Download effects of natural and sexual selection on adaptive population

Evolution, 60(6), 2006, pp. 1242–1253
EFFECTS OF NATURAL AND SEXUAL SELECTION ON ADAPTIVE POPULATION
DIVERGENCE AND PREMATING ISOLATION IN A DAMSELFLY
ERIK I. SVENSSON,1 FABRICE EROUKHMANOFF,1,2
AND
MAGNE FRIBERG1,3
1 Section
for Animal Ecology, Ecology Building, Lund University, SE-223 62 Lund, Sweden
E-mail: [email protected]
2 Département de Biochimie-Génie Biologique Ecole Normale Supérieure de Cachan, 61 Avenue du Président Wilson,
94235 Cachan, Cedex, France
E-mail: [email protected]
3 Department of Zoology, Stockholm University, SE-106 91 Stockholm, Sweden
E-mail: [email protected]
Abstract. The relative strength of different types of directional selection has seldom been compared directly in natural
populations. A recent meta-analysis of phenotypic selection studies in natural populations suggested that directional
sexual selection may be stronger in magnitude than directional natural selection, although this pattern may have partly
been confounded by the different time scales over which selection was estimated. Knowledge about the strength of
different types of selection is of general interest for understanding how selective forces affect adaptive population
divergence and how they may influence speciation. We studied divergent selection on morphology in parapatric, natural
damselfly (Calopteryx splendens) populations. Sexual selection was stronger than natural selection measured on the
same traits, irrespective of the time scale over which sexual selection was measured. Visualization of the fitness
surfaces indicated that population divergence in overall morphology is more strongly influenced by divergent sexual
selection rather than natural selection. Courtship success of experimental immigrant males was lower than that of
resident males, indicating incipient sexual isolation between these populations. We conclude that current and strong
sexual selection promotes adaptive population divergence in this species and that premating sexual isolation may have
arisen as a correlated response to divergent sexual selection. Our results highlight the importance of sexual selection,
rather than natural selection in the adaptive radiation of odonates, and supports previous suggestions that divergent
sexual selection promotes speciation in this group.
Key words.
plasticity.
Adaptive ridge, direct selection, divergence-with-gene flow, fitness valley, indirect selection, parapatry,
Received January 24, 2006.
Several meta-analyses of selection studies in natural populations have recently been published (Hoekstra et al. 2001;
Kingsolver et al. 2001; Hereford et al. 2004), and one conclusion from these studies is that the strength of selection on
phenotypic traits may often be strong in natural populations,
although the definition of weak and strong selection is still
subject to discussion (Hereford et al. 2004). These studies
have also suggested that directional sexual selection on phenotypic traits may be stronger than directional natural selection on the same traits (Kingsolver et al. 2001), although the
effect might have arisen because of the different time scales
over which natural and sexual selection were measured
(Hoekstra et al. 2001). One way to resolve this issue would
be to estimate sexual selection gradients over several different time scales (Kingsolver et al. 2001) and compare the
results within traits with the magnitude of the natural selection gradients.
The strength of directional selection on phenotypic traits
will also provide information about how well adapted populations are to their local environments. Strong directional
selection on phenotypic traits (i.e., selection gradients that
are large in magnitude) implies that populations have not yet
reached their adaptive peaks (Coyne et al. 1997; Kingsolver
et al. 2001). Directional selection gradients thus provide information about two aspects of the adaptive landscape: the
location of the population mean in relation to the adaptive
peak and the steepness of the slope on the landscape (Kingsolver et al. 2001; Hereford et al. 2004). Other aspects of
landscape structure, such as its curvature and the depth of
Accepted April 1, 2006.
fitness valleys require visualization of the adaptive surfaces
(Phillips and Arnold 1989; Schluter and Nychka 1994; Fear
and Price 1998; Arnold et al. 2001). Visualization of adaptive
surfaces separately for different fitness components, such as
survival and mating success may also throw light on the
relative importance of natural and sexual selection in population divergence and speciation (Schluter 2000; Gavrilets
2004). Traditionally, there has been a focus on fitness valleys
between populations and the presence of multiple fitness
peaks (Whitlock et al. 1995), but more recently there has
been an increased attention to the possible existence of adaptive ridges in genotypic or phenotypic space (Schluter 2000;
Gavrilets 2004). Adaptive ridges and other flat regions in the
fitness landscape indicate that several alternative phenotypes
have approximately equal fitness and would suggest that populations may diverge by stochastic forces in addition to the
well-recognized deterministic forces of natural and sexual
selection (Schluter 2000; Gavrilets 2004).
The strength of divergent selection in natural populations
is of interest because of its potential role in generating reproductive isolation between populations (Rice and Hostert
1993; Schluter 2000). According to one common view of the
speciation process, reproductive isolation between populations evolves as a correlated response to divergent natural
selection in different environmental settings (Schluter 2000).
This view is supported by many laboratory experiments on
speciation (Rice and Hostert 1993). Moreover, both phenotypic divergence and reproductive isolation may evolve in
spite of extensive and ongoing gene flow between parapatric
1242
q 2006 The Society for the Study of Evolution. All rights reserved.
SELECTION AND POPULATION DIVERGENCE
populations (Endler 1977; Smith et al. 1997; Hendry et al.
2000; Gavrilets 2004; Coyne and Orr 2004). However, gene
flow constrains adaptive divergence, and the resulting population differences will therefore reflect the balance between
the diversifying effects of natural selection and the homogenizing effects of gene flow (King and Lawson 1995; Hendry
et al. 2002; Hendry and Taylor 2004; Nosil and Crespi 2004).
Although studies of natural and sexual selection have often
been performed in such settings (Kingsolver et al. 2001), they
are seldom integrated with studies on reproductive isolation
(Coyne and Orr 2004). Thus, these different research traditions thus often remain separated from each other.
In this study we present empirical data that should be relevant to the resolution of some of the issues above. Our first
goal is to provide a comparison between the strength of different types of selection. We have studied natural and sexual
selection on morphology in field populations of the bluebanded demoiselle (Calopteryx splendens), an insect with pronounced population divergence in several morphological
traits (Svensson et al. 2004). We proceed by visualizing the
fitness surfaces for natural and sexual selection for two populations, to further investigate the relative importance of natural and sexual selection in morphological population divergence. Finally, we investigate the strength of premating
sexual isolation between several populations of this species.
Results in this study support previous suggestions from comparative studies (McPeek and Brown 2000) that speciation
and adaptive radiation in odonates is primarily driven by
sexual selection and that natural selection has a less pronounced role, compared to its role in other taxa and in other
adaptive radiations (Schluter 2000).
MATERIAL
AND
METHODS
Study Organism, Ecology and Study Populations
Calopteryx splendens is a common damselfly along slow
flowing streams in Europe (Askew 1988). It emerges as an
adult from the larval stage in late May or early June in southern Sweden (Svensson et al. 2004). After reaching sexual
maturity, males and females return to the shores of the water
where mating and oviposition takes place (Corbet 1999).
Males are territorial and polygynous, and may mate with
several females during their life span, although many males
are much less successful and obtain no matings (Corbet
1999). This generates large variance in male mating success,
making this species an excellent model organism for studying
sexual selection in the wild (Svensson et al. 2004).
Previous studies on sexual selection in the genus Calopteryx have revealed that the melanized dark wing patches of
males function in mate choice as species recognition characters and in both con- and hetero-specific male-male competition (Waage 1979; Siva-Jothy 1999; Svensson et al. 2004;
Tynkkynen et al. 2004). The size of the black wing patches
is correlated with male immunological condition, because the
melanin-producing enzyme phenoloxidase (PO) is involved
in the development of melanin as well as having an important
function in insect immune defence (Siva-Jothy 2000). Recent
studies of northern European populations have revealed that
the size of the black wing patch in C. splendens males is
negatively correlated with the relative occurrence of the con-
1243
generic C. virgo, which has entirely black wings (Tynkkynen
et al. 2004). This is partly a result of interspecific antagonistic
male-male interactions in sympatry: here small-patched C.
splendens males are favored because they are attacked less
frequently by C. virgo males (Tynkkynen et al. 2004, 2005).
