SEARCH

SEARCH BY CITATION

Keywords:

  • body size;
  • cricket;
  • encapsulation;
  • genetic correlation;
  • haemocyte load;
  • heritability;
  • immune function;
  • sexual selection

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgments
  9. References

Female house crickets are attracted to male calling song containing a relatively high number of syllables per ‘chirp’, which tends to be produced by large males. In a previous study, we showed that this song characteristic is also positively and independently correlated with haemocyte load, an important determinant of the ability to produce an encapsulation response in insects. Females will therefore tend to select males with high encapsulation ability (and large body size) as mates. The present study demonstrates that variation in haemocyte load and body size, together with a second parameter of immune function (the ability to encapsulate a synthetic substrate), is heritable in the same population. Moreover, all three traits are shown to be positively genetically correlated. In favouring males that produce calling song with the preferred characteristics, females should therefore also tend to produce larger offspring with a greater ability to produce an encapsulation response.


Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgments
  9. References

According to the ’handicap principle’ of sexual selection (Zahavi, 1975, 1977; Andersson, 1982), mating preferences are selected for in females that favour exaggerated, costly secondary sexual traits in males, because those traits provide reliable signals of some aspect of male quality from which females can benefit. This has also been referred to as the ‘honest advertisement’ hypothesis (Johnstone, 1995). Two distinct forms of benefit have been recognized. First, females may accrue direct, phenotypic benefits by mating with males bearing more exaggerated forms of a particular sexual trait. For example, female tree crickets, Oecanthus nigricornis, preferentially mate with males that produce a calling song with a low carrier frequency (Brown et al., 1996). Male tree crickets feed females during copulation with a secretion which enhances their fecundity. Low frequency song tends to be produced by males that are capable of producing a superior quality secretion. Consequently, a preference for low frequency song leads to higher female fecundity.

In many cases, however, there do not appear to be any direct benefits associated with mate choice. This raises the possibility that females can acquire indirect benefits by mating with genetically fitter males. Specifically, if variation in male quality – and ultimately in fitness – is heritable, females may be able to increase the fitness of their offspring by mating with males bearing a ‘quality dependent’ trait (Johnstone, 1995). It is a long standing tenet of quantitative genetic theory that fitness should have zero heritability when a population is at equilibrium (because, by definition, selection should tend to deplete additive genetic variation most rapidly in fitness itself; Fisher, 1930; Falconer & Mackay, 1996). However, several lines of research suggest that sufficient variation may persist in fitness-related traits to generate selection on female mating preferences. Hamilton & Zuk’s (1982) model of pathogen-mediated sexual selection provides perhaps the best known example. Pathogens tend to have shorter generation times than their hosts and can be rapidly selected to overcome their immune defences. Consequently, these organisms may tend to maintain variability in the host population at loci that influence immunity. Thus, the existence of genetic variation in immune resistance would be expected to generate indirect selection on female mating preferences, if females expressing a preference for males with higher resistance produce fitter offspring.

The exaggerated, costly secondary sexual traits that are so often the targets of female mating preferences are particularly suited to the task of reflecting an individual’s past or present infection status. For example, if pathogenic infection reduces the availability of energy or nutrients required for sexual trait expression (e.g. Mappes et al., 1996), males that are genetically more prone to becoming infected may be less able to invest in the sexually selected trait. Folstad & Karter (1992) suggested that males with low resistance ability might be further constrained from investing in sexual traits if sexual trait expression must be traded-off against immune function. For example, if common resource costs underpin each system, only relatively resistant males may be able to risk compromising immune function by allocating essential resources into the preferred trait (Sheldon & Verhulst, 1996).

