Download Chance and Natural Selection

Chance and Natural Selection
Author(s): John Beatty
Reviewed work(s):
Source: Philosophy of Science, Vol. 51, No. 2 (Jun., 1984), pp. 183-211
Published by: The University of Chicago Press on behalf of the Philosophy of Science Association
Stable URL: http://www.jstor.org/stable/187421 .
Accessed: 04/01/2013 16:36
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 and Philosophy of Science Association are collaborating with JSTOR to
digitize, preserve and extend access to Philosophy of Science.
http://www.jstor.org
This content downloaded on Fri, 4 Jan 2013 16:36:51 PM
All use subject to JSTOR Terms and Conditions
Philosophy of
Science
June, 1984
CHANCE AND NATURAL SELECTION*
JOHN BEATTYt
Department of Philosophy
Arizona State University
Amongthe liveliest disputesin evolutionarybiology today are disputesconcerningthe role of chancein evolution-more specifically,disputesconcerning
the relativeevolutionaryimportanceof naturalselectionvs. so-called "random
drift".Thefollowingdiscussionis an attemptto sortout some of thebroadissues
involvedin those disputes.In the first half of this paper,I try to explain the
differencesbetweenevolutionby naturalselectionandevolutionby randomdrift.
On some commonconstrualsof "naturalselection",those two modes of evolutionare completelyindistinguishable.Even on a properconstrualof "natural
selection",it is difficultto distinguishbetweenthe "improbable
resultsof natural
selection"and evolutionby randomdrift.
In the second half of this paper,I discuss the varietyof positionstaken by
evolutionistswith respectto the evolutionaryimportanceof randomdrift vs.
naturalselection. I will then considerthe varietyof issues in questionin terms
of a conceptualdistinctionoften used to describethe rise of probabilisticthinking in the sciences. I will argue,in particular,thatwhatis going on hereis not,
as mightappearat firstsight,just anotherdisputeaboutthe desirabilityof "stochastic"vs. "deterministic"
theories.Modernevolutionistsdo notargueso much
aboutwhetherevolutionis stochastic,but abouthow stochasticit is.
*ReceivedJune 1983;revisedFebruary1984.
tThis articlewas writtenduringthe academicyear 1982-1983, while I was a fellow at
the Centerfor Interdisciplinary
Researchat the Universityof Bielefeld, in Bielefeld, West
Germany.I was part of a group researchproject, organizedby Lorenz Kruger,which
studiedthe rise and role of probabilisticthinkingin the sciences since 1800. I am very
gratefulto the staff of the Center,the facultyof the Universityof Bielefeld, and of course
my fellow probabilistsfor theirthoughtsand for theirfriendship.
RobertBrandon,LorenzKruger,Elliott Sober, KennethWaters, and the referees of
Philosophyof Science all helped me to clarify the issues discussedhere. The residual
unclaritydistinguishesmy contributionsfromtheirs. Some of the residualunclaritymust
be attributedto JonathanHodge's and AlexanderRosenberg'scritiquesof the notion of
"fitness"used here. Theirthoughtfulcritiqueshave, I must admit,left me a bit confused
aboutmy position.
Philosophyof Science, 51 (1984) pp. 183-211.
CopyrightC 1984 by the Philosophyof ScienceAssociation.
183
This content downloaded on Fri, 4 Jan 2013 16:36:51 PM
All use subject to JSTOR Terms and Conditions
184
JOHNBEATTY
1. Introduction. CharlesDarwin'saccountof organicform appealedto
chancein a way that did not settle well with his critics. As DarwinunJohnHerhappilyreportedthe opinionof the greatscientist-philosopher,
schel, "I have heard,by a round-aboutchannel, that Herschelsays my
book 'is the law of higgledy-piggledy.'Whatexactly this meansI do not
know, butit is evidentlyvery contemptuous"(Darwinto Lyell, Dec. 12,
1859, in F. Darwin 1887, Vol. 2, p. 37). In time, though, Darwinwas
praisedratherthanscornedfor his appealto chance.Forinstance,looking
C. S.
backat the turnof the century,anothergreatscientist-philosopher,
Peirce, assessedDarwin'scontributionin this regardmore favorably:
The Origin of Species was publishedtowardthe end of the year
1859. The precedingyearssince 1846 had been one of the most productiveseasons-or if extendedso as to cover the greatbook we are
considering,the most productiveperiodof equal length in the entire
historyof science from its beginningsuntil now. [For]the idea that
chance begets order . . . was at that time put into its clearest light.
(Peirce 1893, p. 183)
Sincethe turnof the century,however,andespeciallysince the thirties,
evolutionistshavefurtherappealedto chancein ways thatDarwinhimself
might contemptuouslyhave regardedas higgledy-piggledyviews of nature.Indeed,proponentsof one such appealhave coined the term "nonDarwinianevolution"to distancetheirviews fromhis. Actually,the new
appealsto chancehave been mattersof considerabledispute. And today
those disputesare amongthe liveliest in the alreadylively field of evolutionarybiology.
The following discussionis an attemptto sort out some of the broad
issuesinvolvedin these disputes.The most generalquestionat issue concernsthe relativeevolutionaryimportanceof "randomdrift"vs. natural
selection.But what does that mean?In the first half of this paper(Sections 2-3), I will try to make sense of that question. That will involve
explainingthe sense in which evolution by randomdrift is, properly
to evolutionby naturalselection. Darwindid
speaking,an "alternative"
not conceiveof chanceas anythinglike an alternativeto naturalselection,
butratheras "complimentary"
to naturalselection.Modernevolutionists,
on the otherhand, recognizealternativeas well as complimentaryroles
of chanceandnaturalselection.And yet, on some commonconstrualsof
"naturalselection",it is difficult (if not impossible)even to distinguish
evolutionby randomdrift from evolutionby naturalselection. Thus, in
orderto construeevolutionby randomdrift and evolutionby naturalselectionas properalternatives,the conceptof naturalselectionitself must
first be properlyinterpreted.Unfortunately,though, even on a proper
interpretation
of "naturalselection",it is difficultto distinguishbetween
This content downloaded on Fri, 4 Jan 2013 16:36:51 PM
All use subject to JSTOR Terms and Conditions
CHANCEAND NATURALSELECTION
185
whatI call the "improbable
resultsof naturalselection"andevolutionby
randomdrift.
In the secondhalf of this paper(Section4-5), I discuss the varietyof
positionstakenby evolutionistswith respectto the evolutionaryimportanceof randomdriftvs. naturalselection.I will thenconsiderthe variety
of issues in questionin terms of a conceptualdistinctionoften used to
describethe rise of probabilisticthinkingin the sciences. I will argue, in
particular,thatwhatis going on hereis not, as mightappearat first sight,
just anotherdisputeaboutthe desirabilityof "stochastic"vs. "deterministic"theories.Modemevolutionistsdo not argueso muchaboutwhether
evolutionis stochastic,but abouthow stochasticit is.
2. Chance in Darwinian Evolutionary Theory. Darwin usually invoked "chance"(or "accident")in the contextof discussionsabouthow
variations,the materialsof evolution, arise. His notion of "chancevariation"was especiallyimportantfor the purposeof distinguishinghis own
theoryof evolutionfrom the older "use and disuse"theory (Hodge and
Kohnforthcoming).So a briefdiscussionof the differencesbetweenthose
theoriesmightbe helpful. The paradigmapplicationof the theoryof use
anddisuseis the accountof the evolutionof giraffesfromshorter-necked,
four-footedgrazers.As the accountgoes, the ancestorsof giraffesfound
in whichtheyhadlittleuponwhichto graze,
themselvesin an environment
otherthanleaves of trees. They stretchedtheir necks to reachmore and
moreleaves, andwerephysicallymodifiedin the process.Theiroffspring
inheritedthe modification-i.e., having longer necks as juveniles than
theirparentshad as juveniles. The offspringalso stretchedto reachmore
and more leaves, lengtheningtheir necks even beyond the length inherited. The thirdgenerationinheritedthe furthermodification.And so on,
until all the descendantsof the originalgroup were quite long-necked.
Theimportantpointhereconcernshow the variationsthataid the survival
of theirpossessorsfirstarose-namely, in responseto the survivalneeds
of theirpossessors.Thatis, the chanceof such a variationoccurringwas
increasedby the very fact thatthe organismneededit to survive.
On the Darwinianaccount,we are askedto considerwhat would have
happenedif, amongthe shorter-neckedancestors,some slightly longerneckedoffspringhappened"by chance"to be born.Thosewho happened
to have the slightly longer necks would be able to reach slightly more
food, and hence would slightly outsurviveand outreproducethe others.
Assumingthat neck lengthwas inheritable,then, a slightly greaterproportionof the next generationthan of the previousgenerationwould be
longer-necked.This processalone wouldresultin an ever-increasingfrequencyof the slightlylonger-neckedindividuals.But now considerwhat
would happenif, among those slightly longer-neckedorganisms, some
This content downloaded on Fri, 4 Jan 2013 16:36:51 PM
All use subject to JSTOR Terms and Conditions
JOHNBEATTY
186
slightly even longer-neckedoffspringhappenedby chance to be born.
the others.Thus,
Thesein turnwouldslightlyoutsurviveandoutreproduce
the proportionsof longer-and longer-neckedindividualswould increase
fromgenerationto generation,untilthe presentproportionswere reached.
