The Optimal Balance between Size and Number of Offspring

Author(s): Christopher C. Smith and Stephen D. Fretwell

Source: The American Naturalist, Vol. 108, No. 962 (Jul. - Aug., 1974), pp. 499-506
Published by: University of Chicago Press for American Society of Naturalists
Stable URL: http://www.jstor.org/stable/2459681
Accessed: 04-10-2015 22:00 UTC

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.

American Society of Naturalists and University of Chicago Press are collaborating with JSTOR to digitize, preserve and
extend access to The American Naturalist.

http://www.jstor.org

This content downloaded from 128.103.149.52 on Sun, 04 Oct 2015 22:00:37 UTC
All use subject to JSTOR Terms and Conditions
Vol. 108, No. 962 The American Naturalist July-August 1974

THE OPTIMAL BALANCE BETWEEN

SIZE AND NUMBER OF OFFSPRING
CHRISTOPHER C. SMITH AND STEPHEN D. FRETWELL*

Division of Biology, Kansas State University, Manhattan, Kansas 66502

In searchingfor patternsof energyexpenditureby parents on their off-

spring,two relationshipsseem intuitivelyobvious. (1) As the energyex-
pended on individual offspringis increased,the number of offspringthat
parentscan produceis decreased. (2) As the energyexpendedon individual
offspringincreases,the fitnessof individual offspringincreases. The rela-
to visualizeis the one betweenthe energyexpended
tionshipwhichis difficult
on individual offspringand the fitnessof the parent whose genotypeinflu-
ences how energyis distributedto the young. Here we presenta graphical
model to representthe relationshipbetweenenergyexpended on individual
offspringand the fitnessof parents.The validity of the graphical model is
then demonstratedby standard analytical procedures.

THE MODEL

The model is based on the one assumptionthat at any point in an or-

ganism's life historythere is an optimumpercentageof available energy
that should be divertedto reproductionto maximizethe parents' total con-
tributionto futuregenerations.Energy available for reproductionis thus
limitedto a set finiteamount at any given time. Our model demonstrates
how thisfiniteamountcan be distributedinto different sizes and numbersof
offspring. The validityof the basic assumptionis discussedlater.
The abscissa of the model in figure1 plots both the size and numberof
individual offspring.Their relationshipon the abscissa is determinedby
the basic assumptionof a set finiteamount of energyavailable for repro-
duction.The ordinateplots the fitnessof individual offspring. Straightlines
throughthe origin are fitnessfunctionsor lines of equal fitnessfor the
parent. As we show below, these fitnessfunctionsare similar to those of
Levins (1968), although their orientationis different.Greater slopes of
fitnessfunctionscorrespondto higherparental fitness.No causal relation-
ship betweenthe abscissa and ordinate is implied in the fitnessfunctions.
We do not wish to imply that expectedfitnessof offspring is linearly pro-
portionalto theirenergyexpense over a large range of energyexpenses in
any real population. Fitness functionsare simply a consequence of the
values of the ordinateand abscissa.
* Order of authorship was determined by a flip of a coin.

499

This content downloaded from 128.103.149.52 on Sun, 04 Oct 2015 22:00:37 UTC
All use subject to JSTOR Terms and Conditions
500 THE AMERICAN NATURALIST

20SUCCESSFUL
0.363 OFFSPRING
0103~/,1

0O.2
6"/
CAD

C= 0.0

0 5
Iy=EFFORT/OFFSPRING. 10 15 20 25
OFOFFSPRING
Ny=NUMBER co 200 100 67 50 40
FIG. 1.-Adaptive functions of parentalstrategies.The verticalaxis describes
parent types with offspringof a certain fitness,as noted. The horizontal
axis describesparent types that put a certainfractionof available energy
into each offspring(or equivalentlyproducea certainnumberof offspring
fromthe 1,000 units of energyavailable). The diagonal lines in the two-
dimensionalspace are sets of parentaltypesall of whosemembersare equally
successfulin contributing
to futuregenerations. Diagonal linesof steeperslope
representmore successfulparenttypes.

