Stearns - 1976 - Life-History Tactics A Review of The Ideas

Life-History Tactics: A Review of the Ideas
Author(s): Stephen C. Stearns

Source: The Quarterly Review of Biology, Vol. 51, No. 1 (Mar., 1976), pp. 3-47
Published by: The University of Chicago Press
Stable URL: http://www.jstor.org/stable/2825234 .
Accessed: 23/07/2013 14:32
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .
http://www.jstor.org/page/info/about/policies/terms.jsp
.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact support@jstor.org.
The University of Chicago Press is collaborating with JSTOR to digitize, preserve and extend access to The
Quarterly Review of Biology.
http://www.jstor.org
This content downloaded from 129.173.72.87 on Tue, 23 Jul 2013 14:32:41 PM

All use subject to JSTOR Terms and Conditions
VOL.51, NO. 1 March 1976
THE QUARTERLY REVIEW

of B JOLOGY
LIFE-HISTORY TACTICS: A REVIEW OF THE IDEAS
By STEPHEN C. STEARNS*
ofAnimalResourceEcology,University
Institute Vancouver,B.C.,
ofBritishColutmbia,
Canada V6T I W5
ABSTRCT
traitsare
This revieworganizesideas on theevolutionof life histories.The keylife-history
broodsize,sizeofyoung, theagedistribution ofreproductive effort,theinteraction ofreproductive
effortwithadultmortality, and thevariation in thesetraitsamongan individual's progeny.
Thegeneral theoreticalproblem is topredictwhich combinations oftraitswillevolvein organisms
livingin specified circumstances.
Firstconsider singletraits.Theorists havemadethefollowing predictions: (1) Vhereadult
exceedsjuvenile mortality, theorganism should reproduce onlyoncein itslifetime. Where juvenile
exceedsadultmortality, theorganism shouldreproduce severaltimes.(2) Brood sizeshould maximize
thenumber ofyoung surviving tomaturity, summed overthelifetime oftheparent. Butwhen
optimum brood-sizevariesunpredictably in time,smaller broodsshouldbefavored becausethey
decreasethechances oftotalfailureona givenattempt. (3) In expanding populations, selection
shouldminimize age at maturity. In stablepopulations, whenreproductive success depends on
size,age,or socialstatus, or whenadultexceeds juvenilemortality, thenmaturation shouldbe
delayed, as it shouldbe in declining populations. (4) Youngshouldincrease in sizeat birth
withincreased predation risk,anddecrease in .sizewithincreased resource availability.
Theorists havealso predicted thatonlyparticular combinations of traitsshouldoccurin
specified circumstances. (5) In growing populations, age at maturity shouldbe minimized,
reproductive effortconcentrated earlyin life,and broodsizeincreased. (6) Oneviewholdsthat
in stableenvironments, latematurity, multiple broods, a few,largeyoung, parental care,and
smallreproductive efforts shouldbe favored (K-selection).In fluctuating environments, early
maturity, manysmallyoung, reduced parental care,and largereproductive efforts shouldbe
favored (r-selection).(7) But another viewholdsthatwhenjuvenilemortality more
fluctuates
thanadultmortality, thetraitsassociated withstableand fluctuating environments shouldbe
reversed.
Weneedexperiments that theassumptions
test andpredictions reviewed here,more comprehensive
theorythatmakesmorereadily falsifiablepredictions, and examination ofdifferent definitions
offitness.
address:Dept.of Zoology,University
*Current of California, 94720
California
Berkeley,

4 THE QUARTERL Y REVIEW OF BIOLOGY [VOLUME 51
GLOSSARY
Bet-hedging:
An alternativeexplanatuin for the trends
explained by r-selectionand K-selection,based on 1 = f e-r I.xbxdx.
the analysisof fluctuationsin adult and juvenile
mortalities.
r-selection:
Eithera labelappliedto thecombination
b.: Average numberof young born to a femaleaged of earlymaturity,many,smallyoung,a shortlife,
x. and largereproductive (as in "r-selected"),
efforts
Costof reproduction:The marginal increase in adult or an impliedexplanationof whythosetraitsare
mortalitybetween time t and time t + 1 caused foundtogether. Whichmeaningis beingused is
by the decision to commita certain proportionof rarelyspecified.
available resourcesto reproductionat time t. effort:
Reproductive The proportionof resources di-
Fitness:Somethingeveryoneunderstandsbut no one vertedto reproduction,
summedover the time
can define precisely.
in question.
interval
Iteroparity:
Repeat reproduction,givingbirthseveral Reproductive
value:The presentvalueof futureoff-
timesin a lifetime.The perennial habit. springimpliedby thecurrentpopulationgrowth
rate,or, theaveragenumberof younga female
K: Equilibrium density,the number of individuals (aged x) can expectto have overthe restof her
presentunder constantconditionswhen population life,discountedbackto thepresent.Usuallydenot-
fluctuationshave died down. May be lower than ed V. or v./vo.
saturationdensity,the numberof individualsthat
the environmentcould conceivablymaintain. The big-bangreproductive
Semnelparity: pattern;giv-
ingbirthonlyonce and committingsuicidein the
K-selection:
Either a labelapplied to the combination process.Whendone in the firstyearof life,the
of late maturity,few,large young,a long life,and annualhabit.
small reproductiveefforts(as in "K-selected"),or
Stable age distribution:
The unique distributionof
an impliedexplanation of whythose traitsare found
organismsamongage classesthata populationwill
together.Which meaning is being used is rarely
reachon a fixedschedulerelatingfecundityand
specified.
mortality
to age.
I.: Probabilityof survivingto age x. Tactic:A setofcoadaptedtraitsdesigned,bynatural
r: The intrinsicrateof naturalincreaseunder a stable selection,to solveparticularecologicalproblems.
age distribution,as defined by Lotka's equation, A complexadaptation.
A
I NTRODUCTION
Scope and Purpose tangled thicketthat blocked furtherprogress.

To date, few workershave shown any inclina-
T THE centerof the evolutionary tion to start the pruning and cutting process
paradigm lies the definition of of experimentationthatwill shape thistangled
fitness. Fit organisms are defined thicketof ideas into clear explanations.
as those better represented in future Throughout this essay, I shall refer to this
generations than their relatively area of biology as the theory of life-history
unfit competitors. Evolution places heavy em- tactics,or life-historytactics,for short. The
phasis on reproduction, for the way an organism phenomena studied naturallyelicit a research
reproduces affects profoundly its contribution viewpointthatcombinesthestudyof reproduc-
to future generations. Reproduction is not the tion, growth, and genetics in an ecological
only component of fitness, but it is certainly setting to produce hypotheses concerning
an important one. evolutionarychanges. I define a tacticas a set
Between 1954 and 1973, intense theoretical of coadapted traitsdesigned, by natural selec-
activityproduced a welter of hypotheses relating tion,to solve particularecologicalproblems.For
observed trends in reproduction to various example, the life-history tacticof a population
causes. The ideas put forth a posteriori as of fish living in a lake might consist of its
explanations of field observations rapidly out- age-specificdistributionof growthrates,repro-
stripped their empirical base and grew into a ductive efforts,progeny produced and their

MARCH 1976] LIFE HISTORY TACTICS 5
size, and the genetic systemunderlyingthose Such traitstake a long time to evolve. Other
traits. traitsvary,and can be fine-tunedby selection
Anotherreviewoflife-historytacticsis waiting to adapt individualsbelongingto differentpop-
to be written.It would take the hypotheses ulations of a single species to varyingcircum-
reviewed in this paper, or their survivingde- stances. Such traitscan evolve rapidly.For the
scendants, and test them against the massive basic traits,the best typeof explanation we can
body of evidence accumulated on Drosophila, generate argues from broad comparisons
Tribolium,cultivatedplants,domestic livestock, among groups of taxa that have encountered
and exploited fish populations. I have not differentconditionsover long periods of evolu-
attemptedto summarize that body of data in tionarytime. This is the comparativeapproach
this paper, which has grown bulky withoutit, toan explanationof adaptation.But the variable
but I hope someone will. traitsare open to explanations that proceed
Thus this paper has a restricted goal: to through deductive models to the test of risky,
reviewmost of the currentideas about life-his- potentially falsifiable hypotheses in selection
torytactics,and some of the data bearing on experiments. The immediate evidence from
them. I seek not to evaluate the ideas against intraspecific selection experiments is much
theevidence,but to bringorder to a fieldwhose strongerthan correlativeevidence from inter-
natural complexityhas been compounded by specific comparisons. For we can tightlycon-
a proliferationof viewpoints. I have tried to trol the situationsencountered by the animals,
writethepaper thatI wanted to read, but could and since the animals all belong to a single
not find,when I started my own study of life species, the rest of the biology of the animals
histories.I have addressed myselfprimarilyto is controlled.
those seeking an introductionto the field. The Conclusionsreached fromexperimentaltests
reviewincludes papers that come to my atten- of riskyhypothesescan advance our knowledge
tion prior to May, 1974. by pointing out weaknesses in our models. It
is much more difficultto falsifya prediction
The Problem made at the interspecificlevel, where one can
always argue that confoundingeffectsobscure
How can we predict what measurable traits the main trend. But generalizationsmade from
populations will evolve in any particular real experimentscarried out on a single species are
situation?Three thingsstrikeme as important. subject to question, perhaps more so than
First,the answer depends to a large extent on generalizations based on correlations drawn
the quality of informationwe can get. Thus from many species. The more certain we are
we should develop hypotheseswithour capaci- that we know what is causing an evolutionary
tiesformeasurementin mind. Second, we must trend, the less certain we are that the cause
agree on whatwe mean by "evolutionarysitua- is general. And the more certain we are that
tion" and "measurable traits."In the language a trend is in fact general, the less certain we
of systemsanalysis,we must define concretely can be about itscause or causes. In livingwithin
the state variables and the systemparameters. these limits,I feel stronglythat we must do
In thisreview,I willconsidersome reproductive experimentswherever possible, and base our
statevariables(biologicaltraits)and some system argumentson the best evidence available when-
parameters(environmentalconditions).Others ever experimentsare not possible, rather than
have yetto be defined.Third, in order to discuss resort to appealing to the beauty of untested
these issues, we must agree on the general speculations.
nature of the problems facingevolving organ- This paper reviews both comparative and
isms.Williams(1966a) put it well: "The central experimentalexplanationsof life-history tactics.
biological problem is not survivalas such, but Our knowledge will advance most rapidly
design for survival." through the interactionof theoryand experi-
In dealingwiththeseproblems,itis important ment at the intraspecificlevel. Many traitsmay
to keep certaindistinctionsin mind. Some traits appear to be fundamental,incapable of rapid
are so basic to the biology of a species that evolution,and subject only to the comparative
theydo not vary withinpopulations. All spar- approach. But we should adopt the attitude
rows have wings, and all men have two legs. that any variable traitcan evolve rapidly and

6 THE QUARTERLY REVIEW OF BIOLOGY [VOLUME 51
is subject to experimentation,in order to mini- numbers of individuals, as defined in the

mize the number of manipulable traitsthat go Lotka-Volterraequations. We can define r by
unmanipulated.Nevertheless,our appreciation a mathematicalrelationshipamong physiolog-
of a species's biologywould be incompleteand ical parameters (see below, p. 7 ff.), but it is
unsatisfyingwithout complementaryexplana- probablybest measured for populations in the
tions pitched at the interspecificlevel. field as the rate of increase in numbers at
In order to arrive at testablepredictions,we low population density. The two definitions
commonly assume that selection will push a have no necessarylogical relationship.It is likely
population towards a combination of traits that every field measurement will yield a dif-
which represents an optimal tactic for that ferentvalue of r, while in theoreticalwork r
situation.However, nothingguaranteesthatthe is usuallyregarded as a fixedparametercharac-
population will attain the optimal tactic. For teristicof a species, population, or genotype.
as the environment changes, the population Similar commentsapply to K.
evolves,and as the population evolves the indi- In life-history
theory,ris used as a phenotypic
vidual organism'sperceptionand definitionof parameter (as it is in human genetics), not as
the environmentchanges. We can assume that a population parameter. The usual model
selectionelicitslocal optima fromthe variability implicitydeals with a population as though it
available in the population, but we have no were a set of asexually reproducingclones (Cl,
guarantee that the local optimum attained will C2, ... Q), each possessing a differentr (rl,
match predictionsmade from models that do rV ... rn), and asks which clone will win out
not account for the limitsof natural variability. in competition under given conditions. This
We can only measure life-historytactics on comment applies to traits other than r, and
populations,not on individuals,forthevariance I emphasize thatmost models dealt within this
among individual patterns is one of the most paper make thisimplicitassumption.It is done
important elements of a tactic, and is only forreasons of mathematicaltractability, forthe
measurable on populations. Nevertheless,such alternativeof dealing withsexual recombination
variance is the product of selection acting on explicityis overwhelminglycomplex. But the
individual organisms, not on populations of assumptionseriouslylimitsthe models that use
them. By defining tacticas I have, I certainly it. As a weak link in the chain of evolutionary
have not meant to circumventthe importance logic, our failure to deal realisticallywith sex
of individual selection,which remainsthe basis deserves wider discussion.
of the phenomena to be explained. This paper requires some understandingof
Withrespectto reproduction,severalbiologi- demography and population models. Among
cal decisions are important.How many times recenttreatmentsfor biologists,those in Krebs
should I tryto reproduce during mylife? How (1972) and Mertz (1970) are particularlylucid.
many young should I have per brood? How Laughlin (1965), Caughley (1966), and Caugh-
old should I be when I firstreproduce? Should leyand Birch (1971) have givenuseful critiques
I have a few large or many small young? How of the variousdefinitionsof population growth-
much variabilityin each of these traitsshould ratesand theirrelativemeritsin differentcon-
there be among my progeny? In essence, the texts.
theoryof life-history tacticstriesto predict the
best decisions to make in the face of problems
A BRIEF HISTORY
posed by differentsituations. After giving a
short historicaldevelopment of the subject, I The theoryof life-history tacticssplitslogical-
will consider the firstthree questions in turn, ly into three parts: (1) definingimportantbio-
then deal with the last two, and others, in a logical traits; (2) representing these traits in
more complex context. mathematicalformand developing mathemat-
Throughout this paper I use r to indicate ical tools for their formal manipulation; and
intrinsicrate of natural increase under a stable (3) deducing predictionsfromthe mathematical
age distribution,Roto indicatethe multiplicative relations defined by the theory. These three
rate per generation, and K to represent the areas have developed neither in parallel nor
carrying capacity of the environment in in series,but in a complex web.

The StableAge Distribution The ComparativeStudyofLife Tables
In 1907, Lotka established that there is a Raymond Pearl's influential books (1922,
certain age distributionto which a population 1925, 1928, 1939) emphasized the comparative
will tend, given a fixed life table, a fixed sexapproach. He showed that maximum duration
ratio, and a fixed schedule of age-specificfe- of life and the age-specificschedule of births
cundity.(Euler anticipatedthisresult in 1760.) and deaths varyboth withinand between spe-
In 1911, Sharpe and Lotka proved that this cies. In several species, including man and
age distributionis stable, and in 1925, in The Drosophila,life-historytraits show fairlyhigh
ElementsOf PhysicalBiology,Lotka summarized heritability. In 1935, Pearl and Miner reviewed
contemporarydemographicknowledge.In that thelifetablesof fourinvertebrates, as measured
book he suggested that the birth-rateis not in the laboratory. They proposed that there
onlyan adaptive traitcapable of being modified are, in general, five possible typesof mortality
by natural selection, but is adjusted to an curves,of which three are common in nature:
intermediateoptimum (see Chapter X, p. 128, thosewhere the major portionof mortalityfalls
Dover edition). late in life,earlyin life(j-shaped), or constantly
throughoutlife (diagonal).
AlthoughLotka and Pearl had urged a com-
The IntrinsicRate of Increase,r parative approach, most life-historystudies
prior to 1940 concentratedon human popula-
How is the growthrateof a population related tions. Deevey (1947) introduced ecologists to
to itsage structure,and to the differentproba- comparative demography. He reviewed the
bilitiesof givingbirthand of dyingat different constructionof life tables by three different
ages? Define lx as the probabilityof surviving methods,and pointedout thatthefewlifetables
to age x, and b. as the instantaneousbirth-rate available on differentspecies could be broken
(bxdxis the number of female offspringborn into Pearl's three patternsof survivorship.He
to femalesof age x to x + dx). Assume a stable raised the hope, now partiallyrealized, of dis-
age distribution.Now consider the number, cerning general ecological relationships
no(t),of newbornindividualsat timet. In order through a study of life tables. Following Dee-
to be newborn at time t, an individual must vey'spaper, the constructionof lifetables,either
have had parents who were newborn at time as an end in itselfor to estimate r, became
t- k, and these parents must have survived popular. Caughley (1966) and Caughley and
to age k and have given birth.Thus, Birch (1971) have recently offered valuable
f
criticismsof the techniquesused in constructing
no(t) = no(t - k)lkbkdk.
life tables and the problems encountered in
*0 interpretingthem.
If we assume that the number of newborn
progenygrowsexponentiallyin time,
Value
Reproductive
no,(t)= kert, then
Fisher (1930) defined and emphasized the
ker = ker(t-k) hkbkdk, and importance of the relative reproductivevalue
of differentage classes. He developed the idea
by analogy with the concept of present value
1= erklkbkdk. of money investedat compound interest.The
0 reproductivevalue, V. = v./v., of a female
This equation, known as Lotka's equation or of age x, relative to that of a female at birth,
the characteristicequation, was firstderived by consists of the average number of young a
Lotka in 1913 (Lotka, 1913), The characteristic female of that age can expect to have over the
equation,whichrelatesthe lifetable and fecun- remainder of her life, discounted back to the
dity schedule of a population in stable age present. The formula is worth reproducing.
toitsgrowthrate,r,has been central MacArthurand Wilson (1967, pp. 89-91) give
distribution
to most subsequent theoreticalwork. a derivation:

V./vo= (e'/1 x) f0
ert Itbtdt.
ever be able to measure a single representative
r.
Vandermeer (1968) and Caughley (1970)
Females at maximal reproductive value have both commented on the meaning of
should contributemost, per capita, to popula- reproductivevalue. Caughley has pointed out
tion growth,and, therefore,should be most that it can be interpretedeither as "(a) the
sensitiveto naturalselectionin a growingpopu- number of females alive at some future time
lation (Cody, 1971). They should, forexample, thatwill be descended froma female currently
be at optimalage fordispersal (MacArthurand aged x, relative to the number of surviving
Wilson, 1967). The idea of reproductivevalue descendents of her newborn contemporary,"
is significantbecause itallows us to characterize (p. 214) or as "(b) the numberof females born,
differentage classes as being "worth"more or at the momentof measurement,to femalesaged
less in termsof theircontributionto the intrinsic x or older, per female aged x" (p. 214). Inter-
growthrate of the population, r. However, the pretation (b) is Vandermeer's version of a
reproductivevalue of a female is the present "reproductivevalue" withan intuitivemeaning
value of her futuredaughters only when both that is easily calculable and not too far wrong,
the female and her daughters are members of so long as the age distributionis close to stable.
a population in stable age distribution(Leslie, As both Leslie (1948) and Caughley (1970)
1948). noted, interpretation(a) is true only for popu-
Leslie defined the reproductivevalue for a lations in stable age distribution.
whole population, V, as the sum of the repro-
ductivevalues of all age classes. In the discrete NarrowingtheViewpoint:Leslie and Cole
form, Between 1910 and 1940, the idea emerged
00 00 ~erx m0 that a life historyis a set of adaptive traits
V= E v./v. = E - E e-rk lkbkdk.
connected by relations that can be mathemat-
x=O x=O x k=x+1 ically analyzed. Concurrently,demographers
and statisticianswere developing mathematical
Thus V, like r,permitscomparison of the value
tools that permit extensive and sophisticated
of differentage structuresto growing popula-
manipulationof model life histories.Two au-
tions. For continuous breeders, Fisher (1930)
thors catalyzed the theoretical work. Leslie
proved that
(1945, 1948) showed that, given the current
dVt population age structureand a knownschedule
r
dt= Vt. of survivaland number of birthsper thousand
dt=
femalesof reproductiveage, one could predict
This remarkable equation holds for any age the detailed future growth of the population
distribution,stable or not, and implies that the by means of matrix methods. [Lewis (1942)
reproductivevalue of a population always in- got the same result,but publishedin a relatively
creases at the rate (r) at which population inaccessiblejournal.] This step introducedline-
numberswillgrowwhen thestableage distribu- ar transformations to demography,and opened
tion implied by the currentlifetable is reached the subject up to the analytic techniques of
(Mertz, 1970). For the discrete case, Leslie stochastic process theory. Renewal theory
(1948, pp. 220-221) devised the analogous constitutesthe other main line of mathematical
proof. Let A be the population projection thoughtthatis applicable to life histories.Key-
matrixand V(t) be a row vectorof reproductive fitz(1968) outlined the formaldevelopment of
values for each age class. Then matrix methods and renewal theory and
V(t)A= V(t+ 1), compared theirusefulnessin differentsettings,
and while Bartlett (1970) has presented a useful
reviewof stochasticmodels in demography.
V(t + 1) = X V(t) Cole (1954) emphasized the importance of
where X is the dominant eigenvalue of the age at firstreproductionin determiningpopu-
matrix A (X = er). The idea is beautiful, but lationgrowthrate, r. In examiningthe sensitiv-
in naturespatial and temporalvariabilityin the ityof r to changes in birth-rateand age at first
environmentmake it unlikelythat one would reproduction,he unveiled an apparent para-

dox. Givingbirthto n + 1 offspring,and then offeringmaximum growth,where the species

dying, is equivalent, in so far as population produces Bp offspringper female every year
growthis concerned, to survivingindefinitely startingat one year of age and continuingfor
and giving birth to n offspring at regular eternity,withno mortality.Then we have
intervals.The hidden assumptionthatgenerat- N(t + 1) = BpN(t) + N(t)
ed his apparent paradox has been revealed
(Charnov and Schaffer,1973), but Cole's defi-
= (Bp+ 1) N(t)
nition of the theoretical problems attracted and
manyworkers,to the greatbenefitof the field.
ln(Bp+ 1) = r.
Thus, Cole concluded, ifthe annual and peren-
Summary
nial populations are to increase at the same
By 1954, the main ideas of life-history
theory rate,
were well defined. Subsequent efforthas con- ln(Ba) = r = ln(Bp + 1),
centrated on the four areas reviewed below.
or
We willsee thatpeople firstbroke life histories
down into independent traits,and considered Ba= BpB+ 1.
those in isolation. Recently, the process of
reassemblingtacticsfrom sets of independent The Controversy
traitshas begun. The areas to be reviewed are:
(1) the argument over whether an organism Murdoch (1966) criticizedCole's conclusion
should reproduce once or many times; the on the grounds that the model that produced
adaptive significanceof variationsin (2) clutch it assumed an oversimplifiedgoal for organ-
size, (3) age at firstreproduction,and (4) size isms: maximizinggrowthrate. Whenever there
of young; and (5) the evolutionof reproductive is a trade-offbetweenadult survivaland repro-
tacticsas systemsof coevolved adaptive traits. ductiveeffort,and the environmentis variable,
an adult that decides not to reproduce may
THE NUMBER OF CLUTCHES IN A LIFETIME have a better chance of survivingto the next
breeding season than would the young pro-
A ParadoxicalResult duced if the adult had decided to breed. Gadgil
and Bossert (1970) used the discrete form of
Cole (1954) initiateda controversywith the Lotka's equation in attemptingto analyze Cole's
followingstatement: statement.They assumed that the population
Foran annualspecies,theabsolutegain intrinsic under consideration is not growing, that the
in
populationgrowthwhichcould be achievedby femalesare just replacingthemselves,and that
changingto the perennialreproductivehabit although there is no mortalityafter maturity,
wouldbe exactlyequivalent toaddingone individ- thereis juvenile mortality.Thus foran annual,
ual to theaveragelittersize (p. 18). = lxb.= lb1 1
He arrived at this conclusion on the basis of 1 =e-r11 b
the followingargument.Consider a population
of annual organismswithno juvenile mortality, r = ln(1lbl).
one that maturesin the summerand dies after For a perennial species,
reproducing. When it has reached its stable
growthrate, the followingrelationshipholds: 1 = E e-- 1.b.dx = 11bE e--,
1
N(t + 1) = erN(t) 1
which implies that

= BaN(t)
01
where Ba is the average female birth-rateper r = ln(l1b1+ 1), since e-rx
capita, and N(t) and N(t + 1) are population 1er- 1
size in successiveyears. Thus, Now since 11b1 1, r ln(211 bI). From this
ln(Ba) = r. result,theymodified Cole's statementto read:
Now consider an extreme case of iteroparity, For an annual speciesthe absolutegain in the

Malthusianparameterwhichcould be achieved N(t + 1) = Ba C N(t)

bychangingto the perennialreproductive habit
would be approximately equivalentto doubling for the annual, and
theaveragelittersize (p. 11). N(t + 1) = Bp C * N(t) + P * N(t)
But surely this is nonsense, for the annual = (Bp' C + P) N(t)
population is just replacing itself(r = 0), while for the perennial. Thus
the perennial population is growingquite rap-
idly (r = ln2 = 0.69). They are comparing apples Aa =Ba* C kp=(Bp*C+P),
and oranges. Their conclusion results from a and if the populations are to increase at the
rather free use of approximations and from same rate,
setting11b1 = 1 forthe perennials,when in fact
Boa = Bp+ (P/C).
their replacement assumption implies some-
thingquite different: In both Cole's and Bryant's models, P = C,
giving
lXbX= 1. B.=-Bp+ I
which is Cole's result. Thus, Charnov and
Bryant (1971) criticizedGadgil and Bossert SchafferrestateCole's resultin thisfashion:
for inducing a general result from a special
case, and dealt with Lotka's equation by using Foran annualspecies,theabsolutegaininintrinsic
a constant littersize, b, but an exponentially populationgrowthratethatcan be achievedby
distributedsurvivorship,e-x, where u is the changingto the perennialreproductivehabit
wouldbe exactlyequivalentto addingP/C indi-
mortalityrate:
vidualsto theaveragelittersize(p. 792).
-
1 = (r+u)x They also worked out the more complicated
case of delayed reproduction.
whichgivesr = ln(b + 1) - u, forthe perennial. Harper (1967), in examining the conditions
For an annual withthe same pre-reproductive under which the effectivereproduction (re-
mortalityas the perennial, cruitmentof one-year-olds per female) of a
biennial plant is equal to that of an annual,
r= ln(b) - u.
dealt withessentiallythe same ideas as Charnov
Fromthislastexpression,Bryantconcluded that and Schaffer,and made a similar prediction.
Cole's resultwas correct. But because he confounded the effectsof de-
layed reproduction and the adult/juvenile
The Effectsof VaryingAge-SpecificMortality survivorshipratio,and because he did not work
with the characteristicequation, he failed to
Notice that Cole and Bryant both assumed recognize the generalityof his results. Both
equal adult and juvenile mortalityrates (no Murdoch (1966) and Cody (1971) recognized
mortalityin either stage for Cole), whereas the importanceof the adult/juvenile survivor-
Gadgil and Bossert made the unnatural as- ship ratio for life-historytactics,and arrived
sumption of some juvenile mortalityand no at a less precise statementof the conclusions
adult mortality.By examiningvariationin both presented here.
adult and juvenile mortality,Charnov and To summarizethe analysismade by Charnov
Schaffer (1973) derived Cole's and Bryant's and Schaffer: they emphasize the importance
results as special cases. They compared two of separatingjuvenile and adult mortality.Age-
populations, one annual, the other perennial, specificmortalitydistributionsplay a large role
producing Ba and Bp young per litter,respec- in determiningthe optimal life-historytactic.
tively.The proportion of offspringsurviving Charnov and Schaffer predict that iteroparity
for the first year is C for both, while the will be favored, for both early and delayed
perennials have an adult survivalrate of P per reproduction,by adult survivalrates which are
year. Thus, the recurrencerelationsfor popu- high relativeto juvenile survivalrates.
lation growth,given in terms of the number The model used by Charnov and Schaffer
next year N(t + 1), and the number this year, exemplifies the implicit assumption that r is
N(t), are: a phenotypiccharacteristic and thatpopulations

consistof clones of asexually reproducingindi- ratio of average juvenile mortalityto average

viduals. They have swept the complicationsof adult mortalityas
sex and geneticsunder the theoreticalrug, and
have chosen to base theirconclusionson a very
simple comparison of the effectsof varying
juvenile and adult mortalities.Although pure
theoreticiansmightfindthesecriticismsserious, [En(j) q 0)A/ n (j)4
they lose much of their strength when one
recallstwo points.No one has adequately dealt where q(j) is the probabilityof dying in the
withthe problems sex and genetics pose to a intervalj to j + 1.
definitionof fitness. I cannot fault Charnov In examining the literature, I uncovered
and Schaffer, or others who use the same about fiftylife tables, from which I selected
approach, for failing to solve a fundamental twenty-three. I selected only lifetables calculat-
problem that has puzzled so many. Secondly, ed from data gathered in the field on animals
suppose theirpredictions,based on admittedly livingin seasonal environmentswitha definite
simplistic models, prove useful? Surely the breeding season, withsample sizes larger than
proper grounds forrejectinga speculation that 150 animals, and where the authors did not
is logicallysound withinthelimitsof itsassump- confoundmortalitywithmigrationor dispersal.
tionsare empirical.That Charnov and Schaffer There remain threepossible problemswiththe
chose to conceptualize life-historyprocesses in tables selected: (1) there may have been inade-
a very simple model is not a valid criticism. quate replication,i.e. calculated from a single
If their predictionswork, then perhaps com- sample, or at a single place, or for a single
plexityis not necessary.These commentsapply year; (2) assumptions may have been made
equally to all the theoretical work, reviewed about the fecundityschedule, rather than es-
below, on r-and K-selection,age at firstrepro- timationsfrommeasurements;or (3) theremay
duction,and bet-hedging. have been perturbations of the population
owing to human intervention, e.g., fromfishing
or huntingpressure.
The Evidence
The resultsare presented in Fig. 1 and Table
A directtestof the Charnov-Schafferpredic- 1. The tablealso noteswhichof the above three
tion would involve selection experiments on criticismsapply to each species. I have ranked
short-livedorganisms, preferablyin the field. the data and calculated Kendall's rank-correla-
Althoughsuch experimentshave yetto be done, tion coefficient,tau. There is a strongenough
we can neverthelessgain some impression of correlation (tau = 0.26, z = 1.78, p = 0.0375)
the plausibilityof the hypothesisthrough the between the degree of iteroparityand the ju-
analysisof published lifetables. If the hypothe- venile/adult mortalityratio to argue that a
sis is correct,then there should be a positive selection experimentwould be worth doing. I
correlation between the average number of certainlydo not claim that the data in Table
breeding seasons (a measure of "degree" of 1 reveal a "natural law." Problems in measure-
iteroparity)and the ratio of average juvenile mentmake the data weak, thereis considerable
to average adult mortality. scatter,and I am sure that many forces other
Let us define average number of breeding than mortalitypatterns have influenced the
seasons as data. That we can only draw weak conclusions
fromthe data available should spur more rigor-
+ 1) ous work.
En(j)(j-(o
forb(j)/=O,
OtherForcesSelectingforRepeatReproduction
E n(j)
Forces other than age-specificmortalityalso
where a is age at firstreproduction, n(j) is influence iteroparity. Holgate (1967) and
the number of females in the cohort aged j, Murphy(1968) pointed out thatiteroparitywill
and b(j) is the average number of female be directlyselected for when the risk of total
progeny born to females aged j. Define the reproductivefailurein any given year is signifi-

un
z 7
n
(n ~ 6 Laughing
gull
6 0 Laughing gull * African elephant
C9
z
LLI 5 - Wildebeeste
w 0 Cave bear
0 Black rhinoceros
0
Co 4
Dall mountain sheep 0 Himalayan thar
0
L 3 0 Domestic sheep 0 Pacific sardine
0 Red deer 0 Barnacle

D 0 Blackbird
Z 2 Starling * Song thrush 0 Pika
w11 American robin 0 Robin
(D
r 1 Pine
i looper
LJ-J 45 year sockeye salmon
> t ~Winter moth
<Eye-spotted bud moth
iPstol casebearer
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0

AVG. JUV. MORT. / AVG. ADULT MORT
FIG. 1. EVIDENCE BEARING ON COLE'S PARADOX.

The average number of breeding seasons is positivelycorrelated with the juvenile/adult mortalityratio.
See Table 1 for references,text for explanation and criticism.
cant. Iteroparitymay also arise concomitantly of producing. (2) Birds lay as many eggs as
withparental care when post-reproductivesur- they can cover. (3) Organisms adjust the size
vival is favored. Strathmann(1974), in consid- of the clutch on the basis of informationre-
ering the short-termadvantages of large-scale ceived through social mechanisms to balance
dispersalof the siblinglarvae of marine inverte- the prevalent level of mortalityand produce
brates, has suggested that iteroparitywould an "optimum" population size (Wynne-Ed-
promotedispersal by samplinga series of times wards, 1962; Skutch, 1967). (4) Parents pro-
and conditionsfor releasing planktoniclarvae. duce, on theaverage, themostproductiveclutch
Whenever larval dispersal is advantageous, he size, defined as that clutch size which results
predicted more intense selection for iter- in the mostyoung survivingto maturity.There
oparityin species withshorterpelagic stages. is an optimum,or most productive,clutch size
correspondingto any particularset of ecological
conditions,and it is determined primarilyby
CLUTCH SIZE the amount of food the parents are capable
of bringing to the young (Lack, 1947, 1948,
The Hypotheses 1954). (5) Clutchsize is determinedbya balance
of allocations of limited resources to competi-
There are at least five hypotheses on the tion, predator avoidance, and reproduction
evolution of clutch size. (1) The organism has (Cody, 1966). A sixthpointof viewplaces clutch
as many young as it is physiologicallycapable size in a more comprehensive context. It is

TABLE 1
A comparison ratiosand breedingseasons
ofmortality
CRITICISM
COMMON NAME SCIENTIFIC NAME J/A MORT. NO. SEAS. REFERENCE (see text)
Red Deer Cervuselephas 0.092 2.4874 Lowe, 1969 1,3

Dall Mountain Ovis d. dalli 0.554 4.0109 Deevey, 1947; Geist, 1971 1
Sheep
Himalayan Thar Hemitragusjemlahicus 1.995 3.9209 Caughley, 1966 1,3
Wildebeeste Connochaetestaurinus 1.120 4.8380 Watson, 1970 1
American Robin Turdusm. migratorius 0.888 1.6899 Deevey, 1947; Lack, 1954 1,2
Starling Sternusvulgaris 0.939 1.8443 Deevey, 1947; Lack, 1954 1,2
Robin Erithacusrubicula 1.315 1.8234 Deevey, 1947; Lack, 1954 1,2
Song Thrush Turdusericetorum 1.168 1.9861 Deevey, 1947; Lack, 1954 1,2
Blackbird Turdusmerula 1.294 2.2750 Deevey, 1947; Lack, 1954 1,2
Laughing Gull Larus argentatus 1.056 6.1615 Kadlec & Drury, 1968 2
Barnacle Balanus glandula 2.468 2.4575 Connell, 1970 2
Pika Ochotonaprinceps 1.780 2.0276 Millar & Zwickel,1972 1,2
Black Rhinoceros Dicerosbicornis 1.324 4.4553 Goddard, 1970 1,3
Pine Looper Bupalus piniarus 0.977 1.0000 Klomp, 1966 3
Four-Year Sockeye Onchorhyncus nerka 0.982 1.0000 Hanamura, 1966 2,3
Salmon
Five-Year Sockeye 0. nerka 0.976 1.0000 Hanamura, 1966 2,3
Salmon
Domestic Sheep Ovis aries 0.398 3.0181 Hickey, 1960 1,3
Cinnabar Moth Tyreajacobaeae 0.987 1.0000 Dempster, 1971 3
PacificSardine Sardinopscaerulea 3.007 2.9575 Murphy, 1967 2,3
WinterMoth Operophterabrumata 0.996 1.0000 Embree, 1965 1,3
Diamondback Moth Plutellamaculipennis 0.933 1.0000 Harcourt, 1963 3
Eye-Spotted Spilonatacoellata 0.972 1.0000 LeRoux, Paradis, & 3
Bud Moth Hudon, 1963
PistolCasebearer Coleophoraserratella 0.870 1.0000 LeRoux, Paradis, & 3
Hudon, 1963
AfricanElephant Loxodontaafricana 2.951 5.8162 Petrides & Swank, 1966 1,2
discussed below under the heading of, "Repro- among ornithologists,but has been subsumed,
ductiveTactics." with modifications,in a more comprehensive
The firstand second hypothesesare at best theory. In this section I shall discuss some
special cases, for theyare contradictedby the criticismsof Lack's hypothesis,review Cody's
evidence (Lack, 1948, 1954; Cody, 1966, 1971; early ideas on clutch size (Cody, 1966), and
Klomp, 1970). Wynne-Edward'shypothesisre- then present three recent models that predict
quires group selection,which may or may not clutch sizes smaller than the one that fledges
occur in nature.But group selectionwould only the most young.
rarelyovercomeindividualselection(cf. Levins,
1970; Wilson, 1973), and in any case should ProblemswithMaximizingProductivity
be invoked only when individual selectioncan-
not account forthe phenomena observed (Wil- There are two aspects to Lack's hypothesis
liams, 1966a). However, much of the evidence that are best kept separate. On the one hand,
adduced by Wynne-Edwards and Skutch in Lack proposed that, withina population in a
arguing against Lack's hypothesisneeds expla- single geographicallocation,the average clutch
nation, and may have found it in the theories size willbe the one thatfledgesthe mostyoung,
of reproductivetactics.Lack's hypothesisthat as a result of selection. On the other hand,
clutchsize is adjusted to maximize the number Lack explained the known geographical trend
ofyoungfledgedheld thefieldforabout fifteen toward larger clutches at higher latitudes as
years (1950-1965), still has many adherents the result of longer daylengths that allow

parentsmore timeto get food and thus increase havior, since if only individual selection is
the size of the most productiveclutch. operating,the individual who "cheats" and has
There is some evidence that birds do tend a large clutchwillwinout. Since group selection
to produce a clutch near the most productive probably rarely occurs, and when present is
of the clutch sizes available to them. Klomp less efficientthan individualselectionin chang-
(1970), in an exhaustive review of clutch size ing gene frequencies, we may disregard
in birds, concluded that evidence from the six Wynne-Edwards's theory. If clutch size de-
well-studiedspecies supportsLack's hypothesis. creases over evolutionarytime as population
Where there are discrepancies, the most fre- densityincreases, it probably does so because
quent clutch size is smaller than the one that selectionactson individualsunder such circum-
fledges the most young (four well-studiedspe- stances to favor small clutches, for any of a
cies), except for the Common Heron. Several large number of reasons reviewed below and
models, reviewed below (p. 15 ff.), suggest in the section on r- and K-selection and bet-
ultimatefactorsthatfavorreduced clutchsizes. hedging.
Drury(1961) studiedthebiologyof passerines Skutch's (1949, 1967) ideas are less easily
in the Canadian Arctic.He noted that,although disposed of. He maintained that under stable
therewere significantdifferencesin clutch size conditions populations are operating at close
between the southern and northern parts of to saturationdensitymost of the time. Under
thebird'srange, the southernpopulations,with these conditions (e.g., in the humid tropics)
a clutchsize of 4 to 5 eggs, had enough available there is no point in raising as many young as
daylight(20 hours) to feed a clutchsize as large you can feed,since all those beyondthe number
as the ones that northern populations raised needed for replacement will die. Skutch sug-
(6 to 7 eggs). Drury suggested that the larger gested that, instead, clutch size is adjusted to
clutch sizes in the north could be explained balance mortalityunder saturationconditions.
(1) as adaptations to increased r, allowing for In counteringLack's food-limitation argument,
rapid recovery from disasters induced by cli- he cited several cases of two species of birds
maticextremes;and (2) as the resultof concen- of the same genus, living in the same habitat,
tratingthe annual production of young into thatraise the same clutch size to maturityeven
one, rather than two, successive clutches, and though both parents feed the young in one
thus allowing equivalent production over the species, and only one parent in the other. In
shorterarcticsummer. countering Lack's most-productive-clutchhy-
On the other hand, Hussell (1972), in a more pothesis, he cited the advantages that small
exhaustive field study of arctic passerines, clutch sizes may have in reducing the risk of
sought evidence that the increase in clutch size predation to parents and young: fewer visits
with latitude was related to the day-length to the nest are required, and that decreases
available for food collection. He found that the chance of the nest being found. Further-
variation in clutch size with latitude, as well more,in those species Skutchcited as evidence,
as with other factors,was related to environ- the young do not give begging calls, and the
mental factorsinfluencingthe food-gathering food itemsbroughtto the nest are quite large,
potentialof parents. a conditionwhichfurtherreduces the necessity
Two older theories on clutch size should be for repeated feeding flightsand decreases the
considered. One can be disposed of easily. probabilityof detectionby predators.
Wynne-Edwards(1962) argued againstLack on Ricklefs (1969) stated as a generality that
the basis of the existence of broad negative predation is the major cause of death for eggs
correlationsof clutch size withpopulation den- and nestlingbirds, regardless of geographical
sity in birds and mammals. Wynne-Edwards location,habitat,or nestsite,and thatstarvation
explained these correlationsby the infamous is less frequentlya source of mortalityin the
theory(cf. Williams,1966a) thatthroughsocial humid tropics. This generalization lends
interactionsthe animals notice they are ap- support to Skutch'semphasis on the problems
proachingcriticaldensityand self-regulatetheir encountered by nesting birds in dealing with
population density by reducing clutch size to predators. Fretwell (1969) has suggested that
preventa population crash. Group selection is the hypotheses of Lack and Skutch can be
required for the evolution of this type of be- reconciled by supposing that adults tryto pro-

