Comparative evidence supports a role for reproductive allocation in the evolution of female ornament diversity

1. Sexually selected ornaments are highly variable, even among closely related species, and the ultimate causes of variation in ornament evolution are unclear, including in rare cases of female ornament expression. One hypothesis is that differences across species in female reproductive allocation may help to explain patterns of female ornament expression among insects with nuptial gifts.


Introduction
In some unusual mating systems, female fitness is limited by male monopolisation of resources required for reproduction, and females may consequently compete for mates. In such mating systems, females can evolve secondary sexual traits if the advantages of winning contests for mates is sufficiently large (Gwynne & Simmons, 1990;Clutton-Brock, 2009;Herridge et al., 2016). However, although sexual selection and ornament expression are common in males of many taxa (Janicke et al., 2016), even when females experience strong sexual selection they rarely have extravagant ornaments (Amundsen, 2000). One possible explanation for this disparity is that female fitness tends Correspondence: Frederick Hunter, School of Biological and Environmental Sciences, University of Stirling, Altchroskie, Enochdhu, Blairgowrie, Perthshire PH10 7PB, U.K. E-mail: freddie_hunter@ live.co.uk to be resource-limited to a greater degree than male fitness; by definition, females must make large investments in eggs, which might trade off with any investment in ornamentation (Fitzpatrick et al., 1995). Male choice for adorned females is probably also constrained by trade-offs between ornaments and offspring, as males should prefer mates who invest in offspring rather than ornaments. Moreover, when females store sperm, attractive females may actually present a higher risk of sperm competition, such that males might avoid rather than prefer showy females . Together, these arguments make rare species with male choice for showy female ornaments perplexing, and good candidates for testing theories about what regulates interspecific diversity in male choice and ornament expression.
Male insects often provide nutrition to females during courtship in the form of 'nuptial gifts', and these material donations are commonly used to initiate or accelerate egg production (Lewis et al., 2014). The degree to which females rely on nuptial gifts for egg production should covary with the sexual receptivity of females, because hungry females might use sex as a foraging technique. An increase in sexual receptivity can lead, in turn, to increased competition among females, especially if the preferred mating rate of females begins to exceed the rate at which males can provide gifts (Simmons & Gwynne, 1993;Arnqvist & Nilsson, 2000). Increased sexual selection on females arising from this enhanced competition could, under some circumstances, lead to the evolution of extravagant traits that improve female attractiveness and therefore female access to limiting nuptial gift nutrients.
While the role of nuptial gifts in promoting female contests and ornament evolution is relatively uncontroversial, the extravagance of ornaments often varies even among closely related species that share the same geographic distribution and mating behaviour, which remains unexplained (Downes, 1970;Cumming, 1994). Cumming (1994) has hypothesised that the presence and level of expression of female ornaments may depend on the intensity of female competition for nuptial gifts, which in turn might be determined by the degree to which females rely on male gifts for egg development. The allocation of resources to eggs prior to mating may therefore mediate the intensity of female competition for nuptial gifts, influencing the strength of selection on female investment into ornamental traits.
Female insects can allocate resources obtained during both larval and adult stages to reproduction, and the timing of acquisition and source of resources have important consequences on the reproductive and foraging strategies of animals (Boggs, 1997). Exclusively using larval-derived resources to invest into egg development is termed autogeny, and is common among Ephemeroptera and many species of Lepidoptera and Diptera (Engelmann, 1970). By contrast, anautogeny describes the condition in which females require some adult nutrients to mature eggs, as in mosquitoes that act as vectors for human diseases including malaria. The level of dependence on adult-derived resources for egg production varies continuously across anautogenous species (Boggs, 1981;Jervis, 2012). Although variation in anautogeny remains largely unexplained, the availability of resources at different stages in insect life cycles probably strongly influences the pattern of resource partitioning, and therefore resources available for reproductive allocation (Jervis et al., 2001). Nuptial gifts, being often of substantial size, may influence the partitioning of resources towards growth and reproduction in life stages prior to adult stages. This may occur because of female expectation of nuptial resources in the adult stage. However, to our knowledge, the extent to which selection arising from the presence of nuptial gifts affects resource partitioning has not been investigated.
By definition, anautogenous species should have females that depend more on adult-derived resources for egg production as compared with others. In these species, therefore, nuptial gifts (should they be present) may represent essential resources for female reproductive success (Fritzsche et al., 2016). As such, females of anautogenous species may experience greater competition for mating in order to acquire nuptial gifts, and subsequently have elevated selection for sexual trait investment (Cumming, 1994).
