The evolution of female promiscuity poses an intriguing problem as benefits of mating with multiple males often have to arise via indirect, genetic, effects. Studies on birds have documented that multiple paternity is common in natural populations but strong evidence for selection via female benefits is lacking. In an attempt to evaluate the evidence more broadly, we review studies of multiple paternity in natural populations of all major groups of nonavian reptiles. Multiple paternity has been documented in all species investigated so far and commonly exists in over 50% of clutches, with particularly high levels in snakes and lizards. Marine turtles and lizards with prolonged pair-bonding have relatively low levels of multiple paternity but levels are nevertheless higher than in many vertebrates with parental care. There is no evidence that high levels of polyandry are driven by direct benefits to females and the evidence that multiple paternity arises from indirect genetic benefits is weak. Instead, we argue that the most parsimonious explanation for patterns of multiple paternity is that it represents the combined effect of mate-encounter frequency and conflict over mating rates between males and females driven by large male benefits and relatively small female costs, with only weak selection via indirect benefits. A crucial step for researchers is to move from correlative approaches to experimental tests of assumptions and predictions of theory under natural settings, using a combination of molecular techniques and behavioural observations.
The link between multiple mating and within-clutch multiple paternity is not straightforward, however (Dunn & Lifjeld 1994; Griffith 2007). For example, under selective fertilization (cryptic female choice; Eberhard 1996) multiple mating does not necessarily lead to multiple paternity and clutch sizes necessarily limits the extent to which multiple mating is reflected by multiple paternity. Thus, categorizing female mating strategies (e.g. monandry vs. polyandry) solely upon patterns of paternity (e.g. Richard et al. 2005; Eizaguirre et al. 2007) confuses patterns with processes and may lead to erroneous conclusions. Furthermore, there is often a discrepancy between the weight of field and laboratory evidence for multiple mating (Sakaluk et al. 2002). For example, whereas there is a huge amount of literature on multiple mating and sexual selection from laboratory experiments in insects, documentation of rates of multiple paternity (or mating) in natural populations of insects remain comparatively rare and most studies in wild populations have been conducted on birds (reviewed in Griffith et al. 2002). Nevertheless, a broad taxonomic approach is required to tackle the complexities of female mating strategies and multiple paternity in the wild as any taxon will exhibit specific shared characteristics that may make one or more explanations particularly likely or tractable. Birds, for example, are atypical vertebrates with respect to breeding (e.g. widespread biparental care) and reproductive traits [e.g. often lack of an intromittent organ and sequential (allochronic) ovulation of eggs], which could influence both selective processes on multiple mating and the degree to which it leads to multiple paternity. Thus, addressing the patterns and processes of multiple paternity in natural populations in a variety of organisms may lead to insights that are hidden from a specific taxonomic perspective. Here we conduct the first comprehensive review of the patterns of multiple paternity in wild nonavian reptiles, a diverse group of vertebrates, and a critical overview of the underlying processes in an attempt to provide a current consensus and framework for future research.
A brief overview of reptilian biology
Reptilia (excluding birds) is a diverse group of ectotherm animals comprising tuataras, crocodilians, turtles and squamates (lizards and snakes) (see Hedges & Polin 1999 and Townsend et al. 2004 for phylogenies). The extraordinary diversity in reproductive traits exhibited by reptiles makes it very difficult to generalize even within well defined taxonomic groups such as snakes. However, parental care of hatchlings is absent or rudimentary in all species and males do not provide any direct resources to the female before, during or after mating (although they may provide indirect resources via territory quality). Mating systems are generally categorized by intense male–male competition for females and, in many species, female- or resource-defence polygyny (e.g. Stamps 1977; Martins 1994; Shine 2003). Reproductive intervals range from days and weeks to years within lizards and freshwater turtles (e.g. Cogger 1978; Pearse & Avise 2001), whereas they are more consistently long (≥ 1 year) in snakes, marine turtles and crocodilians (e.g. Licht 1984; Seigel & Ford 1987). Breeding normally occurs according to a seasonal pattern even in tropical species (e.g. James & Shine 1985; Seigel & Ford 1987) but sperm production, mating and egg production can be decoupled in time (Seigel & Ford 1987; Aldridge & Duvall 2002). Clutch size is extremely variable, ranging from one to over 50 in squamates (Fitch 1970), whereas turtles and crocodilians can lay over 100 eggs in a single clutch (e.g. Greer 1975).
