Вы находитесь на странице: 1из 26

Society of Systematic Biologists

The Limits of Cladism


Author(s): David L. Hull
Source: Systematic Zoology, Vol. 28, No. 4 (Dec., 1979), pp. 416-440
Published by: Taylor & Francis, Ltd. for the Society of Systematic Biologists
Stable URL: http://www.jstor.org/stable/2412558 .
Accessed: 21/07/2014 17:05
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .
http://www.jstor.org/page/info/about/policies/terms.jsp

.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact support@jstor.org.

Taylor & Francis, Ltd. and Society of Systematic Biologists are collaborating with JSTOR to digitize, preserve
and extend access to Systematic Zoology.

http://www.jstor.org

This content downloaded from 142.51.1.212 on Mon, 21 Jul 2014 17:05:11 PM


All use subject to JSTOR Terms and Conditions

THE LIMITS OF CLADISM


DAVID

L. HULL

Abstract
Hull, David L. (Department of Philosophy, Universityof Wisconsin, Milwaukee, Wisconsin
53201) 1979. The Limits of Cladism. Syst. Zool. 28:416-440.-The goal ofcladistic systematics
is to discern sister-grouprelations (cladistic relations) by the methods of cladistic analysis
and to represent them explicitly and unambiguously in cladograms and cladistic classifications. Cladists have selected cladistic relations to represent fortwo reasons: cladistic relations
can be discerned with reasonable certaintyby the methods of cladistic analysis and they can
be represented with relative ease in cladograms and classifications. Cladists argue that features of phylogeny other than cladistic relations cannot be discerned with sufficientcertainty
to warrantattemptingto represent them in either cladograms or classificationsand could not
be represented if they could. I argue that a better alternative is to work toward improving
methods of cladistic analysis (or else to supplement them with other methods) so that such
features of phylogeny can be discerned and to devise methods of representationcapable of
representing them in both cladograms and classifications. However, cladograms and classifications cannot represent everythingabout phylogeny. It is better to represent one or two
aspects of phylogeny explicitly and unambiguously than nothing at all. [Cladism; classification; evolution; species.]

which attemptsto represent only one as


pect of phylogeny is likely to be simpler
than one which attemptsto representtwo
simultaneously. Given any particular
goal, the simplicity of the rules capable
of accomplishing this goal depends both
Charles Naudin's "simile of tree and classifica- on the inherent limitations of the mode
tion is like mine (and others), but he cannot, I
of representationand the ingenuityof the
think, have reflected much on the subject, oth- taxonomist. For example, the traditional
erwise he would see that genealogy by itself
might lend itself to
does not give classification" (Darwin, 1899, Linnaean hierarchy
features of phylogcertain
representing
2:42).
eny more readily than others,but nothing
precludes a taxonomistfromintroducing
One possible goal forbiological clas- additional representationaldevices to resificationis to "represent"phylogeny,to move these limitations.The end result is,
constructclassificationsso that certain of course, more complex rules forclassifeaturesofphylogenycan be read offun- fying. Systematists are thus forced to
equivocally. However, Darwin was quite strikea balance between how much they
correctwhen he noted thatgenealogy by wish to represent and how complicated
itself does not give classification. Differ- they are willing to make their principles
ent taxonomists may select differentas- of classification and resulting classificapects of phylogeny to represent. One tions.
In recent years a school of taxonomy
mightchoose order of branching,another
degree of divergence, another amount of has arisen whose members have seriousdiversity,and so on. Whether or not the ly taken up the challenge of attempting
rules for classifying become simpler, as to represent explicitlyand unambiguousDarwin hoped they would, depends on ly something about phylogeny. The
the features of phylogeny which the tax- school is cladism, the aspect of phylogeonomist chooses to representand the par- ny which the cladists have chosen to rep-ticularset of principles which he formu- resent is order of branching. The cladists
lates to represent them. A classification have given two reasons forchoosing this
416
"Our classificationswill come to be, as faras they
can be so made, genealogies; and will then truly
give what may be called the plan of creation. The
rules for classifying will no doubt become simpler when we have a definite object in view"
(Darwin, 1859:486).

This content downloaded from 142.51.1.212 on Mon, 21 Jul 2014 17:05:11 PM


All use subject to JSTOR Terms and Conditions

1979

LIMITS

OF CLADISM

417

particularfeatureof evolutionarydevel- sen, 1977; Wiley, 1979). Do the evoluopmentto represent.First,the Linnaean tionaryphenomenawhichcladisticanalhierarchy,as a list of indented names, ysis cannot discern actuallyexist? As I
lends itselfmorenaturallyto expressing understandevolution and evolutionary
discontinuousthan continuousphenom- theory,two of these phenomenaclearly
ena. Whetheror not evolutionis gradual occur (ancestorsgiving rise to descenor saltative,phylogenyis largelya matter dants and reticulateevolution),one is
ofsplittingand divergence.A hierachyof quite likely(multiplespeciation),and the
discretetaxanames is well-calculatedto fourthis doubtful(speciationwithoutderepresentsuccessivesplitting.
It is notas viation, although the unique derived
well-calculatedto representvaryingde- charactermaybe all butundetectable).Is
grees ofdivergence.Second, the cladists not the inabilityof cladisticanalysis to
argue thatorderof branchingcan be as- discern the preceding phenomena, ascertainedwithsufficient
certainty
to war- sumingthey take place, a limitationof
in themethodsofcladisticanalysis?Instead
rantthe inclusionof such information
a classificationwhile other aspects of of decliningto deal withsuch phenomphylogenycannotbe. The methodwhich ena because the methods of cladistic
cladistshave devisedto discoverorderof analysiscannotdiscernthem,perhaps a
branchingis cladisticanalysis.
betterstrategy
mightbe to attemptto imCladists arguethat,by and large,all a proveupon themethodsofcladisticanalhas togo on is thetraitsofthe ysis or to supplementthem with other
systematist
specimens before him, whether these methods.The crucial question with respecimens representextant or extinct spect to cladism,as I see it, is what is
forms.By studyinghis specimens,he can cladisticanalysis?Whatare its limits?Is
discovernestedsetsofcharacters.
Unique cladisticanalysisone wayofascertaining
derived charactersdistinguisha mono- phylogeny,the only way of ascertaining
phyleticgroupfromits nearestrelatives; phylogeny,the only way of obtaining
sharedderivedcharacterscombinethese knowledgeof any phenomena?
monophyletic
groupsintomoreinclusive
One factabout cladismwhich complimonophyletic
groups.In thiswaythesys- cates attemptsto answer the preceding
tematistcan ascertainnested sets of sis- questions is thatcladism,like all scienter-groups.This method,so cladists ar- tificmovements,is neither immutable
gue, cannotbe used to discerna variety normonolithic.At any one time,cladists
ofotherrelations-speciationwithoutthe can be founddisagreeingwitheach other
appearance of at least one unique de- about particularprinciplesand conclurived character, reticulate evolution, sions. In addition,if one tracesthe demultiple speciations,and the ancestor- velopmentofcladismthroughtime,from
descendantrelation-nor can any other its inception in the works of Hennig
method.
(1950), throughits introductionto EnSeveralquestionsariseat thisjuncture. glish-speaking systematists (Hennig,
If the Linnaean hierarchycannotrepre- 1965, 1966; Brundin, 1966; Nelson,
sent certainfeaturesof evolutionaryde- 1971a, 1971b, 1972a, 1972b, 1973a,
velopmentverywell, is not thata limi- 1973b, 1973c, 1974) to the latest publitationoftheLinnaeanhierarchy?
Instead cations of present-daycladists, signifiofrefusingto representwhata particular cantchangescan be discerned.The most
in significant
has difficulty
systemofrepresentation
change whichhas takenplace
representing,
a betteralternativemight in cladism is a transitionfromspecies
be to abandon or improvethatsystemof being primaryto charactersbeing priThis is one avenue which mary.For Hennig(1966, 1975) and Brainrepresentation.
the cladiststhemselveshave taken(Grif- din (1972a), the basic units of cladistic
fiths,1976; Hennig,1966; L0vtrup,1973; analysisare species, characterizedby at
Nelson, 1972a, 1973c; Pattersonand Ro- least one unique derivedcharacter.Now,

This content downloaded from 142.51.1.212 on Mon, 21 Jul 2014 17:05:11 PM


All use subject to JSTOR Terms and Conditions

418

SYSTEMATIC

ZOOLOGY

VOL.

28

for a growingnumber of cladists, any theyhave not developed by progressing


monophyleticgroupwhich can be char- up the periodictable, fromhydrogento
acterizedby appropriatetraitscan func- helium,lithium,beryllium,etc.
From the beginning,Gareth Nelson
tion in cladistic analysis,regardlessof
whetherthatgroupis moreinclusive or (1973c) seems to have been developing
less inclusive than traditionally-defined
two notionsof cladisticanalysissimultaspecies. Cladistic classificationsdo not neously,one limitedto historicallyderepresenttheorderofbranchingofsister- veloped patterns(cladism with a small
groups,but the order of emergence of 'c'), the othera moregeneralnotionapunique derived characters,whetheror plicable to all patterns(Cladism with a
not the developmentof these characters big 'C'). His methodof componentanalhappens to coincide with speciation ysis is a general calculus fordiscerning
events. It is not the emergenceof new and representing
patternsofall sorts.For
species which is primarybut the emer- example,in a recentdiscussion,Nelson
gence of new traits(Tattersalland El- (1979:28) statesthata "cladogramis an
dredge, 1977:207; Rosen, 1978:176). In atemporalconcept ... a synapomorphy
general,cladists seem to be movingto- scheme." It remainsto be seen whether
ward the positionthatthe particularsof all ofthe termsof cladisticanalysisfrom
evolutionarydevelopmentare not rele- "sister group " to "synapomorphy" can
vantto cladism.It does notmatterwheth- be definedso thatnone of themneceser speciationis sympatricor allopatric, sarilypresupposesa temporaldimension.
saltative or gradual, Darwinian or La- AlthoughI thinkthatthe principlesof
marckian,
just so long as it occursand is cladisticanalysiscan be extendedto any
1974; system which genuinely evolves (i.e.,
predominantly
divergent(Cracraft,
Bonde, 1975; Rosen,1978; and forthcom-changesthroughtimeby means of moding worksby Nelson and Platnick,El- ificationthroughdescent), I will limit
and others).
myselfin this paper to phylogenyand
dredgeand Cracraft,
The writingsof most cladists have biological evolution.I also will not adbeen concernedwith developingmeth- dress myselfto Nelson's more general
to discuss the princiods ofascertainingthe cladisticrelations notion.Attempting
betweenbiologicaltaxa,but Platnickand ples ofcladisticanalysisas theyapply to
Cameron (1977) have shown thatthese phylogenyis a sufficiently
difficult
task
same techniquescan be used to discern withoutattemptingto presentinterprethe cladisticrelationsexhibitedby lan- tationsoftheseprincipleswhichare also
guagesand texts(see also Kruskal,Dyen, consistentwith Nelson's more general
and Black, 1971; Haigh, 1971; and Nita, program.
1971).In generalcladisticanalysiscan be
PRINCIPLES AND
METHODOLOGICAL
used to discover the cladistic relations
OBJECTIONS
between any entitieswhich change by
means of modificationthroughdescent
In spite of certainobservationswhich
(Platnick,1979).As generalas thisnotion cladistshave made periodicallyaboutthe
of cladistic analysis is, it still retainsa evolutionaryprocess, the principles of
necessarytemporaldimension.Transfor- cladisticanalysis do not concernevolumation series must be established for tion but the goals of biological classifinotjust an abstractatemporal cationand the means by whichtheycan
characters,
series like the cardinal be realized.These principlesare primartransformation
numbersor the periodictable,but a se- ily methodological.As I see it, the goal
in time.For of cladism is to representcladisticrelariesofactualtransformations
example, the vast majorityof matterin tions in both cladogramsand classificathe universehappens to be hydrogen.In tionsas explicitlyand unambiguouslyas
isolated pockets of the universe,more possible. Thus, cladistic methodology
complexmolecules have developed, but has twoparts:rulesfordiscerningcladis-

