Sex-specific selective pressures on body mass in the greater white-toothed shrew, Crocidura russula

Authors


Caroline Reuter, Department of Biological Sciences, Imperial College London, Silwood, Park Campus, Ascot Berks SL5 7PY, UK.
Tel.: ++44 20 7594 2256; fax: ++44 20 7594 2339;
e-mail: c.reuter@imperial.ac.uk

Abstract

The direction, intensity and shape of viability-, sexual- and fecundity selection on body mass were investigated in a natural population of the greater white-toothed shrew (Crocidura russula), combining parentage assignment through molecular techniques and mark–recapture data over several generations. A highly significant stabilizing viability selection was found in both sexes, presumably stemming from the constraints imposed by their insectivorous habits and high metabolic costs. Sexual selection, directional in both sexes, was twice as large in males than in females. Our results suggest that body mass matters in this context by facilitating the acquisition and defense of a breeding territory. No fecundity selection could be detected. The direction of sexual size dimorphism (SSD) was in agreement with the observed pattern of selective pressures: males were heavier than females, because of stronger sexual selection. SSD intensity, however, was low compared with other mammals, because of the low level of polygyny, the active role of females in territory defense and the intensity of stabilizing viability selection.

Introduction

Body mass is a key life-history trait because it strongly affects features that bear direct and diverse connections to fitness, such as resource acquisition, metabolic rate, mating success or longevity (McNab, 1971; Roff, 1992; Stearns, 1992). As a consequence, body mass undergoes a variety of selective pressures, often with opposite effects, and its evolutionary stable value is bound to reflect a subtle balance between multiple forces.

These selective forces can be categorized into three main classes, namely viability-, sexual- and fecundity selection (e.g. Arnold & Wade, 1984a,b; Blanckenhorn, 2000; Kingsolver et al., 2001). Viability selection includes all forces stemming from the ecological interactions an organism develops with its environment. The optimal exploitation of a given ecological niche, in terms of resource acquisition, metabolic rate and predator avoidance, is expected to impose a stabilizing selective pressure on body mass (Peters, 1983; Schmids-Nielsen, 1984). In several bird species, for instance, equilibrium body mass was shown to result from a trade-off between the risk of starvation at lower mass, and the risk of predation at higher mass (e.g. Adriaensen et al., 1998).

However, the two other classes of selective forces may interfere with viability selection and drive body mass away from its ecological optimum. Sexual selection, more likely to act on males, often favours large individuals, either because they dominate in intrasexual competition (del Castillo et al., 1999; Forslund, 2000; Hagelin, 2002), or because they are preferred by opposite-sex partners (Balmford et al., 1992; Patton & Smith, 1993). Fecundity selection, more likely to matter in females, also promote large individuals, because they may accommodate more eggs, feed more offspring (Barbraud et al., 1999; Reeve & Fairbairn, 1999) or produce fitter offspring (Bowen et al., 2001).

Insofar as males and females share the same ecological niche, viability selection affects both sexes equally. Sex differences in body mass may thus shed some light on the intensities of sexual- or fecundity selection, which are more likely to display sex-specificity. Sexual size dimorphism (SSD), defined as the difference in body size between males and females, arises in many taxa, often with taxon-specific differences in the direction of the bias, because selective pressures equilibrate differentially. Female-biased SSD generally predominates among invertebrates and ectothermic vertebrates, where fecundity selection on females is likely to exceed sexual selection on males (Arak, 1988; Shine, 1988; Wiklund & Karlsson, 1988; Head, 1995; Fairbairn & Preziosi, 1996; Reeve & Fairbairn, 1999). By contrast, male-biased SSD often predominates among birds (Björklund, 1990; Webster, 1992) and mammals (McElligott et al., 2001; Kraaijeveld-Smit et al., 2003), because sexual selection on males normally exceeds fecundity selection on females. Although the males of polygynous mammals are likely to encounter the highest levels of sexual selection (Fisher & Lara, 1999; McElligott et al., 2001; Preston et al., 2003), some degree of sexual selection is also expected in monogamous species, stemming from competition for access to better breeding territories and/or better partners.