One group of taxa with nuptial gift giving that exhibit extraordinary interspecific diversity in the extravagance of female ornaments are the dance flies (Diptera; Empididae; Empidinae) (Collin, 1961;Downes, 1970;Cumming, 1994). Despite closely related species having similar courtship behaviour, females of different species display varying levels of ornamentation (Funk & Tallamy, 2000;Murray, 2015). Ornamental traits include pinnate leg scales, enlarged and/or darkened wings, and inflatable abdominal sacs, although their presence and extravagance vary substantially among taxa (Collin, 1961;Downes, 1970). Within-species variation in ornament size is known to influence male mate choice (Funk & Tallamy, 2000: Murray et al., 2018. Ornaments appear to have evolved independently multiple times across the dance fly phylogeny (Murray, 2015). We do not yet know what regulates these many evolutionary transitions in ornament expression across dance flies, but we expect higher levels of ornament expression in species subject to more intense sexual selection on females. Therefore, it is reasonable that the fitness of females in species with substantial sexual ornaments is more strongly constrained by dietary protein than is the case with unadorned species. In such species, investment in ornaments may be justified because the returns on investment in ornaments (through the accrual of nuptial gifts) more than offset the cost of construction (Fitzpatrick et al., 1995).
In roughly a third of dance fly species of the subfamily Empidinae, females are more numerous than males in lek-like mating swarms (Downes, 1970;Cumming, 1994). Males approach swarms with prey items as nuptial gifts, typically another dipteran (Cumming, 1994). Males typically assess females from below, apparently evaluating the gravidity of females (Funk & Tallamy, 2000). In most species, females do not appear to hunt as adults, and therefore derive all of their dietary protein from nuptial gifts.
The strong link between mating and foraging in dance flies suggests that egg development may be copulation-dependent (Cumming, 1994), but this hypothesis has never been rigorously tested. One limitation has been an inability to rear dance flies in the laboratory, and thereby study ovarian physiology in individuals of known age and mating history. Additionally nuptial gifts vary in size and quality, and therefore females may not receive the same volume of resources in each mating (Svensson et al., 1990), which may cause reproductive fitness to depend on nuptial gift characteristics.
In this study, we circumvent our inability to rear and manipulate flies by predicting that mating status (rather than mate number) covaries with egg development in ornamented and non-ornamented taxa. It is possible to distinguish mated from non-mated wild females by inspecting their sperm storage organs for the presence of sperm, and therefore to know whether a female has or has not received adult dietary protein, in the form of at least one nuptial gift. In order to assess the temporal dimension of ovarian maturation, we first adopt and validate an ageing method for wild-caught dance flies. We compare two species that are locally abundant for long periods near our university in central Scotland, facilitating the collection of mated and non-mated individuals at a range of different ages. The two species are Empis aestiva, which has females with extensive pennate scales on their mid-and hind legs (see Fig. 1), and Rhamphomyia crassirostris with no obvious sexual ornaments on either sex (see Fig. 1). In both species, males obligately provide nutritious nuptial gifts to females (males are not known to provide non-nutritious 'sham' gifts in these species). We predicted that, when unmated, the eggs of ornamented E. aestiva females would develop at a slower rate than the eggs of unadorned R. crassirostris females, whereas mated females with access to dietary protein might show no such differences.

Aim 1: Validating methods for ageing wild flies
The thoracic apodemes of flies are known to continue growing even after eclosion, leaving evidence of time passed since eclosion (Schlein & Gratz, 1973). A distinct line marks the extent of the apodeme structure upon eclosion (see Supporting information, Fig. S1). After eclosion, the density of the cuticle deposited at the cortical part of apodemes is influenced by temperature. Daily temperature fluctuations cause banding to occur, and therefore theoretically the number of apodeme bands covaries with the number of days since eclosion (Schlein & Gratz, 1973;Johnston & Ellison, 1982). However, it is likely that cuticle is not laid down on the apodemes indefinitely, and that the age at which the apodemes cease to grow varies among species and even individuals. Using apodeme bands to age flies has been validated for a number of Dipteran species, but not previously for any empid species (Neville, 1983).
In order to test whether our focal species produce apodeme bands that reliably reflect adult age, individuals of known age were required. We were unable to collect a sufficient number of the two focal species (the precise habitats of larval empids remain unclear, and our sampling was unable to improve this knowledge, but it is likely that larval empids are relatively well dispersed as they are thought to be predators). We therefore determine whether the thoracic apodemes of dance flies accumulate daily cortical growth bands using a range of Empidinae species. We also assessed which particular regions of the thoracic apodemes have the most distinguishable bands.