This content downloaded from 142.51.1.212 on Mon, 21 Jul 2014 17:05:11 PM


All use subject to JSTOR Terms and Conditions

1979

LIMITS

OF CLADISM

419

tic relationsand rules forrepresenting truthand cogencyof argument,


but they
them in cladogramsand classifications. are a good deal moreinterestedin truth
For example, Nelson raises two objec- thanin cogencyof argument.If a line of
tionsto theancestor-descendant
relation: reasoningwhich led to a particularconfirst,"ancestralspecies cannotbe iden- clusion turnsout to be somewhatless
tifiedas such in the fossilrecord"(Nel- thanperfect,it reallydoes notmatterall
son, 1973b:311), and second, "they are thatmuch,just so long as the conclusion
also inexpressible in classifications" depicts realitywithgreaterfidelitythan
(Nelson, 1972a:227).
previousattempts.
One sourceofthe confusionwhichhas
Scientistshave limitedpatience when
accompaniedthe controversy
over clad- it comes to discussingarguments.That
ism is the interpretation
of theirmeth- patience is even morelimitedwhen the
odologicalprinciplesas empiricalbeliefs arguments
are methodological.
As strange
about evolution.The claim thatdichoto- as it may sound comingfroma philosomous speciationnever occurs is an em- pher,I am highlyscepticalof methodopirical claim to be tested by empirical logical principlesand prescriptions,
esmeans.The claimthatwe can neverdis- pecially when theyare presentedin the
tinguishbetween dichotomousspecia- midstofscientificdisputes.They tendto
tion events and more complex sorts of be suspiciouslyself-serving,
designedto
speciation is a methodological claim put one's opponents at a disadvantage
about the limitsof our methodsof sci- while shoringup one's own position.Inentificinvestigation.Finally, the claim variably the advocates of a particular
thatmultiplespeciation,if it occursand methodologycan know what theyneed
if it can be discerned,cannotbe repre- to knowwhile theiropponentscan never
sented unambiguously in cladograms hope to know what theyneed to know.
is a commentabout For example,the pheneticistsclaim that
and/orclassifications
the limitationsof these modes of repre- we can analyzetraitsintounitcharacters,
sentation.Once the principlesofcladism but we can neverhope to establishgenare recognizedforwhat theyare, meth- uine, evolutionarytransformation
series
odologicalprinciples,thelogic ofthecla- amongcharacters.None too surprisingly,
disticpositionon a varietyof issues be- thepheneticistsneed to knowtheformer
comes muchclearer.
but not the latter.Only the cladistsand
Scientistsare interestedin truth,not evolutionistsneed to knowthe unknowAbsolute Truth,but truthnevertheless. able. Similarly,thecladistsclaimthatwe
Althoughphilosophershave yet to pro- can establishtransformation
series with
duce a totallyunproblematicanalysisof sufficientcertaintyto warrantthe role
science as the pursuitof truth(see, for which theyplay in cladisticsystematics
example,Laudan, 1977),I thinkthatsci- but that ancestor-descendantrelations
entistsare correctin the focus of their are unknowable.As luck would have it,
attention.Scientistsalso present argu- cladistsneed to knowthe former
but not
mentsto buttresstheirempiricalclaims the latter.Only the evolutionistsneed to
and tryto make theirargumentsas good knowthe unknowable.
as possible. Unfortunately,
scientistsare
As cynical as the preceding remarks
rarelyable to presenttheirargumentsin maysound,I thinktheyhave somevalidthe unproblematicformsthus far ana- ity.It is no accidentthatall fourof the
lyzed successfully by mathematicians phenomenawhichthecladistsclaimcanand logicians.The reasoningwhichgoes not be known,cannot be discernedby
into the formulation
and testingof sci- thetraditional
meansofcladisticanalysis
entifictheoriesis a good deal morecom- and cause problemsforthe explicitand
plex and informalthan any explicitra- unambiguous representationof sistertionalreconstruction
has yetbeen able to group relationsin both cladogramsand
capture.Scientistsare interestedin both classifications.
It is also no accidentthat

This content downloaded from 142.51.1.212 on Mon, 21 Jul 2014 17:05:11 PM


All use subject to JSTOR Terms and Conditions

420

SYSTEMATIC

one of the phenomena which cladists


claim cannot be known (the ancestor-descendant relation) plays a central role in
the research program of their chief rivals-the evolutionists. I do not mean to
imply that the preceding remarks apply
uniquely to the cladists. They apply to
scientists in general. If I were investigating the pheneticists or the evolutionists,
comparable observations would apply as
readily to them. These are the sorts of
games which scientists (not to mention
philosophers) play with each other.
In general, I think it is very bad strategy forproponents of a particular scientific research program to stake their future on epistemological considerations,
especially on our inabilityto know something. Phenomena which scientists in
one age claim can never be known often
become common knowledge at a later
date. The history of science is littered
with the bodies of scientists who staked
the success of their movements on what
we can never know. I agree with Einstein
(1949:684), who, in response to philosophical criticismsof his work, stated:
Science without epistemology is-insofar as it is
thinkable at all-primitive and muddled. However, no sooner has the epistemologist, who is
seeking a clear system, foughthis way through
to such a system,than he is inclined to interpret
the thought-contentof science in the sense of his
system and to reject whatever does not fit into
his system.The scientist,however, cannot afford
to carry his strivingfor epistemological systematic that far.

I hardly want to argue against methodological rigorin science, but I also do


not want to see scientificprogress sacrificed to it. Invariably methodological rigor is a retrospective exercise, carried on
long afterall the Nobel prizes have been
won. From the point of view of current
methodologies, scientists will, as Einstein (1949:684) noted, appear to the
"systematic epistemologist as a type of
unscrupulous opportunist." Instead of
cladists insisting that certain aspects of
phylogeny can never be known and
could not be represented cladistically if

ZOOLOGY

VOL.

28

they could be known,a wiser strategy


would be to attemptto devise methodsof
analysiscapable of discerningthese features of evolutionarydevelopmentand
to
techniquessufficient
representational
representthem.In thispaper,I have attemptedto producean internalcriticism
of cladism; thatis, I have accepted the
goals of cladismand have set myselfthe
task of deciding which methodological
principlesare centralto the undertaking,
which peripheral,and which could be
modifiedor abandoned withoutloss and
possiblywithsome gain.
TREES,
CLADOGRAMS, PHYLOGENETIC
SCENARIOS
AND EVOLUTIONARY

A curious featureof scientificdevelopment is the frequencywith which a


new movementis named by its opponents.Social Darwinistsno morewanted
to be called Social Darwiniststhancladists have wanted to be called cladists.
Twentyyearsago, JulianHuxley (1958)
coined the terms"clade" and "grade" to
distinguishbetweengroupsoforganisms
witha commongeneticoriginand groups
levels of ordistinguishedby different
ganization. Later Mayr (1965:81) and
Camin and Sokal (1965:312) introduced
the term"cladogram"to referto a diagram "depicting the branchingof the
treewithoutrespectto rates
phylogenetic
of divergence"(Mayr,1978:85). Finally,
Mayr(1969:70) inventedthe term"cladism"as a substitutefor"phylogeneticsysby Hennig
tematics,"thenamepreferred
and his followers.Graduallythe cladists
themselveshave come to use the termto
referto themselves,albeit grudgingly.
Whetherthe Hennigianschool of systematicsis called "phylogenetic"or "clabutthepredistic"is notveryimportant,
cise natureofcladogramsas theyfunction
in cladisticanalysisis. In fact,uncertaintyoverwhatitis thatcladogramsare supposed to depict and how theyare supposed to depict it has been the chief
source of confusionin the controversy
overcladism.Like all terms,themeaning
of "cladogram"has changedthroughthe

This content downloaded from 142.51.1.212 on Mon, 21 Jul 2014 17:05:11 PM


All use subject to JSTOR Terms and Conditions

1979

LIMITS

OF CLADISM

421

years.It no longermeansto cladistswhat numbers of organisms, and so on. HowMayr,Camin and Sokal proposed. The ever, as a diagram in two-dimensional
meanings of "cladism" and "cladistic space, a tree can include only so much
analysis" have changed accordingly.In informationbefore a point of diminishing
an attemptto reduce the confusionover returnssets in. Eventually attemptsto inthe meaning of "cladogram," cladists clude additional sorts of informationrehave introducedthe distinction
between sult in the loss of information.If trees are
cladograms,phylogenetictrees,and evo- supposed to be systems of information
lutionaryscenarios.'Once again,the par- storage and retrieval, the information
ticulartermsused to markthese distinc- must be retrievable.
tions are not important;the distinctions Scenarios, as cladists use the term,are
themselvesare. It is also true that the not diagrams. They are historical narracladists have devised these distinctions tives which attemptto describe not only
for their own purposes. Others might which groups gave rise to which (the sort
wantto drawotherdistinctions
orto draw of information contained in trees) but
these distinctions differently.In any also the ecological changes and evolucase, all threeappellationsreferto rep- tionary forces which actually produced
resentations,
notto thephenomenabeing the adaptations which characterize the
represented."Cladogram"and "tree" re- organisms discussed. Romer's (1955:57)
ferto two sortsofdiagrams,while "scen- storyof the role which the drying up of
ario" refers to a historical narrative ponds and streams played in the transicouched in ordinarybiologicallanguage. tion from the crossopterygians to early
Phylogenetictreesare designedto de- amphibians is by now a classic example
pict phylogeneticdevelopment,indicat- of an evolutionary scenario. Because
ing whichtaxaare extinct,whichextant, scenarios are presented in ordinary lanwhich gave rise to which,degrees of di- guage-supplemented
with a host of
vergence,and so on. As diagramstheydo technical biological terms-the phenomnot include discussionsof the methods ena which scenarios can describe are limused to constructthem,the naturalpro- ited only by the limitations of language.
cesses which produced the phenomena
Hennig's early cladograms (Hennig,
theydepict,and a varietyofotherconsid- 1950:103; 1966:59, 71) give every aperationsof equal importance.A tree is a pearance of being highly stylized trees.
diagram,not an entiretaxonomicmono- The circles arrayed along the top of the
graph.All sortsofconventionshave been diagram apparently represent extant
devised to representphylogenyin a dia- species, while those at the nodes repregrammaticform;forexample,lines usu- sent extinct,common ancestors (see Fig.
ally representlineages, forksrepresent la). At one stage in the development of
speciationevents,the slantand lengthof the cladogram, Nelson (1972b, 1973b) ara line reflectthe rapiditywithwhich di- gued thatthe circles at the nodes did not
vergencetookplace, and the termination represent real common ancestors but
ofa line indicatesextinction.
Othertech- "hypothetical ancestors." The term is
niques of representation
have also been problematic. All taxa thatever existed are
used on occasion; forexample,dots in- equally real. Only our ability to discern
dicatingquestionableconnections,lines them varies. We can usually discern exof varyingthicknessesreflecting
relative tant species with greater certaintythan
1 The distinctionbetween cladograms, trees, and
scenarios discussed in Tattersalland Eldredge (1977)
was taken froma manuscriptby Gareth Nelson. See
Mayr (1978) for early definitionsof such terms as
i'cladogram.. "phenogram," and "phylogram."

extinct forms,but the postulation of any


taxon, whether extantor extinct,involves
highly complex inferences and requires
evidence which is frequently quite difficult to obtain. (Recall the objections
which pheneticists raised to the biologi-

This content downloaded from 142.51.1.212 on Mon, 21 Jul 2014 17:05:11 PM


All use subject to JSTOR Terms and Conditions

422

SYSTEMATIC

VOL. 28

ZOOLOGY

ciationevent(thesplittingofone species
intotwo),or the emergenceof an evolutionarynovelty(a unique derived character), or both. If the emergence of a
B
C
A
unique derived characteroccurs always
evolution of the cladogram. (a) A and only at speciation,thenthe two anFIG. 1.-The
cladogram in which the darkened circles at the tercoincide.Ifnot,not.Similar
mini represent extant species, while those at the swersalways
hold forthe term"cladistic
nodes representcommon ancestors. (b) A cladogram observations
in which the darkened circles at the termini rep- relation."It can referto orderofsplitting
resent species, both extant and extinct, while the of taxa,or to orderof appearanceof evocircles at the nodes represent "hypothetical com- lutionary
novelties,orboth.The decision
mon ancestors," i.e., morphotypes.(c) A cladogram
in which the darkened circles at the termini rep- hingeson how "operational"one wishes
resent species, both extant and extinct,while the to make cladisticanalysis.Since specianodes represent speciation events and/orthe emer- tion eventsare inferred
by means of the
gence of unique derived characters.
appearance of evolutionarynovelties,
one inferential
stepcan be eliminatedby
talkingonlyaboutthe emergenceofevocal species concept even when it is ap- lutionarynovelties and not speciation
plied to extantforms.)If extinctformsare events.

