Download Competition Theory, Hypothesis-Testing, and Other Community

The University of Chicago
Competition Theory, Hypothesis-Testing, and Other Community Ecological Buzzwords
Author(s): Daniel Simberloff
Source: The American Naturalist, Vol. 122, No. 5, A Round Table on Research in Ecology and
Evolutionary Biology (Nov., 1983), pp. 626-635
Published by: The University of Chicago Press for The American Society of Naturalists
Stable URL: http://www.jstor.org/stable/2460845 .
Accessed: 11/12/2014 15:31
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 [email protected].
.
The University of Chicago Press, The American Society of Naturalists, The University of Chicago are
collaborating with JSTOR to digitize, preserve and extend access to The American Naturalist.
http://www.jstor.org
This content downloaded from 169.230.243.252 on Thu, 11 Dec 2014 15:31:41 PM
All use subject to JSTOR Terms and Conditions
Vol. 122, No. 5
The AmericanNaturalist
November 1983
COMPETITION THEORY, HYPOTHESIS-TESTING, AND OTHER
COMMUNITY ECOLOGICAL BUZZWORDS
DANIEL SIMBERLOFF
Departmentof Biological Science, Florida State University,Tallahassee, Florida 32306
SubmittedOctober 12, 1982; Accepted April 19, 1983
Roughgarden's (1983) and Quinn and Dunham's (1983) thoughtfulessays on
how to do communityecological research both fall into two parts: (1) general
discussions of the meritsor lack thereofof various philosophicalstances toward
communityinvestigation;and (2) technicalcriticismsof research,includingmuch
of my own work, guided by philosophicalviews that theyfindpoorly suited to
communityecology. Accordingly,my response will be in two corresponding
sections.
PHILOSOPHY
Roughgardenbelieves thatan investigatorestablishesa factin sciencejust as in
everydaylife, "by buildinga convincingcase forthe fact" (p. 583), throughhis
native abilities, common sense, and experience. In everydaylife, he feels, we
rarelyif ever adhere to formalrules in constructinga convincingcase, so it is
rarelyifever appropriateforscientiststo abide by formalrules. I am sympathetic
to Roughgarden's desire to ignore the entire corpus of epistemologyand the
philosophyof science. Afterall, so much formalwritingby philosophersseems
tedious, contrived,and supercilious."Every philosophy... is liable to degenerate in such a way that its problems become practicallyindistinguishablefrom
pseudoproblems, and its cant, accordingly,practically indistinguishablefrom
meaninglessbabble" (Popper 1963, p. 7). Further,as Roughgardenimplies,we
seem to get along quite well in everydaylifeby usingcommon sense and experience to constructa world view, so why not apply the same approach to our
scientificendeavors and let tendentiousphilosophers argue about how many
angels can dance on thehead of a pin? If onlyforthetimeand mentaleffortsaved,
this is a temptingproposition.
Unfortunately,it is probably an overly simplisticone as well. For example,
using common sense, experience, and native abilities,millionshave builta convincingcase forthe existence of a deity,some of themproceedingvery systematicallyto thisconclusion (e.g., James 1902). Othershave concluded by the same
methodthatno such being exists, while agnosticshave argued,equally commonsensically and (to themselves,at least) cogently,that there is no possibilityof
Am. Nat. 1983. Vol. 122, pp. 626-635.
? 1983 by The Universityof Chicago. 0003-0147/83/2205-0005$02.00.
All rightsreserved.
626
This content downloaded from 169.230.243.252 on Thu, 11 Dec 2014 15:31:41 PM
All use subject to JSTOR Terms and Conditions
COMPETITION THEORY AND TESTING HYPOTHESES
627
correctlyreachingeitherof theothertwo positions.Are scientistsdoomed to such
impasses, or are thereformalproceduresthatpromisea more satisfactoryresolution? At least where experimentis possible, it seems to me that applicationof
Popper's procedure (1963, 1972), clearly stated hypotheses and rigorous attemptedfalsification,is more likely to get us closer to an accurate account of
Platt(1964) providesa nontechnicaldescriptionof
nature,and to do so efficiently.
how thisprocedure,whichhe terms"stronginference,"has spurreda remarkable
series of successes in areas of physicsand molecularbiologyand argues thatit is
the strictformalizationthatis responsible.
