Sex-Ratio Control Erodes Sexual Selection, Revealing Evolutionary Feedback From Adaptive Plasticity

Download as pdf or txt
Download as pdf or txt
You are on page 1of 6

Sex-ratio control erodes sexual selection, revealing

evolutionary feedback from adaptive plasticity


Tim W. Fawcett1,2,3, Bram Kuijper1,4, Franz J. Weissing, and Ido Pen
Theoretical Biology Group, University of Groningen, 9700 CC, Groningen, The Netherlands
Edited by Stuart West, Oxford University, Oxford, United Kingdom, and accepted by the Editorial Board August 1, 2011 (received for review April 11, 2011)

evolutionary equilibrium
plasticity sex-ratio bias

| Fisher process | good genes | phenotypic

onspicuous male ornaments such as brightly colored or


elongated feathers, loud vocalizations, and complex courtship dances are the hallmark of sexual selection, maintained
despite their obvious costs because females nd them attractive
(1). This evolutionary force has profound implications for many
life-history decisions, including which sex of offspring to produce
and how to invest in them (24). Provided the heritable benets
of ornamentation are to some degree sex-limited, selection
favors a conditional strategy of sex allocation: Females mated to
attractive, highly ornamented males should overproduce sons,
whereas those mated to unattractive males should overproduce
daughters (3). This pattern of sex allocation has been supported
by a number of theoretical (57) and empirical (810) studies,
but no one has considered how it might feed back to alter
sexual selection. Here, we investigate the dynamic interplay
between sexual selection and sex-ratio adjustment. We rst develop a simple mathematical model in which male ornamentation, female preference, and the sex-allocation strategy can
coevolve and use this model to determine the direction of selection acting on all of these traits. We then extend our analysis
to more biologically realistic conditions by using a series of
individual-based, evolutionary simulations, incorporating continuous variation in ornamentation and preference, a nite
population size, and stochastic factors such as genetic drift. This
dual approach allows us to uncover the evolutionary forces
linking sexual selection and sex allocation. After analyzing this
coevolutionary feedback process in depth, we show how the same
www.pnas.org/cgi/doi/10.1073/pnas.1105721108

principle extends to a wide range of other contexts in which selection favors phenotypic plasticity in response to a directionally
selected trait.
Model
Basic Scenario. For the sake of tractability, we consider just two
types of males, which differ in their ornamentation (5, 11, 12):
those of type 0 lack ornamentation, whereas those of type 1 are
ornamented to a degree given by the evolvable trait t (t $ 0).
Ornamentation is costly in that it reduces survival to adulthood,
with the relative survival of type-1 (compared with type-0) males
given by vm1 (vm1 # 1). This cost reects the energy or resources
invested in the development of secondary sexual traits or an
associated predation risk of being conspicuous (1).
Females are of one type only and have an evolvable preference
p (p $ 0), which is costly and lowers their survival to vf (vf # 1).
This cost may arise because the female has to invest in sensory
apparatus for assessing males, or because she incurs a higher
predation risk while choosing a mate (13). Her preference makes
her more inclined to mate with an ornamented than a nonornamented male, resulting in a proportion of females that
mate with the former type (note that is not xed but depends
on p, t, and the relative frequencies of type-0 and type-1 males).
Consequently, the expected number of mates is q1 for an ornamented male and q0 for a nonornamented male, with q1 $ q0.
Crucially, females can adjust offspring sex ratios in relation to
their partners ornamentation: Sex allocation is determined by
the evolvable traits s0 and s1, where s0 is the proportion of sons
produced when mated to a nonornamented male and s1 the
proportion when mated to an ornamented male. Ornamentation
is heritable from father to son, except when mutations occur:
With probability 0, the son of a nonornamented male is ornamented, and with probability 1, the son of an ornamented male
is nonornamented. In common with standard models of sexual
selection (14, 15), we assume that mutations are biased toward
the loss of ornamentation, i.e., that 1 > 0. This bias prevents
xation of the male ornament and thereby preserves the benet
of female choice (15). Individuals are assumed to die before their
offspring become reproductively mature, so that generations are
nonoverlapping. Fig. 1 summarizes the sequence of events in our
model. Table 1 lists the variables and parameters with their
associated symbols.