Causes of multiple paternity in reptile populations: the role of sperm storage
Multiple paternity can arise via two routes: mating with more than one male during the same reproductive cycle, or mating with one or more males during each reproductive cycle coupled with sperm storage across reproductive cycles. Sperm storage is widespread in all major reptilian taxa (reviewed in Schuett 1992; Olsson & Madsen 1998; Sever & Hamlett 2002; Uller et al. forthcoming). Its evolutionary causes are debated, however. In some species, low population densities could lead to a low rate of mate encounter and therefore a risk of sperm depletion, which would tend to favour prolonged sperm storage (Gist & Congdon 1998). This hypothesis is not strongly supported, however, because mate-encounter rates should be high for many species with capacity for sperm storage (for example, population densities of both lizards and snakes are often extraordinarily high, in particular during the mating season). In this latter category, selection for cryptic female choice could have driven the evolution of sperm storage if it enables female control of paternity and the retention of ‘optimal’ sperm across reproductive cycles (Olsson & Madsen 1998). However, even in reptiles with prolonged storage of sperm, sperm storage organs are relatively undifferentiated (Sever & Hamlett 2002) compared to, for example, the case in insects (e.g. Pitnick et al. 1999). This may suggest that the main reason for retention and survival of sperm in species where sperm storage is not obligate (as would be the case when mating occurs only in autumn and fertilization only in spring; Schuett 1992; Aldridge & Duvall 2002) can be explained by a strong selection for sperm longevity resulting from sperm competition and a quick turnover of reproductive cycles in females, with only weak selection on sperm storage in females.
Sperm stored over a long period of time is known to be capable of fertilizing eggs (e.g. Cuellar 1966; Pearse et al. 2002; Olsson et al. 2007) and multiple paternity resulting from mixing of stored sperm from matings in previous years and recently inseminated sperm has been documented in captive snakes (Agkistrodon contortrix; Schuett & Gillingham 1986) and painted turtles (Chrysemys picta, Pearse et al. 2001). However, although sperm storage may strongly affect the pattern of multiple paternity and compromise inferences about multiple mating (i.e. bias estimates upwards or downwards), its importance is difficult to evaluate due to the lack of studies. We will return to this issue when discussing the processes behind the documented patterns below.
Patterns of multiple paternity
Estimates of levels of multiple paternity in clutches from female reptiles in natural populations (and semi-natural enclosures) are summarized in Table 1 (see Appendix 1 for methods and statistical analyses). All else equal, the degree of multiple paternity across species should be positively correlated with the probability of mate encounters. Although it is difficult, and potentially misleading to generalize across species, species with low population densities during the reproductive season (e.g. sea turtles) and species with strong pair-bonding (social monogamy; Bull 2000; Chapple 2003) should have lower mate encounters during the female receptive phase than other reptiles. These two predictions seem to be upheld: (i) studies of turtles generally show lower estimates of multiple paternity than studies of squamates (turtles: 0.42 ± 0.070, N= 22; squamates: 0.55 ± 0.044, N= 35; χ2= 6.22, P= 0.013, N= 57; using only the largest data set per species: turtles: 0.42 ± 0.081, N= 11; squamates: 0.52 ± 0.056, N= 24; χ2= 3.82, P= 0.05, N= 35); and (ii) lizards of the closely related genera Egernia, Oligosoma and Tiliqua generally show lower levels of multiple paternity than other squamates (0.21 ± 0.061 vs. 0.61 ± 0.042; χ2= 20.38, P < 0.001, N= 35; using only the largest data set per species: 0.23 ± 0.071 vs. 0.59 ± 0.058; χ2= 10.22, P = 0.001, N= 24). Although these patterns presumably reflect comparatively low mate encounter and an unusually high degree of social structure (pair formation; Bull 2000; Chapple 2003; Chapple & Keogh 2006), respectively, the limited phylogenetic distribution of these traits among species for which we have available data precludes any further phylogenetically controlled test of the hypotheses.
