TBE 2015 Fisherian Fies: benefits of female choice in a lekking sandfly by T. M. Jones, R. J. Quinnell and Andrew Balmford Lyrics
We experimentally investigated the fitness consequences of female mate choice in order to test the relative importance of three competing but non-exclusive hypotheses for the maintenance of pronounced female mating preferences on leks: that females benefit directly; that they gain indirect Fisherian benefits by producing more attractive sons; or that they benefit indirectly because preferred males possess `good genes' that confer increased viability on their sons and daughters. We allowed lekking female sandflies, Lutzomyia longipalpis, to choose between males of varying attractiveness to females, and monitored the consequences for their own survival and reproductive success as well as for their o¡spring. In contrast to the predictions of the direct-beneits model, we found no clear sire effect on the fecundity or survival of the females themselves; females mating with more attractive males did survive longer after oviposition, but never long enough to undertake a second batch of egg laying.We also found no evidence that females gained good-genes benefits in terms of enhanced o¡spring survival. However, we did find that generally attractive males fathered sons who were then chosen when they in turn formed leks. Although not completely precluding other benefits, our results indicate that Fisherian benefits are at least partly responsible for maintaining female choice at L. longipalpis leks. These findings indicate the importance of testing all putative benefits concurrently in exploring the maintenance of female mate choice.
Keywords: female choice; Fisherian benefits; good genes; direct benefits; lekking; lek paradox
1. INTRODUCTION
Why mate choice is most pronounced on leksöarenas
where displaying males apparently o¡er discriminating
females nothing but spermöremains one of evolutionary
biology's most enduring puzzles (Darwin 1871;
Kirkpatrick & Ryan 1991; Andersson 1994). Three non-
exclusive explanations suggest choice could increase
female ¢tness directly (by helping females avoid males
with low fertility or sexually transmitted diseases, for
instance), or indirectly, either because male attractiveness
is itself heritable (so choosy females produce attractive
sonsöthe Fisherian process (Fisher 1930)) or because it
signals heritable viability (so females acquire good genes
for both their sons and daughters (Reynolds & Gross
1990; Kirkpatrick & Ryan 1991; Andersson 1994)).
There is now growing evidence for both direct and
indirect bene¢ts of mate choice in various lekking and
non-lekking species (table 1). However, demonstrating
that a particular bene¢t exists does not by itself mean
that it plays the major role in maintaining female choice
in a particular species. For example, if direct bene¢ts are
present, they are likely to have a much greater e¡ect on
female ¢tness than any indirect bene¢ts (Price 1993;
Kirkpatrick & Barton 1997). Thus evidence that indirect
bene¢ts underpin female choice would be most
convincing if coupled with evidence of a lack of direct
bene¢ts. Moreover, discriminating between indirect-
bene¢t models is potentially confounded by the fact that
both Fisherian and good-genes models can accommodate
heritability of male attractiveness (Balmford & Read
1991). In the former, it drives the initial evolution of the
female preference (although it may not even be detectable
nowadays (Grafen 1990)); in the latter, it can arise from a
correlation between attractiveness and viability. Thus
evidence for a Fisher e¡ect is unambiguous only if
coupled with the demonstration that good-genes bene¢ts
are weak or unimportant. Similarly, because (like good-
genes models) Fisherian models predict that preferred
traits will become costly and only expressed by high-
quality males, then provided viability is heritable, choosy
females might nowadays acquire good genes even if their
preferences arose initially (and are largely maintained)
by a Fisherian bene¢t (Balmford & Read 1991). These
considerations mean that studies can best provide support
for either indirect model if all three possible models are
tested simultaneously, direct bene¢ts are absent (or very
limited), and one indirect bene¢t is relatively strong and
the other weak or absent (for similar arguments, see
Iwasa et al. 1991; Kirkpatrick & Ryan 1991; Andersson
1994). So far few studies have achieved this (but see
Moore 1994;Whittier & Kaneshiro 1995).
Here we present the results of an experiment that
attempts to test all three models simultaneously in a
lekking sand£y, Lutzomyia longipalpis (Diptera: Psycho-
didae). Phlebotomine sand£ies of the L. longipalpis species
complex form leks (Bradbury 1981) on or near their
vertebrate hosts (Jones 1997); in the laboratory, these can
be replicated (with or without a host) in small net cages
(Jarvis & Rutledge 1992; Jones 1997). L. longipalpis leks
are relatively mobile in space. Males exhibit territorial
behaviour, defending areas approximately 2 cm in radius
(Jones 1997). Male mating success within leks is highly
skewed and largely determined by female choice (Jones
1997): females typically sample several males before
mating with just one, and can readily terminate courtship
by depressing their abdomens or moving away. Female
choice is correlated with male production of a phero-
mone, which is thought to be dispersed by a wing-fanning
display (Jones & Hamilton 1998). Females are haemato-
phagous, requiring a bloodmeal for egg development.
This process takes approximately three days, after which
females lay a single batch of 40^60 eggs. Neither sex
contributes to parental care.
Our experiment explored the consequences of a
female's choice of mate in terms of her own ¢tness and
that of her o¡spring. Females were permitted to mate
freely in aggregations of ¢ve males that di¡ered in their
attractiveness to females, where attractiveness is de¢ned
as the number of females a male successfully mates with.
