You are on page 1of 9

OIKOS 95: 476–484.

Copenhagen 2001

Unbeatable strategy, constraint and coevolution, or how to resolve


evolutionary conflicts: the case of the fig/wasp mutualism

Marie Charlotte Anstett

Anstett, M. C. 2001. Unbeatable strategy, constraint and coevolution, or how to


resolve evolutionary conflicts: the case of the fig/wasp mutualism. – Oikos 95:
476–484.

Constraints and evolution are central for the resolution of conflicts between mutual-
istic species and for the stability of mutualisms. However, proximal causes of the
benevolence of mutualistic species are often unknown. Monoecious fig species and
their specific pollinators are in conflict on the use of fig ovaries, which can either
produce one seed or host one pollinator larva. Here, I provide new data showing
that, probably because of space constraints during the development of both seeds and
wasps, ovaries vary in their quality as a substrate for pollinator development
depending on their location within the fig inflorescence. This constraint may be
responsible for the stability of the fig/wasp mutualism. Moreover, population density
of pollinators and density dependent selection on the pollinators could be sufficient
to explain the observed highly variable seed/wasp ratios produced by figs.

M. C. Anstett, CNRS-CEFE, groupe coé6olution, 1919 route de Mende, F-34293


Montpellier Cedex 5, France (anstett@cefe.cnrs-mop.fr).

In the gradient from antagonistic to mutualistic interac- mamura 1993, Yamamura, 1996), even when transmis-
tions (Bronstein 1994), mutualisms are characterized by sion remains partly horizontal (Lipsitch et al. 1996).
a reciprocal exploitation in which the net result is The importance of understanding this evolutionary
positive for both species (Thompson 1982). Many sym- trend towards benevolence is illustrated by the evolu-
biotic mutualisms evolved through the evolution of tion towards benevolence and mutualism of some
parasite benevolence (Margulis and Fester 1991, Black- strains of fungal pathogens considered for use in bio-
stone 1995). Various models have been developed to logical control (Freeman and Rodriguez 1993, Hall-
explain the evolution of mutualism (Frank 1997). Many mann and Sikora 1994).
are based on the fact that when parasites are vertically However, many mutualisms have a purely horizontal
transmitted their fitness may be correlated with that of transmission (Douglas 1994). For these interactions, the
their host (Toft and Karter 1990), even if this is not conditions necessary for the evolution of benevolence
always true (Kover and Clay 1998). When this correla- are less well known, especially when conditions for
tion exists, under certain conditions benevolence may interdemic selection (Dugatkin et al. 1992) or for the
evolve (Yamamura 1993, Frank 1997). Such an evolu- evolution of reciprocal altruism (Trivers 1971) do not
tionary scenario has been experimentally tested (Bull et apply. In these cases, benevolence may be due to a
al. 1991), and results consistent with its predictions constraint that limits the exploitation of the partner
have been observed in the field (Herre 1993). Evolution (e.g., a trade-off between virulence and transmissibility
towards vertical transmission, and then subsequently to [Lipsitch et al. 1995], or unrecognition in an op-
mutualism is also predicted under some conditions (Ya- timal foraging game [Mesterton-Gibbons 1991]). The