Table 1. Multiple paternity in wild reptiles. Studies based on free-ranging animals in outdoor enclosures are indicated with*
The low levels of multiple paternity in social skinks notwithstanding, multiple paternity in squamates is extraordinarily common, often occurring at levels far above 50% of clutches in natural populations (Table 1). Despite the caveats regarding presence of sperm storage, this suggests a high degree of multiple mating with different males during each ovarian cycle by female lizards and snakes of virtually all species studied to date. In many species of squamates, the receptivity period is short and a high incidence of multiple paternity should reflect a high mate encounter, either generated by high population densities or by active female strategies to ensure multiple mating.
A further caveat is that multiple paternity will more easily be detected in large, compared to small clutches, both at the intra- and interspecific levels. In the overall data set, there was a positive correlation between the average clutch size and multiple paternity in squamates, but this was mainly driven by the small clutch size for pair-bonding skinks [testudines: r=–0.02, P= 0.95, N= 16; Squamates: r= 0.44, P= 0.023, N= 26; Squamates excluding Egernia and Tiliqua (clutch size estimates missing for Oligosoma, Table 1): r= 0.31, P= 0.16, N= 22]. The same pattern was found when we used the mean value per species from all studies (testudines: r= 0.67, P= 0.070, N= 8; squamates: r= 0.48, P= 0.043, N= 18; Squamates excluding Egernia and Tiliqua: r= 0.37, P= 0.20, N= 14; Fig. 1).
Mate encounter may also be affected by the mating system. In particular, strong territoriality will tend to reduce the number of males that females will encounter during receptivity. Indeed, the two species of lizards with comparatively low levels of multiple paternity in natural populations (Ctenophorus pictus and C. ornatus) are both highly territorial (and closely related). However, New World territorial iguanids (Sceloporus virgatus and Uta stansburiana) show a high frequency of clutches with multiple paternity as do the territorial wall lizard (Podarcis muralis) (Table 1).
Processes underlying multiple paternity
Multiple mating, and therefore multiple paternity, can arise for a number of reasons, some of which are adaptive to females, some of which are not (see Arnqvist & Nilsson 2000; Jennions & Petrie 2000; Birkhead & Pizzari 2002; Simmons 2005 for succinct reviews). Benefits are often divided into two categories—direct and indirect benefits—that broadly correspond to benefits arising from paternal contributions to egg production or parental care and those arising from increased genetic quality, complementarity or variation (genetic bet-hedging) of the offspring. Most direct benefits are unlikely to play an important role in reptiles for two reasons: (i) there is virtually no paternal care among reptiles (but see Shine 1988 for rare exceptions), which renders such explanations unlikely; and (ii) there is no evidence that ejaculates are utilized as resources and, even if they were, ejaculate sizes are unlikely to be large enough to significantly contribute to female resource levels (Olsson & Madsen 1998), in particular since many reptiles tend to rely heavily on stored resources for reproduction (i.e. they are commonly ‘capital’ rather than ‘income’ breeders; Bonnet et al. 1998).
However, a third group of benefits can also be classified as direct, i.e. ensuring the presence of sufficient sperm numbers to allow fertilization of eggs. Sperm limitation can arise if some males are infertile (e.g. due to a lack of mature sperm, Olsson & Madsen 1996; Olsson & Madsen 1998; Roig et al. 2000) or if some copulations result in transfer of insufficient sperm numbers (Ridley 1988; Török et al. 2003). The degree to which multiple mating in female reptiles ensures fertilization success has not yet been systematically addressed. However, the number of matings (with the same or different males) was negatively correlated with the incidence of infertility in the Swedish common lizard (Uller & Olsson 2005) and, in sand lizards, females risk mating with infertile males if they emerge from hibernation before maturation of spermatozoa is complete (Olsson & Madsen 1996; see also Olsson & Shine 1997). Furthermore, evidence from turtles suggests that sperm storage ensures access to sperm for production of successive clutches when mate-encounter rates are low (Pearse & Avise 2001), suggesting that sperm limitation could be a real problem for females in natural populations of some reptiles. A potential additional consequence of risk of sperm limitation is that females should mate with all available males, thereby reducing selection for precopulatory mate choice.
