Theoretical Population Biology 85 (2013) 49–57

Contents lists available at SciVerse ScienceDirect

Theoretical Population Biology

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

Dwarf males, large hermaphrodites and females in marine species: A

dynamic optimization model of sex allocation and growth
Sachi Yamaguchi a,∗ , Kota Sawada b , Yoichi Yusa c , Yoh Iwasa a
a
Department of Biology, Faculty of Sciences, Kyushu University, Fukuoka 812-8581, Japan
b
Department of Evolutionary Studies of Biosystems, The Graduate University for Advanced Studies, Kanagawa 240-0193, Japan
c
Department of Biology, Faculty of Sciences, Nara Women’s University, Nara 630-8506, Japan

article info abstract

Article history: In this study, we investigate the evolutionarily stable schedule of growth and sex allocation for marine
Received 7 September 2012 benthic species that contain dwarf males. We consider a population in an ephemeral microhabitat that
Available online 14 February 2013 receives a constant supply of larvae. Small individuals can immediately reproduce as a dwarf male or
remain immature and grow. Large individuals allocate reproductive resources between male and female
Keywords: functions. The fraction c of newly settled individuals who remain immature and the sex allocation of
Dynamic programming
large individuals m are quantities to evolve. In the stationary ESS, if the relative reproductive success
Sex allocation
Ephemeral microhabitat
of dwarf males is greater than the survivorship of immature individuals until they reach a mature
size, then the population is a mixture of females and dwarf males. If the opposite inequality holds, the
population is dominated by hermaphrodites and lacks dwarf males. There is no case in which a mixture of
hermaphrodites and dwarf males to be the ESS in the stationary solution. The ESS can be solved by dynamic
programming when the strategies depend on the age of the microhabitat (c (t ) and m (t )). Typically, the
ESS schedule begins with a population composed only of hermaphrodites, which is replaced by a mixture
of dwarf males and hermaphrodites and then by a mixture of dwarf males and pure females. The relative
importance of these three phases depends on multiple parameters.
© 2013 Elsevier Inc. All rights reserved.

