize thatthe within-population concept of adaptation via natural selection is aviable hypothesis accounting for the deterministic part of much ge-netic, behavioral, and morphological change in evolution. Howeverwe strongly disagree that a smooth extrapolation of within-population(rnicroevclutionary) processes is a logical or effective integration ofsuch within-population mechanisms with among-species phenom-ena (macroevolution).

The major conceptual problem with such a direct "syntheticist''extrapolation of corroborated microevolutionary processes as mech-anisms for the production of macroevolutionary patterns, is that itcollides with a major, equally well-corroborated hypothesis of sys-tematic biology: species are real, discrete entities in nature andhave, among other things, origins, histories, and extinctions. The as-

j


mathematically. The standard nee-Darwinian paradigm, in basic out-line, seems well corroborated. Natural selection is the dynamic ef-fecting the deterministic component of such change. Generationaldata are needed to test such hypotheses; thus microevolutionarytheory and the phenomena associated with it (especially natural se-lection) are beyond the normal purview of systematic biology. (ii) Hy-potheses concerning mechanisms of speciation require detailedphylogenetic trees for testing-a problematical requirement. None-theless, the broad outlines of the mechanisms of speciation seemfairly well worked out, mostly from pattern analysis by systematistsworking in the field. (iii) Hypotheses concerning macroevolutionaryprocesses require a cladogram (t.e.. a well-corroborated hypothesisthat a group is monophyletic) plus distributional data pertaining tocomponent species. The data of systematics seem most appropriateto the testing of macroevolutionary hypotheses. The dynamic in-volved seems to be a process of differential species survival (t'spe-cies selection"-an analogue, to be sure, of natural selection). Interms of the Linnaean hierarchy, there is nothing more to macroevo-lution than species, inasmuch as taxa of higher rank than species donot exist in the same sense as do species, and thus can in no waybe construed as evolutionary units; rather they are (when monophyle-tic, t.e., correctly delineated) expressions of the branching patternproduced by many speciation events through time.

3. We then argue, following Stanley (1975), Salthe (in press),and earl ier workers, that, to some extent at least, the phenome-nological levels of evolution are "decoupled." l.e.. that each levelhas its own internal dynamic and thus works at least partially in-dependently of its neighboring upper or lower levels. Each level canbe understood on its own terms, without complete reference to theterms of the next lower level. The levels are only partly reducible or"decomposable." Arguments and test examples are presented insupport of the decoupling hypothesis.

However, once decoupling is established, we examine the na-ture of the interactions between the levels. The levels are by nomeans totally divorced, and phenomena at one level are fre~uent!y"resolved" at successively lower levels. Specifically, adaptation vianatural selection may account for all observed morphologiesthroughout the biotic world since its inception; but it does so only inthe sense that species are always adapted to their own niche (by


sumption that Iife has evolved, from the standpoint of systematics,implies that species evolve from other species. Theories of specia-tion abound. Speciation is not, fundamentally, a process of adapta-tion. Therefore, a theory emphasizing adaptation at its core cannotproperly be extrapolated smoothly from the level of microevolution tothe level of macroevolution. The conflict with the syntheticist form ofmacroevolutionary theory, which is a direct, wholesale extrapolationof within-species, rnicroevotutionary theory arises from the necessaryobliteration of species as the basic evolutionary units in the synthe-ticist view. In the transformational approach, the emphasis is on the(adaptive) change of the intrinsic features of a taxon into some otherform, retrospectively classified as a new species, and perhaps anew genus, family, order, etc. This emphasis on the transformation ofmorphology as the basic question of evolution permeates not onlyayntheticist formulations of macroevolutionary theory, but also thewritings of such nonconformists as Grasse, Goldschmidt, and ethers.Most such theories represent a direct fusion of population (and otherareas of) genetics with the data and ideas of paleontology. We havedubbed all such formulations, syntheticist or not, the "transforma-tional" approach to evolution.

The "taxic'' approach, on the other hand, recognizes that spe-cies are evolutionary units. A macroevolutionary theory is incompleteif it does not take this generalization into account. We have reviewedthe major contributions of both approaches to macroevolutionarytheory, showing how the preponderance of work has argued fordirect extrapolation of within-species phenomena as a realisticmechanism for the production of among-species phenomena, andalso showing that some biologists have seen the conceptual dif-ficulty if speciation theory is omitted.

2. We further argue that if evolutionary theory is to be improvedand better integrated, its phenomenological components must firstbe recognized. We agree with some previous authors who recognizea distinction between the levels of (a) rnicroevolution (change ofgene content and frequency within species) and (b) macroevolution(change of species composition in time and space within a mono-phyletic group). Speciation is the process separating the two phe-nomenological levels.

With regard to these levels: (i) Hypotheses concerning rnicroevo-lutionary phenomena can be tested experimentally and modelled

i

IIJ

r


definition, nearly), possessing anatomical. physiological, and be-havioral properties appropriate to each individual niche. In-terspecific interactions (e.g, competition) may be resolved, in part,by microevolutionary changes in sympatric populations of one ormore of the species involved. At the higher level of macroevolution,each component species is assumed to be well adapted to its niche:the process of differential species survival therefore involves fun-damentally a rearrangement of niche space through time. This maybe worked out, in part, by microevolutionary change, but at base ispatently not a matter of progressive adaptive change in gene contentand frequency within a species in response to new or changing envi-ronmental conditions.

4. The basic outline of a testable theory of macroevolution isthus indicated. Such a theory requires (a) a well-corroborated clade-gram (to insure that one is dealing with a monophyletic taxon and toevaluate distributions of pleslomorphies and apomorphies of compo-nent species) and (b) a pattern of distribution in space as well as intime (if available) of the individual species sampled. We recognizethe strong "random" component of such shifts in diversity patternswithin monophyletic groups through time, but argue that a deter-ministic view is a priori at least equally plausible.

We then present an outline of a specific macroevolutionarytheory. It contains little that is new, its originality residing in its form.The major virtue we claim for the theory is that it is testable, i.e..predictions concerning component species of a monophyletic groupexhibiting a particular pattern of diversity can be generated and tes-ted, and that it explicitly incorporates the origin and extinction ofspecies

The theory draws on ecological niche theory as the conceptualnexus between morphology (adaptation) on the one hand, and spe-cies diversity (seen as an interplay between rates of speciation andrates of extinction) on the other hand. This is to say that both adapta-tion (as reflected in anatomical and behavioral specializations whichspecies possess) and diversity (number of species living in a givenarea) are related to the occupation and exploitation of ecologicalniches. We then review some of the theoretical work on the controlsof speciation and extinction. The central hypothesis, in this connec-tion, is that broad-niched, relatively eurytopic species (a) tend toreact to interspecific competition by mutual exclusion, (b) have low


extinction rates (i.e. relatively long stratigraphic ranges), (c) tend tooccur over wide geographic ranges (relative to the collective rangeof all of the species within the monophyletic group), and (d) retainplesiomorphies, i.e., species tend to resemble one another closely,and few if any species develop autapomorphies which can be hy-pothesized to be evolutionary specializations related to the occupa-tion of a particular niche. Relatively stenotopic species (a) tend toreact to interspecific competition by subdivision of niche space, (b)have high extinction rates, (c) exhibit relatively narrow geographicdistributions, and (d) tend to develop anatomical specializationswhich appear as autapomorphies.

With the general notion of factors controlling rates of speciationand extinction in mind, we constructed a table (table 6.1) on whichvarious macroevolutionary patterns ("adaptive radiations," "arrestedevolution," "steady state," and "trends") are characterized in termsof (a) diversity patterns, (b) ratio of speciation to extinction, and (c)absolute rates of speciation as criteria for recognition of the patternitself. We then give a conjectural characterization of the apparentmode of "species selection" for each macroevolutionary pattern. Thetheory is evaluated in light of the predictions concerning the comp,o-nent species which flow from it for each case, The ~redicti~ns .In-

clude geographic distributions, stratigraphic range (i.e. extinctionrate of species), relative eurytopy or stenotopy of component spe-cies, and the distribution of plesiomorphous and apomorphous con-ditions within the species. We test the theory with particular casesrelevant to the four macroevolutionary patterns tabulated. We do notclaim that the theory is highly corroborated. Our n:ain co~cern,rather, is to elaborate a theory of macroevolution consistent with ~hephenomenological levels we recognize in evolution-~ theory Whl.chboth explicitly incorporates the existence and evolutl~n of speciesand which is in a form whereby it can be tested and either corrobo-rated or rejected. Success in meeting these two objectives far out-weighs the importance of the eventual fate of the theory Itself.