The field and house crickets (Orthoptera: Gryllinae) provide a good example of a taxon where females frequently demonstrate mating preferences for males capable of expressing a particular trait, and yet apparently receive little or no direct benefit through doing so (Zuk & Simmons, 1997). Females select males from within calling aggregations on the basis of specific calling song characteristics. Female house crickets, Acheta domesticus, favour the calling songs of large males, which contain more syllables per ‘chirp’ (Gray, 1997). Recently, Ryder & Siva-Jothy (2000) demonstrated that this song character is also correlated with underlying variation in immune function in A. domesticus once variation in male size has been controlled for. In particular, males that call with more syllables per chirp tend to have higher ‘haemocyte loads’. Haemocyte load has a strong influence on the ability of an insect to produce an encapsulation response (e.g. Salt, 1963; Götz, 1986; Eslin & Prévost, 1996, 1998; Kraaijeveld et al., 2001). Encapsulation serves as the front line of defence against multicellular insect pathogens such as parasitoids, nematodes and fungi, which can cause significant mortality in natural populations of crickets (Walker & Masaki, 1989). Briefly, layers of haemocytes (immunologically reactive blood cells) attach to the surface of the pathogen once it enters the haemocoel. Haemocytes within the inner layers of the developing capsule then undergo a process of melanization and produce cytotoxic compounds which are thought to contribute to the pathogen’s death (Gillespie et al., 1997; Ryder, 1999).

There are few obvious direct benefits associated with the preference of female A. domesticus for males that produce calling song with more syllables per chirp. However, given that this song character is correlated with immune function in our study population, a preference for males that produce more syllables per chirp may lead to the production of offspring with greater pathogen resistance ability if immune function is heritable. The aim of the present study, therefore, was to estimate the narrow-sense heritabilities for, and the respective genetic correlations between, male size and two separate measures of the ability to produce an encapsulation response, within the same population. First, we measured haemocyte load, because this parameter is strongly associated with encapsulation ability in insects and has been shown to be correlated with the preferred song characteristic. Secondly, we measured the magnitude of the encapsulation response produced towards a synthetic substrate implanted into the haemocoel. Synthetic substrates have been used successfully in a number of insect species to provide a standardized immune challenge against which the encapsulation abilities of different individuals can be compared (e.g. König & Schmid-Hempel, 1995; Gorman et al., 1996). Although synthetic, these substrates provide biologically meaningful information because the basis of nonself recognition is relatively nonspecific in insects (Gillespie et al., 1997). Acheta domesticus have been shown to produce a well-defined, haemocytic encapsulation response towards synthetic substrates (Ryder, 1999; Ryder & Siva-Jothy, 2000).

Methods

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgments
  9. References

Insects

Acheta domesticus were derived from stock cultures that have been maintained in the Department of Animal & Plant Sciences at The University of Sheffield, England, for 10 years. The stocks were originally derived (approximately 200 individuals) from a commercial supplier (Blades Biological, Leeds, England) and have since been maintained in an insectary under hygienic conditions with shelter and ad libitum food and water. A single breeding stock was maintained at a density of approximately 300 adults (1:1 sex ratio). Eggs were collected from the breeding stock on a fortnightly basis and used to establish fresh larval cultures (2–3 tanks per egg batch, with between 100 and 300 individuals per tank) with which to maintain the breeding stock.

Experimental set up

Heritabilities were estimated for haemocyte load, the ability to encapsulate a synthetic substrate, and body size using a nested full-sib, half-sib breeding design (Lynch & Walsh, 1998). Appropriate sample sizes were determined using a power analysis (see below). Twelve virgin adult males (‘sires’) were selected at random from stock cultures. Each male was placed into its own container, supplied with ad libitum food and water, in a constant temperature room maintained at 28 °C with a 12:12 h light:dark regime. Ten virgin adult females (‘dams’), also selected at random, were placed into each container and left undisturbed for 7 days to allow mating to proceed. Each female was then removed and transferred to its own separate container of oviposition substrate, at which point a second group of 10 virgin adult females were introduced to each male’s container and allowed to mate. The eggs from all females were incubated separately for approximately 12 days, at which point hatching began. Hatching was allowed to proceed for 4 days, after which time most eggs had hatched successfully. Twenty hatchlings from each female’s clutch were selected at random and transferred to a new container, i.e. one container of 20 full-sib offspring per female. These were maintained in the constant temperature room with access to shelter and ad libitum food and water. Once the developing nymphs had reached the final larval instar (at which time females posses a partially developed ovipositor), six males were selected at random and the remainder were discarded. Only males were included in the subsequent analysis in order to eliminate sex-linked effects.