1. Introduction 1995a,b; Zardus and Hadfield, 2004; Buhl-Mortensen and Høeg,

Males that are much smaller than females or hermaphrodites
are widely observed among animals, particularly among marine
benthic species. These males are called ‘‘dwarf males’’ (Ghiselin,
1974; Vollrath, 1998; Yusa et al., 2012), and they occur together
with larger females (dioecious) or hermaphrodites (androdioe-
cious). Whereas some dwarf males are free-living (e.g., in spiders),
others are often attached to the body of a female or hermaphrodite
(e.g., in barnacles, Darwin, 1851; Høeg, 1995a,b, bivalves, Turner
and Yakovlev, 1983, annelids, Vollrath, 1998; Rouse et al., 2004,
and angler fish, Pietsch, 2005). By being physically close to their
potential mate, dwarf males might have an advantage in mating
access over their free-living competitors (Ghiselin, 1974).
Dwarf males often appear in populations of animals that are
symbiotic with other animals. For example, crustaceans that live
on arthropods, cnidarians, and turtles include several species
with dwarf males (e.g., Cash and Bauer, 1993; Crisp, 1983; Høeg,
∗ mortality. However, the studies by Yamaguchi et al. (2008, 2013)
Corresponding author.
E-mail addresses: sachi.dwarfmale@gmail.com (S. Yamaguchi),
sawada_kota@soken.ac.jp (K. Sawada), yusa@cc.nara-wu.ac.jp (Y. Yusa),
yohiwasa@kyudai.jp (Y. Iwasa).
0040-5809/$ – see front matter © 2013 Elsevier Inc. All rights reserved.
doi:10.1016/j.tpb.2013.02.001
2006; Yusa et al., 2001, 2010). Parasitic bivalves are another
example (Ó Foighil, 1985; Goto et al., 2007). However, other
animals with dwarf males utilize limited microhabitats, such as
sunken wood or vertebrate bones (bivalves: Turner and Yakovlev,
1983; annelids: Rouse et al., 2004). The habitats for such species
are often ephemeral, and even the molting of the host may be fatal
for symbiotic species (but see Itani et al., 2002). Therefore, the
possibility of habitat loss might affect the life history and sexual
systems of these species.
The evolution of dwarf males, particularly in barnacles, has been
studied using mathematical models (Charnov, 1982, 1987; Yam-
aguchi et al., 2008; reviewed in Yamaguchi et al., 2012). Charnov
(1982, 1987) and Yamaguchi et al. (2008, 2012) predicted that
limited mating opportunities favor dwarf males in androdioe-
cious/dioecious systems because hermaphrodites should allocate
less resources to male function in smaller mating groups. In ad-
dition, Yamaguchi et al. (2008, 2013) suggested that the benefits
of smallness, such as early maturation, might play an important
role when growth is limited by poor food availability and/or high
were based on heavy numerical analyses of dynamic programming
computation, and the mathematical results are difficult to follow
because of the complexity of the model.
50 S. Yamaguchi et al. / Theoretical Population Biology 85 (2013) 49–57
In this paper, we study a simple evolutionary game model
for marine organisms. This model can explain the conditions for
different patterns of sex expression and life history, including
dwarf males. We consider a population consisting of two size
classes. Small individuals may function as a dwarf male and
immediately achieve reproductive success. Alternatively, these
individuals may remain immature and grow quickly. When small
individuals reach a sufficiently large size, they can potentially
function as hermaphrodites. Small individuals can become a dwarf
male or a fast-growing immature individual, and large individuals
choose the reproductive allocation between male and female
functions. We construct this evolutionary game model and solve
the ESS for the model.
We initially consider the case in which strategies are indepen-
dent of time. In the stationary situation, the numbers of large in-
dividuals, dwarf males, and immature small individuals reach a
stationary state in which the new recruitment from settled lar-
vae are balanced with mortality and growth. The ESS is then one
of two situations: when growth is sufficiently fast, no dwarf males
appear and all large individuals are hermaphrodites. By contrast,
when growth is slow, some of the newly settled individuals be-
come dwarf males but the rest remain immature, and the latter
individuals become pure females when they reach adult size. The
population is a mixture of dwarf males and large females. There
is no case in which the hermaphrodites and dwarf males coex-
ist in the ESS of the model. However, this conclusion is not con-
sistent with observations of some goose (lepadomorph) barnacles,
in which large hermaphrodites coexist with dwarf males (Darwin,
1851; Klepal, 1987; Yusa et al., 2001).
To address this problem, we subsequently develop a dynamic
optimization model in which strategies may depend on the age of
the habitat. Initially, the population colonizes an empty space. In
addition, if there is a predictable maximum age of the microhabitat
T (e.g., the inter-molt duration of a host crab), the choice of the
organisms should be affected by the time until T . We can solve
the optimal schedule by dynamic programming. First, we derive
three ordinary differential equations for the numbers of dwarf
males, immature individuals and large individuals, respectively.
We also derive three ordinary differential equations for future
reproductive success when the current state is specified. We
discuss the conditions for different situations, such as a population
dominated by hermaphrodites, a mixture of hermaphrodites and
dwarf males, and a mixture of females and dwarf males. We
examine the dependence of these patterns on the stability of the
microhabitats, the relative success of a dwarf male compared with
a large reproductive individual, and the growth rate of immature
individuals.
2. Model
We consider marine benthic species living in an ephemeral
microhabitat, including barnacles living on crabs or sea turtles
(e.g., Chelonibia patula: Crisp, 1983; Chelycypraea testudinaria:
Zardus and Hadfield, 2004; Octolasmis warwickii: Yusa et al.,
2010). Barnacle larvae are assumed to settle at a constant rate
and, following metamorphosis, they become sessile. Barnacles
may function as dwarf males immediately, or they may remain
immature and continue growing. When barnacles reach a large
size, they may engage in reproductive activity as hermaphrodites
or females. However, if the individual functions as a dwarf male, it
cannot grow and must remain a dwarf male throughout its life.
For mathematical simplicity, we assume that there are only two
size classes, small and large. We express the growth of an immature
individual as a transition to the large size following a Markovian
jump at a rate g. This growth rate g increases with food availability
in the environment. The mean time for an immature individual to
remain in the small size class is 1/g, which is shorter in a more
productive environment.
The microhabitat for the species (e.g., the shell of a crab)
becomes available at time 0 and disappears at a fixed time T , which
is the maximum longevity of the microhabitat. In addition, the
microhabitat is lost randomly at rate µ.
Mating occurs within the microhabitat. A large individual
chooses its sex allocation to maximize its own reproductive
success. The individual can be a simultaneous hermaphrodite
engaging in both male and female reproductive activities. The
relative fractions of the reproductive effort to male and female
functions are m and 1 − m, which satisfy 0 ≤ m ≤ 1. Here,
m = 0 implies that large individuals are pure females. The sex
allocation should be chosen by the large individual to maximize
the lifetime reproductive success, which follows a sex allocation
game considering the presence of dwarf males.
If a large individual is a pure male, it can achieve a larger male
reproductive success than a dwarf male. We denote the ratio of the
reproductive achievement of a dwarf male to that of a large, fully
active male by α , which is a constant smaller than 1.
2.1. Population dynamics
Let D(t ), U (t ), and H (t ) be the numbers of dwarf males,
immature small individuals, and large individuals, respectively.
These variables are functions of time t, which is the age of the
microhabitat. During the period in which the microhabitat (or
the shell of the crab) is maintained, these variables satisfy the
following differential equations.
dD
= s(1 − c (t )) − uD(t ), (1a)
dt
dU
= sc (t ) − (u + g )U (t ), (1b)
dt
dH
= gU (t ) − uH (t ). (1c)
dt
Here, s is the rate of larval settlement, and c (t ) is the fraction
of newly settled larvae that remain immature to grow. The value
of c (t ) should satisfy 0 ≤ c ≤ 1. 1 − c (t ) is the fraction of
newly settled individuals that immediately become dwarf males.
In general, c (t ) can change with the age of the microhabitat
t. Each individual should choose to maximize its own lifetime
reproductive success. u is the mortality, and we assume that it is
the same between different individuals. g is the growth rate of
immature individuals.
3. ESS for time-independent strategies in a stationary popula-
tion
We consider the simplest case in which the numbers of large
individuals, immature small individuals and dwarf males are
stationary values and the strategies are time-independent. This
situation occurs when T is very large and the effects of the initial
state and final time are weak. The ESS values of c and m are
denoted by c ∗ and m∗ . The equilibrium condition of the population
dynamics equations (1) leads to
s s sg
D̂ = (1 − c ∗ ), Û = c∗, and Ĥ = c ∗ . (2)
u g +u u(g + u)
Among the newly settled larvae, fraction c would remain
immature, whereas fraction 1 − c would become dwarf males.
We now consider the optimal choice of the fraction of newly
settled individuals. Let VD be the expected lifetime reproductive
success for a dwarf male and VU be the success for an immature
small individual. In the ESS, newly settled individuals should
 S. Yamaguchi et al. / Theoretical Population Biology 85 (2013) 49–57 51

choose the option that provides a larger lifetime reproductive

success. Therefore, we have the following relationship:

c ∗ = 1, if VD < VU , (3a)
c ∗ = 0, if VD > VU , (3b)
0 < c ∗ < 1, only if VD = VU . (3c)
The expected future reproductive success of a dwarf male is:
∞
α