Indirect benefits of multiple mating can arise via at least four mechanisms that could be relevant for reptiles (Jennions & Petrie 2000). The most easily grasped is when there is a second-male fertilization advantage and females mate multiply to ‘trade up’, i.e. when they meet a male that is of higher quality than the one(s) they have mated with before (Halliday 1983; Jennions & Petrie 2000; see Pitcher et al. 2003 for empirical evidence in guppies). The other benefits arise from an advantage of simultaneous presence of sperm from multiple males before and during the process of fertilization. One is by promoting sperm competition among ejaculates (e.g. Curtsinger 1991; Parker 1998). This will ensure that eggs are fertilized by sperm from males who are successful in sperm competition situations and, if this trait is heritable, their sons will inherit a high competitive ability under sperm competition (the sexually selected sperm hypothesis, Keller & Reeve 1995). Alternatively, success under sperm competition may be (genetically) correlated with genetic quality more generally (i.e. promoting sperm competition increases the chances of good genes to the offspring; Jennions & Petrie 2000; see also Sheldon 1994). The third mechanism is that multiple mating allows cryptic female choice among ejaculates to increase offspring genetic quality or complementarity (Zeh & Zeh 1996; Tregenza & Wedell 2000). The fourth hypothesis is based on a benefit of multiple paternity per se. Under this hypothesis, genetic (and presumably phenotypic) diversity among offspring is favoured because of unpredictable fluctuations in selective pressures (genetic bet-hedging; Yasui 1997, 1998, 2001).
We now discuss the relevance of each of these hypotheses based on patterns of multiple paternity and the biology of reptiles.
Trading up. The trading up hypothesis predicts that a female will be more likely to remate if the second (and third, etc.) male is of higher quality than the previous male(s). Direct evidence in favour of this hypothesis in reptiles is lacking although positive correlations between multiple mating or multiple paternity and hatching success or offspring viability (Madsen et al. 1992; Olsson et al. 1994; Blouin-Demers et al. 2005; Eizaguirre et al. 2007; Zbinden et al. 2007) is consistent with this, as well as some other hypotheses relating to variation in intrinsic male quality (see below). However, no studies have shown that dead, aborted, or otherwise unviable offspring are sired by particular males, as would be predicted if this was the case. There is also little evidence that females actively seek partners, and that different males give rise to offspring of different phenotypes within a clutch (but see Calsbeek & Sinervo 2004; Stapley & Keogh 2005). Furthermore, the only experimental studies to date to investigate female mating behaviour in relation to mating status and male quality did not find an increased probability of remating for females presented with males of higher quality (females always mated indiscriminately; Olsson et al. 1996; Olsson 2001). The weak evidence for widespread importance of precopulatory mate choice in reptiles (Tokarz 1995; Olsson & Madsen 1995) may also make this hypothesis unlikely to apply in many species. However, the lack of evidence for mate choice may also reflect a lack of well-designed studies on female choice in reptiles, in particular with respect to olfactory cues (e.g. Lopez et al. 2003; Olsson et al. 2004; but see Jansson et al. 2005). Combining behavioural studies of mate preferences with molecular assignment of paternity could provide important insights into the role of the female in postcopulatory processes.
Promoting sperm competition. The second hypothesis assumes that males who are good at sperm competition give rise to offspring (or at least males) that are of high quality (Yasui 1997; Jennions & Petrie 2000). This assumption is untested in reptiles and its verification would require studies documenting that fertilization success in sperm competition is a heritable trait or that it is correlated with genetic quality more generally. No studies of the heritability of fertilization success exist in reptiles and there is also precious little evidence for intraspecific variation in sperm traits and their heritability (Schulte-Hostedde & Montgomerie 2006; Uller et al. forthcoming). However, positive correlations between number of partners (but not number of fathers) and offspring viability (e.g. Olsson et al. 1994) are consistent with this theory. It is also worth noting that intralocus sexual conflict may lead to different fitness returns from sons and daughters in relation to male sperm competition success (Pizzari & Birkhead 2002; Arnqvist & Rowe 2005), which implies that reaping such genetic benefits must involve sophisticated sex allocation decisions (Pischedda & Chippindale 2006; Fawcett et al. 2007). This could involve matching sperm from different males to offspring sex (Calsbeek & Sinervo 2004) or a more general sex ratio adjustment at the level of the clutch (Burley 1981; Olsson et al. 2005a). However, contrary to predictions from the genetic quality theory, sex ratios are biased towards daughters when female sand lizards mate with high-quality males (Olsson et al. 2005a, b; see also Calsbeek & Sinervo 2004 for a potential example of adaptive within-clutch sex ratio adjustment in U. stansburiana). Thus, it is crucial that the fitness returns on investment into sons and daughters in relation to paternal quality is put under empirical scrutiny rather than simply relying on theoretical model assumptions (Olsson et al. 2005a, b; Pischedda & Chippindale 2006).