Measurement of immune function and male size

Males were measured for each trait whilst between the ages of 10 and 30 days post-eclosion. Two parameters of immune function were estimated for each male: the ability to produce an encapsulation response towards a synthetic substrate (nylon); and ‘haemocyte load’– the number of haemocytes per unit volume of haemolymph. Measurement details and spatial and temporal repeatabilities for each parameter are given in Ryder & Siva-Jothy (2000). Briefly, males were removed from the constant temperature room and chilled on ice for 20 min. Each was swabbed with 70% ethanol and a single length of surface-sterilized 2 mm nylon monofilament (0.16 mm diameter) was inserted into the haemocoel through a small incision made between the eighth and ninth abdominal sternites. Males were then returned to the constant temperature room and left undisturbed for 48 h. Each was weighed at this time to the nearest mg on a Mettler (Switzerland) PM480 balance. Fresh weight at this age is highly correlated with fresh weight at eclosion (J.J. Ryder, personal observation), which Gray (1997) showed to be the strongest predictor of body size in A. domesticus. In order to estimate haemocyte load, males were again chilled on ice for 20 min and a 2 μL haemolymph sample was taken from a severed hind leg wound using a Gilson P10 pipette; this was pipetted immediately into 18 μL of chilled anticoagulant and mixed thoroughly. Haemocyte load was estimated by determining the number of haemocytes per mm3 of haemolymph using a haemocytometer under a compound microscope (Leitz Diaplan, Switzerland). Encapsulated nylon implants were recovered under a dissecting microscope (Wild M8, Switzerland). The volume of haemocytic material encapsulating the implants (‘capsule volume’) was taken to provide an estimate of the ability of encapsulate this substrate and was estimated using the same technique reported in Ryder & Siva-Jothy (2000).

Statistical analysis

In order to determine a sample size that would be sufficient for detecting modest heritabilities (given the putative correlations between immune function, size, and fitness) a power analysis was performed according to the procedures outlined in Lynch & Walsh (1998). This indicated that a design using 12 sires, each mated to 14 dams, with four offspring per dam (total of 12 × 14 × 4=672 offspring) would have 95% power to detect a heritability of 0.3. Logistic considerations dictated that this was the largest workable sample size. At the start of the experiment, therefore, 20 females (in two groups of 10) were placed into each male’s container in order to ensure that the minimum number of 14 mated females would be available; the offspring of 14 of these 20 females were reared to adulthood. Similarly, six male offspring were collected from each female in case any failed to survive eclosion; only four were included in the analysis.

Heritabilities were estimated separately for each trait following a nested analysis of variance (Lynch & Walsh, 1998). The significance of the sire and dam mean squares was determined by the appropriate F-test in the ANOVA. The total variance in each trait was partitioned into an among-sire component and an among-dam, within-sire component, each yielding a different estimate of heritability. The among-sire component provides a basis for estimating ‘narrow-sense’ heritability (h2), because it only contains variance arising from additive genetic effects (assuming epistatic effects to be negligible); the among-dam, within-sire estimate also contains variance arising from nonadditive genetic, maternal and common environmental effects. Standard errors were calculated according to Lynch & Walsh (1998). In order to determine whether any of the traits were genetically correlated, a nested analysis of variance was carried out for the cross-products derived from each trait combination (Lynch & Walsh, 1998): haemocyte load and capsule volume; haemocyte load and male size; and capsule volume and male size. The genetic correlations between these traits were estimated according to Lynch & Walsh (1998). Approximate standard errors were calculated according to Falconer & Mackay (1996). Additive genetic and residual coefficients of variation (CVA and CVR, respectively) were also calculated for each trait after Houle (1992), where CVA=100 (sqrt VA)/trait mean; CVR=100(sqrt [VP − VA])/trait mean (VA=additive genetic variance; VP=total phenotypic variance). Analyses were performed using Minitab statistical software (version 12.1 for the PC). All data were checked for normality and homogeneity of variances (Underwood, 1998).

Results

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgments
  9. References

Heritabilities

There was a highly significant effect of the ‘sire’ factor for each trait (haemocyte load, capsule volume and male size; Table 1) demonstrating the presence of heritable variation among males. Additive genetic effects comprised an estimated 29.7% of the total phenotypic variation for haemocyte load, 20.3% for capsule volume and 45.6% for male size (sire estimates of ‘narrow-sense’h2).