VD = e−(u+µ)t λ(1 − m∗ )Ĥ dt , (4)
0 α D̂ + m∗ Ĥ
where λ is the number of eggs laid by a female-acting individual
per unit time, and λ(1 − m∗ )Ĥ is the total number of eggs produced
per unit time in the population. This value is multiplied by the
relative contribution of a single dwarf male among the total males
functioning in the entire population, which includes D̂ dwarf males
and Ĥ large individuals. α is the relative reproductive success
of a dwarf male compared to a large individual. Eq. (4) is the
integral of the expected reproductive success of a single dwarf
male multiplied by e−(u+µ)t , which is the probability that the Fig. 1. The evolutionarily stable solution of the stationary model. The horizontal
dwarf male will survive until age t. Note that reproduction will axis is the growth rate of immature small individuals g. When the growth rate is
high (g > (u + µ)α/(1 − α)), the ESS population consists only of hermaphrodites,
be terminated either by the death of the individual or loss of the
and they show equal sex allocation. When the growth rate is low (g < (u +
microhabitat. Eq. (4) becomes µ)α/(1 − α)), the ESS population is a mixture of dwarf males and pure females.
Many variables change abruptly at the boundary between these two cases. There
αλ (1 − m∗ ) Ĥ is no case in which a mixture of hermaphrodites and dwarf males is the ESS. The
VD = • . (5a) parameters are α = 0.5, u = 0.3, µ = 0.1, s = 300.
u + µ α D̂ + m∗ Ĥ
In a similar manner, we can derive the expected future reproduc- When there is no dwarf male (D̂ = 0), the ESS sex allocation is
tive success of an immature small individual as follows: to invest equally in male and female functions. When there are
 ∞
g some dwarf males, large individuals are hermaphrodites but their
VU = e−(u+µ)t e−gt gdt • VH = • VH , (5b) sex allocation is biased toward the female function (m∗ < 0.5), as
0 g +u+µ shown in Eq. (6a). When the number of dwarf males increases, the
where VH is the expected reproductive success of a large individual. female bias becomes stronger, and all the large individuals finally
Eq. (5b) is derived from the assumption that growth occurs in become pure females, as shown in Eq. (6b).
Markovian jumps with a rate g, considering the survivorship of The maximization symbol in Eq. (5c) indicates that the focal
the small individual. Note that g /(g + u + µ) is the probability individual makes the optimal choice of sex allocation under the
for a newly settled, immature small individual to become a large constraint 0 ≤ m ≤ 1. Because the optimal value of m is equal
individual before being killed. Eq. (5b) states that VU is the product to m∗ , we can derive the maximum symbol in Eq. (5c) as equal to
of this probability and the future reproductive success of a large λ. Therefore, we have VH = λ/(u + µ).
individual VH , which is calculated as follows: Combining these equations, we can derive the ESS. Among
 the three cases in Eq. (3), the case with Eq. (3b) does not lead
 ∞
to a feasible solution because all the newly settled individuals
VH = e−(u+µ)t
max λ(1 − m) + λ(1 − m∗ )Ĥ become dwarf males and none will become a large individual. This
0 0≤m≤1
case would result in an all-male population, in which the male
reproductive success is zero because of the absence of individuals

m
× dt . (5c) functioning as females. Therefore, in the ESS, either Eq. (3a) or
α D̂ + m∗ Ĥ Eq. (3c) holds. We can then derive the following solution (Fig. 1):
[Case 1] When α < g +u+µ holds, we have an ESS with c ∗ = 1. All
g
The first term within brackets is the female reproductive success
of a large individual, which produces λ(1 − m) eggs per unit small individuals remain immature, and there are no dwarf males.
time. The second term is the male’s reproductive success, i.e., the The population composition is
expected number of eggs sired by the focal individual whose share s sg
of the male reproduction is m/(α D̂ + m∗ Ĥ ). The total amount of D̂ = 0, Û = and Ĥ = . (7)
g +u u( g + u)
sperm should be calculated exactly as (H − 1)m∗ + m because
the sperm production of the focal individual also contributes to the The large individuals are all hermaphrodites with the sex allocation
total amount. Here, we assume that the mating group is sufficiently m∗ = 1/2 (equal investment to male and female functions), and
large and the contribution of the focal individual is negligible when the inequality VD < VU in Eq. (3a) holds.
[Case 2] When α > g +u+µ holds, we have an ESS in which
g
summing the amount of sperm. Under this assumption, (H −
1)m∗ + m can be approximated by Hm∗ . This assumption indicates 0 < c ∗ < 1 holds. Here, the dwarf males coexist with the
that local mate competition does not occur in this model. large individuals. Being a dwarf male is more advantageous than
The optimal value of m that attains the maximum in Eq. (5c) is growing fast and becoming a large functional male. Therefore,
equal to m∗ . From this value, we can derive the following condition: m∗ = 0 holds, which implies that large individuals are pure
females. From the equation VD = VU in Eq. (3c), the following
1 α D̂ equality holds:
m∗ = − if α D̂ < Ĥ , (6a)
2 2Ĥ
Ĥ g
m∗ = 0 if α D̂ > Ĥ . (6b) = . (8)
D̂ g +u+µ
52 S. Yamaguchi et al. / Theoretical Population Biology 85 (2013) 49–57

From Eq. (2), we have the fraction of newly settled individuals that fast growth leads to the identical relationship as in Eq. (3). Here, the
remain immature only difference is future reproductive success, VD (t ) and VU (t ), and
g +u the strategy c ∗ (t ) may depend on the age of the microhabitat t.
c∗ = . (9) We can derive the differential equation for future reproductive
2g + 2u + µ successes VD (t ), VU (t ), and VH (t ) by considering the events that
The population composition is: occur during a short time interval 1t. For a dwarf male, we have
the following formula:
s(g + u + µ) s
D̂ = , Û = , and
VD (t ) = RsD (t )1t + (1 − u1t − µ1t )VD (t + 1t ),
u(2g + 2u + µ) 2g + 2u + µ (11a)
(10)
Ĥ =
sg
. where RsD (t ) is the reproductive success and is given by
u(2g + 2u + µ) α
RsD (t ) = λ 1 − m∗ (t ) H (t ) • .
 
(11b)
The inequality distinguishing these two cases compares α , the α D(t ) + m∗ (t )H (t )
relative reproductive contribution of a dwarf male to a large
Eq. (11b) is the total number of eggs produced by large individuals,
individual, and g /(g + u + µ), which is the probability that a
λ(1 − m∗ (t ))H (t ), multiplied by the relative share of a single dwarf
small immature individual will reach a large size before its death
male in the total male reproductive success. In a similar manner,
or the loss of the microhabitat. If the former is larger than the
we have the recursive formula for an immature individual:
latter, being a dwarf male is more advantageous than being a
large male, and the male function is only performed by dwarf VU (t ) = g (t )1tVH (t + 1t ) + (1 − (u1t
males in the ESS. Instead, if the latter is greater than the former, + µ1t + g (t )1t ))VU (t + 1t ), (11c)
functioning as a dwarf male is less advantageous than growing into
a large individual. In this case, no dwarf male exists in the ESS, and and the one for a large individual (which is either a hermaphrodite
the male function is performed by hermaphrodites. Fig. 1 shows or a female):
that as the parameters smoothly change, the ESS solution changes VH (t ) = max RsH (m, t )1t + (1 − u1t − µ1t )
discontinuously between these two cases at the boundary. 0≤m≤1