,,------References

Adanson, M. 1763, Families des plantes. vel. 1. Vincent, Paris.Agassiz, J. L. R. 1844. Recherches sur les poissons fossiles. Neuchiitel,

Switzerland.Anderson. D, T. 1973. Embryology and phylogeny in Annelids and Arthro-

pods. Pergamon Press, Oxford.Ashlock, P. D. 1971 Monophyly and associated terms. 5yst. Zool. 20:63-69.Ashlock, P. D. 1972. Monophyly again. Syst. zo». 21:430-37.Ashlock, P. D. 1974. The uses of cladistics. Ann. Rev. fcol. Syst. 5:81-99.Ayala, F. J., ed. 1976. Molecular Evolution. Sinauer Ass., Sunderland, Mass.Bassindale, R. 1936, The developmental stages of three English barnacles,

Balanus balanoides .(Linn.), Chthamalus stellatus (Poll), and Verrucastroemia (O.F. Muller). Ptoc. Zoot. Soc. London 1936:57-74.

Bell, B. M. 1976. Phylogenetic implications of ontogenetic development inthe Class Edrioasteroidea (Echinodermata). J. Paleontol. 50:1001-19.

Bock, W. J. 1963. The cranial evidence for ratite affinities. Ptoc. 13th Int.

Omithol. Congr., vet. 1, pp. 39-54,Bock, W. J. 1965, The role of adaptive mechanisms in the origin of higher

levels of organization. Syst. Zoo/. 14:272-87.Bock, W. J. 1970. Microevolutionary sequences as a fundamental concept in

macroevolutionary models. Evolution 24:704-22.Bock, W. J. 1972. Species interactions and macroevolution. Evol. Bioi.

5:1-24.Bock, W. J, 1973. Philosophical foundations of classical evoiutionary clas-

sification. Syst. Zool. 22:375-92,Bock, W. J. 1977. Foundations and methods of evolutionary classification. In

M. K. Hecht, P, C. Goody, and B. M. Hecht (eds.). Major Patterns in Ver-tebrate Evolution, pp. 851-95. Plenum, New York.

Bock, W. J. 1979. The synthetic explanation of macroevolutionary change-areouctlonfstrc approach. In J. H. Schwartz, and H. B. Rollins (eds.),Models and Methodologies in Evolutionary Theory, pp. 20--69. Bull. Car-

negie Mus. Nat. Hist. 13.Bock, W. J., and G. Von Wahlert. 1963. Two evolutionary theories-a discus-

sion. Brit. J. Phil. Sci. 14:140-46.Bonde, N. 1977. Cladistic classification as applied to vertebrates. In M. K,

332 References References 333

Hecht, P. C, Goody, and B. M. Hecht (eos.), Major Patterns in VertebrateEvolution, pp. 741-804. Plenum, New York.

Boucot, A. J. 1975. Evolution and Extinction Rate Controls. Elsevier, NewYork.

Boucot, A. J. 1978- Community evolution and rates of cladogenesis. Evol.BioI. 11:545-655.

Bretsky, P, W., and D. M. lorenz. 1969. Adaptive response to environmentalstability: a unifying concept in paleoecology. Proc. North Amer, Pa/eon-tal. Conv., part E, pp. 522-50.

Bretsky. P. w., and D. M. lorenz. 1970, An essay on genetic-adaptivestrategies and mass extinctions, Bull, Geol. Soc. Amer. 81 :2449-56,

Bretsky, S. S. 1979, Recognition of ancestor-descendant relationships in in-vertebrate paleontology. In J. Cracraft and N. Eldredge {eds.}, Phy-logenetic Analysis and Paleontology, pp. 113-63. Columbia Uruv.Press, New York.

Brinkmann, R. 1929. Statistische-biostratigraphische Untersuchungen anmittel jurassischen Ammoniten iiber Artbegriff und Starn-mesentwicklung. Abh. Ges. wzss. G6ttingen, Math-Phys. K1., N. F., vet.8, pt. 3,

Brown, W. L, and E. O. Wilson. 1956. Character displacement. Syst, Zool.5:49-64.

Brundin, L 1966. Transantarctic relationships and their significance, as evi-denced by chironomid midges. Kungl. Svenska Vetenskap. Handl.11:1-472.

Buck, R. C" and D. L Hull. 1966. The logical structure ofthe linnaean hier-archy, Syst. Zool. 15:97-111.

Burkhardt, R W., Jr. 1977, The Spirit of System: Lamarck and EvolutionaryBiology. Harvard Univ. Press, Cambridge.

Bush, G. L 1975. Modes of animal speciation. Ann. Rev. Ecol. Syst.6:339---64,

Cain, A. J" 1954, Animal Species and Their Evolution. Hutchinson and Co.,london, (Reprinted 1960, Harper, New vork.)

Cain, A. J., and G. A. Harrison. 1960. Phyletic weighting. Proc. Zool. Soc,London 135:1-31.

Carruthers, R. G, 1910, On the evolution of Zaphrentis delanouei in lowerCarboniferous times. Quart. J. Geol. Soc. London 66:523-38-

Carson, H. L. 1975, The genetics of speciation at the diploid level. Amer.Nat. 109:83-92.

Coleman, W. 1964. Georges Cuvier Zoologist. Harvard Univ. Press, Cam-bridge, Mass.

Coiless, D. H. 1977, A cornucopeia of categories. Syst. Zool. 26:349-52.Cracraft, J. 1974a. Phylogeny and evolution of the ratite birds. Ibis

116:494---521Cracraft, J. 1974b. Phylogenetic models and classification. Syst. Zool.

23:71-90.

Cracraft, J. 1978. Comparative biology and brain evolution. Sysl. Zool.27:260-64. d I

Cracraft J, 1979. Phylogenetic analysis, evolutionary models an p~ eon-tology. In J, Cracraft and N. Eldredge (eds.), Phylogenetic AnalySIS andPaleontology, pp. 7-39. Columbia Unlv. Press, N~ York. ,

Crowson, R. A, 1970. Classification and Biology. Heinemann Educational

Books, London. M I EvolCrowson, R. A. 1972. A systematist looks at cytochrome c. J. o. .

2:28--37. . .' F'r AlcanOuenot, L 1932. La Genese des eeoecee animales. Llbreme e IX ,

Paris. . , 67 Ath neumDarwin, C. 1859. On the Origin of Species. Facsimile ed., 19, e ,

New York. Ox! ddeBeer G 1958 Embryos and Ancestors. Oxford Univ. Press, o~'.

" . if t! PrfnctpresDeleon, E. 1977, Catarrhine phylogeny and crass! rca Ion: ,methods and comments. J. Human Evo/. 6:433-.59. . . of the

Derson E and P Andrews 1975, Evolution and Interrelationships,., .' I (eds.) phylogenycatarrhine primates, In W. P. Luckett, and F, S. Sza ay "of the Primates, pp. 405-46, Plenum, New York, . f hominid

Deleon E N Eldredqe and I Tattersall. 1977. Reconstruction 0, " . " ., I is J Humanphylogeny: A testable framework based on cladistic ana ys ' .

Evol. 6:263-78. Col mota umvDobzhansky, T. 1937. Genetics and the Origin of Species. u .

Press, New York. . . '2nd rev. eo..Dobzhansky, T, 1941. Genetics and the Of/gm of SpeCies,

Columbia Univ. Press, New York. . . . 3rd rev. ed.,Dobzhansky, T. 1951. Genetics and the Orlgm of SpeCies,

Columbia Univ. Press, New York Columbia Univ.Dobzhansky, T, 1970. Genetics of the Evolutionary Process.

Press, New York, . 1977 Evolu-Dobzhansky, T., F, J. Ayala, G. L. Stebbins, and J. W. Valentine. .

tion. W. H. Freeman, San Francisco. . hi s of the but-Ehrlich, P. R., and A. H. Ehrlich. 1967. The phen.e.ti~ re~atl~~~esis. Syst. zo».

tertnes: I. Adult taxonomy and the nonspeclhclty yp

16:301-17. . . I nons ScienceEhrlich, P. R., and P. H. Raven. 1969. Differentiation of popu a I .

165:1228-32. I anian gastropodEldredge, N. 1968. Convergence be,:""een two p~n~sY;aleontol. 42:188--

species: A multivariate mathematical approac ..

96. in Paleozoic inver-Eldredge, N. 1971. The allopatric model and phylogeny

tebrates. Evolution 25:156-67. . ot Phacops rana (GreenEldredge, N. 1972a. Syst,ematic~ and evolutlon(Trilobita) from the Middle

1832) and Phacops 10wenSlS Delo, 1935 NtH' t 147'45-114Devonian of North America. Bull. Amer. Mus. a. IS. ' .

334 ReferencesReferences 335

Eldredge, N. 1972b. Morphology and relationships of Bouleia Kozlowski,1923 (Trilobita, Calmoniidae). J. Paleontol. 46:140-151.