Investigating the forces of viability-, sexual- and fecundity selection in the field is not an easy task, because it requires long-term monitoring of populations and information on pedigrees to assess parentage. Sexual selection has often been evaluated indirectly, based on operational sex-ratio (ratio of fertile females to mature males), but this may not always reflect the actual degree of competition between individuals (Webster, 1992; Mitani et al., 1996). Complete studies of selective pressures on body mass in mammals are scarce: existing studies often focus on a subset of selective forces, or on one sex only. Fecundity- and viability selection were recently investigated in a female-biased dimorphic species, the yellow-pine chipmunk, Tamias amoenus (Schulte-Hostedde et al., 2002). Similarly, the role of sexual selection was studied in male-biased dimorphic species such as the fallow deer, Dama dama (McElligott et al., 2001) and the marsupial Antechinus agilis (Kraaijeveld-Smit et al., 2003). We are not aware, however, of any mammalian study in which all three selective forces were simultaneously taken into account on both sexes.

In the present study, we combined molecular techniques of parentage assignment with a mark–recapture study over several generations, in order to evaluate longevity, mating success and lifetime reproductive output in a natural population of a small insectivorous mammal, the greater white-toothed shrew (Crocidura russula). This approach, still rarely explored (but see Meriläet al., 1997) enabled us, first to document the patterns of SSD in this species, secondly to quantify the global selective pressure on body mass through its correlation with lifetime reproductive success (LRS), and thirdly to decompose this pressure into its main components, namely viability-, sexual- and fecundity selection, in order to compare their relative contribution in shaping and maintaining the observed pattern of sexual dimorphism.

Material and methods

Study species and field sampling

Crocidura russula is a small annual insectivorous mammal, widespread in south-western Europe. The species is anthropophilic in the northern part of its distribution (including the study area) where it occurs in discrete populations inhabiting villages and suburbs. Pairs rear up to four litters (two on average) from March to September. Less than 5% of adults survive winter to enjoy a second reproductive season (Jeanmaire-Besançon, 1986), so that generation overlap can be safely neglected. Weaning occurs at day 20 after birth, at which time some female-biased natal dispersal occurs (Favre et al., 1997; Balloux et al., 1998). Juveniles weigh 7–9 g at weaning, and reach a stationary body mass of 10–12 g within 2 months.

Our study population occupies an area of c. 250 × 350 m on the campus of the University of Lausanne (Switzerland) at Dorigny (6° 34′E, 45° 31′N, 400 m above sea level). One hundred and eighty longworth traps, baited with Tenebrio molitor larvae, were set at all potentially favourable breeding sites. Subsequently, trapping sessions were performed once a week during four successive years (1998–2001). At first capture, individuals were sexed and marked by toe clipping. Toes were kept frozen (−20 °C) before DNA extraction. At each capture individuals were weighed (to the nearest 0.5 g) and their breeding status was determined (juveniles are greyish and lighter than adults, breeding males present visible lateral glands and breeding females have visible teats). All individuals were released at the site of capture. As female mass drops from about 20 to 14 g at parturition, weekly trapping allowed quite precise estimation of the dates of birth of juveniles in most cases.

Body mass values used in all subsequent analyses are individual averages over the whole adult lifetime, i.e. excluding values taken before day 60 after birth. For females, we also excluded all measurements taken during pregnancy or lactation (from 30 days before parturition to 20 days after parturition). An index of SSD was estimated for all years pooled together (1999–2001), as I = (Mh − Ml)/Ml, where Mh measures the body mass of the heaviest sex and Ml that of the lightest sex. This index is arbitrarily reported as positive when females are larger and negative when males are larger (Fairbairn, 1997).