In addition, there is no case in which the ESS population × VH (t + 1t ), (11d)

contains a mixture of dwarf males and hermaphrodites (Fig. 1). This
where the reproductive success of a large individual is:
conclusion is consistent with the argument that the coexistence of
dwarf males and hermaphrodites is difficult to explain (reviewed RsH (m, t ) = λ(1 − m) + λ(1 − m∗ (t ))
in Yamaguchi et al., 2012). However, many barnacle species have m
both hermaphrodites and dwarf males (Buhl-Mortensen and Høeg, × H (t ) • . (11e)
α D(t ) + m∗ (t )H (t )
2006; Yusa et al., 2010), which cannot be explained by this simple
model. By calculating the limit when 1t is very short, we have the
following differential equations:
4. Dynamic optimization
dVD λf ∗ (t )H (t )
− =α• − (u + µ)VD (t ), (12a)
One possible explanation for the observed mixture of dwarf dt α D(t ) + m∗ (t )H (t )
males and hermaphrodites in some barnacles may be the non- dVU
stationary nature of the microhabitat. In the following, we consider − = g (VH (t ) − VU (t )) − (u + µ)VU (t ), (12b)
dt
the case in which the processes in the model depend on t, the age  
of the microhabitat. The number of different individuals follows dVH f ∗ (t )H (t )
Eq. (1). At time t = 0, the microhabitat becomes available, and − = λ max f + m •
dt m+f =1 α D(t ) + m∗ (t )H (t )
no individual exists (D(0) = U (0) = H (0) = 0). In addition,
there is the final time t = T , at which the microhabitat is
− (u + µ)VH (t ). (12c)
lost, and the random loss of the microhabitat occurs at a rate By definition, we also have the terminal conditions: VD (T ) =
µ. Close to the end of the microhabitat T , the optimal decision VU (T ) = VH (T ) = 0.
of the individuals is different from a decision much earlier than The procedure to determine the ESS schedule c ∗ (t ) and m∗ (t )
T because the shortage of time remaining for growth may favor (for 0 < t < T ) are explained in detail in Appendix.
immediate reproduction as a dwarf male instead of the growth
option accompanied by a time delay. Therefore, in this setting, 5. ESS life history and sex allocation
the strategies of the individuals should depend on the age of the
microhabitat (c ∗ (t ), m∗ (t ), for 0 ≤ t ≤ T ) To derive the optimal We now discuss the ESS schedule of growth and sex allocation,
solution depending on t, we adopt a dynamic programming in which the strategies (c ∗ (t ) and m∗ (t )) may depend on the age of
calculation. the microhabitat t.

4.1. Recursive formula for future reproductive success 5.1. Optimal schedule of sex allocation

In a typical step of dynamic programming calculations, we
consider future reproductive success provided that the individual
will make the best choice in the future until the end of its life.
Future reproductive success is also constrained by the individual’s
current size and time. Let VD (t ), VU (t ), and VH (t ) be the expected
future reproductive successes for a dwarf male, immature small
individual, and large individual, respectively. All of these values
depend on time t.
The optimal choice for newly settled individuals between
functioning as a dwarf male or remaining immature and aiming for
When the strategies depend on t, the ESS behavior of the
population is affected by the initial and terminal conditions.
Examples of the ESS solutions for three different growth rates are
shown in Fig. 2. There are three components for each growth rate:
the top component shows the strategies c ∗ (t ) and m∗ (t ) for 0 <
t < T ; the middle component is the population dynamics; and the
bottom component is the future reproductive success.
Let us illustrate the solution by focusing on the three compo-
nents in the middle column for g = 0.5. The top component
indicates that c ∗ (t ) = 1 holds in the initial phase, which implies
 S. Yamaguchi et al. / Theoretical Population Biology 85 (2013) 49–57 53

Fig. 2. The ESS schedule of growth and sex allocation. The horizontal axis is time, t, defined as the age of the microhabitat. The left, middle, and right columns correspond
to the results of three different growth rates of immature individuals, g = 0.3, g = 0.5, and g = 0.8. The top section indicates the ESS strategies (c ∗ (t ), m∗ (t )). The middle
section is the population dynamics (D(t ), U (t ), H (t )), and the bottom is for the future reproductive success of a dwarf male and an immature individual (VD (t ), VU (t )). The
parameters are α = 0.5, u = 0.3, µ = 0.1, s = 300, T = 3650.

that all of the small individuals choose to grow and no dwarf males
are formed. Thus, the large individuals are hermaphrodites with
equal sex allocation (m = 0.5). The middle component illustrates
the population dynamics. Note that no dwarf male is produced in
the initial phase. The bottom component illustrates VD (t ) < VU (t )
in this initial phase, which implies that being a dwarf male is less
advantageous than being an immature fast-growing individual.
Toward the end of the period (near T ), c ∗ (t ) = 0 holds, which
implies that all the newly settled individuals should choose to
become dwarf males. This choice is because the advantage of the
option with fast growth is jeopardized by the shortage of time until
the end of the season. The bottom component illustrates VD (t ) >
VU (t ), which implies that being a dwarf male is more advantageous
than being an immature individual. From the top component, we
can observe that large individuals should be pure females (m = 0).
Thus, the population is a mixture of dwarf males and pure females.
In between these two phases, there is a period in which
hermaphrodites coexist with dwarf males. The top component
indicates that 0 < c ∗ (t ) < 1 and c ∗ (t ) declines with time, which
implies that some intermediate fraction of the newly settled larvae
become dwarf males, whereas others should remain immature.
Large individuals should allocate some resources to the male
function (m > 0), which implies that they are hermaphroditic.
The three components on the left side of Fig. 2 are for the ESS
solution with a slower growth rate (g = 0.3). The initial period
of only hermaphrodites is short, c ∗ (t ) begins to decline and the
dwarf males begin appearing earlier than in the case with g = 0.5
shown in the middle column. For g = 0.3, c ∗ (t ) declines with time
and becomes zero. The sex allocation of large individuals, m∗ (t ),
declines and becomes zero in g = 0.3 earlier than when g = 0.5.
After this time, the population is composed of pure females and
dwarf males.
For g = 0.3, there is a relatively long period of time in which
c ∗ (t ) is an intermediate value (0 < c ∗ (t ) < 1) and VD (t ) = VU (t )
holds, as shown in the bottom component. This case makes a sharp
contrast with the case of g = 0.5, where c ∗ (t ) switches from 1
to 0. Thus, a slower growth rate of small individuals results in an
optimal schedule with a longer last phase, in which the population
is a mixture of dwarf males and pure females.
The three components on the right side of Fig. 2 are for the op-
posite situation, in which the growth of small individuals is faster
54 S. Yamaguchi et al. / Theoretical Population Biology 85 (2013) 49–57