Now it is truethatPopper's procedurecould well be partof a commonsensical
constructionof a convincingcase, so the two approaches need not be viewed as
mutuallyexclusive alternatives.It is equally truethatthe search forconfirmatory
and is veryseductive.I suspect
evidence is easier thanthe search forfalsification
that in everyday experience it is by far a more common modus operandi than
and I feelthatin communityecologyitalso holds sway. This
attemptedrefutation,
is the detrimentI see to rejectingformalizationin favorof commonsense. If one
agrees that a procedure is effectiveand logically correct, its formalizationis
simplyan aid to force us continuallyto match our actions and thoughtsto the
procedure.An analogyto probabilitytheoryis apt here. Most people feelthatthey
outcomesin
have an intuitiveor commonsensenotionofthelikelihoodof different
simple but well-framedquestions thatarise in everyday,life. A friendlygame of
blackjack is likelyto be played, if not completelyintuitively,at least withlittle
more system than recalling the preceding few cards played, and with some
attentionto hunches. Yet we are very inclined to rely on formal statistical
procedureswhen questions become more complex, even thoughthe procedures
themselvesrelyon the same principlesthatwe ascribeto commonsense. Ifwe set
out to make our livingby beatingthehouse consistentlyat blackjack,we would be
muchmorelikelyto succeed ifwe rigorouslyemployeda countingprocedureand
not the methodswe use at home. In exactly thisway, formalizationof scientific
procedures in communityecology need not be antitheticto common sense, but
oughtto help us to keep trackof what we have and have not establishedin our
studyof what mustbe a vastlymorecomplicatedentitythanmolecularbiologists
or blackjack players deal with. Also, if one agrees with me that confirmatory
evidence per se is not verycompellingsince one can always findsome (Northrop
in allocat1948; Popper 1963), formalizationoughtto help us to be more efficient
ing our research time.
It is ironic that Roughgardenshould call for adherence to common sense as
opposed to the Popperian procedures that my colleagues and I have advocated
while a numberof persons are contendingthat Popper's ideas have application
well beyond the realms of naturalscience, includingday-to-dayactivity.The art
criticGombrich(1960, 1973) and neurophysiologist
Gregory(1973), forexample,
suggestthatperceptionitselfis a tacitconstructionof a hypothesis,withimaginationand/orexperiencefillingin thosepartsof a perceptthathave notbeen directly
establishedby the senses. Popper himself(1972) notedthispsychologicalvariant
of his refutationprocedure. When a perceptualhypothesisis incorrect,it is an
illusionthatcan mislead us in our everydayactions. Our sensationsof the world
This content downloaded from 169.230.243.252 on Thu, 11 Dec 2014 15:31:41 PM
All use subject to JSTOR Terms and Conditions
628
THE AMERICAN NATURALIST
constantlyprovide data thatcan be used to assess the hypothesis.To the extent
that we ignore conflictingdata and accept those data that tend to confirmour
illusion,we are notthinkingefficiently
and our everydayactionsare morelikelyto
be misguided.There is nothingmysteriousabout this psychologicalmodel; it is
simplymeantto show thatin everydaylifeas in science, "buildinga convincing
case" by common sense and experienceis not a trivial,automaticallyefficacious
matter.And in everydaylifeas in science, applicationof a falsification
procedure
is likelyto help.
Hintikka (1969, 1975) suggests a "possible worlds" model of epistemology,
whereinany quest for knowledge,be it in everydayor scientificaffairs,begins
witha large numberof depictionsof reality(possible worlds), all consistentwith
"facts" that we agree are "established." Any advance consists in eliminating
some possible worlds,therebynarrowingthe scope ofpotentiallyvalid depictions.
Such a model, whose similarityto Popper's prescriptionforscientificresearchis
apparent, seems eminentlyreasonable to me and indicates what constitutesan
efficientprocedure in science as in everyday life: one that invalidates a large
numberof hithertopossible worlds. It also suggeststo me thata major pitfallin
both everyday and scientificreasoningis failureto posit initiallya sufficiently
large universe of possible worlds, so that the real world is not among those
initiallypresent.
Particularlywithrespectto evolutionarycommunityecology theory,I feel that
the Popperianprescriptionshould be moderated,as Lakatos (1970) suggests,into
a "sophisticated falsificationism."Afterall, the historyof science is rifewith
observationsthat originallyseemed anomalous but in the lightof later developmentswere shownto be quite consistentwiththehypothesesthatspawned them.