Finally, comparisons of wing remnants of C. splendens males
that were killed by bird predators (mainly pied wagtails Motacilla alba) have revealed that natural selection acts on wing
length and wing width (E. Svensson and M. Friberg, unpubl.
data). Thus, multiple selection pressures are acting on the
morphological traits of C. splendens, and evidence for divergent natural and sexual selection has recently been presented in south Swedish populations (Svensson et al. 2004).
In southern Sweden, populations of C. splendens have diverged extensively in their morphological traits, including
the melanized wing patch, in spite of gene flow between
populations (average FST 5 0.05; range ; 0–0.13; Svensson
et al. 2004). A significant correlation (r 5 0.39: P 5 0.005)
between the phenotypic and molecular-genetic distances of
populations (Svensson et al. 2004) indicates that adaptive
population divergence in this species is partly constrained by
gene flow (Hendry et al. 2002). Based on these values of FST
(Svensson et al. 2004), the estimated number of migrants
(Nm) between our populations range from 1.7 to 919.4 (mean
5 31.00 6 15.75). In addition, natural and sexual selection
differ in magnitude and direction between closely located
populations (Svensson et al. 2004; Tynkkynen et al. 2004),
and these multiple features of C. splendens ecology and genetics therefore make it a suitable model species to investigate the interactions between divergent selection, gene flow,
and morphological divergence.
Field Study of Natural and Sexual Selection
We studied natural and sexual selection in two populations
during the summers of 2001–2003 (Klingavälsån) and 2002–
2003 (Höje Å) outside Lund, southern Sweden. During these
summers, we captured a total of 1645 males at Klingavälsån
(2001: 374; 2002: 486; 2003: 785) and 1445 males at Höje
Å (2002: 453; 2003: 992). We measured 12 morphological
traits on these males (Tables 1–2). Because recent meta-analyses (Kingsolver et al. 2001; Hereford et al. 2004) recommend that future selection studies present mean trait values
and their phenotypic variances and covariances, we present
them in Appendices 1 and 2, respectively.
All captured males were individually marked with unique
three-digit color codes on the last three abdominal segments
(Svensson et al. 2004). Marked males were released in the
field again after they were measured, except for those that
were included in reciprocal transplant experiments (see below). Of the males released locally, a total of 612 (Klingavälsån) and 571 males (Höje Å) were later observed at least
once in the field. We base all of our estimates of natural and
sexual selection on this sample of resighted males. Emigration of males to other sites is unlikely to substantially bias
our selection estimates, because mature males that have been
resighted once usually remain resident at their streams, and
both populations are surrounded by inhospitable habitat of
agricultural fields without suitable aquatic habitats for Calopteryx. Although we have marked and released several thou-
1244
ERIK I. SVENSSON ET AL.
sands of mature Calopteryx males at both Klingavälsån and
Höje Å, not a single male (apart from our experimental immigrant males; see further below) has been observed to migrate between Klingavälsån and Höje Å, which are located
about 30 km from each other. This strongly suggests that
dispersal is indeed limited to other phases of the life cycle,
that is, immediately after emergence from the larval stage.
Therefore, our estimates of local survival at both sites are
likely to reflect true survival and are unlikely to be biased
by emigration between sites.
We performed daily field work during June–August to
quantify male life span and mating success in the wild and
relate this to variation in morphological traits (Svensson et
al 2004). Field work in all seasons continued until all marked
males disappeared from each site. In this study, we present
selection gradients based both on a long-time scale (days)
and a shorter time scale (minutes and hours, from time budget
data) to investigate the role of time interval over which sexual
selection is measured. Long-term sexual selection estimates
were obtained by dividing the number of observed copulations for each male by the number of days he was observed
in the field (life span), whereas short-term sexual selection
estimates were obtained by dividing the total number of copulations for each male by the time he was actually observed
in the field during time budget studies (mean minutes observed 5 15.84 60.61; n 5 1153 males). Both types of sexual
selection estimates are thus measures of ‘‘mating rate’’ (i.e.,
number of copulations per unit time) and these estimates are
therefore not inflated by differences in life span between
males, which could potentially confound the effects of natural
and sexual selection (Hamon 2005; Hamon and Foote 2005).
Both long-term and short-term estimates of male mating rates
are from 2001–2003 (Klingavälsån) and 2002–2003 (Höje
Å).
These two sites were chosen as target populations because
they differ ecologically, which should maximize the power
to detect divergent selection. At Klingavälsån, C. splendens
are sympatric with C. virgo, and sexual selection does at least
in some years favor smaller wing patches in the males (Svensson et al. 2004). At Höje Å, C. virgo is absent and here sexual
selection favors large wing patches (Svensson et al. 2004).
These two populations also differ markedly in the intensity
of avian predation: at Klingavälsån, avian predation pressure
from pied wagtails (Motacilla alba) is high, judged by the
extensive occurrence of wing remnants at ‘‘slaughter stations.’’ In contrast, at Höje Å predation from wagtails is less
intense and wagtails are observed much less frequently than
at Klingavälsån (mean number of wagtails observed per day:
Klingavälsån: 4.92 60.70; Höje Å: 0.52 6022; paired t-test
comparing the localities with respect to the number of birds
observed each day: t48 5 5.726, P , 0.001). We emphasize
that this almost tenfold difference in wagtail occurrence could
potentially lead to divergent natural selection through mortality effects on the males. Higher wagtail occurrence at Klingavälsån could also indirectly fuel divergent sexual selection
by affecting operational sex ratios and male or female mating
behaviors.
Survival in a New Environment: Reciprocal
Transplant Experiments
In 2002 and 2003, we performed reciprocal transplant experiments between Klingavälsån and Höje Å to estimate survival of immigrant and resident males in their native and
novel habitats. A total of 507 individually marked males were
transported either from Höje Å to Klingavälsån (2002: 116;
2003; 137) or from Klingavälsån to Höje Å (2002: 122; 2003:
132). These experimental ‘‘immigrants’’ were released in
their new populations, whereafter we estimated their survival
through the regular field observations.
Handling and transportation of the experimental immigrant
males took about three hours. To investigate whether there
could be any negative effects of this extra handling time of
experimental males, we kept some resident males as
‘‘shams’’ (n 5 392) for three hours and compared their survival with resident males that we released immediately after
marking (‘‘controls’’; n 5 1747). We found no difference
between these two male categories in establishment success,
or in the average life span in the field of males that were
established and resighted (GLZ-models: P . 0.10 in both
populations and during both years). We therefore conclude
that the extra handling time of the experimental immigrant
males did not have any confounding effect on their survival
in the novel habitat.
Courtship Success of Resident and Immigrant Males
We presented resident and immigrant males (n 5 240
males) to local females at seven of our 12 study populations
(Svensson et al. 2004). These experiments were performed
to quantify male courtship success towards local females and
thereby estimate the degree of premating sexual isolation
between populations. The seven study populations included
Klingavälsån and Höje Å (see above), which were also part
of a reciprocal experiment in both directions. The other five
populations were Omma, Världs Ände, Rosendala, Gemla,
and Härnäs (Svensson et al. 2004).
We used a 0.5 m long thread to tie males at the thorax,
without binding their legs, and tied the other end of the thread
to a 1.5 m long bamboo stick. Tied males were thereafter
used during presentation sessions that lasted for 10–30 min.
During these sessions, the tied male was put in close proximity to target females. Each female was mature and resting
on vegetation when the presentations started. The tests were
performed during midday (10:00 to 16:00), in calm and sunny
weather conditions during 2003 and 2004.