Eldredge, N. 1974a. Stability, diversity and speciation in Paleozoic epeiricseas. J. Paleontol. 48:540-548.

Eldredge, N, 1974b. Testing evolutionary hypotheses in paleontology: Acomment on Makurath and Anderson (1973), Evolution 28:479-81.

Eldredge, N. 1975, Survivors from the good old, old, old days. Nat. Hist.84(2):60-69.

Eldredge, N. 1979a. Ctaotsm and common sense. In J. Cracraft and N. El-dredge (eds.), Phylogenetic Analysis and Paleontology, pp. 165-97.Columbia Univ. Press, New York.

Eldredge, N. 1979b, Alternative approaches to evolutionary theory, In J. H.Schwartz and H. B, Rollins (eds.), Models and Methodologies in Evolu-tionary Theory, pp. 7-19; Bull. Carnegie Mus. Nat. Hist. 13.

Eldredge, N" and l. Brantsa. 1980. Calmoniid trilobites of the Lower Devon-ian Scaphiocoelia Zone of Bolivia, with remarks on related species.Bull. Amer. Mus. Nat. Hist. 165(2):181-290,

Eldredge, N. and J. Cracraft. 1979, Introduction to the symposium, In J. Cra-craft and N. Eldredge (eds), Phylogenetic Analysis and Paleontology,pp. 1-5, Columbia Univ. Presss, New York.

Eldredge, N., and M. J. Eldredge. 1972. A trilobite odyssey. Nat. Hist.81(10):52-59,

Eldredge, N, and S. J. Gould. 1972. Punctuated equilibria: An alternative tophyletic gradualism. In 1. J. M. Schopf (ed.), Models in Paleobiology,pp. 82-115, Freeman, Cooper and Co" San Francisco.

Eldredge, N., and S, J. Gould. 1974, Reply to Hecht. Eval. BioI. 7:303-8.Eldredge, N.. and A. R. Ormiston. 1979. Biogeography of Silurian and De-

vonian trilobites of the Malvinokaffric Realm. In J. Gray, and A. J. Boucot(eos.), Historical Biogeography, Plate Tectonics and the Changing Envi-ronment, pp. 147-67 Oreg. State Univ. Press, Corvallis,

Eldredge, N" and I. Tattersall 1975, Evolutionary models, phylogenetic re-construction, and another look at hominid phylogeny. In F S, Szalay(eo.). Approaches to Primate Paleobiology; Contr. Primato!. 5:218--42,

Englemann, G. F. and E. O. Wiley. 1977. The place of ancestor-descendantrelationships in phylogeny reconstruction. Syst. Zool. 26: 1-11

Farris, J, S 1974. Formal definitions of paraphyly and polyphyly. Syst. Zoot.23:548-54.

Farris, J. S. 1976, Phylogenetic classification of fossils with recent species.Syst. Zool. 25:271-82.

Farris, J, S. 1977. On the phenetic approach to vertebrate classification, InM. K. Hecht, P. C, Goody, and B. M. Hecht (eds.), Major Patterns in Ver-tebrate Evolution, cp. 823-50, Plenum, New York,

Fryer, G" and 1. D. lies. 1969, Alternative routes to evolutionary success asexhibited by African cichlid fishes of the genus Tilapia and the speciesflocks of the great lakes, Evolution 23:359--69.

Gaffney, E. S. 1975. A phylogeny and classification of the higher categoriesof turtles. Bull, Amer. Mus. Nat. Hist. 155:387--436.

Gaffney E. S, 1977. The side-necked turtle family Chelidae: A theo.ry of rela-tionships using shared derived characters, »mer. Mus. Novltates No.

2620:1-28. tiGaffney, E. S. 1979, An introduction to the logic of phylogeny reconstruc IO~.

In J, Cracraft and N. Eldredge (eds.), phylogenetic AnalYSIS and Paleo -tology, pp. 79-111. Columbia Unfv. Press, New York, I

Ghiselin, M. 1. 1974. A radical solution to the species problem, Syst. Zoo.

23:536-44. . ceptGhiselin, M, 1. 1977. On paradigms and the hypermodern species con .

Syst. Zool. 26:437-38, ., ' . 0 in'sGhiselin, M. 1., and L. Jaffe. 1973. PhyIOgen~,tic classlflcatl~n 31;--40arw

"Monograph on the Sub-Class Cirrepedla, Syst. Zool. 22,1 " tGingerich, P. D. 1976. Paleontology and phylogeny: patterns.of ;~~I~~~.n a

the species level in Early Tertiary mammals, Arner. J. SCI, 2 .Gingerich P D 1979 The stratopbenetic approach to phylogeny recon-

, '. . d N Eldredgestruction in vertebrate paleontology. In J. Cracraft an. b.

. I P 41-77 Colum ra(eds.), Phylogenetic AnalySIS and Paleonto ogy, p . .Univ. Press, New York. . d 'mate

Gingerich, P. D., and M. Schoeninger. 1977. The tossn record an pn

phylogeny, J. Human Evol. 6:484-505: . U iv PressGoldschmidt, R. 1940. The Material BaSIS of EvolutiOn. Yale ruv. ,

New Haven, Conn. . s roupGorman G C 1968 The relationships of Anolis of the roquet specie g.

, ". . d f d! play behavior Brevlora,(Sauria: lquantdae}: III., Comparative stu yo IS .

no. 284, pp. 1-31. f HarvardGOUld, S, J. 1977a. Ontogeny and phylogeny, Belknap Press 0

Univ. Press, Cambridge, Mass, Hallam (ed.),GOUld, S. J. 1977b. Eternal metaphors of paleonIOI~9Y' In A. 1-26. EI-

Patterns of Evolution as Illustrated by the FOSSil Record, pp.

sevier, New York. .' he tempo andGOUld, S. J" and N. Eldredge. 1977. Punctuated eqUillb~I~: t

mode of evolution reconsidered. Paleobiology 3,1155. d D S Sim-GOUld, S. J., D. M. Raup, J. J. Sepkoskt Jr., T. J. M.. Schopf, an nd' r~ndom

berloff. 1977. The shape of evolution: a comparison of real a

clades. Paleobiology 3:23---40. . men! Bioi J.Grant, P, R, 1972. Convergent and divergent character displace . .

Linn. Soc. 4:39-68. . . s New York.Grant, V, 1963. The Origin of Adaptations. Columbia Un IV. Pres, ,G W H F eman San r-encrsco.rant, V, 1977. Organismic Evolution. ' -.'. re .' ,Grasse, P. P. 1973. L'Evolution au vivant. Albin Michel, P~rls. i s in Macro-Greenbaum I F and R J Baker 1976. Evolutionary ratatlonsb PIS st

tus (M~m'm~'lia; Ch;ro~tera); 'biochemical variation and karyo ogy. Y .

Zool. 25:15-25.

References 337336 References

Greenwood, P. H. 1974. The crchno fishes of Lake Victoria, East Africa: thebiology and evolution of a species flock. Bull. Brit. Mus. Nat. Hist.,suppl 6, pp. 1-134. , .

Greenwood, PH., R. 5, Miles, and C Patterson, eos. 1973, Interrelationshipsof Fishes. Academic Press, London.

Grene, M. 1959. Two evolutionary theories. Brit. J. Phil. Sci. 9:110-27,185-94.

Griffiths. G. C D. 1973, Some fundamental problems in biological classifica-tion, Syst. Zoot. 22:338-43.

Haldane, J. B. S. 1932. The Causes of Evolution. Harper, New York.Hampe, A. 1959. Contribution a l'etude du oeveloppement et de la requla-

non des detlcrences et des excedents dans la patte de I'embryon depoufet. Arch. Anat. Microsc. Morph. Exp, 48:345-78.

Harper, C, W., Jr. 1976. Phylogenetic inference in paleontology, J. Pa/eontol.50:180-93.

Hecht, M. K. 1965. The role of natural selection and evolutionary rates in theorigin of higher levels of organization. Syst. Zoot. 14:301-t7.

Hecht, M. K. 1976. Phylogenetic inference and methodology as applied 10the vertebrate record. Evo/. Bioi. 9:335---63.

Hecht, M. K" and J. L. Edwards. 1976, The determination of parallel ormonophyletic relationships: the proteid salamanders-a test case.Amer. Nat. 110:653-77.

Hecht, M, K" and J. L Edwards. t977, The methodology of phylogenetic in-ference above the species level. In M. K. Hecht, P, C, Goody, and B, M.Hecht (eds.), Major Patterns of Vertebrate Evolution, pp. 3-51. Plenum,New York.

Hennig, W. 1950. GrundziJge einer Theone der phylogenetischen Systema-tik. Deutscher Zentralverlag, Berlin.