Pedigree determination

We used genetic analyses to reconstruct the pedigree of the population for the 1998–2001 breeding seasons. DNA was extracted using either the standard method of Sambrook et al. (1989) or the salting-out procedure of Miller et al. (1988). All individuals were genotyped at 12 microsatellite markers (loci 9, 17, 23, 45, 53, 54, 57, 72 of Favre & Balloux (1997) as well as the loci 24, 41b, 49 and 52 (Genbank accession numbers AY034426, AY034427, AY034428, AY034429 respectively). Parental assignment was carried out using the software Probmax (Danzmann, 1997) combined with temporal and spatial information (more details in Duarte et al., 2003). The large amount of genetic variance in the population provided a very high exclusion probability (0.998, Jamieson, 1994). Of 729 juveniles caught during the 1998 to 2001 breeding seasons, 503 (70%) were attributed to a local breeding pair. The remaining 226 individuals were immigrant in the study site. Pedigree enabled us to quantify the number of mates and of surviving offspring for each breeder.

Selection analyses

The global selective pressure on body mass, as well as its component forces, namely viability-, sexual- and fecundity selection, were quantified through the 1999–2001 breeding seasons. The 1998 season was excluded from the analyses because of an incomplete body mass data set. The intensity of global selection was estimated through individual LRS, estimated from trapping data and parentage assignment. Viability selection was quantified through longevity, obtained from mark–recaptured data. As the time of the year at which individuals are born is likely to affect their longevity, month of birth was introduced as a covariate in analyses. The intensity of sexual selection was measured through individual mating success (number of mates with which the focal individual successfully reared at least one offspring). Fecundity selection was estimated, for mated individuals only, via the number of weaned offspring per litter. As juveniles cannot be caught before they leave the nest for limited exploratory excursions (from day 13 to day 20 after birth), this effective fecundity accounts for possible mortality before weaning.

Standardized selection gradients were estimated for each sex and each year separately according to Arnold & Wade (1984a,b) and Lande & Arnold (1983). We also combined data across years. Standardized selection gradients are based on regressions of relative fitness (inline image, where wi is the absolute fitness of the ith individual and inline image is the mean fitness of individuals in the year considered) on standardized body mass values (inline image, where xi is the body mass of the ith individual, and inline image and SDx are respectively the mean body mass and its standard deviation for the year considered). Directional (linear) selection gradients for sexual-, fecundity-, viability- and global selection (βsβfβv and βLRS respectively) were estimated from the linear model w ′ =c + βz. These gradients are mathematically equivalent to standardized selection differentials, i (Endler, 1986). To investigate possible stabilizing or disruptive (nonlinear) selection, we calculated standardized nonlinear selection gradients by fitting quadratic regressions of the form: w ′ = c + β1z + β2z 2. Coefficients of nonlinear selection (γsγfγv and γLRS respectively) are given by 2β2 (Fairbairn & Preziosi, 1996). Stabilizing selection is indicated by a negative nonlinear selection, whereas a positive coefficient denotes disruptive selection. Selection gradients (both linear and nonlinear) and their significance were assessed with robust regression techniques. Data were resampled 1000 times with replacement, and selection gradients recalculated for each new data set. Standard errors of selection gradients were obtained as the standard deviations of the new data set containing the 1000 selection coefficients obtained by the bootstrap. All statistical analyses were conducted using R 1.7.1 (Ihaka & Gentleman, 1996).

As we estimated two selection gradients for each year (β and γ) the appropriate critical α value should be divided by two; hence for each analysis of selection α =0.025. Because such correction is not standard in estimating selection gradients, we also provide significance at a critical α of 0.05 (Fairbairn & Preziosi, 1996).

Results

Sexual dimorphism

Males were significantly heavier than females for the years 2000 and 2001 (2000: 11.1 ± 1.0 g vs. 10.6 ± 1.1 g for males and females respectively, t-tests on log-transformed data, d.f. = 148, P < 0.01; 2001: 11.0 ± 1.2 g vs. 10.0 ± 0.9 g, t-tests on inverse-transformed data, d.f. =164, P < 0.01), and marginally so for 1999 (11.3 ± 1.3 g vs. 10.9 ± 1.4 g, t-test on log-transformed data, d.f. =141, P = 0.09). The trend was highly significant when all 3 years were combined, males weighing 11.2 ± 1.2 g and females 10.5 ± 1.2 g (t-test on log-transformed data, d.f. = 441, P < 0.01). This amounts to a mean sexual dimorphism index of I = −0.06.