We quantified the females’ sexual responses using an 11degree nominal scale ranging from 0 (female attacks the
male) toward 10 (tandem formation and/or successful copulation of the male). The scale (for details, see Supplementary
Material 1 available online at: http://dx.doi.org/10.1554/06036.1.s1) is similar to the approach used by previous workers
investigating sexual isolation in damselflies (Paulson 1974;
Waage 1975). In this study, we included several additional
steps to obtain a fine-grained scale that included all the discrete and well-defined precopulatory behaviors that have been
described in the genus Calopteryx (Corbet 1999).
We performed 2258 presentations (of 160 resident males)
toward females during the summers 2003 and 2004 (mean
SELECTION AND POPULATION DIVERGENCE
number presentations per resident male 5 14.11; SE 5 4.22).
To avoid pseudoreplication due to multiple presentations to
multiple females by each male, the average female response
scores to each male were used. These mean response scores
were approximately normally distributed (Supplementary
Material 1 available online), justifying the use of parametric
statistical tests. The average female response towards males
was low, and hence most male approaches were rejected by
females, reflecting the general mating biology of this species.
Male mating harassment of females is frequent in damselflies
(Corbet 1999), and many, perhaps most, male mating attempts by males are rejected by females in the field (E. I.
Svensson, unpubl. data). Hence, our female response scale
(Supplementary Material 1 available online) of male courtship success should be interpreted both as female propensity
to mate as well as indicating female resistance toward unwanted mating attempts from males. Female response scores
can also be viewed as a form of proximity score to a given
male, because only males that were permitted to approach a
female at very close distance were subsequently able to contact and obtain tandem position (Supplementary Material 1
available online).
Statistical Analyses
We analyzed data using general linear models (GLMs) and
generalized linear models (GLZs) using the STATISTICA
software package (Statsoft 2003). Standardized linear selection gradients for all 12 traits were quantified for both Klingavälsån (2001, 2002, and 2003) and Höje Å (2002 and
2003). We thus regressed relative fitness against standardized
trait values (mean zero; unit variance) of all 12 traits and
estimated the standardized selection gradients as the partial
regression coefficients for each trait in multiple regression
models (Lande and Arnold 1983; Svensson et al. 2004). Standardized linear selection gradients of this kind separate direct
selection on a trait from the fitness effects of other correlated
traits. Standardized selection differentials and standardized
selection gradients are highly correlated across studies (Kingsolver et al. 2001) and this was the case in our study (E. I.
Svensson, unpubl. data). All our results and conclusions hold
for both standardized selection gradients (the metric presented in this paper) and standardized selection differentials.
The relationships between standardized selection differentials, standardized selection gradients and curvilinear selection coefficients (stabilizing and disruptive selection) will be
presented and discussed in detail in a future paper. In this
study, we used two different fitness components to estimate
linear selection gradients: minimum life span in the field (last
date observed minus date of marking; a measure of survival
and hence natural selection) and mating rate (number copulations/life span; a measure of sexual selection).
Selection gradients were initially estimated as the parametric slopes of fitness regressed againt trait values, using
ordinary GLMs. All significant estimates were subsequently
checked using GLZ (poisson error, logit link, type-3 errors)
because mating rate was highly skewed, with a few males
obtaining most matings. We included the year (2002 and
2003) as a main effect to take into account annual variation
when estimating overall selection gradients. We did not in-
1245
clude two-way interaction terms between year and traits, because such models would become extremely large and complex (13 main terms and 12 two-way interactions). Interaction
terms involving year are also of questionable biological significance in this system. Year effects mainly result from differences in visibility of the damselflies to observers in different years, that is, the available observation time and probability of detecting male copulations in the field.
To investigate natural and sexual selection on overall morphology in Klingavälsån and Höje Å, we performed a principal component analysis on all 12 traits for both populations
combined. Because the different traits are not necessarily
diverging independently due to trait correlations, the use of
principal components should overcome such statistical difficulties arising from multicollinearity. We employed the
nonparametric projection pursuit regression (PPR) approach
on the first two principal components (PC1 and PC2) and
visualized the three-dimensional fitness surfaces using multivariate spline surfaces (Schluter and Nychka 1994). Prior
to the analyses, we performed grid search to find the best
value of the smoothing parameters (l) that minimized the
general cross-validation (GCV) score. We subsequently used
the obtained value of the smoothing parameter when estimating the final surface (Schluter and Nychka 1994). Spline
surfaces were visualized using the maximum stiffness-option
in the ‘‘Surface Plot’’ module in STATISTICA.
In these PPR regressions, we searched through the fitness
surfaces in the PC1/PC2 plane in 5000 random directions to
find the two directions (a1 and a2) that showed the maximum
fitness increase. Standard errors and associated significance
levels for a1 and a2 were obtained through 100 bootstrap
replications of the data. The two directions were subsequently
imposed on the fitness surfaces to visualize multivariate selection (Schluter and Nychka 1994).
All other statistical results in this study come from ordinary
parametric tests (ANOVAs, t-tests, repeated-measures analyses) and all P-values refer to two-tailed tests. Standard errors
are presented whenever means are presented. All relevant
variables were checked for the assumption of normally distributed residuals and equal variances. Selection gradients
differ somewhat from previous estimates from these populations, because we included one additional year (2003), included more traits (12 instead of 2–4) and also estimated
sexual selection gradients on a longer time scale (days), compared to our previous study (Svensson et al. 2004). In addition, we used a logit link function in the GLZ-models instead of the probit link function used in our earlier analyses
(Svensson et al. 2004).
Results from logit models and the parameters obtained
from ordinary parametric regression analysis were very similar or identical, and we only present the estimates from the
parametric models. The use of logistic regression has been
suggested as an alternative to parametric regression, but this
approach is limited to situations when the dependent fitness
variable has an dichotomous outcome (e.g., death or survival)
and these parameters cannot be directly incorporated into
models for predicting microevolutionary change, but have to
be translated into linear coefficients (Janzen and Stern 1998).
In contrast, parameters obtained from parametric regression
1246
ERIK I. SVENSSON ET AL.
TABLE 1. Factor loadings and associated test statistics (eigenvalues) for the principal component analysis of the 12 traits included
in the fitness surfaces (Fig. 2; Table 2). Shown are the first two
principal components, which jointly explained 70.74% of the total
morphological variation. Trait abbreviations as in Appendix 1.
FIG. 1. Natural selection gradients and sexual selection gradients
(absolute values zbz: means 6 95 % CI) on morphological traits
across three years (2001–2003) and two populations (Klingavälsån
and Höje Å) in Calopteryx splendens.
analyses can be directly incorporated in such models (Lande
and Arnold 1983).
RESULTS
Comparison of selection gradients measured in our two
different populations over several seasons revealed that sexual selection was indeed stronger than natural selection across
all traits, and the difference between the three selection categories was highly significant (Fig. 1; repeated measures ANOVA: effects of selection category: F2,116 5 23.974, P ,
0.001; population: F1,58 5 0.335, P 5 0.56; selection category
3 population: F2,116 5 0.569, P 5 0.57). Moreover, sexual
selection was stronger than natural selection, irrespective of
the time-scale over which sexual selection was estimated
(minutes/hours or days; Fig. 1), and both sexual selection
categories are higher in magnitude than the natural selection
category (Tukey HSD posthoc tests; P , 0.001). In contrast,
there was no significant difference between the two sexual
selection categories (Tukey HSD posthoc tests; P 5 0.13).
Hence, sexual selection was stronger than natural selection
on these traits, independent of the time scale over which
sexual selection was estimated (Hoekstra et al. 2001; Kingsolver
et al. 2001). Natural and sexual selection gradients for all years
and populations are provided in Supplementary Material 2 available online at: http://dx.doi.org/10.1554/06-036.1.s2), to facilitate future meta-analyzes (Kingsolver et al. 2001).