Table 1.   Nested analyses of variance for haemocyte load, capsule volume and male size, showing heritabilities (h2) ± SE. Thumbnail image of

The standard errors for each of these estimates were quite high, making it difficult to compare heritabilities between the three traits with any degree of certainty. Moreover, the associated confidence intervals (approximated by multiplication of each standard error by 1.96; Lynch & Walsh, 1998) just encompass zero in each case. This was surprising given the highly significant sire effects noted above. However, our failure to obtain significant heritability estimates per se (with confidence intervals clearly above zero) may result from a degree of non-normality within each data set. Although the data were generally normally distributed within sires (Anderson–Darling normality test), they were slightly but significantly non-normal when the data were pooled for each trait (male size was slightly platykurtotic; haemocyte load and capsule volume were slightly leptokurtotic; capsule volume was also slightly positively skewed). Transformation proved unsuccessful in each case. As the precision of variance estimates derived from an ANOVA can be sensitive to non-normality (Sahai & Ageel, 2000), this may have lead us to under-estimate the heritabilities, or over-estimate the respective standard errors. However, in the present context, the presence of additive genetic variation is more important than its absolute magnitude. ANOVA is robust to reasonably small deviations from normality, particularly where sample sizes are large and the design is balanced (Underwood, 1998): the F-test in the ANOVA thus provides a robust demonstration of such variation, despite the lack of precision reflected in the standard errors.

The among-dam, within-sire heritability estimates lend support to this view. These accounted for 75.1% of the total phenotypic variation for haemocyte load, 31.4% for capsule volume and 102.7% for male size, with approximately equivalent standard errors. Although these estimates reflect additional contributions from nonadditive genetic, maternal and common environmental sources of variance, that the associated standard errors (and confidence limits) were comfortably above zero suggests that there was a significant contribution from additive genetic effects.

Correlations between traits

There were significant covariances among-sires for each combination of traits (Table 2). The estimates for the genetic correlations were all relatively high and positive, with reasonably low (approximate) standard errors. Once again, the slight non-normality of the pooled data may have reduced the precision of our estimates for the genetic correlations. However, our main concern was the detection of genetic correlations, where present, together with a knowledge of their sign.

Table 2.   Nested analyses of covariance for each combination of traits, showing genetic correlations (rA) ± approximate SE. Thumbnail image of

The strength of the genetic correlations was in marked contrast to that of the respective phenotypic correlations (rP), which, where significant, were only weakly positive (haemocyte load and capsule volume: rP=0.181, d.f.=670, P < 0.001; haemocyte load and male size: rP=0.163, d.f.=670, P < 0.001; capsule volume and male size: rP=0.059, d.f.=670, P=0.126). However, the estimated environmental correlations (rE; the correlation of the environmental and nonadditive genetic deviations; Falconer & Mackay, 1996; Lynch & Walsh, 1998) between these traits were also low (haemocyte load and capsule volume: rE=0.183; haemocyte load and male size: rE=0.066; capsule volume and male size: rE=−0.033). Environmental and nonadditive genetic sources of variation (which clearly comprise a substantial proportion of the total phenotypic variance for each trait; Table 1) thus appear to have influenced each trait differently, and probably weakened the respective phenotypic correlations.

Additive genetic and residual coefficients of variation

The heritability estimates shown in Table 1 appear to mask considerable variation in the respective contributions of additive genetic and residual sources of variation. Thus, capsule volume, which had the lowest heritability, had the highest additive genetic coefficient of variation (Table 3). The low heritability for this trait may, therefore, have resulted from relatively high residual (environmental and nonadditive genetic) variation. Conversely, male size, which exhibited the highest heritability, had the lowest additive genetic coefficient of variation. The coefficients for haemocyte load, which had a heritability intermediate between those for capsule volume and male size, were also intermediate.