We then tested the proposed bene¢ts associated with each
model (as summarized in table 1) by examining how the
survival and reproductive success of the females and their
o¡spring compared with the general attractiveness of
their mate and their father, respectively.
2. METHODS
Experimental £ies were obtained from a 22-generation
captive colony originating from the Salvaterra district of
Marajo¨, Brazil, and reared using standard culturing techniques
at the Instituto Evandro Chagas, Bele¨m (Killick-Kendrick et al.
1977; Jones 1997). Adults were maintained in small cages
(15 cm15 cm15 cm) and provided with cotton wool swabs
soaked in 30% sugar solution. Larvae were reared in Petri
dishes (diameter 10 cm, height 1.5 cm) on a diet of dried liver
powder. The population was maintained at ca. 500 adults by
regulating the numbers of females feeding in each generation.
Virgin £ies were obtained by releasing adults into single-sex
cages less than 12 h after emergence (Chaniotis 1967). Females
were blood-fed on an anaesthetized host three days after emer-
gence (Jones 1997, after Flecknell 1987).
Our experiment began with 25 trials in which we sequentially
exposed ten virgin, blood-fed, 4-day-old females to caged leks
consisting of ¢ve individually marked, 4^6-day-old males.
Marking was achieved using £uorescent powders that have no
detectable impact on a male's mating success or immediate
survival (Jones 1997). Females were introduced singly, allowed to
mate once, and then removed. The distribution of matings was
highly overdispersed, with females exhibiting consistent prefer-
ences for particular mates (frequency distribution of males with
0, 1, 2, 3, 4 or 44 matings, compared with Poisson: 25
106.2,
p50.001). Next, to augment the number of females that chose
generally unattractive mates, we regrouped those males
(N30) that obtained the fewest matings (usually zero) in the
¢rst trials into six new ¢ve-male leks. Ten further virgin females
were introduced singly to each of these new leks. The females
again mated non-randomly (25
12.6, p50.05). Conditions were
thus similar in both sets of trials and all experimental females
experienced mate choice.
To test the three hypothesized bene¢ts of mate choice at leks
(table 1), we maintained those experimental females that
survived to oviposition, and compared measures of their ¢tness
and that of their o¡spring with the overall attractiveness of the
males that they chose (de¢ned as the number of matings they
achieved in the ¢rst round of trials). To avoid possible effects of
sperm depletion, we excluded any females not among the ¢rst
four mates of a given male (although this is in fact conservative,
as male fertility declines only after six consecutive matings
(Jones 1997; see also later)). After these exclusions, our sample
consisted of 186 females.
(a) Direct bene¢ts
To test for potential direct bene¢ts of mate choice, females
were individually placed in small, glass, oviposition tubes at the
same densities as those used in standard culturing (Jones 1997).
They were then observed daily, and records taken of the time to
oviposition, the number of eggs laid, and the time to death. A
female's total survival was compared to the mating success of
her mate in a generalized linear model with a ¢tted Weibull
distribution, to allow for age-dependent survival (Pinder et al.
1978). The signi¢cance of each parameter was checked by step-
wise deletion, adding signi¢cant terms (at p50.05) back into the
model before removing the next parameter (after Crawley 1993).
More detailed measures of survival (between mating and ovipo-
sition, and survival from the start of oviposition) were compared
to male mating success using parametric correlations, because
the low numbers of days survived did not permit formal survival
analysis. Likewise, variation in the numbers of eggs laid was
analysed by correlation. In all cases, male attractiveness was set
as the independent variable.
(b) Good-genes bene¢ts
Because of a lack of genetic markers for L. longipalpis, we were
unable to assess o¡spring survival with trials using standardized,
unrelated competitors (Partridge 1980; Taylor et al. 1987). We
thus used a more conservative method to test whether females
gained good-genes bene¢ts from mate choice (see also Simmons
1987; Norris 1993; Moore 1994). We transferred their eggs to
Petri dishes and compared the subsequent survival of families
with paternal attractiveness. Eggs and larvae were monitored
daily to record the numbers surviving to successive instars.
O¡spring of di¡erent females were kept separate until adult-
hood, so that we could compare their survival with the precise
mating success of their fathers. To achieve this while still
matching egg and larval densities to those seen in standard
culturing, we switched to using smaller Petri dishes (diameter
5 cm, height 1.0 cm). Under these conditions, mean survival of
our experimental progeny was comparable to that recorded for
the parent colony (t-tests across the six developmental stages
from egg to adult emergence: all n.s.). The proportion of larvae
surviving each successive instar was calculated. Where data
could be transformed to normality, these proportions were
compared to paternal mating success using parametric correla-
tions.Where transformation to normality was not possible, non-
parametric correlation tests were done. As with tests for direct
bene¢ts, male mating success was set as the independent vari-
able for all comparisons.
