You are on page 1of 3

Journal of Theoretical Biology 359 (2014) 233235

Contents lists available at ScienceDirect

Journal of Theoretical Biology


journal homepage: www.elsevier.com/locate/yjtbi

Total reproductive values for females and for males in sexual diploids
are not equal
Alan Grafen a,b,n
a
b

St. John's College, Oxford OX1 3JP, United Kingdom


Zoology Department, University of Oxford, South Parks Road, Oxford OX1 3PS, United Kingdom

A U T H O R - H I G H L I G H T S







Fisher's famous sex ratio argument has a hitherto unnoticed lacuna.


I perform an age- and sex-structured population calculation of reproductive value.
Fisher's result is recovered with the additional assumption of demographic stability.
Total reproductive value of a sex is proportional to its average age at parentage.
Total reproductive values of female and male offspring are equal.

art ic l e i nf o

a b s t r a c t

Article history:
Received 10 February 2014
Received in revised form
1 May 2014
Accepted 12 May 2014
Available online 21 May 2014

A very simple mathematical exposition of reproductive value in an age- and sex-structured sexual
diploid population employs reproductive value as the probability that a random gene in a distant
generation traces its ancestry to a given individual or set of individuals today. Although the total
reproductive values of all females and that of all males are not in general equal, but instead proportional
to the average age of a new mother and a new father, respectively, Fisher's equal-investment conclusion
for the sex ratio remains valid because the total reproductive value of age-zero females equals the total
reproductive value of age-zero males. However, the conclusion is seen to require an extra assumption,
namely stability of the age-distribution.
& 2014 Elsevier Ltd. All rights reserved.

Keywords:
Sex ratio
Equal investment
Diploidy
Overlapping generations
Structured populations

1. Introduction
Fisher (1930) introduced the concept of reproductive value, and
made two quite different applications of it, both with far-reaching
effects. Dividing the population into age classes, he attributed a
reproductive value to individuals of different ages, and this
quantity is central in modern demography. The central point of
the argument in the second application begins by considering how
parents invest in offspring, and continues
Let us consider the reproductive value of these offspring at the
moment when this parental expenditure on their behalf has
just ceased. If we consider the aggregate of an entire generation
of such offspring it is clear that the total reproductive value of
n
Correspondence address: St. John's College, Oxford OX1 3JP, United Kingdom.
Tel.: 44 1865 277438.
E-mail address: alan.grafen@sjc.ox.ac.uk

http://dx.doi.org/10.1016/j.jtbi.2014.05.021
0022-5193/& 2014 Elsevier Ltd. All rights reserved.

the males in this group is exactly equal to the total value of all
the females, because each sex must supply half the ancestry of
all future generations of the species (Fisher, 1930, p. 142),
and Fisher goes on to draw the basic sex ratio result that equal
investment is the outcome of evolution under assumptions including panmixia and diploidy. With discrete, non-overlapping generations, it does indeed follow that the female offspring and male
offspring must have exactly equal contributions to future generations, but with overlapping generations, it is even possible, for
example, that this particular generation of males never mates, and
the females mate with older males. Formal treatments of this
argument (e.g. Shaw and Mohler, 1953; Bodmer and Edwards,
1960) usually assume discrete non-overlapping generations. Fisher's argument therefore seems to depend on non-overlapping
generations, but is frequently applied in other situations (West,
2009). The current paper calculates reproductive values for a
population divided simultaneously by age and sex, works out the

234

A. Grafen / Journal of Theoretical Biology 359 (2014) 233235

consequences for aggregate reproductive values of all females


versus all males and female offspring versus male offspring, and
then considers the robustness of the equal investment result.

2. Notation and derivation


We consider a diploid dioecious population in discrete time
with overlapping generations. We assume that the population is
sufciently mixing that it makes sense to select a random gene in
some very distant generation, and dene the probability that its
unique ancestral path passes at time t to an offspring (a 0) from
an individual of a given age, a A f1; 2; 3g and sex s A ff ; mg, say
utsa. We consider the paths to pass just after time t, so that at time
t itself the path is with the donor not the recipient of the gene.
Autosomal genes are studied, so whenever a gene's ancestral path
jumps from one individual to a parent, there are equal chances
that the parent is female and male, and we assume that the
probability distributions of the ages of the new parents at time t
are ptfa and ptma (so a ptsa 1). We assume that the gene and its
alleles are selectively neutral. A path that jumps from a parent of
age a to an offspring must be present in the parent all the way
back to age zero, when it passes to one of the parent's own
parents. Let the chance that an ancestral path moves between
individuals at time t be ht, so the chance that it remains within the
same individual is 1  ht . We can write this as
utsa

ht
p :
2 tsa

The probability that a random ancestral path at time t is present in


an individual of a given age a and sex s, vtsa, comprises two
mutually exclusive possibilities, that the ancestral path passes to a
descendant in this period, and that the path is present at time t 1
in an individual of age a 1. An ancestral path is always somewhere at each time, so s;a vtsa 1. The probabilities of mutually
exclusive events add, so
1

vtsa utsa vt 1;s;a 1 ut i;s;a i


i0

and in particular for a 0


1

vts0 ut i;s;i :
i0

We now assume constant demography, so that none of the


quantities varies with time, and we accordingly omit the t-subscript.
The above equations now imply
1