To ensure that the flies caught for this experiment were of known age, we used emergence traps (ground area 1 m × 2 m, height 1 m). The traps were deployed from 13 June to 20 July 2016 at three locations in central Scotland, U.K.: Stirling University campus (56 ∘ 14 ′ 81.10 ′′ N, 003 ∘ 90 ′ 52.02 ′′ W), Kippenross forest (56 ∘ 17 ′ 04.07 ′′ N, 003 ∘ 95 ′ 52.56 ′′ W) and a field near Enochdhu (56 ∘ 74 ′ 20.21 ′′ N, 003 ∘ 52 ′ 21.00 ′′ W). The traps were searched twice per day, once before 12.00 hours and again after 18.00 hours, which ensured that any flies found in the traps had emerged within the last 24 h. We aimed to collect as many flies as could be caught in the traps during the period. Flies in the dance fly subfamily Empidinae were retained and kept alive for between 0 and 10 days, as experimentally determined after capture using a random number generator (pilot experiments suggested that individuals were unlikely to live in captivity beyond 10 days). The captive flies were placed in individual plastic containers closed with cotton stoppers. Each container had two cuvettes placed inside, one with cotton wool soaked in water and another with sugar granules. The containers were kept sheltered, but outside, so that the flies would experience natural daily temperature fluctuations, which are critical for the formation of the thoracic bands (Schlein & Gratz, 1973;Johnston & Ellison, 1982;Neville, 1983). On the appointed day of sacrificing, the flies were killed by freezing (temperature ≤ −12 ∘ C).
The method used to dissect flies and count apodeme bands was modified based on the protocol published in Schlein and Gratz (1973). The head, abdomen, wings, and legs of the flies were removed and the thorax bisected. The bisected thorax halves were placed in distilled water and, using a dissecting microscope (Leica MZ12, Wetzlar, Germany) and titanium forceps, the anterior and posterior thoracic apodemes extracted. The majority of the muscle tissue was removed from the apodemes with forceps, and the rest was dissolved in potassium hydroxide (10%) for 150 min. The specimen was then rinsed in distilled water, stained in potassium permanganate (10%) for 2 min and rinsed again in distilled water. The anterior and posterior apodemes were bisected mid-sagitally with a razor blade to form four fragments. The two fragments of the anterior apodeme were bisected transversely into rostral and caudal halves. The resulting six fragments (two from the posterior and four from the anterior apodemes) were then mounted in DPX medium (Fisher Scientific, Loughborough, U.K.) on a microscope slide.
Taking care not to count the eclosion line, the bands on all six fragments of apodeme were counted using a light microscope (Olympus BX-41, Tokyo, Japan; see Fig. S1 for the locations of bands on the three apodeme fragments). To prevent bias, we counted blind to the knowledge of how many days the fly had been kept alive. Recounting to quantify repeatability using intraclass correlation was performed blind to both the real adult age and the previous count.
Statistics were performed in r version 3.3.2 (R Core Team, 2016). To validate the age estimates obtained from apodeme bands, we used major axis (model 2) linear modelling to regress the maximum number of bands counted from any apodeme against known age. Major axis regression was used because we recognise the possible error in both the predictor and response, and because least-squares regression is known to bias slope estimates negatively in such cases. Because bands are often hard to distinguish, and their formation is unlikely to be as regular as theorised (especially in Scotland where daily temperature fluctuations are often modest), we minimised the resulting bias by using maximum band number instead of the average across apodemes. We used age as the y-axis in this case because we experimentally manipulated the age of these flies, having collected them on the day of eclosion; therefore the band number was the response variable of interest even though our motivation is to assess whether band number reflects age. Models were validated by visually assessing diagnostic plots to confirm normality of residuals and homoscedascity, and to ensure that no records had unduly high influence. We used the ICC package for r to obtain repeatabilities (Wolak et al., 2012).