After adult emergence, we selected two sons and two daugh-
ters each from those females whose mates had achieved the
highest, median or lowest number of matings in the ¢rst round
of trials (means (s.d.) of 4.9 (0.2), 2.0 (0.1) and 0.1 (0)
matings respectively, for high, average and low attractiveness
sires). O¡spring of males that could not be unambiguously
assigned to one of these three categories were discarded. The
culture conditions applied to all these adult o¡spring were the
same as those experienced by their parents. One set of sons and
one set of daughters were monitored daily to investigate survival
of non-reproducing o¡spring. The second set of daughters was
blood-fed at three days and allowed to oviposit, in order to test
for possible sire e¡ects on daughter fecundity; the ¢nal set of
sons was used to look for Fisherian bene¢ts (see below). The
survival of all adult o¡spring and the fecundity of fed daughters
were analysed in the same way as for their mothers. Because
paternal identity was no longer known, the measures we
obtained on o¡spring viability as adults were compared not
with paternal attractiveness per se, but rather with the group to
which their father belonged, which was entered as a categorical
variable with three levels.
(c) Fisherian bene¢ts
We tested for Fisherian bene¢ts by using the last set of sons to
establish 38 three-male leks, each consisting of one son of a male
of high, average and low attractiveness. Six fresh virgin females
were introduced singly to each lek and allowed to mate; the
resulting skew in mating success and female choice matched that
seen in ¢ve-male leks (Jones 1997). The attractiveness of sons was
assessed in the same way as that of their fathers. The proportion
of matings obtained by a son was then analysed in a general
linear model with a binomial error distribution, and with
paternal group again as a categorical variable with three levels.
3. RESULTS
If females visiting leks gain direct ¢tness bene¢ts from
mate choice, those mating with generally attractive males
should exhibit higher survival or fecundity than other
females (Kirkpatrick & Ryan 1991). Our results suggest
that this is not the case (for a detailed statistical break-
down, see table 2). There was no link between the total
survival of females, or the time between mating and
oviposition, and the mating success of their mates.We did
¢nd a di¡erence in survival after oviposition: females
mating with the most attractive males survived an
average of one day longer than those that chose the least
attractive males. However, as no females survived to
undertake a second gonotrophic cycle, the biological
signi¢cance of this result is unclear. There was also no
evidence that females laid more eggs in response to
choosing more attractive mates (table 2 and ¢gure 1a),
suggesting that there is no fecundity advantage to mate
choice. These results were not biased by female mating
order: the number of previous matings a female's chosen
male had achieved did not in£uence her probability of
survival (21
0.38, N186, n.s.), or her fecundity
(r0.10, N186, n.s.).
Although there was no evidence that females mating
with generally attractive males increased their ¢tness
directly, there was evidence of indirect bene¢ts. When
introduced to leks consisting of three sons, virgin females
once again showed clear preferences for particular males
(23
12.7, p50.01). Moreover, in agreement with Fish-
erian models, the mating success of sons increased with
the attractiveness of their fathers (signi¢cance of paternal
group as a non-directional variable in general linear
model: F2,113
6.55, p50.02; ordered heterogeneity test
(Rice & Gaines 1994): rsPc
0.96, p50.01; ¢gure 1b).
Speci¢cally, sons of more attractive fathers obtained far
more than one-third of all the matings in their trial (one-
sample t-test on logistically transformed data: t2.93,
d.f.37, p50.01). Although these results provide impor-
tant support for a Fisherian interpretation of female
preferences, they could conceivably have a good-genes
explanation (with inherited high viability enabling sons of
attractive males to invest heavily in preferred traits).
In an attempt to distinguish between these alternatives,
we compared o¡spring survival to paternal attractiveness.
We found no evidence that a female's mate choice
in£uenced the viability of her o¡spring when sibs were
reared in family groups (for a detailed breakdown, see
table 3). Survival of o¡spring from egg hatch to adulthood
(table 3 and ¢gure 1c), survival across each developmental
stage, and the number of adult o¡spring each female
produced were all independent of paternal mating success.
Likewise, the survival as adults of sons and both reprodu-
cing and non-reproducing daughters did not increase with
sire attractiveness (table 3 and ¢gure 1d). There was a weak
trend for survival to decrease with sire mating success in
the group of daughters that was allowed to reproduce, but
after correction for multiple comparisons (Rice 1989) this
pattern was non-signi¢cant. Finally, the fecundity of
reproducing daughters was again unrelated to their
mothers' choice of mate. These results appear to contradict
key predictions of good-genes models.
4. DISCUSSION
In 1994, Malte Andersson wrote that `no critical test has
been performed that supports Fisherian sexual selection
and excludes the alternatives, or estimates their relative
importance' (Andersson 1994, p.52). Here, we have
attempted to assess the relative importance of the three
main models proposed for the maintenance of female
choice at leks.We found that females mating with generally
attractive males did not obtain any clear-cut direct bene¢ts
from their choice of mate. Similarly, we found no evidence
to suggest that families of o¡spring sired by more attractive
males exhibit higher viability, at any stage of development.
However, we did ¢nd that sons of attractive fathers them-
selves obtained disproportionately more matings. The
evidence presented thus shows that females gain indirect
bene¢ts from mating with particularly attractive males,
and highlights a Fisherian mechanism as being at least in
part responsible for the current maintenance of female
mate choice in L. longipalpis.
