*Correspondence author and present address: Institute of Evolutionary Biology, University of Edinburgh, West Mains Road, Edinburgh EH9 3JT, UK; E-mail: email@example.com
1Differential male reproductive success is commonplace in mammals and frequently attributed to variation in morphological traits that provide individuals with a competitive advantage in female defence mating systems. Other mammalian mating systems, however, have received comparatively little attention and correlates of male reproductive success in them are less well understood.
2We studied a free-ranging population of North American red squirrels (Tamiasciurus hudsonicus Erxleben) exhibiting year-round individual territoriality. Males must temporarily vacate their territories to locate spatially dispersed receptive females, thereby setting the stage for a scramble competition mating system.
3We predicted that both male annual mating success (measured as the number of females copulated with) and annual reproductive success (measured as the number of offspring sired) would be positively correlated with both search ability (measured as the number of oestrous females located over the mating season) and effort (measured as mating season home range size), generating directional sexual selection on these two metrics.
4Mating season home ranges of males showed, on average, an almost 10-fold increase relative to those measured during the nonmating season, while those of females showed a more moderate twofold increase and both annual mating and reproductive success of males was positively correlated with search ability and search effort.
5The spatial dispersion of females, resulting from the strict territorial social structure of red squirrels, gave rise to a predicted scramble competition mating system. Furthermore, the strength of sexual selection on behavioural traits in this mating system equalled previous estimates for morphological traits in female defence mating systems.
The concept of sexual selection was originally formulated by Darwin (1871) to account for the maintenance of conspicuous sexually dimorphic traits not favoured by natural selection. It is not surprising then that sexual selection theory has primarily advanced through the study of model systems with prominent morphological dimorphism. In mammals, differences in reproductive success among males can be pronounced (LeBoeuf 1974; Clutton-Brock, Guinness & Albon 1982) and traditional explanations for these asymmetries rely on variation in traits that enable males to defend sexually receptive females from rival males (Clutton-Brock et al. 1979; Clutton-Brock 1988; Andersson 1994). Although the overt nature of male–male combat, and its associated traits, have led to female defence being the most studied of all mammalian mating systems, it is important to realize that it represents only one point along a continuum. If females are clumped in space and/or come into oestrus asynchronously, dominant males should be able to defend access to available females, giving rise to a female defence mating system (Emlen & Oring 1977; Ims 1988). Conversely, if females are spatially dispersed and/or breed synchronously they become economically indefensible and a scramble competition mating system is thought more likely (Ims 1988; Murphy 1998).
We studied a free-ranging population of North American red squirrels (Tamiasciurus hudsonicus Erxleben; hereafter red squirrel) to provide information on its mating system and test the hypothesis that male searching behaviour is under positive sexual selection. Red squirrels at our Yukon study site rely primarily on the seeds of white spruce (Picea glauca Moench), which they hoard in a central food cache (midden; Smith 1968). These middens are thought to be necessary for overwinter survival (Rusch & Reeder 1978; McAdam & Boutin 2003) and form the centre of individual territories, defended against members of both sexes year round (Smith 1968; Price, Boutin & Ydenberg 1990). As a result, males must temporarily vacate their territories to locate spatially dispersed receptive females, thereby setting the stage for a scramble competition mating system. Studies of other Sciurids, however, (e.g. Farentinos 1972; Thompson 1977; Koford 1982; Wauters, Dhondt & Devos 1990; Koprowski 1993; Schulte-Hostedde & Millar 2004) have suggested elements of both female defence and sperm competition. In those species exhibiting a female defence component, males congregate in the vicinity of females approaching oestrus and more dominant males attempt to maintain exclusive reproductive access. In addition, intraspecific variation in testis size has previously been shown to influence reproductive success (Schulte-Hostedde & Millar 2004). Therefore, we also examined to what extent, if any, female defence and sperm competition may play a role in the mating system of red squirrels. Finally, we quantified the strength and shape of selection acting on male searching behaviour with standardized selection gradients (Lande & Arnold 1983).