Real populationshave a curve of expectedfitnessfor individual offspring

(Wy) in relationto their energyexpenseswhich we assume to be similar
in shape to the solid curve in figure2. The curve leaves the abscissa at some
point greaterthan zero because an offspring must at least have the energy
of a completeset of geneticinformationbeforeit has greaterthan zero fit-
ness. The curve is definedonly for offspringsizes up to that size where all
parental investmentgoes to just one offspring.Beyond that size, expected
fitnessof the young is irrelevant,since the parent could not produce even
one. We consider in detail curves sufficiently convex for the interceptof
the derivativeto pass throughzero for at least one point on the curve.
Apparentlythis case applies to mostof nature,as mostclutchesare greater
than one, but we cannot prove this. Such convexitymight derive bio-
logically fromtherebeing some limit to the fitnessof individual offspring
that cannotbe reachedno matterhow greatthe parental investment.
Then, when the clutch is small enough (Iy large enough), a unit de-
crease in the clutch will produce a relativelylarge increase in fy but a
relativelysmall increasein Wy,producingconvexity.If such convexitydoes
not obtain, then, assuming a continuous curve, one can prove from the
graphicalanalysisthatonlya singleoffspring shouldbe producedper effort.
To determinewhichenergyexpenseper offspring gives the highestparen-
tal fitness,one can draw a series of fitnessfunctionsintersectingthe curve.
The steepestfitnessfunctionthat intersectsthe curve is the fitnessfunction
tangentto the curve (fig. 2). Thereforethe optimumenergyexpense per
offspringis at the point where a fitnessfunctionis tangent to the curve

This content downloaded from 128.103.149.52 on Sun, 04 Oct 2015 22:00:37 UTC
All use subject to JSTOR Terms and Conditions
 SIZE AND NUMBER OF OFFSPRING 501

0.3-

-
40. 0.2
Be

0.1

0 5 10 15 20 25

1.0

,,, 0.8
0.6-

an 0.4

? 0.2-

0.00 5 10 15 20 25
Iy EFFORT/OFFSPRING
FIG. 2.-Fitness set analysis of parental strategy. The curved line in the
upper part of the figureis the set of possible combinations of parental types,
each type expending a certain amount of effortper offspringand producing off-
spring of a certain fitness.The shape of this curve is typical of those that will
produce clutches of more than one offspring.The curve can be thought of as
a cause-effect relationship (effort per offspring determines the offspring's
fitness). Two adaptive functions, drawn as in fig. 1, intersect the curve of
possible parental types. The intersection involving the adaptive function of
highest slope determinesthe optimal parental type defined in terms of effort
per offspring.Intersections of other adaptive functions determine the rela-
tionship between parental fitness and parental type (effort per offspring),
plotted in the lower graph. Note that all parental types in a single adaptive
function are definedto have equal fitness,so that the two intersectionsof any
lower than optimal adaptive function produce points on the bottom graph
with the same value of parental fitness.

of fitnessfor individual offspring.The ratio of the slope of a fitnessfunc-

tion throughanotherpoint on the curve to the slope of the fitnessfunction
tangentto the curve gives thefitnessto the parent of expendingthe amount
of energyper offspring representedby that otherpoint (fig.2). The model,
therefore,allows one to visualize in two dimensionsthe fitnessaccruing to