Fig. 3. The ESS schedule of growth and sex allocation. The horizontal axis is time t. See Fig. 2 for an explanation. The left and right columns are the results for different rates
of microhabitat loss: µ = 0.05 and µ = 0.3, respectively. Dwarf males appear earlier and pure females dominate earlier as the microhabitat becomes less stable (larger µ).
The other parameters are identical as in the standard case of Fig. 2 (g = 0.5).

(g = 0.8). Here, the initial phase of c ∗ (t ) = 1 and m∗ (t ) = 0.5
lasts longer, and the population only consists of hermaphrodites
with equal sex allocation.
The overall picture remains well explained by the criteria for
stationary solutions that separate the two situations. When the
growth of immature individuals is fast compared to the relative
success of the dwarf males, α < g /(g + u + µ), the population is
dominated by hermaphrodites most of the time. Dwarf males are
produced only near the end of the period (see Fig. 2 for g = 0.8).
By contrast, when the growth of immature individuals is slow, α >
g /(g + u + µ), the dwarf males begin appearing earlier (see Fig. 2
for g = 0.3). However, the ESS time-dependent solution is affected
by the effect of an initial empty population at t = 0, and the
expectation of the final time t = T . Consequently, the population
changes the composition with respect to sex expression. The
first phase is dominated by hermaphrodites without dwarf males
followed by a second phase containing a mixture of dwarf males
and hermaphrodites that are finally replaced by a mixture of dwarf
males and pure females. Different parameters, such as growth rate,
the relative success of dwarf males, and mortality, should affect the
importance of these three phases.
5.2. Dependence on the stability of the microhabitat
Fig. 3 illustrates the effect of different rates of microhabitat
loss. The three panels on the left side of Fig. 3 are for a stable
microhabitat with µ = 0.05, and the three panels on the right
side of Fig. 3 are for an unstable microhabitat with µ = 0.3. In
the unstable microhabitat, dwarf males appear earlier and pure
females are more likely to appear than in the stable microhabitat
if all other parameters are identical.
5.3. Dates for switching population compositions
In the initial period, there are no dwarf males. Toward the end
of the period of the microhabitat, all the new settlers become
dwarf males. Therefore, there is a date on which the dwarf males
begin appearing, and we denote this by t1 . With the increasing
 S. Yamaguchi et al. / Theoretical Population Biology 85 (2013) 49–57
a In the stationary solution studied in the earlier sections, both t1
b dwarf male immediately or remain immature and grow, and large
Fig. 4. Dates when the composition of individuals with different sex expressions
changes. (a) t1 , the first day on which dwarf males appear. (b) t2 , the first day on
which large individuals become pure females. The horizontal axis is the relative
advantage of the dwarf males α(g + u + µ)/g. Before t1 , no dwarf males exist,
and large individuals are hermaphrodites with equal sex allocation. Between t1
and t2 , the population is a mixture of dwarf males and hermaphrodites. After t2 ,
the population is a mixture of dwarf males and pure females. In the stationary
solution, the ESS should change discontinuously at α(g + u + µ)/g = 1. In the
time-dependent solution, dwarf males and hermaphrodites coexist around when
α(g + u + µ)/g = 1. The parameters are α = 0.1 − 0.5, 1α = 0.1, g =
0.1 − 0.9, 1g = 0.2, u = 0.1 − 0.5, 1u = 0.1, µ = 0.0 − 0.4, 1µ = 0.1.
number of dwarf males in the population, the ESS sex allocation
of the large individuals m∗ (t ) decreases, which indicates that their
sex allocation becomes female-biased. Finally, m∗ (t ) may become
equal to zero, and the initial date on which this occurs is denoted
by t2 . These critical dates mark the change in the population
composition; before t1 , no dwarf male exists, between t1 and t2 , the
population is a mixture of dwarf males and hermaphrodites, and
after t2 , the population is a mixture of dwarf males and females.
Fig. 4 illustrates these critical dates. The horizontal axis is the
relative advantage of the dwarf males, α(g + u + µ)/g.
Fig. 4(a) shows the initial date of the dwarf males’ appearance,
t1 . If the relative advantage of the dwarf males is low, they appear
only toward the end of the period (when t1 is near T ). By contrast,
if the relative advantage of the dwarf males is larger, they appear
earlier, even on the first day (t = 0).
Fig. 4(b) shows the initial day that pure females appear, t2 . This
switching date also shows a similar dependence on the relative
advantage of dwarf males but sharply changes near the boundary.
55
and t2 are equal to the final time T if the relative advantage of dwarf
males is smaller than 1; however, both t1 and t2 are equal to zero
(the initial time) if the relative advantage is larger than 1. There-
fore, the difference between the time-dependent solution in Fig. 4
and the stationary solution occurs near the boundary between the
two cases, in which the relative advantage of the dwarf males is
nearly equal to 1. A period can then be observed when the dwarf
males coexist with hermaphrodites (between t1 and t2 ).
6. Discussion
Dwarf males are prevalent among marine invertebrates (Voll-
rath, 1998) and coexist with large individuals that are females or
hermaphrodites (Vollrath, 1998; Weeks et al., 2006; Yusa et al.,
2012). In the present paper, we study the evolutionarily stable
schedule of growth and sex allocation for marine benthic species
that may contain dwarf males. Small individuals can function as a
individuals allocate reproductive resources between male and fe-
male functions.
The overall behavior of the model can be understood by the
following inequality:
g
α> , (13)
g +u+µ
where α is the relative reproductive success of a dwarf male
compared to a large individual, g is the growth rate of an immature