In the reciprocal transplant experiment, aimed to investigate survival of immigrant and resident males, we found no
evidence for an interaction between phenotype origin and
population (log likelihood 5 25864.69; x2 5 1.020; P 5
0.31). Longevity in the field was similar for both resident
and immigrant males at both Höje Å (mean field life span in
days: 3.83 60.21 for residents and 4.47 60.44 for immigrants) and Klingavälsån (mean field life-span: 4.81 60.23
for residents and 4.71 60.50 for immigrants).
The first two principal components explained 46.42%
Trait
PC1
PC2
PCS
THW
THL
FWL
FWW
FPL
HWL
HWW
HPL
TOL
ABL
MASS
Eigenvalue
% explained variation
0.008
0.008
0.022
0.147
0.035
0.284
0.143
0.032
0.261
0.184
0.156
0.002
3.971
46.42
20.008
0.025
0.003
0.171
0.043
20.328
0.160
0.036
20.284
0.365
0.318
0.001
2.081
24.32
(PC1) and 24.32% (PC2) of the total morphological variation,
respectively (Table 1). Visualizing the fitness surface for natural selection (life span) and sexual selection (mating rate)
against PC1 and PC2 revealed striking differences between
the surfaces for survival and mating rate. The fitness surfaces
for survival (Fig. 2A, B) were rather flat for both Höje Å and
Klingavälsån. These flat fitness surfaces lacked pronounced
curvature or clear fitness peaks, indicating that variation in
morphology had small consequences for survival. This confirms our previous conclusion from the comparison between
selection gradients (Fig. 1) that natural selection on these
morphological traits is weak to moderate. The general impression of flat fitness surfaces and weak natural selection
on morphology was also confirmed by the results from the
PPR analyses (Table 2). Out of the eight PPR coefficients
for how survival was affected by PC1 and PC2, only one was
significant (Table 2).
In striking contrast to these rather flat fitness surfaces, we
found evidence for adaptive surfaces with considerably more
curvature and with clear fitness peaks when we visualized
variation in mating rate in relation to variation in the two
principal components (Fig. 2C, D). These surfaces indicated
that sexual selection in the two different populations promoted divergent selection on morphology, because the two
population peaks were located in different regions of the PC1/
PC2-plane (Fig. 2C, D). Note that the extraction of these
principal components was performed on the total sample of
all males from both populations, hence these fitness surfaces
are directly comparable between the two populations. The
presence of two fitness optima with different locations in the
PC1/PC2 plane was confirmed in the PPR analysis (Table 2).
The PPR analysis and the spline surfaces indicated the presence of a fitness peak in a region of low values of PC1 and
PC2 at Klingavälsån (Fig. 2C) and another peak at intermediate values of PC1 and high values of PC2 at Höje Å
(Fig. 2D). It is particularly notable that the two fitness surfaces were similar in shape and structure but differed in orientation of the peaks in relation to PC1 and PC2: The fitness
peaks and the directions of maximum fitness increase were
almost orthogonally oriented toward each other (Table 2).
FIG. 2. Fitness surfaces (splines) for survival and mating rate in relation to the first two principal components of 12 morphological traits (PC1 and PC2) for males at
Klingavälsån (A, C) and Höje Å (B, D). The three-dimensional surface plots are shown in the left panel and the two-dimensional surface plots with individual data points
are shown in the right panel. The three-dimensional surface plots have been rotated for visual clarity. The location of each population (mean phenotypic value) is indicated
in each panel by circles. Although the two populations may appear close to each other, they differ significantly in the morphological traits (Appendix 1), as well as in their
mean values of PC1 and PC2 (Wilk’s l 5 0.992; F2,1019 5 3.945; P 5 0.020). Shown in the two-dimensional plots are also the two directions for maximum increase in fitness
(a1 and a2), obtained from projection pursuit regression analysis (Table 2). Solid line, both coefficients for PC1 and PC2 significant; dashed line, one of the coefficients
significant; dotted line, none of the coefficients significant. The intersection of the two fitness directions in each graph coincides with the data point with the highest estimated
(‘‘Y-hat’’) fitness value. Note differences in scale on vertical axis between fitness surfaces for survival and mating rate.
SELECTION AND POPULATION DIVERGENCE
1247
1248
ERIK I. SVENSSON ET AL.
TABLE 2. Coefficients for the directions of maximum fitness increase in the PC1/PC2-plane (Fig. 2). Coefficients were obtained from
projection pursuit regression (Schluter and Nychka 1994), and their standard errors, in parentheses, were obtained through 100 bootstrap
replications. Significant coefficients are indicated in bold.
Fitness
component
Survival
Population
Klingavälsån
Höje Å
Mating rate
Klingavälsån
Höje Å
Principal
component
PC1
PC2
PC1
PC2
PC1
PC2
PC1
PC2
One reflection of this orthogonal direction of the fitness surfaces was that three of the four possible PPR coefficients
differed in sign between the populations (Table 2). Five of
eight of these PPR coefficients were significant for these
sexual selection surfaces, in striking contrast to only one of
eight for the natural selection surfaces (Table 2 and above).
At Klingavälsån, sexual selection appears to favor damselflies with shorter wings, shorter abdomen, and shorter total
length than at Höje Å, where instead larger damselflies were
favored (Fig. 2C,D, Table 1). There also appears to be a
tendency for sexual selection towards shorter forewing and
hindwing patch lengths at Klingavälsån, although this pattern
was less pronounced than for the other traits (Fig. 2C; Table
1). These results are of interest given the sympatric occurrence of C. virgo at Klingavälsån, and are partly consistent
with results from Finnish C. splendens populations occuring
in allopatry and sympatry with C. virgo (Tynkkynen et al.
2004, 2005).
To experimentally investigate the magnitude of sexual isolation between populations, we performed presentation experiments in the field in which we presented resident and
immigrant males to local females. Across all of our popu-
Direction a2
Direction a1
20.034
0.681
0.471
20.568
20.462
0.830
0.660
20.202
(0.636)
(0.362)
(0.486)
(0.467)
(0.211)*
(0.232)***
(0.329)*
(0.644)
0.382
20.672
0.066
0.671
0.411
0.599
20.303
0.651
(0.502)
(0.387)
(0.671)
(0.310)*
(0.194)*
(0.659)
(0.611)
(0.332)*
lations, there was a highly significant difference in male
courtship success between native and immigrant males, demonstrating significant sexual selection against migrants and
premating isolation (Fig. 3; paired t-test: t6 5 3.162, P 5
0.020). In six of seven cases, immigrant males had lower
courtship success than resident males, whereas in the remaining case immigrants had (nonsignificantly) higher success (Fig. 3). The average success of immigrant males compared to resident males, measured across all populations was
0.85 (SE 5 0.046).
These data suggest that sexual selection plays an important
role in population divergence and that divergent sexual selection has caused incipient sexual isolation between these
populations. In our two most intensively studied populations
(Klingavälsån and Höje Å), we analyzed the reciprocal effect
of presenting males of each category in both populations.
Here we found evidence for a significant population interaction (F1,86 5 13.878, P , 0.001; Fig. 4) between male
category (immigrant or resident) and male source population
(Klingavälsån or Höje Å). This interaction effect implies that
sexual selection is divergent between the two populations.
DISCUSSION
FIG. 3. Differences (means 695 % CI) in male courtship success
(female response; y-axis) between resident males and experimental
‘‘immigrant’’ males in seven different populations of Calopteryx
splendens. Female courtship response towards residents and immigrants was quantified on a nominal scale using well-described
and distinct precopulatory behaviors (Supplementary Material 1
available online). Each male was presented to several different females in tethering experiments in the field, and the average female
response was used as a measure of male courtship success.
Our data suggest that directional sexual selection arising
from mating success is stronger than directional natural selection arising from mortality, at least when measured during
the adult part of the life cycle (Figs. 1 and 2). Furthermore,
this is not simply an artifact of the different time intervals
over which sexual selection is estimated (Hoekstra et al.