This content downloaded from 128.103.149.52 on Sun, 04 Oct 2015 22:00:37 UTC
All use subject to JSTOR Terms and Conditions
502 THE AMERICAN NATURALIST

both offspringand parent fromvariation in the energyexpended per off-

spring.
Now, the graphical analysis given above can be presentedanalyticallyto
prove that the point of contactof the sloping line throughthe origin with
the young's fitnesscurve is the optimumoffspring size for the parent.
Let Wy be the fitnessof the young. We assume that energyinvestment
per young (Iy) increasesthe young's fitness,but that the rate of increase
eitheractually reacheszero or gets close to zero for some high but realistic
value of energyinvestment.This impliesa curve of fitnessof young on cost
of young (fig.2) thatis convexforall Iy greaterthan someparticularvalue.
Let Ny be the numberof youngproducedby the parent,and Ip be the total
energyinvestedby the parent.We assume that
Ny=Ip/ly. (1)
This is approximate and simplifying;Ny must in reality have integer
values, and yet Ip/ly can take nonintegervalues, unless we restrictly to
certain values (we assume Ip is fixed). Ricklefs (1968) analyzed com-
promisesthat mightbe made at small clutch sizes, when nonintegervalues
of Ny are optimal.We again proceed,recognizingthat we will obtain non-
integer optimal values for Ny, which require furtheranalysis (see Dis-
cussion).
Parental fitnessforthe clutchis denotedWp, and thatforthe young,Wy.
We claim that

Wp=Wy*Ny =Wy- P (2)

forvalues on the youngfitnesscurve.

We now assume that productionof young in the presentdoes not influ-
ence the futurefitnessof parents.Natural selectionshould then maximize
Wp. We findthe maximumvalue of Wp by maximizingWy/ly in equation
(2) (Ip fixed). But Wy/ly is the tangentof the angle formedby the slop-
ing line throughthe origin and by the ly axis. Since the tangent of an
angle and the angle are positively monotonic,maximizing Wy/ly also
maximizes a. This proves that the line with the maximum slope, which
intersectsthe young's fitnesscurve,also touchesthat curve at the point that
maximizesadult fitness.
These straightlines are analogous to the adaptive or fitnessfunctionsof
Levins (1968), while the curve of offspring fitnesswith size is a fitnessset.
We now establishthis correspondence.
In an adaptive function,points on a single line or curve representalter-
native parent phenotypesall of which have the same fitness.However, all
phenotypesmay not be realizable. The statementof an adaptive functionis
this: if we could find a collectionof parental types with certain specified
properties,all would have the same fitness.Here, the propertiesare fitness
and size of offspring.Parental fitnessis the product of the fitnessof each
of its offspringtimesthe numberof them (eq. [2]).

This content downloaded from 128.103.149.52 on Sun, 04 Oct 2015 22:00:37 UTC
All use subject to JSTOR Terms and Conditions
 SIZE AND NUMBER OF OFFSPRING 503

If all of a group of parents have the same fitness,then Wp k, and

from equation (2),
k
WY ( N ) -(3)
Ny
From (1),
1 ly
by )
N Ip
So, substitutingin (3) we obtain

'WY= k * 1.
'p
is assumed to be constant; and so
Like k, Ip, the total parental investment,
we have derived an equation for a straightline throughthe origin,with
slope k/lp. This straightline is made up of points each of which defines
a parentwithyoung of a certainsize (ly) and a certainfitness(Wy). All
theseparentshave the same fitness(Wp = k), and so the line is an adaptive
function.Adaptive functionswith higherslopes must have higher Wp (Ip
is constant). By Levins' arguments,we know that the optimumphenotype
is found on the fitnessset (the real possibilities),by taking the point of
intersectionwith the adaptive functionof highestfitness(slope).