Hennig, W, 1965. Phylogenetic systematics. Ann, Rev. Ent. 10:97-116.Hennig, W. 1966. Phylogenetic Systematics. Univ. III. Press, Urbana.Hennig, W. 1969. Die Stammesgeschichte der lneektea. Waldemar Kramer,

Frankfurt am Main,Hennig, W. 1975 Cladistic analysis or cladistic classtncanonv-..a reply to

Ernst Mayr, Syst, Zool. 24:244-56,Hopwood. A T. 1950a. Animal classification from the Greeks to Linnaeus, In

Lectures on the Development of Taxonomy, pp. 24-32. Linnean Soc.,London

Hopwood. A T 1950b Animal classification from Linnaeus to Darwin. InLectures on the Development of Taxonomy, pp. 46-59. Linnean Soc"London.

Hull. 0 L. 1964. Consistency and monophyly. Syst. Zool. 13:1-11Hull, D. L. 1965, The effect of essentialism on taxonomy-two thousand

years of stasis (1), Brit, J. Phil. Sci. 15:2-14.Hull, D, L. 1976. Are species really individuals? Syst. Zool. 25:174-91.Hull, D, L. 1978. A matter of individuality. Phil. Sci. 45:335--60.

m with special refer-H I Y J S 1958, Evolutionary processes and taxono y

ux e, . . 'k 1958'21-38ence to grades, Uppsa!a UnJ,v.z: 'h ' st for 'an optimal taxonomy.

Johnson, L. A. S, 1970, Rambow s en ,t e que

Syst. Zool. 19:203-39. I 3' Crustacea. John Wiley andKaestner, A. 1970. Invertebrate Zoology, vet. '

Sons, New York. , ic relationships of the storks.Kahl, M. P. 1971 Social behavior and taxonorm

Living Bird 10:151-70. . h Ciconiidae. The wood-storksKahl, M. P. 1972. Comparative ethology of t e

(genera Mycteria and Ibis). Ibis 114:15-7;he vertebrates. C. V. MosbyKent, G. C., Jr. 1965. Comparative anatomy 0

Co.. St. Louis, Mo. .' 'Pergamon, New York.Kerkut, G. A. 1960, The Implications of EV~lutlon. nt'spatial temporal andKollar, E. J. 197.2. The development of the ~n~=~~.me . ,

phylogenetic factors, Amer. zoot. 12:1 'P ss Cambridge. {Re-Lack, D. 1947. Darwin's finches. Cambridge Un IV, re ,

printed 1961 as Harper TorChb?ok.} H Elliot Trans. MacMillan andLamarck, J, B, 1914. Zoological Phriosophy.. ,

Co., London, th d Isis 58:304--20.Larson, J. L. 1968, Linnaeus and the .natural me '~12-30.Lewontin, A. C, 1978. Adaptati?n. ,SCI. Ame~a2r~~'Linnaei Systema Nat~rae,Linnaeus Carolus (Carl von Linne). 1758. H' t) London (Reprmted

, d B it Mus (Nat. IS., 'Regnum Anlmale, 10th e. fl. .

1939.) John Wiley and Sons, NewLovtrup, S, 1977. The phylogeny of Vertebrata,

York. of Island Biogeogra"MacArthur, A. H" and E, 0, Wilson. 1967. The Theory

phy. Princeton Univ. Press, Princeton, N.J, tmrtlve equids. with notes onMacFadden, B. J. 1976. Cladistic analysis of pn

other pertssodactyts. Syst. Zoo/. 25:1-14. tassrtcat'oo of the Mamma-McKenna, M. C, 1975. Toward a PhyIOgenetd,,)e Phylogeny of the Primates,

lia. In W. P. Lucket and F. S. Szalay (e s. ,

op. 21-46, Plenum, New York. ra- and interspecific variationMakurath, J, H., and E. J. Anderson. ,1973, .lnt

3_

10. .

in gypidulid brachiopods. Evolution. 27 ..30 and the origin and evolutionMargUlis, L. 1974, Five-kingdom claSSificatIOn

of cells, Evol, Bioi. 7:45-78. .' f phyletiC relationships. Syst.Maslin, T, P. 1952. Morphological cnterla 0

Zoo!. 1:49-70. . . s Amer. Nat. 74:249-!8.Mayr, E. 1940. Speciation phenomena In birdS' ecies. Columbia UniV. Press,Mayr, E. 1942. Systematics and the Gflgm of p

New York. (Dover edition, 1964.) . n Harvard Univ. PresS, Cam-Mayr, E. 1963. Animal Species and Evolutlo .

bridge, Mass. . McGraw-Hili, New york.Mayr, E. 1969. Principles of Systematic Zoology.

338 ReferencesReferences 339

Mayr, E. 1970. Populations, Species and Evolution. Harvard Univ. Press,Cambridge, Mass.

Mayr, E. 1974. Cladistic analysis or cladistic classification? Z. Zool. Syst.Evolut.-forsch. 12:94-128.

Meise, ~' 1963. Verhalten der Straussartigen Vogel uno Monophylie derPatttae. Proc. 13th Ornith. Congr. 1.115-25.

Michener, C. D. 1978. Dr. Nelson on taxonomic methods. Syst. Zool27:112-t8. .

Minkoff, E. C. 1965. The effects on classification of slight alterations in nu-merical technique. Syst. Zool. 14:196-213.

Morton, J, E. 1958, Molluscs. Hutchinson and Co., London (Reprinted, 1960,Harper, New York).

Moy-Thomas, J, A., and R. S, Miles. 1971. Paleozoic Fishes. W, B. SaundersPhiladelphia, '

Nelson, G. J. 1969. Gill arches and the phylogeny of fishes, with noteson the classification of vertebrates. Bull. Amer. Mus. Nat. Hist. 141:475-552.

Nelson, G. L 1970. Outline of a theory of comparative biology. Syst. Zool.19:373---84.

Nelson, G, L 1971a. "Cladism" as a philosophy of classification. Syst. Zool.20:373-76,

Nelson, G. J. 1971b. Paraphyly and polyphyly: redefinitions. Syst Zoot.20:471-72. .

Nelson, G. J. 1972a. Comments on Hennig's "Phylogenetic Systematics"and Its Influence on ichthyology. Syst. Zool. 21:364--74.

Nelson, G. J. 1972b. Phylogenetic relationship and classification. Syst. Zool21:227-31 '

Nelson, G, J. 1973a The higher-level phylogeny of vertebrates, Syst Zool22:87-91 . ,

Nelson, G, J. 1973b, Monophyly again?-a reply to P. D. Ashlock, Syst Zool22:310-12. . .

Nelson, G.,J. 1973c, Classification as an expression of phylogenetic rela-tionships. Syst, Zoot. 22:344--59.

Nelson, G, J, 1974, Darwin-Hennig classification: a reply to Ernst Mayr. SystZoot, 23:452-58 .

Nelson, G J. 1978. Ontogeny, phylogeny, paleontology, and the biogeneticlaw. Syst. Zool. 27:324--45.

Nelson, G, J.. 1979. Cladistic analysis and synthesis: principles and defini-tions, with a historical note on Adanson's "Families des ptantes"(1763-1764). Syst, Zool. 28:1-21

Nelson, G, J" and N, I. Platnick. 1980. Cladistics and Vicariance: Pat/ems inComparative Biology. Columbia Umv. Press New Yo'kN I G ' , .

e son, ,J., and~. E. Ro~en, eds. 1980, Vicariance biogeography: Congru-ence of earth history with plant and animal distributions. Columbia Univ.Press, New York.

Parkes, K. C., and G. A. Clark, Jr. 1966. An additional character linking ra-

tites and ttnamous. and an interpretation of their monophyly. Condor68:459--71

Patterson, C, 1977. The contribution of paleontology to teleostean phy-logeny. In M, K. Hecht, P. C. Goody, and B. M, Hecht (eds.), Major Pat-terns in Vertebrate Evolution, pp. 579---643.Plenum, New York.

Patterson,C" and D. E. Rosen. 1977. Review of icthyodectiform and otherMesozoic teleost fishes and the theory and practice of classifying fos-sils. Bull. Amer. Mus. Nat. Hist. 158:81-172.

Plafnick, N. I. 1977a. Paraphyletic and polyphyletic qroups. Syst. Zool.26:195-200.

Plalnick, N. I. 1977b. Cladograms, phylogenetic trees, and hypothesis test-ing. Syst. Zool. 26:438--42.

Ptatnick, N. I. 1977c, The hypochiloid spiders: a cladistic analysis, withnotes on the Atypoidea {Arachnida, Araneae), Amer. Mus. Novitates2627:1-23.

Platnick, N. I. 1978a. Classifications, historical narratives and hypotheses.Syst. Zool. 27:365-69.