2001; Kingsolver et al. 2001). Directional selection has been
suggested to be the primary cause of phenotypic diversification, based on a recent study on the sign of genetic differences between closely related taxa (Rieseberg et al. 2002),
but such indirect inferences can never distinguish between
the relative importance of different forms of selection. Therefore, field studies of natural and sexual selection are necessary complements to indirect inferences about selection from
molecular data (Schluter 2000; Merilä and Crnokrak 2001;
Svensson and Sinervo 2004; Svensson et al. 2004). The typical magnitude of sexual selection gradients in this study
(median zbz 5 0.25; mean 5 0.37 6 0.04; Fig. 1) is almost
40% higher than than the typical sexual selection gradient
reported in Kingsolver et al.’s meta-analysis (median zbz 5
0.18). In contrast, the magnitude of natural selection gradients in our study (median zbz 5 0.08; mean 5 0.11 6 0.01;
SELECTION AND POPULATION DIVERGENCE
FIG. 4. A sexually selected fitness trade-off between two Calopteryx splendens populations differing in morphology and ecology.
Resident males have higher courtship success towards females than
experimental immigrant males, resulting in a significant male category 3 source population interaction (see Results).
Fig. 1) is closer to the typical natural selection gradient in
Kingsolver et al.’s (2001) study (median zbz 5 0.09).
These results and conclusions from the selection gradient
analyses apply only to directional forms of selection. However, our fitness surface analysis of how variation in morphology (PC1 and PC2) affected survival and mating rate
confirmed this picture and revealed considerably stronger
sexual than natural selection on the principal components
(Fig. 2; Table 2). The steep slopes toward the fitness peaks
on the sexual selection landscapes suggest that directional
sexual selection is considerably stronger than directional natural selection (Fig. 2; Table 2). It is interesting that in both
populations that we investigated, the optimal phenotypes favored by sexual selection were at the extreme ends of the
population distributions. This indicates that the average male
phenotype in both populations is quite maladapted with respect to the morphology that would maximize male mating
rate, leading to strong directional sexual selection (Figs. 1,
2C, 2D). The explanations for this could be due to several
different biological mechanisms, such as costs of developing
secondary sexual traits (Kotiaho 2001), sensory bias in local
females favoring male phenotypes outside the current population range (Fuller et al. 2005), strong condition-dependence and high mutational input for secondary sexual traits
(Rowe and Houle 1996), or maladaptive gene flow displacing
populations from their optima (Kirkpatrick 1996). Although
these explanations may all contribute to explain the presence
of strong directional selection on natural populations, none
of them is likely to exclusively explain the patterns in our
data. For instance, gene flow alone could not explain why
our populations are more maladapted with respect to the sexual selection optima than the natural selection optima, although gene flow is a general explanation for the displacement of populations from their adaptive peaks (Coyne et al.
1997). Survival costs of secondary sexual traits is a possibility, but clear fitness peaks for natural selection are lacking
in this study that would support such an explanation.
1249
The fitness landscapes for natural selection were considerably flatter, and variation in morphology thus appeared to
be only weakly correlated with survival (Fig. 2A,B; Table
2). The populations are located in a region with relatively
weak survival selection on morphology, that is, on a flat
fitness landscape (Fig. 2A, B; Table 2). This suggests that
both populations are relatively well adapted to their local
environments and that divergent natural selection is weak. In
contrast, the similar analysis for sexual selection revealed
evidence for divergent selection between populations and fitness landscapes with considerably more curvature (Fig. 2C,
D; Table 2). Visualizing fitness landscapes separately for
different fitness components may thus be rewarding and
should be a useful complement to selection gradient analyses
(Gavrilets 2004). More generally, visualization of fitness
landscapes for two or more traits has a strong tradition in
evolutionary biology (Arnold et al. 2001), and it is increasingly being used in empirical studies to investigate the ecology of fitness valleys and divergent selection between populations (Svensson and Sinervo 2000; Benkman et al. 2003;
Benkman 2003).
Our data suggest that the morphological divergence between populations at Klingavälsån and Höje Å (Appendix 1,
Table 1) is maintained and promoted by strong divergent
sexual selection rather than natural selection (Fig. 2). The
extent to which these morphological differences are heritable
is unknown. We also note that both recent theory and empirical data point to an important role for adaptive plasticity,
particularly in the early stages of divergence (Losos et al.
2000; Badyaev et al. 2002; West-Eberhard 2003; Price et al.
2003).
Our reciprocal transplant experiment revealed no evidence
for lower survival of immigrant males compared to resident
males (see Results). The lack of experimental support for a
strong role for divergent natural selection in this species contrasts markedly with most previous reciprocal transplant experiments performed in both plants and animals (Schluter
2000; Kawecki and Ebert 2004; Nosil et al. 2005), although
it is possible that these previous results are biased towards
systems where divergent natural selection is predicted to be
strong. Based on these previous experimental studies, it has
been suggested that fitness trade-offs between environments
usually result from divergent natural selection (Schluter
2000). However, reciprocal transplant experiments of this
kind are expected to reveal divergent natural selection only
if populations are located on different adaptive peaks separated by a fitness valley (Schluter 2000). In contrast, if populations are located in a trait region connected by phenotypic
intermediates with approximately equal fitness (cf. Fig. 2A,
B) or on an ‘‘adaptive ridge’’ (Schluter 2000), the pattern of
lower immigrant fitness and such phenotype 3 population
interactions are not expected. Thus, the lack of any effect of
our reciprocal transplant experiment on immigrant survival
and the flat fitness surfaces for our natural selection data (Fig.
2A, 2B) provide independent lines of evidence indicating
weak divergent natural selection on morphology in this species.
The data in this study instead suggest the presence of a
fitness trade-off between environments that is caused by divergent sexual selection. Sexual selection on morphology is
1250
ERIK I. SVENSSON ET AL.
divergent between our two populations (Fig. 2B) and females
apparently also prefer local males over immigrant males
(Figs. 3, 4). The magnitude of the sexually selected fitness
trade-off between our populations can be characterized by
the proportion of nonerror variance that is attributable the
interaction term between phenotype and population, following the approach used by Schluter (2000) in his meta-analysis
of published reciprocal transplant experiments. In this study,
the proportion of nonerror variance explained by the interaction (Fig. 4) was about 0.84, which is considerably higher
than the more typical value of 0.50 in the experiments reviewed by Schluter (2000).
This pronounced between-environment trade-off in male
courtship success is of general interest for at least two reasons. First, the fact that females prefer local males over immigrant males indicates some degree of premating sexual
isolation, which could potentially act as a barrier against gene
flow between these populations. This is particularly notable
in light of some degree of gene flow between these same
populations (Svensson et al. 2004). Second, female preference over local rather than foreign males appears to contradict
predictions from at least one model of sexual conflict:
‘‘chase-away sexual selection’’ (Holland and Rice 1998). According to this model of the coevolutionary arms race between the sexes, females should be more susceptible towards
sensory manipulation from males from other populations or
species, and hence we would have expected local females to
show the strongest preference for the immigrant males. Correlative (nonexperimental) data from some previous studies
of birds also suggest stronger female preference for local,
rather than immigrant males (Bensch et al. 1998), as do laboratory experiments on sexual isolation in some species of
fish (Wong et al. 2004; Boughman et al. 2005).
Both observational and experimental data in this study suggest that sexual selection may be more important than natural
selection in this species, at least during the adult part of the
life cycle and in the early stages of evolutionary divergence.
The coexistence of multiple but ecologically similar damselfly species has previously been used as an argument that
evolutionary divergence in damselflies is primarily driven by
sexual selection (McPeek and Brown 2000). Sympatric coexisting damselfly species often differ strikingly in coloration
traits, but less so in ecological traits (McPeek and Brown
2000). An important role for sexual selection in the evolutionary divergence of this group is supported by the observational and experimental field data in this study. In particular, the finding of both strong sexual selection on morphology and sexual isolation between populations indicate
the potential for sexual isolation arising between populations
as a by-product of divergent sexual selection.