DISCUSSION

A numberof life historyphenomenamightappear to contradictour basic

assumptionthat at any stage in an organism's life history an optimum
fractionof its available energyshould be divertedto reproductionto maxi-
mize total reproductionduring its lifetime. E. R. Pianka and D. H.
Feener (personal communication)argue that as an individual's reproduc-
tive value declines after the onset of reproductivematurity,it maximizes
fitnessby divertinga successivelyhigherfraction of its available energy
to reproductionwith each successive,evenly spaced, reproductiveeffort.
Pianka and Feener's argumentson the effectof changing reproductive
value specifywhat fractionof available energyshould be expended on re-
productionat given points througha life history,but not how the fraction
expended should be divided into numbersof offspring.The fact that the
optimumfraction of available energy diverted to reproductionchanges
throughouta life historyin no way negates our model.
For organismslike birds,mammals,and subsocial Hymenoptera(Wilson
1971), whereparentsprogressively provisiontheiroffspringas theydevelop,
energyexpended per individual offspring is considerablygreaterthan that
needed to produce a fertilizedegg. Reproductiveeffortin our model, the
fractionof available energydivertedto reproduction,includes all formsof
parental care. Energy used to provisionprogressivelyor to defend an off-

This content downloaded from 128.103.149.52 on Sun, 04 Oct 2015 22:00:37 UTC
All use subject to JSTOR Terms and Conditions
504 THE AMERICAN NATURALIST

spring is part of the energyexpended upon that individual offspringand

also part of the parent's reproductiveeffort. With progressiveprovisioning,
energy expended per offspringis difficultto measure, but in theory our
model still applies.
This model does not hold when the result of one reproductiveeffort
affectsfitnessof individual offspringin successive reproductiveeffortsof
the same parent. Social Hymenoptera (Wilson 1971) and a few birds
(Lack 1968) demonstratesuch an interactionof successive reproductive
effortswith the parental energy expenditureon individual offspring.In
bothgroups,offspring fromearly clutcheshelp to provisionoffspring from
later clutches.Feeding of siblingsis thoughtto be an evolutionaryconse-
quence of kin selection (Hamilton 1964; Wilson 1971), where the unit
under selectionis the extended genetic familyrather than the individual
parent. Our model thereforeapplies to partitioningof energyby the ex-
tended geneticfamilyratherthan by an individual parent.
For mammalsexhibitingpostpartumestrus,one littermay decrease the
energy available for another. The same problem may occasionally arise
when young fromsuccessivelittersdepend on a parent at the same time,
as in beaver (Bourliere 1956), chimpanzees (van Lawick-Goodall 1968),
and humans. Overlapping parental care of successivelitters generally re-
duces the total energyavailable to each litter.It need not affecthow the
reducedtotal is divided among individual offspring in any litter.We know
of no informationthat shows that less is
energy expended per individual
offspringin successivelitters.
A parameternot included in our model is the period of time over which
energy is collected for a reproductiveeffort.This parameter is directly
related to a significantproblemof the model. How does selectionrespond
to a situationin whichthe energyavailable for reproduction,when divided
by the optimum energy for each offspring,gives a clutch size of 1.5?
The problemof fractionaloffspring is greatestfor small clutches (Ricklefs
1968). In optimal clutches of 1.5, 33% of the reproductiveeffortis unused
by a real clutch of one, while for optimumlittersof 20.5, less than 3%
of the energyis unused by a real clutchof 20. Ricklefs (1968) summarized
extensiveempiricalevidencethat birds with smaller clutches (one or two)
have a greater between-speciesvariance in their developmentrates than
species with larger clutches (three to five). He argues that this difference
in developmentrates could be explained if clutcheswhich would include
fractionalyoungunder maximumdevelopmentrates and maximumfeeding
rates can be increasedto the next largestwholenumberby evolvingreduced
rates of development.Adjusting fractionsin small litterswould lead to a
greaterslowing of developmentrates that would occur for larger litters.
Empirical evidencefrombirds indicatesthat fractionalyoung are adjusted
by changes in the time over which energy is gathered. For our model,
problemsof time of energyaccumulationand fractionaloffspringappear
to cancel each other,the firstbeing adjusted to compensatefor the second.
A flexibilityin the time of energyaccumulationwould give flexibilityto