Author contributions: T.W.F., B.K., F.J.W., and I.P. designed research; T.W.F., B.K., and I.P.
performed research; T.W.F., B.K., and I.P. analyzed data; and T.W.F., B.K., F.J.W., and I.P.
wrote the paper.
The authors declare no conict of interest.
This article is a PNAS Direct Submission. S.W. is a guest editor invited by the Editorial
Board.
1

T.W.F. and B.K. contributed equally to this work.

To whom correspondence should be addressed. E-mail: tim.fawcett@cantab.net.

Present address: School of Biological Sciences, University of Bristol, Bristol BS8 1UG,
United Kingdom.

Present address: Department of Zoology, University of Cambridge, Cambridge CB2 3EJ,


United Kingdom.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.


1073/pnas.1105721108/-/DCSupplemental.

PNAS | September 20, 2011 | vol. 108 | no. 38 | 15925e15930

EVOLUTION

Female choice is a powerful selective force, driving the elaboration


of conspicuous male ornaments. This process of sexual selection
has profound implications for many life-history decisions, including sex allocation. For example, females with attractive partners
should produce more sons, because these sons will inherit their
fathers attractiveness and enjoy high mating success, thereby
yielding greater tness returns than daughters. However, previous research has overlooked the fact that there is a reciprocal
feedback from life-history strategies to sexual selection. Here, using
a simple mathematical model, we show that if mothers adaptively
control offspring sex in relation to their partners attractiveness,
sexual selection is weakened and male ornamentation declines. This
weakening occurs because the ability to determine offspring sex
reduces the tness difference between females with attractive
and unattractive partners. We use individual-based, evolutionary
simulations to show that this result holds under more biologically
realistic conditions. Sexual selection and sex allocation thus interact
in a dynamic fashion: The evolution of conspicuous male ornaments
favors sex-ratio adjustment, but this conditional strategy then
undermines the very same process that generated it, eroding sexual
selection. We predict that, all else being equal, the most elaborate
sexual displays should be seen in species with little or no control
over offspring sex. The feedback process we have described points
to a more general evolutionary principle, in which a conditional
strategy weakens directional selection on another trait by reducing
tness differences.

approach involves three basic steps: (i) Determine the dynamics


of a resident population with trait value x for a given trait of
interest; (ii) determine the invasion tness of a rare mutant with
alternative trait value ^x within this resident population; and (iii)
determine how mutant tness w depends on x and ^x. Below we
outline these steps in detail.
Dynamics of the Resident Population. We rst consider the dynamics of a resident population with ornamentation level t,
preference p, and sex-allocation traits s0 and s1. The numbers of
females, type-0 males, and type-1 males change from one generation to the next according to the transition matrix
A

3
2
1 1 s0 1 s1 vf
q0 1 s0 vf
q1 1 s1 vf
1 4
q1 s1 1 vm0 5:
k 1 s0 1 0 s1 1 vm0 q0 s0 1 0 vm0
2
1 s0 0 s1 1 1 vm1
q0 s0 0 vm1
q1 s1 1 1 vm1
[1]

Fig. 1. Summary of the sequence of events in each generation of our


model. The survival of females (vf) and ornamented males (vm1) to reproduction is reduced by the cost of their preference and ornament, respectively (nonornamented males do not pay a cost; vm0 = 1). Reproducing
females give a proportion of their matings to ornamented males, resulting
in an average number of mates q1 for ornamented males compared with q0
mates for nonornamented males (q1 $ q0). For each offspring produced,
the probability that it is a son is s0 for females with nonornamented partners
and s1 for those with ornamented partners. With mutation probability 0,
the son of a nonornamented male is ornamented, whereas with mutation
probability 1, the son of an ornamented male is nonornamented (1 > 0,
i.e., mutations are biased toward the loss of ornamentation).