The importantpointfor now concernshow, accordingto the Darwinian
account,the variationsthataid the survivalandreproductionof theirpossessorsfirst arise. They do not, as on the use-and-disuseaccount, arise
in responseto the survivaland reproductiveneeds of their possessors.
Onthe otherhand,whetheror not they furtherincreasein frequencydoes
dependon whetherthey serve those needs. But the probabilityof a particularvariationoccurringin an individualis not increasedby the fact
that that variationwould promotethe survivaland reproductionof that
individual.In that sense, it is a matterof "chance"that an organism
would be bornwith a variationthatpromotesits survivaland reproduction, though, again, not a matterof chance that that variationfurther
increasesin frequencyin subsequentgenerations.As Darwinthus distinguishedhis accountfrom the use-and-disuseaccount,the latterexplains
the evolutionof adaptationin a way "analogousto a blacksmithhaving
childrenwith strongarms",while the formerrelies on "theotherprinciple
of those childrenwhich chanceproducedwith strongarms,outlivingthe
weakerones" (Darwin 1839, passage 42). As Darwin elsewhere more
eloquentlyexplainedthe notionof chancevariation,
[Evolutionby naturalselection] absolutelydependson what we in
our ignorancecall spontaneousor accidentalvariability.Let an architectbe compelledto buildan edificewith uncutstones, fallen from
a precipice.The shape of each fragmentmay be called accidental.
Yet the shape of each has been determined . . . by events and cir-
cumstances,all of which dependon naturallaws; but there is no relationbetweenthese laws and the purposefor which each fragment
is used by the builder. In the same mannerthe variationsof each
creatureare determinedby fixed and immutablelaws; but these bear
no relationto the living structurewhich is slowly built up through
the power of selection.
.
.
. (Darwin 1887, Vol. 2, p. 236)
Thereare several, more generalnotions of chance in terms of which
one mighttry to understandthe notionof chancevariation.Thereis, for
instance, the naturaltheological notion of chance vs. intelligently intendedoccurrences.There is also the old Aristoteliannotion of chance
as coincidence.Or we mightinvokethe Laplaceannotionof chance, not
in orderto say anythingaboutthe event in questionitself, butjust as an
admissionof ignoranceconcerningthe causalchainof eventsthatresulted
in thatevent.It seemsto me possibleto makea case for construingchance
variationin each of these ways-Darwin sometimesseems to have fa-
This content downloaded on Fri, 4 Jan 2013 16:36:51 PM
All use subject to JSTOR Terms and Conditions
CHANCE AND NATURAL SELECTION
187
voredone, andthen at othertimes anothermoregeneralmeaning.Sometimes he spoke of chance variationsin contrastwith benevolently intended, useful variations.Sometimeshe meant that it was a matterof
coincidencethata usefulvariationshouldarise. And sometimeshe seems
only to have wantedto leave entirelyopen the questionof how variations
arise-thus makingclearthathis theoryof evolutiondid not rely on the
are suguse and disuse accountof thatprocess. All threeinterpretations
gested, for instance, by the quotationabove! (See also Ghiselin 1969,
Schweber1982, and Sheynin 1980 with regardto these more general
conceptsof chance variation.)But perhapsthe most importantthing to
consideris whatis distinguishablyDarwinianaboutthe notionof chance
variation.And thatis, I think,best broughtout simply in contrastto the
use and disuse theoriesof evolution,from which Darwinwantedto distancehimself.
3. Chance in Modern Evolutionary Theory: Distinctions. 3.1 The
disputeconcerningthe relativeevolutionaryimportanceof naturalselection vs. randomdriftraises genuinelynew issues concerningthe role of
chancein evolution.Whatis not at issue is the role of chancein evolution
as Darwinconceivedit (exceptinsofaras he too narrowlyconceivedit).
Modernevolutionistsassume,thatis, thatvariationsareindeed"random"
in the sense intendedby Darwin.As one contemporaryevolutionistputs
it, "Mutationis randomin [the sense] that the chance that a specific
mutation will occur is not affected by how useful that mutation would be"
(Futuyma1979, p. 249).
ButwhileDarwinconsideredthe originof variationsa matterof chance,
he did not considerthe possibilitythattheirevolutionaryfates might also
be a matterof chance. In particular,he attributedtheirevolutionaryfates
to naturalselectionin or againsttheirfavor. Naturalselectiontook over
wherechancevariationleft off: those organismsthatwere by chancebetter equippedto survive and reproduceactually outsurvivedand outreproducedthe organismsthatwere by chanceless well equipped,andthus
advantageousvariationsincreasedin frequencyfrom generationto generation.In Darwin'sscheme,in otherwords,chanceandnaturalselection
were "consecutive"ratherthan "alternative"stages of the evolutionary
process.With the introductionof the conceptof randomdrift, however,
camethe notionthatthe fate of a chancevariationmightitself be a matter
of chance-i.e., the conceptthat chance and naturalselection might be
alternativeratherthanjust consecutivestagesof the evolutionaryprocess.
But what does it meanto attributethe evolutionaryfate of a variation
to naturalselection vs. randomdrift?The distinctionwill be easier to
make if we confine our discussionto the fates of genetic variations.It
will be useful,in otherwords,to thinkof evolutionarychangesas changes
This content downloaded on Fri, 4 Jan 2013 16:36:51 PM
All use subject to JSTOR Terms and Conditions
188
JOHNBEATTY
in the gene andgenotypefrequenciesof populations.(TheAppendixconsists of a review of genetic terminologythat some readersmight find
useful at this point.) Thus, the kinds of evolutionarychangesthat I will
be talkingaboutare changesof the following sort. Of the alleles (genes)
at a particulargeneticlocus of membersof one generationof a particular
population,50% are of type A and 50% are of type a. Of the genotypes
with respectto that locus, 25% are of the type AA, 50% are of the type
Aa, and 25% are of the type aa. Of the alleles at that locus of members
of a later generationof thatpopulation,80% are of type A and 20% are
of type a. And of the genotypeswith respectto that locus of members
of the latergeneration,70% are of the type AA, 20% are of the type Aa,
and 10%are of the type aa.
Thoughdefinitionsof "evolution"in termsof suchgene- andgenotypefrequencychangesare common(e.g., Wilson and Bossert 1971, p. 20),
I recognizethat we actuallyincludemore thanjust that kind of change
underthe term "evolution"(see Brandon1978). This restriction,however, simplifiesthe following discussionenormously.
3.2 I will rely, for the time being, on your intuitionsabout what it
meansto attributegene- and genotype-frequencychanges to naturalselection. Whatdoes it mean to attributethe same to randomdrift?Since
as early as 1932 (Dubininand Romaschoff1932; see also Dobzhansky
1937, p. 129), a popularapproachto the expositionof randomdrift has
been via a classic means of modelling chance processes-namely, the
blinddrawingof beadsfrom an urn. The beads in this case are allelesthe differentalleles are differentcolors, but they are otherwiseindistinguishableby the blindfoldedsamplingagent. One urnof beadsrepresents
one generationof alleles-a finite number,characterizedby particular
allele frequencies.The next generationof alleles is determinedby a blind
drawingof beads from an urn. This second generationof alleles fills a
new urn, blinddrawingsfrom which determinethe frequenciesof alleles
in the thirdgeneration.And so on. The frequenciesof alleles may differ
from urn to urn-generation to generation-as a result of the fact that
beads sampledby blind drawfrequenciesof otherwiseindistinguishable
ings may not be representativeof the frequenciesin the urnsfrom which
the samplesweredrawn.Theprobabilityof drawinga representativesample froma populationof a givenfinitesize is easy to calculate-the smaller
the population,the smallerthatprobability.
Such blind samplingcould take at least two forms in nature.The first
formis whatmightbe called "indiscriminate
parentsampling".By "parent sampling" I mean the process of determining which organisms of one
generation will be parents of the next, and how many offspring each par-
ent will have. This sort of samplingmight be "indiscriminate"in the
This content downloaded on Fri, 4 Jan 2013 16:36:51 PM
All use subject to JSTOR Terms and Conditions
CHANCE AND NATURAL SELECTION
189
sense that any physical differencesbetween the organismsof one generationmightbe irrelevantto differencesin theiroffspringcontributions.
A forestfire, for instance,might so sampleparents-killing some, sparing some-without regardto physical differencesbetween them. Such
samplingis indiscriminatein the same sense in which the usualmodel of
blinddrawingof beadsfrom an urnis indiscriminate-thatis, any physical differences(e.g., color)betweenthe entitiesin questionareirrelevant
to whetheror not they are sampled.
If a populationis so maintainedat a particularfinite size-i.e., by
samplingparentsindiscriminately-itsgene and genotypefrequencieswill
"drift"from generationto generation.The reason is that the genotype
frequenciesof the parents,weightedaccordingto theirreproductivesuccess, may by chancenot be representativeof the genotypefrequenciesof
the parents'generation.Thereis no formof discriminationto ensurethat
the genotypefrequenciesare representative.To the extent that a parent
sampleis unrepresentative,
and to the extentthatthe gene and genotype
frequenciesof the next generationreflectthe appropriately
weightedgene
andgenotypefrequenciesof theirparents,the next generation'sgene and
genotypefrequenciesmay diverge from those of the previous generation-i.e., an evolutionarychangemay occur.