Plalnick, N. I. 1978b, Gaps and prediction in classification, Syst. Zool.27:472-74.

Platnick, N. I., and W. J. Gertsch, 1976. The suborders of spiders: a cladisticanalysis (Arachnida, Araneae). Amer. Mus. Novitates 2607:1-15.

Ptatmck, N. I., and M. U. Shadab. 1976. A revision of the neotroptcat spidergenus Zimiromus, with notes on Echemus (Araneae. Gnaphosidae).Amer. Mus, Novitates 2609:1-24.

Popper, K. R. 1959. The Logic of Scientific Discovery. Harper Torchbooks,New York,

Prosser, C, L, and F. A. Brown, Jr. 1965. Comparative Animal Physiology.W, B. Saunders, Philadelphia.

Raikow, R. J. 1975. The evolutionary reappearance of ancestral muscles asdevelopmental anomalies in two species of birds. Condor 77:514-17,

Raup, 0, M. 1966. Geometric analysis of shell coiling: general problems.J. Paleontol. 40:1178-90.

Raup, D. M. 1972. Taxonomic diversity during the Phanerozoic, Science177:1065-71.

Raup, D. M, 1977, Stochastic models in evolutionary paleontology. InA. Hallam (ed.) Pat/ems of Evolution, as illustrated by the Fossil Record,pp. 59-78. Elsevier, New York,

Raup, 0, M" and S. J. Gould. 1974, Stochastic simulation and evolution ofmorphology-towards a nomothetic paleontology. Syst. Zoot. 23:305-22.

Raup, D. M., S, J, Gould, T. J. M. Schopf, and D. S. Simberloff. 1973, Sto-chastic models of phylogeny and the evolution of diversity. J. Geol.81:525--42.

Rensch, B. 1960, Evolution Above the Species Level. Columbia Univ. Press,New York.

Romanes, G. J, 1886. Physiological selection: an additional suggestion onthe origin of species. J. Linn. Soc. Zoo/. 19:337-411,

References 341

Simpson, G. G. 1978, Review of "Patterns of Evolution, as Illustrated by theFossil Record," edited by A, Hallam. Nature 273:77-78.

Sneath, P. H. A, and R. R. Sakal. 1973, Numerical Taxonomy. W. H. Free-

man, San Francisco ntSakal. R. R., and J H. Camino 1985. The two taxonomies: areas of agreeme

and conflict. Svsr. Zool. 14:176-95. '\" a cntt-Sakal,R. R., and 1. J. Croveno. 1970. The biological species concep.

cal evaluation. Amer. Nat. 104:127-53. bSakal, R. R., and F. J. Rohlf. 1962, The comparison of dendrograms by 0 -

jective methods. Taxon 11 :33-40. ,Sakal, R. R., and P, H. A. Sneath. 1963, Principles of Numerical Taxonomy.

W. H. Freeman, San Pransctsco. . I I Proc. Nat.Stanley, S, M. 1975. A theory of evolution above the species eve.

scea, Sci. 72:646-50. .' , . the case of theStenzel, H. B. 1949. Successional speciation In paleontology.

oysters of the sel/aeformis stock, Evolution 3:34-50, G H'I! NewStorer, T. I., and R. L. Us inqer. 1965. General Zoology. Mc raw- I ,

York. . nd testing ofSzalay, F. S. 1977, Ancestors, descendants, sister groups a

phylogenetic hypotheses. Syst. Zool. 26:12-18. d f t sy in humanTattersall, I., and N. Eldredge. 1977, Fact, theory an an a

paleontology, Amer. Sci. 65:204-11. I r of the LowerTrueman, A. E, 1922. The use of Gryphaea in the cone a Ion

Lias. Geol. Mag, 59:256-68. , I tructure of theValentine, J. W. 1969. Patterns of taxonomic and ecologlca 2'.684_709

. f Palaeontology 1 . .shelf benthos during Phanerozoic nne. th M rine Biosphere.Valentine, J. W. 1973, Evolutionary Paleoecology of e a

Prentice-Hall, Englewood Cliffs, N,J. eo 1'1-30.Van valen, L. 1973. A new evolutionary law. Eva/. Th ry Illustrated by theVan Valen L 1978 Review of "Patterns of Evolullon, as 17,. . . I 4'210-,

Fossil Record," edited by A. Hallam. Paleoblo ogy hradiatus- Geogn.-Waagen, W. 1869. Die Formenreihe des Ammonites su

Paleont. Beit. 2:181-256. . . ssion of paper by Dr.Waddington, C. H. 1967. Comment made dUring dl~c~ Mathematical Chal-

Eden. In P. S, Moorehead and M. M. Kaplan (e sl ('on p 14 Wistar In-lenges to the Neo-Darwinian Interpretation of Eva U I , . .

sfitute Press, Philadelphia. h raumliche Sonderung.Wagner, M, 1869. Die Enstehung der Arten aero

Benno Schwalbe, Basel. (Not seen.! ' 159'1065-70.White, M. J. O. 1968, Models of speciation. S~lence d cl~ssification: A replyWiley, E. O. 1975. Karl R. Popper, systematics, an.

stsSyst. Zool. 24:233-

to Walter Bock and other evolutionary taxonoml .

43. . h f fossil and RecentWiley, E. O. 1976, The phylogeny ~nd blOg~ograp Y ~us. Nat. Hist Misc.

gars (Actinopterygii: Lep/soste/dae). Unlv. Kans.Publ. 64:1-111.

340 References

Romer, A. S. 1962, The Vertebrate Body. W B, Saunders, Philadelphia.Romer, A S, 1966. Vertebrate Paleontology., 3rd eo. Univ. of Chicago Press,

Chicago,Rosa, D. 1931 t.'Otoaeneee. nouvelle theories de revolution et de te dis-

tribution geographique des etree vivants. Libratrte Felix Arcan. Paris.Ross, H. H. 1974. Biological Systematics. Addison-Wesley, Reading, Mass.Rowe, A. W. 1899. An analysis of the genus Micraster as determined by rigid

zonal collecting from the zone of Rhynchonella Cuvieri to that of Micra-ster cor-anguinum. Quart. J. Geol. Soc, London 55:494-547.

Salthe, S. N. 1975, Problems of macroevolution (molecular evolution, pheno-type definition and canalization) as seen from an hierarchical viewpoint.Amer. Zool. 15:295-314,

Satthe, S. N. In press, An integrated view of evolutionary theory.Schaeffer, B., M. K. Hecht, and N. Eldredge. 1972. Phylogeny and paleon-

tology. Evol, Bioi, 6:31-46.Schindewolf, O. H, 1950. Grundfragen de Peteomotoqte. Stuttgart, Schwer-

zerbartSchnell, G. O. 1970, A phenetic study of the suborder Lari (Aves): 1. Methods

and results of principal components analyses. Syst. Zool. 19:35-57.Shaw, A. B. 1964. Time in Stratigraphy. McGraw-Hili. New York.Shear, W. A. 1975. The opilinoid family Caddidae in North America, with

notes on species from other regions (Opulones, Patpatores, Cad-doidea) J. Arachnol, 2:65-88.

Simpson, G, G 1944. Tempo and Mode in Evolution. Coiumbia Univ. Press,New York.

Simpson, G. G. 1945. The principles of classification and a classification ofthe Mammalia. Bull. Amer. Mus. Nat. Hist. 85:1-350,

Simpson, G G. 1951 The species concept. Evolution 5:285-98.Simpson, G. G. 1953, The Major Features of Evolution. Columbia Umv.

Press, New York,Simpson, G. G. 1959a Mesozoic mammals and the polyphyletic origin of

mammals, Evolution 13:405-14Simpson, G, G, 1959b, The nature and origin of supraspecilic taxa, Cold

Spring Harbor Symp, Quant. BioI. 24:255-71Simpson. G, G. 1959c, Anatomy and morpnotooy: classification and evolu-

tion: 1859 and 1959. Proc. Amer. Phil. Soc, 103:286-306.Simpson. G. G, 1961 Principles of Animal Taxonomy. Columbia Univ. Press,

New YorkSimpson, G G 1963. The meaning of taxonomic statements. In S. L. Wash-

burn (eo.). Classification and Human EVOlution, pp. 1-31 Aldine. Chi-cago,

Simpson, G, G 1975, Recent advances in methods of phylogenetic infer-ence. In W. P. Luckett. and F. S. Szalay (eds.), Phylogeny of the Pri-mates, A Multidisciplinary Approach, pp, 3-19. Plenum, New York.

Simpson, G. G. 1976. The compleat paleontologist? Ann. Rev. Earth Planet.Sci. 4:1-13.

J

342 References

Wiley, E, O. 1978, The evolutionary species concept reconsidered. Syst.Zool.27:17-26.