Overview of the Method. Given this setup, we can use a repro-

ductive value approach (1618) to obtain selection differentials


for male ornamentation, female preference, and conditional sex
allocation, which together describe how the system evolves. This
Table 1. Variables and parameters used in the model
Symbol
yf
ymi
zf
zmi
vf
vmi

qi
si
i

The factor 1/2 is a formality to prevent offspring being counted


twice (once via its mother and once via its father), whereas the
constant k is a scaling factor (equivalent to the average clutch
size) to ensure that the population is stable (in technical terms,
to ensure that the dominant eigenvalue is 1; see refs. 18 and 19).
The leftmost column of A represents the per-capita reproductive
output of females, the center column that of type-0 males, and
the rightmost column that of type-1 males. The three rows represent, from top to bottom, the result of this reproductive output
in terms of surviving females, type-0 males, and type-1 males in
the next generation.
The entries in the matrix are derived from the basic assumptions of our model. To give an example, take the leftmost entry
in the center row, which represents the reproductive contribution
of mothers to type-0 males in the next adult generation. There
are two scenarios in which a female gives birth to a type-0 son:
either she mates with a type-0 male (probability 1 ) and
produces a son (probability s0) who is unaffected by mutation
(probability 1 0), or she mates with a type-1 male (probability
) and produces a son (probability s1) who mutates to a nonornamented state (probability 1). In either case, the survival of
that son to reproductive age is vm0. The other entries in matrix A
are derived by using similar logic, detailed in SI Model.
The relative frequencies of females (yf), type-0 males (ym0),
and type-1 males (ym1) change from one generation to the next
according to the dynamic equation yt+1 = Ayt, where y is the
column vector (yf, ym0, ym1)T representing the relative frequencies
in the current generation (yt) and the next generation (yt+1),
respectively (note that the superscript T indicates transposition).
Explicit equations for yf and ym1 are given in SI Model.
Invasion Fitness of a Rare Mutant. Now we ask whether rare
mutants with different values for the traits of interest can invade
the resident population. The dynamics of mutants are governed
by a matrix similar to A, with the appropriate parameters
replaced by their mutant counterparts:

Meaning
Relative frequency of females
Relative frequency of type-i males (i = 0,1)
Class-specic individual reproductive value of females
Class-specic individual reproductive value of type-i males
Viability of females
Viability of type-i males
Probability that a female mates with a type-1 male
Average number of mates per type-i male
Proportion of sons produced when mated to a type-i male
Probability that the son of a type-i male mutates into the
alternative type

15926 | www.pnas.org/cgi/doi/10.1073/pnas.1105721108

2
^1 ^s0
^1 ^s1 ^vf
1
1 4
^^s0 1 0
^^s1 1 vm0
k 1
2
^^s0 0
^^s1 1 1 ^vm1
1

q0 1 s0 ^vf
q0 s0 1 0 vm0
q0 s0 0 ^vm1

3
^q1 1 s1 ^vf
^q1 s1 1 vm0 5:
^q1 s1 1 1 ^v^m1
[2]

Mutant phenotypes are equipped with a hat (^) to distinguish


^ that
them from resident phenotypes. Note that the probability
a mutant female mates with a type-1 male is distinct from the
corresponding resident probability , because it is determined by
the females mutated preference ^
p. Her viability ^vf also depends
on ^
p, whereas that of mutant males ^vm1 depends on their mutant
level of ornamentation ^t. In contrast, because type-0 males lack
Fawcett et al.

vw
vB
zT y=zT y;
v^x
v^x

[3]

where y represents the relative frequencies of females, type0 males, and type-1 males in the resident population (technically,
a dominant right eigenvector of A), z = (zf, zm0, zm1)T represents
their reproductive values (technically, a dominant left eigenvector
of A) and the derivatives are evaluated at the resident trait values.
Results
Analytical Results. Using the approach (5, 21) outlined in SI
Model, we can use Eq. 3 to obtain the following selection differentials (16) for the traits p, t, s0, and s1, evaluated at the
resident trait values (i.e., where ^p p, ^t t, ^s0 s0 , and ^s1 s1 ):



vw
zm1 zm0
v9f
9
yf zf yf

v^p
q1
q0
vf

[4]

vw q91
v9m1
zm1 ym1
zm1 ym1

q1
vm1
v^t

[5]



vw 1 zm0
1

yf
v^s0
s0 q0 21 s

[6]



vw zm1
1
yf ;

v^s1 s1 q1 21 s

[7]

where s 1 s0 s1 is the average offspring sex ratio.