Blind samplingin naturemight also take the form of "indiscriminate
gametesampling".(This paragraphand the next can be skippedwithout
much loss.) By "gametesampling"I mean the process of determining
which of the two genetically different types of gametes produced by a
heterozygoticparent is actually contributed to each of its offspring. This
sort of samplingmight be indiscriminatein the sense that any physical
differencebetweenthe two types of gametesproducedby a heterozygote
mightbe irrelevantto whetherone or the otheris actuallycontributedto
anyparticularoffspring.Accordingto Mendel'sLaw (see the Appendix),
thereis no physicalbasis for a bias in the proportionsof the two genetically differenttypes of gametesproducedby a heterozygote.What we
are now consideringis that there is also no physical basis for a bias in
the proportionsof gametes that are actuallycontributedto a heterozygote's offspring.
If a populationis so maintainedat a particularfinite size-i.e., by
samplingthe gametes of heterozygotesindiscriminately-the gene and
genotypefrequenciesof the populationwill driftfrom generationto generation.For the gene frequenciesof the gametes that are contributedto
a generationof offspringmay by chancenot be representativeof the gene
frequenciesof the parents'generation.Again, there is no form of discriminationto ensurethat they are representative.Thus, indiscriminate
gametesampling,eithertogetherwith indiscriminateparentsampling,or
alone, can result in an evolutionarychange, a so-called "randomdrift"
of gene and genotypefrequencies.
This content downloaded on Fri, 4 Jan 2013 16:36:51 PM
All use subject to JSTOR Terms and Conditions
190
JOHN BEATTY
Some authorsconstruerandomdriftentirelyin termsof indiscriminate
parentsampling(e.g., Sheppard1967, pp. 126-127). More often, randomdriftis construedentirelyin termsof indiscriminategametesampling
(e.g., Wilson and Bossert 1971, p. 83). But these two processesare importantlysimilaragentsof change-both involvingelementsof randomin the same sense in both cases. Accordingto
ness, and "randomness"
this sense of "randomness",
samplingfroma populationis randomwhen
each memberof the populationhas the same chance of being sampled.
Thisis a commonnotionof randomsampling-e.g., accordingto a philosophicalintroductionto probabilitytheory, a samplingproceduregives
randomsamplesif:
1. Eachmemberof the populationhas an equalprobabilityof being
selectedas the first memberof the sample.
2. Each memberof the population,excludingthose selected as previous membersof the sample, has an equal chance of being selected as the nth memberof the sample(Skyrms1966, pp. 146147; see also Hacking1965, pp. 118-132).
Discussionsof randomdrift, pedagogicalor otherwise, generally do
not includediscussionsof the conceptsof "randomness"and "chance".
Rather,evolutionistshave, in general, simply relied on what is an exemplarymodel of randomsamplingin the literatureof probability,statistics, and philosophy-i.e., the blind drawingof beads from an urn.
Evolutionistsare not, in general,inclinedto defend any particularinterpretationof "randomness".
It is enough,for theirpurposes,to say of any
processin naturethatis sufficientlysimilarto theirbead-drawingmodel
of randomdrift,thatevolutionin this case is as mucha matterof "random
sampling"and "chance"as are the exemplarycases of randomsampling
uponwhich probabilists,statisticians,and philosophersrely.
In whatfollows, in comparingrandomdriftto naturalselection, I will
be discussingonly randomdrift via indiscriminateparentsampling,not
via indiscriminategamete sampling.So, in effect, I will be contrasting
naturalselectionof parentsand indiscriminateparentsampling.I could
alsocontrastnaturalselectionof gameteswith indiscriminategametesampling, but thatwill not be necessaryin orderto makethe kinds of points
thatI want to make.
3.3 Naturalselectionis not randomsamplingin the sense of "random"
just discussed.Naturalselection is, as its name suggests, a discriminate
form of sampling-a samplingprocess that discriminates,in particular,
on the basis of fitness differences.The generalissue involvedin the disputesconcerningthe relativeevolutionaryimportanceof randomdriftvs.
naturalselection, then, boils down to the questionof the relativeevolu-
This content downloaded on Fri, 4 Jan 2013 16:36:51 PM
All use subject to JSTOR Terms and Conditions
CHANCE AND NATURAL SELECTION
191
tionary importance of sampling without regard to fitness differences vs.
sampling with regard to fitness differences. But I cannot unpack this issue
without first discussing briefly the notion of fitness. For according to the
most common conception of fitness, the distinction between random drift
and natural selection of the fittest dissolves, and along with it dissolves
the issue of the relative evolutionary importance of the two supposedly
different sorts of processes.
According to the most common definition of "fitness", the fitness of
an organism is a measure of its actual reproductive success. I. M. Lerner's definition of "fitness" is typical in this regard-according to Lerner,
"the organismswho have more offspring are fitter in the Darwinian sense"
(Lerner 1958, p. 10; see also Waddington 1968, p. 19; Mettler and Gregg
1969, p. 93; Crow and Kimura 1970, p. 5; Dobzhansky 1970, pp. 101102; Wilson 1975, p. 585; Grant 1977, p. 66). Along the same lines, the
fitness of a genotype is the actual average offspring contribution of organisms of that type. For instance, to say that organisms of type AA are
fitter than organisms of type aa is to say that organisms of the first type
actually leave a higher average number of offspring.
But if fitness is so construed, then it is not clear what the supposed
distinction between natural selection of the fittest (i.e., sampling on the
basis of fitness differences), and sampling without regard to fitness differences, amounts to. How could parent sampling be anything other than
fitness discriminating, when the organisms that leave the most offspring
are by definition the fittest? On this conception of fitness, all parent sampling processes are fitness discriminating.
Consider, for instance, the proposed bead-drawing models of selection
that are based on this notion of fitness-those models are indistinguishable from the bead-drawing models of random drift. Consider, in particular, the bead game called "Selection", described by Manfred Eigen and
Ruth Schuster in their otherwise thoughtful book, Laws of the Game (Eigen and Schuster 1981, pp. 49-65). The game requires a checkerboard,
the playing squares of which are identified by coordinates, and are occupied by colored beads, one bead per square. To begin, the beads that
cover the board are of at least two colors-in whatever initial frequencies
desired. The roll of two dice (each of which has the appropriatenumber
of sides) picks out a square on the board. A bead of whatever color occupies that squareis used to replace whatever color is on the square picked
out by the next roll of the dice. This is followed by another color-determining roll, followed by another color-replacing roll. And so on, until
eventually one color fills the entire board.
This is supposed to represent selection of the fittest color, where the
fittest color is, as Eigen and Schuster put it, "the one that turns out to
be the winner" (1981, p. 55). This is also, of course, just indiscriminate
This content downloaded on Fri, 4 Jan 2013 16:36:51 PM
All use subject to JSTOR Terms and Conditions
192
JOHN BEATTY
sampling.Indeed, Eigen and Schustersay of their model that "We are
faced with the paradoxthat the selective process producesa winner in
everygamethoughcompetitorsdo not differfromeachotherat all" (1981,
p. 55). Unfortunately,Eigen and Schusterare not just calling this game
"Selection",they actuallyintendto explicatethe concept of Darwinian
selectionin termsof the game. But if this is a properexplicationof selection, then the problemof the relativeevolutionaryimportanceof random driftvs. naturalselectionis a pseudo-problem-thereis no difference betweenthem.
Thereis, however, anotherconceptionof fitness that keeps the issue
alive. To motivatethis conceptionbriefly, considerthe example of two
geneticallyidenticaltwins, one of whomis struckby lightningandkilled,
the otherof whom is spared.As a result, say, the formerleaves no offspring,while the lattergoes on to be a parent.We balk at the consequencesaccordingto our conceptionof fitness. That is, we hesitate to
attributezero fitness to the formertwin and relativelyhigh fitness to his
geneticallyidentical,but seeminglyluckierbrother.
One way of accommodatingthese intuitionsis to identify the fitness
of an organismnot with the actualnumberof offspringit contributes,but
with the number of offspring that it is physically disposed to contribute
(Brandon1978, Mills andBeatty 1979, Sober 1980, Burian1983). Along
the same lines, we can talk aboutthe fitness of a genotypein terms of
the averagenumberof offspringthatpossessorsof thatgenotypearephysically disposedto contribute.
On this conceptionof fitness, the two physicallyidenticaltwins, who
mustbe physicallydisposedto contributethe same numberof offspring,
are equallyfit. Hence we can say of the lightningthat killed one and
sparedthe other, that it did not sample the twins on the basis of their
fitnesses-it was not an agent of naturalselection. It was clearly an indiscriminatesamplingagent-indiscriminate,thatis, withregardto physical fitness differencesbetweenthe organismssampled.
3.4 Thatis perhapsas muchaboutfitness as we reallyneed to discuss
in orderto makemeaningfulthe issue concerningthe relativeevolutionary
importanceof randomdriftvs. naturalselection. But I am afraidthat, as
muchas I wouldlike the distinctionbetweennaturalselectionandrandom
driftto be a clear-cutone, it is not as clear-cutas the precedingdiscussion
suggests.(ElliottSoberandKennethWatershelpedme considerablywith
the following discussion.)In orderto complicatematterssomewhat,let
me elaboratea bit upon the notionof "fitness".