Materials and methods
A population of red squirrels near Kluane National Park in the south-west Yukon (61° N, 138° W) was studied from 2003–05. Individuals were resident on two 40-ha study grids bisected by the Alaska Highway. Details of the study landscape (Krebs, Boutin & Boonstra 2001) and population (e.g. Humphries & Boutin 2000; Boutin et al. 2006; McAdam et al. 2007) have been reported previously. Briefly, individuals were captured in live traps (Tomahawk Live Trap, Tomahawk, WI, USA) placed on, or in the immediate vicinity of, their middens. All individuals born into the study population were originally handled in their natal nest and had there received unique alphanumeric ear tags. Ages of such focal males (N = 68) were known with certainty. Any immigrating adults received ear tags on first capture and for these focal males (N = 38) age is treated as a minimum estimate (i.e. age = 1 in year of first capture). Restricting our analyses to include only males of known age did not influence any of our main conclusions. Each individual was also given a unique combination of one or two coloured wires, threaded through their ear tags, on first capture of the season to allow for identification from a distance. Male body mass (to the nearest gram, using a 500 g Pesola spring scale) was recorded at each capture in 2003–05, and in 2004 and 2005 testis size was also measured (to the nearest millimetre, using a ruler placed along the long axis of the left testis). Testis size was repeatable across time (Pearson's r64 = 0·46) and the means of all measurements of an individual male (body mass: range = 1–16 measurements; testis size: range = 3–12 measurements) taken during the mating season, for both traits, were used in analyses. The Biosciences Animal Policy and Welfare Committee at the University of Alberta approved all protocols.
mating season observations
Details of the mating season for this population, and observations employed, are provided elsewhere (Lane et al. 2007; Lane et al. 2008). Briefly, mating usually commences in mid-late winter (Lane et al. 2007; S. Boutin, unpublished data) with selection favouring early breeding by females (Réale et al. 2003). Females typically produce a single litter each year after a 35-day gestation period, but will occasionally attempt a second litter after litter loss and, rarely, following a successful litter (Boutin et al. 2006). Behavioural oestrus commences with the receptive female emerging from the nest in the morning and continues until the female retires to the nest in the evening, thereby encompassing the full activity period for 1 day. All resident females were monitored daily for reproductive activity in 2003–05. In 2003, 46 mating chases were monitored through to completion. In 2004 and 2005, owing to logistical constraints, a smaller sample was monitored [23 (2004), 16 (2005)]. In all years, focal females were chosen randomly from the total population.
For focal mating chases, a combination of focal-animal sampling of the oestrous female, scan sampling for attending males and all-occurrence sampling of mating behaviour was employed (Altmann 1974; Martin & Bateson 1986). This protocol allowed for the identification of both attending and copulating males and, by extension, the total number of oestrous females located, and copulated with, by individual males over the mating season. Because a large number of copulations occur out of sight (e.g. in burrows on the oestrous female's midden, under downed woody debris and in snow tunnels), these copulations were judged as occurring when a female was followed by a male to an out-of-sight location and the pair remained there for a minimum of 60 s or if copulatory vocalizations could be heard, regardless of the time spent out of sight. Copulations of tree squirrels generally last less than 60 s (Koprowski 2007) and this duration is adequate for fertilization in red squirrels (J. E. Lane & S. Boutin, unpublished data). This criterion has been used previously in this (e.g. Lane et al. 2007) and other (Waterman 1998) systems.
searching behaviour metrics
Two searching behaviour metrics were quantified: search ability and search effort. The number of oestrous females located by individual males [N = 69 (2003), 47 (2004), 48 (2005) males analysed] over the mating season each year was taken to represent search ability (sensu Schwagmeyer & Woontner 1986). As the number of mating chases followed varied across the 3 years, each male's value was standardized by dividing it by the annual population mean. The home range size of individual males [N = 63 (2003), 41 (2004), 42 (2005)] during the mating season was taken to represent search effort. To examine whether male and female red squirrels exhibit the space-use patterns across seasons typical of scramble competition mating systems, namely that the home ranges of males greatly expand during the mating season while those of females remain relatively consistent, the home range sizes of males during the non-(post-)mating season [N = 57 (2003), 39 (2004), 51 (2005)] and females during both seasons [mating: N = 62 (2003), 49 (2004), 61 (2005); nonmating: N = 53 (2003), 53 (2004), 52 (2005)] were also calculated. Study grids were staked and flagged at 15-m or 30-m intervals to provide spatial references for all location data and three types of spatial data were compiled for home-range analyses. In both seasons, and for both sexes, the trap locations of individuals were recorded and the location of any individuals seen opportunistically (for example, during regular trapping rounds) was noted. A random sample of individuals [males: N = 43 (2003), 32 (2004), 27 (2005); females: N = 69 (2003), 49 (2004), 63 (2005)] was also outfitted with radiocollars (model PD-2C, 4 g, Holohil Systems Limited, Carp, ON, Canada) and radiotelemetry was used to augment our spatial data. During the mating season, point locations were obtained (2003) or behavioural focals conducted (2004 and 2005; recording spatial locations every 30 s for 10 min) for males and point locations in each year (2003–05) were obtained for females. Post-mating season spatial data for females were also provided, through coordination with other projects, in the form of behavioural focals (recording spatial locations every 30 s for 10 min) and nest site locations. Red squirrels become accustomed to human observers rapidly following a brief period of acclimation (Smith 1968). This population has been the focus of research for 20 years, and behavioural focals have been employed since its inception (e.g. Stuart-Smith & Boutin 1995; Humphries & Boutin 2000; Lane et al. 2008). We are thus confident that our presence did not affect natural ranging behaviour of individuals.