Accepted 11 June 2001


Copyright © OIKOS 2001
ISSN 0030-1299
Printed in Ireland – all rights reserved

476 OIKOS 95:3 (2001)


ultimate form of the constraint is the unbeatable strat- The conflict
egy (West and Herre 1994) where one species entirely
The genus Ficus is characterized by a peculiar inflores-
controls the other.
cence, the fig or syconium (Fig. 1), in which flowers are
Most empirical studies of the evolution of benevo-
enclosed. Figs of each species are pollinated only by a
lence focus on host parasite interactions. I adopt a
species-specific wasp that enters the syconium, polli-
different approach, using a mutualism as a model sys-
nates and lays eggs in some fig flowers. Some Ficus
tem and focusing on the conflicts between mutualists.
Studies of the resolution of these conflicts could give species are monoecious, others are (gyno)dioecious.
insights on the conditions and constraints necessary for These two reproductive systems impose different con-
the evolution of benevolence more readily than studies flicts on the mutualism (reviewed by Patel et al. 1993,
on host/parasite interactions, where benevolence does Anstett et al. 1998). Here monoecious species are
not exist. treated. Some traits of dioecious fig pollination systems
Actual causes of benevolence in a mutualism may not have implications for the hypothesis developed and will
be straightforward to determine. For example coevolu- be discussed. In monoecious species, each female flower
tion usually does not occur between two characters, but may either produce one seed or host one pollinator
between two sets of characters, rendering actual con- offspring. Male pollinators are apterous and die just
flicts between mutualists more difficult to distinguish. after they mate with females that developed in the same
Once an evolutionary conflict is identified, several alter- syconium. When female pollinators exit the gall where
native hypotheses may plausibly explain the stability of they developed, they become loaded with pollen (or
the mutualism. This is the case in the fig/wasp pollina- load themselves with pollen for species with active
tion mutualism. pollination) before leaving the natal syconium. Thus
Fig trees are pollinated solely by species-specific female pollinators developing in one syconium are the
wasps, and each wasp larva develops at the expense of only dispersers of the pollen produced by this syco-
one fig ovary (Janzen 1979, Bronstein 1988, Compton nium. Reproductive success of fig trees through male
1993). Thus fig trees and their associated fig wasps are function is (roughly) proportional to their ‘‘produc-
in conflict over predation of fig ovaries (see below for tion’’ of female pollinators.
more details). This conflict was first described more Thus fig trees and wasps are in conflict over the fate
than 20 years ago (Janzen 1979) and since then, eight of female flowers: the fig tree is selected to allocate half
different explanations of the persistence of this mutual- of its reproductive effort to the female function (esti-
ism have been published (reviewed by Kathuria 1999). mated by seed production) and half to the male func-
After a brief review of the three main hypotheses, I will tion (estimated by pollen + male and female wasps,
provide new data that suggest the existence of a trade- Charnov et al. 1976, Herre 1989). But pollinators do
off that could explain why fig pollinators do not occupy not have any individual advantage in not laying eggs in
all fig ovaries and allow some seeds to develop. all female flowers. Why do pollinators not occupy all
fig female flowers?
Three hypotheses are currently debated to explain the
seed/wasp production ratio: the unbeatable-seeds hy-
pothesis (West and Herre 1994), the ovipositor length
optimization hypothesis (Ganeshaiah et al. 1995), and
the limited-egg-supply hypothesis (Nefdt and Compton
1996). All of these are supported by data. Five other
hypotheses have been convincingly rejected on grounds
explained by Bronstein (1992) and Kathuria (1999).
In the ‘‘unbeatable-seeds’’ hypothesis (West and
Herre 1994) ‘‘a certain proportion of flowers cannot be
used by any pollinator or parasite wasps, which results
in unbeatable seeds’’. This hypothesis was based on the
fact that some parasitic wasp species that also develop
at the expense of fig ovaries (but do not pollinate)
usually compete with the pollinators and not with the
seeds (West and Herre 1994), even though some para-
sitic species do compete with seeds (Kerdelhué et al.
Fig. 1. Schematic drawing of the interior of the fig just after 2000). The ovule dimorphism of Ficus asperifolia, a
pollination, showing the irregular organization of the ovaries. (gyno)dioecious species, was wrongly used to support
Ovaries of the outer layer belong to long-styled, short-pedi- the unbeatable-seeds hypothesis in monoecious species.
celled flowers. They usually develop into seeds. Ovaries of the
inner layer belong to short-styled, long-pedicelled flowers. In (gyno)dioecious Ficus, some trees (females) never
They usually host a pollinator larva. host pollinator larvae and only produce seeds, whereas