"hypothetical" because of their inferential basis, then so are extant forms.The


difference between extant and extinct
species is not between observation and
inference, but between inferences of
varying degrees of certainty.
Anotherinterpretationof "hypothetical
ancestor" is thatthe circles which appear
at the nodes of cladograms are not supposed to be taxa at all but "morphotypes," rational constructs characterized
by the definingtraitsof the taxa listed at
the terminiof the cladogram and no others. As Platnick (1977c:356) observes, the
"nodes of cladograms represent only inferred species (or, more accurately, only
minimum sets of synapomorphic characters)." In this sense, hypotheticalancestors are "hypothetical," but they are not
"ancestors." They are not even taxa (see
Fig. lb). In present-day cladograms, all
taxa, whether extinct or extant, appear
along the top of the cladogram. Cladograms would be much less misleading if
they included nothing at the nodes (see
Fig. ic).
If the nodes in a cladogram do not represent real taxa (ancestral or otherwise),
what do the lines and forksin cladograms
represent?Two answers have been given
to this question, which may or may not
'be reducible to the same answer. A fork
in a cladogram can representeithera spe-

CERTAINTY

AND LEGITIMATE

DOUBT

One reasonforthecladists'introducing
the distinction between cladograms,
trees, and scenarios was to clarifythe
technicalsense in whichtheywere using
the term"cladogram."Too manypeople
were misinterpreting
cladogramsas bizarre,highlystylizedtrees.A second reason was to establishthe logical and epistemologicalpriorityof cladogramsover
treesand oftreesoverscenarios.As cladists definethese terms,an inclusionrelationexistsbetweenthem.Scenariosincontained in
clide all the information
trees,and more besides. Trees include
all theinformation
containedcladograms,
and morebesides. Thus, any knowledge
requiredfora cladogramis requiredfor
a tree,and anyknowledgerequiredfora
tree is required fora scenario,but not
vice versa. Thus, cladogramsare more
certainthantrees,and treesmorecertain
than scenarios.As harmlessas the distinctionbetween cladograms,trees,and
scenariosmayseem on the surface,it results in cladisticanalysisbeing in some
sense basic to all evolutionarystudies.
The relationsset out above do obtain
between cladograms,trees,and scenarios-as the cladists definethese terms.
That is one reason why non-cladists
But even
mightpreferotherdefinitions.

This content downloaded from 142.51.1.212 on Mon, 21 Jul 2014 17:05:11 PM


All use subject to JSTOR Terms and Conditions

1979

LIMITS

OF CLADISM

423

if one were to accept the cladists' defi- quisitionis a processofboth"reciprocal


nitions,certainconclusionswhich clad- illumination"(Hennig,1950; Ross, 1974)
istshave claimedfollowfromthepreced- and "reciprocalblundering."Gettingone
ingline ofreasoningdo not.For example, elementrighthelps improveour underin claimingthat"treesshould always be standingoftheotherelements,buterrors
based on cladogramsand thatscenarios feed throughthe systemjust as readily.
should followfromtrees,"Tattersalland Althoughcladistsmightwell admitthat
Eldredge (1977:205) are confusinglogi- knowledge acquisition in general is a
cal and epistemologicalorderwithtem- process of reciprocalillumination,they
poral order.The conclusionto an argu- also tend to be so scepticalof trees and
mentfollowslogicallyfromits premises. scenariosthattheysee all the illuminaThat does not mean that people must tion going in a single direction-from
thinkofthe premisesfirstand onlythen cladograms to trees and scenarios. At
thinkof the conclusion.Sense data are timescladistsseem to arguethatnotonly
priorto all knowledge, should evolutionarystudies begin with
epistemologically
but that does not mean that scientists cladograms but also they should end
should begin everyscientificstudywith there as well. Cladistic relations are
an extensiveinvestigationof theirown knowable;everything
else aboutphylogsense data. If scientistshad actuallypro- eny is unknowable.
ceeded in thisfashion,we would stillbe
For example,Platnick(1977d:439) arawaiting Ptolemaic astronomy,not to gues thatthe onlyway thatphylogenetic
theory.The inferences treescan be testedis by the same means
mentionrelativity
which take place in the actual course of used to test cladograms,"by evaluation
are veryintricate in the lightof newly discoveredcharacscientificinvestigations
and frequently
"feedbackupon each oth- ters."Anythreetaxacan be orderedinto
er,"as Tattersalland Eldredge(1977:205) 22 different
trees.Some ofthesetreescan
thereis considerable be rejected by discoveringappropriate
note. In retrospect,
pointto unravelingthese inferencesand characters,i.e., the appropriateautaposettingthemout in some logicallycoher- morphies and synapomorphies.Other
ent fashion,but the insistencethatsci- trees can be rejected only by claiming
mustalways begin at the begin- thatno autapomorphiesexistforthe relen-tists
ningand proceed step by step according evanttaxa.Since the mostthata systemto some sortoflogical or epistemological atistcan legitimately
claim is thathe has
order would stop science dead in its yetto detectsuch autapomorphies,
he is
tracks.2
notjustifiedin rejectingthesetrees.PlatCladists also make much of the differ-nick (1977d:440) concludes that"phyloences in certaintybetween cladograms, genetictreesare nottestableby character
trees,and scenarios.In actual practice, distributions
and thusthatscientific
physcientistswanderfromone level ofanal- logenyreconstruction
is not possible at
ysis to anotherin the course of theirin- the level ofphylogenetictreesand must
vestigations.As a resultknowledge ac- be restricted
to the level ofcladograms."
Butthesame sortofargument
can be presented againstthe cladists.They do not
2The pheneticists presented similar arguments need to know
ancestor-descendant
relafor the epistemological priorityof phenetic classibut
tions,
do
need
they
to
establish
transfications. According to the pheneticists, the only
seriesand to determinethepoplace at which systematistscan legitimatelybegin formation
is phenetic charactersand the constructionof pure- larityof these series. They do need to
ly phenetic classifications. Once such a phenetic individuatecharacters
and decide which
classification is erected, then a systematistcould
are
etc.
autapomorphic,
synapomorphic,
put various "interpretations"on his classifications
and possibly produce one or more "special pur- For instance, synapomorphiescan be
pose" classifications; see e.g., Sneath and Sokal used to test cladogramsif theyare syn(1973).
apomorphies,but no one has yetto sug-

This content downloaded from 142.51.1.212 on Mon, 21 Jul 2014 17:05:11 PM


All use subject to JSTOR Terms and Conditions

424

SYSTEMATIC

ZOOLOGY

VOL. 28

gest any infallible means for deciding be known(or at least cannotbe known
whetheror not a traitis actuallya syn- withsufficient
to see whythey
certainty)
apomorphy.The same can be said for are so difficult
to discern.Is it thatthey
such relationsas the polarityof transfor-cannotbe known,or thattheycannotbe
mationseries. In fact,Tattersalland El- knownby means of cladisticanalysis?Is
dredge(1977:206) state,"In practiceit is thereno wayto acquireknowledgeofthe
hard, even impossible, to marshal a world other than by cladistic analysis?
strong,logical argumentfora given po- The pointofthissectionis,however,that
larity for many characters in a given the differencein our abilityto ascertain
group." The sortof argumentfromneg- cladisticrelationsand otherevolutionary
ative evidence which the cladists use phenomena is one of degree, not kind.
againstothersat the level of taxacan be Differencesir1degree of certaintyare
used againstthemat the level of traits. neitherveryneat noraesthetically.
pleasOf course,theiropponentsalso mustin- ing, but they are the most thatcladists
dividuateand categorizetraits.Thus, the can justifiablyclaim. Whatthey lose in
cladists' positionis refutedat only one elegance and simplicity,they gain in
level of analysis,while the position of plausibility.
their opponents is refutedat two. Put
Certaincladists,in morerecentpublimore directly,such argumentsrefuteno cations,seem to be movingin precisely
positionswhatsoever.
this direction.For example, Eldredge
The cladistshave weakened the force (1979:169) statesthat,contrary
to his earoftheirarguments
by presentingthemin lier opinions:
a needlessly dichotomousfashion.The
I no longer oppose the constructionof phylogeonlythingthattheyneed to knowto protrees outright,or for that matter, scenarnetic
duce cladogramsare cladistic relations, ios (which
are, afterall, the most fun),but mereand theyare knowable.Otherswho wish
ly point out that, in moving through the more
to produce trees need to know much
complex levels, we inevitablybecome furtherremoved fromthe original data base in adding asmore; e.g., ancestor-descendantrelasumptions and ad hoc (and largely untestable)
tions, the existence of multiplespeciahypotheses. As long as we understand precisely
tion and reticulateevolution,etc. These
what we are doing at each step in the analysis,
phenomenaare totallyunknowable.The
which includes having an adequate grasp of the
cladists need not present such an exprobability that we are wrong and of what the
assumptions are what we have added along the
treme(and suspiciouslyself-serving)
pothere no longer seems to me any reason for
sition. Estimationsof cladisticrelations way,
anyone to tell anyone else what not to do.
are inferencesand as such carrywith
themthe possibilityoferror.However,it
Althoughthoseworkersengagedin the
is certainlytrue that everythingwhich productionof trees and scenariosmight
the cladistsneed to knowthe evolution- differwith Eldredge on the situation
ists also need to know,while the evolu- beingas extremeas he makesitoutto be,
tionists need to know more besides. theytoo are aware ofthe difficulties.
For
These additionalphenomenaalso carry example,Lucchese (1978:716) concludes
withthema certaindegree ofuncertain- a paper on the evolutionof sex chromoty. Hence, evolutionaryclassifications somes withthe followingremarks:
are bound to be moreuncertainthancladistic classifications.The question re... foran individual to professan insightinto the
mainswhetherall evolutionary
phenom- biological changes that have occurred
through
ena otherthan cladisticrelationsare so
time remains an act of faithof considerable magnitude. [Yet, within]the limitationsjust set forth,
uncertainthatno attemptshouldbe made
the purpose of this article has been to spin a relto include information
about them in a
atively reasonable evolutionary tale in the hope
In the succeedingsections
classification.
of expanding the currentperception of a highly
of this paper, I will investigatevarious
significantexample of genetic regulation in euphenomenawhich cladistsargue cannot
karyotes.