Cryptic female choice. Multiple mating to increase the potential for cryptic female choice has been championed as the most likely cause for the evolution of female promiscuity in lizards (Olsson et al. 1994; reviewed in Olsson & Madsen 1998, 2001). This idea fits well with the observation that precopulatory female choice is rare in reptiles (Olsson & Madsen 1995; Tokarz 1995) and was also supported by early experimental and fieldwork in sand lizards (Olsson et al. 1996). However, both this hypothesis, and those relying on ‘good genes’, suggests that paternity will be strongly skewed towards one male (Eberhard 1996). Thus, if selection for multiple mating arises via cryptic female choice or to obtain genetic benefits via promoting sperm competition, female mating rate should be consistently higher than the degree of multiple paternity (i.e. mating is random but fertilization is not). In fact, multiple paternity in lizards and snakes is extraordinarily common and reach higher levels than that documented for any other vertebrate group (Table 1) and it frequently involves more than two males siring offspring within a given clutch (e.g. 1–5 sires in northern water snakes, Nerodia sipedon (Prosser et al. 2002); side-blotched lizards, U. stansburiana (Zamudio & Sinervo 2000); and loggerhead turtles, Caretta caretta (Zbinden et al. 2007)).
The high level of multiple paternity in the wild hence represents a problem for the hypothesis of female control over fertilization. Females may still obtain benefits by mating multiply if fertilization is random, however, since at least some offspring would result from fertilizations by high-quality (or complementary) males. However, assuming random mating and that male quality is evenly distributed (i.e. there is an equal chance of mating with a male being one standard deviation higher than the mean as with a male being one standard deviation lower than the mean), the average fitness of offspring would be the same for singly and multiply mating females. An alternative explanation is that there could exist multiple males within a population that are equally good for a female and that these males may share paternity whereas males of poorer quality fail to fertilize any eggs. But with many males of equal quality present within a given population, selection for cryptic female choice may not be sufficiently strong to allow mechanisms to evolve. Clearly, we need experimental tests of the importance of mating order and individual genotypes and phenotypes for paternity patterns (Olsson et al. 2004) before we can accurately address the role of promoting sperm competition and cryptic female choice in the evolution of female promiscuity. Ideally, this should be combined with measures of female fitness in relation to mating frequency and the degree of multiple paternity [both direct (Fitze et al. 2005; Le Galliard et al. 2005) and indirect via offspring or embryonic survival (Madsen et al. 1992; Olsson et al. 1994; Eizaguirre et al. 2007)]. Surprisingly, despite that the first documentations of a relationship between multiple mating and female fitness via offspring quality was conducted in snakes and lizards (Madsen et al. 1992; Olsson et al. 1994), and that one of the species subsequently showed nonrandom fertilization in mating trials (Olsson et al. 1996, 2004), these studies have not been followed by detailed laboratory and experimental studies that address model assumptions and predictions in other squamates. Importantly, at least in many squamates, statistical correction for mortality during early development (Olsson et al. 1999; Simmons 2007) can be ensured since resorption of eggs does not occur (Blackburn 1998; Blackburn et al. 2003) and all eggs are normally oviposited at the same time, which generates reliable estimates of embryonic mortality.