Table 3.   Additive genetic (CVA) and residual (CVR) coefficients of variation for each trait. Thumbnail image of

Discussion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgments
  9. References

In this study, we have shown that variation in both encapsulation ability (measured as haemocyte load and ‘capsule volume’) and body size is heritable in our study population of A. domesticus. These traits were also shown to be positively genetically correlated. Each trait is known to be correlated with a sexually selected song character in the same population (Ryder & Siva-Jothy, 2000). Females expressing a preference for males that express that song characteristic should therefore tend to increase both the encapsulation ability and size of their offspring. Given that these traits may both be correlated with components of fitness (e.g. encapsulation ability: Eslin & Prévost, 1996, 1998; size: Simmons, 1986, 1988), we suggest that the potential exists for females to accrue indirect benefits through mate choice. However, several caveats to our results warrant further consideration.

Heritability estimation

Our heritability estimates are restricted to a sib analysis that was based on male offspring only. We can only suggest therefore that additive genetic variation in each trait would generate comparable heritabilities for male and female offspring. However, in support of this hypothesis, Kurtz & Sauer (1999) recently detected very similar heritabilities for male and female Panorpa vulgaris offspring for a haemocyte-based immune response (phagocytosis), when offspring were regressed on their fathers. Interestingly, Kurtz & Sauer (1999) failed to detect significant heritabilities for the same trait when offspring were regressed on their mothers, possibly owing to confounding maternal sources of phenotypic variation. Our results also indicate that maternal factors may have had a considerable effect on phenotypic trait values, as the dam heritability estimates for each trait were consistently higher than the respective sire estimates. However, because the dam estimates reflect unknown degrees of nonadditive genetic and/or common environmental variation – in addition to maternal effects – it is not possible to say which made the larger contribution. For body size in particular, it would appear that the combined effect of these factors was substantial (over and above the demonstrable additive genetic effect), as the dam heritability estimate was extremely high (slightly exceeding unity).

The crickets used in the current study were obtained from stock laboratory cultures that were originally derived from a commercial supplier and subsequently maintained under relatively hygienic conditions. It is, therefore, important to consider the biological relevance of the heritabilities presented here, given that a typical field population would probably tend to suffer higher frequencies of pathogenic challenge (Walker & Masaki, 1989). Recent work suggests that immune function may be negatively genetically correlated with other fitness traits in insects (such as larval competitive ability; Kraaijeveld & Godfray, 1997; Fellowes et al., 1998). Under laboratory conditions, lower rates of pathogensis may, therefore, have resulted in net selection for lower resistance, tending to erode additive genetic variation in immune function and thus reduce heritability. However, some selection for resistance is likely to have continued under laboratory conditions. Crickets frequently wound one another when cultured in the laboratory (typically at relatively high densities), usually during or shortly after moulting when the new cuticle is still soft. Wounds provide a point of open entry into the haemocoel for a variety of potential pathogens that may have been present in our cultures, such as bacteria and entomopathogenic fungi. Haemocytes are also intricately associated with wound healing (Danielli et al., 2000), a process which is vital if an insect is to maintain physiological integrity. Thus, although it is difficult to say precisely how our heritability estimates would compare with those that might have been obtained for a freshly collected field population, they do, nonetheless, demonstrate additive genetic variation in a trait that is likely to be closely associated with offspring fitness.

Genetic correlations

Each of the three traits measured in this study were shown to be genetically correlated. This is perhaps not surprising with regard to haemocyte load and capsule volume. Each parameter was intended to provide a measure of the ability to produce an encapsulation response. That two different aspects of encapsulation ability should be influenced by similar genes (and thus demonstrate correlated additive genetic variation) would be expected. More striking, however, are the genetic correlations between immune function (haemocyte load and capsule volume) and male size. Previous work on a related grylline cricket (Gryllus campestris) suggests that male calling song may signal several different components of ‘quality’– in particular, age, size and bilateral symmetry (although female preferences target a different song characteristic in G. campestris;Simmons, 1995; Simmons & Ritchie, 1996). Similarly, a genetic correlation between body size and encapsulation ability in A. domesticus might suggest that these two traits reflect underlying genetic variation in male quality. For example, the expression of each trait may depend substantially on genetic variation in condition and the ability to acquire resources (Houle, 1991; Rowe & Houle, 1996).