First,it is importantto recognizethatthe fitness of an organismis the
numberof offspringit is physicallydisposedto contributein a particularlyspecifiedenvironment.For instance,the numberof offspringthat a
This content downloaded on Fri, 4 Jan 2013 16:36:51 PM
All use subject to JSTOR Terms and Conditions
CHANCE AND NATURAL SELECTION
193
dark-coloredmoth is disposedto contributeis greaterin a darkenvironmentin which it is effectively camouflagedfrom its predators,thanin a
light environmentin which it is more conspicuous.So all attributionsof
levels of fitness mustbe relativeto particularlyspecifiedenvironments.
The specificationof the environment,in turn, involves specifying a
rangeof circumstances,each weightedaccordingto the likelihoodof its
occurrence,or accordingto the likelihoodof the organism(s)in question
experiencingit. So one changesthe environment,so to speak, by changing eitherthe qualityor the weightingof the componentenvironmental
circumstances.The abilityof the darkmoth to surviveand reproducein
a uniformlydarkenvironmentis presumablydifferentfrom its ability in
a three-fourthsdarkenvironment,and that is presumablydifferentfrom
its abilityin a half-darkenvironment,and so on.
The environmentalcircumstancesrelativeto which fitness values are
ascribedto membersof a populationare, ideally, all the environmental
circumstancesrelevantto determiningdifferencesbetweenthe reproductive successesof those organisms.Some of these factorsdiscriminatebetween the organismson the basis of fitness differencesbetween them.
For instance,the combinationof a darkbackgroundand color-sensitive
birds favors the reproductivesuccess of dark moths over light moths,
because of the difference in their color. Some other factors among the
specifiedenvironmental
circumstancesmay be responsiblefor differences
in reproductivesuccess, butnot in connectionwith any fitnessdifferences
betweenthe organismsin question.Forestfires, for instance,might kill
and sparemothswithoutregardto any fitness differencesbetweenthem.
Not only the formerfactors, but also the latterbelong in the environmentalspecificationsrelativeto which fitness values are ascribed.Dark
mothsmay be able to leave more offspringthanlight moths in environmentsin whichthe backgroundis darkandthe majorpredatorsare color
sensitive,andin whichforestfires arerare.But darkandlight mothsmay
have roughlythe same offspring-contribution
dispositionsin an environment in which forest fires are very frequent,even if the backgroundis
otherwisedarkandthe majorpredatorsare color sensitive. Similarly,the
darkand light moths might have roughlyequal fitnesses in an environment in which theirpredatorsare color insensitive.
At any rate, relative to an environmentalspecificationin terms of
weightedenvironmentalcircumstances,there is a range of numbersof
offspringthat an organismis more or less disposed to contribute.For
instance,in an environmentmadeup of 60%dark-coloredtrees and 40%
light-coloredtrees, a darkmothhas a chanceof visitingmoreor less than
60%darktrees, andmoreor less tharn
40%light trees. Given its physical
structure,it may leave somewhatmoreoffspringif it landson more dark
trees and somewhatfewer if it lands on fewer darktrees.
This content downloaded on Fri, 4 Jan 2013 16:36:51 PM
All use subject to JSTOR Terms and Conditions
194
JOHN BEATTY
So for an organism in a particularly specified environment, there is a
range of possible offspring contributions. And for each number within
the range, there is a greater or lesser ability on the part of the organism
to leave that number. Accordingly, we can talk about fitness distributions
relative to specified environments. For instance, relative to a particularly
specified environment, an organism might have a fitness distribution like
the one in Diagram 1. Here the x-axis represents the range of offspring
contributions, and the y-axis represents the strength of the organism's
ability to leave the corresponding number on the x-axis.
DIAGRAM1
Somewhat alternatively, we can view this distribution as a probability
distributionof offspring contributions, where the x-axis again represents
the range of offspring contributions, and the y-axis the probability that
the organism will contribute the corresponding number on the x-axis. The
probabilities should, however, be interpreted as strengths of the organism's physical ability to contribute various numbers of offspring, in line
with a propensity interpretationof probability (Brandon 1978, Mills and
Beatty 1979). We can also construct a fitness distribution for a genotype.
The range of this distribution would consist of the union of the ranges of
possible offspring contributions of the individual members of the genotype. The height of the distribution at any point along the extended range
would represent the group's average probability of leaving that many offspring.
Now the last thing I want to do in connection with this very general
discussion of random drift, natural selection, and fitness, is to consider
a possible scenario that requires that we give a bit more thought to the
distinction between natural selection of the fittest and the indiscriminate
sampling sources of random drift. How distinguishable are they, after all?
Imagine again the case of the light and dark moths and their color-sensitive predators. Imagine too that they inhabit a forest in which 40% of
the trees have light-colored bark and 60% have dark-colored bark. In this
environment, we would say that the dark-colored moths are fitter, since
the forest provides more camouflage for them than for the light moths.
The fitness distributions of finite numbers of the two types might thus
differ in the way shown in Diagram 2.
This content downloaded on Fri, 4 Jan 2013 16:36:51 PM
All use subject to JSTOR Terms and Conditions
CHANCEAND NATURALSELECTION
195
Dark
Light
DIAGRAM2
But supposethatin one particulargeneration,we find, amongremains
of the mothskilled by predators,a greaterproportionof darkmothsthan
was characteristic
of the populationas a whole, and a smallerproportion
of light moths than was characteristicof the population.Say the actual
averageoffspringcontributionsof the two types of mothsin the previous
generation differed as in Diagram 3. And finally, let us say that we find
the remains so distributedin the areas of dark and light trees that we have
reason to believe that the dark moths were perched on light trees when
attacked,and the light moths on dark trees. In other words, the dark
mothschancedto land on light trees more frequentlythanon darktrees,
even thoughthe frequencyof darktrees was greater.
Xd
XI
Xd = Actualaveragenumberof offspringof darkmoths in E
x= = Actualaveragenumberof offspringof light moths in E
DIAGRAM 3
That's the scenario. The question it raises is this. Is the change in fre-
quencyof genes and genotypesin questiona matterof naturalselection,
or a matterof randomdrift?Thatis, is the changein questionthe result
of samplingdiscriminatelyor indiscriminatelywith regardto fitness differences?It is not easy to maintainthatthe samplingwas entirelyindiscriminatewith regardto differencesin survivaland reproductiveability.
At least it is difficult to maintain that the death by predation of conspicis indiscriminate
sampling,whereas
uously darkmothsin this environment
the deathof conspicuously light mothsin the same environmentis selec-
tion. On the otherhand,it is also difficultto maintainthatselectionalone
This content downloaded on Fri, 4 Jan 2013 16:36:51 PM
All use subject to JSTOR Terms and Conditions
196
JOHN BEATTY
is the basis of the change. At least, it is difficult to maintainthat the
fittestwere selected.
The problemhere, I think, is thatit is difficultto distinguishbetween
random drift on the one hand, and the improbable results of natural se-
lectionon the otherhand. Whereverthereare fitness distributionsassociatedwith differenttypesof organisms,therewill be rangesof outcomes
of naturalselection-some of the outcomeswithin those rangeswill be
moreprobablethanothers,but all of the outcomeswithinthe rangesare
possibleoutcomesof naturalselection. And yet some outcomeswithin a
fitnessdistribution(the outer-lyingoutcomesof a bell-shapedfitness distribution,for instance), are in a sense "less representative"of the offabilitiesof the organismsin question.Considera coinspring-contribution
tossing analogy. "Fiftyheads: fifty tails" and "ninetyheads: ten tails"
arebothpossibleoutcomesof one hundredflips of a fair coin, where by
"faircoin" I meana coin equallydisposedto landheadsup and tails up.
And yet "ninetyheads: ten tails" is in a sense less representativeof the
fairnessof the coin. So too in the case of the light and darkmothsjust
discussed,the actualaverageoffspringcontributionsshown in Diagram
3 are possibleresultsof naturalselection, but not very "representative"
of the fitnesses of the light and dark moths. To the extent that those
outcomesare less representativeof the physicalabilitiesof those organisms to surviveand reproducein the environmentin question, any evolutionarychange that results will be more a matterof randomdrift. In
otherwords,it seems thatwe mustsay of some evolutionarychangesthat
they are to some extent, or in some sense, a matterof naturalselection
and to some extent, or in some sense, a matterof randomdrift. And the
reason (one of the reasons)we must say this is that it is conceptually
difficultto distinguishnaturalselectionfromrandomdrift,especiallywhere
improbable results of natural selection are concerned.
Even given these difficultiesof conceptualanalysis, though, the new
conceptionof the role of chance in evolution clearly goes beyond Darwin's conceptions,and raises questionsthat did not arise for him. In
Darwin'sscheme of things, recall, chance events and naturalselection
were consecutiveratherthan alternativestages of the evolutionaryprocess. Therewas no questionas to which was more importantat a particularstage. But now thatwe have the conceptof randomdrifttakingover
whererandomvariationleaves off, we arefacedwithjust such a question.
Thatis, given chance variations,are furtherchanges in the frequencies
of those variationsmore a matterof chance or more a matterof natural
selection?