This content downloaded from 128.103.149.52 on Sun, 04 Oct 2015 22:00:37 UTC
All use subject to JSTOR Terms and Conditions
 SIZE AND NUMBER OF OFFSPRING 505

the numberof youngalong the abscissa of our model and make energycost
per young the only significantvariable on the abscissa (fig. 2).
Althoughour model probably applies to all organisms,it is of greatest
utilityin species with a large clutch size and no parental care. For these
speciesthe relativecost of offspring of differentsizes is closelyestimatedby
their relative caloric content.If the most common (and presumablyopti-
mum) size of offspring such as acorns or salmon eggs contained10 units of
energy,it would be safe to assume,on the basis of our model,that an off-
springcontainingnine units would have a fitnessless than nine-tenthsthat
of the 10-unitindividual. In mostcases fitnesswill be measuredby relative
survival. However,in species such as some desert annual plants, competi-
tion resultsin decreased fecundityratherthan increased mortality(Went
1955). The competitiveadvantage during early growthresultingfrom a
larger parental investment(e.g., larger seed size) may not be expressed
until a seed has grownup to reproduceitself.
Our model integratesassumptionscommonlymade by others. Its main
advantage is in allowing these ideas, which are usually separated, to be
consideredtogether,both graphicallyand analytically,in only two dimen-
sions.Because the model does not specifyhow energyis expended,or what
advantage is gained from it, effectsof competition,predation, and the
physical environmentcan be illustrated simultaneously.The effectsof
these three environmentalfactors can also be isolated and treated inde-
pendently.For example, a 10% decrease in the energy contentof a seed
mightbe totallyat the expense of endospermand embryoor totally from
seed coats or fromsome combinationof the two. If one could find seeds
within a species that expressed all the above alternatives,one could test
to see whetherthe reductionin endospermand embryoled to reduction
of survivalin competitionwithotherplants and reductionin seed coats led
to increasedmortalityfromseed eaters.This lets one measure the adaptive
value of the energyexpended in different tissues. Thus, our model can be
used to expressthe effectof competition,predation,or any otheraspect of
the environmenton the fitnessof offspring.

SUMMARY

The relationshipbetweenthe energy expended per offspring, fitnessof

offspring,and parental fitnessis presentedin a two-dimensionalgraphical
model. The validity of the model in determiningan optimal parental
strategyis demonstratedanalytically. The model applies under various
conditionsof parental care and sibling care for the offspringbut is most
useful for species that produce numeroussmall offspringwhich are given
no parental care.

ACKNOWLEDGMENTS

We thank Eric Pianka for criticismof the manuscript.Reviewers' com-

mentswere helpful in clarifyingsome arguments.

This content downloaded from 128.103.149.52 on Sun, 04 Oct 2015 22:00:37 UTC
All use subject to JSTOR Terms and Conditions
506 THE AMERICAN NATURALIST

LITERATURE CITED

Bourlibre,F. 1956. The naturalhistoryof mammals.2d rev. ed. Knopf, New York.
364 pp.
Hamilton,W. D. 1964. The geneticevolutionof social behavior.I and II. J. Theoret.
Biol. 7: 1-52.
Lack, D. 1968. Ecological adaptationsfor breedingbirds. Methuen,London. 409 pp.
Levins, R. 1968. Evolution in changingenvironments. Princeton UniversityPress,
Princeton,N.J. 120 pp.
Ricklefs,R. E. 1968. On the limitationof brood size in passerinebirds by the ability
of adults to nourishtheiryoung.Proc. Nat. Acad. Sci. 61: 847-851.
van Lawick-Goodall,J. 1968. The behaviorof free-livingchimpanzeesin the Gombe
StreamReserve.Anim. Behav. Monogr.1:161-311.
Went,P. 1955. The ecologyof desertplants. Sci. Amer. 192: 68-75.
Wilson,E. 0. 1971. The insectsocieties.Belknap,Cambridge,Mass. 548 pp.

This content downloaded from 128.103.149.52 on Sun, 04 Oct 2015 22:00:37 UTC
All use subject to JSTOR Terms and Conditions