Primes (9) denote differentiation with respect to the trait
under consideration.
At the equilibrium for the sex-allocation traits s0 and s1, the
selection differentials given by Eqs. 6 and 7 must be zero (20),
and so zm0 =q0 1=21 s and zm1 =q1 1=21 s. Thus, we
have zm0 =q0 zm1 =q1 , which implies that the rst term on the
right of Eq. 4 vanishes as well. Assuming that v9f dvf =dp is
negative, i.e., that female choice is costly (13), it follows that the
selection differential for p is negative. Hence, at the sex-allocation equilibrium, selection cannot sustain a costly female preference. Sex-ratio adjustment dependent on male ornamentation
erodes the female preference to zero, and as a result, male ornamentation will evolve to zero as well.
Numerical Results. A numerical implementation of this analytical

model, illustrated in Fig. 2, shows how conditional sex-ratio adjustment erodes sexual selection (see SI Model for full details of
the calculation). Initially, we x the sex-allocation traits at s0 =
s1 = 0.5, such that offspring sex ratios are unbiased (Fig. 2, left of
the vertical dashed lines). Under these conditions, male ornaFawcett et al.

mentation and female preference evolve away from their survival


optima (at zero elaboration and zero preference, respectively) to
a stable, exaggerated level (Fig. 2A), following predictions from
standard models of sexual selection (14, 15, 22, 23). Then, from
the point indicated by the dashed lines, we allow the sex-allocation traits s0 and s1 to evolve. Conditional sex-ratio adjustment
evolves as predicted by theory (5, 6): Females mated to highly
ornamented males have more sons than those mated to lessornamented males (Fig. 2B, to the right of the dashed line).
[Note that s1 is prevented from deviating too far from 0.5 because of counterselection to restore an even population sex ratio,
because type-1 males vastly outnumber type-0 males (6).] As
biased sex allocation develops, however, this strategy weakens
sexual selection, leading to a gradual decline in male ornamentation and female preference (Fig. 2A, right of the dashed line).
Thus, sexual selection favors conditional sex allocation, but
this plasticity then erodes sexual selection. Two main processes
are responsible for this erosion. First, sex-ratio adjustment allows
females with unattractive partners to mitigate the tness disadvantage of low male ornamentation. In simple terms, ending
up with an unattractive male is not so disastrous if a female
can skew offspring production toward daughters. This plasticity
reduces the tness benet of female choosiness, which is selected
against because of its costs. Second, because choosier females
tend to mate with more ornamented males and, therefore, produce mainly sons, their strong preference genes will rarely be
expressed by their offspring. This masking lowers the average female preference in subsequent generations and, thereby, reduces

EVOLUTION

ornamentation altogether, their viability vm0 and per-capita


number of mates q0 are the same as for resident type-0 males and
so are left without a hat. The sex-allocation traits s0 and s1 receive
hats in the rst column, representing the reproductive output of
mutant females, but not in the second and third columns because
mutant males are assumed to mate with resident females only
(due to the rarity of mutant females). Note that when the mutants
trait values are the same as those of the resident (x ^x), matrices
A and B are identical.
The ability of mutant individuals to invade the resident population is given by their tness w, which is the dominant eigenvalue of matrix B. Assuming mutations of small effect, the
selection differential vw=v^x expresses how w depends on ^x, the
mutant value for the trait of interest. According to a standard
result (20) from evolutionary invasion analysis, this is

Fig. 2. Sex-ratio adjustment erodes sexual selection (numerical results). A


shows the level of male ornamentation (t, blue) and female preference
(p, pink; note that this partly obscures the blue line), whereas B shows the
proportion of sons produced by females mated to nonornamented (s0, light
green) and ornamented (s1, dark green) males. Offspring sex ratios are initially unbiased (s0 = s1 = 0.5), but they are allowed to evolve from the point
indicated by the vertical dashed lines. Male ornamentation and female
preference reach a stable level of exaggeration in the absence of sex-ratio
bias, then decline to zero as conditional sex allocation develops (subject
to counter selection on s1 to restore an even population sex ratio; ref. 6) and
s0 and s1 reach their optima. For the example shown, cp = 0.2, cf = 0.001, cm =
0.1, 0 = 0.02, and 1 = 0.3; the starting values for ornamentation and
preference were t = 0.5 and P = 1.5.