OIKOS 95:3 (2001) 477


‘‘male’’ trees never produce seeds (they have non-func- data from 12 crops at the same date [Bronstein and
tional ovaries) and only host pollinators (Kjellberg et Hossaert-McKey 1996]), showing that egg supply is not
al. 1987, Patel and McKey 1998). Ovaries of the female always limiting.
trees have an integument circumvallating the nucellus, Within a fig, there are several irregular layers of
which is the common situation of anatropous ovules ovaries. Because style and pedicel are both approxi-
(Verkerke 1987a, b). The integument may mechanically mately straight, with all stigmata being at the same
protect the ovaries against predation by pollinator lar- height during fig receptivity, lengths of style and pedicel
vae. Ovaries of the ‘‘male’’ trees have a thicker raphe of each flower are highly negatively correlated. Thus
and shorter integument, two characteristics that are female flowers of the outer layers (near the fig wall)
very likely to facilitate oviposition and development of have a long style and a short pedicel and female flowers
pollinator larvae by providing more food for the larva of the inner layers (near the fig lumen) have a short
just after hatching and by facilitating movement of the style and a long pedicel (Fig. 1). Because ovary organi-
larvae to the endosperm (Verkerke 1987a). However, zation is irregular, style length distribution is unimodal.
no such dimorphism exists among the ovaries of a Seeds and pollinators can develop in any layer of
monoecious fig (Verkerke 1987a). All ovaries of monoe- ovaries (Kerdelhué and Rasplus 1996, Nefdt and
cious figs have the same structure as those of ‘‘male’’ Compton 1996). However, ovaries of the outer layer
figs of dioecious species. They thus have the same (near the fig wall) are more likely to develop into seeds,
adaptations for pollinator development, and lack the whereas ovaries of the inner layer are more likely to
ovary protection of a continuous integument present in host a wasp larva (pollinator or parasite). This is the
flowers of female figs. In the absence of description of only point that was observed by all the authors that
any actual biological traits characterizing these puta- proposed the above hypotheses (West and Herre 1994,
tively ‘‘unbeatable flowers’’ in any monoecious fig, this Ganeshaiah et al. 1995, Nefdt and Compton 1996) and
hypothesis is not testable or refutable. others (Kerdelhué and Rasplus 1996).
When the pollinator lays eggs, it inserts its ovipositor This spatial distribution is thought to be the result of
through the style, and precisely deposits each egg next a choice by the pollinator during oviposition that could
to the nucellar epidermis (Verkerke 1987a). Thus the be based on style length and shape. This choice could
relative distributions of style length and ovipositor also be due to differences in the quality of flowers
length give the percentage of flowers accessible for egg within each fig. Understanding the causes of this choice
laying. Comparisons of ovipositor length and style may help us to examine the three existing plausible
length show that in monoecious species female wasps hypotheses for why pollinators do not lay eggs
have access to only 50% to 90% of female flowers everywhere.
(Nefdt and Compton 1996, Kathuria 1999). The o6ipos- Here I address the following questions: Does the fate
itor-length-optimization hypothesis (Ganeshaiah et al. of an ovary depend on its position (i.e. ovary layer, as
1995, Kathuria et al. 1995) supposes a cost of a longer defined by pedicel length, see below) or on female
ovipositor, making it too costly to evolve an ovipositor flower style length? Does duration of pollinator devel-
capable of reaching all female ovaries. In this hypothe- opment depend on the position of its gall? Does gall
sis, the number of wasp eggs is never limiting, which is seed size vary with the position of the ovary? In light of
clearly an oversimplification (Nefdt and Compton these new results, I will discuss the reasons why pollina-
1996). Moreover, this hypothesis does not explain the tors are benevolent to their host tree and whether this
highly variable seed/wasp ratios observed (from 0 to benevolence is due to an evolutionarily stable strategy,
100% of seeds within a crop, Bronstein and Hossaert- an evolutionary constraint or an unbeatable strategy of
McKey 1996), and a cost of having a longer ovipositor one of the partners.
has never been demonstrated.
The limited-egg-supply hypothesis (Nefdt and Comp-
ton 1996) suggests that on average, there are not
enough eggs deposited in one syconium to occupy all
Materials and methods
female flowers. This is clearly true in some cases
(Anstett et al. 1996, Nefdt and Compton 1996), but the Ficus microcarpa (Linn. f.) is native to Asia and intro-
hypothesis cannot explain why some seeds are still duced as an ornamental tree in many countries (McKey
produced when the number of eggs is clearly saturating 1989). Its pollinator, Eupristina 6erticillata (Waterston)
(Anstett et al. 1996, Bronstein and Hossaert-McKey has also established in many areas of introduction
1996). Moreover, the number of foundresses, and thus (McKey 1989). On three different trees planted in
the egg supply, greatly varies within individual figs Sevilla (Spain), I collected figs just before wasp emer-
(from 1 to 24 in F. aurea [Bronstein and Hossaert- gence, then placed them in glass vials. The number of
McKey 1996] while five foundresses are clearly saturat- foundresses per fig is unknown since they can re-exit
ing [Anstett et al. 1996]), and among crops (mean the fig they visit (pers. obs.), as in many species (Giber-
number of foundresses per fig ranges from 1.4 to 7.6, nau et al. 1996). When male wasps started to emerge,

478 OIKOS 95:3 (2001)


Table 1. Results of a multiple regression explaining pedicel length. This model explains 67.4% of the total variance.