The complexityof communitiesand difficulty
of manipulatingthemdictatethata
theorynot be immediatelydiscardedbecause a singleobservationseems to refute
it. As Lakatos (1970, p. 179) says, "one must treat budding programmesleniently,"especially while theyseem to be growingand generatingresearchthat
anticipates new facts. But eventuallythere must be a willingnessto confront
theorywithcontradictory
data in a decisive manner;Lakatos (1970, p. 179) allows
only that "a budding research programme . . . should be sheltered for a while
. . . " (myitalics). My complaintis thatcommunitytheoryhas been shelteredfor
longenoughand has littlemoreto show forthisforbearancethana proliferation
of
theory.This is not a novel concern (cf. McFadyen 1975; Smith 1976; Futuyma
1975; Brown 1981; Pielou 1981),and I do notmean to be dogmaticabout it. It may
be, as Roughgardenfeels, thatwe are on thethresholdof dramatic,rigoroustests
ofcommunitytheorythatwillfinallyprovidecommunityecologywithwell-tested,
predictivemaxims.If so, thispast decade and a halfof intensetheoreticalactivity
will have been well spent. If not, a large fractionof a generationof evolutionary
ecologistswill have spentits timemanipulatingmodels thatare practicallyirrelevant to nature.
Having statedmyreasons fordesiringsome sortoffalsification
procedure,even
ifdelayed, I findmuchin Roughgarden'sphilosophicalsectionwithwhichI agree.
Roughgardenis correctto distinguishbetweenmathematicaland biologicalproofs
in that, exactly as Lakatos emphasizes, subsequent observations can always
This content downloaded from 169.230.243.252 on Thu, 11 Dec 2014 15:31:41 PM
All use subject to JSTOR Terms and Conditions
COMPETITION THEORY AND TESTING HYPOTHESES
629
change our interpretation
of a physical observation;but as Lakatos argues, this
condition dictates only patience and leniency, not eternal immunityfrom attemptedfalsification.I also agree withRoughgarden(cf. Simberloff1980) thatthe
variedand complex natureof communitiespreventsa simple"litmustest" forthe
historicoccurrence of coevolution or interspecificcompetition.Popper (1972,
1980) feels that his procedure applies to both singularand historicevents, but
surelyfalsification
is vastlymoredifficult
withoutreplication.This difficulty
in no
way obviates theultimateattemptto refutea theory,however,and italso suggests
monumentalmodestyin assessing the scope and value of a model.
I even agree with Roughgardenand with Quinn and Dunham that we were
incorrectto suggest (Strong et al. 1979) that the hypothesisof no population
interactionshas "logical primacy" as a null hypothesis.I do feel, however,that
such a hypothesisis an apt startingpoint. There is such a plethoraof information
in the populationbiology literatureon the ecological and evolutionaryresponses
of individualspecies to habitatvariationthatit seems reasonable to assume that
the habitat must always affecta population's densitywhetheror not species
interactionsare important.In fact, as an appeal to Occam's razor (admittedlya
psychologicallyattractiveratherthan a logically necessary principle)it seems
appropriateto ask if habitat variationmightnot explain a species' distribution
withoutany need to invoke otherspecies' distributionsat all.
If thiswere so, it would certainlynot demonstratethatspecies interactionsare
not important,nor have my colleagues or I ever suggestedthatit would. But it
shouldforceus to recognizethatour universeof "possible worlds" is largerthan
we had thought.If, on the other hand, we reject this null hypothesis(as my
colleagues and I have oftendone), we have successfullynarrowedthisuniverse.
By contrast,approachingthismatteras a problemof everydaycircumstances,as
Roughgardensuggests, seems oftento have led to no null hypothesisat all, in
which instancewe cannot in principlehave narrowedthe universe. Othertimes,
"common sense" has led to an initialuniversethatexcludes an eminentlyreasonable galaxy ofpossible worlds. Pielou (1981, p. 24) observes thattheoreticians
modelingmultispeciessystemsroutinelyassume thatcompetitionis occurringand
that such models ". . . shackle thought. With depressing frequency, they cause
studentsto assertthatthisor thatprocess must(or cannot)be takingplace, merely
because some model or othersays it must(or cannot). So-called counter-intuitive
models have thedesirableeffectofjoltingthoughtout oftherutscreatedby earlier
models .... ."