Home range sizes were estimated with minimum convex polygons (MCP). While 95% MCPs are traditionally used for home range analyses (White & Garrott 1990) and are less vulnerable to being confounded by the number of data points used in the calculation, off-territory excursions, most relevant to our hypotheses, are largely excluded from these calculations. All data were thus retained and 100% MCPs were calculated. To cope with its potential confounding influence, the number of data points used (range: 3–485) was included as a covariate in all analyses following Kenward (2001).
molecular analyses and paternity assignment
Details of the molecular methods for microsatellite loci isolation and paternity assignment have been provided elsewhere (Gunn et al. 2005; Lane et al. 2008). Briefly, DNA was extracted from preserved tissue using either an acetate-alcohol precipitation protocol (Bruford et al. 1998) or DNeasy Tissue extraction kits (QIAGEN, Venlo, The Netherlands), and polymerase chain reaction (PCR) amplification was performed for a panel of 16 microsatellite loci [for details of microsatellite loci used, including the number and size range of alleles at each locus, as well as observed (HO) and expected (HE) heterozygosities, see Lane et al. (2007)]. Separate analyses were conducted for each of the years 2003, 2004 and 2005. Maternity was determined by behavioural observation at the nest and paternity was assigned at 95% confidence using Cervus 2·0 (Marshall et al. 1998). We were able to assign paternity to 327 of 507 (64%) of the offspring produced over the 3 years. In the other 180 cases the delta value, provided by Cervus 2·0, was too low to be able to assign paternity at 95% confidence.
Both male relative annual mating (i.e. annual number of females copulated with divided by the population mean) and relative annual reproductive success (i.e. annual number of offspring sired divided by the population mean) were calculated. It was necessary to calculate the relative success of males, rather than analyse the raw data, because we pooled data across 3 years that varied in theoretical maximum success of males. For example, because we observed the mating chases of 46 females in 2003, the theoretical maximum mating success of males (46 females copulated with) was double that of males in 2004 when 23 mating chases were followed. The opportunity for sexual selection on both success metrics (I) as the square of the coefficient of variation [(standard deviation/mean)*100] (Crow 1958; Arnold & Wade 1984a,b) was also calculated.
statistical and selection analyses
A linear mixed-effects model (LME) was used to determine whether sex and/or season influenced home range size. Home range sizes varied over five orders of magnitude and were thus log-transformed before analysis as suggested by Kenward (2001). The log-transformed number of data points used in the MCP calculations was included as a covariate. Sex and season were fitted as fixed effects and the sex*season interaction was included. Individuals (male: N = 129; female: N = 113) are represented in multiple years and across both seasons. Individual ID was therefore fitted as a random effect. Year (2003, 2004, or 2005) and study grid were included as three and two-level, respectively, categorical explanatory terms in the model.
To investigate whether either of our search metrics explained male success, generalized linear mixed-effects models (GLMMs) with Poisson error structures were used. In all cases, study grid was tested as a two-level categorical variable and minimum age, (minimum age)2 and body mass were included as covariates. Testis size was also fitted as a covariate in analyses of male relative annual reproductive success. Treating testis size as the residual from a mass/testis size regression did not change our main conclusions. The quadratic term for age was included to model any potential curvilinear relationship in male performance with age (i.e. senescence). In our first two analyses, whether either search ability or search effort influenced relative annual mating success was tested. In the next two analyses, whether either of these two variables influenced relative annual reproductive success was tested. The number of females that a male located divided by the annual population mean (i.e. search ability) and the log-transformed home range sizes (i.e. search effort), respectively, were fitted as the fixed effects. The number of females copulated with divided by the annual population mean represented the dependent variable in the mating success analyses and the number of offspring sired divided by the annual population mean represented the dependent variable in the reproductive success analyses.