The relative importance of natural and sexual selection in
speciation processes has been subject to several recent reviews, but a paucity of empirical data preclude any general
conclusions (Schluter 2000; Coyne and Orr 2004). An important role of natural selection is universally acknowledged
in adaptive radiations (Schluter 2001), but the role of sexual
selection remains more controversial (Coyne and Orr 2004).
Closely related species do often differ more markedly in male
secondary sexual characters than in other traits (Darwin 1871;
West-Eberhard 1983; Price 1998; Panhuis et al. 2001), which
suggest an important role for sexual selection in evolutionary
divergence. However, these interspecific differences may
have arisen secondarily, after speciation was already accomplished (Butlin 1987, 1989), making conclusions from comparative studies open to alternative interpretations. Furthermore, divergent sexual selection is supposed to result in incipient species that are only weakly ecologically differentiated, in contrast to divergent natural selection (Coyne and
Orr 2004). How such ecologically similar incipient species
formed by sexual selection could coexist and be able to
evolve reproductive isolation before extinction is a problem
for speciation models based solely on sexual selection (Schluter 2000; Coyne and Orr 2004).
Reproductive isolation readily evolves in allopatry as a
correlated response to divergent natural selection (Rice and
Hostert 1993; Schluter 2000). There is accumulating empirical evidence for the evolution of reproductive isolation as a
correlated response to divergent natural selection (Jiggins et
al. 2001; Nosil et al. 2002). In theory, divergent sexual selection could also promote the development of sexual isolation between populations (Lande 1982; Iwasa and Pomiankowski 1995; Day 2000), although the empirical evidence
for this effect in nature is much more limited than for natural
selection (Seehausen et al. 1997). In C. splendens, divergent
sexual selection between parapatric populations may have
resulted in partial sexual isolation between populations as a
correlated response (Figs. 2B, 3, 4). Alternatively, such premating isolation may result from either direct selection on
female mate preferences (Servedio 2001, 2004) or indirect
selection through so-called ‘‘good genes’’ (Kirkpatrick and
Barton 1997). Although our data does not permit us to conclusively confirm or reject any of these possibilities, recent
theory and empirical evidence strongly suggest that indirect
selection is a relatively weak force in the evolution of mate
preferences (Kirkpatrick and Barton 1997; Möller and Alatalo
1999; Arnqvist and Kirkpatrick 2005; Orteiza et al. 2005).
Because the magnitude of indirect fitness benefits of female
mate choice is usually small and on the order of only a few
percent (Möller and Alatalo 1999), the strong sexual selection
coefficient (20.15) against immigrant males in this study
(Fig. 3) seems more likely to either be a correlated response
to sexual selection or a result of direct selection on female
preferences.
The novelty of this study is that we have disentangled the
relative role of natural and sexual selection in the phenotypic
divergence of natural populations, and related this to the development of sexual isolation. Previous studies of free-living
vertebrate populations have shown that directional natural
selection may drive population divergence but the extent of
divergence is constrained by gene flow over both large and
small geographical distances (Smith et al. 1997; Hendry et
al. 2000, 2002; Badyaev et al. 2002). However, we are unaware of studies that have demonstrated such a predominant
role for directional sexual selection, in combination with such
a pronounced degree of sexual isolation between closely located, parapatric populations. Previous studies on sexual isolation and other forms of reproductive isolation have focused
mainly on the two extreme cases of full gene flow (sympatry)
or zero gene flow (allopatry) between populations (Coyne
and Orr 2004). More recently, it has been suggested that
SELECTION AND POPULATION DIVERGENCE
future studies should focus on the more common scenario of
intermediate gene flow, or parapatry (Gavrilets 2004; Coyne
and Orr 2004), as we did in this study. To our knowledge,
field studies on the extent of premating isolation between
parapatric populations are still relatively few, except in the
context of hybrid zones or between highly diverged taxa. In
light of a recent focus on the role of natural selection in
evolutionary divergence (Schluter 2000), time is now ripe to
evaluate the relative importance of natural and sexual selection, as well as interactions between these different forms of
selection. A major challenge of future studies will be to quantify the strength and evaluate the relative importance of natural selection, sexual selection (this study) and sexual conflict
(Gavrilets 2000; Svensson et al. 2005) in the divergence of
natural populations.
ACKNOWLEDGMENTS
We are grateful to R. Härdling, T. Gosden, C. Benkman,
A. Hendry, and an anonymous referee for constructive criticisms on the first draft of this manuscript, and to all the field
assistants that participated in this project during 2001–2003.
This study was financially supported by grants from the
Swedish Research Council (VR), The Swedish Council for
Environment, Agriculture, and Spatial Planning (FORMAS),
and the Swedish Royal Academy of Science (KVA) to EIS.
LITERATURE CITED
Arnold, S. J., M. E. Pfrender, and A. G. Jones. 2001. The adaptive
landscape as a conceptual bridge between micro- and macroevolution. Genetica 112–113:9–32.
Arnqvist, G., and M. Kirkpatrick. 2005. The evolution of infidelity
in socially monogamous passerines: The strength of direct and
indirect selection on extrapair copulation behavior in females.
Am. Nat. 165:S26–S37.
Askew, R. R. 1988. The dragonflies of Europe. Harley Books, Colchester, U.K.
Badyaev, A. V., G. E. Hill, M. L. Beck, A. A. Dervan, R. A. Duckworth, K. J. Mcgraw, P. M. Nolan, and L. A. Whittingham. 2002.
Sex-biased hatching order and adaptive population divergence in
a passerine bird. Science 295:316–318.
Benkman, C. W. 2003. Divergent selection drives the adaptive radiation of crossbills. Evolution 57:1176–1181.
Benkman, C. W., T. L. Parchman, A. Favis, and A. M. Siepielski.
2003. Reciprocal selection causes a coevolutionary arms race between crossbills and lodgepole pine. Am. Nat. 162:182–194.
Bensch, S., D. Hasselquist, B. Nielsen, and B. Hansson. 1998. Higher
fitness for philopatric than for immigrant males in a semi-isolated
population of great reed warblers. Evolution 52:877–883.
Boughman, J. W., H. D. Rundle, and D. Schluter. 2005. Parallel
evolution of sexual isolation in sticklebacks. Evolution 59:
361–373.
Butlin, R. 1987. Speciation by reinforcement. Trends Ecol. Evol. 2:
8–13.
———. 1989. Reinforcement of pre-mating isolation. Pp. 158–179
in D. Otte and J. A. Endler, eds., Speciation and its consequences.
Sinauer, Sunderland, MA.
Corbet, P. S. 1999. Dragonflies: behaviour and ecology of Odonata.
Harley Books, Colchester, U.K.
Coyne, J. A. and H. A. Orr. 2004. Speciation. Sinauer, Sunderland,
MA.
Coyne, J. A., N. H. Barton, and M. Turelli. 1997. Perspective: a
critique of of Sewall Wright’s shifting balance theory of evolution.
Evolution 51:643–671.
Darwin, C. 1871. The descent of man and selection in relation to
sex. Murray, London.
Day, T. 2000. Sexual selection and the evolution of costly female
preferences: spatial effects. Evolution 54:715–730.
1251
Endler, J. A. 1977. Geographic variation, speciation, and clines.
Princeton Univ. Press, Princeton, NJ.
Fear, K., and T. Price. 1998. The adaptive surface in ecology. Oikos
82:440–448.
Fuller, R. C., D. Houle, and J. Travis. 2005. Sensory bias as an
explanation for the evolution of mate preferences. Am. Nat. 166:
437–446.
Gavrilets, S. 2000. Rapid evolution of reproductive barriers by sexual
conflict. Nature 403:886–889.
Gavrilets, S. 2004. Fitness landscapes and the origin of species.
Princeton Univ. Press, Princeton, NJ.
Hamon, T. R. 2005. Measurement of concurrent selection episodes.
Evolution 59:1096–1103.
Hamon, T. R., and C. J. Foote. 2005. Concurrent natural and sexual
selection in wild male sockeye salmon, Oncorhynchus nerka. Evolution 59:1104–1118.