df Type III SS F P

Style 1 77.5 72.3 B1×10−4


Ovary length 1 16.2 15.1 B1×10−4
Tree 2 10.2 5.12 B9×10−3
Fig(tree) 13 40.7 2.9 B5×10−4
Style×content 1 45.6 42.6 B1×10−4
Ovary length×content 1 109.6 102.2 B1×10−4
Style×fig (tree) 15 35.5 2.21 B6×10−3
Error 347 371

each fig was immediately cut open and put into 70% Results
ethanol, with all the wasps. I thus stopped emergence of
wasps at a point when some females had already exited Does the content of an ovary depend on its
their galls whereas others were still inside. For five figs position (i.e. ovary layer or pedicel length) or on
per tree, I measured style length, pedicel length, seed or style length?
gall length and width for at least 30 flowers per fig.
There was an overall correlation between style and
Since styles and pedicels are approximately straight
pedicel length (r= − 0.643, p B10 − 4, n=398). A mul-
(especially for pedicels), and since they stay intact dur-
tiple regression analysis (Table 1) shows that pedicel
ing fig development, measurement is quite easy in spite
length depended on style length and ovary (i.e. seed or
of their small size. During fig receptivity all stigmata
gall) length. Pedicel length also depended on tree, fig
are at the same height and form a stigmatic platform
(tree) and an interaction term style × fig (tree). It also
for the pollinators. However, after pollination the style
depended on the interactions ovary length × content
no longer has a function. When wasps emerge, stigmata
and style × content (Table 1), showing that the regres-
may be at any height and do not form a platform
sion line was different for flowers containing a seed and
anymore. On the other hand, pedicels are likely to
flowers containing a gall. Pedicel length decreased more
ensure circulation of nutrients during the whole devel- with ovary length for seeds than for galls. This may be
opment of the fig. When wasp emerge, pedicels are still due to very different size (see below) and shape of galls
straight and perpendicular to the fig wall and thus and seeds, seeds being longer and larger. Fig. 2 shows
provide an accurate measurement of the distance be- the variability of the regression between style and
tween the ovary and the fig wall, i.e. of ovary position. pedicel length for each fig and each content (seed or
I also recorded content of each flower: seed, female wasp). Besides the overall correlation between pedicel
pollinator inside a closed gall, female pollinator inside a and style length, the high number of variables signifi-
gall with a hole, empty gall. No parasites were present cantly explaining pedicel length indicates the intricate
in these figs. Within a fig, male pollinators exit their developmental factors responsible for the final pedicel
galls first, then mate females by piercing a small hole in length.
the gall of each female wasp and inserting the abdo-
men. Thus a female pollinator inside a gall with a small
hole is a mated female. It is only after being mated that
female pollinators exit their galls and then their natal
fig.
Continuous variables (seed or gall length and width)
were analysed using general linear models (proc GLM,
SAS 1999) and type III SS. Qualitative variables with
two levels (seed or wasp, closed or open gall, full or
empty gall) were analysed by logistic regression (bino-
mial distribution and logit link function, proc GEN-
MOD, SAS 1999), and qualitative variables with three
levels (closed, open or empty gall) by a log linear model
(Poisson distribution and log link function, proc GEN-
MOD, SAS 1999). For these analyses LR statistics for
type III analysis are provided (proc GENMOD, SAS
1999). I used forward model selection, keeping only
significant variables and interactions in the models (p B
0.05, unless to show the non-significance of some vari- Fig. 2. Regression lines between style and pedicel length for
each fig (black lines for flowers containing a wasp and gray
ables). In all analyses backwards selection gave the lines for flowers containing a seed), showing the overall nega-
same final model. tive correlation and the variability of the relationship.

OIKOS 95:3 (2001) 479


Table 2. LR statistics for type 3 analysis of a logistic regres-
sion of the content of the ovaries (seed or wasp). Only
significant variables are retained in the model. The content of
the ovaries depended on pedicel length and on the fig but not
on style length. n = 403.