I agree withRoughgarden,too, thatit is a distractionto focus moreon testinga
model thanon testinghypothesesabout processes in an actual system.
simplifying
My question to him then is, How many of the pages on evolutionaryecology
theoryin, for example, The American Naturalist, are misguidedlyfocused on
theoryand how many are actually requiredto presentthe model thatmotivates
testsof hypotheses?Roughgardenfeelsthatthemajorculpabilityfortheory'sbad
press rests on the "always naive" and "usually incorrect" set of "popular
impressionsabout 'what theoryactually does say' " (p. 598). I presumethatby
"popular impressions" he refers to interpretationsby us benighted nonmathematiciansand by the nonecologistswho have to determinewhat practical
This content downloaded from 169.230.243.252 on Thu, 11 Dec 2014 15:31:41 PM
All use subject to JSTOR Terms and Conditions
630
THE AMERICAN NATURALIST
lessons ecology yields.I would pointout firstthatthereis some blame attachingto
the modelersthemselves(cf. Levin 1975) and second, thatthe bloated theoretical
literaturehas consequences beyondbeinga chore forecologiststo wade through.
Simplyby persistenceand volume alone, a fieldacquires both a lifeof its own
and at least temporaryrespectabilityamongnonexperts.That ecological modeling
forits own sake is now a recognizeddisciplineis witnessedby the emergenceof
journals (e.g., Ecological Modeling, Theoretical Population Biology) devoted
primarilyor solely to modelingand scarcely at all to whetherthe models correspond to nature. Now, I agree withRoughgardenthattheoreticaland empirical
research need not march in lockstep; however, I would emphasize his qualifier:
"Nonetheless, theoreticalecology is ultimatelyvalued by itsutilityin understandingecological processes in nature" (p. 597). I suspectthata good manyecological
modelingattemptsare not motivatedby thisultimategoal. Pielou (1981) observes
thatmany ecological models seem to have been constructedwithno purpose in
mind,but I believe she overstatesthis case. No biological purpose perhaps, but
some purpose nonetheless.One would surelynot go throughthe troubleof doing
theresearch,writingthepaper, and attendingto thenumeroustediousdetailsthat
ensureitspublicationifone onlywantedto fillup time.Many purposesin addition
to biological ones may be served by doingsuch work: amourpropre,professional
prominenceand its attendantrewards,even such mundanemattersas tenureand
promotioncan all benefitfromdoingthiswork. Such goals are neitherblameworthynor unique to ecological modeling;Merton (1973) makes a strongcase that
they motivateall scientists. The importantpoint is that such factors imparta
tenacityto a disciplinesuch thatit withersslowlyeven when it turnsout to yield
few substantiveresults.
Alas, it is no easier for nonecologiststo assess the validityof an ecological
school thanit is forecologists to judge the relativemeritsof different
approaches
to cosmology or cancer etiology. So long as its practitionershave academic
credentials and are sufficiently
adamant and vocal, a patina of respectability
accrues to a school in the eyes of nonspecialists,even scientists.I would argue
thatthe enhanced statureof ecology in the eyes of othersciences thatRoughgarden perceives (but does not document)is, ifit exists at all, not informedapprobationof ecology's accomplishments.More likely,it would be simplya comfortable
feelingthatsome ecologists are speakingin the mathematicaltermsthatphysical
scientistsunderstand,ratherthanin Latin nomina. "Physics-envy" (Cohen 1971)
is misguided; ecologists' proper goal should be not approbationfromphysical
scientistsbut a firmunderstanding
of naturalprocesses, to thepointwherewe can
predictthe outcome of specifiedecological processes and answer many of the
specificecological questions of directapplicationthatcurrentlybesiege us.
Respectabilityamong nonspecialistshas consequences beyond the enhancementof self-esteem,unfortunately.
In the absence of a clear consensus among
workersin the field,a recommendationsupposedlydictatedby "theory" can be
promoted as policy when "theory" could as well be writtenor construed to
generate exactly the opposite recommendation.For one example, the International Union for Conservation of Nature and Natural Resources (1980) in its
overview statementabout what measures we mustimmediatelytake to stem an
This content downloaded from 169.230.243.252 on Thu, 11 Dec 2014 15:31:41 PM
All use subject to JSTOR Terms and Conditions
COMPETITION THEORY AND TESTING HYPOTHESES
631
imminentdisastrous increase in extinctionrate, states thatrefugedesign criteria
and managementpractices should accord with the equilibriumtheoryof island
biogeography.This is in spite of the factsthat(a) the theoryitselfis increasingly
viewed as not a particularlyfaithful
respresentationof mostnature(Gilbert1981);
(b) the main recommendationadvocated in the name of theory-single large
refugesratherthan clusters of small ones-is not a consequence of the theory
(Simberloffand Abele 1976; Higgs 1981); and (c) mostempiricaldata thatbear on
thematterdo not supporttherecommendation(Simberloff
and Abele 1982). Were
thisrecommendationtakenseriously,it could reallyhinderconservationefforts.I
could easily cite other examples in such practical areas as pest and disease
control,food production,etc., whereecological modelshave been used to support
prominentrecommendationsthat could as well be wrong as right.My point,
however,is not to indicateweaknesses of specifictheoreticalrecommendations.
