Polyandry-fecundity relationship in insects: methodological and conceptual problems

Authors


Correspondence: Dr. Luis Miguel Torres-Vila, Servicio de Sanidad Vegetal, Consejería de Agricultura DRMAyE, Avda. de Portugal s/n, 06800 Mérida (Badajoz), Spain. Tel.: +34 924 00 25 30; fax: +34 924 00 22 80; e-mail: luismiguel.torres@juntaextremadura.net

Abstract

Polyandry is perhaps the most puzzling component of mating systems because the fitness benefits for females of mating with more than one male during lifetime are poorly understood. The occurrence and extent of polyandry varies considerably both among and within species, and a positive association between polyandry and fecundity is widespread but not universal. The scenario is further complicated because the scientific literature on this issue includes studies that are often inconclusive or contradictory even for the same target species. A previous meta-analysis detected the crucial importance of two usually neglected aspects that potentially bias the interpretation of primary studies about the polyandry–fecundity relationship: the methodological approach – experimental or descriptive – and the polyandry concept itself – realized or potential. In this paper, we experimentally test the effect of these aspects with the moth Lobesia botrana. We used an innovative protocol in which the experimental and the descriptive methods were conducted simultaneously on the same target population and the results were then interpreted from the perspective of both concepts of polyandry. The results clearly showed that 1) the conclusions about the polyandry–fecundity relationship were strongly dependent on the methodological approach used and 2) the concept of polyandry invoked by the researcher was a confounding effect that potentially biases data interpretation. We suggest that greater attention must be paid to intraspecific variation among females in their propensity to remate. The differentiation in experimental studies between potentially polyandrous and monandrous phenotypes could greatly improve our knowledge about the maintenance of female mating polymorphism in most species and the adaptive significance of polyandry.

Introduction

The mating system of an animal species consists of a number of strategies selected to maximize reproductive success and individual fitness. One crucial component is the lifetime number of matings of each sex. The best male strategy is in almost all cases to acquire as many mates as possible (polygyny) to maximize paternity. In contrast, the fitness benefits for females of mating with more than one male during her lifetime (polyandry) are far less obvious. Female remating rate has been shown to vary considerably both among and within insect species (Drummond, 1984; Eberhard, 1985; Ridley, 1988; Vahed, 1998; Arnqvist & Nilsson, 2000; Torres-Vila et al., 2004), and the occurrence of polyandry therefore puzzles evolutionary biologists.

A huge research effort across most insect taxa has been carried out to explain the adaptive significance of polyandry. Two main approaches have been considered. The first focuses on male nuptial gifts acting as paternal investment and any direct (material/energetic) benefits derived from nutrient acquisition via ejaculate to enhance fecundity and/or offspring quality. Direct benefits for females also include the potential to counteract infertile first matings, to offset sperm ageing and to acquire additional sperm to ensure fertilization of the full egg complement. The second is connected with sperm competition and/or cryptic female choice and deals with the indirect (genetic) benefits of multiple paternity, enhancing mean offspring fitness either through the increase of genetic diversity among progeny or through the differential acquisition of ‘good genes’ that elevate mean offspring fitness. Indirect benefits also include to avoid female–male genetic incompatibility or take selective advantage of bet-hedging strategies (Parker, 1970; Walker, 1980; Ridley, 1988; Hunter et al., 1993; Boggs, 1995; Keller & Reeve, 1995; Eberhard, 1996; Choe & Crespi, 1997; Birkhead & Møller, 1998; Vahed, 1998; Yasui, 1998; Arnqvist & Nilsson, 2000; Jennions & Petrie, 2000; Kondoh, 2001; Simmons, 2001; Zeh & Zeh, 2001). However, discriminating between direct and indirect benefits of polyandry may be difficult, and a recent meta-analysis suggests that there is still weak experimental evidence for genetic benefits to polyandry (Slatyer et al., 2012).

Several potential costs to female remating have been identified that counterbalance potential benefits, including increased exposure to male-borne parasites and diseases, increased risk of predation and sexual conflict over mating rates due to time and energy expenditure or male-produced physical injury (Walker, 1980; Thornhill & Alcock, 1983; Drummond, 1984; Ridley, 1988; Hunter et al., 1993; Keller & Reeve, 1995; Choe & Crespi, 1997; Yasui, 1997; Birkhead & Møller, 1998; Arnqvist & Nilsson, 2000; Siva-Jothy, 2000; Tregenza & Wedell, 2000). It is a demanding task to test and quantify these benefits and costs to understand the sexual selection mechanisms driving the evolution of polyandry. An even more convoluted scenario arises if the evolutionary interests of both sexes at mating are in a conflict that results in antagonistic coevolution (Chapman et al., 2003; Vahed, 2007).