PNAS | September 20, 2011 | vol. 108 | no. 38 | 15927

the tness benet of male ornamentation. In effect, the conditional strategy of sex allocation reduces the heritability of both low
attractiveness and strong preferences, undermining selection to
invest in costly ornamentation.
Individual-Based Simulations. Using individual-based computer
simulations, we can extend this analysis to a more realistic situation where male ornamentation and female preference vary
continuously and the evolutionary dynamics are subject to stochastic demographic factors. We simulated a nite population in
which a costly male ornament and a costly female preference
could change over time through selection and mutation (see
Materials and Methods for full details). As in the earlier numerical
results, evolutionary change in the male ornament and female
preference follows predictions from standard analytical models of
sexual selection (14, 15, 22, 23), with both traits quickly evolving to
a stable, exaggerated level (Fig. 3A, left of the dashed line).
Similar patterns are seen regardless of whether male ornamentation is an arbitrary Fisherian trait (Fig. 3) or is a conditiondependent indicator of good genes (see additional simulation
results in SI Results and Fig. S1).
We then allowed a conditional strategy of sex-ratio adjustment
to evolve by incorporating two additional traits, s and s+ (6).
These traits determine a females sex-allocation strategy, with s
(0 # s # 1) being the chance of producing a son when mated to
a male with below-average ornamentation and s+ (0 # s+ # 1)
that when mated to a male with above-average ornamentation.
Starting from a situation in which offspring sex ratios are unbiased (s = s+ = 0.5), conditional sex-ratio adjustment gradually
develops as predicted by theory (5, 6): Females mated to highly
ornamented males overproduce sons, whereas those mated to
less-ornamented males overproduce daughters (Fig. 3B). [Note
that with continuous variation in male ornamentation, s and s+
become biased to a similar extent (6).] This strategy then
weakens sexual selection, leading to a gradual decline in male
ornamentation and female preference (Fig. 3A, right of the
dashed line).
When male ornamentation is a condition-dependent indicator
of good genes (23), sexual selection is weakened to a lesser extent than when it is a purely Fisherian trait (SI Results). In the
former case, the heritable benets for a female who mates with
an attractive male are not entirely sex-limited; although only her
sons can prot from their fathers ornamentation genes, both her
daughters and her sons will inherit his genes for viability. Thus,
even when females exert a great degree of control over the sex of
their offspring, it still pays to mate with more ornamented males.
This difference notwithstanding, for both Fisherian and good
genes models of sexual selection, ornamentation and preference
are substantially reduced as conditional sex allocation develops.
To check that sex-ratio adjustment is directly responsible for
this decline, we ran another set of simulations in which strategies
with varying degrees of sex-ratio bias were introduced partway
through (Fig. 4). Initially, with the sex ratio xed at 0.5, the male
ornament and female preference quickly evolve to a stable, exaggerated level as before. We then introduced a biased sexallocation strategy for all females, causing them to produce more
sons when mated to an attractive partner and more daughters
when mated to an unattractive partner. The effect on sexual
selection is dramatic. For a moderate degree of bias, s+ = 0.7
and s = 0.3, very rapidly the ornament and preference drop to
approximately one-half of their original level of expression.
Adjustment strategies involving weaker biases result in a smaller
drop, whereas with stronger biases the decline in ornamentation
is even sharper (Fig. 4).
Our simulation results conrm that the equilibrium levels of
female preference and male ornamentation are substantially
lower when sex-ratio adjustment is possible. In effect, sexual
15928 | www.pnas.org/cgi/doi/10.1073/pnas.1105721108

Fig. 3. Coevolutionary dynamics of sexual selection and sex allocation (individual-based simulations). A shows the level of male ornamentation
(t, blue) and female preference (p, pink) under Fisherian sexual selection,
whereas B shows the female sex-allocation strategy in the same set of simulations based on traits s+ and s, where s+ (dark green) is the probability of
producing a son when her partner has above-average ornamentation and s
(light green) is the probability of producing a son when he has below-average
ornamentation. Offspring sex ratios are initially unbiased (s+ = s = 0.5), but
they are allowed to evolve from the point indicated by the vertical dashed
lines. All values are shown as the mean (solid line) SD (stippling) from 20
replicate simulation runs. For parameter values, see Materials and Methods.