duce as many breeding offspringas they can, variables. It is difficultto see how an animal's
but thatwherevermortalityselectivelyimpinges abilityto compete or to avoid predators could
on excess offspringthrough social dominance be a simple function of energy allocation.
or other effects,then the birth-ratecan be Morphological design seems to me more
adjusted downward, not in order to balance important.
the death-rate,but to maximize the number Nevertheless,Cody predicted that in envi-
of survivingprogeny. ronments he defined as unstable (temperate,
Skutch's explanations of lower clutch sizes arctic) most energy will go into increasing the
in the tropics are not the only ones. One reproductiverate, and thus increasing clutch
alternativeexplanation is that where replace- size,at leastin part.In environmentshe defined
ment is important,a few high-qualityyoung as stable (the tropics, islands, coastlines), we
are preferable to many low-qualityyoung, as should findsmallerclutches.Finally,hole-nest-
has been suggestedfor fishby Svardson (1949) ing species, which are relativelypredator-free,
and Williams (1966a). In fact, a bet-hedger but more so in the tropicsthan in the temper-
would produce just a few more high-quality ate areas, should have larger clutches in all
young than are necessaryfor replacement,in areas, but would exhibit less of an increase of
order to take advantage of occasional fluctua- clutch size with latitude. All of Cody's predic-
tions in conditions to increase the number of tions were corroborated by the data he cited,
survivingyoung. The number of additional but I question his definitionsof stabilityand
youngproduced willdepend on the probability instability.We have few, if any, good data on
of a change in conditions. how stable differentenvironmentsappear to
Much oftheargumentoverLack's hypotheses organisms. Without an independent measure
has stemmedfrommisinterpretations of Lack's of stability,Cody's argument is circular. Nor
intent,fromhis overlystronginsistenceon food did he consider alternativeexplanations,which
as the proximal limiton clutch size, and from exist in abundance (cf. Table 5).
the fact that in the 1950's and 1960's, when
the argument was at its peak, the concept of ModelsFavoringSmall Clutches
a tradeoff between reproductive effort and
parentalsurvivalhad notenteredthe discussion. Cody predicted clutch sizes smaller than the
I am sure Lack was on the righttrackin saying most productive clutch size on the basis of a
that birds produce clutch sizes that leave the trade-offbetween reproduction, competition,
most survivors.The only real changes in our and predatoravoidance. There are at least three
thinkingabout clutch size have come, first,in othertheoreticalmodels thatpredictclutchsizes
how we count survivors-not just fledglings smallerthanthe mostproductive.In thatsense,
from a single clutch but all young produced all threeoffermodificationsto Lack's hypothe-
over the life span of the parent; secondly, in SiS.
recognizingthat factorsother than food, such The first,developed by Charnov and Krebs
as predationor parental survival,may be limit- (1973) and based on the centralidea in Williams
ing in many cases; and thirdly,in recognizing (1966b), makes the assumption that there is
thatan alternativeexplanation arises when we a positive correlationbetween clutch size and
ask not how to maximizethe number of young adult mortality.They argued that for a peren-
produced, but how to minimize-the probability nial species, where the act of reproduction
of leaving no young at all. impliesa mortalityrisk,large clutchsizes reduce
the parents'schances to surviveand reproduce
A Trade-OffModel in the future. The lifetime contribution to
future generations is optimized by producing
Restating much of what Skutch had said, clutchessmallerthanthe mostproductivesingle
Cody (1966) reasoned that organisms have a clutch(Fig. 2). Table 2 citesevidence thatthere
limitedamount of energy,and thattheyallocate is a trade-offbetween reproduction and adult
it to reproduction,competition,and avoidance mortality.However, thispredictiondepends to
of predation. Energy not needed for competi- a certain degree on the shape of the trade-off
tion can go into reproduction, and so forth: function,which has never been measured in
there are trade-offs among the three state any species. Its measurementwill be one step

reproduction,what conditions will be like for

the rest of the year. Since making too large
S ~~~s a reproductiveeffortmayresultin disaster(with
all the young and perhaps the parents dying)
whereas laying a less than optimal clutch at
(A)L b.a leastresultsin some young,theorganismshould
always hedge its bet on the side of a smaller
than optimal clutch, rather than in the other
direction(Fig. 3). Mountford(1968) describe$i
G)~~~~b
an alternative mechanism that produces the
same result. If one assumes that organisms
produce a normal distributionof clutch sizes
over their lifetime,and that the proportion
0=~~~~6 recruited decreases with clutch size, then
compares a series of similarlyshaped distribu-
tions of clutch size that have differentmeans,
one discoversthatthedistributionthatproduces
(B)7U|(g :: the most recruitsover the organism's lifetime
O b.s has a mode thatis less than the most productive
single clutch.
Mountford(1973) emphasized thatifthe goal
is to minimizethe probabilityof extinction,then
the tacticsselectedwillusuallydifferfromthose
based on the traditionalcriterionof maximizing
growthrate. His model deals with populations,
Clutch-size,b not individuals, as the units of selection. In
FIG. 2. A GRAPHICAL ARGUMENT ON OPTIMIZING
general, there are good biological reasons not
CLUTCH SIZE
to do this: selection acts on individuals, not
(A) The relationship between clutch size, b, and on groups. But under special circumstances,
survivalrate for the firstyear of life,s (broken line). the idea becomes more plausible and brings
The resulting b s or production curve (solid line) up the third argument, put forthby Gilbert
has a singlemaximumat b0.Under Lack's hypothesis and Gutierrez(1973). They dealt witha popu-
this is the clutch favored by natural selection. lation broken up into groups, each of which
(B) The impactof the cost of reproduction.If adult undergoes independent growth and eventual
mortality,q (dashed line), increases with b,then the density-dependentlimitation.Under these spe-
clutch bo*that mnaximizes the measure of fitness,a, cial circumstances,how can the organism avoid
is alwayssmallerthan b0.The optimalclutchis found
selection for ever-increasingfecundity,which
by constructingthe line parallel to q (q', dotted line)
which just intersectsthe b*s curve (solid line). In
would greatlyincrease the chances of group
general, clutches smaller than bo will have a higher extinction?They point out thatthe total fitness
fitnessthan those larger. of an individualover the entiregrowingseason
is made up of two factors,its fecundityduring
the growthphase and itscontributionto group
in testing both this and other theories on life fecundityduring the density-limited phase. In-
history phenomena (cf. Williams 1966a, b; dividual selection will act to maximize total
Schaffer, 1972, 1974a). fitness.When there is a correlation between
The second modification, suggested by group fecundityand probabilityof extinction,
Cohen (1967) and by Boer (1968) but most optimum individual fecundityis less than the
thoroughly developed by Mountford (1973), most productive individual fecundity. This
involves the idea of hedging bets in the face model works for an aphid population, where
of uncertainty. Suppose environmental condi- the membersof groups are closelyrelated,but
tions vary from year to year, and that the forsexuallyoutcrossingorganismsthe situation
organism cannot be certain, at the time of changes, and "cheaters" may win.

TABLE 2
Evidencethatthereis a real costassociatedwithreproducing
TYPES OF EVIDENCE SPECIES REFERENCE
(1) Increase in adult mortality Rocky Mountain Sheep: Ovis dalli Geist, 1971
during and afterreproduction: carabid beetles: Agonumspp. Murdoch, 1966
relativeto those that do not littoralgastropods Fotheringham,1971
reproduce Shaskyus and Ocenebraspp.
house sparrow: Passerdomesticus Summer-Smith,1956
European blackbird: Turdusmerula Snow, 1958
isopod: Armadillidiumvulgare Paris & Pitelka, 1962
meadow voles: Microtusagrestis Clouch, 1965
pandalid shrimp: Pandalus borealis Allen, 1959
Olive Baboon: Papio anubis Berger, 1972
minnow: Pimephalespromelas Markus, 1934*
bug: Dysdercus fasciatus Clark & Sardesai, 1959*
(2) A decrease in growthrate barnacles: Elimniusmodanus Crisp & Patel, 1961
during and afterreproduction: Balanus balanoides Barnes, 1962
relativeto those that do not sugar cane: Saccharumofficinarum
reproduce
(3) A reductionin reproductive guinea pig Loeb, 1917
effort,or an increase in sterility sheep and cows Hammond, 1958*
in underfed
or infertility Lamming, 1969*
animals. pigs Duncan & Lodge, 1960*
domestic fowl Brody, 1945; Peterson et
al., 1960; Kurnick et
al., 1961
rainbow trout: Salmo gairdneri Scott, 1962
*cited in Calow, 1973
Summary the evolutionaryprocessesinvolved.Naturalse-

lectiondoes notoperateto maximizethe repro-
Lack's original hypothesishas been reinter- ductiverate,butfavorsthosecharacteristics
which
preted in later models that alter it without allowtheindividuals
possessingthemto maximize
destroyingit. In the process, emphasis has theiroverallgeneticcontributionto subsequent
shifted to a differentway of looking at the breedingpopulations.This can be achievedby
evolutionof life-history
traits.Instead of isolat- changesin anyaspectofthereproductive strategy
ing one adaptive trait and looking for the (Hussell1972,p. 353).
ecological interactionsthat determine its ob-
served variation,we must consider organisms At least five theoreticalmodels incorporate
as systemsof interacting,coadapted traits,with mechanismsthat account for the reduction of
trade-offsamong them. As Hussell put it, clutch size below the most productivesize: (1)
trade-offsbetween demands for resources on
. . .' a distinctionshouldbe made betweenthe the part of reproductionversusother functions
factorsresponsible fortheshort-term regulation (Cody, 1966; Skutch, 1967); (2) trade-offsbe-
ofpopulations and thoseresponsible forthelong- tween clutchsize and adult mortality(Charnov
termevolution ofreproductive ratesand theother and Krebs, 1973); (3) bet-hedgingin the face
aspectsof thelifecycle.In theshorttermbirds of uncertaintyabout conditions during the
are probablyreproducing as rapidlyas possible
breeding season (Boer, 1968; Holgate, 1967);
withinthe limitsset by theircurrentgenetically
controlled capabilities.
Populationsizeswouldthen (4) the interactionof a normal distributionof
be regulatedby subsequentdensity-dependentclutch sizes with a probabilityof recruitment
mortality. But the idea of maximization of the that declines with increasing clutch size
reproductive totheenvironmental (Mountford, 1968); and (5) a positive correla-
rate,in relation
food supplyhas no meaningin the contextof tion between clutch size and the probabilityof

growthrate, r, to changes in the age at first

reproduction, the age at which reproductive
value startsto decline (the turnoverpoint),and
0
-~~b the age at last reproduction. He found that
- -- - - - r was most sensitiveto a decrease in age at
firstreproduction,less sensitiveto an equivalent
decrease in the turnover point, and quite in-
sensitiveto a decrease in the age at last repro-
duction (holding total numberof progenycon-
b1 bo b2 b3 stant in all cases). Decreasing the age at first
Clutch- Size (b) reproductionfrom12 to 9.8 days was equivalent
to doubling the total number of progeny pro-
duced.
FIG. 3. BET-HEDGING REDUCES AVERAGE OPIMUM On the other hand, Hamilton (1966) noted
CLUTCH SIZE that,duringa populationdecline,the individual
Assumeagain,as in Fig. 2, a trade-off between should delay reproduction to the modal age
clutchsize and survivalof young.Define symbols of the stable age distributionin order to slow
as in Fig. 2. If thereis year-to-year in the rate of decline. Mertz (197 la, b) made the
variability
theoptimalclutchsize,selectionfavorsbet-hedging same point, and suggested that the delayed
witha smallerclutch.Let b0 be the optimalclutch reproduction and longevityof the California
size thisyear,withb s curve1. If in the nextyear
condor may representan adaptive response to
changedconditions resultinb s curve2,thena smaller
clutch,b1, will offermaximumfitness,while an a long period of decline in size of population.
equivalently largerclutch,b2, offersmuchreduced Mertz suggested that populations which ex-
fitness;a verylargeclutch, b3,wouldresultina serious perience long periods of decline punctuated
mortality riskto the parentand no youngat all. by brief episodes of expansion will evolve de-
The analysisgivendependson theshape of the b s layed reproduction,but offeredno method of
and q curves,whichcertainly varyfromspeciesto quantifyingtheprediction.Accordingto models
species,placeto place,and timeto time. of island populations proposed by MacArthur
and Wilson (1967), a colonizing species multi-
extinctionwhere the population is broken up plies rapidly in a new habitat, then declines
intogroups of closelyrelatedorganisms(Gilbert slowly.During the decline it sends off propa-
and Gutierrez,1973). gules to other habitats. According to Mertz,
such populations should delay reproduction,
but according to MacArthur and Wilson,
AGE AT FIRST REPRODUCTION Lewontin, and others, they should have early
reproduction.(I thankG. C. Williamsforpoint-
Cole (1954) and Lewontin (1965), among ing thisout to me.)
others, have pointed out that, in a growing The differentpredictionsstemfromdifferent
population, selection will push the age at first estimates of the frequency of colonizing epi-
reproduction to the physiological minimum. sodes. We need a comprehensive theoretical
This can perhaps best be seen by analogy with treatmentof the impactof differentfrequencies
compound interestin banking: it will always of colonizing episodes on the evolutionof life-
pay to get your money in the bank as soon historytactics.Such an analysis should reveal
as possible so that the compounding interest a criticalregionin the frequencyof colonization
rate will get to work most rapidly. below which earlyreproductionis favored,and
More specifically,Cole (1954: Figs. 4 and 5) above which delayed reproduction is favored,
pointed out that,other thingsbeing equal, the given a declining population. Another analogy
advantage of a lower age at firstreproduction withbanking is useful here. If the interestrate
is greater for animals with large clutch sizes is positive, it pays to get your money in the
than for those with small ones, and greater bank as soon as possible.When thebank,instead
for semelparous than for iteroparous organ- of giving interest,levies a monthlycharge on
isms. Lewontin(1965) used a simulationmodel moneydeposited, but you know you eventually
to deal with the sensitivityof the population must deposit the money, then it pays to delay

putting your money in the bank as long as viewpointson reproductivetacticsoverlap and

possible. interpenetrateextensively.
In fact, population growth rates fluctuate
around zero. Rarely will a population experi-
ence a long, uninterruptedperiod of increase FitnessAs A FunctionofLife HistoryTraits
or decline. Thus pure r-selection,or pure selec-
tion for delayed reproduction as proposed by The characteristicsof a population have
Hamilton and Mertz,representunrealisticthe- evolved togetheras interrelatedtraits. Evolu-
oretical limiting cases rarely approached in tionarychanges in one traitmayimplycompen-
nature. When the growth rate is confined to satoryor opportunisticchanges in anothertrait.
a narrowband around zero, I expect entirely The most revealing question to ask of such a
differentselectionforcesto dominate. systemis this:How sensitiveis fitnessto a change
Other authors have discussed factorswhich in any one of the traits?Is this sensitivityitself
favor delaying the age at firstreproduction, a functionof the other traits?We may consider
e.g., Harper (1967), Murphy(1968), Gadgil and as most importantthat trait to whose change
Bossert (1970), Cody (1971), Geist (1971), fitnessis mostsensitive.Giventhatwe are willing
Schaffer(1972), Schafferand Reed (1972), and to accept a singlefitnessmeasure as appropriate,
Schafferand Elson (1975). Some of these will we may rank the importanceof any set of traits
be discussed in more detail in the section on whose relationshipto a given fitnessmeasure
reproductivetactics(below). They maybe sum- can be preciselyexpressed. Cole (1954), Lewon-
marized as follows. (1) If juvenile survival is tin (1965), and Meats (1971) have used this
greater than adult survival (because juveniles procedure to explore the sensitivityof rate of
live in a protected habitat, or because there increase (Ro or r) to changes in life history
is a significanttrade-offbetween reproduction parameters.
and adult survival),then it is advantageous to Cole (1954) found that r is quite sensitive
delay reproduction to a certain point. Healy to changes in age at firstreproduction (A),
(1974) has documented reproductivetrendsin and that r is much more sensitiveto a given
Massachusetts newts (Notophthalmus) that percentage change in A when A is low.
support this view. (2) If reproductive success Moreover, r is much more sensitiveto a change
depends on age, size, or social status,as it does in A when the birth-rateis high. Thus, one
in many hierarchicalsocieties (e.g., mountain could expect to find age at firstreproduction
sheep, baboons, and seals) and in species where and birth rate under strong selection when
it takes considerable effortjust to reproduce populations are growingrapidly.
at all (e.g., salmon and eels), then delayed Lewontin (1965) was concerned with those
maturitywill again be favored. combinationsof lifehistorytraitsthatproduced
individuals fit for colonization. He assumed
exponential population growth, a stable age
REPRODUCTIVE TACTICS distribution,and overlapping generations,and
took r as his fitnessmeasure. He then modeled
So far we have been considering different a life historyusing V(x) = lxbxas a triangular
aspects of a life historyas more or less isolated functionof age x (see Fig. 4), and defined A,
and independent phenomena. Now let us put T, and W as age at firstreproduction,age at
themtogetheras a systemof interrelatedadap- peak reproduction,and age at last reproduc-
tivetraitsforminga set of reproductivetactics. tion, respectively.To derive an expression for
There are at least four differentviewpointsin r in termsof these parameters,Lewontinnoted
the literatureon how to look at reproductive fromthe geometryof Fig. 4 that
tactics.How sensitiveare fitnessmeasures to 2S(W- x)
changesin differentparametersofa lifehistory? V(x) = , x2T
How do r- and K-selection and bet-hedging (W- T)(W- A)
affectlifehistories?How do lifehistoriesevolve 2S(x-A)
in variable environments?And how do trade- V(x) =- , x< T
(T- A)(W- A)
offs between reproductive effort and other
commitmentsaffectlife histories?These four He then substitutedthese expressions in