A positive association between polyandry and fecundity is probably the most quantitatively important, short-term process sustaining the direct benefits of multiple mating in insects. Primary studies and narrative reviews (Ridley, 1988; Vahed, 1998 and references therein) as well as three meta-analyses (Arnqvist & Nilsson, 2000; Torres-Vila et al., 2004; South & Lewis, 2011) support the claim that females gain directly from multiple mating across species in terms of increased reproductive output. However, a careful inspection of available studies still reveals that a positive correlation between polyandry and fecundity is not universal, at least for the Lepidoptera. Therefore, the occurrence of monandrous and polyandrous females in certain species is poorly explained in terms of reproductive output. The scenario is further complicated because the scientific literature on this topic includes numerous studies that are rather inconclusive or contradictory, even those on the same target species.

Two distinct methods, usually the so-called experimental and descriptive (Lehner, 1996), are typically used to study the effect of female remating on fecundity (Torres-Vila et al., 2004). The experimental method involves the creation of two groups of randomly chosen females. In the control group, females are allowed to mate only once. In the test group, females are allowed to mate multiply. The experimental method implies that the researcher has designed the experiment a priori and randomly established female groups to be compared. In this study, we will refer to these groups as mating opportunities (see below). In contrast, the descriptive method implies that all tested females are provided with mating opportunities throughout their lifetime. The number of matings for each tested female is determined by visual observations or by waiting until she dies and then counting the number of spermatophores she contains. The descriptive method implies that the groups to be compared are established a posteriori and that the researcher does not determine their membership. In this study, we will refer to these groups as female phenotypes (see below). For both methods, the mean fecundity of females from the two groups is compared to assess the effect of remating on fecundity.

The descriptive method does not usually allow a causal relationship between polyandry and increased fecundity to be inferred. Hence, the experimental method is usually regarded as the preferred way to test whether remating enhances fecundity because it eliminates confounding factors that might covary with female mating rate and bias the interpretation of a purely descriptive study. In a landmark paper, Ridley (1988) discussed most of the advantages and handicaps of each methodological approach. The main outcomes of the meta-analysis by Torres-Vila et al. (2004) illustrate well the state of affairs in the Lepidoptera: 1) studies where female remating increased fecundity were more often established in polyandrous species when the experimental method was used, 2) the experimental method tended to be utilized more often for polyandrous species and 3) remating increased fecundity irrespective of the test method in polyandrous species but conclusions were method-dependent in monandrous species as fecundity did not increase with remating in descriptive studies. From these results, the crucial importance of two usually neglected aspects potentially biasing the interpretation of results from primary studies becomes clear. The first aspect is the methodological approach chosen by the researcher – experimental or descriptive – to measure the benefits of polyandry. The second aspect, closely related to the first, is the polyandry concept itself: realized or potential. Realized polyandry refers to the number of times females do mate whereas potential polyandry refers to the number of times females would mate if males were always available. Researches have not usually distinguished between these two concepts. Female post-mating receptivity has often been overlooked, and therefore, females that naturally chose to be monandrous and females forced to be monandrous have been considered equivalent.

This paper deals with the experimental verification of the importance of these two aspects based on meta-analysis predictions using as a model species the European grapevine moth Lobesia botrana Den. et Schiff. (Lepidoptera: Tortricidae). We performed an innovative protocol to explore the effect of female remating on fecundity. The experimental and the descriptive methods were conducted simultaneously on the same population to exploit the advantages and to handle the constraints of each method, and the results were then analysed and interpreted from the perspective of both concepts of polyandry: realized or potential.

Materials and methods

Insects and general procedure

Adults were obtained from a rearing colony maintained on a semisynthetic diet at 22 ± 1°C, 60 ± 10% r.h. and a L16 : D8 h photoperiod (1000 lux photophase luminosity). To insure virginity of test moths, pupae were isolated in glass tubes (70 × 9 mm in diameter) stoppered with cardboard plugs. Newly emerged moths (< 24-h old) to be used in tests were collected daily and sexed. Tests were performed in a controlled environmental room under the same conditions but with an L(15 + 1) : D8 h photoperiod. The first 15 photophase hours (day) were at 1000 lux luminosity, and the last hour was at 25 lux to simulate dusk. Artificial dusk was necessary to observe adult behaviour as this is the daily period during which vital activities (flight, feeding, calling, mating and egg-laying) take place in L. botrana. The photoperiod in the rearing and the test rooms was synchronized.