Hendry, A. P., and E. B. Taylor. 2004. How much of the variation
in adaptive divergence can be explained by gene flow? An evaluation using lake-stream stickleback pairs. Evolution 58:
2319–2331.
Hendry, A. P., J. K. Wenburg, P. Bentzen, E. C. Volk, and T. P.
Quinn. 2000. Rapid evolution of reproductive isolation in the wild:
evidence from introduced salmon. Science 290:516–518.
Hendry, A. P., E. B. Taylor, and J. D. McPhail. 2002. Adaptive
divergence and the balance between selection and gene flow: lake
and stream stickleback in the misty system. Evolution 56:
1199–1216.
Hereford, J., T. F. Hansen, and D. Houle. 2004. Comparing strengths
of directional selection: how strong is strong? Evolution 58:
2133–2143.
Hoekstra, H. E., J. M. Hoekstra, D. Berrigan, S. N. Vignieri, A.
Hoang, C. E. Hill, P. Beerli, and J. G. Kingsolver. 2001. Strength
and tempo of directional selection in the wild. Proc. Natl. Acad.
Sci. USA 98:9157–9160.
Holland, B., and W. R. Rice. 1998. Perspective: chase-away sexual
selection: Antagonistic seduction versus resistance. Evolution 52:
1–7.
Iwasa, Y., and A. Pomiankowski. 1995. Continual change in mate
preferences. Nature 377:420–422.
Janzen, F. J., and H. S. Stern. 1998. Logistic regression for empirical
studies of multivariate selection. Evolution 52:1564–1571.
Jiggins, C. D., R. E. Naisbit, R. L. Coe, and J. Mallet. 2001. Reproductive isolation caused by colour pattern mimicry. Nature 411:
302–305.
Kawecki, T. J., and D. Ebert. 2004. Conceptual issues in local adaptation. Ecology Letters 7:1225–1241.
King, R. B., and R. Lawson. 1995. Color-pattern variation in Lake
Erie water snakes: the role of gene flow. Evolution 49:885–896.
Kingsolver, J. G., H. E. Hoekstra, J. M. Hoekstra, D. Berrigan, S.
N. Vignieri, C. E. Hill, A. Hoang, P. Gibert, and P. Beerli. 2001.
The strength of phenotypic selection in natural populations. Am.
Nat. 157:245–261.
Kirkpatrick, M. 1996. Genes and adaptation: a pocket guide to the
theory. Pp. 125–146 in M.R. Rose and G.V. Lauder, eds. Adaptation. Academic Press, San Diego, CA.
Kirkpatrick, M., and N. H. Barton. 1997. The strength of indirect
selection on female mating preferences. Proc. Natl. Acad. Sci.
USA 94:1282–1286.
Kotiaho, J. S. 2001. Costs of sexual traits: a mismatch between theoretical considerations and empirical evidence. Biol. Rev. 76:
365–376.
Lande, R. 1982. Rapid origin of sexual isolation and character divergence in a cline. Evolution 36:213–223.
Lande, R., and S. J. Arnold. 1983. The measurement of selection on
correlated characters. Evolution 37:1210–1226.
Losos, J. B., D. A. Creer, D. Glossip, R. Goellner, A. Hampton, G.
Roberts, N. Haskell, P. Taylor, and J. Ettling. 2000. Evolutionary
implications of phenotypic plasticity in the hindlimb of the lizard
Anolis sagrei. Evolution 54:301–305.
McPeek, M. A., and J. M. Brown. 2000. Building a regional species
pool: diversification of the Enallagma damselflies in Eastern North
America. Ecology 421:904–920.
Merilä, J., and P. Crnokrak. 2001. Comparisons of genetic differentiation at marker loci and quantitative traits. J. Evol. Biol. 14:
892–903.
1252
ERIK I. SVENSSON ET AL.
Möller, A. P., and R. Alatalo. 1999. Good-genes effects in sexual
selection. Proc. R. Soc. Lond. B. 266:85–91.
Nosil, P., and B. J. Crespi. 2004. Does gene flow constrain adaptive
divergence or vice versa? A test using ecomorphology and sexual
isolation in Timema cristinae walking-sticks. Evolution 58:
102–112.
Nosil, P., B. J. Crespi, and C. P. Sandoval. 2002. Host-plant adaptation drives the parallel evolution of reproductive isolation. Nature 417:440–443.
Nosil, P., T. H. Vines, and D. J. Funk. 2005. Perspective: reproductive
isolation caused by natural selection against immigrants from divergent habitats. Evolution 59:705–719.
Orteiza, N., J. E. Linder, and W. R. Rice. 2005. Sexy sons from
remating do not recoup the direct costs of harmful male interactions in the Drosophila melanogaster laboratory model system.
J. Evol. Biol. 18:1315–1323.
Panhuis, T. M., R. Butlin, M. Zuk, and T. Tregenza. 2001. Sexual
selection and speciation. Trends Ecol. Evol. 16:364–371.
Paulson, D. R. 1974. Reproductive isolation in damselflies. Syst.
Zool. 23:40–49.
Phillips, P. C., and S. J. Arnold. 1989. Visualizing multivariate selection. Evolution 43:1209–1266.
Price, T. 1998. Sexual selection and natural selection in bird speciation. Philos. Trans. R. Soc. Lond. B. 353:251–260.
Price, T. D., A. Qvarnstrom, and D. E. Irwin. 2003. The role of
phenotypic plasticity in driving genetic evolution. Proc. R. Soc.
Lond. B. 270:1433–1440.
Rice, W. R., and E. Hostert. 1993. Laboratory experiments on speciation: what have we learned in 40 years? Evolution 47:
1637–1653.
Rieseberg, L. H., A. Widmer, A. M. Arntz, and J. M. Burke. 2002.
Directional selection is the primary cause of phenotypic diversification. Proc. Natl. Acad. Sci. USA 99:12242–12245.
Rowe, L., and D. Houle. 1996. The lek paradox and the capture of
genetic variance by condition dependent traits. Proc. R. Soc. Lond.
B. 263:1415–1421.
Schluter, D. 2000. The Ecology of Adaptive Radiation. Oxford Univ.
Press, Oxford, U.K.
———. 2001. Ecology and the origin of species. Trends Ecol. Evol.
16:372–380.
Schluter, D., and D. Nychka. 1994. Exploring fitness surfaces. Am.
Nat. 143:597–616.
Seehausen, O., J. J. M. van Alphen, and F. Witte. 1997. Cichlid fish
diversity threatened by eutrophication that curbs sexual selection.
Science 277:1808–1811.
Servedio, M. R. 2001. Beyond reinforcement: the evolution of premating isolation by direct selection on preferences and postmating, prezygotic incompatibilities. Evolution 55:1909–1920.
———. 2004. The evolution of premating isolation: Local adaptation
and natural and sexual selection against hybrids. Evolution 58:
913–924.
Siva-Jothy, M. 1999. Male wing pigmentation may affect reproductive success via female choice in a Calopterygid damselfly (Zygoptera). Behaviour 136:1365–1377.
———. 2000. A mechanistic link between parasite resistance and
expression of a sexually selected trait in a damselfly. Proc. R.
Soc. Lond. B. 267:2523–2527.
Smith, T. B., R. K. Wayne, D. J. Girman, and M. Bruford. 1997. A
role for ecotones in generating rainforest biodiversity. Science
276:1855–1857.
Statsoft, I. 2003. STATISTICA (data analysis software system). Vers.
6. Available at: www.statsoft.com.
Svensson, E. I., and B. Sinervo. 2000. Experimental excursions on
adaptive landscapes: density-dependent selection on egg size.
Evolution 54:1396–1403.
———. 2004. Spatial scale and temporal component of selection in
side-blotched lizards. Am. Nat. 163:726–734.
Svensson, E. I., L. Kristoffersen, K. Oskarsson, and S. Bensch. 2004.