Variable df x2 P

Pedicel length 1 109 B10−4


Fig (tree) 14 44 B10−4
Pedicel×tree 2 9 0.01

did not have a significant effect (df = 1, x2 = 0.27,


p \ 0.6). However, in a simple logistic model with
style length and fig as the only independent variables,
style length did explain the content of the ovaries
(df = 1, x2 = 62.9, pB0.0001). In this case, style
length did not enter into the final model because the
Fig. 3. Style length distribution according to ovary content. part of variance it explains is included in the variance
Both seeds and pollinators develop in flowers of all style explained by the variable pedicel length (since pedicel
length. The mode of the style length distribution is smaller for
pollinators than for seeds. n= 117 for flowers containing a and style length are correlated). This underlines the
seed and n =343 for flowers containing a pollinator. importance of measuring both variables. Fig. 4 illus-
trates that flowers containing a seed had a shorter
pedicel.
Female flowers of the entire range of style lengths
can be occupied by either a seed or a pollinator (Fig.
3). Pollinators occupied flowers of the entire range of Does rapidity of pollinator emergence depend on
pedicel lengths, but seeds mainly occupied flowers
the position of its gall?
with short pedicels (Fig. 4). However some seeds did
(rarely) develop in flowers with long pedicels. The different gall states (closed, with hole and a pol-
The content of the ovaries (seed or gall) depended linator inside or empty) reflect the rapidity of emer-
significantly on the fig and on pedicel length, with a gence, i.e. the time of wasp development plus the
marginal effect of pedicel ×tree (Table 2, two figs time spent waiting for males when development is
without seeds were excluded from this analysis). already completed. The gall state depended on the fig
When added to this model, the variable style length and on pedicel length (Table 3). When added to this
model, style length did not significantly affect gall
state (df =1, x2 = 0.25, p\ 0.6). Pedicels of empty
galls were longer than those of galls with a hole and
pollinators inside, which were in turn longer than
those of closed galls (Fig. 5). Thus wasps of the inner
layer of ovaries emerged first. This is due to the fact
that females occupying flowers with longer pedicels
were mated first (Table 4), and exited their gall first
(Table 5). I found no significant effect of style length
on these variables. Thus rapidity of pollinator emer-
gence depended on the position of the ovary within
the fig.

Table 3. LR statistics for type 3 analysis of a log linear model


of the gall state (closed, with a hole, empty). Pedicel length
significantly explained the gall state. For this analysis, I only
retained figs with galls in all three states. n = 94.

Fig. 4. Pedicel length distribution according to ovary content. Variable df x2 P


Both seeds and pollinators develop in flowers of all style
lengths. However, very few seeds develop in flowers with a Pedicel length 1 6.43 0.011
long pedicel. n = 117 for flowers containing a seed and n =273 Fig (tree) 3 6.78 0.079
for flowers containing a pollinator.

480 OIKOS 95:3 (2001)


Table 5. LR statistics for type 3 analysis of a logistic regres-
sion on whether or not the pollinator had exited from its gall.
Above a tree and a fig effect, pedicel length significantly
explained the pollinators’ exit status.

Variable df x2 P

Tree 2 5.9 0.05


Fig (tree) 12 56.2 1×10−3
Pedicel length 1 9.7 B10−4

characteristics but also to the quantity and quality of


pollinator visits.

An advantage of developing in the inner ovaries


Fig. 5. Percentage of the different gall states for each pedicel
length class, showing that galls with a wasp already exited Several advantages are associated with development in
have longer pedicels than galls with a mated wasp inside. the inner galls. First, inner galls are larger than outer
Virgin wasps tend to occupy galls with the shortest pedicels.
Sample size is in parentheses above the bars. galls, and are likely to host larger pollinators. Larger
pollinators have a higher probability of reaching a
Do gall and seed size vary with the position of receptive syconium and laying eggs (Herre 1989). More-
the ovary? over, larger species of pollinators have a higher number
of eggs (Herre, 1989). This pattern is likely to be
Both gall and seed size (height ×width) depended on maintained at the intra-specific level and to add an
the tree, on the fig and on pedicel length but not on advantage of larger size. Second, female pollinators
style length (Tables 6, 7). However, gall size increased developing in the outer galls should have a lower
with pedicel length (Fig. 6a) whereas seed size decreased probability of mating. When I opened the figs, most
with pedicel length (Fig. 6b). Thus the outer ovaries males had already exited, and at that time, the proba-
(Fig. 1) produced many large seeds and a few small bility of mating of virgin females should be low. Un-
galls, whereas inner ovaries produced few small seeds mated females can lay eggs but will only have sons,
and many large galls. Once more, position of the ovary resulting in zero fitness if the female is the only
(i.e. pedicel length) and not style length was responsible foundress. Third, females developing in the inner
for these differences. ovaries will exit the fig first, just after the males, which
could possibly aid in escaping ant predation. Duration
of wasp emergence from a single fig was about 1 h.
However, after emergence pollinators of Ficus micro-
Discussion
carpa live less than 5 h (pers. obs.). A small difference
In this study I demonstrate that pollinators developing in the timing of emergence is likely to have a strong
in flowers with long pedicels and short styles (i.e. with impact on pollinator fitness.
ovaries near the lumen of the fig) will tend to develop in After pollination, enlarging seeds and galls fill the
a larger gall. Female wasps developing in the innermost lumen of the fig, leading to its disappearance and to the
layer of ovaries will be mated first, and will exit their tight packing of the ovaries. In some species fruits may
gall, and probably their natal fig, earlier than those sometimes even burst before maturity (F. racemosa;
developing in ovaries in other positions. I will discuss pers. obs.). At the beginning of fruit development, just
below how this results in a fitness advantage in laying after pollination, there is some free space near the fig
eggs in flowers that will develop in the inner layers. In wall and near the lumen. Near the fig wall only between
most of our analyses there is either a tree and/or a fig
effect. These are likely to be due not only to plant Table 6. Results of a GLM model explaining gall size. Gall
size depended on the tree, on the fig and on pedicel length, but
not on style length. This model explains 40.7% of the total
Table 4. LR statistics for type 3 analysis of a logistic regres- variance.
sion of mating status of female pollinators. Whether the
female was mated or not depended on pedicel length and on Variable df Type III SS F P
the host tree. n=126.
Pedicel 1 33.1 25.6 B10−4
Variable df x2 P Style 1 1.1 0.9 0.33
Tree 2 63.5 24.6 B10−4
Pedicel length 1 15.7 B10−4 Fig (tree) 13 76.4 4.5 B10−4
Tree 2 13.0 B2×10−3 Error 253 326.8