species should show much less geneticvariance

1=J e- V(x)dx in age at firstreproduction, which is under
0
strong selection pressure, than they show for
integrated,and obtained a complicatedimplicit fecundity.
expressionforrthatcould be evaluated numer- Meats (1971) extended Lewontin's analysis
ically. to a wider range of values of r and examined
He took as his baseline the case where A the separate effectsof M and N, mortalityand
= 12 (age at firstreproduction),T = 23 (age natality,forboth discreteand overlappinggen-
at peak reproduction), W = 55 (age at last erations. He found thatwhen prereproductive
reproduction), r = 0.30, and S = 780 (total mortalityis high (ca. 0.80-0.99), the growth
eggs expected per female over her lifetime). rate of an annual population (Ro) is mnuch more
The change in S required to change r from sensitiveto changes in mortalitythan natality;
0.30 to 0.33 is +670, nearlya doubling to 1350 when juvenile mortalityis low (ca. 0.01-0.60),
eggs. The changes in the other parametersthat the effectsof a given percentage change in
would be equivalentin effecton rto thischange mortalityor natality are nearly equivalent.
in fecundityare a 1.55 unit rigid translation Using Lewontin's example as a baseline, he
of the V(x) triangle to the left, a 2.20 unit found that as r declines below 0.05, it becomes
decrease in age at firstreproduction, a 5.55 less sensitiveto changes in age at firstreproduc-
unit decrease in age at peak reproduction,and tion (A), peak reproduction(T), and last repro-
a 21.00 unitdecrease in age at lastreproduction. duction (W). When r drops below about 0.025
r is most sensitiveto a given change in units in Meat's model it becomes more sensitive to
of time when fertilityis high and age at first changes in birth-ratethan to age at firstrepro-
reproduction low, and is least sensitivewhen duction. r shiftssensitivity fromT to birth-rate
fertilityis low and age at firstreproduction at about 0.030, and from W to birth-rateat
high, as Cole found. But, in general, r is most about .050. Thus, at low values of r,Lewontin's
sensitiveto changes in age at firstreproduction. resultsare qualified or reversed.
From this, Lewontin predicted that colonizing
r- and K-Selection,
or Bet-Hedging?
Since life-history
traitsare intimatelyinterre-
lated, attemptsto explain variabilityin life-his-
tory traits should model the important traits
and their interactions. Two theoretical ap-
proaches have begun this process. They have
led to contradictorypredictions.One approach
travels under the name "r- and K-selection";
V(x I call the other approach "bet-hedging."Advo-
cates of r- and K-selection deal with models
in which mortalityand fecundityschedules do
not fluctuate.Advocates of bet-hedgingtryto
A/0 deal withthe consequences of just such fluctua-
tions.
r- and K-Selection
D A T W The idea of r-and K-selectionoriginatedwith
Age Dobzhansky (1950), who proposed that natural
selectionoperates in a fundamentallydifferent
way in the tropics than it does in temperate
FIG. 4. LEWONTIN'S MODEL OF A LIFE HISTORY
V(x)= I.b. is, in general,triangular.A, age at areas. He argued thatin temperateareas physi-
maturity; T, turnover point;W,age oflastreproduc- cal factorsare most frequentlylimiting.These
tion;S,totalnumberofoffspring; x,age; I., probabil- actin density-independent fashion,selectingfor
ityofsurviving toage x; bx,averagenumberofyoung lower age at first reproduction and larger
producedbya femaleaged x.(AfterLewontin, 1965.) clutches(Schmalhausen, 1949). But, he argued,

in tropicalareas biologicalinteractionspredom- viewpoint,generated markedlydifferentpre-

inate,leading to selectionforabilityto compete dictions.Murphycompared two populations in
and to avoid predation. Skutch (1949, 1967) a computer simulation. Population 1 (r =
has made similar suggestions. MacArthur 0.3188, alpha = 3) had later age at firstrepro-
(1962) gave theoreticalsupport to these ideas duction, better adult survival rates, lower fe-
by deducing that, in density-dependentsitua- cundity,and a higher carryingcapacity than
tions,naturalselectionwillfavorgenes thathave Population 2 (r = 0.5472, alpha = 2). The two
a highercarryingcapacity,K, and bysuggesting populations were linked by Ricker's (1954)
that in such situations K can replace r, the reproductiveequations. When the interaction
Malthusianparameter,as a fitnessmeasure. On coefficientswere equal and there was no envi-
the other hand, Lewontin (1965) emphasized ronmentalvariability,the simulationmodel be-
that in colonizing situations(where limitation haved realistically.Population 2 (short-lived,
tends to be density-independent)age both at high r) increased more rapidlythan Population
firstreproduction and at the turnover point 1 (long-lived, lower r), then went extinct as
in reproductiveeffortshould be minimized,and Population 1 grew past the carryingcapacity
the clutchsize increased, in that order. of Population 2. Next Murphy altered the
MacArthur and Wilson (1967) drew these interactioncoefficientsso that the two popula-
ideas together and coined the terms "r-selec- tions coexisted in a stable environment(Pop.
tion" for selection in environments favoring 1 = 2364 individuals,Pop. 2 = 6757 individu-
rapid population growth,and "K-selection"for als). He started both populations with 1000
selection in saturated environments,favoring individuals.Then after26 timeunits,when they
abilityto competeand to avoid predation.Their had almost reached equilibrium,he introduced
discussion provoked the spurt of papers pub- uniformrandom variationin reproductivesuc-
lishedon the topicin the last fiveyears (Pianka, cess. In all cases, Population 1 increased in
1970, 1972; King and Anderson, 1971; Rough- numbers and dominated, while Population 2
garden, 1971; Hairston, Tinkle, and Wilbur, decreased in numbers, but did not go extinct
1970; Cody, 1971; Gadgil and Solbrig, 1972; (Fig. 5).
Abrahamson and Gadgil, 1973). This result goes counter to the r-selection
Table 3 summarizes the correlatesof r and argument,but seems to be supported by data
K-selection in the environment and in the on herring-likefish,whose reproductivespan
organsim(modifiedaccordingto Pianka, 1970). is stronglycorrelatedwithvariationin spawning
Mostof the relationshipsfollowin a straightfor- success (Murphy, 1968). Murphy's ideas are
ward fashion from the arguments given by quite similar to those presented in Fig. 3, a
Schmalhausenand Dobzhansky,or are outlined resemblance which argues that environmental
above where I considered the traitsone at a variabilityselects for reduced clutch size, and
time.The theoryis qualitative,not quantitative, his data suggest that the effectcould be quite
and admits comparisons only within limited strong. Note that variation in reproductive
groupings. But it does predict the association success is essentiallysynonymouswithvariation
of the biological traitsconstitutinglife-history in juvenile mortality.
tacticsinto two groups: (1) r-selection:
early age Schaffer (1974b) extended and reinforced
at first reproduction; large clutch size; se- Murphy'sconclusionswitha simple population
melparity;no parental care; a large reproduc- model. Consider a population without age
tive effort; small, numerous, offspring; low structure.Then
assimilationefficiency;and a short generation
X= B + P
time; (2) K-selection:delayed reproduction;
iteroparity; small clutches; parental care; where B is the numberof offspringthatsurvive
smaller reproductive effort;a few, large off- fromtimetto timet + 1, and Pis the probability
spring; and high assimilationefficiency. thatthe female herselfsurvivesto breed again.
Both B and P are functionsof reproductive
Bet-Hedging effort,E. In a fluctuatingenvironment,breed-
ing success and adult survivalvary from year
Murphy (1968) and Schaffer (1974b), who to year. In such circumstances,the long-term
examinedthesame trendsfromthe bet-hedging rate at which a population increases, X, is the

0~
U CO/) 4 c
Z ~~~~~~~c C
C~~~~~~~~~t
C~~~~~~~~~~~~~~~~~~/
biO 0
ct~~~~
t- t- N ~ ~ ~ COOC A
Ct(tc
C~~ C-~~~~ ~ CU> CZObQ
V C~~~~~~t /
I b~~~~~~~~~~~~~~~~~~~14c0
r 6V
1.4 ~ C -1C~-
V ~~~~~~~~~cO
0 -e~~~~~~~~~~~~
C ~~~~~~~~~~~C
0 o~~~~~~~~~~~~~~ .,07
Z CO Q)V ~~~~~~~~~~~~~~~C
~iCO.,V
Ct 0C H bc C
C C~~~~~~cO
C4-d nC
C-~~~~~~~~~~~~~-
~
-~~~~~~~~~
0 ~ ~ ~ ~ ~ ~ .
-e C~~~~~~~C/ /

- -
0
--
- t
: u~~~~ 1?-~~~~
-S -0
bD
C.
At g Q Q S i E _
CZS V
v~~~~~~~~~~~ Ut
4-
0~~~~~~~~~~
V V M ; V C-C/
Q-
-eV~~~~~~~~~~~~~~~~~~~~~~~-
JV CVt 7~
r a 0
VO - -
C V~~~~~~~~~~~~~~-
V~~~~~~~~~~~~~~~b -
u~~~~~
$-
cli 0
1-4 cPbL- (
CE:Z E,_
g, xi

SUMMARY OF MURPHY'S MODEL

POPULATION 1: long lived POPULATION 2: short lived
r=.3188 Ix mx r= .5472
ix Mx
1.0- / 2.0 1.0- 2.0
Ix mx, lx mx Ix I Mx, Ixmx

lx mx lxmx
-00.5-
0.5- 1.0 "-1.0
2 3 4 5 6 7 8 9 1011 12 23 4 5 6 7 8
AGE AGE
*-long lived short lived

8- _
6-
U) 16X
06
Cl)i 0-
9 0.
w 2-
z
0 Z2-6 =_
c 0-
o4- I ._ X
0 25 50 75 100 125 150 175 200 225

TIME INTERVAL
SIMULATION(averageof20 iterations)
FIG. 5. THE IMPACT OF VARIATIONS IN RECRUITMENT ON LIFE HISTORIES

Simulationof the impact of random variationin reproductivesuccess (juvenile mortality)on a long-lived,
slowly growing population, compared with a short-lived,rapidly growing population. At all three levels
of variationin reproductivesuccess (4x, 8x, and 16x), the long-livedpopulation becomes dominant about
75 time-intervalsafter the introductionof variabilityat the 26th time-interval.(Adapted from Murphy,
1968).
product of the various rates of increase in randomly distributed and occur with equal
differentenvironments(el, e2, ... en), raised frequency:
to a power qi equal to the frequency of the X2 = AgXb.
environment.Therefore,
= 1
A Al2 q3
Let s measure the departure of good and bad
3
2 ..
years fromthe mean. Then,
and
00 Xg= B(1 + s) + P,
In (X) = lqn (XA) (Levins, 1968). Xb = B(1 - s) + P,
and
Schaffer considered the simplest case, where
two environmentalstates, good and bad, are A2 = AgAb =(B + p)2 - S2B2

dA
For the optimum E, = 0, and
dE
dP dB dB
=-[1 -{s2B/(B+P)}]
dE dE d2
d
Since 0 < [1-{S2 B/(B + P)}] < I for s2 < dE ')\' y
(P/B + 1), optimal effort,E, varies inversely
withs (Fig. 6a).
On theother hand, ifenvironmentalvariabil- dE
ityaffectsadult, ratherthan juvenile, survival,
Xg= B + P(1 + s),
o fE 1
Xb = B + P(1 -s),
E
and
2 = (B+ p) 2 - S2 p2.
(B) ,wd
Thus,
dB dP dB
- [1-{s2P/(B + P)}]
dE dE
_dP
Now so long as 0 < [1 - {S2 P/(B + P)}] d dE
< 1, or S2 < (B/P+ 1), increased s favors in- dE
creased reproductiveeffort(Fig. 6b). (-s2PdP
To summarize Schaffer'sargument, a fluc-
tuating environment that has its impact on - - 2 '~' B-P'dE
juvenile mortalityfavorsreduced reproductive
effort,smaller clutches, and longer-lived or- o EEf 1
ganisms.But environmentalvariabilitythat af-
fectsadult survivalfavors increased reproduc- E
tive effort,larger clutches,and short-livedor-
ganisms. Table 4 contrasts the bet-hedging FIG. 6. THE IMPACT OF VARIATIONS IN MORTALITY ON
LIFE HISTORIES
predictionswiththose made by r- and K-selec-
(A) Juvenilemortalityvariable.When variations
tion.
in mortality
affectthenumberof youngthatsurvive
Cody (1971), Gadgil and Solbrig (1972), and to breed,B, thenthe optimumreproductive effort
Abrahamson and Gadgil (1973) have checked in a fluctuating
environment,Ef,is smallerthanthe
the predictions of r- and K-selection against optimaleffortin a constant
environment, E.
field data on birds and wildflowers,and they (B) Adultmortality variable.When the variation
are reasonably well borne out. However, in mortalityaffectsparentalsurvival,P, then the
Menge's (1974) workon intertidalstarfishshows optimalreproductive in a fluctuating
effort environ-
thatother factorscan complicatethe situation. ment,Ef,is largerthan the optimalreproductive
In starfishliving on the open coast, there is effortin a constantenvironment. s measuresthe
a trade-offbetween reproductive effort and departure ofbothgoodandbad yearsfromthemean.
adult mortality, caused (AdaptedfromSchaffer, 1974b;see text.)
specificallythe mortality
by fluctuationsthat create the "r-situation."
Life HistoryTacticsIn VariableEnvironments
Thus, organismsliving in a situationthat one
would expect to select for an increased repro- This section is organized on the basis of a
ductiveeffortactuallyshow a decreased repro- classificationof types of variabilityin physical
ductiveeffort(cf. Williams,1966b). I know of and biologicalenvironments.The criterionused
no published work presentingdata that would is that of the types of informationabout the
allow us to discriminatebetweenthe alternative future provided by the historiesof different
explanations listed in Table 4. Such work is environments(Fig. 7). With each typeof envi-
badly needed. ronmentalvariationthere are associated some

TABLE 4 springhatchingtimes(also see Williams,1975).

The contrasting ofr- and K-selection
predictions and
There has been no formal theoreticalinves-
bet-hedging tigation of optimal life-historytacticsin Type
1 environments.However, Levins (1968, Chap.
(1) r AND K-SELECTION AND BET-HEDGING WITH ADULT
2) has made a preliminarysuggestion that a
MORTALITY VARIABLE developmental switch that produces one
StableEnvironmentsFluctuating Environments phenotype above a certain threshold and an-
other phenotype below that threshold will be
Slowdevelopment Rapiddevelopment optimal in such environments(fine-grainedin
and latematurity and earlymaturity space, coarse-grainedin time). The aphids and
Iteroparity Semelparity
some open-watercladocerans fitthe prediction,
Smallerreproductive Largerreproductive
effort effort but the pelagic tunicates change life-history
Feweryoung Moreyoung tactics more nearly as a continuous function
Longlife Shortlife of food supply,withno thresholdphenomena.
(2) BET-HEDGING WITH JUVENILE MORTALITY VARIABLE Other cladocerans apparently produce
ephippia and enter a restingstage long before
Earlymaturity Late maturity
the seasons change, perhaps to avoid biological
Iteroparity Iteroparity
Largerreproductive Smallerreproductive rather than climaticstress.
effort effort Fleas living on rabbits have available a de-
Shorterlife Longerlife pendable signal predictingthe onset of a colo-
Moreyoungper Feweryoungper nizing phase: the estrus cycle of their host
brood brood (Rothschildand Ford, 1972). They have evolved
Fewerbroods Morebroods obligate synchronization with their host's
breeding cycle. (However, some questions re-
biologicalexamples and a discussionof whatev- main open. What happens to fleas living on
er theory has been constructedto date. The male rabbits?).Synchronizationis only possible
classificationis not logically exhaustive, but because fleas have a capacityforrapid numeri-
neitherare the types of variationencountered cal response that enables them to achieve
in nature. synchronywithin the period of the rabbit's
breeding cycle.
Type 1: Cyclic,Period Fixed and Long
(>>T)
Multivoltine temperate insects, temperate Type 2: Cyclic,Period Short (<T)
cladocerans, pelagic tunicates (Heron, 1972a,
b), and a host of other animals inhabitenviron- 2a. Cycle predictable
mentsin which regular periods of severe stress
are punctuated by conditions favorable for Large organisms living in stable, seasonal
colonizingepisodes. These circumstancesfavor environments encounter this situation, e.g.,
low age at firstreproduction,large clutch size, most terrestrialvertebratesand trees-also or-
and parthenogenesis during the colonizing ganismsthateitherliveor spawn in the intertidal
phase, followed either by (1) sexual mating at zone, e.g., many molluscs, barnacles, starfish,
the onset of stressto produce a diapause form crabs,and the California grunion.Since condi-
capable of resistingthe stress period (aphids, tions are predictable, the situation selects
cladocerans); or (2) developmentalplasticityin breedingat theoptimaltime,withlittlevariance
life-historyparameters,producing longer-lived in breeding time (Cohen, 1971). MacArthur
iteroparousforms (pelagic tunicates).The sig- (1968) has pointed out that where the cycle
nificanceof the timingof matingin the former the population
affectsage classes differentially,
is not known. Some persons think that pre- will evolve towards a state where its age dis-
diapause sex is adaptive because it increases tributioncycleswiththe environment,withthe
the genetic variabilitypresent in the overwin- most resistantage classes dominating at any
teringclutchofany givenparent.Added genetic time. Certainlypart of this synchronizationis
variabilitycould spread the risk of succumbing achieved by selection of breeding time within
to uncertain winter conditions and uncertain the cycle, but I know of no examples that

Type 1
z
0
I- -i T generation length
D
C) Type 2a
0 K
n-
z
Type 2 b
T
0
ry Type 2c
LL
z
o Type 2d
L>L T
0
>- Type 3
:D
TIME
7. A CLASSIFICATION OF FLUCTUATING ENVIRONMENTS
FIG.
By considering the variabilityand predictabilityof the environment,relative to the generation time of
a population, one finds six general classes. The shaded areas indicate regions of unpredictabilityin either
amplitude or period of fluctuations.See text for explanation.