Di¡erential maternal investment (a potential problem
elsewhere (Partridge 1980; Simmons 1987)) is unlikely to
confound the observed correlation between the attractive-
ness of fathers and sons. Sand£ies are highly fecund, so
we would expect di¡erences in reproductive e¡ort to be
mediated via variation in egg number rather than quality
(Simmons 1987; Petrie 1994), yet we found no evidence
that females laid more eggs for attractive mates. A more
subtle maternal e¡ect might arise because, although we
allocated some females to groups of generally unattractive
males, all our females chose their mates themselves (cf.
Moore 1994; Petrie 1994). Fisherian models rely on
linkage disequilibrium between genes coding for female
preferences and for preferred traits. Thus maternally
rather than (or as well as) paternally inherited genes
might be responsible for the attractiveness of those sons
whose mothers chose generally attractive mates. Never-
theless, it is hard to explain such a pattern of inheritance
without invoking an underlying Fisherian mechanism. It
would be worthwhile replicating this study using ¢rst-
generation captive stock to rule out the possibility that
the observed heritability of male attractiveness might be
the result of the rapid evolution of trait and preference in
the laboratory (Shelly et al. 1994).
A common criticism of laboratory-based studies of
evolutionary questions is that an inability to detect
variations in ¢tness may be a result of experimental
conditions. Our tests of putative direct and good-genes
bene¢ts are potentially vulnerable to the same criticism.
For example, most direct bene¢ts accrued by choosy
females should be re£ected in increased survival or
immediate fecundity.We found no evidence of this, but it
is conceivable that the enhanced post-oviposition survival
detected here might translate into a direct bene¢t if it
appreciably increases the probability of free-ranging
females entering a second gonotrophic cycle. Likewise,
certain postulated direct bene¢ts of mate choice (such as
avoidance of predators or sexually transmitted diseases)
could not be detected in our system. However, there is
little empirical support for the predator- or disease-avoid-
ance hypotheses in the current lekking literature
(Ho« glund & Alatalo 1995).
Moving on to good-genes bene¢ts, it could be argued
that these do accrue to choosy female sand£ies but were
masked here by the `welfare state' conditions of our
experiment. However, genetic di¡erences in o¡spring
survival are frequently detected in laboratory experiments
(Simmons 1987; Taylor et al. 1987; Reynolds & Gross 1992;
Moore 1994), and importantly in this particular case,
conditions were clearly not favourable enough to mask
genetic di¡erences in the attractiveness of sons. In
addition, we are con¢dent that, despite a large amount of
within-group variation, our methodology was capable of
detecting meaningful di¡erences in o¡spring survival
across treatment groups if these existed. Power analysis
(Cohen 1988) shows that with our sample size and
variation in survival, we had a 95% chance of detecting a
correlation coe¤cient between o¡spring survival and
mating success of as little as 0.26. To put this in context,
the median r-value between measures of o¡spring
survival and sire attractiveness reported by ¢eld studies
claiming good-genes e¡ects is 0.47 (von Schantz et al.
1989; Norris 1993; Petrie 1994; Hasselquist et al. 1996;
Sheldon et al. 1997); the equivalent ¢gure from the
laboratory is 0.65 (Taylor et al. 1987; Reynolds & Gross
1992; Moore 1994). Thus good-genes e¡ects in lekking
L. longipalpis would have to be considerably weaker than
those found in other species to have been missed by our
experimental protocol.
Figure 1. Tests of proposed bene¢ts of mate choice at sand£y leks. (a) Number of eggs laid by females in relation to the attrac-
tiveness of the males they chose. (b) Proportion of matings (mean s.e., calculated using logistically transformed data) achieved
by sons of males of low, average and high attractiveness, competing in three-male leks. (c) Proportional survival of o¡spring of
females in relation to the mating success of their mates. (d) Survival as adults (mean s.e.) of non-reproducing sons and daugh-
ters and reproducing daughters sired by males of low, average and high attractiveness. Detailed statistical breakdowns are given
in tables 2 and 3.
A ¢nal consideration is that larvae in this study only
had the opportunity to compete with full sibs. It is
possible that competition within a family of low-quality
individuals and within a family of high-quality indivi-
duals may yield similar mean survival rates, thereby
obscuring true ¢tness di¡erences: low-quality individuals
may only survive when they have low-quality competi-
tion. However, several studies of other species have shown
good-genes e¡ects despite rearing young alongside their
sibs (Simmons 1987; Norris 1993; Moore 1994). Moreover,
laboratory studies of L. longipalpis suggest that eggs and
larvae are likely to be aggregated in family groups in the
¢eld (Elnaiem & Ward 1992; Dougherty et al. 1993, 1994).
Thus, even if rearing in family groups did mask some
underlying di¡erences in o¡spring competitive ability in
our experiment, such di¡erences might be equally
masked (and unavailable to choosy females) in free-
ranging populations.