OIKOS 95:3 (2001) 481


Table 7. Results of a GLM model explaining seed size. Seed wasps emerge from their galls, the fig enlarges and the
size depended on the tree, on the fig and on pedicel length, but lumen reappears in the centre. When male wasps exit, it
not on style length. This model explains 50.1% of the total
variance. should be easier for them to move in the lumen than
between compacted galls and seeds, which would explain
Variable df Type III SS F P why they mate first with females in the inner layers of
Pedicel 1 45.8 10.6 0.001 ovaries. As a consequence, these females will be the first
Style 1 11.0 2.55 0.11 to exit their galls and then the fig. These space constraints
Tree 2 86.8 10.0 B10−4 could explain the observed differences in gall size and
Fig (tree) 11 205.9 4.32 B10−4 development, and why these differences are more closely
Error 95 411.2
related to pedicel than to style length.

pedicels is there any free space, whereas ovaries touch


each other and are more compacted. Thus ovaries can How can foundresses choose flowers with a long
develop either towards the lumen or towards the fig wall. pedicel?
This would explain why galls, which develop mostly in
inner flowers, are bigger when they are nearest the centre, Space constraints do not explain why pedicel length also
while seeds, which develop mostly near the fig wall, are better explains the distribution of seeds and wasps than
bigger when they are nearest the fig wall. Just before does style length. During oviposition, foundresses could
conceivably assess style length and shape, but not pedicel
length. Foundresses could respond to length and curva-
ture of the style as an indicator of pedicel length, but this
seems unlikely. An alternative hypothesis is that there is
a differential development of the pedicel depending on
the position or on the content of the ovary. Due to the
same space constraints as discussed above, inner ovaries
would grow and be pushed towards the lumen, resulting
in a longer pedicel, whereas outer ovaries would develop
towards the fig wall and would have a shorter pedicel.
This consideration, and the different regression lines
between style and pedicel length, suggest that pedicel
growth is smaller for ovaries that contain a seed. Thus
foundresses preferentially lay eggs in flowers with a short
style (the content of the ovary depends on style length
when pedicel is not in the model, see Results). These
flowers have also a long pedicel and will grow and be
pushed towards the inside of the fig during development.
According to this reasoning, the pedicel of long-styled
flowers (especially those that contain a seed) will grow
less during post-pollination development. This could
explain why pedicel length is a better predictor of ovary
content than is style length.

A new hypothesis explaining pollinator


benevolence
These results confirm that pollinators can develop in all
flowers, whatever the lengths of their styles (Fig. 3) and
pedicels (Fig. 4). They also show that flowers are of
different quality for pollinator development depending
on their position (i.e. pedicel length). The preference of
foundresses for short-styled, long-pedicelled flowers, il-
lustrated by Figs 3 and 4, is now explained, since there
is a cost in laying eggs in ovaries close to the fig wall.
Fig. 6. a) Relationship between gall size and pedicel length; b) The only data supporting the ‘‘unbeatable-seeds’’
Relationship between seed size and pedicel length. For both
graphs each line represents one fig and the bold line the hypothesis (West and Herre 1994) is the apparent com-
relation for all flowers. petition between fig pollinators and some species of