Wiley, E. O. 1979. Ancestors, species, and craooorame-vemarke on thesymposium. In J. Cracraft and N. Eldredge (eos.), Phylogenetic Analysisand Paleontology, pp. 211-25. Columbia Univ. Press, New York.

Williams, G. C, 1966. Adaptation and Natural Selection. Princeton Univ.Press, Princeton, N,J.

Williams, H. S. 1910, The migration and shifting of Devonian faunas. PopularSci. Monthly 77:70-77.

Willis, J, C. 1940, The Course of Evolution. Cambridge Univ. Press. Cam-bridge.

wnmot. A. J. 1950. Systematic botany from Linnaeus to Darwin. In Lectureson the Development of Taxonomy, pp. 33-45. Linnaean Soc., London.

Winsor, M. P. 1976. Starfish, Jellyfish, and the Order of Life. Yale Univ.Press, New Haven, Conn.

Wolf, A. 1930. Textbook of Logic. Collier Books, New York.Wright, S. 1931. Evolution in mendelian populations. Genetics 16:97-159.Wright, S, 1932. The roles of mutation, inbreeding, crossbreeding, and se-

lection in evolution. Proc. Vllnt. Congr. Genetics, 1:35&-66.Wright, S. 1945 Tempo and mode in evolution: a critical review. (Review of

"Tempo and Mode in Evolution" by G, G. Simpson.) Ecology 26:415-19.Wright. S. 1956. Modes of selection. Amer. Nat. 90:5-24.Wright, S. 1967 Comments on the preliminary working papers of Eden and

Waddington. In P, S, Moorehead and M. M. Kaplan (eds.) MathematicalChallenges to the Neo-Darwinian Interpretation of Evolution,pp 117-120, Wistar Institute Press, Philadelphia.

Index

AJnot-Agroups, 152, 158-65, 183, 190,210--11, 238-39, 266, 320; defined,158

Adanson, M., 9adaption, 11, 13-15, 17, 125, 246, 248,

253, 255, 26Hi2, 265-66, 268, 268 f.270,27&, 278-79, 296, 302, 327; rela-tionship to diversity, 304

adaptive:divergence, 201, grid, 261, 261n,263; landscape, 251-52, 255, 271-73;peaks, 263; radiation, 273, 312, 322,329; value, 251; zones, 31411, zonesdefined, 261

Agassiz, J, L. R., 148, 149n, 255allopat-e speciation, see speciationallopatry, 295; defined, 103amino acid sequences, 48anagenesis, 13, 265ancestors, 9-10, 89, 118, 124-46, 181,

184,~12;classification of, 186-88, 190;speCified, 239; suoreeoectnc. 11, 239

anCestral-descendanthypotheses: testingOf, 128-46

an~estry,5,7,215; common, 10; tests of,36; see a/so ancestors; ancestral-

descendant hypothesesAnderson, E. J., 286Andrews, P., 81apomorphy, 33; defined, 31; see a/so

~napOmorphyAristotelean tnt C·.. In"lng and classification158--£0' ,arrested evolution, 320, 329Ashlock, P. D., 21cn 221 230aula "POmorphy,38; defined, 33; see a/so

synaPOmorphyAyala, F. J., 47

Baker, R. J., 47Batten, R. L., 288-89biogenetic law, 59---61,see a/so ontogenybiogeography: historical, 1biological species concept, see speciesbiostratigraphy, 1Bock, W, J., 9, 13, 35--37,80, 172, 193-94,

19411, 198-99, 201-3, 207-10, 222,275--76, 276n, 306, 309, 312-13

Bonde, N., 218, 226Boucot, A. J., 26Sn, 303bradytely, 320-22branching diagrams, 169, 171, 173;

polylomous, 144; see a/so cladogram;phenogram; tree

Brarusa, L., 315Bretsky, P. W., 308, 310Bretsky, S, S" 4, 114Brinkmann, R., 117nBrown, F, A., Jr.. 45Brown, W. L., 295Brundin, L., 188, 218Bryan, W. J" 255r1Buck, R. C., 169Burkhardt, R. W., Jr., 14BnBUsh, G. L., 122-26, 270, 294

Cain, A. J., 94, 116, 175Camin, J, H., 176Carruthers, R. G" 117nCarson, H, L., 11911, 126category, taxonomic, 166--68character, 30, 3On, 43---44; analysis, 23-24,

26-27,29---30,33,53; congruence, 12:derived, 31 (see a/so apomorphy;synapomorphy); displacement. 295,313; distribution, 23, 26-27, 271; extrin-

character (Continued)sic, 1, 4, 44, 49, 103, 127, 138, 142:intrinsic, 1,4,6,14,16-17,20,44,100,102, 116-18, 122, 134, 248, 293; levelsof distribution, 63; modification, 23-24,277; nested sets of. 14, 16; primitive, 31(see also plesiomorphy; syrnptesiomcr-phy); reversal, 133, 136; set-defining,174; weighting, 9, 12, 66, 73, 189, 190

character state, 30, 3iX1;see also charac-ter

character transformation, 13-15, 26-27,31-34; analysis of, 55-66: ontogeneticcriterion, 58-62: outgroup comparison,63--66; paleontological criterion, 55--58

chromosome structure, 47ctvonocune, defined, 56clade, 13, 266clad ism, see cladisticscladistics, 6, 9, 11, analysis, 16, 19-85,

179cladistic hypotheses, 19-85, 113: testing

of, 67-70creooqerests, 13ctadoqram. 10, 16-17, 20, 22-23, 25, 27,

33,40-41,43-44,67,113,127,130,137,143, 145, 178, 178-79, 184, 188, 215,238, 280, 282; construction of, 29-30;defined, 10: general characteristics of27-29: hierarchical levels of, 28; predic~tive aspects of, 40; relationship to clas-sification, 175, 191,211, 213-15; sets of,37

Clark, G. A., .Jr., 80classification, 6, 8-11, 17, 53, 147----48,

150, 165,211, 241, 253: Darwin's view149-58: defined, 6; evolutionary, 191 etseq, 201-2, 206, 208-10, 239: expres-sing set-membership, 194, 219, 224:hierarchical, 224; indented-list 229-32239: information content of, '13, 170:17iX1,171-73,190, 192, 194, 194n 201226; information retrieval, 201, inf~rma:ten storage, 194n, 201, logical structureof, 215; natural, 210, 238-39: of unre-solved trichotomy, 236, 238; phylogen-etrc. 152, 215, et seq" 224, 228 230234,238-39: predictions of, 194,202-3:239: purposes of, 193; search for pat-tern, 148; topological structure, 169,172, 214; use of branching diagrams,

344 IndexIndex 345

175, 203; use of cladograms, 218;use of incertae sedis, 238

clustering, 176, 182, 192Coleman, W., 148Corless, D. H., 8common ancestor, 35, 214; see also an-

cestors; ancestrycommunity composition, 301comparative biology, 1-3, 35competition: interspecific, 295-96convergence, 9, 12-13, 38, 62, 66-77:

defined, 8, 38Cracraft, J" 4, 10, 37, 80, 218, 226, 228Crovello, T. J., 42, 93Crowson, R. A., 48, 221-22Cuenot, L, 249Cuvier, G., 148, 149n, 159, 162---63,255

Darwin, C., 20, 35, 92, 14917, 150, 152-53,156-57, 157n, 160, 181,26&1,272

deBeer, G., 59deCandolle, A. P.. 149Delson, E., 58, 81dendrograms, 9; see also cladogramderived characters, 33; see also apomor-

phy; character; synapomorphydescendants, 9, 89, 139: see also ances-

tors: ancestral-descendant; ancestrydescent, 5, 7deterministic processes, 251, 267n, 298,

300-1, 327developmental data, 28; see also charac-

ter transformation; ontogenydifferential species survival, 274; see also

species selectiondirectional selection, 267;seea/so natural

selection; selectiondistribution, ecological correlates, 306,

309-10; morphological correlates, 307diversity, 311, predictions about morphol-

ogy, 308: relationship to adaptation,304: species, 303: taxonomic, 15,303

DNA hybridization, 47-48Dobzhansky, 1.,47,95,124,245,251-53,

255, 272, 273'l

Ehrlich, A. H., 178Ehrlich, P. R., 122, 178Eldredge, N., 4, 10, 54, 56, 82, 110-11,

117, 117n, 131, 134, 141,247,258,265n, 267, 267n, 270, 274--75, 276n,285-86,298, 293, 309, 315, 317, 318,31&1,321, 321n, 323

electrophoresis, 47-48Engelmann, G. F., 134essences, as real entities, 147essentialism, 147eurytopy. 304, 306-7, 309, 311, 313,