individual, u is mortality and µ is the loss of the microhabitat. If
this inequality holds, namely, if a dwarf male is more advantageous
than being a large individual, the population is composed of dwarf
males and females. If the opposite inequality holds, the population
consists of large individuals who are hermaphrodites, and no
dwarf males are produced. The two phases are not connected
smoothly. When the parameters change gradually, many variables
change abruptly at the boundary between the two phases (Fig. 1).
Inequality (13) implies that the relative reproductive success of a
dwarf male is greater than the survivorship of a small immature
individual until it achieves a large size. In this stationary model,
there is no case for a mixture of hermaphrodites and dwarf males
to be the ESS.
We then studied the case in which strategies may depend on the
age of the microhabitat (c (t ) and m(t )). Here, c (t ) is the fraction
of newly settled larvae that choose to remain immature to grow.
m(t ) is the male allocation of large individuals. The population col-
onizes an empty space at t = 0, and the microhabitat will be lost
at time t = T . The ESS can be solved by dynamic programming.
An abrupt change in the optimal schedule often appears in dy-
namic optimization problems. This general phenomenon is closely
related to the ‘‘bang–bang control’’ of reproductive resource alloca-
tion, which shows a sudden shift from vegetative growth to repro-
ductive growth, in the optimal life history (Cohen, 1971; Iwasa and
Roughgarden, 1984; Iwasa and Cohen, 1989; Perrin, 1992). Again,
the overall behavior of the model can be understood in terms of
the relative magnitude of both sides of the inequality (13), al-
though the pattern is affected by the initial and final conditions
(see Fig. 2). Notably, the ESS population may include a mixture of
hermaphrodites and dwarf males.
Because the model only includes three states, dwarf males,
immature small individuals, and large individuals, we have three
ordinary differential equations for the future reproductive success
given that the current state is specified. This method allows a
mathematical analysis of the optimal schedule (see Appendix).
The studies by Yamaguchi et al. (2008, 2012, 2013) on the evo-
lution of dwarf males adopt a numerical simulation of compu-
tational techniques of dynamic optimization. These models are
56 S. Yamaguchi et al. / Theoretical Population Biology 85 (2013) 49–57
flexible and can incorporate many factors that might affect sex al-
location and growth. However, the conclusion is rather qualitative.
In the present paper, we adopted a different approach; we consider
a simple model that can be fully analyzed mathematically.
Another novelty of our model is that we hypothesized that
individuals behave optimally under each given environment. In
other words, this model applies to species with environmentally
determined sex allocation or sex determination. In fact, many
animals with dwarf males are thought to have an environmentally
determined sex system (Vollrath, 1998; Rouse et al., 2004; Yusa
et al., 2010). This study suggests that individuals that settle on
transient microhabitats should become dwarf males and therefore
may explain why the proportion of dwarf males differs greatly
among microhabitats (e.g., Yusa et al., 2010).
Many theoretical studies on the condition for the evolution of
dwarf males in marine organisms were based on the sex alloca-
tion game (Charnov, 1987; Yamaguchi et al., 2008; reviewed in
Yamaguchi et al., 2012). Some studies incorporate an important
aspect of the mating system of barnacles, in which the large indi-
viduals extend their penis to contact mating partners. Combining
this result and the fact that mating only occurs between different
individuals, Charnov (1982, 1987) predicted female-biased sex al-
location when the number of candidates for mating partners is lim-
ited. This situation should produce an opportunity for dwarf males
to be advantageous (Charnov, 1982, 1987). Therefore, among bar-
nacles, dwarf males are expected to be more prevalent for species
with smaller mating group sizes, which are defined as the num-
ber of large (non-dwarf) individuals that can donate sperm to one
another (Charnov, 1980, 1982, 1987). This prediction is consis-
tent with a broad pattern of observed taxonomic distributions in
species with dwarf males (Yusa et al., 2012).
The model studied in the current paper includes a number
of assumptions that makes mathematical analysis feasible, and
these might be mitigated in future studies. Some directions for the
extension of the model are suggested below.
(1) Density dependence. In the current paper, the reproductive suc-
cess and other fitness components are assumed to be independent
of density. However, it is likely that the number of large individuals
H (t ) might affect many processes of small individuals, such as the
relative success of dwarf males α and growth rate of immature in-
dividuals g. α is most likely a decreasing function of H (t ) because it
is more difficult for dwarf males to share in fertilization in the pres-
ence of many large individuals. This result may explain why dwarf
males evolve in species with a smaller abundance of large individ-
uals (Yusa et al., 2012). However, the growth rate of immature indi-
viduals is also likely to be suppressed by the presence of many large
individuals, which makes g a decreasing function of H (t ), and re-
sults in a prediction in the opposite direction. However, if the high
productivity of the environment (e.g., high food availability) pro-
duces both the fast growth of immature individuals and high den-
sity of large individuals in the habitat, then the absence of dwarf
males will be expected in the area with a high density of large in-
dividuals. Further study is necessary to know how the high density
of large individuals may affect the condition for dwarf males to be
advantageous.
(2) Local mate competition. Barnacles mate by extending their