Rather,I want to suggestthatexaggeratedclaims forecological models can have
consequences in the real world.
Finally, I must demurfromRoughgarden'sassertionthatthe Popperianphilosophical stance is moot and irrelevantto the practiceof scientificresearch. I can
onlyhope thatmyabove discussionof thebenefitsof systematization
and rigorin
reasoningabout empiricalfacts, admittedlyas much an assertionas Roughgarden's advocacy of common sense, is the more cogent.
Nor can I concurwithQuinn and Dunham that,effectiveas Popper's procedure
is in some sciences, it is inappropriatein ecology and evolutionbecause the latter
fieldsare characterizedby multiplecausality. I agree that any naturalhistoric
phenomenonis very likely to have been affectedby a numberof forces, but I
suspect that this complicationonly demands more ingenuityon the part of researchers in framingunambiguous hypotheses. Afterall, molecular biological
processes are also affectedby a varietyof inputs,yet Quinn and Dunham do not
contestPlatt's estimation(1964) of the success of stronginferencein thatfield.
Many of the crowningachievementsof molecular biology were achieved in
spite of complicated inputs, by clever experimentalprocedures that could
unambiguouslytest well-chosenhypotheses.In geneticsit is routineto use mutantsthatcannotmanufacturea particularmoleculein orderto show an effectthat
would otherwisebe masked. Averyet al. (1944) demonstratedthatDNA was the
active agentin bacterialtransformation
by usinga Pneumococcus strainthatdoes
not manufacturea polysaccharidecoat. Benzer (1962) used mutantsof T4 phage,
the rII types,thatcould not grow on strainK of Escherichia coli to demonstrate
manyaspects of genetic structure.Many otherdefinitiveexperimentsin physiology and moleculargeneticsrestedon inhibiting
or repressingagentsthataffected
one or more molecular functionswhile leaving others intact. Analogous techniques should allow ecologiststo testhypothesesin similarlyincisivefashion.For
example, at a stroke,Keith (1963) was able to refutethehypothesisthatsnowshoe
hare cycles are part of an intrinsicpredator-prey
oscillationwhen he foundthat
hare populationsoscillate on AnticostiIsland, in the absence of lynx.Underwood
(1978) caged three grazing gastropod species in a varietyof densities to show
clearly competitiveeffectsof Cellana tramoserica on Bembicium nanum, and
equally clearly the absence of othercompetitiveeffectsamong the species. For
This content downloaded from 169.230.243.252 on Thu, 11 Dec 2014 15:31:41 PM
All use subject to JSTOR Terms and Conditions
632
THE AMERICAN NATURALIST
Underwood's resultas forKeith's the key to a clear resultwas an unequivocally
simplified,by whatevermeans, to
falsifiablehypothesisand a systemsufficiently
allow an unambiguoustest of the hypothesis.
It is not necessary, fora hypothesisto be falsifiable,thatit state thatforce A
and only force A is acting,while the alternativehypothesisposits only force B.
Rather,it is at least as much a matterof statingthe hypothesesand choosingthe
tests in such a way that theiroutcome is unambiguous.It does not make sense
(nor have mycolleagues or I attemptedthis) to hypothesizethatcompetitionand
only competitionis responsiblefora communitypattern,and to pose as alternatives that predation and only predation, weather and only weather, etc., are
task in communityecology to state
uniquelyresponsible.It maybe a verydifficult
informative
hypothesesthatallow unambiguousanswers,but it is probablyworth
theeffort(Strong1980). If the bone of contentionis some claimedhistoricor very
infrequentlyoccurringphenomenonthe resolutionof the matterby Popperian
methodswill be even harder,but I doubtthatthe inductiveor descriptivemethod
advocated by Quinn and Dunham-estimating the probable contributionsof a
numberof a priori"causes"-will provide an easier path. Such a methodseems
to me to resembleRoughgarden'scommonsenseapproach, and would thus suffer
the same shortcomings.As I said above, I could well be wrong,and only the
collectivejudgmentof ecologists over the nextfew years will determineifeither
variantproduces convincingand usefulecological descriptions.