Bet-hedging. The fourth indirect benefit of multiple mating (and paternity) occurs when the creation of genetically (and phenotypically) diverse offspring within a clutch is favoured directly, commonly referred to as bet-hedging (Yasui 1997, 1998, 2001). Bet-hedging can evolve under fluctuating environmental conditions when the optimal phenotype is difficult to predict at the timing of mating (or fertilization) since it reduces variance in female reproductive success (i.e. maximizes geometric mean fitness, Yasui 1998, 2001; Fox & Rauter 2003; Sarhan & Kokko 2007). Although theoretically plausible, and sometimes referred to as a potential explanation for high levels of multiple paternity (e.g. Calsbeek et al. 2007), the conditions that favour genetic bet-hedging are quite restrictive (Yasui 2001; Sarhan & Kokko 2007). Furthermore, there is little evidence in any taxa that paternal genotypic variation gives rise to sufficient phenotypic variation to be of importance under fluctuating selection. An empirical problem is that a robust test requires a comparison of fitness of females that produce offspring from multiple fathers with those producing single-fathered clutches across breeding attempts that reflect the natural variation in selective regimes. This is clearly a daunting task for any field based research project regardless of the study organism and there is currently no available evidence that allow an evaluation of its importance for reptilian systems (see Sarhan & Kokko 2007 for a possible example in a butterfly).
Multiple paternity can arise without direct or indirect benefits to females (Lee & Hays 2004; Arnqvist & Kirkpatrick 2005). Selection on males to mate multiply is uncontroversial and will be particularly strong in mating systems that lack paternal care (Andersson 1994). Thus, multiple mating and multiple paternity will arise whenever the realized mating system is closer to the male optimum than to the female optimum. Assuming that there are no direct or indirect benefits to females from mating with multiple males and no cost of mating for males (which is doubtful if sperm is costly, see Olsson et al. 1996), the level of multiple mating in a population will be set by the frequency of mate encounter and the costs of mating to females. However, even in the presence of female costs, males may be able to enforce copulations. For example, in garter snakes (Thamnophis sirtalis), males are able to induce cloacal gaping and thereby allow intromission by pressing the female to the ground (Shine et al. 2003; Shine & Mason 2005). The level of multiple paternity will subsequently be affected by processes of sperm competition, which in the absence of any female-driven processes might constitute a ‘fair raffle’ (Parker 1998).
Under this scenario, multiple paternity in the wild will largely reflect the probability of encountering different males during each reproductive cycle. Importantly, mate encounter rates are likely to be higher on average in nonterritorial species and species with high population densities and lower in species that form more stable bonds between males and females (regardless of whether this is driven by strong territoriality or other factors). The patterns of multiple paternity across reptilian taxa are broadly consistent with this hypothesis. For example, among lizards, pair-bonding social skinks show lower degrees of multiple paternity than does reasonably ecologically similar lacertids and agamids and, in the latter, some territorial species have notably low incidence of multiple paternity for squamates (see Table 1 for references). Furthermore, populations of olive ridley sea turtles that breed and nest in aggregations have a higher proportion of multiply sired clutches than populations with solitary nesting and, across turtle studies, the proportion of multiply sired clutches is positively correlated with breeding population size (Jensen et al. 2006).
However, interspecific patterns should be treated with caution for a number of reasons:
1There is insufficient data at present to allow a proper phylogenetically controlled test of the relationship between categories, such as pair-bonding, territorial and nonterritorial lizards.
2The sample sizes for many species are low, suggesting large confidence intervals. Reassuringly, however, levels of multiple paternity for some species represented by low sample sizes have been confirmed with more exhaustive data (e.g. loggerhead and leatherback turtles, see Table 1 for references).
3The ecology of many species is simply not sufficiently well known to allow separation into distinct categories. For example, the actual opportunities for multiple mating may not be accurately reflected by male territoriality if females actively pursue multiple mating, as has been shown in some birds (e.g. superb fairy wrens; Malurus cyaneus, Cockburn 2004). Furthermore, although sand lizards are not territorial, they commonly guard females after copulation (Olsson et al. 1996; Gullberg et al. 1997), thereby essentially reducing the probability of female remating in the same way as would territoriality (but levels of multiple paternity are still high; Gullberg et al. 1997; Olsson et al. unpublished data).