selection undermines itself by favoring a conditional strategy of


sex-ratio adjustment based on male attractiveness.
Discussion
Previous theory (5, 6) has conrmed the empirical suggestion (3)
that variation in male sexual displays favors conditional sex allocation by females. Here, we have shown an unexpected consequence of this process: that by reducing the tness difference
between females with attractive and unattractive partners, this
sex-allocation strategy undermines the same selective force that
created it, causing male ornamentation to decline. Moreover,
because choosier females tend to mate with more highly ornamented males and, therefore, produce sons, their stronger preference genes are likely to be masked in the next generation,
weakening sexual selection still further. Our evolutionary simulations predict a lower level of sexual display than in cases where
facultative sex-ratio adjustment is not possible. This nding
implies that, all else being equal, the most exaggerated secondary
sexual traits should be seen in species with little or no control over
offspring sex. For instance, we might expect that species with
genotypic sex determination will have more exaggerated sexual
ornamentation than closely related species with temperaturedependent sex determination, assuming that the latter mechanism
affords parents greater control over the sex of their offspring.
It is known that the evolution of phenotypic plasticity in a
quantitative trait can alter the evolution of the average phenotype for that trait (2427). Here, we have shown a related effect:
that plasticity in one trait (sex-ratio bias) can alter the evolution
of another trait (ornamentation) on which it is conditional. We
propose that this phenomenon is not restricted to sex allocation,
but is an example of a more general principle. Whenever heritable variation in tness is maintained for a given trait, selection
Fawcett et al.

example, through niche construction, then the strength of selection on dispersal will be weakened. Selection favors a strategy
that mitigates the tness disadvantage of staying in the natal
habitat, and this plasticity erodes directional selection on the
ability to disperse away from that habitat.
These diverse examples show that our model applies to a broad
range of contexts. The evolutionary feedback process we have
described is likely to be a widespread and important force maintaining phenotypic variation in the face of directional selection.

Fig. 4. Stronger degrees of sex-ratio bias have stronger eroding effects on


sexual selection. The plots show the level of male ornamentation (t, blue)
and female preference (p, pink). Offspring sex ratios are initially unbiased
(s+ = s = 0.5), but from the point indicated by the vertical dashed line,
females use a xed sex-allocation strategy conditional on the males ornamentation. Three different degrees of sex-ratio bias are shown: s+ = s = 0.5
(5050, i.e., no bias); s+ = 0.7, s = 0.3 (7030); and s+ = 0.9, s = 0.1 (9010). All
values are shown as the mean (solid line) SD (stippling) from 20 replicate
simulation runs. For parameter values, see Materials and Methods.

should favor any conditional strategy that improves the tness


prospects of the least successful phenotypes, but in doing so, it
erodes selection on the trait.
To illustrate the general nature of our argument, we give
examples from a range of contexts that do not involve sex allocation. The rst concerns kleptoparasitism (28), in which one
animal steals food that a conspecic has caught before the latter
can eat it. Selection for good hunting skills is expected to be
strong in any predatory species, but there may still be substantial
variation in hunting success because of mutations in polygenic
traits affecting the development of motor skills. If poor hunters
adopt kleptoparasitic behavior, however, this conditional strategy will reduce tness differences based on hunting success and,
thereby, weaken selection on hunting ability. Combined with the
costs incurred by parasitized hunters, the weakened selection may
lead to a decline in hunting skills that, in turn, will reduce the
benets of stealing. Thus, selection on hunting ability and kleptoparasitism interact in a highly dynamic fashion.
The second example involves polygynous mating systems in
which access to females is determined by male dominance relations. In such systems, there will be strong selection for male
characteristics related to dominance, such as large body size. Slight
differences between males in these characteristics early in life may
largely determine their relative positions in the dominance hierarchy, leading to substantial differences in lifetime reproductive
success. If small males adopt a sneaker tactic (29), however,
allowing them to achieve signicant reproductive success by subversive means, this conditional strategy will reduce tness differences between males of high and low dominance rank and, thereby,
weaken selection on body size. This weakening of selection, in turn,
will alter the selectionmutation balance, allowing greater levels of
genetic variation for body size to persist in the population.
Our nal example deals with costly dispersal. In many plant
and animal species, dispersal away from the natal habitat may be
favored despite the energetic cost or mortality risk associated
with this movement. The benets of dispersal will typically be
frequency-dependent, with the greatest pressure to disperse occurring when most individuals stay at home. However, if individuals that forgo dispersal can adapt better to overcrowding, for
1. Andersson M (1994) Sexual Selection (Princeton Univ Press, Princeton, NJ).
2. Trivers RL, Willard DE (1973) Natural selection of parental ability to vary the sex ratio
of offspring. Science 179:90e92.
3. Burley N (1981) Sex ratio manipulation and selection for attractiveness. Science 211:
721e722.
4. West SA (2009) Sex Allocation (Princeton Univ Press, Princeton, NJ).