4. Chance in Modern Evolutionary Theory: Issues. 4.1 As I suggestedearlier,thereare at least two versionsof this generalquestion.At
This content downloaded on Fri, 4 Jan 2013 16:36:51 PM
All use subject to JSTOR Terms and Conditions
CHANCE AND NATURAL SELECTION
197
issue in the first version is the extent of so-called selectively "neutral"
mutations.The mostlikely candidatesfor this title werediscoveredduring
the molecular-biologyboom of the fifties and sixties. It was discovered
then that differentpermutationsof the sequencesof bases that make up
DNA code for differentaminoacids. But it was also discoveredthatthe
sequenceof bases that codes for a particularamino acid can sometimes
be changedin a certainway such thatthe same aminoacid is produced.
These were called "synonymous"changes, in the informationlanguage
of the field. It is certainlya plausibleenoughsuggestionthatsynonymous
changeshave no effect on the numbersof offspringwhich organismsare
physicallydisposedto contribute.Claimsfor neutralityhave, however,
also been extendedto grosserphenotypicdifferences,from single aminoacid differencesbetweenproteinsof the same functionalfamily, to differencesbetweenbloodtypes(Wright1940),to differencesbetweenbanding
patternson snail shells (Diver 1940). Manysuch changeshave not, upon
firstthought,had any imaginableeffect on the numbersof offspringthat
theirpossessorswere physicallydisposedto contribute.
Samplingamongsuch alternativeswouldbe indiscriminatewith regard
to fitnessdifferences,simplybecausetherewouldbe no such differences.
As JackKing andThomasJukesexpressedthe basic idea in theirclassic
positionpaper,"Naturalselectionis the editor,ratherthanthe composer,
of the genetic message. One thing the editor does not do is to remove
changeswhich it is unableto perceive"(King and Jukes 1969, p. 788).
As a result of this sort of indiscriminatesampling, the frequenciesof
neutralgeneticalternativesmay drift.In the wordsof the self-styled"neutralists",the "fates"of these geneticalternativesare mattersof "chance"
(e.g., Nei 1975, p. 165).
very sophisticated,
Theneutralisttheoryof evolutionis mathematically
thoughsome of its featurescan be discussed informally.The theory is
in the sense that, given the gene
"stochastic"ratherthan "deterministic",
and genotypefrequenciesof one generation,and values for all the other
variablesof the theory(like the size of the population),one can at best
calculatethe probabilitiesof possible gene and genotypefrequenciesin
successivegenerations.It is not possible to predictone specific gene or
genotypefrequencyfor each successive generation.
Diagram4 is an exampleof the sorts of calculationsthat the theory
allows(fromRoughgarden1979, p. 62). In this case, we aretalkingabout
an infinitenumberof populations,each of which has a constantsize of
8 individuals,and each of which starts off with gene frequenciesof
0.5 A and0.5 a-i.e., 8 A alleles and 8 a alleles. This is representedby
the top distribution,a barthatsignifiesthat 100%of the populationshave
8 A genes at time t = 0. Accordingto the theory, over the course of
generations,the distributionof populationsspreadsout with respect to
This content downloaded on Fri, 4 Jan 2013 16:36:51 PM
All use subject to JSTOR Terms and Conditions
198
JOHN BEATTY
1.0
.8
t= 0
.6
i
.2
l
.8
.6
.4
.2
l Il
t=
ll
1
0
.8
t= 4
.6 _t=
4~
.4
0 _
C
L
VI4-----~-4--I-Li
iiiii
I
.8
.o
t= 7
.6
.4
?
.2
0
.8
t= 10
.6
.4.2
1
.8
t= 20
.6
.4
.2
0
.
0
l
1
2
3
4
5
6
7
8
9
10 11 12 13 14 15 16
Numberof A alleles in populationi
DIAGRAM4
the frequency of A allels in each, until finally all of the populations have
either no A alleles, or 100% A alleles (i.e., either 0 or 16 A alleles). We
cannot predict whether or not a particularpopulation will have, say, 50%
A and 50% a alleles after one generation, but we can predict the probability that it will be 50: 50, in terms of the proportion of populations
that will still be 50: 50 after one generation. The probability is about
0.2. Diagram 5 is a somewhat more general way of representing what
Diagram 4 represents. This is the first of such diagrams, from Sewall
Wright's pioneering 1931 work on drift. According to the diagram, the
frequencies of populations with all or none of a particular allele increase
faster, the smaller the population size N.
That is, of course, just a glimpse of the general theory of the random
This content downloaded on Fri, 4 Jan 2013 16:36:51 PM
All use subject to JSTOR Terms and Conditions
199
CHANCE AND NATURAL SELECTION
"Rateof decay"of
variability
1/2N
I
4
0
0
0.25
0.5
0.75
1.0
Allele frequency
DIAGRAM
5
drift of frequencies of selectively neutral genetic alternatives. But even
that glimpse is important for our purposes. It is important, too, that we
consider the sorts of issues that divide investigators in this area. There
are more specific as well as more general issues. The more specific issues
concern whether or not a change in frequency of a specific set of genetic
alternatives can be ascribed to their selective neutrality and the consequent random drift of their frequencies. Complementarily, one might expect that the more general issues concern whether all or no evolutionary
changes-or whether all of a certain kind or none of a certain kind of
evolutionary change-are due to random drift alone. It is important to
recognize, however, that even the most general issues surrounding this
version of the importance of random drift do not boil down to questions
of all or none, but to questions of more or less. I cannot improve on
This content downloaded on Fri, 4 Jan 2013 16:36:51 PM
All use subject to JSTOR Terms and Conditions
200
JOHNBEATTY
DouglasFutuyma'sassessmentof this situation,so I will simply quote
it. (I especiallylike this quotationbecause it is from a textbookin evolutionarybiology, and communicatesto studentsof evolutiona valuable
lessonconcerningthe natureof evolutionarytheorizing,and of the nature
of disputesin thatfield.)
The answerto the question,Is enzymepolymorphismdue to selection
or drift?will, of course,be: Both. But this is reallyno answerat all.
No neutralistwould deny that some few enzymepolymorphismsare
maintainedby selection, and even the staunchestselectionistwould
not deny that some amino acid substitutionsmust have such trivial
effectsthatthey have vanishinglysmalleffects on fitness andso fluctuate in gene frequencyby drift alone. The real question is, What
to each factor . . .? (Futuyma
fractionof the variationis attributable
1979, p. 340)
Similarly,MasatoshiNei, a principlefigure in the neutralistdisputes,
characterizesthe strongestneutralistposition as only "a majorityrule"
(1975, p. 165). As he explains,even the strongestneutralistpositiondoes
not rule out selectionwith regardto some sets of genetic alternatives.It
assertsinsteadthat the majorityof gene-frequencychanges involve selectivelyneutralalternatives.Motoo Kimura,perhapsthe principalneutralist,has emphasizedover andover againthe slackin his relativelyvery
strongposition.As he recentlyreported,along these lines,
. . .in one of my paperson the neutraltheory (Kimura1968), I
wrote: "the recent findings of 'degeneracy'of DNA code, that is,
existence of two or more base tripletscoding for the same amino
acid, seem to suggest that neutralmutationsmay not be as rare as
previouslyconsidered."However,I also addeda note of caution:"It
is importantto note thatprobablynot all synonymousmutationsare
neutral,even if most of them are nearlyso.'"Perhapsthis was a pertinentstatementin lightof the recentfindingthatsynonymouscodons
or unequalfashion (Kimura1981).
are often used in "non-random"
(Kimura1982, p. 8)
4.2 Theproponentsof the secondversionof the importanceof random
driftdiscouragedividingevolutionarychangesinto those due entirelyto
naturalselectionandthose due to randomdriftwith no role for selection.
They encourage,instead, analysisof the evolutionaryeffects of natural
selectionand indiscriminatesamplingacting concurrently.That is, they
ratherencouragedividingevolutionarychangesinto those due predominantlyto naturalselection,those duepredominantlyto randomdrift, and
those in which naturalselection and randomdrift both play important
This content downloaded on Fri, 4 Jan 2013 16:36:51 PM
All use subject to JSTOR Terms and Conditions
CHANCE AND NATURAL SELECTION
201
roles. Consider, for instance, Wright's criticism of the evolutionary perspective of R. A. Fisher and E. B. Ford-that they misconstrued the
importance-of-driftproblem:
They hold that fluctuations of gene frequencies of evolutionary significance must be supposed to be due either wholly to variations in
selection (which they accept) or to accidents of sampling. The antithesis is to be rejected. The fluctuations of some genes are undoubtedly governed largely by violently shifting conditions of selection. But for others in the same populations, accidents of sampling
should be much more important and for still others both may play
significant roles. It is a question of the relative values of certain coefficients (Wright 1948, p. 291, emphases added).
Wright's criticism was twofold. First, he was warning against reducing
the importance-of-driftquestion to an all-or-none neutralism issue-i.e.,
to an issue concerning whether all or no evolutionary changes are due to
randomdrift with no role for selection. This is an issue that I just claimed
was not really an issue with regard to the neutralist version of the importance of drift. And Fisher and Ford also denied being concerned with
it (Fisher and Ford 1950, p. 175). But Wright was also warning against
reducing the importance-of-drift question to even a more-or-less neutralism issue-i.e.,
to an issue concerning whether more or fewer evolutionary changes are due to random drift with no role for selection. Wright
considered that the proper version of the importance of random drift relative to natural selection was one that concerned the concurrent relative
roles of indiscriminate and selective sampling in accounting for evolutionary change with regard to each set of genetic alternatives. As Theodosius Dobzhansky, following Wright, also expressed this version,
An evolutionary change need not be due either to directed or to random processes [for example, either to natural selection or to indiscriminate sampling]; quite probably it is the result of a combination
of both types. The theoretically desirable and rarely achieved aim of
investigation is to quantify the respective contributions of the different factors of gene frequency change, as well as their interactions.