Aim 2: The effect of age and mating on ovarian maturation
In order to investigate the relationship between ovarian development and female resource allocation, we collected data on egg area (as a proxy for the stage of vitellogenesis), mating status (mated or unmated, based on the absence or presence of sperm in the spermatheca) and estimates of age using thoracic apodeme counts from wild females of unknown age captured in mating swarms or on vegetation using hand-held nets. We aimed to collect 30 unmated and mated females of both species. Flies were dissected using a Leica MZ12 light microscope and titanium forceps. The abdomen was removed and placed in distilled water and opened. We photographed between three and five of the largest eggs (dance flies mature eggs in clutches, such that within-female variation in egg size is negligible; Luc F. Bussiere, pers. obs.) using a microscope-mounted camera (Olympus SP-500UZ) and measured egg area using imagej version 1.51 h (Schindelin et al., 2015).
Like most other insects, female dance flies are able to store sperm, and eggs are fertilised only at the time of oviposition. We checked the mating status of females by splitting the spermatheca with forceps and visually assessing the presence of sperm (females captured while mating invariably contain sperm, suggesting that failures to transfer ejaculate are rare in these taxa, Luc F. Bussiere, pers. obs.). As females in the two species being investigated have never been seen mating without a nuptial gift (Luc F. Bussiere, pers. obs.), by comparing ovarian development in mated and unmated females, we are able to distinguish the development of eggs in females with and without access to adult dietary protein. The thoracic apodemes of the species were dissected and stained using the method outlined earlier. We used the maximum number of bands counted on any of the apodeme types as our best estimate of age.
To test whether the two species of dance fly differ in the effect of mating status on egg development, we fitted a multiple regression model including a three-way interaction between age (in days), mating status (mated or unmated as a categorical variable) and species (E. aestiva or R. crassirostris). The two species differed substantially in the size of mature eggs, so average egg sizes were standardised independently for each species, which facilitates interpretability of model coefficients (Schielzeth, 2010) by providing a strong test of whether the interaction between mating status, species and age represents a difference in the rate of maturation, rather than merely reflecting the different egg sizes across species. We assessed model quality by visually inspecting diagnostic plots. In order to achieve homoscedascity and improve fit, we used a natural log-transformed average egg size for each female in our model.

Results
Among flies caught in emergence traps (see Table 1) and reared in captivity for up to 10 days, the numbers are of bands counted on the rostral and caudal parts of their anterior apodemes ranged from zero to eight and zero to five, respectively. For the posterior apodemes the number of bands ranged from zero to nine. The repeatabilities in Table 2 show that estimates of age are most consistent across blind trials for the maximum band number, followed by the rostral fraction of anterior apodemes, then posterior apodemes, and finally the caudal part of the anterior apodemes. We therefore estimated the age of field-caught dance flies using the maximum number of bands counted on any apodeme. This subset of counts produced an age-predicting model with the lowest Akaike information criterion and highest R 2 when compared with models using counts from only one apodeme type (see Fig. 2).
The age estimates for R. crassirostris females ranged from 0 to 10 days, with a broadly Gaussian distribution notwithstanding the strict bound at zero (see Fig. 3). Nineteen of the 69 females were unmated, and average egg sizes ranged from 0.0092 to 0.1305 mm 2 , with mean (± SE) of 0.0500 ± 0.0036 mm 2 . The age estimates for E. aestiva ranged from 0 to 5 days (see Fig. 3). Eighteen of 49 females were unmated. Average egg sizes ranged from 0.0068 to 0.0523 mm 2 , with a mean of 0.0181 ± 0.0014. The effect of mating status on egg development differed between the two dance fly species. In the unadorned R. crassirostris, eggs increased gradually in size for both mated and unmated females (although the intercept for unmated females was lower, as might be expected if nuptial gifts boosted vitellogenesis). By contrast and as predicted, unmated E. aestiva demonstrated no change in egg size with age (partial F-test of whether the removal of the three-way interaction term results in a poorer model fit: F = 4.1919, P = 0.0430; see Fig. 4).

Discussion
We examined egg development in mated and unmated dance flies from two species that contrast sharply in female ornamentation, and found evidence consistent with an association between anautogeny and ornamentation, supporting the hypothesis that interspecific differences in dance fly sexual ornament expression derive at least in part from differences in ovarian physiology. Mindful of the limitations on inference that are inherent with two-species comparisons, we discuss the implications of our findings for mating system diversity in this group and female ornament expression in general, as well as the validated ageing method that we modified from previous work.

How mate acquisition relates to investment in ornaments
Sexually selected ornaments are highly variable, even among some closely related species (Murray, 2015). Ornaments evolve as a consequence of competition for mates, but the underlying causes of competition are unclear in many cases. Female ornaments are particularly curious adaptations, because general conditions that favour their evolution appear to be exceedingly rare (Clutton-Brock, 2009). In spite of this rarity, some taxa, such as the dance flies, possess extravagant variation in female ornaments, which challenges our understanding of the general rules that are thought to regulate diversity in mating systems (Janicke et al., 2016).