Selection on the two searching traits was calculated as standardized linear and nonlinear selection gradients (Lande & Arnold 1983). The standardized linear (i.e. directional) selection gradients were obtained from separate multiple regressions of annual reproductive success on either the search ability or the search effort of males as well as average body mass and testis size. Trait values were standardized to a mean of zero, and relative fitness was calculated within each year to avoid confounding selection gradients with covariances between environmental conditions, trait values and reproductive success. Because standardizing trait values and relative fitness within the year were inherent in our selection analyses, the selection gradients were calculated on the raw data rather than the values divided by the population mean as in our GLMM analyses. Quadratic i.e. stabilizing or disruptive) and correlational i.e. indirect) selection gradients were estimated in separate linear models that included all linear, quadratic and pairwise cross-products of traits. Standard errors were generated by jackknifing and the probability values associated with the jackknife estimates are reported, rather than using the sequential Bonferroni correction, which is not appropriate for correlated variables (Rice 1989). All analyses were implemented in r (version 2·6·0; R Core Development Team, Boston, MA, USA) and values are presented as mean ± 1 S.E.
Mating commenced in mid-late winter of each year (2003: 19 February; 2004: 5 March; 2005: 21 March). In 2003 and 2004, the mating seasons extended to the late spring (2003: 4 June; 2004: 19 May), while in 2005, it extended to mid-summer (17 July). The latter mating season was characterized by five preceding years of low food and was preceded a mast crop of high food abundance. Many females (as described in Lane et al. 2008) lost their earlier litters and re-cycled and females also re-cycled following successful litters (as described in Boutin et al. 2006) this year. Females entered oestrus asynchronously, resulting in temporally dispersed reproductive opportunities for males and a male-biased daily operational sex ratio (25 males/female; Lane et al. 2008).
The home ranges of the two sexes showed dissimilar patterns across the two seasons (sex*season: LME: F1,409 = 112·87, P < 0·001; Fig. 1). During the non-(post-)mating season, male home ranges averaged 6194 ± 887 m2 (N = 147) and during the mating season, they expanded by almost 10-fold (mean = 59827 ± 3416 m2; N = 177). These values represent multiples of 1·5 and 14·6, respectively, of average male territory sizes (i.e. actively defended, core, areas; LaMontagne 2007). Female home ranges, by contrast, varied less throughout the year (Fig. 1). During the nonmating season, female home ranges averaged 8470 ± 781 m2 (N = 158), while during the mating season they expanded to 16189 ± 1891 m2 (N = 172). These values represent multiples of 2·1 and 3·9 of the average female territory sizes (LaMontagne 2007).
Male relative annual mating success was correlated with our measures of both search ability (Fig. 2a) and search effort (Fig. 2b; Table 1). In contrast, mating season body mass, minimum age and (minimum age)2 showed weaker and inconsistent relationships with relative annual mating success. In the search ability analysis, neither minimum age, nor (minimum age)2 were related to relative annual mating success while mating season body mass was negatively correlated. In the search effort analysis, however, minimum age and (minimum age)2 were correlated with relative annual mating success, while mating season body mass was not correlated.
Table 1. Coefficients, standard errors and statistical significance (including Wald statistics and P values) for terms included in the two GLMM models (search effort and search ability) on male mating success in North American red squirrels
Relative number of oestrous females located (number of oestrous females located/annual population mean)
Log (home range size)
Male relative annual reproductive success was similarly correlated with our measures of both search ability (Fig. 2c) and effort (Fig. 2d; Table 2). In the search ability analysis, minimum age and (minimum age)2 were correlated with relative annual reproductive success while mating season body mass was not. In the search effort analysis, however, none of minimum age, (minimum age)2 or mating season body mass were related to relative annual reproductive success. Testis size was unrelated to the relative annual reproductive success in the search ability analysis but was positively correlated in the search effort analysis.