(Dobzhansky 1970, p. 231)
As for the various ways in which selective and indiscriminate sampling
can act concurrently, several possibilities come to mind. The change in
frequencies of A and a genes, and AA, Aa, and aa genotypes over the
course of generations may be due predominantly to natural selection during some generations, and to indiscriminate sampling during others. Indiscriminate and selective sampling are only loosely "concurrent"in this
case. Or the changes in frequencies of those genes and genotypes may
This content downloaded on Fri, 4 Jan 2013 16:36:51 PM
All use subject to JSTOR Terms and Conditions
202
JOHN BEATTY
be due to a combination of discriminate parent sampling and indiscriminate gamete sampling, or vice versa. Or the changes may be due to sampling in some geographical subpopulations by entirely indiscriminate
sampling agents, like forest fires, while the rest of the population is sampled selectively. Or the changes may be due to the sort of more-or-less
selective and more-or-less indiscriminate samplings that I discussed in
Section 3.4.
The general theory that describes the concurrent roles of random drift
and natural selection is, like the neutralist theory, stochastic rather than
deterministic-again in the sense that, given the gene and genotype frequencies of one generation, and values for all the other variables of the
theory (including, again, size of the population, and in this case fitness
values as well), one can at best calculate the probabilities of possible
gene and genotype frequencies in successive generations.
The general theory includes, among many other things, variations on
the U-shaped curves that describe the effects of random drift without selection. For example, Diagrams 6(a) and 6(b), again from Wright's pioneering 1931 work, show the effects of the same variety of selection
pressures in small (6a), and then in large (6b) populations. As in Diagrams 4 and 5, we are dealing with proportions of populations and their
gene frequencies. So all we can predict about a population with the help
of these diagrams is the probability that it will have a particular combination of gene frequencies. Also, as in Diagram 4, all the populations in
question had 50: 50 gene frequencies at time t = 0. Diagrams 6(a) and
I'
O
~
~~
0.5
~
I~S
.
1.1005
s= 2/N'
0
0.5
1.0
0
6(a): Population size N
0.5
6(b): Population size N'
DIAGRAM6
This content downloaded on Fri, 4 Jan 2013 16:36:51 PM
All use subject to JSTOR Terms and Conditions
1.0
=
4N
CHANCE AND NATURAL SELECTION
203
6(b) show the distributions at some much later time. When there is no
selection-represented by the case where the selection coefficient, s, equals
zero-the characteristicU-shaped distributionresults, althoughit "squeezes
out" more populations with intermediate gene frequencies in the case of
the smaller populations. For both the large and small populations, greater
selection results in a greater skewing of the curve-that is, more and
more of those populations have greater and greater frequencies of the
gene whose frequency is represented on the x-axis. But this shows up
more clearly in the case of the larger population, where in the case of
strong selection for the allele in question (i.e., where s = 2/N), almost
all of the populations have between 70% and 100% of that allele; whereas
in the case of weaker selection (i.e., where s = 1/2N), many more of
the populations have less than 70% of the allele in question.
There again you have only a glimpse of a theory, but hopefully a glimpse
that will be of some use. Again, somewhat aside from the theory, there
are more or less specific and more or less general issues that divide investigators in this area. Specific issues with regard to this version of the
importance of drift are those that concern the relative, concurrent importancesof selective and indiscriminatesampling in accountingfor changes
in frequency of a specific set of genetic or genotypic frequencies (e.g.,
Kerr and Wright 1954, Dobzhansky and Pavlovsky 1957, Lamotte 1959).
The general issues are, as in the case of the neutralism version of the
importance of random drift, more or less issues rather than all or none
issues. In other words, the issues do not simply boil down to whether
indiscriminate sampling has a certain relative importance in every evolutionarychange, but whether it has a certain relative importance in many
evolutionary changes. For instance, Wright, who comes as close as anyone to defending a general position on the relative importance of random
drift, nevertheless warns, "There is no theoretical necessity for supposing
that evolution has proceeded in the same way in all groups. In some it
may proceed largely under direct selection pressure following change of
conditions, in other cases it may be determined by random differentiation . . ., with or without selection" (Wright 1940, p. 181). And as Dobzhansky so nicely summed up this version of the importance of random
drift, and the general issues associated with it,
Since evolution as a biogenic process obviously involves an interaction of all of the [agents of evolutionary change, including natural
selection and indiscriminate sampling], the problem of the relative
importanceof the different agents presents itself. For years this problem has been the subject of discussion. The results of this discussion
so far are notoriously inconclusive. . . . One of the possible sources
of the situation may be that a theory which would fit the entire living
This content downloaded on Fri, 4 Jan 2013 16:36:51 PM
All use subject to JSTOR Terms and Conditions
204
JOHN BEATTY
world is in general unattainable, since the evolution of the different
groups may be guided by different agents. To a certain extent this
possibility is undoubtedly correct. In species that occur in great abundance in a fairly continuous area of distribution, the population size
factor is bound to be less importantthan in species that are subdivided
into numerous local colonies each having a small effective breeding
population. [For] organisms whose environment is in the throes of a
cataclysmic change, natural selection and mutation pressure are more
important than [for] organisms living in a relatively constant environment. Nevertheless, one can hardly eschew trying to sketch some
sort of a general picture of evolution. (Dobzhansky 1937, p. 186)
I have tried, so far, to distinguish two versions, and two sets of issues,
concerning the evolutionary importance of random drift. But, of course,
the distinction can be collapsed. As one investigates smaller and smaller
fitness differences, one moves from an investigation of the second version
of the importance of drift, to the first. Whether-or to the extent thatresearchprogramsinto the two versions of the importance of random drift
have maintained separate identities in spite of this sort of conceptual continuity, or whether this continuity has, at times, favored a unified program, is something that I can only say that I would like to know more
about.
5. Chance in Nature. At any rate, given a suitable interpretationof fitness and natural selection, there are two more or less distinguishable versions of the evolutionary importance of indiscriminate vs. selective sampling. Now if we step back from these positions and do not concentrate
so much on the variety of issues involved, we see in the development of
evolutionary biology what we see in the development of so many scientific disciplines in the nineteenth and twentieth centuries, namely the
rise of stochastic theories. We have already discussed, very briefly, the
respects in which evolutionary theories that encompass genetic drift are
stochastic vs. deterministic. It might also be helpful to discuss the stochasticity of evolutionary theory by comparing the stochastic "version"
of evolutionary theory with the effectively deterministic version that is
presented in the opening chapters of most textbooks (prior to the introduction of the stochastic version).
The deterministic version, like the stochastic version, is based on the
so-called "Hardy-WeinbergLaw". According to that principle, in an infinitely large population in which individuals mate randomly with respect
to their geneotypes, and in which there is no selection, mutation, or migration, the gene and genotype frequencies remain at equilibrium. More
specifically, if the only alleles at a locus are A and a, and if we represent
the proportion of A genes by p and the proportion of a genes by q, the
This content downloaded on Fri, 4 Jan 2013 16:36:51 PM
All use subject to JSTOR Terms and Conditions
CHANCE AND NATURAL SELECTION
205
genotypicfrequenciesof the populationwill remainin the equilibrium
frequencies,
p2AA 2pqAa: q2aa.
Now assumefor the momentthatpopulationsize is not a variableof
the theory-that the theoryonly applies to populationsof infinite size.
of gene and geIn an infinitepopulation,any "local"misrepresentations
notypefrequenciesdue to indiscriminatesamplingwouldbalanceout. So
by insistingthat the populationbe infinitelylarge, evolutionby genetic
driftis effectivelyruledout.
Law, given valuesfor p and
Giventhis versionof the Hardy-Weinberg
q in one generation,and given 1) magnitudesof selectionfor or against
the variousgenotypes, 2) rates of mutationwith respectto the various
alleles, 3) magnitudesof preferentialmating between the differentgenotypes,and 4) magnitudesof migrationof the differentgenotypes, we
can calculatethe exact gene and genotypefrequenciesin the succeeding
generations.For instance,if the frequenciesof A and a are p and q respectively,andif the relativeaveragefitnessesof the genotypesAA, Aa,
andaa are 1, 1, and 1 - s respectively,then the changein frequencyof
A (tip) in one generationis,
Ap = (spq2)/(1 - sq2).
Thetheoryof evolution,thusconstrued,wouldbe neatlydeterministic,
but would be grossly incompletewith respectto the many finite populationsthatsupposedlyalso undergoevolutionarychange!To coverthese,
populationsize mustbe consideredas a variable.We can then formulate
the sortsof stochasticextensionsof the theorythatwe have alreadydiscussed. Withoutthem, the theoryof evolutionwould be awfully incomplete.
andquantummechanics,stochastic
As in the cases of thermodynamics
theoriesplay a very importantrole in evolutionarybiology. In spite of
all the disputesconcerningthe evolutionaryimportanceof randomdrift,
no one denies that evolutionarytheory shouldbe stochastic.But just to
say thatevolutionarytheoryis stochastic,and thatis that, is to overlook
an importantrespect in which stochasticitycontinuesto be an issue in
evolutionarybiology. I do not meanthatthe issue has been transformed
into the sort of stochasticity vs. in-principle stochasticity issue charac-
teristicof quantummechanics.I have in mindyet anotherrespectin which
the stochasticityissue in evolutionarybiology is far from being settled.