We tested Cumming's (1994) prediction that interspecific variation in how and when female dance flies allocated resources to eggs play a key role in mediating the intensity of female contests Table 3. Parameter estimates describing the effect of age, mating status, and species on egg development (× denotes interaction terms). The model uses treatment contrasts to compare categories; the reference levels are Empis aestiva and mated flies (see Schielzeth, 2010). This model explained a significant fraction of the variation in natural log-standardised egg sizes (mm 2 ) (R 2 -adjusted = 0.1756, P = 0.0002, n = 118). for mates. Although we did not directly manipulate allocation (such experiments remain impossible given the current state of knowledge for culturing dance flies), we nevertheless generated a priori predictions for patterns of ovarian maturation in females of two different species that differed in ornament expression. We found that, as predicted, the eggs of unmated females of the ornamented species developed at much slower rates compared with the unornamented species. This pattern suggests that females of the ornamented species rely to a greater degree on male nuptial gifts compared with females from unadorned species (see Table 3), which provides long-awaited support for Cumming's hypothesis, and can help to explain the otherwise confounding diversity of sexual trait expression among dance flies : Houslay & Bussière, 2012. In light of the observational nature of our work, there are several factors in addition to the hypothesised mechanism that could be at play. For example, we cannot rule out the possibility that females may receive nutrition or simulants other than prey item nuptial gifts from mating, as is common in other insect species, which may also influence the development of eggs (Lewis et al., 2014). However, even if this were true, it would not explain the difference in ovarian maturation seen among unmated females in the two focal species. Likewise, our estimates of egg size, mating status, and age were all measured with error. While these sources of error do not lead to systematic biases in our estimates, they nevertheless make our central conclusions tentative, especially in light of the marginal significance of our key result. More information on how these errors contribute to patterns in the data would be welcome. For example, we did not know the number of times a female had mated, and the volume of resources acquired by the female during each mating. Some of these matings may have been too recent for the effective conversion of nuptial gifts to egg maturation, and others may have resulted in minimal resource intake. More detailed information on mating history rather than mere mating status (mated or unmated) would clearly provide more resolution for this kind of analysis.
Another possibility is that species differences in mating system caused differences in the representativeness of subsamples of unmated females. For example, if male choice is stronger in the ornamented E. aestiva, it might lead to a stronger difference in average condition between the mated and unmated fractions of females. The fact that low-condition females in that species are less likely to mature eggs over time could therefore be due to their lower condition rather than to species differences in reproductive allocation. Although we cannot rule out this possible alternative explanation, it is worth noting that the proportions of unmated females were similar in the two species, suggesting no large difference in the chances of mating (19/69, or 27.5%, of females for R. crassirostris, compared with 18/49, or, 36% for E. aestiva). It seems unlikely that this small difference explains our observed differences in ovarian physiology.
Another intriguing contrast across species involved the difference in female age range, as estimated by apodeme bands. Empis aestiva ages ranged from 0 to 5 days and R. crassirostris from 0 to 10 days. Although these differences in apparent age structure are intriguing and unexplained, they may simply reflect differences across species in the deposition of cuticular bands. Once again, we think this is unlikely, because if many individuals were older than the maximum number of bands possible, we would expect a left-skewed distribution of apparent ages, which we did not observe (Fig. 3).
In species with relatively brief adult life span, it is probably important for females to mate quickly in order to produce eggs and oviposit. Selection for females to mate quickly after emerging may act as an important factor leading to the expression of ornaments. In the apparently longer-lived R. crassirostris, females may be able to remain virgins for a longer period. Comprehensively disentangling the possible causes of species differences in sexual receptivity requires more information on the individual species in question, as well as more comparative work on further species. Herridge (2016) found significant differences in the number of matings obtained by females of three nuptial gift-giving species of dance fly, including our study species E. aestiva. Interestingly, mate number in these species did not straightforwardly reflect ornament expression. Although the most ornamented species (R. longicauda) does appear to have the most sexually receptive females, female E. aestiva appeared to mate less often than in a completely unornamented species (E. tessellata). The longevity of R. longicauda and E. tesselata are unknown, but it is possible that the combination of female size, longevity and reproductive allocation affects sexual receptivity, which helps to further explain cross-species differences in female ornament expression.