Viability selection

Results on lifespan globally point to a stabilizing viability selection for body mass (Fig. 1b). Data for females show a negative nonlinear gradient for each year (γv =−0.23 ± 0.09, P < 0.05 for 1999, γv = −0.17 ± 0.07, P = 0.05 for 2000, and γv = −0.19 ± 0.07, P < 0.05 for 2001). The trend is highly significant when all 3 years are pooled (γv = −0.17 ± 0.04, P < 0.001). Lifespan was thus shorter for individuals lighter or heavier than the population mean. Directional selection gradients in females were not significant for any year. On average, female lifespan was equal to 6.1 ± 3.8 months (range 2.0–19.7). Remember however, that individuals dying before being 2 months old are excluded from analyses.

Figure 1.

Viability selection for body mass. Linear (a) and nonlinear (b) selection gradients and their SE for males (in black) and females (in white). Sample sizes are given in the top of graphs. Cubic spline visualization of viability selection on body mass (plain line) ±1 SE (dotted lines) for males (c) with λ = −2 and females (d) with λ = 0 (Schluter, 1988). *P < 0.05, **P < 0.01, ***P < 0.001, (*) indicates that the P-value is not significant at the 0.025 level (see Material and methods).

In males, stabilizing selection was also highly significant in 1999 (γv = −0.24 ± 0.08, P < 0.001), and when all data were pooled (γv = −0.19 ± 0.05, P < 0.001, Fig. 1b). In 2000, a significant positive directional trend was also observed for males (βv = 0.29 ± 0.07, P < 0.001, Fig. 1a). Males lived on average 6.4 ± 3.7 months (range 2.0–25.1). A graphical visualization of longevity in function of body mass is shown in Fig. 1c and d.

Sexual selection

Over the 4 years, the mean sex-ratio, measured as the proportion of males, was slightly female biased (0.45 ± 0.05), so that, when all adults are accounted for, males had on average access to 1.23 ± 0.05 females (and females to 0.82 ± 0.03 males). Excluding nonbreeding individuals, males mated with 1.92 ± 1.10 females (range 1–6), with nearly half of the breeding males (48%) mating with two or more females.

Heavier males had an advantage in access to females (Figs 2a and c). Directional sexual selection through mating success was positive and significant in all years at the 0.025 level. The βs were equal to 0.83 ± 0.19 in 1999 (P < 0.001), 0.64 ± 0.23 in 2000 (P < 0.01) and 0.71 ± 0.13 in 2001 (P < 0.001). On average over the 3 years, a body mass increase of 1 SD enhanced male mating success by 0.7 partners. The effect disappeared when the analysis was restricted to breeding males, showing that the weight differential characterized the contrast between breeding and nonbreeding males. Neither stabilizing nor disruptive selection was apparent in any of the years (Fig. 2b).

Figure 2.

Sexual selection for body mass. Linear (a) and nonlinear (b) selection gradients and their SE for males (in black) and females (in white). Sample sizes are given in the top of graphs. Cubic spline visualization of sexual selection on body mass (plain line) ±1 SE (dotted lines) for males (c) with λ = 2 and females (d) with λ = 1 (Schluter, 1988).

Excluding nonbreeding individuals, females bred with 1.30 ± 0.61 males (range 1–5). About 28% (22 of 79) of the breeding females mated with two or more partners. Polyandry was mainly sequential, but multiple paternity occurred in 37% (eight of 22) of the litters of multiply-mated females.

Heavier females also gained more partners than lighter ones (Fig. 2a and d). Directional selection gradients for mating success were significant at the 0.025 level for 1999 and 2001 (βs = 0.41 ± 0.15 and 0.53 ± 0.16 in 1999 and 2001 respectively, both P < 0.01), and at the 0.05 level for 2000 (βs = 0.37 ± 0.18). On average over the 3 years, females gained 0.4 partners for 1 SD increase of body mass. The effect also disappeared when the analysis was restricted to breeding females, showing that the weight differential characterized the contrast between breeding and nonbreeding females. Nonlinear selection gradients were not significant for any of the years (Fig. 2b).