In conclusion, we found evidence for Fisherian but no
other detectable mate-choice bene¢ts in L. longipalpis. We
therefore suggest that heritability of male attractiveness
is an important factor maintaining female mating
preferences in this species (although this does not of
course preclude the possibility that other adaptive or
non-adaptive mechanisms were responsible for the origin
of these preferences (Balmford & Read 1991)). Other
studies, particularly those claiming good-genes e¡ects,
have often not even looked for Fisherian bene¢ts, and we
suggest that indirect ¢tness gains via increased attrac-
tiveness of sons may turn out to be more widespread
than generally thought. We believe that future studies
should no longer argue for the importance of particular
mate-choice bene¢ts unless they also test for alternatives.
The approach adopted here, with concurrent tests of all
the main postulated bene¢ts (and with further modi¢ca-
tion of the protocol for assessing o¡spring viability, if
possible using genetic markers) provides a robust way
forward for studies into the maintenance of female
mating preferences.
Keywords: female choice; Fisherian benefits; good genes; direct benefits; lekking; lek paradox
1. INTRODUCTION
Why mate choice is most pronounced on leksöarenas
where displaying males apparently o¡er discriminating
females nothing but spermöremains one of evolutionary
biology's most enduring puzzles (Darwin 1871;
Kirkpatrick & Ryan 1991; Andersson 1994). Three non-
exclusive explanations suggest choice could increase
female ¢tness directly (by helping females avoid males
with low fertility or sexually transmitted diseases, for
instance), or indirectly, either because male attractiveness
is itself heritable (so choosy females produce attractive
sonsöthe Fisherian process (Fisher 1930)) or because it
signals heritable viability (so females acquire good genes
for both their sons and daughters (Reynolds & Gross
1990; Kirkpatrick & Ryan 1991; Andersson 1994)).
There is now growing evidence for both direct and
indirect bene¢ts of mate choice in various lekking and
non-lekking species (table 1). However, demonstrating
that a particular bene¢t exists does not by itself mean
that it plays the major role in maintaining female choice
in a particular species. For example, if direct bene¢ts are
present, they are likely to have a much greater e¡ect on
female ¢tness than any indirect bene¢ts (Price 1993;
Kirkpatrick & Barton 1997). Thus evidence that indirect
bene¢ts underpin female choice would be most
convincing if coupled with evidence of a lack of direct
bene¢ts. Moreover, discriminating between indirect-
bene¢t models is potentially confounded by the fact that
both Fisherian and good-genes models can accommodate
heritability of male attractiveness (Balmford & Read
1991). In the former, it drives the initial evolution of the
female preference (although it may not even be detectable
nowadays (Grafen 1990)); in the latter, it can arise from a
correlation between attractiveness and viability. Thus
evidence for a Fisher e¡ect is unambiguous only if
coupled with the demonstration that good-genes bene¢ts
are weak or unimportant. Similarly, because (like good-
genes models) Fisherian models predict that preferred
traits will become costly and only expressed by high-
quality males, then provided viability is heritable, choosy
females might nowadays acquire good genes even if their
preferences arose initially (and are largely maintained)
by a Fisherian bene¢t (Balmford & Read 1991). These
considerations mean that studies can best provide support
for either indirect model if all three possible models are
tested simultaneously, direct bene¢ts are absent (or very
limited), and one indirect bene¢t is relatively strong and
the other weak or absent (for similar arguments, see
Iwasa et al. 1991; Kirkpatrick & Ryan 1991; Andersson
1994). So far few studies have achieved this (but see
Moore 1994;Whittier & Kaneshiro 1995).
Here we present the results of an experiment that
attempts to test all three models simultaneously in a
lekking sand£y, Lutzomyia longipalpis (Diptera: Psycho-
didae). Phlebotomine sand£ies of the L. longipalpis species
complex form leks (Bradbury 1981) on or near their
vertebrate hosts (Jones 1997); in the laboratory, these can
be replicated (with or without a host) in small net cages
(Jarvis & Rutledge 1992; Jones 1997). L. longipalpis leks
are relatively mobile in space. Males exhibit territorial
behaviour, defending areas approximately 2 cm in radius
(Jones 1997). Male mating success within leks is highly
skewed and largely determined by female choice (Jones
1997): females typically sample several males before
mating with just one, and can readily terminate courtship
by depressing their abdomens or moving away. Female
choice is correlated with male production of a phero-
mone, which is thought to be dispersed by a wing-fanning
display (Jones & Hamilton 1998). Females are haemato-
phagous, requiring a bloodmeal for egg development.
This process takes approximately three days, after which
females lay a single batch of 40^60 eggs. Neither sex
contributes to parental care.
Our experiment explored the consequences of a
female's choice of mate in terms of her own ¢tness and
that of her o¡spring. Females were permitted to mate
freely in aggregations of ¢ve males that di¡ered in their
attractiveness to females, where attractiveness is de¢ned
as the number of females a male successfully mates with.
We then tested the proposed bene¢ts associated with each
model (as summarized in table 1) by examining how the
survival and reproductive success of the females and their
o¡spring compared with the general attractiveness of
their mate and their father, respectively.
2. METHODS
Experimental £ies were obtained from a 22-generation
captive colony originating from the Salvaterra district of
Marajo¨, Brazil, and reared using standard culturing techniques
at the Instituto Evandro Chagas, Bele¨m (Killick-Kendrick et al.