penises and inseminating other individuals within reach. There-
fore, we expect that the reproductive success of a male should
be limited; the male reproductive success cannot increase linearly
with the investment to male function. It is notable that dwarf males
appear in this model, even in the absence of local mate compe-
tition. Therefore, this model may help to understand the occur-
rence of dwarf males among animals without a penis (e.g., bivalves:
Turner and Yakovlev, 1983; annelids: Vollrath, 1998; Rouse et al.,
2004).
(3) Continuous size. We only considered two size classes (small and
large) and represented growth as a transition from a small size
class to a large size class. A more realistic model is to consider a
continuous trait of body size w , which increases with time. The
reproductive behavior and growth rate will then be provided as a
function of size w . A comparison between the cases with two size
classes and a model with many continuous size classes is a notable
theme of future study (see, e.g., Yamaguchi et al., 2013).
(4) Selfing. We assumed that the newly settled individuals all mate
with one another. In other words, we assumed outcrossing, which
is a good assumption for barnacles (Yusa et al., 2012). However,
Weeks et al. (2006) mentioned that hermaphrodites do not always
mate with one another and undergo selfing in all androdioecious
animals except for barnacles. We must therefore modify the model
to include the possibility of selfing and inbreeding depression,
which implies that the viability of the offspring due to selfing is
lower than that of offspring due to outcrossing.
(5) Growth of dwarf males. In this paper, the dwarf males are
assumed to not grow. However, dwarf males in some barnacle
species have the potential to grow (Annandale, 1905; Hiro,
1933) and become large hermaphrodites (Crisp, 1983; Zardus and
Hadfield, 2004; Yusa et al., 2010). The model may incorporate
the possibility for dwarf males to grow at a speed slower than
immature individuals.
In spite of its many limitations, an advantage of the simple
model developed in this paper is its tractability. The model demon-
strates how the choice of growth and sex allocation is coupled in
many marine sessile species.
Acknowledgments
This work was performed under the support of the Environment
Research and Technology Development Fund (S9) of the Ministry of
the Environment, Japan, a Grant-in-Aid for Scientific Research (B)
to Y.I., and another for JSPS Fellows to S.Y. We are very thankful
to the following people for their very useful comments: Eric L.
Charnov, Y. Tachiki, and K. Uchinomiya.
Appendix
How to obtain the ESS schedule.
The ESS schedule (c ∗ (t ), m∗ (t ), for 0 < t < T ) is one
that satisfies all of the following conditions: (1) the population
dynamics equations (1) and their initial conditions, (2) the optimal
choice condition equation (3) in which the variables depend on
time t, (3) the ordinary differential equations (12) and their final
condition for the future reproductive successes satisfying the three
states, and (4) the optimal choice of sex allocation m(t ) that
achieves maximization in Eq. (12c).
The ESS sex allocation is provided by Eq. (6), where m, D and
H are all dependent on t. Because of this condition, the first term
of Eq. (12c) is simply equal to λ after calculation of the optimal
choice of sex allocation. The differential equations (12b) and (12c)
can then be solved by backward integration using the terminal
condition. Thus, we obtain VU (t ) for 0 < t < T .
The calculation can be performed by searching for function
c ∗ (t ) for 0 < t < T by iteration. We initially established this
value as an initial condition, for example c (t ) = constant for
0 < t < T . Using this trial function c (t ), we can calculate the Eqs.
(1) population dynamics and obtain D(t ), U (t ), H (t ) for 0 < t < T .
We then calculate sex allocation m(t ) for 0 < t < T from Eq. (6).
By integrating Eqs. (12a) we can calculate the future reproductive
success of a dwarf male VD (t ) for 0 < t < T . The result of VD (t )
can be compared with the previously calculated VU (t ) to determine
whether they satisfy Eq. (3) at each t. In general, if Eqs. (3) are not
 S. Yamaguchi et al. / Theoretical Population Biology 85 (2013) 49–57
satisfied, we then change the trial function c (t ) using the following
method:
c next (t ) = c (t ) + w eight ∗ (VU (t ) − VD (t )),
where w eight is a positive constant controlling the speed of
convergence. Using the revised c (t ), we repeat the same procedure
because Eq. (A.1) makes c (t ) approach the relation provided by Eq.
(3). After some number of iterations, c (t ) no longer changes, and
Eqs. (3) sufficiently and accurately hold. We then regard this value
as the ESS solution.
Numerical procedure to obtain the ESS schedule.
Here, we explain the numerical scheme we used to calculate
the ESS schedule of reproduction and growth. We regard the basic
model as the discrete dynamics with the unit time step short (such
as one day).
Let z be the length of one step, which is, for example, 1/365
if one day is one step and the time unit of the model is one year.
τ is an integer close to T /z. We prepare vectors (arrays) of size
τ + 1 for D[i], U [i], H [i], c [i], m[i], VD [i], VU [i], VH [i], where i =
0, 1, 2, . . . , τ , which indicates D(iz ), U (iz ), H (iz ), c (iz ), m(iz ),
VD (iz ), VU (iz ), and VH (iz ), respectively.
(1) Calculate VU and VH .
Set VU [τ ] = VH [τ ] = 0. Subsequently, perform the following
calculation for i in decreasing order (backward calculation): i =
τ , τ − 1, τ − 2, . . . , 2, 1.
VU [i − 1] = (1 − uz − µz − gz )VU [i] + gzVH [i]
VH [i − 1] = (1 − uz − µz )VH [i] + λz .
(2) Set the initial values of c.
c [i] = c initial, for i = 0, 1, . . . , τ − 1, τ .
(3) Population dynamics
Set D[0] = U [0] = H [0] = 0.
Perform the following calculation for i in increasing order: i =
0, 1, 2, . . . , τ − 1.
D[i + 1] = D[i](1 − uz ) + sz (1 − c [i])