It is kept alive by the varietyof issues and disputesconcerningthe importanceof randomdrift.In orderto see the connection,it will be useful
to considerthe disputesconcerningthe importanceof randomdriftin the
contextof other, somewhatsimilardisputesin evolutionarybiology.
This content downloaded on Fri, 4 Jan 2013 16:36:51 PM
All use subject to JSTOR Terms and Conditions
206
JOHN BEATTY
Evolutionistsrecognizea varietyof modes of evolution-a varietyof
ways in which the gene and genotypefrequenciesof a populationmay
be changed(basically,any of the ways in which Hardy-Weinbergequilibriummight be upset). There is evolutionvia the genetic mutationof
one allele to another,evolutionvia the migrationof geneticallydifferent
types of organismsinto or out of the population,evolutionvia the preferentialmating of organismswith particulargenotypes, and of course
evolutionvia randomdrift, and evolutionvia naturalselection. Evolutionarytheorydescribesthe separateand concurrentoutcomes of these
variousprocesses,dependingon the values of the relevantvariables.So,
for instance,evolutionarytheorydescribeswhat happensin a large populationwhen thereis no migration,no mutation,andmoderateselection;
what happensin a small populationwhen there is moderatemigration,
moderatemutation,and no selection;etc.
Agreementwithregardto the theoryas describedstill leaves open some
veryimportantquestions-namely, questionsconcerningthe relativeevoof the variousmodesof evolutionin nature-hence,
lutionaryimportances
the magnitudesof the relevantvariablesin nature. It is remarkablehow
many modernevolutionistshave been of the opinion that the work that
"remainsto be done" in evolutionarybiology is just that of measuring
these magnitudes.Timofeef-Ressovsky's1940 assessmentof the workto-be-doneis representativeof this opinion, and has been echoed often
since then. As Timofeefexplainedthe situation,evolutionarytheoristsof
the likes of R. A. Fisher, J. B. S. Haldane,Sergei Tschetverikov,and
Sewall Wrighthad succeededin
. . .showing us the relativeefficacy of variousevolutionaryfactors
underthe differentconditionspossiblewithinthe populations(Wright
1932). It does not, however,tell us anythingaboutthe real conditions
in nature,or the actualempiricalvalues of the coefficients of mutation,selection, or isolation[limitationof populationsize]. It is the
task of the immediatefutureto discover the orderof magnitudeof
the coefficientsin free-livingpopulationsof differentplants and animals; this should form the aim and content of an empiricalpopulation genetics (Buzzati-Traverso,Jucci, and Timofeef-Ressovsky
1938). (Timofeef-Ressovsky1940, p. 104)
And as A. J. Cain expressedthe same positionmuch more recently,
The researchesof R. A. Fisher,J. B. S. Haldane,andSewall Wright
laid the foundationsof the mathematicsof populationgenetics. ...
What they did was to provide a mathematicaltheory covering all
possible contingencies,from which quantitativepredictionsof both
deterministicand stochasticprocesses could be made. That was a
greatgain, but it does not tell us what of all these possibilities are
This content downloaded on Fri, 4 Jan 2013 16:36:51 PM
All use subject to JSTOR Terms and Conditions
CHANCE AND NATURAL SELECTION
207
actually exemplified in the wild-what, in short, are relevant to the
actual process of evolution. (Cain 1979, pp. 599-600)
So evolutionists can agree on theory-on what are the possible modes
of evolution, and how evolution occurs given various possible magnitudes
of the relevant variables-and at the same time disagree vehemently as
to the actual magnitudes of these variables and thus the relative importances of these modes of evolution in nature. Since Darwin, evolutionists
have continually haggled over such matters. Neo-Lamarckians concerned
themselves for a long time with the relative evolutionary importance of
an evolutionary factor that is no longer included in evolutionary theorynamely, the inheritance of acquired characters. Darwin and Moritz Wagner argued about the importance of migration relative to selection (see
Sulloway 1979). In the early twentieth century, William Bateson and W.
F. R. Weldon argued over the importance of mutation size and pressure
relative to selection (see Provine 1971). There are even disputes as to
which is the most important kind of selection. For instance, there was a
long, complicated, and even bitter controversy between Dobzhansky and
H. J. Muller as to the relative evolutionary importance of selection in
favor of heterozygotes vs. selection in favor of homozygotes (see Lewontin 1974). Representative of this interest of evolutionary biologists in
ranking the modes of evolution according to their importance is the title
of a book by A. L. and A. C. Hagedoorn, published in 1921: The Relative
Value of the Processes Causing Evolution.
These are rarely all-or-none issues. The disputants defend the importance of their favorite modes of evolution without ruling the others entirely out of the question. As Stephen Gould and Richard Lewontin characterize such issues, "In natural history, all possible things happen
sometimes; you generally do not support your favoured phenomenon by
declaringrivals impossible in theory" (Gould and Lewontin 1979, p. 151).
The neo-Lamarckians recognized selection, though their neo-Darwinian
opponents did not reciprocate. Darwin did not altogether deny the evolutionary effects of migration, nor did Wagner completely rule out selection. Bateson certainly did not ignore selection, nor did Weldon ignore
mutation. Dobzhansky always admitted cases of selection in favor of
homozygotes, and Muller always admitted cases of selection in favor of
heterozygotes.
The structure of these disputes thus has the important effect of not
rendering entirely fruitless those parts of evolutionary theory that deal
with the supposedly less importantmodes of evolution. So those parts of
the theory are rarely threatened by elimination. These disputes can take
place, in other words, without bringing any part of evolutionary theory
into question, either for reasons of incorrectness or lack of fruitfulness.
The issues concerning the relative evolutionary importance of random
This content downloaded on Fri, 4 Jan 2013 16:36:51 PM
All use subject to JSTOR Terms and Conditions
208
JOHN BEATTY
drift are very similar:What is more importantin accountingfor evolutionarychange,indiscriminatesamplingin finite populations,or discriminatesampling?As has been discussed,these issues are not all-or-none,
but more-or-lessissues. As Gouldcharacterizesthe randomdrift issues,
"Thequestion,as with so manyissues in the complexsciences of natural
history,becomesone of relativefrequency"(Gould1980, p. 122). Thus,
no neutralistdenies selection. No selectionistdenies neutralmutations.
No one who investigatesthe concurrentaction of naturalselection and
randomdriftbelievesthateitherof those factorsis alwaysoverwhelming.
The stochastictheorythatdescribesthe evolutionof neutralalternatives,
and that describesthe concurrenteffects of differentdegrees of indiscriminatesamplingand naturalselection, is thus not at issue.
But preciselybecausethe actualrelativemagnitudesof populationsize
and indiscriminatesamplingon the one hand, and selectionpressureon
the otherhand,are at issue, stochasticityis still an issue in evolutionary
biology. Evolutionarybiologists are contentto have a stochastictheory
of evolution,but are not at all in agreementconcerninghow importanta
role to attributeto chancein accountingfor actualevolutionarychange.
The stochasticityissue in evolutionarybiology is decidedlynot whether
chanceplays a role in evolutionarybiology, but to what extent.
I will close with reference,once more, to the long-standingdispute
betweenWrightand Fisherconcerningthe relativeevolutionaryimportanceof randomdriftvs. naturalselection. In his forthcomingbiography
of Wright,WilliamProvinethoughtfullyanalyzesthe developmentof their
agreementsanddisagreements.As Provinepointsout, WrightandFisher
wereultimatelyable, in publishedworkand correspondence,to workout
differencesconcerningthe mathematicaltheory involved. For instance,
Wrightsucceededin convincingFisherthat the rate of decay of neutral
variabilitywas 1/2N (Diagram5) ratherthan 1/4N as Fisher originally
thought.
As Wrightacknowledgedin his review of Fisher'slandmark,The Genetical Theory of Natural Selection (Fisher 1930), "our mathematical re-
sults on the distributionof gene frequenciesare now in completeagreement, as far as comparable"(Wright1930, p. 352). But as Wrightalso
madeclear, agreementwith regardto mathematicaltheorystill left room
for considerabledisagreement.And the most importantsource of disagreementwas Fisher's assumptionof large populationsizes in nature,
whicheffectivelyruledout the actualimportanceof evolutionby random
drift. Again, this issue was somewhatdetachablefrom the mathematical
theory.As for the real disagreement,as Wrightlaterput it, "Itis a question of the relativevalues of certaincoefficients"(1948, p. 291). Thus,
evolutionistslike WrightandFishercouldagreethatsmallpopulationsize
andchanceareimportantin theory,all the while disagreeingconsiderably
This content downloaded on Fri, 4 Jan 2013 16:36:51 PM
All use subject to JSTOR Terms and Conditions
CHANCE AND NATURAL SELECTION
209
as to theirimportancein nature.It is in this sense that the stochasticity
issues in evolutionarybiology are still far from being resolved.
REFERENCES
Brandon,R. (1978), "AdaptationandEvolutionaryTheory",Studiesin Historyand Philosophy of Science 9: 181-200.