This content downloaded from 142.51.1.212 on Mon, 21 Jul 2014 17:05:11 PM


All use subject to JSTOR Terms and Conditions

1979
THE PRINCIPLE

LIMITS

OF CLADISM

OF DICHOTOMY

Few principles attributed to cladistic


taxonomists have caused more consternation and confusion than the claim that
all cladograms and classificationsmust be
strictlydichotomous. When cladism was
firstintroduced to English-speaking systematists,cladists and anti-cladists alike
agreed that the principle of dichotomy
was "essential to the philosophy of Hennig and Brundin " (Nelson, 1971a:373;
see also Darlington, 1970:3; Mayr,
1974: 100; Bonde, 1975:302; Platnick,
1977d:438). Although Cracraft(1974:79)
agrees that "phylogenetic classification
sensu Hennig and Brundin would appear
to require the assumption of dichotomous
branching," cladistic classification in a
more general sense "does not necessitate
dichotomous branching, and the exact
patternofbranchingis determinedby the
manner in which shared derived character-states cluster taxonomic units." The
question remains why the principle of
dichotomyhas seemed so central to most
cladists and continues to seem so to
some.
Although no cladist has ever maintained that the principle of dichotomy is
an empirical claim about the speciation
process, few cladists have been able to
resist the temptation to justifyit by reference to empirical considerations. For
example, Hennig (1966:210, 211) begins
by statingthatdichotomyis "primarilyno
more than a methodological principle"
and then goes on to add, "A priori it is
very improbable that a stem species actually disintegratesinto several daughter
species at once." Cracraft (1974:79) interrupts the discussion quoted above
with the remarkthat the principle of dichotomy "can be justified on theoretical
grounds not associated with classification." Bonde (1975:302) agrees. Although
dichotomyis a methodological principle,
in nature "speciations are probably nearly always dichotomous."
Whether multiple speciation seems
possible or impossible depends on the
unit of time one selects to define "simul-

425

taneous." As Hennig (1966:211) notes,


the issue of dichotomous speciation can
be trivialized by defining "simultaneous" too narrowly. If it is defined in
termsof split seconds, then multiple speciation is impossible. Conversely, if "simultaneous" is defined too broadly,
everythingcan be made to occur "at the
same time." The question of dichotomous speciation can be made significant
only by the specificationof a unit of time
which makes sense in the context of the
evolutionary process. In the absence of
such a unit, the notion of simultaneity
can be expanded or contractedat will and
the issue decided by fiat (see Cracraft,
1974:74).
Whether multiple speciation as an empirical phenomenon seems plausible or
implausible depends on the model (or
models) of speciation which one holds. If
speciation is viewed as the gradual splitting and divergence of large segments of
a species, as Hennig (1966:210) maintains, then multiple speciation looks less
likely than if it is viewed as a process in
which small, peripheral populations become isolated and develop into separate
species, as Wiley (1978:22) supposes.
But, as the cladists have reminded us
often enough, dichotomy is a methodological principle. If speciation were always dichotomous, this fact about evor
lution would certainly lend support to
dichotomyas a methodological principle,
but good reasons might exist forproducing exclusively dichotomous cladograms
and classifications even if speciation is
not always dichotomous.
Cladists have presented two sorts of
justification for their preference for the
principle of dichotomy: our inability to
distinguish dichotomous speciation from
more complex sortsof speciation and our
inability to represent unequivocally
more complex sorts of speciation in
cladograms and classifications. If one
limits oneself just to traditionally-defined
traits (e.g., presence of various sorts of
appendages, dentitions, etc.) and if one
assumes that these traits have been appropriately individuated and identified,

This content downloaded from 142.51.1.212 on Mon, 21 Jul 2014 17:05:11 PM


All use subject to JSTOR Terms and Conditions

426

SYSTEMATIC

ZOOLOGY

VOL.

28

thenthe argumentwhichthecladistsput groups under study are legitimate. Many


If a trichotomiesare very likely to be resolvforthis reasonablystraightforward.
cladiststartswithanytwospecies chosen able into dichotomies afterfurtherstudy,
at randomand comparesa thirdspecies many of the groups treated as single may
to them,two ofthese species will be cla- have to be split into two or more groups,
disticallymore closely related to each and vice versa. However, as the groups
other than either is to the remaining are studied more exhaustively, the legitspecies. If the cladist continuesto add imacy of continued doubt decreases. If
species to his studyone at a time,he will afterexhaustive study,three species conproduce a consistently dichotomous tinue to appear to be related trichotocladogramuntilone oftwothingsoccurs: mously, the claim that they mightactualeither he happens upon a genuine in- ly be related by a pair of dichotomies
stance of multiplespeciationor else he begins to ring rather hollow. Descartes
is unable to resolve this complex situa- has shown where that sort of mindless
tionintotheappropriatesequence ofsuc- scepticism leads. If a systematistcan becessive dichotomies.Because dichotomy come reasonably certain that he has coris the simplesthypothesis,it should al- rectly discerned a genuine dichotomy, I
evidence permitting. see no reason why he cannot also attain
ways be preferred,
But as the cladistssee it, given the sort that same level of certaintyin the idenand tification of trichotomies-even using
of data available to the systematist
the resolvingpower of his methodsof just the traditional methods of cladistic
analysis,he will alwaysend up wherehe analysis. But systematistsare not limited
begins-with doubt or with dichotomy. just to the study of ordinary taxonomic
Never is a systematist
justifiedin con- traits.There is no reason fornot using the
cluding thata speciationevent is genu- methods of cladistic analysis on other
inely trichotomous
because he is never sorts of evidence, e.g., the data of historjustifiedin dismissingthepossibilitythat ical biogeography. As Platnick and Nelan apparenttrichotomy
is reallya pair of son (1978: 10) remark:
unresolveddichotomies.
Trichotomous cladograms are of no signifiIf the precedingis a faircharacterizacance as tests (or as initial hypotheses) unless the
it has two
tion of the cladists'argument,
cladograms for all available test groups are triweaknesses.First,the cladistsare selecchotomous, in which case we may suspect that
an event disconnected an area into three smaller
tive in acknowledgingpossible sources
areas simultaneously. Testing this hypothesis by
of error.Afterall, a systematist
cannot
biotic
relationships seems impossible (because
guaranteethata traitwhichhe considers
we have no way of distinguishing those cladoto be unique and derived actually is.
grams reflectingactual trichotomiesfrom those
There is alwaysthe possibilitythathe is
reflecting only our failure to find the relevant
synapomorphies that would resolve a dichotodividinga singlegroupintotwoor lumping two groupsinto one. If a trichotomy mous cladogram), but it should be subject to independent testing by data fromhistorical geoloor
representseithera genuinetrichotomy
gy, which can either be in accordance with such
two unresolveddichotomies,then a dia synchronous tripartitedisconnection of the total area or not.
chotomycould just as well represent
a lumpedtrieithera genuinedichotomy,
If multiple speciation can occur in naor a singlelineage dividedmischotomy,
takenly into two. When a systematist ture and if in certain circumstances it can
claims thata particularspeciationevent be discerned, then cladists would be
is trichotomous,
he maybe mistaken,but wise to devise methods of representing
he mayalso be mistakenin claimingthat multiple speciation in their cladograms
it is dichotomous.Duringtheearlystages and classifications.One theme which the
of investigation,
doubtsabout the actual cladists repeatedly voice is that anything
cladisticrelationswhichcharacterizethe represented in a cladogram and classifi-

This content downloaded from 142.51.1.212 on Mon, 21 Jul 2014 17:05:11 PM


All use subject to JSTOR Terms and Conditions

1979

LIMITS

OF CLADISM

427

cation should be representedexplicitly


and unambiguously.
Ifcladistswere willing to arguethattrichotomous
speciation
never took place, then trichotomiesin
A
B
7 c
cladograms would not be ambiguous.
FIG. 2.-The ambiguitybetween genuine trichotThey would always representa pair of omies and unresolved dichotomies. (a) A cladounresolveddichotomies.As thingsstand gram which represents unambiguously a pair of
now, they are ambiguous,representing successive dichotomies; epistemological doubts, if
eithera genuinetrichotomy
orelse a pair any, are not indicated. (b) A cladogram which is
used to represent a pair of unresolved
of unresolveddichotomies.Whetheror commonly
dichotomies, a genuine trichotomy,common ancesnot cladiststhinkthatinstancesof mul- tryand reticulate evolution. (c) A cladogram which
tiple speciation can be discerned, the indicates doubt about the actual relations between
adoptionof such an ambiguousmode of the species indicated. Such cladograms might in
representationseems counter-produc- time resolve to (a) a pair of dichotomies or (b) a
trichotomy.Common ancestry and reticutive.A betteralternativewould be to re- genuine
late evolution must be represented in some other
serve trichotomiesfor representingtri- but equally unambiguous fashion.
chotomousspeciationeventsthewaythat
dichotomiesare used to representdichotomousspeciationevents,epistemo- ANCESTOR-DESCENDANT
RELATIONS
logical doubtsto one side, and to devise
Although
some
disagreement
exists
another way of indicating incomplete
over
the
occurrence
of
multiple
speciaknowledge.For example,ifa systematist
is reasonablysure thattwo species are tion, everyone agrees thatcertain species
each other'sclosest relatives,he can in- which once existed no longer exist and
dicate this by a dichotomy.If he thinks that some of these species gave rise to
--they
mightbe butis notsure,he can con- Recent species. The cladists' complaints
nectthemby dots.Similarremarksapply about the ancestor-descendant relation
to trichotomies
(see Fig. 2). No systemof cannot possibly be interpretedas empircan representeverything,ical. They are clearly methodological.
representation
but if doubt is important
enoughto rep- Once again the distinctionmust be made
resent in a cladogram,it is important between difficultiesin discerning ancestor-descendant relations and difficulties
enoughto representunequivocally.
Similarremarksare equally applicable in representing them. Farris (1976:272)
to classifications.For example, Nelson acknowledges this distinction when he
(1973c) distinguishesbetween the two presents two reasons fornot treatingfossortsofrepresentation
commonlyused in sil species as ancestral to later species:
the Linnaean hierarchy-subordination First, there is no obvious way of using a classification to represent ancestor-descendant relaand sequencing-and shows how comtionships.A taxonomichierarchywithits wellbinationsof these two conventionscan
nestedtaxais naturallysuitedonlyto the reprerepresenta varietyof evolutionaryphesentationof sister-group
relationships.Second
nomena includingtrichotomous
speciawhilesister-group
relationships
maymoreorless
readilybe establishedthroughthe detectionof
tion. More recently,Wiley (1979) has
apomorphoussimilarities,ancestor-descendant
suggested using sedis mutabilis whenrelationships
maynotbe so established.
ever the orderof listingtaxa in a classificationmeansnothing.In additionto the
A second distinction of equal impordifficulties
whichtherecognition
ofmultiple speciationraisesforcladogramsand tance is between the recognition of exit also poses problemsfor tinct species as species and deciding
classifications,
the properdefinitionof monophylyand which species gave rise to which. Those
related terms (Wiley, 1977; Platnick, cladists who are willing to recognize ex1977c).
tinct species as species while maintain-

This content downloaded from 142.51.1.212 on Mon, 21 Jul 2014 17:05:11 PM


All use subject to JSTOR Terms and Conditions

428

SYSTEMATIC

ZOOLOGY

VOL.

28

ing thatthe ancestor-descendant


relation enough and with sufficient
certaintyto
is unknowableare put in the positionof justifysystematists'
attempting
to repreexplaininghow we can knowthe former sentthemin theirclassifications.
An even
but notthelatter(Engelmannand Wiley, more fundamentalquestion is whether
1977). Those cladistswho arguethatnei- biological species are sufficiently
importheris knowableare sparedthisdilemma tantto the evolutionaryprocess to warbut are faced withthe problemof what rantthe attentiontheyare given. But in
to do withfossils.L0vtrup(1973) argues any case, the existenceand extentof bithatextinctspecies are so poorlyknown ological species are almost always inthattheyshouldbe totallyexcludedfrom ferredindirectlyfromcharacterdistribuclassifications.Crowson (1970) sees the tions. Hence, mistakes will be made.
in ourabilityto recognizeex- When these species are extinct,mistakes
differences
tinctand extantformsto be sufficientlyare even more likely and our abilityto
greatthattheyshouldbe includedin sep- correctthemeven morelimited.No one
arate classifications.Nelson (1972a:230) could possiblyclaimthatall pterodactyls
reasonsthatfossilgroupsshouldbe fitted belonged to a single species, but finer
intoclassifications
whichalso containre- distinctions
are highlyproblematic.
cent groupsbut distinguishedby some
Thus, those cladists who continue to
convention. Finally, Rosen (1978:176) maintainthatbiological species are the
agrees that extant and extinct forms ultimateunitsin theirinvestigations
are
should be distinguishedbut that they forcedto admitthattheircladogramsand
shouldbe treatedas thesame sortsofen- classificationsmightbe mistaken.They
tities-groups distinguishedby unique mighthave lumped several species into
derivedcharacters.In thissectionI will one or divided a single species intotwo
deal firstwith difficulties
in discerning or morespecies. For extinctforms,these
species and ancestor-descendantrela- mistakesare difficult,
if not impossible,
tions and then with problemsof repre- to remedy.Cladistswho have abandoned
sentation.
the biologicalspecies conceptare spared
Hennig (1966, 1975) and Brundin this problem. They are not classifying
(1972b) maintain that the biological species but are groupingorganismsacspecies conceptis centralto phylogenet- cording to the possession of particular
ic systematics.
Fromthe beginning,crit- traits.The only doubt associated with
ics ofthebiologicalspecies concepthave such an exercise is whetherthe organargued thatit is not sufficiently
"opera- isms actuallyhave the traitattributedto
tional" even forextantspecies. Perhaps them,whetheror notit is actuallya trait
it is possible to ascertainwhich organ- and not several,and whetherit has the
isms actually have mated with one distribution
attributed
to it.The question
anotherto produce fertileoffspring
and remainswhetherit is any easier to indiwhich organismscannotmate witheach viduatetraitsthanit is to individuatetheotherand/orproducefertileoffspring,
but oreticallysignificant
taxa.The two seem
it is impossibleto discoverwhich could both conceptuallyand inferentially
very
have done so but did not. When the closelyconnected.
species under investigationare extinct, If extinctspecies cannotbe recognized
so the criticsargue,nothingcan be dis- as genuinegroupsofsome sort,theycancovereddirectlyabout theirbreedingre- not be recognizedas sistergroups.But
lations.It is certainlytruethatthe bio- assumingthatextinctformscan be reclogical species concept is not very ognized as extinctspecies or as extinct
operational,but as I have argued else- groupsdelimitedby a particular
trait,can
where, no theoreticallysignificantcon- ancestor-descendantrelations between
cept in science is (Hull, 1968,1970).The extinct and extant formsalso be disinterestingquestion is whetherbiologi- cerned?Cladistsare all butunanimousin
cal species can be discerned often claiming that they cannot. As Nelson