On average females underwent a weaker sexual selection pressure than males, but the difference was only significant when data were combined across years (t-test on the two classes regression coefficients; 1999: d.f. = 114, n.s; 2000: d.f. = 88, n.s; 2001: d.f. = 131, n.s.; across years: d.f. = 341, P < 0.01).

Fecundity selection

Body size was apparently not under fecundity selection. In females, the directional selection coefficient was positive and significant at the 0.05 level in 2000 only (βf = 0.22 ± 0.08, Fig. 3a) and the trend disappeared when all years were pooled (βf = 0.11 ± 0.06, n.s). Females produced on average 2.0 ± 1.0 surviving offspring per litter (range 1–5.5) and males 2.5 ± 1.5 (range 1–6.8; calculated from individual mean values). In males, the directional selection gradients were significant in none of the years (Fig. 3a). The cubic spline fitness functions for both sexes are shown in Fig. 3c and d. Nonlinear selection gradients were all nonsignificant for both males and females (Fig. 3b).

Figure 3.

Fecundity selection for body mass. Linear (a) and nonlinear (b) selection gradients and their SE for males (in black) and females (in white). Sample sizes are given in the top of graphs. Cubic spline visualization of fecundity selection on body mass (plain line) ±1 SE (dotted lines) for males (c) with λ = 7 and females (d) with λ = 2 (Schluter, 1988).

Global selection

Heavier males sired more offspring than lighter ones (Fig. 4a and c). Directional selection gradients for the number of offspring were positive and significant at the 0.025 level in all years. The βLRS were equal to 0.75 ± 0.29 in 1999 (P < 0.05), 0.64 ± 0.32 in 2000 (P < 0.01) and 0.99 ± 0.20 in 2001 (P < 0.001). On average over the 3 years, a body mass increase of 1 SD enhanced male fitness by 0.9 offspring. Neither stabilizing nor disruptive selection was significant in any of the years (Fig. 4b).

Figure 4.

Global selection for body mass. Linear (a) and nonlinear (b) selection gradients and their SE for males (in black) and females (in white). Sample sizes are given in the top of graphs. Cubic spline visualization of lifetime reproductive success selection on body mass (plain line) ±1 SE (dotted lines) for males (c) with λ = 2 and females (d) with λ = 6 (Schluter, 1988).

Heavier females also had greater fitness than lighter ones (Fig. 4a and d). Directional selection gradients on global fitness were positive and significant at the 0.025 level for 1999 and 2000 (βf = 0.56 ± 0.21, P < 0.001 and 0.42 ± 0.18, P < 0.01 in 1999 and 2000 respectively), and at the 0.05 level for 2001 (βf = 0.33 ± 0.18). On average over the 3 years, females increased their fitness by 0.6 offspring for 1 SD increase of body mass. Nonlinear selection gradients were not significant for any of the years (Fig. 4b).

Males underwent a higher LRS selection pressure than females in all years, the difference being significant in 2001 and when data were combined across years (t-test on the two classes regression coefficients, 1999: d.f. =114, n.s; 2000: d.f. = 88, n.s; 2001: d.f. = 131, P < 0.05; across years: d.f. = 341, P < 0.05).

Discussion

Our results point to a strong and highly significant stabilizing viability selection on body mass in both males and females. Values provided here might actually represent lower bounds: as individuals dead before day +60 were excluded from analyses, the intensity of selection might be underestimated. Selection against lighter individuals was not unexpected given the increased fasting endurance generally associated with heavier mass (Goodman et al., 1984). The energetic costs of living are very high in shrews (Genoud, 1985) because their small size and high surface to volume ratio boost thermoregulation costs (Peters, 1983). Fontanillas et al. (2004) showed that heavier individuals of C. russula benefited from an increased thermogenic capacity associated with a better ability to accumulate brown fat, and were more likely to survive winter in semi-outdoor conditions.