4Multiple paternity is not the same as multiple mating. We have already discussed how indirect benefits to females can drive biased probability of fertilization (and the limited evidence thereof) but it is equally true that clutch sizes and sperm storage may confound the links between paternity and mating strategies. Although there was only weak evidence that clutch size explains variation in multiple paternity among species when pair-bonding skinks (with low clutch sizes and low rates of multiple paternity) were excluded, larger females with larger clutches have higher incidence of multiple paternity in some species (e.g. Chrysemys picta: Pearse et al. 2002; Lacerta vivipara: Eizaguirre et al. 2007) and it is uncertain to what extent this reflects differences in mating strategies. Furthermore, sperm storage cannot be ruled out as an important cause of variation among or within species. For example, in the painted dragon, Ctenophorus pictus, males mating early in the season (before ovulation of the first clutch) gain a large proportion of paternity also in later clutches (Olsson et al. 2007), thereby reducing levels of multiple paternity both within and across clutches (Olsson et al. unpublished data). There is also commonly a substantial variation among populations within the same species in key predictors, such as population density (e.g. Pearse et al. 2006), which makes comparison across species difficult.
5Finally, local and annual variation in climatic conditions (e.g. temperature, sunshine) will strongly affect reptilian activity patterns and thereby the strength of pre- and postcopulatory sexual selection (Uller et al. forthcoming), which may cause variation in multiple paternity at spatial and temporal scales (Prosser et al. 2002).
The last problem suggests that the preferred approach to test the role of population parameters on mating and paternity patterns would be to use experimental manipulation within species, in particular species that show large variation in key characters in natural populations. Recent experimental studies of the common lizard, L. vivipara, have taken important steps towards an understanding of sexual conflict for mating and paternity patterns. Studies using experimental enclosures where the adult sex ratio was manipulated showed that, contrary to predictions, the proportion of clutches with multiple paternity was not affected by the availability of males (Fitze et al. 2005; see also Laloi et al. 2004), but that the incidence of multiple paternity was not consistent across female age categories (Richard et al. 2005). Middle-aged females showed the lowest incidence of multiple paternity (Richard et al. 2005), which could reflect the outcome of female choice and sexual conflict. However, the studies also suggested that male harassment was an important factor in determining the degree of multiple paternity, since multiple paternity clutches in male-biased enclosures had more fathers per clutch on average than multiple paternity clutches in female-biased enclosures (Fitze et al. 2005). Furthermore, female reproductive success was reduced in male-biased enclosures, presumably via increased male harassment (Fitze et al. 2005; Le Galliard et al. 2005; see also Richard et al. 2005; Uller & Olsson 2005; Eizaguirre et al. 2007). However, the lack of data on individual interactions in all L. vivipara studies (all data comes from paternity analyses) makes it difficult to evaluate the relative importance of male and female mating strategies and postcopulatory processes for the observed patterns of paternity. Thus, behavioural studies of male and female reproductive strategies are a necessary and invaluable complement to molecular methods of paternity assignment (Griffith 2007).
As the studies of L. vivipara and T. sirtalis indicate, the cost of mating and mate choice may be of critical importance for the evolution of male and female mating strategies and, consequently, patterns of multiple paternity (see Kotiaho & Puurtinen 2007). Costs of mating in males are often considered to be negligible (at least in relation to the benefits; e.g. Bateman 1948; Parker 2006), but studies of the adder, Vipera berus, suggest that sperm production can infer a substantial cost (Olsson et al. 1996). Sexually transmitted disease may also impose an important selective pressure on the sexual behaviour of both sexes (Boots & Knell 2002; Knell & Webberley 2004). Unfortunately, our knowledge of sexually transmitted diseases in wild animals (including reptiles) remains poor (Knell & Webberley 2004), which precludes a detailed discussion of its role in the evolution of mating strategies. Additional costs to female reptiles include (i) physical and physiological harm; (ii) reduced time for other activities; and (iii) predation (e.g. Shine et al. 2000, 2003, 2004; Le Galliard et al. 2005). Interestingly, in L. vivipara, population sex ratios (in semi-natural populations) have a strong effect on female fitness by reducing female survival and clutch size in male-biased populations (Fitze et al. 2005; Le Galliard et al. 2005), presumably because of repeated mating attempts by males that incur both physical harm (the male bites the female on her abdomen during copulation and mating is unusually long in L. vivipara, Bauwens & Verheyen 1985; Heulin 1988), and reduced time for other activites, such as foraging. This suggests that costs of mating may be substantial in this species. However, since females may not necessarily have mated more frequently, or with more males, in male-biased populations, the costs may simply arise from repeated mating attempts and female resistance rather than successful matings (Fitze et al. 2005). Thus, not only mating per se, but resistance to mating, is an important cost to females that should be considered both theoretically and empirically.