Brandon,R. (1978), "Evolution",Philosophyof Science45: 96-109.
Burian,R. (1983), "Adaptation",in M. Grene (ed.), Dimensionsof Darwinism. Cambridge:CambridgeUniversityPress.
A., Jucci, C. and Timofeef-Ressovsky,N. W. (1938), "Geneticsof
Buzzati-Traverso,
Populations", Ricerca Scientifica, Year 9, Vol. 1.
to GeneralDiscussion",Proceedingsof theRoyalSociety
Cain,A. J. (1979), "Introduction
of London B205: 599-604.
Crow, J. F. and Kimura, M. (1970), An Introduction to Population Genetics Theory. New
York:Harperand Row.
Darwin,C. (1839), "N Notebook",in H. E. Gruberand P. H. Barrett,Darwin on Man.
New York:Dutton, 1974.
Darwin, C. (1859), On the Origin of Species. London: Murray. Cambridge: Harvard Uni-
versityPress(facsimileedition), 1959.
Darwin, C. (1887), The Variation of Animals and Plants under Domestication, 2nd ed.
(2 vols.). New York:Appleton.
Darwin, F. (ed.) (1887), The Life and Letters of Charles Darwin (3 vols.). London: Murray.
Diver, C. (1940), "TheProblemof Closely RelatedSnailsLiving in the Same Area", in
J. S. Huxley (ed.), TheNew Systematics.Oxford:Clarendon.
Dobzhansky, T. (1937), Genetics and the Origin of Species. New York: Columbia Uni-
versityPress.
Dobzhansky, T. (1970), Genetics of the Evolutionary Process. New York: Columbia Uni-
versityPress.
Dobzhansky,T. and Pavlovsky,0. (1957), "An ExperimentalStudy of InteractionBetween GeneticDrift and NaturalSelection",Evolution11: 311-319.
Dubinin,N. P. and Romaschoff,D. D. (1932), "The GeneticStructureof Species and
their Evolution" (in Russian). Biologichesky Zhurnal 1: 52-95.
Eigen, M. and Winkler,R. (1981), Laws of the Game. New York:Knopf.
Fisher, R. A. (1930), The Genetical Theory of Natural Selection. Oxford: Oxford Uni-
versityPress.
Fisher,R. A. and Ford, E. B. (1950), "The 'Sewall WrightEffect"', Heredity4: 117119.
Futuyma,D. J. (1979), EvolutionaryBiology. Sunderland,Massachusetts:Sinauer.
Ghiselin, M. T. (1969), The Triumph of the Darwinian Method. Berkeley: University of
CaliforniaPress.
Gould,S. J. (1980), "Isa New andGeneralTheoryof EvolutionEmerging?"Paleobiology
6: 119-130.
Gould, S. J. and Lewontin,R. C. (1979), "The Spandrelsof San Marcoand the Panglossian Paradigm: A Critique of the Adaptationist Programme", Proceedings of the
Royal Society of London B205: 581-598.
Hacking, I. (1965), Logic of Statistical Inference. Cambridge:Cambridge University Press.
Hagedoorn, A. L. and A. C. (1921), The Relative Value of the Processes Causing Evo-
lution.The Hague:Nijoff.
Hodge,J. andKohn,D. (Forthcoming),"TheImmediateOriginsof the Theoryof Natural
Selection", in D. Kohn (ed.), The Darwinian Heritage: A Centennial Retrospect.
Princeton:PrincetonUniversityPress.
Studiesof the Distributionof Gene
Kerr,W. E. and Wright,S. (1954), "Experimental
Frequencies in Very Small Populations of Drosophila melanogaster", Evolution 8:
172-177.
This content downloaded on Fri, 4 Jan 2013 16:36:51 PM
All use subject to JSTOR Terms and Conditions
210
JOHN BEATTY
Kimura, M. (1981), "Possibility of Extensive Neutral Evolution under Stabilizing Selection with Special Reference to Non-Random Usage of Synonymous Codons", Journal
of Molecular Evolution 17: 121-122.
Kimura, M. (1982), "The Neutral Theory as a Basis for Understanding the Mechanism of
Evolution and Variation at the Molecular Level", in M. Kimura (ed.), Molecular
Evolution, Protein Polymorphism, and the Neutral Theory. New York: Springer.
King, J. L. and Jukes, T. H. (1969), "Non-Darwinian Evolution", Science 164: 788-798.
Lamotte, M. (1959), "Polymorphism of Natural Populations of Cepaea nemoralis", Cold
Spring Harbor Symposia on Quantitative Biology 24: 65-84.
Lerner, I. M. (1958), The Genetic Basis of Selection. New York: Wiley.
Lewontin, R. C. (1974), The Genetic Basis of Evolutionary Change. New York: Columbia
University Press.
Mettler, L. E. and Gregg, T. G. (1969), Population Genetics and Evolution. Englewood
Cliffs, New Jersey: Prentice Hall.
Mills, S. K. and Beatty, J. (1979), "A Propensity Interpretation of Fitness", Philosophy
of Science 46: 263-286.
Nei, M. (1975), Molecular Population Genetics and Evolution. New York: Elsevier.
Peirce, C. (1893), "Evolutionary Love", Monist 3: 176-200.
Provine, W. B. (1971), The Origins of Theoretical Population Genetics. Chicago: University of Chicago Press.
Provine, W. B. (Forthcoming), Sewall Wright: Geneticist and Evolutionist. Chicago: University of Chicago Press.
Roughgarden, J. (1979), Theory of Population Genetics and Evolutionary Ecology: An
Introduction. New York: Macmillan.
Schweber, S. (1982), "Aspects of Probabilistic Thought in Great Britain During the Nineteenth Century: Darwin and Maxwell", in A. Shimony and H. Feshback (eds.), Physics and Natural Philosophy. Cambridge: MIT Press.
Sheppard, P. M. (1967), Natural Selection and Heredity. London: Hutchinson.
Sheynin, 0. B. (1980), "On the History of the Statistical Method in Biology", Archive
for History of the Exact Sciences 22: 323-371.
Skyrms, B. (1966), Choice and Chance. Belmont, California: Dickenson.
Sober, E. (1980), "Evolutionary Theory and the Ontological Status of Properties", Philosophical Studies 40: 147-161.
Sulloway, F. J. (1979), "Geographical Isolation in Darwin's Thinking: The Vicissitudes
of a Crucial Idea", Studies in the History of Biology 3: 23-65.
Timofeef-Ressovsky, N. W. (1940), "Mutations and Geographical Variation", in J. S.
Huxley (ed.), The New Systematics. Oxford: Clarendon.
Waddington, C. H. (1968), "The Basic Ideas of Biology", Towards a Theoretical Biology
1. Chicago: Aldine Publishing Company, pp. 1-41.
Wilson, E. 0. and Bossert, W. H. (1971), A Primer of Population Biology. Sunderland,
Massachusetts: Sinauer.
Wilson, E. 0. (1975), Sociobiology. Cambridge: Harvard University Press.
Wright, S. (1930), "The Genetical Theory of Natural Selection: A Review", Journal of
Heredity 21: 349-356.
Wright, S. (1931), "Evolution in Mendelian Populations", Genetics 16: 97-159.
Wright, S. (1932), "The Roles of Mutation, Inbreeding, Crossbreeding, and Selection in
Evolution", Proceedings of the Sixth International Congress of Genetics ]: 356-366.
Wright, S. (1940), "The Statistical Consequences of Mendelian Heredity in Relation to
Speciation", in J. S. Huxley (ed.), The New Systematics. Oxford: Clarendon.
Wright, S. (1948), "On the Roles of Directed and Random Changes in Gene Frequency
in the Genetics of Populations", Evolution 2: 279-294.
APPENDIX
Recallthatthe hereditarymaterialof sexualorganismscomes in pairsof morphologically
similar,or homologous,chromosomes.The genes residelinearlyalong the chromosomes;
This content downloaded on Fri, 4 Jan 2013 16:36:51 PM
All use subject to JSTOR Terms and Conditions
CHANCE AND NATURAL SELECTION
211
A
a
the two genes that lie opposite one another on homologous chromosomes are said to occupy
the same chromosomal location, or "locus", and are said to be "alleles" with respect to
that locus. In the diagram, A and a are alleles at a particular locus. The two alleles at a
locus may be materially different, as in the case of A and a (as the different symbols are
supposed to represent), or they may be materially identical. Different combinations of
alleles may result in alternative states of the same general character, like blue eye color
vs. brown eye color.
The particular combination of alleles that an organism has is called its genotype. We
can speak of the genotype at a particular locus-for instance, we can speak of the Aa
genotype in the case at hand-or we can speak of the entire genotype. If the two alleles
at a locus are different, the organism is said to have a "heterozygotic" genotype with
respect to that locus. Otherwise, it would be said to have a "homozygotic" genotype with
respect to that locus.
Finally, the gametes that a sexual organism produces (i.e., the sperm/pollen or egg cells
that it produces) each appropriatelycontain one allele from each locus. According to "Mendel's Law", there is no physical basis for a bias in the number of gametes containing one
or the other allele of a heterozygotic organism. In other words, in the long run, a heterozygotic organism should produce gametes, half of which contain one allele, and the other
half of which contain the other allele.
This content downloaded on Fri, 4 Jan 2013 16:36:51 PM
All use subject to JSTOR Terms and Conditions