I cannotrefrainfromprotestingabout two pointsoffact.First,nowheredid we
(or Popper or Lakatos, for that matter)argue that scientificprogresswill occur
fasterifone investigatesalternativehypothesesratherthanthe principalthesisof
interest. This view of procedure as an either/orphenomenon erects a false
dichotomy.Our contentionhas been thattheremustbe morethanone hypothesis
(nulland alternatives,ifyou will) such thattheempiricaldata soughtwillarguefor
one over the others: the very proceduredescribed by Platt (1964) formolecular
biology. Second, Quinn and Dunham state, "Popper (1959, but see 1983a, 1983b)
argued explicitlythat estimates of probabilityare unfalsifiable,and thus not
subjectto scientifictestundercriteriaof disproof" (1983, p. 603). Popper (1972, a
of
revision of 1959) states: "I believe that my emphasis upon the irrefutability
probabilistic hypotheses . . . was healthy: it laid bare a problem which had not
been discussed previously .... Yet my reform ... changes the situation entirely.
For this reform . .. amounts to the adoption of a methodological rule . .. which
makes probabilityhypothesesfalsifiable"(p. 191).
TECHNICAL
CRITIQUE
As sympatheticas I am to Roughgarden'sdesire to obviate the philosophyof
science, I cannot agree with much of his technicalcritique. A fairfractionof it
seems to me to exemplifythe shortcomingsof a commonsensicalapproach. For
example, Roughgarden'soutline of coevolutionarytheoryappears (by virtueof
niches' beingto therightor leftofone another)to assume a one-dimensionalniche
space, surelyan unwarrantedrestrictionon the numberof possible worlds. That
one-dimensionalresults need not triviallyand automaticallytranslateinto mul-
This content downloaded from 169.230.243.252 on Thu, 11 Dec 2014 15:31:41 PM
All use subject to JSTOR Terms and Conditions
COMPETITION THEORY AND TESTING HYPOTHESES
633
tidimensionalones is well known(e.g., May 1974). And nowherein his criticismof
Connell's protocol does Roughgardensuggestany conceivable observationthat
would be taken to indicate that exploitationcompetitionwas not importantin a
particularsystem. Then why do any field or laboratoryresearch at all? It is
certainlytruethatthe removalexperimentadvocated by Connell (1980) cannot in
principleexorcise the ghost of competitionpast, but it seems as if no data can,
even though we are not dealing here with a logical proof. At least Connell's
procedurehas the virtueof clarifyinghow naturepresentlyoperates.
Roughgardenagain seems to narrowtheuniverseof possible worldsunwarrantedly when he criticizesour nullhypothesesbecause "No biologicalprocesses are
exhibitedthatproduce the distributionspredictedby [them]" (p. 591). Although
we do not attemptto account foreveryforceactingon a species, norfortheexact
mechanismby whichforcesact, it is clear thatwe are askingifspecies' individual
responses to the physical environmentsufficeto explain their distributions.
Roughgarden similarlyomits likely importantforces from his coevolutionary
model. (Is he saying,e.g., thatweatheris unimportant,
thatthehabitatis homogeneous?) He is also not yeryspecificabout mechanisms(e.g., is the competition
exploitationor interference?).Generallyhe seems to be arguingthata model, to
be useful,mustbe realistic;but he has not noticedthatall models are unrealistic
by virtueof being abstractions(Simberloff1982); thatthe dividingline between
"realistic" and "unrealistic" is subjective (Pielou 1981); and that unrealistic
models are at least as useful as realisticones in advancingour understandingof
natureor generatingmechanistichypotheses(Pielou 1981; Simberloff1982).
In testingnull and alternativehypothesesbased on more or less unrealistic
models, what is importantis that differentdegrees of realism not affectthe
models' performancesin thearea of concern.Roughgardensubsequentlyseems to
accept this view when he avers that most communityecology models aim to
simplify,and that"we fillthe gaps in the simplifieddescriptionwithassumptions
thatmay actuallybe false iftakenliterally,but whichare hoped somehow not to
be too misleading" (p. 594). I do not see that Roughgarden's revolutionary
models depict the "real processes thatbringspecies to islands" any more than
ours do, yet he feels we mustmodel these forcesto determinewhat distributions
of data are expected in the absence of interspecific
competition.Whymusthe not
also model themto determinewhat distributions
are expected in the presence of
interspecificcompetition?
In no sense have my colleagues and I ever said the world is withoutstructure.