Mating opportunity

We created pairs consisting of one 1- to 2-day-old virgin female and one 2- to 3-day-old virgin male. Moth pairs were caged in 22-mL clear plastic containers as mating and oviposition chambers and provided with water ad libitum through a soaked cotton wick. The sexual activity of moths was continuously observed during the first 1-h dusk period, and mated pairs were noted. On the following morning, unmated pairs (< 5%) were rejected, and all males were removed leaving the females isolated. Mated females were then randomly distributed before the onset of the second dusk to establish the two experimental groups differing in mating opportunity.

In the first group (hereafter referred to as control group or once-mated group), mated females remained isolated throughout their lifetime. Females were daily observed at dusk, and the eventual occurrence of recalling behaviour associated with a new period of release of sex pheromone (denoting female receptivity) was recorded for each female. Calling females were identified by the characteristic position of the wings, slightly raised, the abdomen turned downwards and the pheromone gland protruding and clearly visible at the abdominal tip (Torres-Vila et al., 2002).

In the second group (hereafter referred to as test group or ‘at will’ mated group), the protocol was similar except that when a mated female resumed calling, a new virgin male was immediately introduced into the container. Remating was noted, and male was removed on the next morning again leaving the female isolated. This procedure of providing and removing a new virgin male was repeated every time a female recalled. Females were always isolated after each mating as male–female cohabitation, particularly at high population densities and male-biased sex ratios, can unnaturally increase female remating propensity (Eberhard, 1985) and therefore overestimate polyandry as showed in L. botrana (Torres-Vila et al., 1997). When necessary, females were dissected after death under a stereomicroscope, and spermatophores in the bursa copulatrix were counted to verify the success of the observed matings and eventually to account for unobserved ones, especially in those females exhibiting a high remating rate. In both female groups, fecundity (viable eggs), refractory period (the number of days elapsed between the first mating and recalling or remating), lifetime mating number (test group) and longevity (days) were recorded for each tested female.

Female phenotype

The protocol classified females into two subgroups (hereafter referred to as female phenotypes) within each experimental group that differed in mating opportunity. Note that the term ‘phenotype’ is properly applied here as polyandric behaviour in this species has both an environmental and genetic basis (Torres-Vila et al., 1997, 2002). Mated females that remained refractory throughout their lifetime (i.e. never exhibit recalling) were scored as monandrous. Females that resumed calling were scored as polyandrous. Note that the true polyandry rate may be overestimated if female remating is merely a result of unrecorded first mating failure (Ridley, 1988). In L. botrana, specifically, mated but unfertilized females (laying no viable eggs or no eggs at all) always recall and remate (Torres-Vila et al., 1997). To avoid this potential skew, we confirmed the fertilization of all tested females after the first mating. In the control group, there was no major problem as females only mated once. In the test group, however, it was necessary to differentiate the eggs laid after the first mating from those laid after the second and eventual successive matings. To do this, all eggs laid after the first mating were marked with a permanent pen on the outer wall of the container before the introduction of the second male. Eggs were incubated after the death of female in the test room at 22°C for 10 days, and the hatching of marked eggs was verified a posteriori. Mated but unfertilized females after first mating (< 4%) were excluded from both groups for data analysis.

Moth strain

The same protocol was carried out using two moth strains differing in their polyandry level, a naturally monandrous (M) strain and an artificially selected polyandrous (P) strain (Torres-Vila et al., 2002). Nearly 500 and 900 females were tested in the P and M strains, respectively, as female number had to be greater in the M strain to ensure a sufficient sample size of polyandrous females could be tested. In the P strain, we obtained 493 mated and fertilized females (398 of them recalled), and in the M strain, we obtained 868 mated and fertilized females (231 of them recalled). Consequently, polyandry frequency was about 27% in the M strain and 81% in the P strain. Female sample size for each combination of mating opportunity, female phenotype (or refractory period) and strain are given in Fig. 1 and 2.

Figure 1.