This content downloaded from 142.51.1.212 on Mon, 21 Jul 2014 17:05:11 PM


All use subject to JSTOR Terms and Conditions

1979

LIMITS

OF CLADISM

(1973b:311) states, "Indeed, I have assumed that ancestral species cannot be


identifiedas such in the fossil record,and
I have pointed out that this assumption
is fundamentalto Hennig's phylogenetic
systematics." Two considerations seem
central to the cladists' claim that ancestor-descendantrelations are unknowable.
First, any distribution of characters
which would imply that one taxon was
ancestral to another would equally imply
that one was the sister group of the other,
and no other way exists for deciding
ancestor relations. Second, given the vast
number of species which must have existed fromthe beginning of time and the
relatively few which have leftrecords, it
seems very unlikely that very many actual common ancestors have been discovered.
I find the second argument more persuasive than the first.It is always possible
that what appears to be a genuine ancestor is really a sister group and that the
two share an unknown common ancestor.
Of course, it is always possible thata trait
which has been recognized as a single
traitis actually two, that a traitwhich is
considered to be derived really is not,
that a derived characterwhich is thought
to be unique really is not, etc., etc. The
possibility of error pervades all of science. What is needed is some reason to
believe that not only is it possible forthe
two groups to be sister groups instead of
one being ancestral to the other,but that
one hypothesis is more probable than the
other. Reference to the vast number of
species which have existed and the few
which have left any traces of their existence does just that. If paleontologists
claim that they have identifiedan extinct
species which is in a direct line of descent to some earlier or later species,
then such claims are on the face of it extremelyunlikely.
For example, when Romer claims that
amphibians arose from ancient crossopterygians,he surely could not have meant
that one of the fossils already collected
represented the actual stem species from
which all amphibians arose and that he

429

knewpreciselywhichfossilsthesewere.
Perhapspaleontologistshave been overly enthusiastic in identifyingmissing
linksand commonancestors,but I findit
hard to believe thattheyhave been this
foolhardy.As Harper (1976) and others
have pointed out, paleontologistsrarely
specifya particularspecies as a common
ance*stor.Almost always a genus or
highertaxonis specified.When paleontologistsclaim that one genus or other
highertaxonarose fromanother,theydo
not mean thathighertaxa "evolve," althoughat one time macroevolutionwas
more popular than it is now. All they
mean is thatthe species whichis ancestral to this highertaxon,if it were ever
discovered,would be placed in thetaxon
whichhas been specifiedas being ancestral.For example,whetheror not specimens of the actual species which gave
riseto amphibianshave been discovered,
whetheror nottheycould be recognized
as such iftheywere,Romeris committed
to the view thatthey would be placed
Nor are paamongthe crossopterygians.
leontologistsnecessarily committedto
Even
ofmonophyly.
Simpson'sdefinition
thoughthey may be willing to specify
only a genus or otherhighertaxonas a
commonancestor,theymaystillhave the
goal of makingall highertaxa monophyletic at the species level and would redrawthe boundariesof theirhighertaxa
if theybecame convincedthata higher
taxonwas descended fromtwo or more
species, even thoughthese species were
all includedin a highertaxonofequal or
lower rank.But it mustalso be admitted
thatclaiminga highertaxonas a common
ancestoris not as empiricalas claiming
thata particularspecies is. Higher taxa,
as theyare usuallyconstructed,
are largely a functionofthe principlesof classification adopted (Engelmann and Wiley,
1977).
Harper(1976) also presentseightprinciples which he thinkscan help distinrelationsfromancesguish sister-group
relations.All of Harper's
tor-descendant
principlescannotbe discussedhere,but
one is sufficiently
importantto warrant

This content downloaded from 142.51.1.212 on Mon, 21 Jul 2014 17:05:11 PM


All use subject to JSTOR Terms and Conditions

430

SYSTEMATIC

mentioning-the role of fossils. Does not


the existence of fossils in certain strata
and their absence in others, especially
when a fairlyextensive fossil record is
available, imply something about possible ancestor-descendant relations? Hennig thinksit does. For example, he (Hennig, 1966:169) argues thatthe "sequence
in which the characters in question
evolved" is "sometimes clarified by fossils." He also thinksthat fossils can help
determine the minimumage ofthe monophyletic groups to which they belong
(Hennig, 1975:112). Nelson (1969:245) is
a good deal more sceptical: "If a given
fossil could be demonstrated to have
been a representativeof a population ancestral to a Recent species," it might be
of some significance,but such an "ancestor-descendant relationship strictly
speaking cannot be demonstrated." Even
though Nelson (1972b:367-370) admits
that data concerning "relative stratigraphic position" might narrow the
"range of possible relationships held by
the taxa in question," he recommends
divorcing "problems of relationships
fromdata concerning stratigraphicdistribution of fossils." He concludes that
ancestor-descendant relations are both
unknown and unknowable in an "empirical sense," that is, "by way of inference
fromobservation of studymaterial" (Nelson, 1973b:311).
Following Popper (1959, 1972), several
cladists have argued that hypotheses
about ancestor-descendant relations are
unscientific because they are unfalsifiable (Wiley, 1975; Engelmann and Wiley,
1977; Platnick and Gaffney,1977, 1978;
Cracraft,1978; Nelson, 1978a; Patterson,
1978). Because these hypotheses are unscientific,they should play no role in science. At one time Bonde (1975) agreed
with this line of reasoning, but in the
meantime he (Bonde 1977:772) has noticed something else about ancestor-descendant hypotheses: they are bolder
than sister-group hypotheses. Because
cladograms claim less than trees and
trees claim less than scenarios, scenarios
are bolder than trees and trees are bolder

ZOOLOGY

VOL.

than cladograms. And, according to Popper, bolder hypotheses are preferable to


less bold hypotheses. Thus, two of Popper's desiderata come into conflict.If all
the relevant data were available, bolder
hypotheses would be easier to falsifybecause they imply more. The problem is
that in this case, the evidence necessary
to test the bolder hypotheses is much
harder to obtain than the evidence necessary to test the less bold hypotheses.
And what is worse, mistakes in ascertaining ancestor-descendant relations introduce much more serious errors into a
study than mistakes in ascertaining sister
groups (Engelmann and Wiley, 1977).
As I mentioned earlier, I am highly
suspicious of scientists claiming that
something cannot be known, especially
when they do not need to know it and
their opponents do. In this case, I think
thatthe cladists have exaggerated the difficultiesin making reasonable inferences
about prabable ancestors for a second
reason as well-difficulties in representation which common ancestors pose for
both cladograms and classifications. If
common ancestors cannot be recognized
with reasonable certainty,then the evolutionists are in trouble. If they can be,
then cladists are in trouble. As Tuomikoski (1967:144) remarks, "Hennig and
Brundin are aware of the factthat such a
linear classification of monophyletic
groups is capable of expressing only horizontal sister relationships, not vertical
mother-daughterrelationships." If ancestor-descendant relations can never be
known with sufficientcertaintyto warrant postulating them, then the inability
of cladograms and classifications to represent them is no great drawback, but
Wiley (1977) has suggested that a better
strategywould be to devise representational devices capable of representing
them just in case. So far,several systems
of representation have been suggested
for distinguishing extinct from extant
groups, and the difficulties which attempts to distinguish between sistergroup relations and ancestor-descendant
relations in cladograms and classifica-

This content downloaded from 142.51.1.212 on Mon, 21 Jul 2014 17:05:11 PM


All use subject to JSTOR Terms and Conditions

1979

LIMITS

OF CLADISM

431

tionshave been pursued at greatlength


1974; Patterson,
(Nelson,1973c;Cracraft,
1976; Pattersonand Rosen, 1977; Wiley,
in clado1979).For exampletrichotomies
B
A Y
C
gramsare alreadyambiguous,representof evolution. (a)
possible
patterns
FIG.
3.-Three
speciationevents
ingeithertrichotomous
A phylogenetic tree in which a single lineage
or a pair of unresolveddichotomies.At- changes in time without splitting. (b) A phylogetemptingto representcommonancestry netic tree in which a single ancestral lineage splits
in this same way would introduceyet into two descendant lineages and both descendant
anotherdimensionof ambiguity.But I lineages diverge fromthe ancestral lineage. (c) A
tree in which splittingoccurs but only
see no reason why some systemof rep- phylogenetic
one of the descendant lineages diverges fromthe
disresentationcannotbe developed to
ancestral lineage.
tinguishthese three possible situations
in bothcladogramsand classifications.
sificationbut forthe purposesof formuPHYLETIC EVOLUTION AND SPECIATION
latingand testinghypothesesabout the
Cladists are frequentlyinterpretedas evolutionary
as sciprocess.As important
claimingthatphyleticevolutiondoes not entificclassifications are, scientifictheooccur. Once again, the actual thrustof ries are even more important.
theirdiscussionof phyleticevolutionis
Eldredge and Gould (1972) have armethodological,not empirical.Figure 3 gued that phyletic evolution rarely, if
representsthreepossible patternsofevo- ever, occurs and that the evidence for it
lution:gradualchangein a lineage with- is largely illusory. However, regardless
out the lineage splitting,splittingfol- of whether or not phyletic evolution oclowed by both descendant lineages curs, the cladists maintain that a single
diverging,and splittingfollowedby only lineage should not be divided into
one of the descendantlineages diverg- species no matter how much it might
ing. Simpson (1961) has argued that change. New species are to be recogsometimeslineages change sufficientlynized only when splittingoccurs. Accordthroughthe courseoftheirdevelopment ing to Hennig (1966, 1975), species are
so thatlaterstagesshouldbe considered not classes of similar organismsbut welldistinctspecies even thoughno splitting integrated gene pools. In phyletic evohas taken place. The inevitable opera- lution, the lineage changes throughtime
tionalist objections have been raised but retains its integration. Hence, it
against Simpson's suggestion: that no should be considered a single species. As
way exists for dividing a Platnick (1977a:97) argues:
non-arbitrary
continuouslyevolving lineage into distinctspecies and thatlineages cannotbe
To attempt to divide a species between speciation events would indeed be arbitrary:we would
established in the firstplace because
thereis no way ofknowingwhichorgan- not call an individual person by one name at age
10 and a differentname at age 30. Dividing
isms are ancestralto which. I have alspecies at their branching points, however, beat
sufficient
readydiscussedthese topics
comes not only non-arbitrarybut necessary: we
lengthboth in thispaper and elsewhere
would not call a son by the same name as his
(Hull, 1965,1967,1970).Ifthecontinuity father.
to
of phyleticevolutionwere sufficient
In this instance, I think Hennig is
preclude "objective" subdivision,then
all our measurementsof physical space right,not because of any epistemological
would be equally suspectbecause space or methodological reasons, but because
is even morecontinuous.Perhapsthe es- his decision is consistent with the most
tablishmentof lineages is a riskybusi- theoretically appropriate definition of
important "species." Species are integrated linness,butitis also an extremely
notforthepurposesofclas- eages developing continuously through
undertaking,

This content downloaded from 142.51.1.212 on Mon, 21 Jul 2014 17:05:11 PM


All use subject to JSTOR Terms and Conditions

432

SYSTEMATIC

time (Simpson, 1961; Mayr, 1963; Ghiselin, 1974; Hull, 1976, 1978; Wiley,
1978). Cohesiveness at any one time and
continuitythroughtime are what matter,
not phenotypic or even genotypic similarity.Some lineages may diverge extensively through time without splitting;
some not. It does not matter.A continuously evolving lineage should no more
be divided into distinct species than an
organism undergoing ontogenetic development should be divided into distinct
organisms. These same considerations
apply to the other situations shown in
Fig. 3.
Cladists argue thatnew species should
be recognized only when splitting occurs. In non-saltative speciation, temporal continuityis maintained, but the cohesiveness of the lineage is disrupted.
They also argue that whenever splitting
takes place, the ancestral species must be
considered extinct. Bonde (1975:295)
characterizes Hennig's position as follows:
By the process of speciation the ancestral species
is split into two new species called sisterspecies
(or daughterspecies) .... That the two sisterspecies are new and the ancestral species becomes
extinct at the speciation is most practical for a
consistent terminology,and it also follows frorn
Hennig's species concept, and fromthe definition of phylogenetic relationship below (cf.
Brundin, 1972a: 118).