U [i + 1] = U [i](1 − uz − gz ) + sz ∗ c [i]
H [i + 1] = H [i](1 − uz ) + gz ∗ U [i].
(4) Sex allocation of large individuals
Perform the following calculation for i = 0, 1, 2, . . . , τ − 1, τ ,
1 Turner, R.D., Yakovlev, Y., 1983. Dwarf males in the Teredinidae (Bivalvia,
if α D[i] < H [i], then m[i] = (1 − α D[i]/H [i]); else m[i] = 0.
2
(5) Calculation of future reproduction of a dwarf male
Set VD [τ ] = 0.
Perform the following calculation for t in decreasing order
(backward calculation)
i = τ , τ − 1, . . . , 2, 1.
If D[i − 1] < 0.001, then success = α ;
Else if D[i − 1] < H [i − 1]/α, success = α
Else, success = H [i − 1]/D[i − 1]
VD [i − 1] = success∗ λz + (1 − uz − µz )VD [i].
(6) Recurrence formula for c
c [i] ← c [i] + w eight ∗ (VU [i] − VD [i]),
for i = 0, 1, . . . , τ − 1, τ .
(7) Go back to (3). Repeat until c [i] does not change any more.
57
References
Annandale, B.A., 1905. Malaysian barnacles in the Indian museum, with a list of the
(A.1) Indian pedunculata. Mem. Asiat. Soc. Bengal. 1, 73–84.
Buhl-Mortensen, L., Høeg, J.T., 2006. Reproduction and larval development in three
scalpellid barnacles, Scalpellum scalpellum (Linnaeus 1767), Ornatoscalpellum
stroemii (M. Scars 1859) and Arcoscalpellum michelottianum (Seguenza 1876),
Crustacea: Cirripedia: Thoracica: implication for reproduction and dispersal in
the deep sea. Mar. Biol. 149, 829–844.
Cash, C.E., Bauer, R.T., 1993. Adaptations of the branchial ectoparasite Probopyrus
pandalicola (Isopoda: Bopyridae) for survival and reproduction related to
ecdysis of the host, Palaemonetes pugio (Caridea: Palaemonidae). J. Crustac. Biol.
13, 111–124.
Charnov, E.L., 1980. Sex allocation and local mate competition in barnacles. Mar.
Biol. Lett. 1, 53–57.
Charnov, E.L., 1982. The Theory of Sex Allocation. Princeton University Press,
Princeton.
Charnov, E.L., 1987. Sexuality and hermaphroditism in barnacles: a natural selection
approach. In: Southward, A.J. (Ed.), Barnacle Biology. In: Crustacean Issues,
vol. 5. A. A. Belkema, Rotterdam, pp. 89–103.
Cohen, D., 1971. Maximizing final yield when growth is limited by time or by
limiting resource. J. Theoret. Biol. 33, 299–307.
Crisp, D.J., 1983. Chelonobia patula (Ranzani), a pointer to the evolution of the
complemental male. Mar. Biol. Lett. 4, 281–294.
Darwin, C., 1851. The Lepadidae. In: A Monograph on the Sub-Class Cirripedia,
vol. 1. The Ray Society, London.
Ghiselin, M.T., 1974. The Economy of Nature and the Evolution of Sex. University of
California Press, Berkeley.
Goto, R., Hamamura, Y., Kato, M., 2007. Obligate commensalism of Curvemysella
paula (Bivalvia: Galeommatidae) with hermit crabs. Mar. Biol. 151, 1615–1622.
Hiro, F., 1933. Report on the cirripedia collected by the surveying ships of the
imperial fisheries experimental station on the continental shelf bordering
Japan. Rec. Oceanogr. Works Jpn. 5, 11–84. 3 plates.
Høeg, J.T., 1995a. Sex and the single cirripede: a phylogenetic perspective.
In: Schram, F.R., Høeg, J.T. (Eds.), New Frontiers in Barnacle Evolution.
In: Crustacean Issues, vol. 10. A. A. Belkema, Rotterdam, pp. 195–207.
Høeg, J.T., 1995b. The biology and life cycle of the Rhizocephala (Cirripedia). J. Mar.
Biol. Assoc. U. K. 75, 517–550.
Itani, G., Kato, M., Shirayama, Y., 2002. Behaviour of the shrimp ectosym-
bionts,Peregrinamor ohshimai (Mollusca: Bivalvia) and Phyllodurus sp. (Crus-
tacea: Isopoda) through host ecdyses. J. Mar. Biol. Assoc. U. K. 82, 69–78.
Iwasa, Y., Cohen, D., 1989. Optimal growth schedule of a perennial plant. Am. Nat.
133, 480–505.
Iwasa, Y., Roughgarden, J., 1984. Shoot/root balance of plants: optimal growth of a
system, with many vegetative organs. Theor. Popul. Biol. 25, 78–105.
Klepal, W., 1987. A review of the comparative anatomy of the males in cirripedes.
Oceanogr. Mar. Biol. Ann. Rev. 25, 285–351.
Ó Foighil, D., 1985. Form, function, and origin of temporary dwarf males in
Pseudopythina rugifera (Carpenter, 1864) (Bivalvia: Galeommatacea). Veliger 27,
245–252.
Perrin, N., 1992. Optimal resource allocation and the marginal value of organs. Am.
Nat. 139, 1344–1369.
Pietsch, T.W., 2005. Dimorphism, parasitism, and sex revisited: modes of
reproduction among deep-sea ceratioid anglerfishes (Teleostei: Lophiiformes).
Ichthyol. Res. 52, 207–236.
Rouse, G.W., Goffredi, S.K., Vrijenhoek, R.C., 2004. Osedax: bone-eating marine
worms with dwarf males. Science 305, 668–671.
Pholadacea). Science 219, 1077–1078.
Vollrath, F., 1998. Dwarf males. Trends Ecol. Evol. 13, 159–163.
Weeks, S.C., Benvenuto, C., Reed, S.K., 2006. When males and hermaphrodites
coexist: a review of androdioecy in animals. Integr. Comput. Biol. 46, 449–464.
Yamaguchi, S., Charnov, E.L., Sawada, K., Yusa, Y., 2012. Sexual systems and
life history of barnacles: a theoretical perspective. Integr. Comput. Biol. 52,
356–365.
Yamaguchi, S., Yusa, Y., Sawada, K., Takahashi, S., 2013. Sexual systems and dwarf
males in barnacle: integrating life history and sex allocation theories. J. Theoret.
Biol. 320, 1–9.
Yamaguchi, S., Yusa, Y., Yamato, S., Urano, S., Takahashi, S., 2008. Mating group size
and evolutionarily stable pattern of sexuality in barnacles. J. Theoret. Biol. 253,
61–73.
Yusa, Y., Takemura, M., Miyazaki, K., Watanabe, T., Yamato, S., 2010. Dwarf males of
Octolasmis warwickii (Cirripedia: Thoracica): the first example of coexistence
of males and hermaphrodites in the suborder Lepadomorpha. Biol. Bull. 218,
259–265.
Yusa, Y., Yamato, S., Marumura, M., 2001. Ecology of a parasitic barnacle,
Koleolepas avis: relationship to the hosts, distribution, left–right asymmetry and
reproduction. J. Mar. Biol. Assoc. U. K. 81, 781–788.
Yusa, Y., Yoshikawa, M., Kitaura, J., Kawane, M., Ozaki, Y., Yamato, S., Høeg, J.T., 2012.
Adaptive evolution of sexual systems in pedunculate barnacles. Proc. R. Soc. B
279, 959–966.
Zardus, J.D., Hadfield, M.G., 2004. Larval development and complemental males in
Chelonibia testudinaria, a barnacle commensal with sea turtles. J. Crustac. Biol.
24, 409–421.