482 OIKOS 95:3 (2001)


parasitic wasps. However, the existence within the fig of Bronstein, J. L. 1994. Conditional outcomes in mutualistic
flowers of different quality as larval hosts for both interactions. – Trends Ecol. Evol. 9: 214 – 217.
Bronstein, J. L. and Hossaert-McKey, M. 1996. Variation in
pollinators and parasites, is sufficient and more parsi- reproductive success within a subtropical fig/pollinator mu-
monious to explain this competition. Thus no data tualism. – J. Biogeogr. 23: 433 – 446.
exclusively support the unbeatable-seed hypothesis. Bull, J. J., Molineux, I. J. and Rice, W. R. 1991. Selection of
The ovipositor-length-optimization hypothesis (Gane- benevolence in a host-parasite system. – Evolution 45:
875 – 882.
shaiah et al. 1995, Kathuria et al. 1995) is lacking a Charnov, E. L., Maynard Smith, J. and Bull, J. J. 1976. Why
demonstrated cost of the ovipositor length. Different be an hermaphrodite? – Nature 236: 125 – 127.
quality of flowers depending on their position could fill Compton, S. G. 1993. One way to be a fig. – Afr. Entomol. 1:
151 – 158.
this gap. An ovipositor too long could increase the Douglas, A. E. 1994. Symbiotic interactions. – Oxford Univ.
percentage of eggs laid in long-styled, low-quality flow- Press.
ers and could entail a cost of ovipositor length. More Dugatkin, L. A., Mesterton-Gibbons, M. and Houston, A. I.
work is needed to test for the existence of such a cost. 1992. Beyond the Prisoner’s Dilemma: toward models to
discriminate among mechanisms of cooperation in nature.
Based on these hypothesis I propose a new explana- – Trends Ecol. Evol. 7: 202 – 205.
tion of pollinator benevolence in the fig/wasp mutual- Frank, S. A. 1997. Models of symbiosis. – Am. Nat. 150,
ism that synthesizes the ovipositor-length-optimization Suppl.: S80 – S99.
Freeman, S. and Rodriguez, R. J. 1993. Genetic conversion of
and the limited-egg-supply hypothesis. Selection on a fungal plant pathogen to a nonpathogenic, endophytic
ovipositor length could be frequency dependent: in the mutualist. – Science 260: 75 – 78.
absence of competition for egg laying sites (i.e. in Ganeshaiah, K. N., Kathuria, P., Shaanker, R. U. and Va-
single-foundress figs) there would be an advantage in sudeva, R. 1995. Evolution of style-length variability in figs
and optimization of ovipositor length in their pollinator
having a short ovipositor (long enough to permit laying wasps: a coevolutionary model. – J. Genet. 74: 25 – 39.
of all the foundress’ eggs but short enough to prevent Gibernau, M., Hossaert-McKey, M., Anstett, M. C. and Kjell-
laying of eggs in the outer layers of ovaries) whereas in berg, F. 1996. Consequences of protecting flowers in a fig:
a one way trip for pollinators? – J. Biogeogr. 23: 425 – 432.
situations with competition (i.e., several foundresses), Hallmann, J. and Sikora, R. A. 1994. Occurrence of plant
pollinators with a longer ovipositor would be favoured parasitic nematodes and non-pathogenic species of Fusar-
since they would have access to more ovaries. The ium in tomato plants in Kenya and their role as mutualistic
resulting ovipositor length would depend on fig struc- synergists for biological control of root-knot nematodes. –
Int. J. Pest Manage. 40: 321 – 325.
ture (style length distribution, number of female flowers Herre, E. A. 1989. Coevolution of reproductive characteristics
per fig) and on fig wasp population dynamics (distribu- in 12 species of New World figs and their pollinator wasps.
tion in space and time of the number of foundresses per – Experientia 45: 637 – 647.
Herre, E. A. 1993. Population structure and the evolution of
fig). Thus the proximate cause of seed/wasp production virulence in nematode parasites of fig wasps. – Science 259:
ratios at the syconium level would depend on the 1442 – 1445.
number of foundresses per fig: with one or few Janzen, D. H. 1979. How to be a fig. – Annu. Rev. Ecol. Syst.
foundresses the number of eggs would be limiting, 10: 13 – 51.
Kathuria, P. 1999. The pollination biology of monoecious fig
whereas with many foundresses, wasp production trees in India. – PhD thesis, Univ. of Leeds.
would be limited by the ovipositor length. This hypoth- Kathuria, P., Ganeshaiah, K. N., Shaanker, R. U. and Va-
esis seems congruent with currently published data but sudeva, R. 1995. Is there dimorphism for style lengths in
monoecious figs? – Curr. Sci. 68: 1047 – 1049.
needs further testing. Kerdelhué, C. and Rasplus, J.-Y. 1996. Non-pollinating Afro-
tropical fig wasps affect the fig-pollinator mutualism in
Acknowledgements – I thank Martine Hossaert and Doyle Ficus within the subgenus Sycomorus. – Oikos 75: 3 – 14.
McKey for helpful comments on the manuscript, and Edmond Kerdelhué, C., Rossi, J. P. and Rasplus, J.-Y. 2000. Compara-
Dounias for preparing Fig. 1. tive community ecology studies on old world figs and fig
wasps. – Ecology 81: 2832 – 2849.
Kjellberg, F., Ibrahim, M., Raymond, M. and Valdeyron, G.
1987. The stability of the symbiosis between dioecious figs
References and their pollinators: a study of Ficus carica L. and
Blastophaga psenes L. – Evolution 41: 693 – 704.
Anstett, M.-C., Bronstein, J. L. and Hossaert-McKey, M. Kover, P. X. and Clay, K. 1998. Trade-off between virulence
1996. Resource allocation: a conflict in the fig/fig wasp and vertical transmission and the maintenance of a virulent
mutualism? – J. Evol. Biol. 9: 417 –428. plant pathogen. – Am. Nat. 152: 165 – 175.
Anstett, M.-C., Hossaert-McKey, M. and Kjellberg, F. 1998. Lipsitch, M., Herre, E. A. and Nowak, M. A. 1995. Host
Figs and fig pollinator: evolutionary conflicts in a co- population structure and the evolution of virulence: a ‘‘law
evolved mutualism. – Trends Ecol. Evol. 12: 94 – 99. of diminishing returns’’. – Evolution 49: 743 – 748.
Blackstone, N. W. 1995. Perspective: a units-of-evolution per- Lipsitch, M., Siller, S. and Nowak, M. A. 1996. The evolution
spective on the endosymbiont theory of the origin of the of virulence in pathogens with vertical and horizontal
mitochondrion. – Evolution 49: 785 –796. transmission. – Evolution 50: 1729 – 1741.
Bronstein, J. L. 1988. Mutualism, antagonism, and the fig-pol- Margulis, L. and Fester, R. (eds) 1991. Symbiosis as a source
linator interaction. – Ecology 69: 1298 –1302. of evolutionary innovation. – MIT Press.
Bronstein, J. L. 1992. Seed predators as mutualists: ecology McKey, D. 1989. Population biology of figs: application for
and evolution of the fig/pollinator interaction. – In: conservation. – Experientia 45: 661 – 674.
Bernays, E. (ed.), Insect-plant interactions IV. CRC Press, Mesterton-Gibbons, M. 1991. An escape from ‘the prisoner’s
pp. 1– 44. dilemma’. – J. Math. Biol. 29: 251 – 269.