Inferring adaptive patterns from two-species comparisons is difficult (Garland & Adolph, 1994). We recognise that our study would benefit from comparing additional species; however, the collection of sufficient numbers of mated and unmated females of even the two species used here was difficult, despite the fact that these species are abundant, and we know a reasonable amount about both their mating swarm locations and habitat use. Although we remain cautious in our conclusions, we agree with Cooper (1999) that two-species comparisons help to identify promising avenues of research.
Interspecific variation in anautogeny remains largely unexplained (Jervis et al., 2001). However, resource allocation to reproduction is expected to relate to the nutritional ecology of the adult (Jervis et al., 2001). The quantity, quality, and predictability of resources available to the adult in the environment are factors that probably influence the evolution of reproductive allocation across life-history stages (Karlsson, 1994). This fact supports the notion that nuptial gifts are in a unique position to influence both sexual selection and life history (Lewis et al., 2014).

Validation of an ageing method for wild dance flies
A technique for ageing wild subjects is one of the prerequisites for measuring egg development rates in dance flies, which have heretofore been difficult to culture under laboratory conditions. We have demonstrated that using thoracic apodeme bands works at least as well for the Dipteran subfamily Empidinae as has previously been shown for other groups of flies (Neville, 1983), in spite of the relatively slight daily fluctuations in temperature that are routine in Scottish summers. We hope this demonstration will stimulate more investigations of an ecological and evolutionary nature on the demographic patterns and life-history traits of wild flies, which remain heavily reliant on laboratory studies (see Bonduriansky & Brassil, 2002 for an exception).
In previous studies that evaluated the accuracy of apodeme bands to age other species of fly, flies were kept in controlled-temperature regimes (Neville, 1983); hence flies experienced constant temperature fluctuation between day and night for an extended period of time, which probably influenced how distinguishable the bands were (Johnston & Ellison, 1982;Schlein & Gratz, 1973). In our study, recently emerged flies were kept outside and experienced natural daily fluctuations in temperature. Although it was difficult to count bands in some specimens, which inflated uncertainty around age estimates, our estimates were nevertheless consistently close to the ages of known-age specimens (see Fig. 2). Regardless of the exact linear modelling approach (ordinary least squares, major axis, or forcing the intercept through the origin), our slope estimate was consistently > 1, indicating that band numbers are a good proxy for age but are likely to systematically underestimate it. This is not surprising as it is easier to imagine environmental conditions that obscure bands than those that might create additional bands.
Because of an inability to collect large numbers of individual species in emergence traps, we had to pool individuals of all dance fly species for our analysis. It is, of course, possible that species differ in the rate of deposition of cuticle, although the theoretical link between temperature and band deposition makes it unlikely that any such difference is a cause of bias in our study. Nevertheless, it would be useful to supplement our data with collections of more known-age flies in the future to test this possibility.
It is also possible that the deposition of cuticle on apodemes ceases in the adult at a particular age, which is known to occur in several Drosophila species (Johnston & Ellison, 1982), and that the age at which this occurs varies across species and individuals. Our data in Fig. 2 show a linear relationship between band number and age, with no identifiable upper limit to age resolution (if the number of bands were saturating, we might have expected a non-linear relationship between age and band number). However, only one fly was kept alive for 10 days, and it exhibited nine bands on its posterior apodeme. Whether band deposition continues beyond the maximum of nine we observed is unclear. However, in the vast majority of flies used to measure egg development, individuals had fewer than nine bands and so were unlikely to be older than the maximum age that this method may provide. Furthermore, the histograms describing the distributions of bands in wild flies were not left-skewed (as might be expected if there were many individuals at or near an upper bound of band number), which suggests that most wild females were still accumulating bands at the time of capture (see Fig. 3).
The validation of an ageing technique for further groups of wild insects may be important to future investigations in evolutionary ecology. Estimates of age and longevity of wild populations are often the focus of studies of natural and sexual selection (Endler, 1986). The apodeme band ageing technique may be successfully applied in studies examining differences in the age structure of the sexes. For example, male dance flies are hypothesised to incur higher mortality rates than females due to the male sex-specific behaviour of acquiring nuptial gifts (Murray, 2015). However, the extent to which sexual differences may incur survival costs in other contexts remains unclear [e.g. due to the risk of entanglement for ornamented females in spider webs; see, e.g., Gwynne et al. (2015)]. The consequences of sex differences in mortality could be evaluated by comparing the age profiles of the sexes in more wild populations.