Maintenance of additive genetic variation

In Hamilton & Zuk’s (1982) hypothesis of pathogen-mediated sexual selection, host–pathogen co-evolution maintains the heritable variation in fitness that is necessary to generate (indirect) selection on female mating preferences. However, in the present context, it is perhaps unlikely that host–pathogen coevolution could maintain much heritability in either haemocyte load or the ability to encapsulate a synthetic substrate (especially given that rates of pathogenesis were artificially low in our cultures). These parameters provide a measure of nonspecific components of immune function and are more likely to reflect resistance towards generalist pathogens than those that have evolved specific counter-defence mechanisms (Ryder & Siva-Jothy, 2000).

Instead, therefore, we suggest that other mechanisms may be more important for the maintenance of heritable variation in our study system. First, recent studies demonstrate that antagonistic pleiotropic effects between encapsulation ability and other fitness-related traits can contribute to the maintenance of heritable variation in insect immune function. Kraaijeveld & Godfray (1997) and Fellowes et al. (1998) demonstrated that selection for increased encapsulation ability (parasitoid resistance) led to reduced larval competitive ability in Drosophila melanogaster, probably because of antagonistic pleiotropic effects of the genes coding for higher encapsulation ability (see also Boots & Begon, 1993). Antagonistic pleiotropy may tend to maintain additive genetic variation if selection pressures vary either spatially or temporally. In our study population, density was allowed to fluctuate within the larval cultures over a range of approximately 100–300 individuals. Similar genetically based trade-offs may thus conceivably have occurred within our population as a result of varying intensity of larval competition (Simmons, 1987a). Second, theoretical work by Rowe & Houle (1996) suggests that characters that show condition-dependent expression may be especially likely to reflect heritable fitness differences (see also Houle, 1992). Both immune function and body size are influenced by nutritional stress in A. domesticus (J. J. Ryder, unpublished data) and thus show a degree of condition dependence. Finally, although persistent female choice for high quality males would be expected to deplete genetic variation, variation in the level of female ‘choosiness’ itself may slow the erosion of additive variation (e.g. Poulin & Vickery, 1996). Although Gray (1997) demonstrated that female A. domesticus orientate preferentially towards male calling song that contains a relatively high number of syllables per chirp, this study was based on a sample of relatively young virgin females. When older virgin females were used, they were found to be unselective with respect to ‘attractive’ and ‘unattractive’ male calling song (Gray, 1999). However, additional factors such as nutritional condition and (female) body size did not influence the female preference.

Conclusions

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgments
  9. References

In conclusion, this study has shown that variation in both immune function and body size is heritable in our study population. As both of these traits are correlated with the expression of a sexually selected song character in males (Ryder & Siva-Jothy, 2000), and are in themselves genetically correlated, females should have the opportunity to obtain indirect benefits through mate choice. Body size, which is correlated with components of fitness in a variety of organisms, including crickets (e.g. Simmons, 1988), and has been shown to be heritable in several other gryllines (e.g. Simmons, 1987b; Simons & Roff, 1994), may be a relatively common target of female mating preferences in this taxon. A number of studies have also shown that variation in encapsulation ability and haemocyte load can be heritable in insects (e.g. Carton & Boulétreau, 1985; Carton et al., 1992; Kraaijeveld & Godfray, 1997; Fellowes et al., 1998; Kraaijeveld et al., 2001). However, to our knowledge, only one other study has examined the genetic basis of variation in immune function in insects in the context of a sexually selected male character. Kurtz & Sauer (1999) detected heritable variation in two other parameters of immune function in a similar study of the scorpion fly, Panorpa vulgaris. However, it was unclear whether females were able to increase the ability of their offspring to produce an immune response by mating with more attractive males, largely because of insufficient statistical power. In contrast, the results presented here suggest that female house crickets should be able to increase the ability of their offspring to encapsulate pathogens by mating with males that express the preferred song characteristic.

Acknowledgments

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgments
  9. References

We are extremely indebted to Nick Colegrave and Sara Knott at ICAPB, The University of Edinburgh, for their help with the power analysis. We also thank two anonymous referees, who provided valuable criticism which greatly improved the manuscript, and Dave Coltman for providing statistical advice. J.J.R. was supported by a BBSRC studentship. M.T.S-J. was partially supported by NERC grant no. GR3/12121.

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgments
  9. References