Fecundity (viable eggs) and longevity (days) of female Lobesia botrana depending on mating opportunity (females mated at will vs. females mated once) and female phenotype (polyandrous vs. monandrous) in two laboratory strains differing in polyandry level (polyandrous (P) and monandrous (M) strains with 81% and 27% polyandry, respectively). Note that sample size (numbers near data points) in the M strain does not reflect a 27% polyandry level because data from monandrous females in that strain were only recorded in a random subsample (see text). Asterisks indicate significant differences at the < 0.05*, < 0.01** and < 0.001*** levels after one-way anovas comparing female phenotype within mating opportunity and mating opportunity within female phenotype (see text and Tables 1 and 3 for a complete statistical analysis). Vertical bars indicate the standard error of the mean (SE).

Figure 2.

Fecundity (viable eggs) of female Lobesia botrana depending on mating opportunity (females mated at will vs. females mated once) and refractory period (1, 2, 3, 4, 5, 6, 6<  days and no-recalling) in two laboratory strains differing in polyandry level (polyandrous (P) and monandrous (M) strains with 81% and 27% polyandry, respectively). Note that sample size (numbers near data points) in the M strain does not reflect a 27% polyandry level because data from no-recalling (monandrous) females in that strain were only recorded in a random subsample (see text). There was a significant mating opportunity x refractory period interaction in both strains (see Table 2 for a complete statistical analysis), so that we computed one-way anovas (see text) comparing the refractory period within mating opportunity (different letters denote significant differences at the < 0.05 level after a LSD test) and mating opportunity within refractory periods (asterisks indicate significant differences at the < 0.05*, < 0.01** and < 0.001*** levels, respectively). Vertical bars indicate the standard error of the mean (SE).

Data analysis

Data were analysed using multiple-way analysis of variance (anova). Three-way anovas were computed to assess the effect of mating opportunity, female phenotype and moth strain on female lifetime fecundity (viable eggs) and female longevity (days). Levels for each studied factor were mating opportunity (at-will-mated and once-mated females, 2 levels), female phenotype (polyandrous and monandrous females, 2 levels) and moth strain (polyandrous and monandrous strain, 2 levels). The refractory period was recorded for each female, so that similar three-way anovas were used to assess the effects of mating opportunity, female refractory period and moth strain. Statistical levels for female refractory period were 1, 2, 3, 4, 5, 6, 6 < days and no-recalling (8 levels). In females mated more than twice (test group), we used as refractory period the time elapsed between the first and second mating. No-recalling females were included in this analysis to allow a direct connection between the anovas computed using either female phenotype or refractory period data for each tested female. This poses no problem in statistical or biological terms because no-recalling females can simply be viewed as individuals whose refractory period exceeds their lifespan. Further two- and one-way anovas were computed when necessary. Model I anovas were used throughout as all studied factors were considered to have fixed effects. Statistical tests were performed using Systat (2000) software.

Results

A three-way anova showed that all three studied factors, mating opportunity, female phenotype and moth strain, significantly affected female fecundity (Table 1). Fecundity in the M strain was slightly, but significantly, higher than in the P strain, in the range of 10–25 eggs more depending on the combination of mating opportunity and female phenotype (Fig. 1). Fecundity differences were partly attributed to female body weight as a posterior comparison between strains of two random subsamples showed that M strain females (8.6 mg) were significantly heavier than P strain females (7.5 mg) (F1,78 = 24.12, < 0.001).

Table 1. Three- and two-way Model I anovas of the effects of mating opportunity (MO), female phenotype (FP), strain (S) and their interactions on fecundity of female Lobesia botrana
Sourced.f.MS F P
  1. Main factors and their statistical levels are as follows: mating opportunity (MO): once-mated and at-will-mated females (2 levels); female phenotype (FP): monandrous and polyandrous females (2 levels); strain (S): polyandrous and monandrous strains (2 levels) with polyandry levels of 81% and 27%, respectively. MS: mean-squares.

Both strains
MO177548.5033.86< 0.001
FP132880.5114.36< 0.001
S146883.4520.47< 0.001
MO × FP124354.6810.640.001
MO × S1811.960.360.55 ns
FP × S13820.681.670.20 ns
MO × FP × S154.510.020.88 ns
Error8072290.13  
Polyandrous strain
MO147621.3421.82< 0.001
FP129876.3413.69< 0.001
MO × FP113500.296.190.013
Error4892182.04  
Monandrous strain
MO130916.6112.59< 0.001
FP17067.212.880.09 ns
MO × FP110936.164.450.036
Error3182456.34  