321-22, 328-29evolution, 2, 6, 87, 89, 241; modern

definition, 245; relationship toclassification, 14917; synthetic theory, 7,92n,118

evolutionary analysis, see phenomenolog-ical levels of

evolutionary mechanisms, 13,242-43evolutionary novelties, 10, 16, 18, 21-23,

26-28, 125; nested sets of, 10-11, 87;see also character transformation,synapomorphy

evolutionary pattem: tests of, 325evolutionary process, 4--5, 12, 242, 246evolutionary systematics, 35; see also

classification; systematicsevolutionary theory, 3, 242; characterized,

244-45: epistemological aspects, 13:ontological aspects, 13; synthesis of, 7;see also nee-Darwinism: synthetic ism

extinction, 299, 310: differential, 293: ratesof, 310, 313, 321, 323, 329: see alsospecies selection

falsification, 5, 69nFarris, J. S., 12, 17On, 176, 178, 209, 21iX1,

229-30, 239features, see charactersfitness, 251, 266, 269fossil groups, classification of, 233-36fossil record, 19,57; gaps in, t tg.see ezso

paleontologyFryer, G., 308

ecology, 301, relationship to distribution,306; relationship to morphology 302-3305 ' ,

ecosystem, 271, 303Edwards, J. L., 66, 71-72

Gaffney, E. S., 5, 37, 50, 55, 67, 74, 77, 79gaps, in fossil record, 116genealogy, 2, 12, 201, 241, see also

cladisticsgeneralization, ecological, see eurylopy

genetic drift, 298Gertsch, W. J., 83Ghiselin, M. 1.. 88, 157, 245Gingerich, P. D., 55--56, 136, 266, 291Goldschmidt, R., 249, 254, 263'J, 326Gorman, G. C., 46Gould, S. J" 59, 601, 117, 117n, 131,250,

263'l, 265n, 267, 267n, 274--75, 285,293-94, 299, 323

grades, 13, 265, 268ngradualism, see phyletic evolutionGrant, P. A., 295Grant, V., 275, 323Grasse, P. P" 249, 326Greenbaum, L F" 47Greenwood, P. H., 164,308Grene, M., 14Griffiths, G. C, D .. 222groups: nested, 28: see also monophyletic

groupsgroup selection, 273-74; see also selec-

tion; species selection

Haeckel, E.. 59Haldane, J. S, S., 274Harper, C. W., Jr. 56Harrison, G, A, 176Hecht, M, K" 54, 56, 66, 71-72, 323Hennig, W" 2, 10-11, 31-32, 37, 42n, 51,

153, 21On, 218, 220-24hierarchical arrangement, 6, 176hierarchical levels, 21, 29hierarchical pattern: nested sets, 153hierarchy: nature's, 7, 10, 20: oftaxa, 166,

168, 171, taxonomic, 230higher taxa, defined, 249-50history: evolutionary, 242: relationship

with process, 242holophyletic: defined, 210-11n, see also

monophylyhomology, 35-40, 62. 69, 71. analysis of,

37-38: comparative methodology,36---37: defined, 36: operationaldefinition, 35; phylogenetic criterion.35: problem of, 36: test of, 35

Hopwood, A. T., 149n, 159, 163Hull, D. L" 88, 147, 169, 230. 245, 275Huxley, J S, 13, 265hybridization, 100-1hypotheses, 5: phylogenetic, 10: see also

cladistics

..-~-----r346 Index

byporheticc-deductive methods. 5. 18. 19,67

lies, T, D.. 300immunological reaction, 48information storage, see classificationintergroup selection, 252, 273, 275; see

also selection; species selectionisolating mechanisms, 124, 270

Jaffe, L, 157Johnson, LAS., 17Ch, 176, 178

Kahl, M. P" 46karyotypes, 47Kent, G. C" Jr., 164Kerkut, G. A., 49key innovations, 313

Lack, D., 306Lamarck, J 8., 149n, 159, 163Larson, J, L, 148Lewontin, R C., 272lile cycle. 99lineage: splitting 01, 244; subdivision of,

181Linnaean classification, 168, 171, 173.

188; composition of. 238; logical struc-ture of, 211, see also classification

Linnaean hierarchy, 41, 89, 90, 165.169-70,182,184,191. 201, 215; logicalstructure of, 165--71, 191, topology of,226

Linnaeus. C" 87, 147, 14&1, 159,162,167,210,218

Lorenz, D. M., 308. 310Levtrup. S., 44-45, 49, 164, 249Lyell, C., 26&1

MacArthur, R. H" 303McKenna, M, C., 164, 222. 292macroevolution, 24&1, 247-48, 254, 263,

266, 268, 271-72, 27&1, 277, 279-80,294. 298, 325-26, 328; characterized,259-60, 297; defined. 15-16, 247; de-terministic theory, 301 (see also deter-ministic processes); ecological predic-tions. 31Q--11; patterns, 17, 312-13:rates of, 281, testing of hypotheses,282-83,327-28; theory and informationin, 305

Makurath, J. H., 286Margulis, L., 161-62Maslin, T. P" 54Mayr, E., 6--9, 11, 14,41-42,71,92-93,

122, 126, 152, 157, 168, 182-84,188-89,192-94,197,197 f, 202-3, 207,239, 244, 254, 275, 27&1, 293

Meise, W., 80Michener, C, D., 194n, 211microevolution, 17, 253, 266, 277-78, 280,

294, 325-26, 328; defined, 15, 243;processes, 27&1

Miles, R S., 164Minkoff, E. C" 176mode, of evolution, 249, 260, 269monophyleticgroups(taxa), 12-13. 17, 53,

165, 220, 241, 250; classification of,229; nested sets of, 14

monophyly, 12, 42, 42n, 5Q--51. 21Q--l1n;defined, 10, 39; intemesting statementsot re

morphocline: defined, 54; polarity of, 54;see also character transformation

morphology: relationship to ecology,302-3, 305

Moy-Thomas, J, A., 164mutation, 119

natural classification, 147-48, 157, 161,163,165; see also classification

natural groups, 147--50, 156, 16Q--tll, seealso monophyletic groups

natural hierarchies, 271natural selection, 13, 15,17,119, 124,243,

246,24&1.248, 25Q--51, 260, 266, 269,278-79, 294, 327; see also selection;species selection

natural system, 149n, 156nature: hierarchical structure of, 239Nelson, G, J., 10, 12, 37, 50, 121, 152, 184,

188,190, 196--200,207-9. 21Ch, 212n,218, 226

neo-Darwlnlsm, 17, 116-19, 119'1, 246,2460, 254, 266, 27&1, 278, 325, 327

neoteny, 60, 6Chnested sets of taxa, 169-70New Systematics, 7niche space, 125niche theory, 265, 302, 304, 306--7,

31Q--l1, 328; relationship to mac-roevolution, 309

nonmonophyly, 12, 39, 21iX1; see alsoconvergence; monophyly

numerical taxonomy, 6, 8-9, 35-36, 48,174, 176, 178--79, 201-2; and naturalclassification, 176; see also phenetlcs

ontogenetic data. 60ontogenetic transformation, 23-24; see

also character transformationontogeny, 58, 101; use in phylogenetic

reconstruction, 5H3operationallaxonomic units(OTUs), s.see

also numerical taxonomyorder, natural, 1, 3, 8Ormiston, A A., 315, 317, 31&'1cnrceerecuco. 275; see also natural

selection; selectionoutgroup comparison, 26-28, 63-66; see

also cladistic analysisoverall similarity, 184n; see also phe-

retics: phenogram: similarity

paadomorphoeis. 6Chpaleontology, 55-58, 10Q--2, 11Q--12,

116--20. 131-36, 138-46, 233-38, 254 etseq,

parallelism, 8, 71-74; see also con-vergence; homology

parapatric speciation, see speciationparapatry. defined, 103paraphyly, 21Q--l1n: see also monophylyParkes, K, C., 80parsimony, 67. 70, 73. 179pattern, 1-2, 4, 12,20;analysis, 3, 12, 148;

evolutionary, 325; phylogenetic, 4-5, 14Patterson, C., 14911, 164, 222, 226, 233,

237, 239phenetic classification, 17Ch; see also

classification; numerical taxonomyphenetic relationships, defined, 175phenetics. 6, 9, 11, 178: see also numeri-

cal taxonomyphenograms, 9, 175, 176--79, 190, 238;

and classification, 175-79phenomenological levels of evolution, 17.