1977; Jones 1997). Adults were maintained in small cages
(15 cm15 cm15 cm) and provided with cotton wool swabs
soaked in 30% sugar solution. Larvae were reared in Petri
dishes (diameter 10 cm, height 1.5 cm) on a diet of dried liver
powder. The population was maintained at ca. 500 adults by
regulating the numbers of females feeding in each generation.
Virgin £ies were obtained by releasing adults into single-sex
cages less than 12 h after emergence (Chaniotis 1967). Females
were blood-fed on an anaesthetized host three days after emer-
gence (Jones 1997, after Flecknell 1987).
Our experiment began with 25 trials in which we sequentially
exposed ten virgin, blood-fed, 4-day-old females to caged leks
consisting of ¢ve individually marked, 4^6-day-old males.
Marking was achieved using £uorescent powders that have no
detectable impact on a male's mating success or immediate
survival (Jones 1997). Females were introduced singly, allowed to
mate once, and then removed. The distribution of matings was
highly overdispersed, with females exhibiting consistent prefer-
ences for particular mates (frequency distribution of males with
0, 1, 2, 3, 4 or 44 matings, compared with Poisson: 25
106.2,
p50.001). Next, to augment the number of females that chose
generally unattractive mates, we regrouped those males
(N30) that obtained the fewest matings (usually zero) in the
¢rst trials into six new ¢ve-male leks. Ten further virgin females
were introduced singly to each of these new leks. The females
again mated non-randomly (25
12.6, p50.05). Conditions were
thus similar in both sets of trials and all experimental females
experienced mate choice.
To test the three hypothesized bene¢ts of mate choice at leks
(table 1), we maintained those experimental females that
survived to oviposition, and compared measures of their ¢tness
and that of their o¡spring with the overall attractiveness of the
males that they chose (de¢ned as the number of matings they
achieved in the ¢rst round of trials). To avoid possible effects of
sperm depletion, we excluded any females not among the ¢rst
four mates of a given male (although this is in fact conservative,
as male fertility declines only after six consecutive matings
(Jones 1997; see also later)). After these exclusions, our sample
consisted of 186 females.
(a) Direct bene¢ts
To test for potential direct bene¢ts of mate choice, females
were individually placed in small, glass, oviposition tubes at the
same densities as those used in standard culturing (Jones 1997).
They were then observed daily, and records taken of the time to
oviposition, the number of eggs laid, and the time to death. A
female's total survival was compared to the mating success of
her mate in a generalized linear model with a ¢tted Weibull
distribution, to allow for age-dependent survival (Pinder et al.
1978). The signi¢cance of each parameter was checked by step-
wise deletion, adding signi¢cant terms (at p50.05) back into the
model before removing the next parameter (after Crawley 1993).
More detailed measures of survival (between mating and ovipo-
sition, and survival from the start of oviposition) were compared
to male mating success using parametric correlations, because
the low numbers of days survived did not permit formal survival
analysis. Likewise, variation in the numbers of eggs laid was
analysed by correlation. In all cases, male attractiveness was set
as the independent variable.
(b) Good-genes bene¢ts
Because of a lack of genetic markers for L. longipalpis, we were
unable to assess o¡spring survival with trials using standardized,
unrelated competitors (Partridge 1980; Taylor et al. 1987). We
thus used a more conservative method to test whether females
gained good-genes bene¢ts from mate choice (see also Simmons
1987; Norris 1993; Moore 1994). We transferred their eggs to
Petri dishes and compared the subsequent survival of families
with paternal attractiveness. Eggs and larvae were monitored
daily to record the numbers surviving to successive instars.
O¡spring of di¡erent females were kept separate until adult-
hood, so that we could compare their survival with the precise
mating success of their fathers. To achieve this while still
matching egg and larval densities to those seen in standard
culturing, we switched to using smaller Petri dishes (diameter
5 cm, height 1.0 cm). Under these conditions, mean survival of
our experimental progeny was comparable to that recorded for
the parent colony (t-tests across the six developmental stages
from egg to adult emergence: all n.s.). The proportion of larvae
surviving each successive instar was calculated. Where data
could be transformed to normality, these proportions were
compared to paternal mating success using parametric correla-
tions.Where transformation to normality was not possible, non-
parametric correlation tests were done. As with tests for direct
bene¢ts, male mating success was set as the independent vari-
able for all comparisons.
After adult emergence, we selected two sons and two daugh-
ters each from those females whose mates had achieved the
highest, median or lowest number of matings in the ¢rst round
of trials (means (s.d.) of 4.9 (0.2), 2.0 (0.1) and 0.1 (0)
matings respectively, for high, average and low attractiveness
sires). O¡spring of males that could not be unambiguously
assigned to one of these three categories were discarded. The
culture conditions applied to all these adult o¡spring were the
same as those experienced by their parents. One set of sons and
one set of daughters were monitored daily to investigate survival
of non-reproducing o¡spring. The second set of daughters was
blood-fed at three days and allowed to oviposit, in order to test
for possible sire e¡ects on daughter fecundity; the ¢nal set of
sons was used to look for Fisherian bene¢ts (see below). The
survival of all adult o¡spring and the fecundity of fed daughters
were analysed in the same way as for their mothers. Because
paternal identity was no longer known, the measures we
obtained on o¡spring viability as adults were compared not
with paternal attractiveness per se, but rather with the group to
which their father belonged, which was entered as a categorical
variable with three levels.