Selective pressures against larger individuals, by contrast, may stem in natural conditions from constraints imposed by their ecological niche. Shrews are insectivorous predators, preying actively upon a variety of small invertebrates (Myriapoda, Isopoda, Lepidoptera larvae, Gastropoda and Araneae; Genoud & Hutterer, 1990). Prey are caught by foraging in grass and leaf litter, wood and compost piles, stone walls, as well as underground (soil crevices, runways, tunnels and burrows of rodents; Churchfield, 1990). This feeding ecology necessarily imposes strong constraints on body size (Ochocinska & Taylor, 2003). Predation risk and metabolic costs presumably also play a role of their own: large individuals might be less agile or more visible to predators (Andersson, 1994), and may require more energy or suffer from greater heat stress (McNab, 1971; Furuyama & Ohara, 1993). Although energy needs relative to body mass are smaller in large individuals, their absolute values are larger (McNab, 1971; Churchfield, 1990) which may become crucially important when resource supply is low. Scarcity of food, especially in winter, appears as an important factor selecting for small structural body size in shrews (Churchfield, 2002; Ochocinska & Taylor, 2003), as otherwise evidenced by the so-called Dehnel's phenomenon (winter decrease in structural body size) that characterizes many Soricinae (Mezhzherin, 1964). Winter may thus simultaneously impose both a negative selection on structural mass, and a positive selection on brown fat accumulation, stemming respectively from the scarcity and from the unpredictability of resources, and thereby contribute to the stabilizing selective pressure on body weight (a complex trait) evidenced here.

Sexual selection, by contrast to viability selection, was directional in both sexes, consistently favouring larger individuals throughout the three breeding seasons considered. Heavier individuals were more likely to mate with at least one partner. However, the number of additional partners was unaffected. Body mass presumably matters in this context by facilitating the acquisition and defense of a breeding territory. Actually, body mass is known to correlate with dominance hierarchy in many species, larger individuals being more likely to outcompete smaller ones in territorial fights (Webster, 1997; Fisher & Lara, 1999; Piper et al., 2000; McElligott et al., 2001). Although sexual selection concerned both sexes in our case, it affected males significantly more than females (about twice as much), which seems readily interpretable on the basis of known behavioural features. For instance, both sexes are known to participate actively in territory defence in C. russula (Cantoni & Vogel, 1989). Males, however, play the main role, as underlined by the large lateral scent glands they develop during the breeding season and use for territory marking (Hausser & Juhlin-Dannfelt, 1979; Churchfield, 1990). Following Greenwood's (1980) resource-competition hypothesis, this predominant role of males in territory acquisition and defence may actually account for their higher philopatry, an interesting convergence with passerine birds (in which male song plays the functional role of lateral glands in shrews).

Body mass was not under fecundity selection. Larger breeding adults did not produce significantly more surviving offspring per litter. This might seem to oppose the results of Genoud & Perrin (1994), who documented an increase in litter size with female size in C. russula. However, these authors also found that juveniles from larger litters were lighter, which may induce increased mortality before weaning. Remember that the present study deals with effective fecundity, estimated at weaning, and therefore also accounts for possible nest mortality. Our present results thus suggest that, in terms of effective fecundity per litter (number of weanlings), neither males nor females benefit from a larger size.

In a review of the general trends of selection on morphological traits in natural populations of invertebrates, vertebrates and plants, Kingsolver et al. (2001) found that directional selection from survival is usually weaker (median of absolute values of β = 0.09) than it is from fecundity or mating success (median |β| = 0.16 and 0.17, respectively). Our results are in good qualitative agreement with this trend, although the strength of sexual selection found in the present study (β = 0.4 in females and 0.7 in males) can be considered as very important: only rarely do estimates of beta exceed 0.5 (Hoekstra et al., 2001; Kingsolver et al., 2001). By contrast, the patterns of nonlinear selection gradients seem to oppose the general trend, in that selection from viability is usually weaker (median of absolute values of γ = 0.02) than it is from fecundity or mating success (median |γ| = 0.14 and 0.16 respectively). The reverse occurs here, as viability was the only fitness component imposing significant nonlinear selection (with values of −0.17 and −0.19 in males and females, respectively). Thus, stabilizing viability selection on body mass can be considered as particularly important in the system under study.