247,254,2637,271-73,276,278.281,283,293---94,329: decoupling of levels,275, 277, 283, 327

phyletic evolution (gradualism). 55,115-17, 117n, 249, 261, 26&'1, 284,286

phyletic sequencing, see sequencing

Index 347phyletic transformation, 116, 244phylogenetic analysis, 179; see also

cladistics, analysisphylogenetic hypotheses, see cladistic

hypothesesphylogenetic pattern, see patternphylogenetic trees, see treesphylogeny, 21, reconstruction of, 6--9. 11,

19; see also cladistics, analysisPlatnick. N. L, 10, 12, 83,108--9,121,131,

134,203. aio-u»plesiomorphy, 3, 33, 63; defined, 31, see

also Character analysis, primitiveplesion concept, 233-34, 237, 239polarity: 56, 63: defined, 54; see also

morphocnne: character transformationpolyphyly, 13, 21Q--l1n; see also

monophylyPopper, K. R, 67prediction, in classification, see

classificationprimitive character, see plesiomorphy;

character; character transformationprocess, 1, 3-4, 12, 148, 242properties, see charactersProsser, C. L., 45punctuated equilibrium, 26&'1, 286

quantum evolution, 249, 252. 261, 263,263'>

random processes, 251, 267n, 298, 301,308, 328

rank (taxonomic), 167, 169. 191, 200ranking, in systematic practice, t 82,

191-92, 197, 220, 222, 224, 226, 228,230; criteria for, 193

rates of evolution, 268: genetic, 269; mor-phological. 269; taxonomic, 268

Raup, D M., 250, 288, 298--99Raven, P H, 122Ray, J., 160reductionism, 15, 17,251,27&1,283reiationship, 51, ancestral-descendant,

184; genealogical, 239; horizontal, 181,188; kinship, 188; phenetic, 175;phylogenetic, 172, 179, 218. 226, 241,similarity, 188; vertical, 181, 188; seealso ancestors

Rensch, B" 253-54, 275reproduction: differential. 269

II

I

..,

I

348 Index Index 349

revision of species, 110Rohlf, F, J., 178Aomanes, G. J., 122Romer, A. S" 164, 198-200, 203, 207--9Rosa, D" 249Rosen, D. E., 121, 222, 226, 233. 237, 239Ross, H. H.. 54, 168Rowe, A W" 117n

saltation, 17, 116, 119, 126, 248, 293Sallhe, S, N., 267n, 271,287-89,294,327scenario,S, 19SChaeffer, B., 54, 56, 323Schindewolf, O. H., 249Schnell, G. D., 176Schoeninger, M., 55-56Schopf, T. J. M" 250selection, 252, 255. 260, 262, 265, 267-68,

26&7, 270, 272, 278, 296: directional,267; intergroup, 273; see also naturalselection: species selection

sequencing: in classification, 110, 169,218,227-28, 233, 237, 239

sets of taxa, 6, 21, membership in, 191,nested pattem, 201, 203

sexual dimorphism, 98, 101Shadab, M, U" 108-9, 131Shaw, A 8., 139Shear, W. A" 112similarity, 1-2, 6, 9-10, 12, 20, 22, 26,

28-30, 35, 43, 76; analysis, 43 et seq.:behavioral, 46; biochemical, 45;coefficient, 76; developmental, 49;genetic, 9, 12, 172, 188, 190, 194n, 201;immunological, 47; in classification,147; kinds of, 7; levels of, 12: nestedpattern 01,2, 6, 11, nonhomologous (seeconvergence): overall, 9, 11, 48, 202,239; phenetic, 172, 175, 194n, 202;physiological, 44-45: weighted, 194:see also character weighting

Simpson, G, G., 8, 11, 13,35-36,41,71,93, 116, 120, 152, 157, 167, 182--83,188-89, 193, 195-97, 196-97n, 203,210, 211n, 228, 230, 239, 244, 24&1,249--50, 252, 254-55, 259-61, 263,263n, 265-66, 268,268n, 273,278,303,313, 320

sister-taxa, 39, 57, 220Sneath, P. H, A, 8, 35, 42, 176Sokal, R. R., 8, 35, 42, 93,176,178

special creation, 2-3specialization, see stenotopyspeciation, 16, 139, 244, 249, 261, 265,

270,274,279,286, 294,326; euocemc,106,121-22,124-25,127,270,294:asarandom process, 274; defined, 114, 121,parapatic, 121, 124-27; rates of, 310,323; relationship to macroevolution,275: sympatric, 121, 124-27; theory,121, 270: transformational, 118-19

species, 9, 11, 16, 42, 87, 93, 104, 113,116,118,248-49,269,271,279; andmacroevolution, 275: as individuals,275; as particles, 299; as reproductivecommunities, 90; as units of selection(see species selection); biologicalconcept, 93; concepts, 92-93, 244; de-fined, 88, 91; differential survival, 282;diversity, 306-9 (see also diver-sity); evolutionary definition, 93;fwoctnetcc-oecucuve approach, 94,97; morphological distinctness, 90: na-ture of, 90; phylogenetic criterion,105--7; reality of, 15, 88, 92n, 168,245-46, 249, 277, 325; recognition,94-112: reproductive criterion, 95, 97,100-1, 105; sibling, 105; terminal, 106

species selection, 250, 273--74, 282, 298,321-24, 327, 329

Stanley,S, M., 250, 274, 277, 284, 321,323, 327

stasigenesis, 13steady state, 321-23, 329stenotopy, 306-7, 309, 311, 313, 322-23,

329; defined, 304Stenzel, H, B" 118steplike evolution, 268nstochastic processes, 267n, 274, 285Storer, T. L, 164stratigraphy: use in tree construction,

184n; see also paleontologystratccneneuc. 136subordination: use in classification, 169,

218,221,224-28,233,237,239sympatric speciation, see speciationsympatry, 103, 295symplesiomorphy, 36-38, 53; defined, 33synapomorphy, 29, 36-38, 40-42, 53, 58,

60,63-65,71,73-74,76,87,106,143,153, 179, 186; and classification, 158;analysis of, 53-66; as tests of cladistic

hypotheses, 67-70; conflicts in pattern,70-74; defined, 33: nested patterns of,5067 90,137,187--88,211,217

syntheti~isrn, 17, 92n, 24&1, 248, 253, 272,27&1, 325

systematics, 3, 5; defined, 2; evolution-ary, 6--7, 9, 11, 36, 66, 152, 188, 192-94,197, 200-2, 208, 229--30, 239 (seealso classification, evolutionary);phylogenetic, 6, 9, 217, 220 (see alsocladistics, classification, phyloge-netic); task ot 6

Szalay, F, 5., 56, 66

Tattersall, L, 10, 82, 134, 258taxa, 12, 166--67,253; ancestral, 11, 190

(see also ancestors); composition ~t.167-68; defined, 6, 41, monophyletic,210 (see also monophyly): nested sets(hierarchy) of, 20, 22, 50, 76, 190, 211,218,238; of higher rank, 250, 271, setdefinition, 44

taxic approach toevolutionary theory, 247,296, 326

taxonomic hierarchy, 87, 173tempo, of evolution, 249; see also

evoiutionary ratestemporai distribution, 104; see also

paleontologythree taxon statement, SO, 63, 65, 143transformation, 245, 247, 260: see also

character transformationtransformational approach to evolutionary

theory, 14-15, 244, 246, 249, 263, 268,271, 296, 326

fransspecific evolution, 254, 272, 275; seealso macroevolution

trees, 33, 179, t82, 184, 184n, 188, 190;

and classification, 175, 182-85, 190,211 as scientific hypotheses, 145; con-stru~tion of, 127-28, 130; evolutionary(phylogenetic), 7,9--11. 13, 16--17, 21,30, 113--14, 127-28, 137, 140, 145, 238,280, 282; subdivision of, 182; testing of,127-28,137-38

trends, evolutionary, 267, 267n, 322-23,329; defined, 274

Trueman, A E., 117ntypology, defined, 7

usincer. R. L, 164

Valentine, J. W" 1, 271, 303Van varen. L, 2657, 268n, 299vicariance, 122, 125von aeer. K., 59, 59n, 60, 83von Wahlert, G" 13

Waagen, w., 120Waddington, C. H" 272Wagner, M" 122Wallace, A. R., 273weighting, see character weightingWhite M J. D" 145Wiley' E. 0" 10, 37, 42, 52, 67, 74, 76.93,

11-4, 133-34, 196--97n, 226, 233, 244Williams, G. C., 274Williams, H. 5" 307Willis, J. C., 249, 263'JWilmott, A, J" 149nWilson, E. 0" 295, 303Winsor, M" 149--50, 162-63Waif. A, 147Woodward, S. P., 149nWright. 5., 251-53, 255, 260, 263'J, 271,

273-74, 298Wright's rule, 275, 298

Phyl~enetic Patterns and the Evolutionary Process. Cracraft80... · Phylogenetic patterns and the...

Documents

Transcript of Phyl~enetic Patterns and the Evolutionary Process. Cracraft80... · Phylogenetic patterns and the...