According to Hennig, the two sorts of


splittingdepicted in Figs. 3b and 3c are
to be treated in exactlythe same way; i.e.,
as a mother species becoming extinct as
it divides into two daughter species. It
does not matterthat in one instance both
daughter species diverge fromthe mother species, while in the second only one
does. As much trouble as it may cause the
systematist attempting to distinguish
ancestors from descendants, it simply
does not matterthat the mother species
is indistinguishable from one of its
daughter species. I agree with Hennig's
reasoning but not his conclusion. If integrationis what matters,then the differences indicated in Figs. 3b and 3c are
irrelevantforthe individuationof species.

ZOOLOGY

VOL. 28

If divergenceis irrelevantwithoutsplitting,it should be just as irrelevantwith


splitting.
On Hennig'sview,whatshould
really matteris the integrationof the
gene pool. At speciation, so Hennig
claims, the ancestral gene pool totally
disintegrates,
resultingin the extinction
of the ancestral species. Accordingto
Bonde (1-977:754),Hennig's species concept is the"onlylogicalextensionin time
of the concept of the integratedgene
pool. At speciationthisgene pool is disintegratedand two (or more) new sister
species originate, while the original
species becomes extinct."Wiley (1978:
21-22) agrees that "in most cases the
methodologicalnecessityof postulating
extinctionofancestralspecies in phylogas advocatedby Heneny reconstruction
nig (1966) is biologically (as well as
methodologically)sound," but not always.Accordingto his own evolutionary
species concept,an ancestralspecies can
survivea split,ifit can "lose one or more
constituent
populationswithoutlosingits
historicalidentityor tendencies."
Ifspeciationcan occuronlybythemassive disintegration
of the parentalgene
pool, then Hennig's decision always to
treatancestralspecies as extinctis wellfoundedbiologically.However, if Mayr
(1963), Carson (1970), and Eldredge and
Gould (1972) are rightand speciationalways (or usually)takesplace by the isolation of small, peripheralpopulations,
thenit seems veryunlikelythatsuch an
eventwill totallydisrupttheorganization
of the parentalgene pool, and Hennig's
conventionof treatingancestralspecies
as always going extinctupon speciation
loses its empiricalsupport.3As interesting as theseempiricalconsiderations
are,
as I have mentionedseveral times previously, cladists are currentlydisassociating themselves fromany particular
3The arguments presented by Eldredge and
Gould (1972) support Hennig's contentionthatnew
species should be recognized only when splitt'ing
occurs. These same arguments, however, count
against Hennig's contention that ancestral species
always disintegrateupon speciating.

This content downloaded from 142.51.1.212 on Mon, 21 Jul 2014 17:05:11 PM


All use subject to JSTOR Terms and Conditions

1979

LIMITS

OF CLADISM

433

views about the evolutionaryprocess. as integratedgene pools, then species


For example, Bonde (1977:793) states changinggraduallythroughtime should
that"contraryto my earlierbeliefs,the be consideredsingle species and not didetailsoftheprocessofspeciationare not vided successivelyinto distinctspecies.
important
forthe phylogeneticsystemat- New species should be recognizedonly
ic theorybecause the patternsresulting at splitting,
but at splittingthe ancestral
fromallo-, para- and sympatricspecia- species does not always go extinct.How
tionscan be analyzedin the same way in many species are to be recognized at
terms of degrees of phylogeneticrela- splittingdepends on whathappensto the
tionship."If so, thenfactsaboutthe evo- integrationof the parentalgene pool. If
lutionaryprocess cannotbe used to cast it remainslargelyunaffected,then the
doubt on the cladists'researchprogram. ancestralspecies continuesto existas it
Such considerationscannot therebybe buds offdescendantspecies. If not,itbeused to supportit either,and Hennig's comes extinct.Whetheror notthe daughargumentsmustbe abandoned.
terspecies divergeaftersplittingis irrelThe only remainingconsiderationis evant forthe individuationof species as
"consistency of terminology."Hennig integratedgene pools. Divergence may
seems to thinkthatcalling a species by matterfora hostofotherreasons,but not
the same name beforeand afterspecia- for the individuation of species. Of
tion would cause all sortsofterminolog- course,none ofthe precedingis relevant
ical confusion. However, comparable to those cladistswho thinkthatthe recterminological difficultiesare easily ognitionofspecies as some sortofsignifsurmountedat the level of organisms. icant evolutionaryunit plays no role in
When a single Parameciumsplitsdown cladistic systematics.It is also true that
the middle to formtwo new organisms, in the absence of some sort of divereach is considereda distinctorganism.If gence, the systematist
is unlikelyto nowe were proneto name such entities,we tice thatspeciationhas occurred(see latwould give each a separatename. How- er discussionof the Rule of Deviation).
ever, Hydra can continueto existwhile
MONOPHYLY AND RETICULATE
buddingoffotherHydra. Once again, if
EVOLUTION
we were inclinedto,we could give each
of these organismsits own name. The
The term"monophyly"has had an exparentHydrawould retainitsnameeven tremely varied history. According to
thoughit budded offotherdescendant Mayr(1942:280) all taxashouldbe monoHydra. Finally, in sexual reproduction, phyletic.By thishe meansthatall organtwo parentalorganismsproducegametes isms included in a taxonshould be "dewhichcome togetherto forma third.Or- scendantsof a single species." He does
ganisms need not cease to exist when not say, however,whetherall the dethey mate. Queen Victoriaand Prince scendantsof a single species should be
Albertremainedthe.sa-meorganismsas included in a singl.ehighertaxon,nor
they produced child afterchild. Occa- whetherthis stem species itselfshould
sionally,we do call a son by the same be included in this highertaxonor exname as his father.Usuallywe have the cluded fromit.In laterdiscussions,Mayr
good sense to add "Jr.,"but whetherwe, (1969) statesthathe intendedno such imdo ornot,theyremaindistinctorganisms. plications.He also statesexplicitlythat,
If such practicesoccasion so little con- as faras animalsare concerned,reticulate
fusion at the level of organisms,there evolutionmay be disregarded.He justiseems no reasonforthemto introducein- fieshis positionby referenceto how rare
surmountableterminologicaldifficulties genuine hybridspecies are among,aniat the level of species.
mals. Botanists mightchoose to make
In thissectionI have arguedthatifone anotherdecision. Hennig (1966:73) deaccepts Hennig's conceptionof species fines "monophyly"somewhatdifferent-

This content downloaded from 142.51.1.212 on Mon, 21 Jul 2014 17:05:11 PM


All use subject to JSTOR Terms and Conditions

434

SYSTEMATIC

ZOOLOGY

VOL. 28

ly: "A monophyleticgroupis a groupof same as thatofMayrand Hennig:hybrid


species descendedfroma single("stem") species mayoccurin plants,but theyare
species, and which includes all species rare in animals. Simpson also does not
descended fromthis stem species." On specifyany uniformway in which stem
Hennig's definition,all and only those species are to be treated.
Several pointsare at issue in the prespecies (both extantand extinct)which
are descendedfroma singlespecies must ceding discussion:(1) should all taxa be
be includedin a singlehighertaxon(see monophyleticat the level of species, (2)
also Bonde, 1975:293; 1977:757).Hennig if so, how are hybridspecies to be treat(1966:207-208)also recognizesthatthehy- ed, (3) shouldall thedescendantsofa sinbrid origin of species would produce gle stemspecies be includedin the same
"special complications"forthe principle highertaxon,and (4) if so, what should
and echoes Mayrin his re- be done withthestemspecies itself?The
ofmonophyly
is rare cladists find themselves in agreement
sponse thatluckilyhybridization
in animals.Amongplantshe is willingto withrespectto (1) and (3). All species in
countenance polyphyleticspecies but any higher taxon must be descended
not polyphyletichigher taxa (Hennig froma single ancestral species. Con1966:208). Hennig (1966:64, 70-72, 207) versely,all the descendantsof any one
is not clear about the recognitionand stemspecies mustbe includedin a single
placementof "stem species" in his clas- highertaxon.Some disagreementexists
of (2) and (4),
over the propertreatment
sifications.
As much as Mayr and Hennig differ hybrid species and stem species. For
with respect to the principleof mono- thosecladistswho maintainthatthe only
phyly,theyagree thatall species includ- kind of event in the past which can be
ed in a single highertaxonmustbe de- reconstructedwith reasonable certainty
scended froma single stem species. is speciation(orat least theproductionof
Simpson (1961:124) presents a much unique derivedcharacters),commonanbroader definition:"Monophylyis the cestry,whethersingle or multiple,canderivation of a taxon throughone or not be ascertained.The distributionof
is
more lineages (temporalsuccessions of charactersresultingfromhybridization
ancestral-descendant
populations)from indistinguishable fromthat resulting
whichin turnis indisone immediatelyancestral taxon of the froma trichotomy,
same or lowerrank."Accordingto Simp- tinguishablefroma pair of unresolved
if it dichotomies,which in turnis indistinson, a taxonis strictly
monophyletic
arises froma single immediatelyances- guishable fromcommonancestry.Since
tralspecies; itis minimallymonophyletic all three phenomena can produce the
ifit arises froma single immediatelyan- same distribution
oftraits,all such distrias
cestral taxon of its own or lower rank. butionsoftraitsshouldbe interpreted
if unresolved dichotomies-other things
Thus, a genus is strictlymonophyletic
it arises froma single immediatelyan- being equal.
The claim that we can never distincestral species. It is only minimally
monophyleticif it arises fromtwo or guishbetweena genuinecommonancesmore species which are included in the torand a closelyrelatedsistergrouphas
ifit arises some plausibility.The claimthatwe cansame genus. It is polyphyletic
fromtwoor morespecies notincludedin
the same genus. As broad as Simpson's
definitionis, it too is confronted
by the monophyleticfurtherweakens the already loose redifficultyof accommodating hybrid lationship between phylogeny and biological classpecies (Hull, 1964).4 His responseis the sifications which Simpson proposes, Bonde (1975,
1977) interpretsme to be defending Simpson's definition of "monophyly." There is a point to making
one's polemics as polite as possible, but when an
'Althoughthe mainpurposeof Hull (1964) was attack is interpreted as a defense, that point has
to argue thatallowingtaxa to be only minimally been passed.