Nor have we intendedthata patternmustbe manifestbeforeone is licensed to
pursue research into what may have caused the pattern.Our sole concern has
been thatthe fullrangeof forcespotentiallygeneratingthe patternbe recognized
so thatthe search forits causes not be undulycircumscribedinitially.Certainlya
process may be studiedfruitfully
beforeits consequences are well understood,so
long as the process does not imperceptibly
become establishedas fact simplyby
virtueof its studyor its mathematicaltidiness.Roughgardendoes not recognize
thisas a problem: "Sometimes it is obvious thata process is occurring"(p. 592).
No doubt as obvious as is the existence or nonexistenceof God.
The "serious technical" criticismsof our work to whichRoughgardenalludes
This content downloaded from 169.230.243.252 on Thu, 11 Dec 2014 15:31:41 PM
All use subject to JSTOR Terms and Conditions
634
THE AMERICAN NATURALIST
are fullydiscussed by Strongand Simberloff(1981), Simberloff(1982), and Connor and Simberloff(1983), who have shown themto be eitherincorrector of no
consequence withrespect to our conclusions. Nor has Roughgardenstatistically
substantiatedhis claim thatour testsare biased towardtypeII errors.I have yetto
see his analysis of bias or power of any of his models or ours. Related complaints
by Quinn and Dunham are similarlydismissedin the above rebuttals.In particular, the observationthata source pool may alreadybe structuredby competition
was anticipatedby Connor and Simberloff(1979) and Stronget al. (1979), and is
not germaneto the hypothesestheywere testing,nordoes it generatecircularity.
Its sole potentiallydebilitatingeffectis to lower statisticalpower. Withrespectto
presence and absence of species combinationson islands, Connorand Simberloff
(1979) were explicitthatthe occurrencefrequenciesof the different
species might
well be partlydeterminedby interactions,but thatthe question asked was ifthe
co-occurrencedistribution
patternsofthedifferent
species, above and beyondthe
individualfrequencies,demanded an interactiveexplanation: ". . . unless one is
willingto ascribe to competitionthe facts thatislands have different
numbersof
species and that species are found on differentnumbers of islands, the New
Hebrides data still argue heavily against the claim that competitiondetermines
of species on islands" (p. 1,136,myitalics). As for
mostaspects ofthedistribution
the contentionthatthe bill sizes in a species pool may already be structuredby
competition,Stronget al. (1979) did not disputethis.All theydid was to perform
the classical statisticalprocedure of randomization(Bradley 1968; Hendrickson
in bill1981) to see ifobserved subsets of birdson islands were typicallydifferent
size characteristicsfromthose we would have observed ifspecies and races were
independentlyand randomlyplaced on islands. Again, we did not contend that
sizes of birdsin the species pool were not affectedby competition;we asked only
ifthe sizes of whicheverraces coexist, above and beyondthe distribution
of sizes
in the pool, additionallydemand an interactiveexplanation.
LITERATURE CITED
Avery,0. T., C. M. MacLeod, and M. McCarty. 1944.Studieson thechemicalnatureofthesubstance
of pneumococcal types. Inductionof transformation
inducingtransformation
by a desoxyribonucleicacid fractionisolated fromPneumococcus type III. J. Exp. Med. 79:137-158.
Benzer, S. 1962. The finestructureof the gene. Sci. Am. 206:70-84.
Bradley,J. V. 1968. Distribution-free
statisticaltests. Prentice-Hall,Englewood Cliffs,N.J.
Brown, J. H. 1981. Two decades of homage to Santa Rosalia: towarda general theoryof diversity.
Am. Zool. 21:877-888.
Cohen, J. E. 1971. Mathematicsas metaphor.Science 172:674-675.
Connell,J. H. 1980. Diversityand coevolutionof competitors,or theghostof competitionpast. Oikos
35:131-138.
Connor,E. F., and D. Simberloff.1979.The assemblyof species communities:chance or competition?
Ecology 60:1132-1140.
1983. Neutralmodels of species' co-occurrencepatterns.In D. R. Strong,Jr.,D. Simberloff,
L. G. Abele, and A. B. Thistle, eds. Ecological communities:conceptual issues and the
evidence, PrincetonUniversityPress, Princeton,N.J. (in press).
Futuyma,D. J. 1975. Review of M. L. Cody's Competitionand the structureof bird communities.
Q. Rev. Biol. 50:217.
This content downloaded from 169.230.243.252 on Thu, 11 Dec 2014 15:31:41 PM
All use subject to JSTOR Terms and Conditions
COMPETITION THEORY AND TESTING HYPOTHESES
635
Gilbert,F. S. 1980. The equilibriumtheoryof island biogeography:factor fiction?1. Biogeogr.7:209235.