The observed pattern of selective pressures driven by the combination of the three separate forces discussed above is in agreement with the direction of SSD, at least in our population and for the 3 years studied. Sexual selection was significantly larger in males than in females, and so was body mass. The intensity of SSD, however, was quite low (−0.06), compared with other mammals, in which SSD may reach extreme values, as observed in Primates (−0.20 to −1.20, Mitani et al., 1996), Proboscidea (−0.54), or Pinnipeds (−0.81 to −1.98, Weckerly, 1998). The small body size of shrews may play a role, as Rensch's rule states that SSD increases with size where males are the larger sex (Abouheif & Fairbairn, 1997). However, the values documented in other small mammals such as the insectivorous Hottentot golden mole Amblysomus hottentotus (−0.25 to −0.28, Bronner, 1996) or rodents of the genus Microtus (−0.04 to −0.32, Boonstra et al., 1993) or Spermophilus (−0.09 to −0.17, Dobson, 1992) consistently exceed the one reported here.

Several factors may be responsible for the low SSD in C. russula. Mating system might be of importance, considering the low level of polygyny displayed by C. russula in comparison with most mammalian species (Bouteiller & Perrin, 2000). Body mass did not affect the number of partners (as might often be the case in mammals; Coltman et al., 1999; Kraaijeveld-Smit et al., 2003), but more likely the ability to acquire a breeding territory, a prerequisite for any successful access to reproductive status. As females also take part in territory acquisition (Cantoni & Vogel, 1989), they also undergo some sexual selection on body mass, which limits sexual dimorphism. Furthermore, viability selection is likely to play a role as well: the strength of stabilizing selective pressure in both sexes may oppose any deviation of body mass from its ecological optimum. Finally, dimorphism may be restricted by genetic constraints: a strong genetic correlation between the sexes, for instance, should suffice to oppose a sex-divergence in body mass (Lande, 1980). This theoretical expectation is supported by empirical evidence: in Drosophila melanogaster, both sexes increased (respectively decreased) their body size in response to an upward (respectively downward) selection on males only or females only because of a genetic correlation of 0.9 between sexes (Reeve & Fairbairn, 1996). Genetic correlations of this order of magnitude are not uncommon for homologous morphological characters in males and females (Lande, 1980; Reeve & Fairbairn, 1996).

Thus, the weak but significant SSD documented in the present study is likely to stem from sexual selection on body size, which appears significantly stronger in males than in females. Our results suggest that this trait matters because an individual's ability to acquire and defend a territory (and thereby reach breeding status) depends on its body mass, and because males play a more central role in this process. However, the significant (although weaker) role of females in territory defense, as well as the strong stabilizing viability selection on both sexes, limits the divergence between sexes (SSD) and maintains body mass close to its ecological optimum.

The point remains that the global pattern of selection on LRS was directional and positive, namely favouring heavier individuals. Both larger males and females produced more offspring during their lifetime. The question thus arises, why are shrews not larger, given the substantial additive genetic variance measured for body mass (C. Bouteiller-Reuter & N. Perrin, unpublished)? This lack of microevolutionary response seems to be a familiar paradox (e.g. Alatalo et al., 1990; Milner et al., 1999, 2000; Kruuk et al., 2000, 2001, 2002), for which several potential explanations have been proposed (review in Meriläet al., 2001). One is that selection actually fluctuates from year to year, so that short-term microevolutionary trends disappear on the mid-term. However, selection on LRS was found to be directional and positive in both sexes and all 3 years under study. Furthermore, this interpretation would not account for the significant amount of genetic variance found in body size, as fluctuating selection is as efficient as directional selection in removing additive genetic variance. More likely, the answer may lie in the complex nature of the body mass variable. Besides its structural component (likely the one responsible for heritability), mass possesses components more likely to be under environmental determinism (e.g. differences in energy stores, stemming from differences in territory quality). The component of body mass that is actually selected for larger values (presumably energy stores) may thus bear no heritability, which would prevent evolution towards larger values.

Acknowledgments

We thank M. Reuter, W. Blanckenhorn, P. Fontanillas, L. Lawson and L. Braendli who made helpful comments on the manuscript and L.C. Duarte, S. Bocherens and M. Ehinger for their help in field work.

Ancillary