28 THE QUARTERLY REVIEW OF BIOLOGY [VoLuME51
illustratethe more complex adaptation Mac- optimal tactic here is to spread the risk of
Arthurpredicted. germinating(or hatching)too soon or too late
For tropicaltrees that experience a long dry in the cycleby developing a within-clutchpoly-
season, Janzen (1967, 1971a, b, 1972a) posited morphism in age at firstreproduction or in
a tactic that channels effort into vegetative response to environmentalcues, or in both.
growthduring the wet season, when competi- More precisely,the optimal tactic consists of
tion for light is severe, then into reproduction generatinga distributionof hatchingtimes in
during the dry season, when pollination and theclutchthatmatchesthe historicalprobability
dispersal agents are most dependent on the distributionof the optimal date for reproduc-
trees as an energy source. Once the trees are tion. Palmblad (1969) and Marshall and Jain
roughlysynchronized,coevolved synchronyin (1970) have found support for this prediction
the pollinator and disperser populations will in weedy plants.
make it difficultfor trees to reproduce during
thewet season, on the one hand, or to decrease 2c. Start of cycle predictable, conditions un-
the variabilityin floweringand fruitingtime known
too far on the other. Perfect synchronywill
not evolve, because so brief a burst of energy Periodabout equal to lifetime. This situation
in time would not be worth the attention of characterizessome univoltineinsectand annual
specialized, tropicalpollinatorsand dispersers.
plant species towardsthe limitsof their range:
At the same time, some decrease in variability
on mountain tops, at high latitudes,in deserts
in fruitingtime is advantageous to swamp seed(Mayerand Poljakoff-Mayber,1963), where the
predators (Janzen, 1967, 197 lb). probabilityof a disastrous season in which no
Thus a dynamic balance of community-level reproduction is possible becomes significant.
forcesmay determinethe onset of the breeding The populations may be maintainedby migra-
season and the variabilityof reproductiveeffort
tion from areas with more predictable condi-
withinit,both among and withinplant species. tions. Variance in reproductive effortshould
The evolution of a predator-swampingeffect, be high and well correlated with variance in
followingimperfectsynchronyowingto optimi- environmentalquality.
zation of breeding timewithrespectto physical As pointed out above (p. 000 ff.), environ-
factors,can also be invokedto explain the mass,
mental conditionsthat vary fromyear to year
synchronousnestingbehavior (arribadas) of sea
selectclutch sizes both smallerand more varia-
turtles in Costa Rica and the islands of theble than the most productivesize. Some beetles
Indian Ocean (Richards and Hughes, 1972). and birds of prey have developed two different
Periodic cicadas provide another remarkable
methods of generating flexibilityfor dealing
example. There are three species, each splitwith temporal variability.Owls, hawks, and
intotwo races,one witha 13-yearand the othereagles lay theireggs severaldays apart, produc-
witha 17-yeardiapause beforeemergence. The ing clutches in which all the young survive in
larvae live in the soil, sucking on tree roots;
good years, whereas in bad years the oldest
the entire brood emerges within a period of progeny outcompete their siblings for scarce
a fewweeks. Lloyd and Dybas (1966) suggestedfood, and the clutch produces one well-fed
thatthe long larval phase, combined withlarge
fledglinginstead of threeor four starvedweak-
emergence numbers,acts to swamp predators, lings(Burton, 1973). Such flexibilitymaximizes
and thatthe 13-yearand 17-yearprime-number recruitmentin good yearswhileminimizingthe
life cycles evolved as means of escaping from
riskof totalfailurein bad years.Carabid beetles
shorter-livedparasitoids,which would have toof the genus Agonum,livingin isolated patches
achieve diapause periods of similarlength be-
of deep leaf-litterin marshes, use a different
fore any returnwould be realized. tactic for generatingflexibility, but it has the
same result.When poor conditionslower juve-
nile survival,the beetles decrease their repro-
2b. Startof cycle unpredictable
ductive effort.Since reproduction costs these
This environmentis inhabited by some uni- beetles somethingin terms of adult mortality,
voltine insects and annual plants. As Cohen the adults live longer, and normally annual
(1966) and Boer (1968) have pointed out, the individuals survive into the next season to

reproduce when conditions may be better tion in order to minimizethe riskof extinction.
(Murdoch, 1966). The power of this selectionfactorwill increase
Periodlessthanlifetime. Larger, longer-lived as population sizes decrease. In particular,the
species,such as tropicaltreesand birds,experi- proportionof progeny developing each season
ence these conditions. As Cohen (1966) and reflectsa balance betweenminimizingthe prob-
Boer (1968) have emphasized, under these abilityof extinctionand maximizingthe long-
conditionsit pays to produce a clutch smaller term growth rate. The wild oats Avena fatua
than the most productive (see Fig. 3). It also and A. barbatamay provide one example (Mar-
pays to increase the variance in breeding time shall and Jain, 1968, 1970).
among the progeny. Nidicolous birds living in
temperateareas should decrease clutch size as 2d. Start predictable, conditions partially
the season progresses in order to put more known
effortinto each chick and to fledge the clutch
in timeforwinter.About sixtybird species show Large, long-livedvertebrates,some trees,and
a decline in clutch size from the start of the some parasites and commensals inhabit this
season (Klomp, 1970), but whetherthe decline environment.One tacticis to skip reproduction
is adaptive is not known. Several authors have completelyif conditions indicate a bad season
suggested that the decline is not adaptive and ahead; and even, if reproductionis mistakenly
is produced by the smaller effortmade by late attempted, to resorb the embryo or carry it
breeders, who have usually already failed at over in a resting state until better conditions
one clutch,or are breeding for the firsttime prevail,as, e.g., in the case of the red kangaroo
(Klomp, 1970). Holgate (1967) and Murphy in the Australian desert (Short, 1972) or fleas
(1968) have pointed out that in such circum- livingon rabbits(Rothschildand Ford, 1972).
stances iteroparityand longer lifetimes are Cohen (1967) has analyzed diapause strate-
favored because they reduce the risk of the gies in this context. He predicted that the
animals gettingcompletely eliminated in one organismswill lock in on environmentalvari-
bad year. ables thatare correlatedwithfutureconditions,
In focusing on dormancy tactics, and will produce a probabilitydistributionof
Cohen, (1967) postulated that the long-term diapause times in their progeny that reflects
population growth rate reflects the balance the probabilitydistributionof breeding times
betweentheproportionofprogenygerminating predictedby the environmentalcues. However,
in any given year, and the proportion dying the interestsof parents and progeny may con-
during dormancy over the next year if they flict,for each of the offspringshould time its
do not germinate. In other words, spreading emergence from diapause to hit the peak of
the risk is advantageous, but how far the risk the breeding season, rather than arrive either
getsspread is determinedby a balance between early or late as the resultof parental optimiza-
the seriousness of the risk and the cost of tion (cf. Trivers, 1974).
spreading it. More precisely, a mixed tactic, Many temperateconifersare known to have
where only a fractionof the population repro- mast years (years of heavy seed production)
duces, resultswhen the variabilitiesof the out- alternatingwith years in which there is little
comes of reproducing and of not reproducing or no seed production (seed crop "failure").
increaserelativeto the differencebetweentheir Mast years are common in the Douglas Fir
means (Cohen, 1968). In a furtheranalysisof (Pseudotsuga menziesii) in British Columbia
the distributionof reproductiveversus vegeta- (Smith, 1970), and reach their most extreme
tive(somatic) growthin an unpredictableenvi- developmentin the monotypicstands of Pinus
ronment,Cohen (1971) found thatwhen condi- cembrain the Russian taiga, where there are
tions are uncertain, the optimal tactic is to 8- to 11-yearperiods between seed crops (Jan-
spread out the productionof seeds and leaves. zen, 1971a). At least three factors influence
This pattern is found in some desert annual masting. (1) It may take a number of years
plants. of vegetative growth to build up the energy
Mountford (1971) also emphasized the needed to reproduce withany chance of success
importanceto individualsof retaininga certain fora wind-pollinatedspecies thatmustproduce
proportionof progenywithdelayed reproduc- a large pollen crop. (2) Environmentalsignals

available at the startof a breeding season may ment. The conditionsduring the habitable pe-
occasionally indicate favorable conditions riod may be either predictablyfavorable,pre-
ahead. The pine trees would respond to such dictablyunfavorable,or unpredictable. Micro-
signalswitha strong,but incompletelysynchro- organismsinhabitingephemeral environments
nous, response. (3) This imperfectsynchrony are one type of organism encountering these
in mast years will be reinforcedby predators conditions,e.g. terrestrialrotifers,paramecia,
which eliminateseed crops set in offyears,but unicellularalgae, and other species dependent
are swamped by the amount of seed produced on rainfall for the opportunityto grow and
in mast years (Smith, 1970; Janzen, 197 la). reproduce. Inhabitants of temporary ponds
The annual fishesstudied by Wourms (1972) provide anotherexample. When conditionsare
provide a much better and more thproughly favorable, the optimal tactic should be rapid
analyzed example. He said: developmentand a total commitmentof avail-
able energy to reproduction that produces a
Annualfishescan maintainpermanentpopula-
tionsin temporary aquatichabitatssincethepop- restingstage. When conditionsare unpredict-
ulation survivesdry seasons in the form of able, a mixed tacticof producing some proge-
diapausingeggs.Populations persisteventhough ny that hatch at the firstencounterwithgrow-
subjectto erraticenvironmental cyclesand recur- ing conditions, others at later encounters, is
rentecologicalcatastrophes.... favored,e.g., by fairyshrimp.
Survivalstrategyis based on the "multiplier Plantsdependent on fireforseed release and
effect";i.e., interpositionintothedevelopmental germinationmeet these conditionson a larger
pathway ofthree branchpointscontaining diapause time scale. Tactics can take at least two forms:
stagesof prolonged,variabledurationgenerates holding the seeds on the tree untila firepasses
different
eight distributions
oftotaldevelopmental
time.Thus, a singleegg populationof identical through, e.g., lodgepole pine, Pinus contorta,
age can generateseveralsubpopulations, all of on the eastern slope of the Rockies (Smith,
whichdevelopaccordingto different schedules. 1970); or releasingseeds intothesoilbut making
A developmental programis establishedwhich germination dependent on heat shock, e.g.,
permits therepeatedlossofindividual eggsunder many prairie grasses. In both examples, repro-
conditions whichmayinitiatehatching butdo not ductiveeffortis storedin potentialform(seeds)
allowformaturation and successfulreproduction. when conditionsare favorablefor growth,and
The "multiplier effect"augmented byotheradap- development depends on a signal from the
tationsguaranteesthatsome portionof the egg randomlydistributedenvironmentalvariable.
populationwill surviveto reproduce(Wourms,
1972: p. 389).
Wourms's exemplary studies establish clear Timing The AmountAnd Distribution Of Effort
links between the details of the developmental The previous section has emphasized the
biologyof a species and the ecological problems problems of optimizingthe patternin time of
posed by the situationin which it evolved. We reproductionand diapause in thefaceof various
need much more work like his, but two types temporalpatternsofenvironmentalfluctuation.
of information would make his story more This section deals with the same problem in
complete. First, the genetics underlying the homogeneous environments.The question was
distributionof diapause typesshould be eluci- firstposed by Fisher (1930) in one of the most
dated; and secondly,the relationshipbetween frequentlyquoted passages of his often-quoted
the frequencyof diapause typesin local popula- book (Williams, 1966b; Taylor, Gourley,
tions and the distributionpatternsof rainfall Lawrence, and Kaplan, 1974):
should be established on a comparative basis
over a geographical range broad enough to to knownotonlybywhat
It wouldbe instructive
provide a diverse sample of types of climatic physiologicalmechanism a justapportionmentis
made betweenthe nutrimentdevoted to the
variability.
gonadsand thatdevotedtotherestoftheparental
organism,but also what circumstances in the
Type 3: Random In Time andenvironment
life-history wouldrenderprofit-
able the diversionof a greateror lessershare
No theoreticalwork on optimal life-history of the availableresourcestowardsreproduction
tacticshas been done for this type of environ- (p. 47).

I have built this section on a foundation matureone at a time.) Furthermore,Neill notes

provided primarilyby Williams (1966a, b). In that most temperate insects encounter few
his book, Adaptationand NaturalSelection,Wil- predatorsand ample food when theyfirsthatch
liamsdifferentiated betweenthe totaleffortthe in the spring,and thus violate the assumptions
adult puts into its progeny,on the one hand, Williams made in deducing his prediction of
and the partitioningof that effortinto a few viviparity.
large young or many small young. Svardson Gotto (1962) reviewed egg size and egg
(1949) made the same distinction,and antici- number in commensal and parasiticcopepods,
pated many of Williams's ideas, while Lack and suggested the followingstory.Where the
(1954) mentioned the trade-offbetween size host is abundant, readily accessible, and easy
and number of young in discussingclutchsize. to find, and where environmentalconditions
I willfirstreviewtheideas and evidence relevant are stable (calm, sheltered waters),then a few
to the partitioningproblem, then cover the large eggs will be favored,because larger eggs
question of dividingresourcesbetween somatic produce largerlarvae,whichhave a competitive
growthand reproduction. advantage in the scramblefor the few available
sites in a saturatedenvironment.On the other
A Few Large vs. Many Small Young hand, where the host is sparsely distributed,
inaccessible,hard to find,or highlymobile, or
Let us assume that the eggs should be laid where the host's habitat is unstable or difficult
at a size which yields the maximum growth to penetrate (the intertidal, wave-beaten
rate on the parental investment.The effective shores), then many small eggs will be favored,
growthis theaverage growthrateof individuals because theycan sample more of the environ-
towards reproductivematurityminus the loss ment and reduce the risk that the clutch will
to mortality.In a population thatis not growing fail to produce any mature progeny.
rapidly,the numberof offspringdecreases over Gotto's review of the data supports this pic-
timeowing to mortality,and unless the growth ture on the whole, although several poorly
rateof the survivingoffspringmore than makes understood species break the main trends.
up for the loss, the net yield on reproductive Gadgil and Solbrig (1972) would call the group
effortwill be negative. If progeny can grow producing a few large eggs K-selected, and
fasteras larvae outside the parent (when re- those producing many small eggs r-selected.
sources for the young are abundant, and They have suggested that a similar situation
predation pressure is low), then many small obtainsin wildflowers.Price (1973) noted simi-
progenywillbe favored.If resourcesforyoung lar trends in ichneumonid wasps, a family
are scarce, or predator riskto small size classes parasiticon insecthosts.Differentwasp species
is high, then the parent will tend to produce lay theireggs in differentstages of theirhost's
a few large progeny. In such circumstances, development.Those layingon earlyinstarshave
live-bearingwill be favored over egg-layingif many small eggs; those laying on later instars
internal fertilizationis possible, because live have a few large eggs. Price suggests that this
young do not experience the start-uptime of trend can be explained in terms of adjusting
eggs thathave to develop beforetheycan begin the fecundityto the probabilityof survival.To
assimilatingenergyand grow,and because live explain the trend, one could also invoke the
young generally experience lower mortality ideas of Williamsand Svardsonon the competi-
rates than eggs do. (This is an abstractof an tive advantages of a few, large young. There
argumentgiven by Williams,1966a, Chap. 6). is a correlationin mammals between increased
In the light of this argument, Williams was parental care and fewer,larger progeny. Wil-
puzzled by the low incidence of viviparityin liams (1959) showed,that this correlation did
insects,and its absence from birds. However, not exist in the darters,where only the males
Neill suggests (pers. commun.) that viviparity care for the young, and was at least clouded
will not be advantageous to insects that over- in other fish by the confounding correlation
winteras eggs, nor to birds,which must mini- of egg numbers with body size, which is not
mizeweightin order to flyproperly.(Most birds necessarilyrelated to parental care.
will carryone heavy egg in flight,but I know Janzen (1969, 197 1b, 1972b) and Harper,
of no bird that carries a full clutch. The eggs Lovell, and Moore (1970) have reviewed the

problem of what determinesseed size and seed how much effortshould the individualput into
number in plants. Most plants,especiallyin the vegetative as against sexual reproduction at
tropics,exist in a complex coevolutionarycon- differentages and under differentconditions?
text,contendingwithseed and seedling preda- And whyis sexual reproductionassociated with
tors (chieflyfungi, viruses, insects, mammals, dispersal and withresistantpropagules? There
and birds),and benefitingfromdispersalagents is a growing literaturethat focuses on these
(wind, water, insects, birds, and mammals). questions, and since the problems of sexual
Optimizing seed design for one purpose does versus asexual reproduction are complex
not necessarily optimize it for another: an enough to place them beyond the scope of this
intermediate compromise must be reached. paper, I referthe interestedreader to Maynard
(Some plants produce a mixtureof large and Smith(1971), Williamsand Mitton(1973), and
small seeds, but even here decisions must be Williams (1975).
made on dividing resources between the large Sarukhan (1974) and Sarukhan and Harper
and the small seeds-cf. Harper, Lovell, and (1973) studiedthepopulation dynamicsof three
Moore, 1970). For example, many small seeds species of buttercups.Two species could repro-
are favored both for wind-dispersal and for duce vegetatively.Almostall recruitmentto the
avoiding attack by beetles that lay their eggs population on the part of one, Ranunculus
in theseeds. If a seed can be made smallenough, repens,was owing to vegetativereproduction,
it maybe too smallfora larval beetle to develop whereas, of the other two species, one repro-
in it. But large seeds with thick, hard coats duced strictlysexually (R. bulbosus),and one
can escape many seed predators,and have the R. acris,bya mixtureof the two means. Vegeta-
energyreservesto send up a large,competitively tive reproductionproved much more depend-
superior seedling. Many plants produce seeds able than seed production as a means of per-
impregnatedwith noxious chemicals, presum- sistinglocally. The species that put most of its
ablyat some cost.There is considerableevidence effortinto vegetativereproduction, R. repens,
that this also reduces seed predation (Janzen, also produced a small seed crop that had a
197lb). long dormancy period in the soil. The two
There are difficultiesintrinsicto the give- sexually reproducing species produced large
and-take of coevolutionary interactions that seed crops, practicallyall of which germinated
make detailed predictions difficultand the in the next year. It seems likelythat the alloca-
sortingout of the post hoc, plausible explanation of effortbetween vegetative and sexual
tions commonly found in the literaturewell- reproductionwillbe stronglyinfluencedby the
nigh impossible. A great deal depends on the probabilityof local extinction,competition,and
exact nature of the species and the community the problemsassociated withdispersaland ger-
involved. Switchingfrom one size of seed to mination. More exemplary studies like Saruk-
another to escape a size-selectivepredator may han's are needed to enable one to assess the
be no advantage if there is another predator factorsaffectingthisdecision.
in the communityready to switch onto your
seeds as theychange in size. Timing The Amount of ReproductiveEffort
Many plants and some animals (e.g., Hydra
and other coelenterates)can reproduce vegeta- Williams(1966a, b) was the firstto formulate
tively. As Williams (1975) points out, sexual clearlythe problem facinga species or popula-
reproductioncan have several advantages over tionof decidinghow much of currentresources
asexual reproduction. It always generates to commit to reproduction now, inasmuch as
variabilitythrough genetic recombination,and a commitmentof resources to the present car-
usually has evolved to permit dispersal and ries with it some cost to the future. His rule
create a propagule that can be made resistant for optimizing current effort is to increase
to harsh circumstances.Vegetative,or clonal, current commitmentso long as the current
reproduction, on the other hand, preserves profitsmore than outweigh the future losses
successful genotypes, placing progeny in an caused thereby.Table 2 presentsevidence that
environmentthat has been tested and found in the fewcases studied,a greaterreproductive
to be favorable-assuming, of course, that effortdoes increase mortality.Thus, according
conditions do not change. Given that option, to Williams,since reproductivevalue decreases

beyond a certain point with age, reproductive thatmakes the maximumpossible reproductive
effortshould always increase with age. Gadgil effortat the earliest age at which that size is
and Bossert(1970) derived the same conclusion reached. Since reproductivevalue declines with
fromtheirsimulationmodel, but Fagen (1972) age in a growingpopulation,F' (w) willbe lower
constructeda model animal forthe Gadgil-Bos- than B(w) at greater sizes. If reproductive
sert simulation for which reproductive effort potential increases slowly with size, then the
firstdecreased, then increased with age. As size of maximal contributionto fitnesscan be
Schaffer(1974a) has rightlypointed out, this much lower than the size of greatestreproduc-
example reveals the weakness of simulation tive potential.
studies. The resultsare only as general as the For a big-bang reproducer, reproduction
examples used, and it is hard to be sure that should occur at the age and size corresponding
you have explored the parameter space to the peak in the F'(w) versus w curve. For
thoroughly enough to talk about general repeat-reproducers,growthshould continue as
conclusions. long as reproductive potential increases with
If clutch size is taken as proportional to size, stoppingwhen reproductivepotential has
reproductiveeffort,we findthatfield evidence reached a maximum. Sexual maturitywill not
regardingclutchsize in birds shows that repro- be postponed beyond the peak in the F'(w)
ductiveeffortdoes increase withage, with the versuswcurve,but mayoccur earlier,especially
exceptionof the partridgein England (Klomp, if the slope of the curve is gentle. When this
1970). Or if we take the ratio of ovary weight happens, growthwillcontinuebeyond maturity,
to the 2/3 root of body weight (to correctfor and reproductive effortwill increase rapidly
changes in the ratio of digestive surface area with age. Gadgil and Bossert predict that pe-
to body mass) as being proportionalto repro- rennial plants and fish that show this pattern
ductive effort,we then find that reproductive should have a gradual rise in the F'(w) versus
effortincreaseswithage in several fish,Hippo- w curve. But where the contributionto fitness
glossoidesplatessoides,
Melanogrammus aeglifinus, at a given size, F'(w), increases rapidly with
Clupea harengus,and Clupea pallaseii (Gerking, size, growth should cease at maturity,repro-
1959; Iles, 1974). These definitionsof reproductive effortshould be high from the very
ductive effortare inadequate because repro- beginning,and reproductiveeffortshould in-
ductive effortshould properly be considered crease only slowlywithage. Gadgil and Bossert
a rate phenomenon: the rateat whichresources predict thatbirds and mammals that show this
in excess of maintenance requirements are pattern should have a sharp rise in the F'(w)
divertedinto reproductionratherthan growth. versus w curve.
In theabsence of bettermeasures,theydo serve The delayedeffectsof reproduction.Schaffer
to indicate the gross outlines of trends that, (1972, 1974a) looked in detail at the interaction
it may be hoped, will be more fullyand more of a reproductivecommitmentat one age with
rigorouslyinvestigated. the optimal tacticsfor subsequent age classes.
First,in a simple model where he assumed that
Theory survivalrates do not change withage and that
fertilitygrows geometrically,Schafferderived
Gadgil and Bossert (1970), Schaffer (1972, X(E) = b(E) + p(E)g(E)
1974a), and Taylor et al. (1974) have looked
at differentaspects of reproductiveeffort,all fromLotka's equation, where b(E) is the birth-
of them startingfroma focus provided by the rate of individuals in the firstage class, p(E)
work of Fisher and of Williams. is the survivalrate,and g(E) is the rate at which
Growthand reproductive effort.Gadgil and fertilitygrows with age. Clearly, the optimal
Bossert (1970) defined reproductivepotential, effortwill depend on the shape of the b(E)
B, at a givensize, w,as the numberof offspring and p(E) g(E) versus E curves. Schaffer ana-
produced by an organismof size w that makes lyzed these curves graphically,and concluded:
the maximum possible reproductive effort. (1) If both b(E) and p(E)g(E) are convex, the
They defined the maximum potential contri- optimaleffort theanimalseither
isall-or-none;
bution to fitnessat a given size, F'(w), as the commit suicidein reproducing,
or skipa year.
numberof offspringproduced by an organism (2) If both b(E) and p(E)g(E) are concave,the