Fawcett et al.

Details of the Individual-Based Simulations. The individual-based simulations


were similar to those described in an earlier paper (6). In the main text, we
focus on Fisherian sexual selection, whereas the simulations for good genes
sexual selection are presented in SI Results. We modeled a population of
5,000 individuals, each with diploid, autosomal genetic values for the following traits: p, coding for preference (expressed only by females); t, coding
for ornamentation (expressed only by males); and two sex-allocation traits, s
and s+ (expressed only by females). The value for p can take any real number,
whereas t is limited to positive values and s and s+ are limited between
0 and 1. We chose to restrict t to positive values because this range might
better represent certain forms of male display (30), for example, the height
of a plumage crest, but we obtain similar results when male ornamentation
can also take negative values (see additional simulation results in SI Results
and Fig. S2). Female preference and male ornamentation are both assumed
to be costly; survival to maturity is maximized for p = 0 and t = 0 and declines
away from these optima as specied by the functions exp cf p2 and
exp cm t 2 , where cf and cm are positive constants.
For reproduction, females are drawn from the population with a chance
proportional to their survival probability. Each surviving female then samples
10 males, again weighted by survival probability, and chooses one of them on
the basis of his ornamentation. The chance that she picks a given male is
proportional to expcp pt, where cp is a positive constant scaling the importance of ornamentation to female choice. Thus, females with a positive
preference (p > 0) prefer more ornamented males, those females with
a negative preference (p < 0) prefer less ornamented males and those
females with p = 0 mate randomly. To facilitate sexual selection, we started
the simulations with a positive preference (22, 30); the same process occurs
when starting from a situation of random choice, but it takes longer.
Each mating produces a single offspring, whose genetic values are determined by standard Mendelian inheritance. We assume that there is no
genetic dominance and that the loci are unlinked. Offspring sex is determined
by the fathers ornamentation and the mothers sex-allocation strategy: the
probability of producing a son is s+ when the fathers ornamentation level is
above average and s when it is below average. For each trait, we assume
that mutations occur in a small fraction of offspring (with probability p for
p, t for t, and s for s and s+), causing the genetic value to change upward
or downward by an amount drawn from a uniform probability distribution
(up to a certain maximum amount). Upward and downward mutations are
equally likely except in the ornamentation trait t, for which we assume that
a downward mutation bias reduces ornamentation by an average amount g
(6, 15). Reproduction continues until a total of 5,000 offspring have been
produced, at which point all of the adults die and are replaced by the offspring generation. The same cycle of events was repeated for 100,000
generations, which is the timespan depicted in our gures. Computer code
for the simulations is available from the authors upon request.
For the results shown in the main text, the parameter values were cp = 1.0,
cf = 0.001, cm = 0.5, p = t = s = 0.05 and g = 0.02, with the average genetic
 1, t 0, and s s 0:5. Howvalues in the initial population set at p
ever, the eroding effect of sex-ratio adjustment is seen for a wide range of
parameter values, whenever sexual selection leads to exaggerated male
ornamentation.
ACKNOWLEDGMENTS. We thank John McNamara, Tobias Uller, the editor,
and two anonymous referees for constructive feedback on the manuscript.
This research was funded by The Netherlands Organization for Scientic
Research Grant 810.67.021 (to F.J.W.).