vf 0 ufi
i0

1
h 1
h h 1
p p u vm0 :
2 i 0 fi 2 2 i 0 mi i 0 mi

Hence the total reproductive value of all age-zero females


equals the total reproductive value of all age-zero males. We note
that each equals h/2, which makes sense as we are dening
reproductive value as the probability that a random distant gene
traces back to the newborns today. The result is a natural one. The
probability that an ancestral path intersects newborn females
today equals the probability that it jumped into a one-year old
female in one year's time, plus the probability that it jumped into a
two-year old female in two years time, and so on. Now under
stable demography, these probabilities do not depend on the time
at which the jump occurs, and so we may align them all in the
same year, converting a diagonal sum across years into a vertical
sum, which equals the chance that the path jumps into a female of
any age this year. But these vertical sums must be equal for males
and females, for next year's newborns have one father and one
mother, and a gene goes equally likely to one and the other.

Turning to the total reproductive value over all ages, we obtain


1

vsa ua i;s aua;s :


a

a i0

As a usa h=2, we conclude that the ratio of total reproductive


value of all females to the reproductive value of age-zero females
equals the average age of a new mother, and mutatis mutandis for
males; and so the ratio of the total reproductive value of all
females to that of all males equals the average age of a new mother
divided by the average age of a new father. This result has a simple
interpretation. As we trace back the ancestral path of a given allele,
it leaps backwards into female and male bodies, and with equal
probability at each leap. Its average sojourn time in a female body
is the average age of a new mother, and in a male body is the
average age of a new father. Hence the chance of nding it at any
one point in time in a given sex is proportional to the average age
at which the path enters that sex.
Note that the sum of the total reproductive value of females
and of males equals one, as the ancestral path of a future allele
rests in exactly one individual at any given time t. Hence the total
reproductive value of all females equals the average age of a new
mother divided by the sum of the average ages of a new mother
and a new father.

3. Discussion
The total reproductive value of all females and of all males is
not in general equal in dioecious diploids in discrete time with
overlapping generations, casting doubt on a crucial step in Fisher's
sex ratio argument. In view of the empirical success of sex ratio
theory (West, 2009), it is fortunate that his equal investment
conclusion survives a more general analysis. However, it was
necessary to make the assumption of demographic stability to
achieve this result. The total reproductive values of age-zero
females and males are indeed equal in this case, and Fisher's
argument can therefore proceed unhindered from the passage
quoted above to the famous conclusion of equal investment. The
total reproductive values of all females and of all males are shown
very naturally to depend on how old an average new mother and
average new father are.
Sex ratio is a major topic in behavioural ecology today, and the
standard reference is West (2009). Following Hamilton (1967),
many exceptions to Fisher's conclusion are now conceptually wellunderstood and empirically well-documented as not meeting his
assumptions, and sex ratio has the best quantitative match
between theory and data in modern adaptationist biology. One
relevant nding here is that unequal reproductive values of all
females and all males have been noted before, in the study of sex
ratio in Hymenoptera with overlapping generations, in the socalled sphecid and halictid patterns of partial bivoltinism (Seger,
1983; Grafen, 1986), which are simple examples where demographic stability does not hold for every single generation (though
it could be reasonable to assume that it held for pairs of generations so that all even generations were the same as each other, and
all odd generations were the same as each other). However, those
are haplodiploid models, and the phenomenon has not to my
knowledge been noted in diploid models before. Whereas nonoverlapping generations ensure equal total reproductive value of
females and males in each generation, overlap of generations will
generically cause differences to arise between them.
This lacuna in Fisher's argument has not, so far as I know, been
noted before. Fortunately, as we have seen, the major implications
for sex ratio theory are unaffected by studying reproductive value
in a sex- and age-structured population under demographic stability.
But the need for this assumption raises questions. Does random

A. Grafen / Journal of Theoretical Biology 359 (2014) 233235

uctuation in age distribution affect the equilibrium sex ratio?


Does systematic (e.g. seasonal) uctuation in age distribution
affect it? How does any effect depend on an individual's information about the age distribution or the season? In the light of the
answers to those theoretical questions, how reasonable is the
assumption of constant demography in empirical applications?

Acknowledgments
The author thanks Katja Lehmann and Andy Gardner for
comments, and declares no conict of interest.

235

References
Bodmer, W., Edwards, A., 1960. Natural selection and the sex ratio. Ann. Hum.
Genetics 24, 239244.
Fisher, R.A., 1930. The Genetical Theory of Natural Selection. Oxford University
Press, Oxford, UK (see Fisher (1999) for a version in print).
Fisher R.A.,1999. In: J.H. Bennett (Ed.), The Genetical Theory of Natural Selection,
Oxford University Press, Oxford, UK (Variorum Edition of 1930 OUP edition and
1958 Dover edition).
Grafen, A., 1986. Split sex ratios and the evolutionary origins of eusociality. J. Theor.
Biol. 122, 95121.
Hamilton, W.D., 1967. Extraordinary sex ratios. Science 156, 477488.
Seger, J., 1983. Partial bivoltinism may cause alternating sex-ratio biases that favour
eusociality. Nature 301, 5962.
Shaw, R.F., Mohler, J., 1953. The selective signicance of the sex ratio. Am. Nat. 87,
337342.
West, S.A., 2009. Sex Allocation. Princeton University Press, Princeton, New Jersey.