Table 2. Coefficients, standard errors and statistical significance (including Wald statistics and P values) for terms included in the two GLMM models (search effort and search ability) on male reproductive success in North American red squirrels
Relative number of oestrous females located (number of oestrous females located/annual population mean)
Log (home range size)
The opportunity for sexual selection on males was 0·64 for relative annual mating success and 1·52 for relative annual reproductive success, and sexual selection favoured both male search ability and effort (Table 3). The selection gradient for search ability on relative annual reproductive success P = 0·006) was ‘very strong’ (β′ > 0·50; Kingsolver et al. 2001), while that for search effort P < 0·002) fell just below this cut-off. Selection did not show any nonlinear (stabilizing or disruptive) elements.
Table 3. Standardized linear, quadratic and correlational selection gradients (± SE) for male searching ability (number of oestrous females located) and effort (log home range size) based on male annual reproductive success (number of offspring sired). Sample sizes are included in parentheses
Despite this, red squirrels exhibited an almost exclusively scramble competition mating system. Females showed a modest increase in home range size from the nonmating to mating seasons, potentially because they were searching for vacant territories to subsequently relinquish to offspring (Boutin, Larsen & Berteaux 2000) or because they were pilfering food stores from vacant male neighbours (Gerhardt 2005). Male home ranges, by contrast, expanded by almost 10-fold as they vacated their territories to search for females in or approaching oestrus. Furthermore, both male relative annual mating and reproductive success were correlated with male searching behaviour and body mass was, if anything, only weakly negatively correlated with mating success.
A key element of the ecology of red squirrels likely preventing males from attempting to defend females approaching oestrus is the spatial dispersion of females and their territorial social structure. With the higher degree of sociality exhibited by previously studied sciurids, males are able to congregate in the area surrounding a female in the days preceding her oestrus. Under these conditions, older males tend to dominate younger competitors and maintain exclusive access to the female, consequently achieving higher mating success (e.g. Wauters & Dhondt 1989; Wauters et al. 1990; Wauters & Dhondt 1992). For male red squirrels, however, this would require abandoning defence of their territories, while incurring aggression from the territory owners neighbouring the female. Thus, although reproductive opportunities are temporally dispersed, the strict territoriality of red squirrels in our study population prevents males from attempting to defend females approaching oestrus, resulting in the observed scramble competition mating system.
Patterns of reproductive success in red squirrels follow theoretical predictions (Bateman 1948; Trivers 1972; Emlen & Oring 1977). Although the opportunity for sexual selection on males was moderate, relative to previously studied female defence mating systems (e.g. Coltman et al. 2002), it is approximately threefold larger than that for females on annual mating success (0·23) and 10-fold larger than that for females on annual reproductive success (0·16) (J. E. Lane & S. Boutin, unpublished data). In contrast to female defence systems (e.g. Kruuk et al. 2002), however, sexual selection more strongly favoured behavioural, in comparison to morphological, traits.
The strength of selection on both male searching behaviour metrics is similar to previous estimates for secondary sexual characters in female defence mating systems. For example, using lifetime reproductive success, the standardized selection gradient on antler size in male red deer (Cervus elaphus L.) was found to be 0·44 (± 0·18 S.E.) (Kruuk et al. 2002). Whether selection on searching behaviour will be weaker when considering lifetime, rather than annual, reproductive success in red squirrels is currently unknown. On the other hand, the strength of selection we observed may represent an underestimate for scramble competition mating systems in general. In thirteen-lined ground squirrels (Spermophilus tridecemlineatus Mitchill), for example, males who exhibit greater search ability are more likely to arrive early at females’ mating chases (Schwagmeyer et al. 1998) and a queuing convention promotes early mating by these males relative to later-arriving competitors (Schwagmeyer & Parker 1987). Unlike red squirrels, which do not exhibit patterns of sperm precedence (sensu Lacey, Wieczorek & Tucker 1997; J. E. Lane & S. Boutin, unpublished data), a first-male precedence is apparent in thirteen-lined ground squirrels and this further strengthens the competitive advantage to early arriving males (Foltz & Schwagmeyer 1989). Thus, the tight linkages between searching behaviour, annual mating success and, ultimately, annual reproductive success provide consistent directional benefits to early arriving males. We would thus predict that the strength of selection on searching behaviour should be even stronger for thirteen-lined ground squirrels than red squirrels.