(c) Fisherian bene¢ts
We tested for Fisherian bene¢ts by using the last set of sons to
establish 38 three-male leks, each consisting of one son of a male
of high, average and low attractiveness. Six fresh virgin females
were introduced singly to each lek and allowed to mate; the
resulting skew in mating success and female choice matched that
seen in ¢ve-male leks (Jones 1997). The attractiveness of sons was
assessed in the same way as that of their fathers. The proportion
of matings obtained by a son was then analysed in a general
linear model with a binomial error distribution, and with
paternal group again as a categorical variable with three levels.
3. RESULTS
If females visiting leks gain direct ¢tness bene¢ts from
mate choice, those mating with generally attractive males
should exhibit higher survival or fecundity than other
females (Kirkpatrick & Ryan 1991). Our results suggest
that this is not the case (for a detailed statistical break-
down, see table 2). There was no link between the total
survival of females, or the time between mating and
oviposition, and the mating success of their mates.We did
¢nd a di¡erence in survival after oviposition: females
mating with the most attractive males survived an
average of one day longer than those that chose the least
attractive males. However, as no females survived to
undertake a second gonotrophic cycle, the biological
signi¢cance of this result is unclear. There was also no
evidence that females laid more eggs in response to
choosing more attractive mates (table 2 and ¢gure 1a),
suggesting that there is no fecundity advantage to mate
choice. These results were not biased by female mating
order: the number of previous matings a female's chosen
male had achieved did not in£uence her probability of
survival (21
0.38, N186, n.s.), or her fecundity
(r0.10, N186, n.s.).
Although there was no evidence that females mating
with generally attractive males increased their ¢tness
directly, there was evidence of indirect bene¢ts. When
introduced to leks consisting of three sons, virgin females
once again showed clear preferences for particular males
(23
12.7, p50.01). Moreover, in agreement with Fish-
erian models, the mating success of sons increased with
the attractiveness of their fathers (signi¢cance of paternal
group as a non-directional variable in general linear
model: F2,113
6.55, p50.02; ordered heterogeneity test
(Rice & Gaines 1994): rsPc
0.96, p50.01; ¢gure 1b).
Speci¢cally, sons of more attractive fathers obtained far
more than one-third of all the matings in their trial (one-
sample t-test on logistically transformed data: t2.93,
d.f.37, p50.01). Although these results provide impor-
tant support for a Fisherian interpretation of female
preferences, they could conceivably have a good-genes
explanation (with inherited high viability enabling sons of
attractive males to invest heavily in preferred traits).
In an attempt to distinguish between these alternatives,
we compared o¡spring survival to paternal attractiveness.
We found no evidence that a female's mate choice
in£uenced the viability of her o¡spring when sibs were
reared in family groups (for a detailed breakdown, see
table 3). Survival of o¡spring from egg hatch to adulthood
(table 3 and ¢gure 1c), survival across each developmental
stage, and the number of adult o¡spring each female
produced were all independent of paternal mating success.
Likewise, the survival as adults of sons and both reprodu-
cing and non-reproducing daughters did not increase with
sire attractiveness (table 3 and ¢gure 1d). There was a weak
trend for survival to decrease with sire mating success in
the group of daughters that was allowed to reproduce, but
after correction for multiple comparisons (Rice 1989) this
pattern was non-signi¢cant. Finally, the fecundity of
reproducing daughters was again unrelated to their
mothers' choice of mate. These results appear to contradict
key predictions of good-genes models.
4. DISCUSSION
In 1994, Malte Andersson wrote that `no critical test has
been performed that supports Fisherian sexual selection
and excludes the alternatives, or estimates their relative
importance' (Andersson 1994, p.52). Here, we have
attempted to assess the relative importance of the three
main models proposed for the maintenance of female
choice at leks.We found that females mating with generally
attractive males did not obtain any clear-cut direct bene¢ts
from their choice of mate. Similarly, we found no evidence
to suggest that families of o¡spring sired by more attractive
males exhibit higher viability, at any stage of development.
However, we did ¢nd that sons of attractive fathers them-
selves obtained disproportionately more matings. The
evidence presented thus shows that females gain indirect
bene¢ts from mating with particularly attractive males,
and highlights a Fisherian mechanism as being at least in
part responsible for the current maintenance of female
mate choice in L. longipalpis.
Di¡erential maternal investment (a potential problem
elsewhere (Partridge 1980; Simmons 1987)) is unlikely to
confound the observed correlation between the attractive-
ness of fathers and sons. Sand£ies are highly fecund, so
we would expect di¡erences in reproductive e¡ort to be
mediated via variation in egg number rather than quality
(Simmons 1987; Petrie 1994), yet we found no evidence
that females laid more eggs for attractive mates. A more
subtle maternal e¡ect might arise because, although we
allocated some females to groups of generally unattractive
males, all our females chose their mates themselves (cf.