Gombrich, E. H. 1960. Art and illusion: a study in the psychologyof pictorial representation.
Pantheon, New York.
1973. Illusion and art. Pages 193-243 in R. L. Gregoryand E. H. Gombrich,eds. Illusion in
natureand art. Scribners, New York.
Gregory,R. L. 1973. The confoundedeye. Pages 49-95 in R. L. Gregoryand E. H. Gombrich,eds.
Illusion in natureand art. Scribner's, New York.
Hendrickson,J.A. 1981.Community-wide
characterdisplacementreexamined.Evolution35:794-809.
Higgs, A. J. 1981. Island biogeographytheoryand naturereservedesign. J. Biogeogr. 8:1 17-124.
Hintikka,J. 1969. Models formodalities.D. Reidel, Dordrecht,The Netherlands.
D. Reidel, Dordrecht,The Netherlands.
. 1975. The intentionsof intentionality.
InternationalUnion for Conservationof Nature and Natural Resources. 1980. World Conservation
Strategy.IUCN-UNEP-WWF.
James,W. 1902. The varietiesof religiousexperience. Longman's, Green, New York.
Keith, L. B. 1963. Wildlife'sten-yearcycle. Universityof WisconsinPress, Madison.
Lakatos, I. 1970. Falsificationand the methodologyof scientificresearchprograms.Pages 91-196 in
I. Lakatos and A. Musgrave, eds. Criticismand the growthof knowledge. Cambridge
UniversityPress, Cambridge.
Levin, S. A. 1975. On the care and use of mathematicalmodels. Am. Nat. 109:785-786.
MacFadyen, A. 1975. Some thoughtson the behaviourof ecologists. J. Anim. Ecol. 44:351-363.
May, R. M. 1974. Stabilityand complexityin model ecosystems. 2d ed. PrincetonUniversityPress,
Princeton,N.J.
Merton,R. K. 1973. The sociology of science. Theoreticaland empiricalinvestigations.Universityof
Chicago Press, Chicago.
Northrop,F. S. C. 1948. The logic of the sciences and the humanities.Macmillan,New York.
Pielou, E. C. 1981. The usefulnessof ecological models: a stock-taking.Q. Rev. Biol. 56:17-31.
Platt,J. R. 1964. Stronginference.Science 146:347-353.
Popper, K. R. 1959. The logic of scientificdiscovery. Hutchinson,London.
1963. Conjectures and refutations:the growthof scientificknowledge.Harper & Row, New
York.
1972. The logic of scientificdiscovery. 3d ed. Hutchinson,London.
1980. Letter to editor. New Sci. 87:611.
Quinn, J. F., and A. E. Dunham. 1983. On hypothesistestingin ecology and evolution. Am. Nat.
122:602-617.
Roughgarden,J. 1983. Competitionand theoryin communityecology. Am. Nat. 122:583-601.
D. 1980. A succession of paradigmsin ecology: essentialismto materialismand probabiSimberloff,
lism. Synthese43:3-39.
1982. Propertiesof coexisting bird species in two archipelagoes. In D. R. Strong,Jr., D.
Simberloff,L. G. Abele, and A. B. Thistle,eds. Ecological communities:conceptual issues
and the evidence. PrincetonUniversityPress, Princeton,N.J. (in press).
Simberloff,D. S., and L. G. Abele. 1976. Island biogeographytheoryand conservationpractice.
Science 191:285-286.
Am. Nat.
1982. Refuge design and island biogeographictheory:effectsof fragmentation.
120:41-50.
Science 192:546.
Smith,F. E. 1976. Ecology: progressand self-criticism.
Strong,D. R., Jr. 1980. Null hypothesesin ecology. Synthese43:271-285.
Strong, D. R., Jr., and D. Simberloff.1981. Strainingat gnats and swallowing ratios: character
displacement.Evolution 35:810-812.
character
Strong,D. R., Jr., L. A. Szyska, and D. S. Simberloff.1979. Tests of community-wide
displacementagainst null hypotheses.Evolution 33:897-913.
Underwood,A. J. 1978. An experimentalevaluationof competitionbetweenthreespecies ofintertidal
prosobranchgastropods. Oecologia 33:185-202.
This content downloaded from 169.230.243.252 on Thu, 11 Dec 2014 15:31:41 PM
All use subject to JSTOR Terms and Conditions