optimal effort is intermediate(regular b3 = C-pi

V-+,,
iteroparity).
(3) If b(E) isan increasing
sigmoid,and p(E) g(E) where A, B, and C are defined by the last two
is a decreasingsigmoid,the optimaleffort equations. Thus, as Goodman (1971) showed
variesaccordingto theinitialconditions;the for the comparable continuous case,
animalseithermakea totalcommitment (big-
bang suicide),or put forthan intermediate Sb3
levelof effort (iteroparity). ,= -V3+1
In a more complex model incorporatingage

structure,Schaffershowed that optimizingthe This differentialequation defines the locus of
reproductiveeffortis equivalent to optimizing points(b,,p,) such thatr,the population growth
rate, is constant. Now consider trade-offsbe-
b,+p,g,V(i+ 1) tweenbirths,b,and survivalto thenextbreeding
where bXis the numberof female progenyborn season, p, as graphed in Fig. 8. The straight
to a female of age i, pXis the probabilityof lines representthe familyof solutionsto
survivingfromage i to age i + 1, gXrepresents
the factor by which fecundity at age i + 1 8bI
- = -V3+1
exceeds fecundityat age i, and V(i) is repro- bPI
ductive value at age i.
r increases as the lines march out to the right.
That is, the optimal life historymaximizes,at
Thus r is maximized when the trade-offcurve
everyage, the sum of presentbirthsand future of births versus survival intersectswith that
expected births. In essence, the evolutionary member of the familyof straightlines
rule in both models is always to optimize the
currentreproductivevalue based on expected b3= C-Pi VI+1
returns from current reproductive effort:
that is farthestto the right. (I am gratefulto
young born this season plus expected returns
Ric Charnov for showing me this derivation).
fromyoung born in futureseasons, discounted
Schafferwent on to analyze graphicallythe
back to the present,and weightedbythe proba-
coevolution of age-specificeffortsin a three-
bilitythat the parent will surviveto give birth
stage lifehistory.Certainshapes of the relation-
to them.This resultwas anticipatedby Williams
ship betweeneffortat one age and reproductive
(1966b), Hamilton (1966), and Emlen (1970),
effortat the next generatetwo stable equilibria,
all of whom stated it indirectlyor implicitly.
one predicting semelparity,the other itero-
Schafferwas the firstto state it directlyand
parity. Which one evolves will depend on
to begin an analysisof its consequences.
the initialconditions,i.e., the reproductivetac-
To see how this conclusion followslogically
ticsof the ancestor. Schaffersuggeststhat this
from demographic assumptions, consider the
may explain why iteroparous steelhead exist
following reasoning. Using Lotka's charac-
side-by-sidewith semelparous salmon (cf. Fig.
teristicequation,
9). An alternative explanation is that se-
00
melparityhad to be adaptive to be evolved,
1 = E b,I,e-r%
0
but need nothave been adaptive to be retained.
we isolate a single age group, j, "Once all of the life cycle that followsthe first
reproduction has completelyatrophied there
l-l
I1 = bt 1,e-rt + 13.1 P3e-r3 (b3 + pi VI+1)
would be no way of re-establishingit. The
0
change from iteroparityto semelparityshould
Now consider the effectsof varyingbirths at be irreversible"(Williams,pers.commun.). Both
the given age, j; that is, we ask the question, explanations are good examples of the plausi-
"How rapidly does future worth change, and ble, but not testable,interspecificcomparisons
in whatdirection,as a resultofchangingcurrent discussed in the introduction.
reproductivecommitment?"All termsthat do An analyticapproach. Taylor et al. (1974)
not contain b, or p,are constant,so have established several potentially useful
theorems concerning populations in a stable
1 =A+B(b?+pIVI+,)
age distribution.Dealing with Lotka's charac-
and teristicequation as a definitionof r, and with

Fisher's definitionof reproductivevalue, they

proved that:
1. Maximizing thereproductive valueat age zero
is mathematically equivalentto maximizing the
ultimate rateof increase(p. 109). /b= -C PJVJ
2.... in order to maximizethe reproductive
valueat age zero,thereproductive valuemust
be maximized at everyage,takingintoaccount
the organism'sdevelopment up to thattime.
Thisprincipleofsuccessive lifehistoryoptimi-
zationmustresultinthefitteststrategy(p. 111).
In applying their results to life-historypat-
ternsthatincorporatethe effectsof growthand
the costs of reproduction,theyestablished,as 0 P 1
did Schaffer (1974a), that if the number of
offspringis a convex or linear function of
pJ
reproductive effort, then the organism will
either grow or reproduce, but not both, i.e.,
it will adopt a big-bangtactic.Taylor et al. also FIG. 8. A GRAPHICAL ARGUMENT ON OPTIMIZING
REPRODUCTIVE EFFORT
established that if the energy available for
Considera lifehistory in whichwe isolatea single
growthand reproductionis a concave function
age class,j, and allowbirthsand adult survivalto
of biomass, and the mortalityrate is constant, varyforthatage classalone whileholdingthe rest
then selectionfavorsdeterminategrowth.This of the lifehistoryconstant.Then thatcombination
second result refines Gadgil and Bossert's as- ofreproductive as reflected
effort, inbirths(b3),and
sertionthat determinategrowthis a result of adultsurvival tothenextbreedingage (p), isoptimal
a life-historypattern showing a rapid rise in whichmaximizes b,+ V3+l. That pointcan be deter-
reproductivepotentialwithsize. minedgraphically, as shownin the figure,but the
actualvalues,and whetheror not thereare one or
morecombinations of b,and p,3willbe determined
Evidence bytheshapeof thecurve.
Williams(1966a, b) has suggestedthe follow-
ing correlatesof increased reproductiveeffort: turity,small clutches, and long reproductive
a largerratioof progenyweightto female body spans, e.g., raptors and large seabirds. Also,
weight,more clutchesper season, brightercol- for wildflowers,Gadgil and Solbrig (1972)
oration in breeding males, more lengthyand demonstratedthat plants growingin disturbed
elaborate courtship displays, more territorial sites devoted greater effort to reproduction
and aggressivebehavior,viviparityratherthan than did plants growing in undisturbed sites.
oviparity,and more parentalcare. Tinkle (1969) By making appropriate assumptionsabout en-
demonstrated trends in all these parameters vironmentalconditions and the shape of the
that showed, in general, that early-maturing curves relating reproductive effortand adult
lizard species were showing the other signs of survival,one can explain all these observations
increased reproductive effortthat one would as optimizationof currentreproductiveeffort
expect fromWilliams'spredictions. plus expected future returns. In none of the
Tinkle, Wilbur,and Tilley (1970) found that above cases, however, was the shape of the
early-maturing lizards also tend to have several trade-offcurve measured.
smallclutchesper season and larger per-season Schaffer and Tamarin (1973) showed that
fecunditiesthan do the late-maturinglizards, population cycles in voles could be explained
which produce a single, large clutch. These as resultingfromthe lagging of changes in the
general trends are matched in birds (Cody, optimal point for reproductive effortslightly
1971), whoselife-history tacticscan be separated behind density changes, the proximate cause
intotwogroups: (1) thosehavingearlymaturity, beingan increasein juvenile mortalityas density
large clutches, and short reproductive spans, increased. In essence, they showed that their
e.g. passerines, and (2) those having late ma- model could account for the observed fluctua-

been wellenough confirmedto be worthpursu-

--*1 ing. Now we need measurementsof the actual
Semelparity
trade-offsbetween adult survival and repro-
ductive effortto generate the shapes of the
curvesnecessaryto testthe graphicalarguments
I \sz~~~b, =E- p,V,oI put forth by Gadgil and Bossert (1970) and
by Schaffer (1974a). Until someone actually
measures the shapes of the curves in the field,
we will be left with unsatisfactory,post hoc
explanations.
~~~~bp,
DISCUSSION
- - - - - - - _,
This review has dealt with a large volume
r-- - --
Iteroparity or of material,and a discussion could take many
Semelparity Semelparity i directions.I shallrestrictmycommentsto three
areas. First, I will criticizea recent trend in
thelife-history literature.Second, I willexamine
Iteroparity what conclusionswe can draw fromthe review
bL about the general evolutionarycauses of ob-
servedtrendsin life-history phenomena. Third,
I will ask what implicationsthe review has for
Y; - p,Vo the design of theoreticaland experimentalre-
search projects.
pL or Empiricism?
Scholasticism
Since the approach taken in several of the
FIG. 9. A GRAPHICAL ARGUMENT ON REPRODUCING
papers I have reviewed (e.g., Cody, 1966; Mac-
ONCE OR MANY TiMEs
Arthur, 1968; Gadgil and Bossert, 1970;
Schaffer's(1974a) trade-off model predictsthat
Mountford,1971, 1973) representsan unfortu-
under certainassumptions about the shape of the
curvebetweenreproductive
trade-off effortandadult nate trend in ecological studies, I am going
eithersemelparity
survival, or iteroparity
mayevolve to criticizethem quite severely. In so doing,
dependingon initialconditions. The upperexample I criticizethe trend as much as the ideas in
showsconditionsunderwhichonlysemelparity will the particular papers that exemplify it. For
evolve;the lower,either.bt,progenyborn at age example, Gadgil and Bossert presented a
i thatsurviveto reproduce;pt,probability thatthe number of ideas (e.g., that of F') that, while
parentwillsurviveto breedagain; b, - p, trade-off charming and plausible, generate predictions
curveof birthsvs.adultsurvival.See textforderiva- that are absolutely unfalsifiable. Therefore,
tion of b. = C -P V,+1. theydo not constitutescience as I understand
it. Recently,manypapers have appeared which
tions, but that it was not mutually exclusive disregard a consideration of hard evidence in
of othersproposed. Furthermore,Schafferand preference for a discussion of ideas for their
Elson (1975) sh:owedthatmuchof the variability own sake.
in age at firstreproductionin Atlanticsalmon To an alarming degree, the attitude that
could be explained in terms of a trade-off sanctions such speculations in a vacuum of
between the cost of (1) reproductiveeffortas evidence spurs two additional objectionable
measured by the length of the river,and (2) trends.First,theoriesaccumulate, few of them
futurereturnsas measured by the growthrate formulatedin common terms,much fasterthan
at sea. evidence can be assembled to test them. The
result is a confusion of untested ideas which
Summary are judged, not on their abilityto withstand
The theoriesarising fromWilliams'sconjec- empirical tests, but on the difficultyof the
ture have proved to be provocative and have mathematicsused or the obscurityof the theo-

reticaldevelopment. Clearly, such criteriaare The standard organism-environment dichot-

inappropriate. Second, that such papers are omy used in the previous paragraph dan-
acceptedand publishedspursa casual disregard gerously oversimplifiesthe situation. To the
forthehard workinvolvedin rigorousempirical population,the definitionof the "environment"
tests.Not only do theoriesaccumulate, but the depends on the biological characteristicsthat
manner of their accumulation decreases the the population has already evolved. As one
likelihoodthattheywillbe tested.An enormous characteristic changes under selectionpressure,
amount of effortis being put into the develop- e.g., age at firstreproductionand withit gener-
mentof ideas for whichno one has established ation time,the definitionof the "environment"
connections with the real world. If the field changes, perhaps from irrelevantnoise to sig-
is to progress, we must get away from the nificantfluctuation,or from being unpredict-
practicallyScholasticapproach surfacingin such able to being effectivelyconstant.Such changes
papers, and get back to rigorous empiricism. in the definitionof the environmentwill affect
Theoretical work should be carefullydone profoundlythe furtherevolutionof othertraits.
and clearly communicated,and should result For any individual trait,the definitionof the
in predictionsthatare at least potentiallyfalsi- ''environment"includes the current states of
fiable. In all cases, alternative explanations all other traits in the population. Traits co-
should be impartiallyconsidered in the light evolve. As one trait changes under selection
of the best available evidence. That papers that pressure froma classic externalenvironmental
do not meet these criteria are getting past variable, such as an increase in population
referees and editors disturbs me, for it may density,other traitswillundergo compensatory
mean that the nature of science is broadly or opportunisticchanges owing to the redefini-
misunderstood. Our purpose is not to have tionof theirrelative"environments."The work
beautiful ideas, or profoundly obscure ideas, of Lewontin (1965) and Meats (1971) has pro-
but to arriveat explanationsthatmeet the most vided an indicationof wherethisline of thought
rigorouschallenges thatevidence and logic can could lead.
pose. Thus, two factors complicate the simple
cause-effectrelationshipin termsof which we
PatternsOf Causation usually think. In general, several evolutionary
causes, possiblyoperatingat the same time,can
Tables 5 and 6 present an abstract of the produce the same effectin a population's life
hypothesesreviewedabove. Two patternsstand history.Furthermore,a givenexternalenviron-
out from the tables. First,for any given trend mentalfactormaywell affecta numberof traits
in life-historytraits,e.g., an increase in clutch thatwill,as theychange, redefinethe "environ-
size, thereare several plausible hypotheses,not ment" and thus the "optimumtactic."
mutuallyexclusive,thatcould explain the trend, Withinthe biological community,there is a
either singly or in combination. Second, al- subterranean split between those who believe
though there is no standard method for ap- that for every phenomenon there is a single
proaching the evolution of life histories,dif- cause at a given level of explanation,and those
ferentauthors repeatedly emphasize the same who believe that there can be multiple causes
points.The importantlife-history traitsare both for certain phenomena operating at the same
the mean and the variance in age at firstrepro- level of explanation. Because thissplit is rarely
duction, clutch size, size of young, number of articulated,it can block intelligentdiscussion.
clutchesper lifetime,and inter-broodinterval. Furthermore,people who prefer to deal with
A givencombinationof these traitsis a life-his- ideas one at a time may not recognize the
torytactic.The evolution of life-history tactics potential importance of multiple causes. The
is stronglyinfluencedbythepatternof variation problem is serious; it erects mental blinders
in space and time of relevant environmental thatlead researchersto argue withthe mislead-
variables. Among the multitude of environ- ing consequences of unarticulatedpresupposi-
mental factors,those most generally relevant tions, rather than deal directlywith the pre-
to lifehistoriesare food, temperature,breeding suppositionsthemselves,with the logical diffi-
sites, refugia, competitors, and predators. cultiesof certain theories,or withthe admissi-
However, in specific cases other factors may bilityof givenevidence. Althoughthis problem
dominate. is certainlynot unique to life-historytheory,

i:
u
z 00
v
>1 ct C%
xn
v 4-J
4-J CZ CZ
CZ
-14 r- U
u r. -14 : - Z 01
CZ 0 CZ CZ CZ CZ
CZ
> 4-J 0
> CZ
4-i
v v CZ
0 4--J >
CZ v CZ
t4-4
z
>
Q Q u
0
v
> V u v CZ
4-J
CZ 'IC z
r. 4-4 0
CZ u 0 u >1
- "
0
r.
4-.4 t4--l 4"J 4--j j
CZ
6r4 r) 0 4-J
CZ -4
cld CZ 0
Itz
u
4-J 4-J
bo ;:I
4-J 4-i 0 v 14.4
CZ "
in.
z
CZ 0 lz 0.1
a, v
U 4-J CZ .-
> 0
1
A4 Ct V Z z
CZ q; v CZ V -
v
4-J
v CZ 0 CZ v -0 m
44- > li 4 0
4-i
0 0 u 0
-g --( "
CZ 4.j 4-J
C's
> C4-4
0
4-i >
CZ 0 0 CZ
CZ
Zn >, v $n. m 0
v 4-J 4 CZ 4-J
0 CZ
CZ c- CZ
CZ CZ > 7z
v
CZv CZ v C,3 0
> 0 CZ
v 0 in. z u CZ
CZ v
> 0
r. N" CZ
0
v Z
0
CZ
z
c-'s0 CZ
0
CZ 4-;
r. v '4-J 0 bo Ou u 0 V t W .- ,
Cs.- CZ 0
CZ CZ 4-J
Q CZ V u
4-J v , v
od -4-J . r.
CZ CZ
sn U 4-J u
u U
ct 4-J czt CZ u
z
C> cn C z
0
>
CZ r.
4-J
4-J
v . 0
r.
> >
4-i
4-J
z
w z
U) CZ CZ
0 od 0. 4,J
CZ -
C4-1
z
0
CZ
O
CZ u u
biD 0
z
u 4-J
4-J Q I."
v .-4-> CZ u od
4-J
0
.O 0 CZ
u CZ CZ
4-J 4-J
ad
z E 0
Q r. E
t4-4
4-i
4-J 0 r 4-J
CZ
Ci CZ
t4.4
in. z .-Z .-" U u
v z t4-4
> v r
44 z -
4-J
V
1; v ct ..4-J 0I
0 Z ct
CZ
u
z cli CZ
4-J
z v CZ C's v
C's
v z
V 4-J 0 CZ v CZ
0 0 4-J
4-J
4-J 0 0 4-J v CZ
> CZ 4-J
4-J z 4-J (4-4
>
14-0 v
> 4) Q CZ 4-J > t."
CI. CZ
> CZ v u
>1
4-J
z 4,J
CZ CA v " Z
c) - 0 - C).
CZ.- cn 0
CZ CZ Z > CZ v u 41
Z
4--l CZ 0
Ct > z
V) CZ
ct v m Qd V '0
V CZ
v -lc CZ z
CZ 4-J v
v 4-J O Z 4
4.j
4-J CZ
CZ v C-Z Z
4.4 > C-z
ct CZ
4-J b.0 CZ
> ct Od 0
CZE >
.4-J
CZCZ
CA
CZ
u
z
4-i CZ CZ ct
od CZ U
V
" 4-J
od
v cn c) u
-
4--l CZ
U
> u CZ U U Q Z -uZ
z od r. r.
4-J
c)
CZ V CZ Q
C d C
NHJISAS -IvsflvD wailsAs -IvsflvD

C Q0(CQ
C
*Co) Ct V
4- 0'
-~ CtN V C-O
C0 u V
C C~~~~~~~~~~~~~~~C
c C
C b)VC
u
0 0~~~~~~~~
-e~~~~C >~~~~C . V0 -~ C V
CZ CV > V CZ~
~~ ~~~-~~~o0 V
Cd
U
CCbO0~~~~~~~~~
- C
0 C
o-C
~~~~~~
b
~~~~~~~~ ;C C b~'0 bi
o.. ~~~~~ ~~~OCC~~0 0 C J
VC's ZCZ C
4CH ; 0 C OH C
0 7~~~~~ 7 ~ 0
C-.C
Vb 0 -z~
0 ~ ~ ~ V
4-J ~ ~ - ~ ~70 ~ ~ u-0

.C
0 Q0 C
ObCOu0 v
CZ
cn z
z~0
w 04J
Vt bo CA
z ~~~~~~~~~~~0
-CO
U' HC~~~~~~~i
-C' U' C Jcl -C
S
0C
CT
CV C CC CC-Z
OU C U> -e 4-C
4-i 0
CV 0 u
C C b
OC~~~~~~~~~~~~~~~~~, CCf
CV UOC.C~~~- CD
-Tp ~~~ C
~~~~~V CV. . C
~~~~ C~~~~~0-.-CZ
bl
. V C C
0.~~~~~0 ~ ~ ~ ~ H ~~~~~
'C~~~~~~~~0C~~~1
54- U
~~.VV C C ~~~~~~~~~~C~~~CO
CC~~~bi
0 C I 0 0.~~~~~~~~4J
COV'S
ct z CZ U
V ~~~~~~~~~~~~~~~~~~~~~~~~~~~~
C~~~~~~~~~~~0U 0
IaTIlSAS 1VSflVD WJ'~TSASIVSI71VD

C -~
7 N~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~1
~~~~~~~~ 4
~ ~ ~ ~ ~ ~ ~
C,
C~ 7
bP,
NN-
N
-~~-~ ~~~~~~~~~~~~~~~~~~~~~~bi
~ i
>~ -NCZ v
bi-7 CZ ~- r
Z
7 0
"Z 0
_0~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ C130 - C C13 o
Ct -C biO ct z~~~~~ZNN -7;