To study how the interaction between mating opportunity and female phenotype affected fecundity, we conducted two-way anovas separately for each strain. There was significant mating opportunity x female phenotype interaction in both strains (Table 1), so that one-way anovas were performed between female phenotypes within mating opportunities and between mating opportunities within female phenotypes. In females mated at will (test group), there were no significant differences between phenotypes in either the P strain (F1,134 = 0.85, = 0.36) or the M strain (F1,157 = 0.11, = 0.75). By contrast, in females mated once (control group), the monandrous phenotype was significantly more fecund than the polyandrous one, both in the P strain (F1,355 = 26.29, < 0.001) and in the M strain (F1,161 = 6.09, = 0.015) (Fig. 1). Such a response pattern resulted in polyandrous females mated at will laying significantly more eggs than polyandrous females mated once, both in the P strain (F1,396 = 55.36, < 0.001) and in the M strain (F1,229 = 28.27, < 0.001). As one might expect, the fecundity of monandrous females was similar in the two experimental groups (mating opportunities), both in the P strain (F1,93 = 2.16, = 0.15) and in the M strain (F1,89 = 0.73, = 0.40), because all these females had the same phenotype and experienced the same environment and reproductive background despite belonging to different statistical treatments.

Finer analyses revealed a significant impact of the refractory period on the fecundity of polyandrous females. A three-way anova showed that mating opportunity, female refractory period and moth strain all significantly affected female fecundity (Table 2). Following the same protocol described previously, we computed two-way anovas separately for each strain. The response pattern obtained was similar, with fecundity being higher in the M strain (Fig. 2). There was a significant mating opportunity x refractory period interaction in both strains (Table 2), so that one-way anovas were computed between refractory periods within mating opportunities and between mating opportunities within refractory periods. In the test group (females mated at will), female refractory period did not significantly affect fecundity in either the P strain (F7,128 = 0.71, = 0.67) or the M strain (F7,151 = 1.13, = 0.35), producing a more or less flat response line whose slope did not significantly differ from zero (Fig. 2). By contrast, in the control group (females mated once), the refractory period significantly affected fecundity in both the P strain (F7,349 = 21.21, < 0.001) and the M strain (F7,155 = 5.31, < 0.001). The longer the refractory period, the greater the realized fecundity, the response curve following an asymptotic trend to a Y-axis value close to the mean fecundity of females that mated at will in the test group (Fig. 2). A comparison between mating opportunities within refractory periods showed significant differences in fecundity when refractory periods were shorter than 4–5 days, depending on the moth strain. Thus, with a longer refractory period (a lower propensity of female to want to mate again), the effect of mating opportunity became nonsignificant because the fecundity of females mated once reached that of females mated at will (Fig. 2).

Table 2. Three- and two-way Model I anovas of the effects of mating opportunity (MO), refractory period (RP), strain (S) and their interactions on fecundity of female Lobesia botrana
Sourced.f.MS F P
  1. Main factors and their statistical levels are as follows: mating opportunity (MO): once-mated and at-will-mated females (2 levels); refractory period (RP): 1, 2, 3, 4, 5, 6, 6 <  days and no-recalling (8 levels); strain (S): polyandrous and monandrous strains (2 levels) with polyandry levels of 81% and 27%, respectively. MS: mean-squares.

Both strains
MO176558.9839.05< 0.001
RP720488.6210.45< 0.001
S137779.7219.27< 0.001
MO × RP79254.754.72< 0.001
MO × S1438.820.220.64 ns
RP × S7767.320.390.91 ns
MO × RP × S7529.800.270.97 ns
Error7831960.33  
Polyandrous strain
MO136448.2220.59< 0.001
RP715161.198.57< 0.001
MO × RP75489.263.100.003
Error4771769.92  
Monandrous strain
MO141674.7818.46< 0.001
RP79011.613.99< 0.001
MO × RP74666.912.070.047
Error3062257.16  

A further two-way anova just considering the 295 females that mated at will in both strains (test group) showed that the number of matings per female (1 – 4, four levels) had no effect on fecundity (F3,287 = 0.95, = 0.42). These results also confirmed that fecundity in the M strain was higher than in the P strain (F1,287 = 4.50, = 0.035), with the strain × mating number interaction being nonsignificant (F3,287 = 0.29, = 0.83).

Lastly, a three-way anova showed that female longevity was significantly affected by female phenotype but not by mating opportunity or moth strain (Table 3). There was also a female phenotype x strain interaction, so that separate two-way anovas were computed for each strain. Female longevity was significantly higher in polyandrous than monandrous females irrespective of the strain (Table 3, Fig. 1). Similar results were obtained when the effect of female refractory period on longevity was analysed.