5. Pen I, Weissing FJ (2000) Sexual selection and the sex ratio: An ESS analysis. Selection
1:59e69.
6. Fawcett TW, Kuijper B, Pen I, Weissing FJ (2007) Should attractive males have more
sons? Behav Ecol 18:71e80.
7. Blackburn GS, Albert AYK, Otto SP (2010) The evolution of sex ratio adjustment in the
presence of sexually antagonistic selection. Am Nat 176:264e275.

PNAS | September 20, 2011 | vol. 108 | no. 38 | 15929

EVOLUTION

Materials and Methods

8. Ellegren H, Gustafsson L, Sheldon BC (1996) Sex ratio adjustment in relation to


paternal attractiveness in a wild bird population. Proc Natl Acad Sci USA 93:
11723e11728.
9. Sheldon BC, Andersson S, Grifth SC, rnborg J, Sendecka J (1999) Ultraviolet colour
variation inuences blue tit sex ratios. Nature 402:874e877.
10. Pike TW, Petrie M (2005) Offspring sex ratio is related to paternal train elaboration
and yolk corticosterone in peafowl. Biol Lett 1:204e207.
11. Kirkpatrick M (1982) Sexual selection and the evolution of female choice. Evolution
36:1e12.
12. Kokko H, Brooks R, McNamara JM, Houston AI (2002) The sexual selection continuum.
Proc Biol Sci 269:1331e1340.
13. Pomiankowski A (1987) The costs of choice in sexual selection. J Theor Biol 128:
195e218.
14. Bulmer M (1989) Structural instability of models of sexual selection. Theor Popul Biol
35:195e206.
15. Pomiankowski A, Iwasa Y, Nee S (1991) The evolution of costly mate preferences. I.
Fisher and biased mutation. Evolution 45:1422e1430.
16. Taylor PD (1996) The selection differential in quantitative genetics and ESS models.
Evolution 50:2106e2110.
17. Pen I, Weissing FJ (2000) Sex-ratio optimization with helpers at the nest. Proc Biol Sci
267:539e543.
18. Pe I, Weissin FJ (2000) Towards a unied theory of cooperative breeding: The role of
ecology and life history re-examined. Proc Biol Sci 267:2411e2418.

15930 | www.pnas.org/cgi/doi/10.1073/pnas.1105721108

19. Mylius SD, Diekmann O (1995) On evolutionarily stable life histories, optimization and
the need to be specic about density dependence. Oikos 74:218e224.
20. Otto SP, Day T (2007) A Biologists Guide to Mathematical Modeling in Ecology and
Evolution (Princeton Univ Press, Princeton, NJ).
21. Pen I, Weissing FJ (2002) Sex Ratios: Concepts and Research Methods, ed Hardy ICW
(Cambridge Univ Press, Cambridge, UK), pp 26e45.
22. Lande R (1981) Models of speciation by sexual selection on polygenic traits. Proc Natl
Acad Sci USA 78:3721e3725.
23. Iwasa Y, Pomiankowski A, Nee S (1991) The evolution of costly mate preferences. II.
The handicap principle. Evolution 45:1431e1442.
24. Kirkpatrick M (1982) Quantum evolution and punctuated equilibria in continuous
genetic characters. Am Nat 119:833e848.
25. Frank SA, Swingland IR (1988) Sex ratio under conditional sex expression. J Theor Biol
135:415e418.
26. Pl C, Mikls I (1999) Epigenetic inheritance, genetic assimilation and speciation.
J Theor Biol 200:19e37.
27. Lande R (2009) Adaptation to an extraordinary environment by evolution of phenotypic plasticity and genetic assimilation. J Evol Biol 22:1435e1446.
28. Brockmann HJ, Barnard CJ (1979) Kleptoparasitism in birds. Anim Behav 27:487e514.
29. Gross MR (1996) Alternative reproductive strategies and tactics: Diversity within sexes.
Trends Ecol Evol 11:92e98.
30. Kokko H, Jennions MD, Brooks R (2006) Unifying and testing models of sexual selection. Annu Rev Ecol Evol Syst 37:43e66.

Fawcett et al.

You might also like

pFad - Phonifier reborn

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.

Note: This service is not intended for secure transactions such as banking, social media, email, or purchasing. Use at your own risk. We assume no liability whatsoever for broken pages.


Alternative Proxies:

Alternative Proxy

pFad Proxy

pFad v3 Proxy

pFad v4 Proxy