Moore 1994; Petrie 1994). Fisherian models rely on
linkage disequilibrium between genes coding for female
preferences and for preferred traits. Thus maternally
rather than (or as well as) paternally inherited genes
might be responsible for the attractiveness of those sons
whose mothers chose generally attractive mates. Never-
theless, it is hard to explain such a pattern of inheritance
without invoking an underlying Fisherian mechanism. It
would be worthwhile replicating this study using ¢rst-
generation captive stock to rule out the possibility that
the observed heritability of male attractiveness might be
the result of the rapid evolution of trait and preference in
the laboratory (Shelly et al. 1994).
A common criticism of laboratory-based studies of
evolutionary questions is that an inability to detect
variations in ¢tness may be a result of experimental
conditions. Our tests of putative direct and good-genes
bene¢ts are potentially vulnerable to the same criticism.
For example, most direct bene¢ts accrued by choosy
females should be re£ected in increased survival or
immediate fecundity.We found no evidence of this, but it
is conceivable that the enhanced post-oviposition survival
detected here might translate into a direct bene¢t if it
appreciably increases the probability of free-ranging
females entering a second gonotrophic cycle. Likewise,
certain postulated direct bene¢ts of mate choice (such as
avoidance of predators or sexually transmitted diseases)
could not be detected in our system. However, there is
little empirical support for the predator- or disease-avoid-
ance hypotheses in the current lekking literature
(Ho« glund & Alatalo 1995).
Moving on to good-genes bene¢ts, it could be argued
that these do accrue to choosy female sand£ies but were
masked here by the `welfare state' conditions of our
experiment. However, genetic di¡erences in o¡spring
survival are frequently detected in laboratory experiments
(Simmons 1987; Taylor et al. 1987; Reynolds & Gross 1992;
Moore 1994), and importantly in this particular case,
conditions were clearly not favourable enough to mask
genetic di¡erences in the attractiveness of sons. In
addition, we are con¢dent that, despite a large amount of
within-group variation, our methodology was capable of
detecting meaningful di¡erences in o¡spring survival
across treatment groups if these existed. Power analysis
(Cohen 1988) shows that with our sample size and
variation in survival, we had a 95% chance of detecting a
correlation coe¤cient between o¡spring survival and
mating success of as little as 0.26. To put this in context,
the median r-value between measures of o¡spring
survival and sire attractiveness reported by ¢eld studies
claiming good-genes e¡ects is 0.47 (von Schantz et al.
1989; Norris 1993; Petrie 1994; Hasselquist et al. 1996;
Sheldon et al. 1997); the equivalent ¢gure from the
laboratory is 0.65 (Taylor et al. 1987; Reynolds & Gross
1992; Moore 1994). Thus good-genes e¡ects in lekking
L. longipalpis would have to be considerably weaker than
those found in other species to have been missed by our
experimental protocol.
Figure 1. Tests of proposed bene¢ts of mate choice at sand£y leks. (a) Number of eggs laid by females in relation to the attrac-
tiveness of the males they chose. (b) Proportion of matings (mean s.e., calculated using logistically transformed data) achieved
by sons of males of low, average and high attractiveness, competing in three-male leks. (c) Proportional survival of o¡spring of
females in relation to the mating success of their mates. (d) Survival as adults (mean s.e.) of non-reproducing sons and daugh-
ters and reproducing daughters sired by males of low, average and high attractiveness. Detailed statistical breakdowns are given
in tables 2 and 3.
A ¢nal consideration is that larvae in this study only
had the opportunity to compete with full sibs. It is
possible that competition within a family of low-quality
individuals and within a family of high-quality indivi-
duals may yield similar mean survival rates, thereby
obscuring true ¢tness di¡erences: low-quality individuals
may only survive when they have low-quality competi-
tion. However, several studies of other species have shown
good-genes e¡ects despite rearing young alongside their
sibs (Simmons 1987; Norris 1993; Moore 1994). Moreover,
laboratory studies of L. longipalpis suggest that eggs and
larvae are likely to be aggregated in family groups in the
¢eld (Elnaiem & Ward 1992; Dougherty et al. 1993, 1994).
Thus, even if rearing in family groups did mask some
underlying di¡erences in o¡spring competitive ability in
our experiment, such di¡erences might be equally
masked (and unavailable to choosy females) in free-
ranging populations.
In conclusion, we found evidence for Fisherian but no
other detectable mate-choice bene¢ts in L. longipalpis. We
therefore suggest that heritability of male attractiveness
is an important factor maintaining female mating
preferences in this species (although this does not of
course preclude the possibility that other adaptive or
non-adaptive mechanisms were responsible for the origin
of these preferences (Balmford & Read 1991)). Other
studies, particularly those claiming good-genes e¡ects,
have often not even looked for Fisherian bene¢ts, and we
suggest that indirect ¢tness gains via increased attrac-
tiveness of sons may turn out to be more widespread
than generally thought. We believe that future studies
should no longer argue for the importance of particular
mate-choice bene¢ts unless they also test for alternatives.
The approach adopted here, with concurrent tests of all
the main postulated bene¢ts (and with further modi¢ca-
tion of the protocol for assessing o¡spring viability, if
possible using genetic markers) provides a robust way
forward for studies into the maintenance of female
mating preferences.