Table 3. Three- and two-way Model I anovas of the effects of mating opportunity (MO), female phenotype (FP), strain (S) and their interactions on longevity of female Lobesia botrana
Sourced.f.MS F P
  1. Main factors and their statistical levels are as follows: mating opportunity (MO): once-mated and at-will-mated females (2 levels); female phenotype (FP): monandrous and polyandrous females (2 levels); strain (S): polyandrous and monandrous strains (2 levels) with polyandry levels of 81% and 27%, respectively. MS: mean-squares.

Both strains
MO14.560.940.33 ns
FP1335.4668.88< 0.001
S15.871.210.27 ns
MO × FP113.872.850.09 ns
MO × S12.100.430.51 ns
FP × S128.925.940.015
MO × FP × S12.640.540.46 ns
Error8074.87  
Polyandrous strain
MO16.501.240.27 ns
FP1283.6954.14< 0.001
MO × FP12.230.430.52 ns
Error4895.24  
Monandrous strain
MO10.230.050.82 ns
FP182.8219.26< 0.001
MO × FP114.153.290.07 ns
Error3184.30  

Discussion

Our results showed that mating opportunity, female phenotype and their interaction critically affect the relationship between polyandry and fecundity in L. botrana. The main outcomes were as follows: 1) polyandrous females that could remate (i.e. in test group) had higher fecundity than once-mated polyandrous females, 2) polyandrous females that could remate had similar fecundity to that of monandrous females, 3) monandrous females had a higher fecundity than polyandrous females that only mated once, and therefore, 4) potentially polyandrous females denied the opportunity to remate suffered a drastic fecundity reduction. This response pattern was consistent for both strains, although the M strain had a slightly higher mean fecundity because females were a little larger and female size positively correlates with fecundity in L. botrana (Torres-Vila et al., 1999) as in most lepidopterans. Looking at these results, the central question remains without a clear answer: Does polyandry increases fecundity in L. botrana or not? We argue that there are two key issues that could greatly influence the correct interpretation of the relationship between polyandry and fecundity in this and other insect species.

The methodological approach: experimental vs. descriptive

The main outcome of the study was that conclusions about the polyandry–fecundity relationship in L. botrana are strongly dependent on the methodological approach used. According to the experimental method, female remating had an effect as fecundity was significantly higher in the test group (females mated at will) than in the control group (females mated once). In contrast, according to the descriptive method, female remating did not affect fecundity as in the test group the fecundity of polyandrous and monandrous females was similar. Surprisingly, the two methods produced seemingly conflicting results in the same species about the relationship between polyandry and fecundity. This is not a trivial matter because two independent studies that used different methodological approach would have produced opposite results. The scientific literature includes a number of cases of this kind in the Lepidoptera, producing inconclusive or contradictory results on the same target species depending on the method used, not only in monandrous species as Ostrinia nubilalis Hb., but also in polyandrous species as Helicoverpa armigera Hb., Earias insulana Boisd. and Spodoptera frugiperda Smith (Torres-Vila et al., 2004 and references therein).

The inconsistency of the results depending on the method used suggests at first sight a methodological artefact, but there is a simple analytical explanation. A direct comparison between the control and test groups (mating opportunity) following the experimental method is statistically inappropriate if there is a significant interaction between mating opportunity and female phenotype, as shown in this study. The statistical interaction is explained in biological terms because the fecundity of monandrous and polyandrous females is equivalent when they mate at will (test group), but it differs significantly when both female phenotypes are constrained to mate once (control group). The mating opportunity × female phenotype interaction (and its potentially biasing effect on the polyandry–fecundity relationship) is a crucial issue steadily neglected in experimental studies that has only received attention in a few papers (Tamhankar, 1995; Delisle & Hardy, 1997; Cook, 1999; Bergström & Wiklund, 2002).