OIKOS 95:3 (2001) 483


Nefdt, R. J. C. and Compton, S. G. 1996. Regulation of seed Verkerke, W. 1987a. Ovule dimorphism in Ficus asperifolia
and pollinator production in the fig-fig wasp mutualism. – Miquel. – Acta Bot. Neerl. 36: 121 – 124.
J. Anim. Ecol. 65: 170 –182. Verkerke, W. 1987b. Syconial anatomy of Ficus asperifolia
Patel, A. and McKey, D. 1998. Sexual specialization in two (Moraceae), a gynodioecious tropical fig. – Proc. K. Ned.
tropical dioecious figs. – Oecologia 115: 391 – 400. Akad. Wet. C 90: 461 – 492.
Patel, A., Hossaert-McKey, M. and McKey, D. 1993. Ficus- West, S. A. and Herre, E. A. 1994. The ecology of the New
pollinator research in India: past, present and future. – World fig-parasitizing wasps Idarnes and implications for
Curr. Sci. 65: 243 –253. the evolution of the fig-pollinator mutualism. – Proc. R.
SAS 1999. Ver. 8.0. – SAS Inst., Inc., Cary, NC. Soc. Lond. B 258: 67 – 72.
Thompson, J. N. 1982. Interaction and coevolution. – J. Yamamura, N. 1993. Vertical transmission and evolution of
Wiley and Sons. mutualism from parasitism. – Theor. Popul. Biol. 44:
Toft, C. A. and Karter, A. J. 1990. Parasite-host coevolution. 95 – 109.
– Trends Ecol. Evol. 5: 326 –329. Yamamura, N. 1996. Evolution of mutualistic symbiosis: a
Trivers, R. 1971. The evolution of reciprocal altruism. – Q. differential equation model. – Res. Popul. Ecol. 38: 211 –
Rev. Biol. 46: 35 –37. 218.

484 OIKOS 95:3 (2001)

You might also like