This content downloaded from 142.51.1.212 on Mon, 21 Jul 2014 17:05:11 PM


All use subject to JSTOR Terms and Conditions

1979

LIMITS

OF CLADISM

435

not discernhybridspecies does not.We ly. If speciation is on occasion trichotoknowwithas muchcertainty


as we know mous and such events can be discerned,
any empiricalphenomenonthatcertain why not represent them? If common
species are of hybridorigin.We did not ancestors and reticulate evolution can be
observe the hybridizationevent in the discerned, why not represent them?
past,butwe can inferitsoccurrencewith However, the principle of monophyly is
a high degree of certainty.
For example, importantto cladism. In recent years a
if a species were discovered at the large literature has grown up over the
boundarybetween two species withthe "proper" definitionof "monophyly" and
combined chromosomes of these two related terms (Ashlock, 1971, 1972; Nelspecies, the likelihoodis veryhigh that son, 1971b, 1973b; Farris, 1974; Platnick,
this species is a hybridofthe othertwo. 1976, 1977b, 1977c; Wiley, 1977). ScienPerhaps the outcomeof a cladisticanal- tificterms change their meanings as sciysis of these three species would be a ence develops. "Gene" did not mean the
pair of unresolveddichotomies,but that same thing at the turn of the century as
is a problemforcladisticanalysis.
it does today. One importantfunctionof
Anotherreasonwhichcladistshave for the historyof science is to trace such serejectingboth stem species and hybrid mantic changes. However, the results of
species as unknowableis the difficulties such inquiries cannot be used to desigwhich both relationspose forunequivo- nate certain meanings as "proper" and
cal representationin both cladograms others as "improper." I'm not sure what
and classifications.
For example,Nelson Darwin meant by the term"monophyly."
(1979:8) observes:
I'm not even sure he ever used the term.

Changing the definitionsof terms in science haphazardly introduces needless


An instance of hybridization would be represented in fundamental cladograms by non-com- confusion. From the point of view of
binable components that exhibit non-random making assertions about classifications,
replication, and in the general cladogram by tri- the term "holophyly" will do as well as
or polytomies that present conflicting,but non- "monophyly." The desire on both sides
random, possibilities fordichotomous resolution to retain the term
"monophyly" fortheir
(as exemplified in fundamentalcladograms).
concept stems fromthe need to establish
continuityin scientificdevelopment. Just
Instead of introducingyet anotherdi- as continuitymattersin individuatinglinmension of ambiguityinto polytomous eages in biological evolution, it matters
cladograms,a betteralternative
would be in individuating scientific lineages in
to devise distinctmethodsof represen- conceptual evolution. The sortof fighting
tationforeach ofthese distinctphenom- over words which is so common in sciena. Perhapswe will neverknowwhich ence is not an idle exercise.
species are actually stem species, perTerminological squabbles to one side,
haps we will never knowwhich specia- the dispute between Ashlock (1971,
tioneventsare trichotomous,
perhapswe 1972) on the one hand and Nelson
will neverknowwhichspecies are ofhy- (1971b, 1973b) and Farris (1974) on the
brid origin,but just in case somedaywe other hand has some substantive points.
can, why not devise methodsof repre- Cladists want to represent cladistic relasentingthese phenomenain cladograms tions by nested sets of derived traits.To
and classifications?Wiley (1978, 1979) do so, they must exclude what they term
proposesto do just this.
"paraphyletic" taxa. As far as assertive
In the previoussectionsI have argued content is concerned, which definitionof
thatcertainmethodologicalprinciplesat- "monophyly" wins out is irrelevant; the
tributedto the cladistsreallydo notcon- structureofthe resultingclassificationsis
tributeall thatmuchtothelong-term
goal not. Whetherone claims thatall taxa must
of cladism-the representation
of cladis- be holophyletic (sensu Ashlock) or monotic relationsexplicitlyand unambiguous- phyletic (sensu the cladists) is important

This content downloaded from 142.51.1.212 on Mon, 21 Jul 2014 17:05:11 PM


All use subject to JSTOR Terms and Conditions

436

SYSTEMATIC

sociologically. What matterssubstantively is that the taxa themselves be grouped


appropriately.
However much cladists and evolutionists disagree about the proper definition
of"monophyly," they agree on one point:
hybrid species pose a problem. Although
I hate to destroythis rare instance of unanimity, I think they are mistaken. The
goals of neither school of taxonomy require that all species arise from single
ancestral species. All that is needed is
that all species have a single origin. In
most cases, species do have their origins
in the speciation of single ancestral
species. It follows thatthese species also
have single origins. But sometimes a new
species arises fromtwo immediately ancestral species. They nevertheless have
their origins in a single speciation event,
and that is all that is necessary. Both
schools need to require that all taxa be
monophyletic, but only as monophyletic
as nature allows.
THE DEVIATION

RULE

According to Hennig (1966:207),


"When a species splits, one of the two
daughter species tends to deviate more
stronglyfromthe other fromthe common
stem species"
(see also Hennig,
1950:111). Brundin (1972b:71) claims
thatthe deviation rule is "one ofthe most
fundmental aspects of the principle of
life" and essential to cladism. Bonde
(1975:296) views deviation as a common
(though not universal) pattern of evolution, which is important(though not necessary) to cladism. Schlee (1971:5) finds
it "unessential" to the Hennigian approach, while Nelson (1971a:373) concludes that it is both unessential and a
methodological principle. If the rule of
deviation is supposed to be an empirical
claim about evolution, it has some plausibility. As Bonde (1975:296) notes, if
speciation takes place only when an
"atypical" population becomes isolated
in an "atypical" environment,one might
reasonably expect this new daughter
species to diverge more fromthe ancestral species than its sister does.

ZOOLOGY

VOL.

28

The rule of deviation also has a methodological form.Whether or not speciation is always accompanied by the production of at least one unique derived
characterin one of the groups,5ifno such
character is discernible, the groups will
not be discernible by the methods of cladistic analysis. Of course, they will not
be discernible by any other methods
either. It is unlikely (probably impossible) for speciation to occur without at
least one new character developing, but
if "character" is used in its traditional
sense to refer to such things as tooth
structure,feathercolor, type ofblood proteins, peculiarities in mating dance, etc.,
it may be the case that speciation can occur without the production of a distinguishing featureof the sortthata systematist is likely to notice. He can know that
the two groups are two groups because
they do not mate. He may even be able
to tell which group is which because they
inhabit differentranges. But he will not
be able to decide which species a specimen belongs to merely by examining it.
(Of course, one might redefine the term
"character" to include spatial location or
some such.)
For those cladists who see theirtask as
the recognition of species and the grouping of these species into more and more
inclusive taxa, the discovery of the appropriate unique derived characters and
shared derived characters is crucial. It is
very likely that the relevant characters
exist. The task is to discover and categorize them appropriately. The cladists

5The number of characters necessary to distinguish a new taxon depends on a varietyofdecisions.


For example, if one assumes that the ancestral
species has already been delineated and can bud
offnew species withoutitselfgoing extinct,a single
unique derived character will do. If the ancestral
species also must be delineated, a second unique
derived character is necessary. If ancestral species
are always considered to be extinctupon speciation
and speciation is always dichotomous, still another
character is required, and so on. Of course, in traditional cladistic classifications, all these "ancestral" species will be treated as sister groups.

This content downloaded from 142.51.1.212 on Mon, 21 Jul 2014 17:05:11 PM


All use subject to JSTOR Terms and Conditions

1979

LIMITS

OF CLADISM

share this task with all systematists.For


those cladists who see their task as the
recognition of sequences of unique derived charactersand shared derived characters, regardless of the coincidence of
the emergence of these characters with
speciation events, the rule of deviation
becomes a tautology. It reduces to the
claim that when a unique derived character is recognized a new group is recognized; otherwise not. By terming the
rule of deviation, interpretedin this way,
a tautology,I do not mean to denigrate it.
CONCLUSION

In this paper I have examined some of


the principles which cladists have suggested to facilitate the representation of
sister-grouprelations in cladograms and
classifications. Currentlysome disagreement exists among the cladists themselves over the precise makeup of these
principles, but eventually a single, generally accepted, reasonably simple system of conventions is likely to materialize. The consistent application of these
principles will result in classifications
which allow anyone who wishes to read
off fromthem the sister-grouprelations
which went into their construction. I
have also raised questions concerning
some of the prescriptionswhich cladists
have enunciated about what cannot be
known about phylogeny or represented
if it were known. All scientific knowledge is fallible, including the recognition
of sister groups. Perhaps ancestors, hybrid species and multiple speciation
events cannot be discerned with the
same degree of confidence as sistergroup relations, but the contrast is not
between fact and fantasy.-Cladists have
selected sister-group relations because
they are relatively easy to discern using
the methods of cladistic analysis and just
as easy to represent in cladograms and
classifications. Opponents have complained that the price is too high. Being
able to infer sister-grouprelations from
biological classifications is not worth the
increase in complexityand asymmetryof
the resulting classifications. They argue

437

thatthe cladists'rules forclassifyingare


simple enough, but that the resulting
classifications
are not.
Anotheravenue ofattack,employedby
some taxonomists,is to agree that genealogy is worthrepresentingin classificationsbut objectto theaspect selected
by the cladists.Divergenceis also an importantfeatureof evolutionarydevelopment.Instead of constructing
classifications solely on the basis of cladistic
relations,varyingdegrees of divergence
should also be reflected.One problem
with this response is that no methods
have been set out thusfarwhichpermit
the inclusionofbothsortsofinformation
in a single classificationin such a way
thatboth are retrievable.It is one thing
to let a varietyof considerationsinfluence how one constructs
a classification.
It is quite anotherto formulatea set of
principlesso thatotherscan retrievethis
informationfromthe classifications.If
classifications
are to be systemsof informationstorageand retrieval,information
mustactuallybe retrievable.So far,only
Ashlockand Brothers(1979) have taken
up the challenge.A secondproblemwith
this alternativeis thatit is likelyto increase thecomplexity
ofboththerulesof
classificationand the resultingclassifications.If classifications
whichrepresent

only sister-grouprelations are too complicated, classifications which represent


sister-group relations, ancestor-descendant relations and degrees of divergence
are likely to be even more complicated.
Another possible response is that cladists have taken Darwin's suggestion too
literally.Classification should be in some
vague sense "phylogenetic," but biological classifications cannot be made to reflectvery much about phylogeny without
frustratingother functions of scientific
classification. The most that systematists
can hope to do is to weigh a variety of
conflicting goals and produce the best
possible compromise. Clear decisions
between these various alternative goals
are not easy, but ifthe cladists have done
nothing else, they have shown the sorts
of rules which are necessary if biological

This content downloaded from 142.51.1.212 on Mon, 21 Jul 2014 17:05:11 PM


All use subject to JSTOR Terms and Conditions

438

SYSTEMATIC

ZOOLOGY

VOL.

28

are to representexplicitly CROWSON, R. A. 1970. Classificationand biology.


classifications
and unambiguouslya particularfeature AthertonPress,New York.
P. J.,JR. 1970. A practicalcriticism
ofphylogeneticdevelopment.They have DARLINGTON,
of Hennig-Brundin"phylogeneticsystematics"
also posed clearlyand forcefully
a chal- and antarcticbiogeography.Syst.Zool. 19:1-18.
lenge to theirfellowsystematists:
if the DARWIN, CHARLES. 1859. On the Origin of
goal of biological classificationis not to Species,a facsimileofthefirstedition(1859)with
by ErnstMayr(1966). HarvardUnirepresentone or moreaspects of phylo- introduction
versityPress,Cambridge,Mass.
geneticdevelopment,whatis the goal of DARWIN, F. 1899. The Life and LettersofCharles
biologicalclassification?
Darwin. D. Appletonand Company,New York.