The frequency of polyandry of the studied species or population could by itself affect results if an appropriate experimental design is lacking, so that the interaction between mating opportunity and female phenotype remains hidden. The reasoning is as follows. Polyandry in the Lepidoptera is a heritable trait (Torres-Vila et al., 2001, 2002; Wedell et al., 2002) which is also regulated by a large set of physiological and environmental factors (e.g. Torres-Vila et al., 1997, 2005; Larsdotter-Mellström & Wiklund, 2010). Polyandry is not thereby a species-specific trait because the equilibrium frequency may vary among populations depending on the selective pressures imposed by the ecological context. Hence, monandry and polyandry are two intraspecific mating strategies constituting a true balanced polymorphism (Torres-Vila et al., 2002). The observed polyandry frequency may also depend on stochastic processes inherent to the sample size in both field and laboratory researches. All these genetic, environmental and experimental factors will determine actual polyandry frequency in the target population and therefore the relative proportion of each female phenotype within experimental groups. If the frequency of polyandrous females increases, then it will be more likely to find differences between control and test groups. On the contrary, if monandrous females predominate, then differences between mating opportunities might not be significant. It follows that, even with a similar response pattern, the actual polyandry frequency could bias the comparison between control and test groups when the mating opportunity × female phenotype interaction is not controlled for in an experimental study (Torres-Vila et al., 2004). Note that this ‘frequency effect’ arises because there are more (or less) females that benefit from remating irrespective of whether polyandrous females benefit more than monandrous ones. The scientific literature provides some examples. For instance, in Danaus plexippus L. (the monarch butterfly), female remating did not affect fecundity in a population with 63% polyandry (Svärd & Wiklund, 1988) but it did in a population with 95% polyandry (Oberhauser, 1989). The frequency effect stresses again that it is indispensable to control for any hidden interaction in an experimental study. Our results did not, however, reveal a frequency effect presumably because fecundity data were only recorded in a subsample of monandrous females within the control group of the M strain. We conclude that the occurrence of a hidden interaction could help explain the inconsistent and even contradictory results between studies using the experimental method with the same target species.

The polyandry concept: realized vs. potential

The results from L botrana illustrate how the concept of polyandry used by the researcher can have a confounding effect and bias data interpretation, an effect previously detected through meta-analysis (Torres-Vila et al., 2004). This confounding effect is especially problematic when there is a hidden interaction between mating opportunity and female phenotype. First, if our data are viewed from the perspective of realized polyandry, all females in the control group are classified as monandrous as they only mated once, regardless of their post-mating receptivity. Note that, when focusing on realized polyandry, any interaction between mating opportunity and female phenotype goes undetected. As fecundity was significantly higher in the test than control group, and other things being equal, the conclusion would be that polyandry increases fecundity. Second, if one focuses on potential polyandry, the control group is no longer homogeneous because although all females mated once, some did recall (potentially polyandrous) and others did not (truly monandrous). Similar reasoning applies to females mated at will in the test group, as some females did mate multiply (polyandrous) and others did not (monandrous). The fecundity was equivalent in monandrous and polyandrous females mated at will, so the conclusion would be that polyandry does not enhance fecundity. The reasoning can be taken even further if one considers that truly monandrous females had higher fecundity than polyandrous females forced to mate once. From this viewpoint, the final conclusion would be that polyandry reduces fecundity.

Our results show why realized polyandry offers an oversimplified approach to understand the polyandry–fecundity relationship. Monandrous females of L. botrana that are forced into monogamy or chose not to remate had very different reproductive outputs. The recovery of sexual receptivity in females (as evidenced by recalling behaviour) was linked to a drastic fecundity reduction for potentially polyandrous females that were denied the opportunity to remate. The shorter the refractory period, the lower the realized fecundity, with fecundity differences between mating opportunity treatments becoming less marked with refractory periods longer than 4–5 days. Hence, in spite of females being able to alternate oviposition and recalling behaviour in the same dusk period, there was a significant tendency to reduce egg laying when recalling behaviour was triggered. This confirms that, in addition to female phenotype, the refractory period has a great impact on realized fecundity (Ridley, 1988; Vahed, 1998) at least for polyandrous females constrained to mate only once. The fecundity drop in once-mated polyandrous females was proportionally higher in the P than in the M strain, especially for the shortest refractory periods. This fact supports the claim that, as a result of artificial selection for polyandry, P-strain females became more reluctant to lay eggs than M-strain females when they were constrained to mate once.

The potential polyandry approach also reveals that L. botrana females with a polyandrous mating strategy assume a high risk when they recall to look for new mates because their fecundity may be seriously compromized if they fail to remate. The polyandrous strategy is even more puzzling when one considers that, even if polyandrous females do remate, other things being similar, multiple mating will not increase their reproductive output in relation to the monandrous strategy. A mate deficit for females is unlikely in the field because L. botrana males are highly polygynous (Torres-Vila et al., 1995), but female-biased operational sex ratios cannot be excluded, for example, in naturally sparse populations – the Allee effect – (Rhainds, 2010) or at the end of each seasonal flight as a result of protandry (Torres-Vila et al., 2005).