Molecular population divergence and sexual selection on morphology in the banded demoiselle (Calopteryx splendens). Heredity 93:423–433.
Svensson, E. I., J. Abbott, and R. Hardling. 2005. Female polymorphism, frequency dependence, and rapid evolutionary dynamics
in natural populations. Am. Nat. 165:567–576.
Tynkkynen, K., M. J. Rantala, and J. Suhonen. 2004. Interspecific
aggression and character displacement in the damselfly Calopteryx
splendens. J. Evol. Biol. 17:759–767.
Tynkkynen, K., J. S. Kotiaho, M. Luojumaki, and J. Suhonen. 2005.
Interspecific aggression causes negative selection on sexual characters. Evolution 59:1838–1843.
Waage, J. K. 1975. Reproductive isolation and potential for character
displacement in damselflies, Calopteryx maculata and Calopteryx
aequabilis (Odonata-Calopterygidae). Syst. Zool. 24:24–36.
———. 1979. Reproductive character displacement in Calopteryx
(Odonata: Calopterygiidae). Evolution 33:104–116.
West-Eberhard, M. J. 1983. Sexual selection, social competition and
speciation. Q. Rev. Biol. 58:155–183.
———. 2003. Developmental plasticity and evolution. Oxford University Press, Oxford, U.K.
Whitlock, M. C., P. C. Phillips, F. B. G. Moore, and S. J. Tonsor.
1995. Multiple fitness peaks and epistasis. Annu. Rev. Ecol. Syst.
26:601–629.
Wong, B. B. M., J. S. Keogh, and M. D. Jennions. 2004. Mate recognition in a freshwater fish: geographical distance, genetic differentiation, and variation in female preference for local over
foreign males. J. Evol. Biol. 17:701–708.
Corresponding Editor: C. Benkman
APPENDIX 1.
Estimated LS means from a general linear model with population, year, and interactions between year and populations as factors.
Differences between populations for 12 different traits in the selection study. Total sample sizes are 455 (Klingavälsån) and 568 males
(Höje Å), and data are from the seasons of 2002 and 2003.
Mean (SE)
Trait
Abdomen length (ABL)
Forewing length (FWL)
Forewing width (FWW)
Forewing patch length (FPL)
Hindwing length (HWL)
Hindwing width (HWW)
Hindwing patch length (HPL)
Mass (MASS)
Patch color score (PCS)
Total length (TOL)
Thorax width (THW)
Thorax length (THL)
Klingavälsån
35.26
29.97
9.76
15.73
28.93
9.49
15.04
0.122
3.24
43.34
3.86
4.50
(0.08)
(0.06)
(0.02)
(0.10)
(0.06)
(0.02)
(0.09)
(0.001)
(0.07)
(0.09)
(0.01)
(0.01)
Höje Å
35.57
29.97
9.81
15.85
29.08
9.60
15.35
0.118
2.86
43.58
3.81
4.34
(0.07)
(0.05)
(0.02)
(0.08)
(0.05)
(0.02)
(0.07)
(0.001)
(0.06)
(0.07)
(0.01)
(0.01)
Diff.
t
P
20.31
;0
20.05
20.12
20.15
20.11
20.31
0.04
0.38
20.24
0.05
0.16
4.514
0.014
2.489
1.380
2.819
5.691
4.073
4.750
5.980
3.106
6.239
15.688
,0.001
0.99
0.013
0.170
0.005
,0.001
,0.001
,0.001
,0.001
0.002
,0.001
,0.001
APPENDIX 2.
PCS
THW
THL
FWL
FWW
FPL
HWL
HWW
HPL
TOL
ABL
Mass
PCS
THW
THL
FWL
FWW
FPL
HWL
HWW
HPL
TOL
ABL
Mass
1.80986
0.04144
0.05647
0.03068
0.07029
0.20061
0.11845
0.06344
0.22049
0.07528
0.12472
0.00263
0.51366
0.03734
20.06014
20.00710
20.00059
20.12931
20.04391
20.00895
20.15761
20.06130
20.03170
20.00094
PCS
0.23245
0.01756
0.01502
0.07355
0.01985
0.05264
0.06969
0.01695
0.04575
0.08066
0.05961
0.00114
0.33949
0.02355
20.00739
0.02701
0.01652
20.03388
0.01711
0.01186
20.04508
0.04325
0.03494
0.00022
THW
0.22838
0.61685
0.03378
0.06959
0.02208
0.06664
0.06708
0.01942
0.05920
0.09117
0.06788
0.00109
20.37128
20.21293
0.05108
0.03565
0.01414
0.10394
0.05401
0.01878
0.11622
0.06546
0.05614
0.00102
THL
0.02521
0.61361
0.41868
0.81798
0.15227
0.50379
0.70866
0.12002
0.42817
0.61426
0.47647
0.00728
20.01267
0.22504
0.20169
0.61155
0.11737
0.28370
0.54160
0.11051
0.28220
0.49869
0.42534
0.00368
FWL
0.17020
0.48802
0.39144
0.54849
0.09422
0.12226
0.15944
0.07260
0.12507
0.14524
0.12736
0.00166
20.00276
0.35866
0.20850
0.50014
0.09005
0.05744
0.11880
0.07343
0.05461
0.12599
0.11633
0.00103
FWW
0.10658
0.28388
0.25918
0.39813
0.28467
1.95757
0.48342
0.10272
1.61691
0.41768
0.30497
0.00545
20.13942
20.17057
0.35537
0.28033
0.14790
1.67475
0.28238
0.06577
1.43178
0.27520
0.25220
0.00386
FPL
0.10037
0.59946
0.41610
0.89322
0.59211
0.39387
0.76951
0.12925
0.43916
0.56800
0.45391
0.00706
20.07873
0.14324
0.30703
0.88991
0.50869
0.28038
0.60566
0.11688
0.31756
0.48397
0.41917
0.00379
HWL
0.16826
0.45641
0.37708
0.47349
0.84391
0.26195
0.52573
0.07854
0.10655
0.12090
0.10373
0.00127
20.04210
0.26052
0.28009
0.47626
0.82469
0.17129
0.50615
0.08804
0.07530
0.10633
0.11291
0.00101
HWW
0.13084
0.27559
0.25714
0.37792
0.32525
0.92254
0.39964
0.30350
1.56923
0.37223
0.28255
0.00464
20.18290
20.24433
0.42768
0.30012
0.15136
0.92016
0.33937
0.21105
1.44569
0.28195
0.24968
0.00402
HPL
0.05063
0.55069
0.44882
0.61448
0.42810
0.27009
0.58582
0.39030
0.26884
1.22165
0.89866
0.00784
20.07364
0.24265
0.24939
0.54910
0.36152
0.18311
0.53548
0.30856
0.20192
1.34873
0.94054
0.00580
TOL
0.09620
0.46680
0.38327
0.54670
0.43057
0.22620
0.53696
0.38407
0.23406
0.84373
0.92861
0.00632
20.04235
0.21799
0.23786
0.52080
0.37118
0.18660
0.51573
0.36438
0.19884
0.77546
1.09072
0.00406
ABL
0.14610
0.64120
0.44358
0.60063
0.40409
0.29063
0.60046
0.33889
0.27657
0.52925
0.48981
0.00018
20.12303
0.13718
0.42473
0.44065
0.32170
0.27950
0.45702
0.31890
0.31369
0.46807
0.36477
0.00011
MASS
Phenotypic variance-covariance and correlation matrices for all the morphological traits in the divergent selection study. Trait abbreviations as in Table 1. Diagonal: variances.
Below diagonal: covariances. Above diagonal: correlations. Significant correlations (P , 0.05) are indicated in boldface. Data are shown separately for the two populations:
Klingavälsån (upper matrix; n 5 455) and Höje Å (lower matrix; n 5 568). For explanation of trait abbreviations, see Appendix 1.
SELECTION AND POPULATION DIVERGENCE
1253