Although the searching behaviour of males was the primary influence on both measures of success, it had a relatively weaker effect on relative annual reproductive, as compared to mating, success. This is perhaps not surprising, given the high levels of multiple male mating by females (Lane et al. 2008). The mean number of males mated with by females in this population is seven, while the mean litter size is only three (McAdam et al. 2007; Lane et al. 2008). Consequently, a large number of copulating males fail to sire offspring, suggesting that sperm competition is likely an important component of the red squirrel genetic mating system. Testis size was, at most however, a minor influence on relative annual reproductive success. This discrepancy could be due to sperm quality (or compatibility) being a more important determinant than quantity. Females in other species show selection for both good and compatible genes (Olsson et al. 1996; Neff & Pitcher 2005) and although female red squirrels do not bias paternity based on genetic relatedness (Lane et al. 2007), they may use more sophisticated cues (e.g. selection based on MHC genotypes; Neff & Pitcher 2005). Alternatively, the temporal dispersion of reproductive opportunities for male red squirrels may place little draw on their spermatogenesis abilities, thereby requiring comparatively smaller testes and consequently testis size may be an inaccurate reflection of ejaculate investment for this species.
Our analyses revealed that the searching behaviour of male red squirrels is under positive sexual selection. The proximate mechanism(s) underlying the observed variation in searching behaviour, however, is currently unknown. We can envision at least three non-exclusive candidate hypotheses. First, recent work in this population has identified a shy-bold continuum in female temperament that is repeatable across time (Boon, Réale & Boutin 2007). Should males also exhibit this continuum, bold males may be more predisposed to venture off-territory in search of reproductive opportunity. Second, males may vary in spatial ability and/or memory which affects their capacity to find and/or remember the location of females approaching oestrus, as has been shown in other species under both laboratory (Spritzer et al. 2005b) and field conditions (Schwagmeyer et al. 1998). Finally, annual searching behaviour may be a plastic response reflecting temporal variation in condition and/or residual reproductive value. Red squirrels are relatively long-lived for a small mammal, with a maximum lifespan of 9 years (McAdam et al. 2007). Males in relatively poor condition may, consequently, invest less in current, so as to not sacrifice future, reproduction. Future work on inter-individual variation in temperament, spatial ability/memory and life-history patterns on males in this population will test these candidate hypotheses.
Due to the complexities and subtleties inherent in how males achieve success within scramble competition mating systems, our understanding of sexual selection on these traits lags considerably behind relative to other systems (e.g. Maynard Smith 1985; Kruuk et al. 2002). This not only influences our understanding of mammalian sexual selection, but also has important implications for inferences regarding the evolution, genetic structure and ecology of populations. For example, differential patterns of male reproductive success between female defence and scramble competition mating systems should have direct consequences for the fine-scale genetic structure of populations. In female defence systems, reproductive skew accentuates the genetic structure because successful males create clusters of relatives (Nussey et al. 2005), whereas in scramble competition systems the most successful males are those that range the furthest and therefore disrupt the genetic structure. In addition, the more moderate skew among male reproductive success in scramble competition, relative to other mating systems, lessens the influence of male reproductive success on effective population sizes (Sugg & Chesser 1994). These predictions are in accordance with previous analyses showing that, despite strong philopatry, females mate randomly with respect to genetic relatedness (Lane et al. 2007). We suggest that future studies investigate the ways in which males achieve reproductive success in less represented mating systems. A better appreciation of the strength and shape of sexual selection on a more diverse suite of morphological and behavioural traits should aid our understanding of mammalian reproductive ecology and evolution.
We thank the 2003–2005 members of Squirrel Camp for help with fieldwork and to Elizabeth Anderson and Ainsley Sykes for coordinating research efforts. Katie Hartnup assisted with genotyping for the molecular paternity analyses. Andrew McAdam provided instruction on the selection analyses. This study was supported by funds from a Natural Sciences and Engineering Council of Canada (NSERC) Discovery Grant (S.B), National Environment Research Council Standard Research Grant (D.W.C) and a Department of Indian and Northern Affairs, Canada Northern Scientific Training Program Grant (J.E.L). An NSERC Postgraduate Scholarship-B and Alberta Ingenuity Fund studentship supported J.E.L. This is paper no. 44 from the Kluane Red Squirrel Project.