A 0. ~7 Ij - - &
7 ~ ~~ ~ ~ ~ ~ ~ ~ ~~~-7~
0 t

nowhere does it cause more problems. ing outlined above will remain suspect. The
For example, we should remain open to the root of the problem lies in the need to define
possibilitythat the assemblages of traits,the the fitnessof individual organisms,and in the
tactics,predicted by r- and K-selection or by destructionof individual genotypesby sex.
bet-hedging,or theirdescendants,are not iron- For it makes no sense to speak of genotypes
clad entitiespermittingno exceptions. In na- seeking to perpetuate themselvesin a sexually
ture, organismsmay respond with some traits reproducingpopulation, since the processes of
to maximizing r, with others to bet-hedging. segregation and independent assortment of
Both causal systemsmay operate at once, with genes among the progeny continually break
differentoutcomes, on balance, for different down and reshufflethe genotypes into new
traits.The concept of "tactic" as a coevolved combinations. Since the individual organism
assemblage of traitsthat must alwaysbe found acts for a multitude of genes, it does make
togethermay prove misleading. some sense to speak of organisms seeking to
There is another, more serious, logical diffi- perpetuate theirgenes, if not themselves.This
cultywith most of the theoreticalmodels re- compromisehas logical faults,but seems to be
viewed in this paper, and with evolutionary the best we can do at present.
theoryin general. As Kempthorne and Pollak Because fitnesshas not been clearlydefined
(1970) have pointed out, there are grave diffi- forpopulations of sexuallyreproducingorgan-
cultieswith the available mathematicaldefini- isms,confusionarises when two people assume
tions of fitnessfor sexual species. No fitness that theyare both using the same, well-under-
definitionsimultaneouslyaccounts for sexual stood concept of fitness,when in facttheymay
recombination,applies to whole organismsand not be talkingabout the same thing at all. To
their progeny rather than to populations of takean example,usefulforitsclaritybut lacking
single genes, and allows for the differentcon- subtlety,consider the confusionin a discussion
tributionsof male and female offspring.Fur- of trends in reproductive effort when one
thermore,fitnessonly applies to organisms in participantassumes that organisms attemptto
the contextof a given environment.Therefore, maximize r, and the other assumes that organ-
an adequate definitionof fitnessmusttake into isms attempt to minimize the probabilityof
account the existenceof differentgeneral envi- leaving no young at all. In fact, the focus is
ronmentalsituations,and explicitlymodel the on the wrong object. Attentionshould be di-
interactionof environmentand organism. On rected to the assumptions made about the
these grounds, all theoretical work reviewed definition of fitness, assumptions which
in thispaper is suspect. should be testedalong withthe predictionsthat
For example, theoreticiansfrequentlyargue flowfromthem.
as follows. Take a fitnessmeasure (e.g., r or A less serious difficulty,easily avoided by
K) that can only be defined and measured on learningsome biology,arises frequentlyin dis-
a whole population.Find out whatcombinations cussions of the theoretical literature.All op-
of biological traits will maximize that fitness timalitymodels of evolutionaryprocesses share
measure. Then predict that we should expect a common weakness: natural selection favors
individualorganisms,or collectionsof individu- not the optimal trait,but simply the best of
al organisms,to possess those optimal traitsin the available traits. There is a world of dif-
the given situation. ferencebetweenthe optimaland the best avail-
One problem lies with the last step, which able. We can be sure that selection will elicit
involvesa theoreticalfeat akin to the statistical- a local optimum from the variabilityavailable
mechanical derivation of thermodynamiclaws in the ancestral population, but we cannot be
from the behavior of individual molecules. sure thatthe resultwillmatchpredictionsmade
Natural selection acts directlyon individuals by a mathematical model that does not take
and their progeny, not on populations. Until the limitsof natural variabilityinto account.
we can establishthat what holds for collections
of individual organisms holds for the popula- ImplicationsFor Research
tion (which we can measure) and vice versa-
that what holds for the population holds for Of all the hypothesesreviewed above, I re-
individualorganisms-then the line of reason- gard the followingset as best supportedbydata.

In fluctuatingenvironments,age and size at should run for six to ten generationsto detect
firstreproductionshould be respectivelylower a response to selection,it would probably run
and smaller, reproductive efforthigher, size fortwo to threeyearsifexecuted on a poeciliid
of young smaller, and number of young per fish in the tropics,for six months if executed
brood higher, than in constant environments, on Drosophilain the laboratory,for six to ten
wheretheopposite trendsshould hold. We have years if executed on wild oats in countrywith
a fair amount of field evidence, which is only a hard winter,or for twelve to twentyyears
suggestive,not conclusive,to indicatethatthese if executed on pink salmon. Drosophilawould
ideas should be worth testingexperimentally take the least effort,but a poeciliid fishwould
(Gotto, 1962; Cody 1971; Gadgil and Solbrig, providethe bestanalogy tolargerfishand other
1972; Price, 1973; Abrahamson and Gadgil, vertebrates,and the results obtained would
1973); and there is one experiment that has generalize more easily to economicallyimpor-
confirmedseveral of the hypothesesfordande- tant fishspecies.
lions (Gadgil and Solbrig, 1972). The mostexciting,recent,theoreticalmodels
The field evidence consists of correlations make the assumption that reproduction costs
betweenenvironmentalpatternsand reproduc- somethingin termsof subsequent survivaland
tive trends.The argumentsused are the "post futurepossibilitiesfor reproduction (Williams,
hoc, ergo propter hoc" variety,and are less 1966a; Gadgil and Bossert,1970; Schaffer1972,
than convincing. Furthermore,although au- 1974a; Charnov and Krebs, 1973). Predictions
thorsfrequentlyinvokeenvironmentalinstabil- based on thisidea depend on the shape of the
ityto explain the trendstheyobservein life-his- relationshipbetweenreproductiveeffortat time
tory phenomena (e.g., Cody, 1966, 1971), no t,and survivalfromtime tto time t + 1. There
one has actually defined instabilityunambigu- are several approaches to measuringthiscurve,
ously, and then measured it along with the the mostconvincingof whichinvolvesmanipu-
relevantreproductivetraitsduring the process lating the clutch sizes of a bird species that
of selection. has a long period of parental care, and which
Most broad comparisons of reproductive does not put a large proportionof itsreproduc-
trends from field data lack proper control. tive effortinto the production of an egg. If
Many investigatorstry to give evolutionary the experimentercould then measure the dif-
explanationsof phenomena withouteliminating ferencesamong treatmentsin adult survivalto,
possibilitiesthatrequire no evolutionat all. The and capacityto reproduce in, the next breeding
proximalfactorthatmostfrequentlyconfounds season, we would have a reliabletrade-offcurve.
supposed evolutionarytrends is food. Unless A nestingcolonyof cliffswallowsmightpresent
the nutritionalstateof the differentpopulations a good opportunity to do this if the adults
has been assessed and corrected for, or the returnedto the same colonyeach year. Without
geneticbasis of differenceshas been established such tests,the intriguingspeculations of Wil-
throughlaboratorystudies, then any assertion liams (1966a, b), Gadgil and Bossert (1970),
that a difference,observed in the field, has Schaffer(1974a), and Taylor et al. (1974) will
evolved and is an adaptation must be viewed remain unsubstantiated.
withsuspicion. Correlational evidence from field observa-
In order to make progress at this point, we tions will simply not substantiatea theoretical
need carefullycontrolledfield experimentson speculation,since an infinitenumberof models
a short-livedplant or animal. With the predic- can be constructedthat predict the same cor-
tions of Table 4 in mind, the experimenter relation.Only experimentscan trulytesttheory.
should contrasttwo sets of populations (repli- If we want knowledge of the natural processes
cates), one set living in a stable environment, thatshape lifehistories,we willhave tobe willing
the other set in a fluctuatingenvironment.The to do experiments,even if they take years to
experimentershould be sure to measure the execute. The alternative,a fieldfullof interest-
fluctuationsof relevant environmental vari- ing ideas supported in a broad way by correla-
ables, detail the biological mechanisms that tionalevidence, is not just unappealing-it isn't
mediate their impact on the population, and science.
simultaneouslymeasure age (if possible), size, Tests of predictionson reproductivetrends
number of young per brood, size of young, may fail to give any informationbecause the
and reproductiveeffort.Since the experiment causal factors under study were complicated

by parametersaffectingthe behavior,genetics, ACKNOWLEDGMENTS

or developmentof the population. I have some
evidence fromwork in progressthat this is the WilliamE. Neill,Dennis Chitty,George C. Williams,
David B. Mertz,and an anonymous reviewerdid me
case for Gambusia and other poeciliid fish.
a great favor by providing a thorough, constructive
Therefore, we need research that will start to criticismof this paper. Eric L. Charnov deserves
establishthe coevolutionaryconnectionsamong special credit for helping me get started.To James
reproduction,behavior, genetics,and develop- MacLean, Conrad F. Wehrhahn,JudithMyers,J. D.
ment. Most theoreticalmodels reviewed in this McPhail,C. S. Holling, and especiallymywife,Bever-
paper deal withfemalesonlyand assume a stable ley, I owe thanks for encouragement. I also thank
age distribution.Natural populations are rarely the members of the Instituteof Animal Resource
in stableage distribution(Caughley, 1966), and Ecology at the Universityof British Columbia and
the selectiveforcesoperating on males can be the participantsin Organization forTropical Studies
quite differentfromthoseoperatingon females courses 72-3 and 73-2 for enjoyable discussionsand
constructivecriticism.Peter Larkin instigated this
(cf. Darwin, 1859; Trivers, 1972). Theoretical
studyin January,1972, witha recommendationthat
modelsshould explicitlyseparate sexes and deal I spend a year thinkingabout things. Michael T.
withboth. Relaxing the assumptionof a stable Ghiselin brought John Wourms's excitingpapers to
age distributionpresentsa more difficultprob- myattention.I was supported by Isaak Walton Killam
lem; workingout its implicationsis a job for and National Research Council predoctoral fellow-
someone less mathematicallyinept than I am. ships.
LIST OF LITERATURE
ABRAHAMSON, W. G., and M. D. GADGIL. 1973. Growth CAUGHLEY, G. 1966. Mortalitypatternsin mammals.
formand reproductiveeffortin goldenrods (So- Ecology,47: 906-918.
lidago,Compositae). Am. Nat., 107: 651-661. . 1970. A commenton Vandermeer's "pseudore-
ALLEN, J. A. 1959. On the biologyof Pandalus borealis productivevalue." Am. Nat., 104: 214-215.
Kroyer with reference to a population off the CAUGHLEY, G., and L. C. BIRCH. 1971. Rate of increase.
Northumberlandcoast. J. Mar. Biol. Assoc. U.K., J. Wildl.Manage. 35: 658-663.
38: 189-220. CHARNOV, E. L., and J. R. KREBS. 1973. On clutchsize
BARNES, H. 1962. So-called anecdysisin Balanus balan- and fitness.Ibis, 116: 217-219.
oidesand the effectof breeding upon the growth CHARNOV, E. L., and W. M. SCHAFFER. 1973. Lifehistory
of the calcareous shell of some common barna- consequences of natural selection: Cole's result
cles. Limnol.Oceanogr.,7: 462-473. revisited.Am. Nat., 107: 791-793.
BARTLETT, M. S. 1970. Age distributions.Biometrics, CLARKE, B. 1972. Density-dependentselection. Am.
26: 377-385. Nat., 106: 1-13.
BARTHOLOMEW, G. A. 1970. A model forthe evolution CLARKE, K. U., and J. B. SARDESAI. 1959. An analysis
of pinniped polygyny.Evolution,24: 546-559. of the effectsof temperatureupon growthand
BERGER, M. E. 1972. Population structureof olive reproductionof DysducusfasciatusSign. (Hemip-
baboons (Papio anubis)(J. P. Fischer) in the tera,Pyrrhocoridae).Bull. Entomol.Res.,50: 387-
Laikipia Districtof Kenya. E. Afr. Wildl.J., 10: 405.
159-164. CLOUGH, G. C. 1965. Viabilityin wild meadow voles
BOER, P. J.DEN. 1968. Spreading of riskand stabiliza- under various conditions of population density,
tion of animal numbers. Acta Biotheor.,18: 165- season, and reproductive activity.Ecology,46:
194. 119-134.
BRODY, S. 1945. Bioenergetics of Growth.1st ed. Rein- CODY, M. 1966. A general theory of clutch size.
hold, New York. Evolution,20: 174-184.
BRYANT, E. H. 1971. Life historyconsequences of CODY, M. L. 1971. Ecological aspectsof reproduction.
natural selection: Cole's result. Am. Nat., 105: In Farner and King (eds.), Avian Biology,p.
75-77. 462-512. Academic Press, New York.
BURTON, J. A. (ed.). 1973. Owls Of The World:Their COHEN, D. 1966. Optimizing reproduction in a ran-
Evolution,Structure, and Ecology.E. P. Dutton & domlyvaryingenvironment.J. Theoret.Biol., 12:
Co., New York. 119-129.
CALOW, P. 1973. The relationshipbetween fecundity, . 1967. Optimizing reproductionin a randomly
phenology, and longevity:a systemsapproach. varying environment when a correlation may
Am. Nat., 107: 559-574. existbetween the conditionsat the time a choice

has to be made and the subsequent outcome. M., and 0. T. SOLBRIG. 1972. The concept
GADGIL,
J. Theoret.Biol. 16: 1-14. of "r" and "K" selection: evidence from wild
. 1968. A general model of optimalreproduction. flowersand some theoreticalconsiderations.Am.
J.Ecol., 56: 219-228. Nat., 106: 14-31.
. 1971. Maximizing final yield within growthis GEIST, V. 1971 MountainSheep: A Studyin Behavior
limitedby timeor by limitingresources.J.Theor. and Evolution.Univ. of Chicago Press, Chicago.
Biol., 33: 299-307. GERKING, S. D. 1959. Physiologicalchanges accom-
COLE, L. C. 1954. The population consequences of panying ageing in fishes. In G. E. Wolsten-
life historyphenomena. Q. Rev. Biol., 29: 103- holme and M. O'Conner (eds.), The Lifespanof
137. Animals(CIBA Foundation Colloquia on Ageing,
CONNELL, J. H. 1970. A predator-preysystemin the Vol. 5), p. 181-207. Churchill,London.
marine intertidalregion. I. Balanus glandulaand GILBERT, N., and A. P. GUTIERREZ. 1973. A plant-aphid-
several predatoryspecies of Thais. Ecol. Monogr., parasiterelationship.J.Anim.Ecol.,42: 323-340.
40: 49-78. GODDARD, J. 1970. Age criteriaand vital statisticsof
CRISP, D. J., and B. S. PATEL. 1961. The interaction a black rhinocerospopulation. E. Afr. Wildl.J.,
betweenbreedingand growthrate in the barnacle 8: 105-121.
Elminius modestus darwin. Limnol. Oceanogr.,6: GOODMAN, L. A. 1971. On thesensitivity of theintrinsic
105-115. growthrate to changes in age-specificbirth-and
DARWIN, C. 1859. The Origin ofSpecies.JohnMurray, death-rates.Theor.Popul. Biol., 2: 339-354.
London. Gor-ro,R. V. 1962. Egg-numberand ecology in com-
DEEVEY, E. S. 1947. Life tablesfor naturalpopulations mensal and parasiticcopepods. Ann. Mag. Nat.
of animals. Q. Rev. Biol. 22: 283-314. Hist., Ser. 13: 97-107.
DEMPSTER, J. P. 1971. The population ecology of the HAIRSTON, N. G., D. W. TINKLE, and H. M. WILBUR.
cinnabar moth, Tyriajacobaeae L. (Lepidoptera, 1970. Natural selection and the parameters of
Arctiidae). Oecologia,7: 26-67. population growth. J. Wildl. Manage. 34: 681-
DOBZHANSKY, T. H. 1950. Evolutionin thetropics.Am. 690.
Sci. 38: 209-221. HAMILTON, W. D. 1966. The moulding of senescence
DRURY, W. H., JR. 1961. Studiesof thebreedingbiology by natural selection.J. Theor.Biol., 12: 12-45.
of the Horned Lark, Water Pipit, Lapland HAMMOND, J. 1958. Fertility.In A. S. Parkes (ed.),
Longspur, and Snow Bunting on Bylot Island, Marshall'sPhysiology of Reproduction, Vol. 2, p.
NorthwestTerritories,Canada. Bird-Banding32: 648-740. Longmans, Green, London.
1-46. HANAMURA. N. 1966. Sockeye salmon in the Far East.
DUNCAN, D. L., and G. A. LODGE. 1960. Diet in relation Int.N. Pac. Fish Bull., 18: 1-28.
to reproductionand viabilityof young. III. Pigs. HARCOURT, D. G. 1963. Major mortalityfactorsin the
Commonwealth Bur. Anim. Nut. Tech. Com- population dynamicsof the diamondback moth,
mun. No. 21; 106 pp. Plutella maculipennis(Curt.)(Lepidoptera: Plu-
EMBREE, D. G. 1965. The population dynamicsof the tellidae). Mem. Entomol.Soc. Can., 32: 55-66.
wintermoth in Nova Scotia. 1954-1962. Mem. HARPER, J. L. 1967. A Darwinian approach to plant
Ent. Soc. Canada, No. 46; 57 pp. ecology. J.Ecol., 55: 242-270.
EMLEN, J. M. 1970. Age specificityand ecological HARPER, J.L., P. H. LOVELL, and K. G. MOORE. 1970.
theory.Ecology,51: 588-601. The shape and sizes of seeds. Ann. Rev. Ecol.
EULER, L.., 1760. Rechereches generalessur la morta- Syst.,1: 327-356.
lite et la multiplicationdu genre humain. Mem. HEALY, W. R. 1974. Population consequences of al-
Acad. Sci., Berlin,16: 144-164. [In Opera OCmnia, ternativelifehistoriesin Notophthalmus v. Virides-
Series 1, Vol. 7: 79-100.] cens.Copeia, 1974: 221-229.
FAGEN, R. M. 1972. An optimal life historyin which HERON, A. C. 1972a. Populationecologyofacolonizing
reproductiveeffortdecreases withage. Am. Nat., species: the pelagic tunicate Thalia democratica.
106: 258-261. I. Individual growthrate and generation time.
FISHER, R. A. 1930 (2nd ed., 1958). The Genetical
Theory Oecologia,10: 269-293.
ofNaturalSelection.Dover, New York. . 1972b. Population ecology of a colonizing spe-
FOTHERINGHAM, N. 1971. Life historypatternsof the cies: the pelagic tunicate Thalia democratica. II.
littoralgastropods Shaskyusfestivus(Hinds) and Population growthrate. Oecologia,10: 294-312.
Ocenebra poulsoni (Carpenter)(Prosobranchia: HICKEY,F. 1960. Death and reproductive rates of
Muricidae). Ecology,52: 742-757. sheep in relation to flockculling and selection.
FRETWELL, S. 1969. The adjustment of birth-rateto N. Z. J.Agric.Res.,3: 332-344.
mortalityin birds. Ibis, 111: 624-627. HOLGATE, P. 1967. Population survivaland lifehistory
GADGIL, M., and W. BOSSERT. 1970. Life historyconse- phenomena. J. Theor.Biol., 14: 1-10.
quences of naturalselection.Am. Nat., 104: 1-24. HUSSELL, D. J. T. 1972. Factors affectingclutch size