A. 1949. Einstein:reply.In Schilpp,P.


A., (ed.), AlbertEinstein:Philosopher-Scientist.
Open Court,New York.
I would like to thankWalterBock,Niels Bonde,
ELDREDGE, N. 1979. Cladismand CommonSense.
SteveFarris,ErnstMayr,GarethNelJoelCracraft,
In Cracraft,
J.,and N. Eldredge(eds.),Phylogeny
son, NormanPlatnickand Ed Wiley forlong and
and Paleontology.Columbia UniversityPress,
heated battlesof the mostenjoyablesortover the
New York.
exactnatureofcladisticanalysisand cladisticrepELDREDGE, N., AND S. J.GOULD. 1972. Punctuated
resentation.
Needless to say,ourmeetingofminds
equilibria:an alternative
to phyleticgradualism.
has yetto attainthe isomorphism
characteristic
of
In Schopf,T. M. J.,(ed.), Modelsin paleontology.
cladogramsand cladistic classifications.The reFreeman,Cooper, and Co., San Francisco,pp.
searchforthispaperwas supportedin partby NSF
82-115.
grantSOCX75-03535A01.
ENGELMANN, G. F., AND E. 0. WILEY. 1977. The
place of ancestor-descendantrelationshipsin
REFERENCES
phylogenyreconstruction.
Syst.Zool. 26:1-11.
ASHLOCK, P. D. 1971. Monophylyand associated FARRIS,J.S. 1974. Formaldefinitions
ofparaphyly
terms.Syst.Zool. 20:63-69.
and polyphyly.
Syst.Zool. 23:548-554.
ASHLOCK, P. D. 1972. Monophylyagain. Syst. FARRIS, J. S. 1976. Phylogenetic
classification
of
Zool. 21:430-437.
fossilswithRecent species. Syst.Zool. 25:271ASHLOCK, P. D., AND D. J.BROTHERS. 1979. Sys282.
tematizationand higherclassificationin evolu- GHISELIN, M. 1974. A radical solution to the
tionarysystematics
throughcladisticand anagespecies problem.Syst.Zool. 23:536-544.
neticanalysis.Manuscript.
GRIFFITHS, G. C. D. 1976. The future
ofLinnaean
BONDE, N. 1975. Originof"highergroups":viewnomenclature.
Syst.Zool. 25:168-173.
points of phylogeneticsystematics.Problemes HAIGH, J. 1971. The manuscript
linkageproblem.
actuels de paleontologie-evolution des verIn Hodson, F. R., D. G. Kendall,and P. Tautu
tebres.Coll. Internat.C.N.R.S.,no. 218:293-324.
(eds.), Mathematicsin the Archaeologicaland
BONDE, N. 1977. Cladisticclassification
as applied
HistoricalSciences. Univ. Press,Edinburgh,pp.
In Hecht,M. K., P. C. Goddy,and
to vertebrates.
396-400.
B. M. Hecht (eds.), Majorpatternsin vertebrate HARPER, C. 1976. Phylogeneticinferencein paevolution.PlenumPublishingCorporation,
New
leontology.J.Paleo. 50:180-193.
York,pp. 741-804.
HENNIG, W. 1950. Grundzugeeiner Theorie der
BRUNDIN, L. 1966. Transantarctic
relationships phylogenetischenSystematik.Deutscher Zenand theirsignificance,
as evidencedby chiron- tralverlag,
Berlin.
omid midges.K. svenskaVetensk.Akad.Handl. HENNIG, W. 1965. Phylogeneticsystematics.
Ann.
(4)11:1-472.
Rev. Ent. 10:97-116.
BRUNDIN, L. 1972a. Evolution,causal biology,and HENNIG, W. 1966. Phylogenetic
Unisystematics.
classification.
Zool. Scripta.1:107-120.
versityofIllinoisPress,Urbana.
BRUNDIN, L. 1972b. Phylogenetics
and biogeog- HENNIG, W. 1975. "Cladisticanalysisor cladistic
raphySyst.Zool. 21:69-79.
classification?":
a replyto ErnstMayr.Syst.Zool.
CAMIN, J. H., AND R. R. SOKAL. 1965. A method
24:244-256.
fordeducingbranchingsequences in phylogeny. HULL, D. L. 1964. Consistencyand monophyly.
Evolution19:311-326.
Syst.Zool. 13:1-11.
CARSON, H. L. 1970. Chromosometracersof the HULL, D. L. 1965. The effectof essentialismon
origin of species: some Hawaiian Drosophila
Brit.J.Phil. Sci. 15:314-326;16:1-18.
taxonomy.
species have arisenfromsinglefounderindivid- HULL, D. L. 1967. Certainty
and circularity
in evouals in less than a million years. Science
lutionary
taxonomy.
Evolution21:174-189.
168:1414-1418.
HULL, D. L. 1968. The operationalimperativeCRACRAFT, J. 1974. Phylogenetic
modelsand classense and nonsensein operationalism.
Syst.Zool.
sification.
Syst.Zool. 23:71-90.
17:438-457.
CRACRAFT, J. 1978. Science, philosophy,
and sys- HULL, D. L. 1970. "Contemporary
systematic
phitematics.Syst.Zool. 27:213-215.
losophies."Ann.Rev. Ecol. Syst.1:19-54.
ACKNOWLEDGMENTS

EINSTEIN,

This content downloaded from 142.51.1.212 on Mon, 21 Jul 2014 17:05:11 PM


All use subject to JSTOR Terms and Conditions

1979

LIMITS

OF CLADISM

439

D. L. 1976. Are species reallyindividuals?


Mathematics
in theArchaeological
and Historical
Syst.Zool. 25:174-191.
Sciences. Univ. Press,Edinburgh,pp. 401-409.
Phil. PATTERSON,C. 1976. The contribution
HULL, D. L. 1978. A matter
ofindividuality.
ofpaleonSci. 45:335-360.
tologyto teleosteanphylogeny.In Hecht,M. K.,
HUXLEY, J. 1958. Evolutionary
processesand taxP. C. Goody,and B. M. Hecht (eds.), MajorPatonomywithspecial referenceto grades.Uppsala
terns in VertebrateEvolution, Plenum Press,
Univ.Arssks.,pp. 21-38.
New York.
KRUSKAL, J.B., I. DYER, AND P. BLACK. 1971. The
PATTERSON, C. 1978. Verifiability
in systematics.
vocabularymethodof reconstructing
language
Syst.Zool. 27:218-221.
trees. In Hodson, F. R., D. G. Kendall,and P. PATTERSON, C., AND D. E. RoSEN. 1977. Review
Tautu (eds.), Mathematicsin the Archaeological of ichthyodectiform
and otherMesozoic teleost
and HistoricalSciences.Univ.Press,Edinburgh, fishesand the theoryand practiceof classifying
pp. 361-380.
fossils.Bull. Amer.Mus. Nat. Hist. 158:81-172.
LAUDAN, L. 1977. Progressand its problems:to- PLATNICK, N. I. 1976. Aremonotypic
generapossiof
warda theoryof scientificgrowth.University
ble? Syst.Zool. 25:198-199.
CaliforniaPress,Berkeley.
PLATNICK, N. I. 1977a. Review of concepts of
conventionand
species. Syst.Zool. 26:96-98.
LqVTRUP, S. 1973. Classification,
logic.Zool. Scripta.2:49-61.
PLATNICK, N. I. 1977b. Paraphyletic
and polyphyLUCCHESI, J.C. 1978. Gene dosage compensation
leticgroups.Syst.Zool. 26:195-200.
and the evolutionof sex chromosomes.Science PLATNICK, N. I. 1977c. Monotypyand the origin
202:711-716.
ofhighertaxa:a replyto E. 0. Wiley.Syst.Zool.
MAYR, E. 1942. Systematicsand the origin of
26:355-357.
species. Columbia UniversityPress,New York. PLATNICK, N. I. 1977d. Cladograms,phylogenetic
MAYR, E. 1963. Animal species and evolution.
trees,and hypothesistesting.Syst.Zool. 26:438HarvardUniversity
442.
Press,Cambridge,Mass.
MAYR, E. 1965. Classificationand phylogeny. PLATNICK, N. I. 1979. Philosophy
and thetransforAmer.Zool. 5:165-174.
mationofcladistics,Syst.Zool. 28:537-546.
MAYR, E. 1969. Principlesof systematic
zoology. PLATNICK,N. I., AND H. D. CAMERON. 1977.CladisMcGraw-Hill,New York.
tic methodsin textual,linguistic,and phylogeMAYR, E. 1974. Cladisticanalysisor cladisticclasneticanalysis.Syst.Zool. 26:380-385.
Z. f.zool. Systematik
sification.
u. Evolutionsfor- PLATNICK,N. I., AND E. GAFFNEY. 1977. Systematschung.12:94-128.
ics: a Popperian perspective. Syst. Zool. 26:360MAYR, E. 1978. Originand history
of some terms
365.
in systematicand evolutionarybiology. Syst. PLATNICK, N. I., AND E. GAFFNEY. 1978.EvolutionZool. 27:83-87.
arybiology:a Popperianperspective.Syst.Zool.
NELSON, G. 1971a. "Cladism" as a philosophyof
27:137-141.
classification.
Syst.Zool. 20:373-376.
PLATNICK,N. I., AND G. NELSON. 1978.A methodof
NELSON, G. 1971b. Paraphyly
and polyphyly:reanalysisforhistoricalbiogeography.Syst.Zool.
definition.
Syst.Zool. 20:471-472.
27:1-16.
NELSON, G. 1972a. Phylogenetic
relationshipand POPPER, K. R. 1959. The logicofscientific
discovclassification.
Syst.Zool. 21:227-230.
ery. Basic Books, New York,480 pp.
NELSON, G. 1972b. Commentson Hennig's"phy- POPPER, K. R. 1972. Objective knowledge. Oxford
logeneticsystematics"and its influenceon ichUniversityPress, Oxford.
thyology.
Syst.Zool. 21:364-374.
POPPER, K. R. 1976. Unended quest: an intellectual autobiography. Open Court Press, La Salle,
NELSON, G. 1973a. The higher-level
phylogenyof
Illinois, 255 pp.
vertebrates.
Syst.Zool. 22:87-91.
NELSON, G. 1973b. "Monophylyagain?": a reply ROMER A. S. 1955. The vertebrate body. W. B.
Saunders Company, Philadelphia and London.
to P. D. Ashlock.Syst.Zool. 22:310-312.
ROSEN, D. 1978. Vicariant patterns and historical
as an expression explanation in
NELSON, G. 1973c. Classification
biogeography. Syst. Zool. 27:159of phylogeneticrelationship.Syst.Zool. 22:344188.
359.
Ross, H. H. 1974. Biological systematics.AddisonNELSON, G. 1974. Darwin-Hennigclassification:
Wesley Publishing Company, Reading, Massaa replyto ErnstMayr.Syst.Zool. 23:452-458.
chusetts.
NELSON, G. 1978. Classification
and prediction:a SCHLEE, D. 1971. Die Rekonstruktionder Phyloreplyto Kitts.Syst.Zool. 27:216-217.
genese mit Hennig's Prinzip.Aufsatzeu. Red
NELSON, G. 1979. Cladisticanalysisand synthesis:
Senckenberg. Naturforsch.Ges. 20:1-62.
principlesand definitions,
witha historicalnote SIMPSON, G. S. 1961. Principles of animal taxonoHULL,

on Adanson's Familles des Plantes (1763-1764).

Syst.Zool. 28:1-21.
NITA, S. C. 1971. Establishingthe linkageof differentvariantsofa Romanianchronicle.In Hodson, F. R., D. G. Kendall,and P. Tautu (eds.),

my. Columbia University Press, New York, 247


PP.
SNEATH, P. H. A., AND R. R. SOKAL. 1973. Nu-

W. H. FreemanCo., San Franmericaltaxonomy.


cisco.

This content downloaded from 142.51.1.212 on Mon, 21 Jul 2014 17:05:11 PM


All use subject to JSTOR Terms and Conditions

440

SYSTEMATIC

ZOOLOGY

VOL.

28

WILEY, E. 0. 1977. Are monotypicgenerapolyphyletic?:a responseto NormanPlatnick.Syst.


theory,and fantasyin human paleontology.
Zool. 26:352-354.
AmericanScientist65:204-211.
species conTUOMIKOSKI,R. 1967. Noteson someprinciplesof WILEY, E. 0. 1978. The evolutionary
Ann.Ent.Fenn.33:137cept reconsidered.Syst.Zool. 27:17-26.
systematics.
phylogenetic
WILEY, E. 0. 1979. An annotatedLinnaean hier147.
WILEY, E. 0. 1975. Karl R. Popper,systematics, archy,withcommentson naturaltaxa and competingsystems.Syst.Zool. 28:308-337.
and classification:
a replytoWalterBockand othtaxonomists.
Syst.Zool. 24:233er evolutionary
243.

TATTEIRSALL,I., AND N. ELDREDGE. 1977. Fact,

This content downloaded from 142.51.1.212 on Mon, 21 Jul 2014 17:05:11 PM


All use subject to JSTOR Terms and Conditions