These results seem to challenge that spermatophores act as nuptial gifts in L. botrana for three reasons: 1) although all females received only one spermatophore in the control group, fecundity was dependent on female phenotype; 2) a positive correlation between spermatophore number and female fecundity was absent in the test group and 3) longevity was always higher in polyandrous than monandrous females regardless of the number of spermatophores received. Results also contradict previous suggestions that polyandry in insects might be explained solely in terms of direct benefits (Arnqvist & Nilsson, 2000; South & Lewis, 2011). Further research is necessary, however, to definitely elucidate this point as there still may be differences between polyandrous and monandrous females in their metabolic rate, so that while the former require male donations to achieve potential fecundity, the latter just rely on their own resources, as suggested in Pieris napi L. (Wedell et al., 2002). If so, monandrous females could have advantage over polyandrous ones under cold weather conditions (see also Välimäki & Kaitala, 2006), suggesting that monandry could also be adaptive even in species using nuptial gifts (Gwynne, 2008). Polyandry in L. botrana is also a mechanism counteracting infertile first matings (Torres-Vila et al., 1997), but other direct benefits (see Introduction) are unlikely to be important, as, for instance, offset sperm ageing (needless in a short-lived species) and sperm replenishment (monandrous females had sperm enough to realize their potential fecundity).

Rather to the contrary, the scenario seems to suggest that genetic benefits could be involved in keeping the polyandrous strategy if some females indirectly benefit more from remating by increasing the fitness of their offspring. More experimental research is necessary to support this notion, especially because evidence for genetic benefits to polyandry is weak in egg-producing animals, including not only insects, but also fish, reptiles and birds (see Slatyer et al., 2012, for a recent review). On the other hand, potential costs of remating are probably small in L. botrana, including male harassment and sexual conflict over mating rates (it is a cryptic, small and crepuscular species in which mating is under female control through long-range sexual pheromones) and increased exposure to male-borne parasites and diseases (most polyandrous females do not mate more than twice), whereas the effect of increased predation on pairs in copula is unknown (although presumably low since mating occurs at dusk and takes just about 30–45 min). The benefit–cost ratio of polyandry could help clarify the proximate and ultimate selective pressures maintaining female remating behaviour, particularly in monandrous species where male-donated nuptial gifts may be unimportant or lacking.

Concluding remarks

The evolutionary processes maintaining intraspecific female mating polymorphisms are largely unknown, despite variation in polyandry within species is widespread in both arthropods and higher animal taxa (Choe & Crespi, 1997; Birkhead & Møller, 1998). Our results show that the potential polyandry perspective could contribute to clarifying the selective pressures maintaining mating polymorphism in the propensity of females to mate multiply. Thus, the differentiation between potentially polyandrous and monandrous phenotypes in experimental studies could greatly improve our knowledge on the adaptive significance of polyandry. Determining whether a mated female is potentially polyandrous is easy in most moth species because recalling is a conspicuous behaviour that clearly denotes female receptivity. In other insects (as most butterflies and skippers), post-mating receptivity may be less evident because females do not exhibit (re)calling. Even so, female receptivity could be ascertained by observing the daily oviposition pattern or the mate refusal posture displayed when female is solicited by a courting male or a dummy. We need to know whether an interaction between mating opportunity and female phenotype that affects fecundity is widespread across insects and other animals and if there are species-specific response patterns depending on evolution-shaped polyandry rate. This could ultimately help to identify subtle fitness benefits to polyandry in the array of species in which some females remate for reasons other than increasing fecundity. Improved experimental protocols could help to elucidate how genetics and environment interact to determine why polyandrous phenotypes occur within mostly monandrous species (as L. botrana) but also why a number of females are monandrous within polyandrous species. Otherwise stated, why strictly polyandrous and monandrous reproductive strategies are so rare in insects and other animals.

Acknowledgments

I am very grateful to the Integrated Protection staff for their assistance in the moth rearing and laboratory work and to Dr. Michael Jennions for his valuable suggestions and English improvement. The author is still indebted to Dr. Jacques Stockel and Dr. Roger Roehrich for their advice and guidance during the early stages of this study. Financial, personal and material support was provided by the Servicio de Sanidad Vegetal (Gobierno de Extremadura, Spain) and by the former Station de Zoologie de Bordeaux (INRA, France). I also appreciate the valuable comments of three reviewers to improve the original manuscript. I have no conflict of interest to declare.

Ancillary