in arcticpasserines. Ecol. Monogr.,42: 317-364. casebearer, the fruit-treeleaf roller, and the
ILEs,T. D. 1974. The tacticsand strategyof growth European corn borer in Quebec. Mem. Entomol.
in fishes.In F. R. Harden Jones(ed.), Sea Fisheries Soc. Can., 32: 67-82.
Research,p. 331-346. John Wiley & Sons, New LESLIE, P. H. 1945. On the use of matricesin certain
York. population mathematics. Biometrika,33: 183-
JANZEN, D. H. 1967. Synchronizationof sexual repro- 212.
ductionin treeswithinthe dry season in Central .1948. Some furthernoteson the use of matrices
America. Evolution,21: 620-637. in population mathematics.Biometrika, 35: 213-
. 1969. Seed eaters versus seed size, number, 245.
toxicity,and dispersal. Evolution,23: 1-27. LEVINS, R. 1968. Evolutionin ChangingEnvironments.
. 197 la. Escape of Cassia grandisL. Beans from PrincetonUniv. Press, Princeton.
predatorsin timeand space. Ecology, 52: 964-979. . 1970. Extinction.In Lectureson Mathematics in
. 1971b. Seed predation by animals. Ann. Rev. the Life Sciences, Vol. 2 (Some Mathematical
Ecol. Syst.,2: 465-492. Questionsin Biology),p. 75-108. Am. Math. Soc.,
. 1972a. Association of a rainforestpalm and Providence,R.I.
seed-eating beetles in Puerto Rico. Ecology,53: LEWIs, E. G. 1942. On the generationand growthof
258-261. a population. Sankhya,6: 93-96.
. 1972b. Escape in space by Sterculaapetalaseeds LEWONTIN, R. C. 1965. Selection forcolonizingability.
fromthe bug Dysducusfasciatusin a Costa Rican In H. G. Baker and G. L. Stebbins (eds.), The
deciduous forest.Ecology,53: 350-361. GeneticsofColonizingSpecies,p. 79-94. Academic
KADLEC, J. A., and W. H. DRURY. 1968. Structureof Press, New York.
the New England herringgull population. Ecolo- LLOYD, M., and H. S. DYBAS. 1966. The periodical
gy,49: 49-78. cicada problem. II. Evolution. Evolution,20:
KEMPrHORNE, O., and E. POLLAK. 1970. Concepts of 466-505.
fitnessin mendelian populations. Genetics,64: LOEB, L. 1917. The concrescence of folliclesin the
125-145. hypotypicalovary. Biol. Bull., 32: 187-193.
KEYFITZ, N. 1968. Introduction to the Mathematicsof LOTKA, A. J. 1907. Studies on the mode of growth
Population.Addison-Wesley,Reading, Mass. of materialaggregates. Am. J. Sci., 24: 199-216,
KING, C. E. 1964. Relative abundance of species and 375-376.
MacArthur'smodel. Ecology,45: 716-727. . 1913. A natural population norm. II. J. Wash.
KING, C. E., and W. W. ANDERSON. 1971. Age specific Acad. Sci., 3: 289-293.
selection.II. The interactionbetween r- and K- . 1922. The stabilityof the normal age distribu-
during population growth.Am. Nat., 105: 137- tion. Proc.Nat. Acad. Sci., U.S., 8: 339-345.
156. . 1925. Elementsof PhysicalBiology.Dover, New
KLOMP,H. 1966. The dynamicsof a field population York.
of the pine looper, Bupalus piniarus L. (Lep., LOWE, V. P. W. 1969. Population dynamicsof the Red
Geom.). Adv. Ecol. Res.,3: 207-306. Deer (Cervuselephas)on Rhum. J. Anim. Ecol.,
. 1970. Clutch size in birds. Ardea,58: 1-121. 38: 425-458.
KREBS, C. J. 1972. Ecology:The Experimental Analysis MAcARTHUR, R. H. 1962. Some generalized theorems
of Distributionand Abundance.Harper & Row, of natural selection. Proc. Nat. Acad. Sci. U.S.,
New York. 48: 1893-1897.
KURNICK, A. A., H. B. HINDS, M. W. Pasvogel,and B. . 1968. Selection for life tables in periodic envi-
L. Reid. 1961. Dietary energy levels for laying ronments.Am. Nat., 102: 381-383.
hens as related to age and environmentaltem- MAcARTHUR, R. H., and E. 0. WILSON. 1967. Theory
peratures. Poult.Sci., 40: 1483-1491. of Island Biogeography.Princeton Univ. Press,
LACK, D. 1947-1948. The significanceof clutch size. Princeton.
Ibis,89: 302-352, 90: 25-45. MARKUS, H. C. 1934. Life historyof the black-headed
.1954. The NaturalRegulationofAnimal Numbers. minnow, Pimephalespromelas.Copeia, 70: 116-
Oxford Univ. Press, Oxford. 122.
LAMMING, G. E. 1969. Nutritionand reproduction.In MARSHALL, D. R., and S. K. JAIN. 1968. Phenotypic
D. Cuthbertson (ed.), Nutritionof Animals of plasticityof Avenafatuaand A. barbata.Am. Nat.,
AgriculturalImportance,Vol. I. p. 411-453. 102: 457-467.
Pergamon, Elmsford,N.Y. , and -. 1970. Seed predation and dormancy
LAUGHLIN, R. 1965. Capacity for increase: a useful in the population dynamics of Avena fatua and
population statistic.J. Anim. Ecol., 34: 77-92. A. barbata.Ecology,51: 886-891.
LERoux, E. J., R. 0. Paradis, and M. Hudon. 1963. MAYER, A. M., and A. POLJAKOFF-MAYBER. 1963. The
Major mortalityfactors in the population dy- Germination ofSeeds.Macmillan,New York.
namics of the eye-spottedbud moth, the pistol MAYNARD SMITH, J. 1971. The originand maintenance

of sex. In G. C. Williams (ed.), Group Selection, mortalityof certain lower organisms. Q. Rev.
p. 163-175. Aldine-Atherton,Chicago. Biol., 10: 60-79.
MEATS, A. 1971. The relativeimportanceto population PETERSON, C. F., E. A. SAUTER, D. H. CONRAD, and C.
increase of fluctuationsin mortality,fecundity, F. LAMPMAN, 1960. Effect of energy level and
and the timevariablesof the reproductivesched- laying house temperature on the performance
ule. Cecologia,6: 223-237. of white leghorn pullets. Poul. Sci., 39: 1010-
MENGE, B. A. 1974. Effectof wave action on brooding 1017.
and reproductiveeffortin a seastar, Leptasterias PETRIDES, G. A., and W. G. Swank. 1966. Estimating
hexactis.Ecology,55: 84-93. the productivity and energy relationsof an Afri-
MERTZ, D. B. 1970. Notes on methods used in life can elephant population. Proc. NinthInt. Grass-
historystudies. In J. H. Connell, D. B. Mertz, lands Cong.,Sao Paolo, Brazil: 831-842.
and W. W. Murdoch (eds.), Readingsin Ecology PIANKA, E. R. 1970. On "r" and "K" selection. Am.
and EcologicalGenetics,p. 4-17. Harper & Row, Nat., 104: 592-597.
New York. . 1972. "r" and "K" or "b" and "d" selection?
. 1971a. Life historyphenomena in increasing Am. Nat., 106: 581-588.
and decreasing populations. In G. P. Patil, E. PIMLOTr, D. H. 1959. Reproductionand productivity
C. Pielou, and W. E. Waters (eds.). Statistical of the Newfoundland moose. J. Wildl. Manage.,
Ecology,Vol. 2, p. 361-399. Pennsylvania State 23: 381-401.
UniversityPress, UniversityPark. PRICE, P. W. 1973. Reproductivestrategiesin parasi-
. 1971b. The mathematicaldemography of the toid wasps. Am. Nat., 107: 684-693.
California condor population. Am. Nat., 105: RICHARDS, J. B., and D. A. HUGHES. 1972. Some ob-
437-454. servationsof sea-turtlenesting activityin Costa
MILLAR, J. S., and F. C. ZWICKEL. 1972. Determination Rica. Mar. Biol., 16: 297-309.
of age, age structure,and mortalityof the pika, RICKER, W. E. 1954. Stock and recruitment.J. Fish.
Ochotonaprinceps(Richardson). Can. J. Zool., 50: Res. Bd. Can., 11: 559-623.
229-232. RICKLEFS, R. E. 1969. An analysisof nestingmortality
MOUNTFORD, M. D. 1968. The significance of littersize. in birds. Smithsonian Contrib. Zool. No. 9;
J. Anim. Ecology37: 363-367. Washington,D.C.
- 1971. Population survivalin a variable environ- ROTHSCHILD, M., and B. FoRD, 1972. Breeding cycle
ment. J. Theor.,Biol., 32: 75-79. of the flea Cediopsyllasimplexis controlled by
- 1973. The significanceof clutchsize. In Bartlett, breeding cycleof its hosts. Science,178: 625.
M. S., and R. W. Hiorns (eds.), The Mathematical ROUGHGARDEN, J. 1971. Density-dependentnaturalse-
Theoryof the Dynamicsof Biological Populations. lection. Ecology,52(3): 453-468.
Academic Press, New York. SARUKHAN, J. 1974. Studies on plant demography:
MURDOCH, W. W. 1966. Population stabilityand life RanunculusrepensL., R. bulbosusL. and R. acris
historyphenomena. Am. Nat., 100: 45-51. L. II. Reproductive strategiesand seed popula-
MURPHY, G. I. 1967. Vital statisticsof the Pacific tion dynamics.J.Ecol., 62(1): 151-178.
sardine (Sardinopscaerulea) and the population SARUKHAN, J.,and J.L. HARPER. 1973. Studies on plant
consequences. Ecology,48: 731-735. demography: RanunculusrepensI., R. bulbosusL.,
- 1968. Pattern in life historyand the environ- and R. acris L. I. Population flux and survivor-
ment. Am. Nat., 102: 390-404. ship. J. Ecol., 61(3): 675-716.
PALMBLAD, I. G. 1969. Population variationin germi- SCHAFFER, W. M. 1972. Evolutionof optimalreproduc-
nation of weedy species. Ecology,50: 746-748. tivestrategies.PhD Thesis; PrincetonUniversity,
PARIS, 0. A., and F. H. PITELKA. 1962. Population Princeton.
characteristicsof the terrestrialisopod Armidil- - 1974a. Selection for optimal life histories:The
lidium vulgarein California grassland. Ecology, effectsof age structure.Ecology,55(2): 291-303.
43: 229-248. - 1974b. Optimal reproductiveeffortin fluctuat-
PEARL, R. 1922. The BiologyofDeath.J. B. Lippincott, ing environments.Am. Nat., 108: 783-790.
Philadelphia. SCHAFFER, W. M., and P. F. ELSON. 1975. The adaptive
- 1925. The Biologyof PopulationGrowth.Alfred significanceof variations in life historyamong
A. Knopf, New York. local populations of Atlantic Salmon in North
- 1928. The Rate of Living. Alfred A. Knopf, America. Ecology,56: 577-590.
New York. SCHAFFER, W. M., and C. A. REED. 1972. The coevolu-
- 1939. The Natural Historyof Population.Oxford tion of social behavior and cranial morphology
UniversityPress, London. in sheep and goats (Bovidae, Caprini). Fieldiana
PEARL, R., andJ. R. MINER. 1935. Experimentalstudies Zool., 61: 1-88.
on the duration of life. XIV. The comparative SCHAFFER, W. M., and R. H. TAMARIN. 1973. Changing

reproductiverates and population cyclesin lem- TINKLE, D. W., H. M. WILBUR, and S. G. TILLEY. 1970.
mingsand voles. Evolution,27: 111-124. Evolutionary strategies in lizard reproduction.
SCHMALHAUSEN, I. I. 1949. Factorsof Evolution: The Evolution,24(1): 55-74.
TheoryofStabilizingSelection.Blakiston,Philadel- TRIVERS, R. L. 1972. Parental investmentand sexual
phia. selection. In B. Campbell (ed.), Sexual Selection
Sco-rr,D. P. 1962. Effectsof food quantityon fecun- and theDescentofMan-1871-1971, p. 136-179.
dity of rainbow trout, Salmo gairdneri.J. Fish. Aldine-Atherton,Chicago.
Res. Bd. Can., 19: 715-731. 1974. Parent-offspring conflict.Am. Zool., 14:
SHARPE, F. R., and A. J. LOTKA. 1911. A problem in 249-264.
age-distribution.Philos.Mag., 21: 435-438. VANDERMEER, J. H. 1968. Reproductive value in a
SHORT,R. V. 1972. Species differences.In C. R. Austin population of arbitrary age distribution. Am.
and R. V. Short (eds.), Reproduction in Mammals, Nat., 102: 586-589.
Book (Reproductive Patterns),
p. 1-33. Cambridge WATSON, R. M. 1970. Generation time and intrinsic
Univ. Press, Cambridge. rates of increase in wildebeeste (Connochaetes
SKUTCH,A. F. 1949. Do tropical birds rear as many taurinusalbojubatusThomas). J. Reprod.Fert.,22:
young as theycan nourish? Ibis,91: 430-455. 557-561.
. 1967. Adaptive limitationof the reproductive WILLIAMS, G. C. 1959. Ovary weights of darters: a
rate of birds. Ibis, 109: 579-599. test of the alleged association of parental care
SMITH, C. C. 1970. The coevolutionof pine squirrels withreduced fecundityin fishes.Copeia, 1959(1):
(Tamia sciurus)and conifers.Ecol. Monogr.,40(3): 18-24.
349-371. . 1966a. Adaptationand NaturalSelection.Prince-
SNOW, D. 1958. A StudyofBlackbirds. Allen and Unwin, ton UniversityPress, Princeton.
London. . 1966b. Natural selection,the costs of reproduc-
STRATHMANN, R. 1974. The spread of sibling larvae tion, and a refinementof Lack's principle. Am.
ofsedentarymarineinvertebrates.Am. Nat., 108: Nat., 100: 687-690.
29-44. . 1975. Sex and Evolution.Princeton University
SUMMERS-SMITH, D. 1956. Mortalityof the house spar- Press, Princeton.
row. Bird Study,3: 265-270. WILLIAMS, G. C., and J.B. MIrnoN. 1973. Why repro-
SVARDSON, G. 1949. Natural selectionand egg number duce sexually?J. Theor.Biol., 39: 545-554.
in fish.Ann. Rep. Inst.Freshwater Res., Drottning- WILSON, E. 0. 1973. Group selection and its signifi-
holm,29: 115-122. cance for ecology. Bioscience,23(11): 631-638.
TAYLOR, H. M., R. S. GOURLEY, C. E. LAWRENCE, and R. WOURMS, J. P. 1972. The developmental biology of
S. KAPLAN. 1974. Natural selectionof life history annual fishes.III. Pre-embryonicand embryonic
attributes:an analyticalapproach. Theor. Popul. diapause of variable duration in the eggs of
Biol., 5: 104-122. annual fishes.J. Exp. Zool., 182: 389-414.
TINKLE, D. W. 1969. The concept of reproductive WYNNE-EDWARDS, V. C. 1962. Animal Dispersionin
effortand its relation to the evolution of life Relation to Social Behavior. Oliver and Boyd,
historiesin lizards. Am. Nat., 103: 501-516. Edinburgh.


Stearns - 1976 - Life-History Tactics A Review of The Ideas

Hochgeladen von

Dokumentinformationen

Originaltitel

Copyright

Verfügbare Formate

Dieses Dokument teilen

Dokument teilen oder einbetten

Freigabeoptionen

Stufen Sie dieses Dokument als nützlich ein?

Sind diese Inhalte unangemessen?

Copyright:

Verfügbare Formate

Stearns - 1976 - Life-History Tactics A Review of The Ideas

Hochgeladen von

Copyright:

Verfügbare Formate

Life-History Tactics: A Review of the Ideas

Author(s): Stephen C. Stearns

This content downloaded from 129.173.72.87 on Tue, 23 Jul 2013 14:32:41 PM

THE QUARTERLY REVIEW

LIFE-HISTORY TACTICS: A REVIEW OF THE IDEAS

This content downloaded from 129.173.72.87 on Tue, 23 Jul 2013 14:32:41 PM

Scope and Purpose tangled thicketthat blocked furtherprogress.

This content downloaded from 129.173.72.87 on Tue, 23 Jul 2013 14:32:41 PM

This content downloaded from 129.173.72.87 on Tue, 23 Jul 2013 14:32:41 PM

is subject to experimentation,in order to mini- numbers of individuals, as defined in the

This content downloaded from 129.173.72.87 on Tue, 23 Jul 2013 14:32:41 PM

The StableAge Distribution The ComparativeStudyofLife Tables

This content downloaded from 129.173.72.87 on Tue, 23 Jul 2013 14:32:41 PM

This content downloaded from 129.173.72.87 on Tue, 23 Jul 2013 14:32:41 PM

dox. Givingbirthto n + 1 offspring,and then offeringmaximum growth,where the species

which implies that

This content downloaded from 129.173.72.87 on Tue, 23 Jul 2013 14:32:41 PM

Malthusianparameterwhichcould be achieved N(t + 1) = Ba C N(t)

This content downloaded from 129.173.72.87 on Tue, 23 Jul 2013 14:32:41 PM

consistof clones of asexually reproducingindi- ratio of average juvenile mortalityto average

This content downloaded from 129.173.72.87 on Tue, 23 Jul 2013 14:32:41 PM

0 Red deer 0 Barnacle

0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0

FIG. 1. EVIDENCE BEARING ON COLE'S PARADOX.

This content downloaded from 129.173.72.87 on Tue, 23 Jul 2013 14:32:41 PM

Red Deer Cervuselephas 0.092 2.4874 Lowe, 1969 1,3

This content downloaded from 129.173.72.87 on Tue, 23 Jul 2013 14:32:41 PM

This content downloaded from 129.173.72.87 on Tue, 23 Jul 2013 14:32:41 PM

This content downloaded from 129.173.72.87 on Tue, 23 Jul 2013 14:32:41 PM

reproduction,what conditions will be like for

This content downloaded from 129.173.72.87 on Tue, 23 Jul 2013 14:32:41 PM

TYPES OF EVIDENCE SPECIES REFERENCE

Summary the evolutionaryprocessesinvolved.Naturalse-

This content downloaded from 129.173.72.87 on Tue, 23 Jul 2013 14:32:41 PM

growthrate, r, to changes in the age at first

This content downloaded from 129.173.72.87 on Tue, 23 Jul 2013 14:32:41 PM

putting your money in the bank as long as viewpointson reproductivetacticsoverlap and

This content downloaded from 129.173.72.87 on Tue, 23 Jul 2013 14:32:41 PM

species should show much less geneticvariance

This content downloaded from 129.173.72.87 on Tue, 23 Jul 2013 14:32:41 PM

in tropicalareas biologicalinteractionspredom- viewpoint,generated markedlydifferentpre-

This content downloaded from 129.173.72.87 on Tue, 23 Jul 2013 14:32:41 PM

C~~ C-~~~~ ~ CU> CZObQ

This content downloaded from 129.173.72.87 on Tue, 23 Jul 2013 14:32:41 PM

This content downloaded from 129.173.72.87 on Tue, 23 Jul 2013 14:32:41 PM

SUMMARY OF MURPHY'S MODEL

Ix mx, lx mx Ix I Mx, Ixmx

*-long lived short lived

0 25 50 75 100 125 150 175 200 225

FIG. 5. THE IMPACT OF VARIATIONS IN RECRUITMENT ON LIFE HISTORIES

This content downloaded from 129.173.72.87 on Tue, 23 Jul 2013 14:32:41 PM

This content downloaded from 129.173.72.87 on Tue, 23 Jul 2013 14:32:41 PM

TABLE 4 springhatchingtimes(also see Williams,1975).

This content downloaded from 129.173.72.87 on Tue, 23 Jul 2013 14:32:41 PM

This content downloaded from 129.173.72.87 on Tue, 23 Jul 2013 14:32:41 PM

This content downloaded from 129.173.72.87 on Tue, 23 Jul 2013 14:32:41 PM

This content downloaded from 129.173.72.87 on Tue, 23 Jul 2013 14:32:41 PM

This content downloaded from 129.173.72.87 on Tue, 23 Jul 2013 14:32:41 PM

I have built this section on a foundation matureone at a time.) Furthermore,Neill notes

This content downloaded from 129.173.72.87 on Tue, 23 Jul 2013 14:32:41 PM

This content downloaded from 129.173.72.87 on Tue, 23 Jul 2013 14:32:41 PM

This content downloaded from 129.173.72.87 on Tue, 23 Jul 2013 14:32:41 PM

optimal effort is intermediate(regular b3 = C-pi

C C-~~ ~ CU> CZObQ