Multiple paternity is ubiquitous in reptiles and occurs at high levels in all major groups, strongly suggesting high levels of female promiscuity. Although authors frequently suggest that indirect benefits to females are the selective pressure behind multiple mating, the available evidence suggests that it primarily arises through a combination of strong selection on male multiple mating (perhaps more so than in taxa with parental care), low degree of precopulatory mate choice, high mate-encounter rate and a relatively moderate cost of repeated mating to females (Lee & Hays 2004). In fact, since the first evidence for indirect genetic benefits to females via a positive correlation between multiple mating and offspring viability in the adder, Vipera berus (Madsen et al. 1992) and the sand lizard, L. agilis (Olsson et al. 1994, 1996), there has been very little further evidence to suggest that multiple mating or multiple paternity has positive fitness consequences in reptiles (but see Blouin-Demers et al. 2005; Madsen et al. 2005; Eizaguirre et al. 2007 for correlations between multiple paternity and estimates of hatching success or offspring viability). Both the sand lizard and adder studies were conducted on small and inbred populations. In these populations, multiple mating with different males resulted in a higher proportion of viable young (Madsen et al. 1992; Olsson et al. 1994) and, in the sand lizard, more distantly related males sired the majority of offspring both in the field and in staged sperm competition trials (Olsson et al. 1996). However, it is quite possible that a general tendency towards female promiscuity in these species have evolved in the absence of indirect female benefits via strong selection on males (as envisioned above, see also e.g. Shine et al. 2003, 2004) and that benefits only arise secondarily due to population-specific characters, such as low genetic variation and high degree of inbreeding depression (Olsson & Madsen 2001). Indeed, both sand lizards and adders show high levels of multiple paternity even within small populations (above 65% for both species; Gullberg et al. 1997; Ursenbacher et al. unpublished data), suggesting limited paternity bias in relation to genetic compatibility in the wild. Overall, there is currently little evidence that indirect benefits are important selective forces in reptiles in general and, in fact, the high levels of multiple paternity may argue against both cryptic female choice and ‘good genes’ scenarios. Similar conclusions have recently been advocated for birds (Westneat & Stewart 2003; Arnqvist & Kirkpatrick 2005, 2007; Albrecht et al. 2006; Akçkay & Roughgarden 2007; see also Kotiaho & Puurtinen 2007). Thus, we suggest that researchers should avoid relying too heavily on hypothesis of indirect benefits and take seriously models that make fewer assumptions and in which ecological parameters such as mate-encounter rates and costs of mating are of prime interest. Although they may be less spectacular, and require logistically challenging behavioural studies, such approaches are more likely to capture the evolutionary dynamics of mating strategies and multiple paternity in the wild. Nevertheless, our conclusion also reflects the paucity of studies that experimentally test assumptions (e.g. do high-quality males sire high-quality offspring? what is the cost of mating?) and predictions (e.g. do females bias fertilization success according to male quality?) from sexual selection and sperm competition theory. Many reptiles are well suited for studies that can bridge the gap between laboratory studies and fitness estimates in the wild (Wapstra et al. 2007). Given that the evidence is strong that multiple paternity is prevalent in all reptilian taxa, the time is ripe to move on from correlative to experimental approaches and to put theoretical models to the test.
We are grateful to Sylvain Ursenbacher, Dan Warner, Erik Wapstra and Geoff While for providing unpublished data and to Thomas Madsen, Tom Pizzari, Charlie Cornwallis and one anonymous reviewer for discussions. T. Uller was supported by the Wenner-Gren Foundations and the Australian Research Council. M. Olsson was supported by the Australian Research Council.
TU currently holds a lectureship at the University of Oxford. MO is Professor in evolutionary ecology at the University of Wollongong. Both have broad interests in evolutionary biology, with a main focus on life history evolution, sexual selection, and phenotypic plasticity. Most of their collaborative work is conducted on European and Australian Lizards.
Materials and methods for data collection and statistical analyses
Data were collected using literature searches of the ISI Web of Science and Biological Abstracts and by screening all papers initially obtained. Furthermore, we contacted researchers we knew had data and asked them to provide unpublished results. Information on the mean clutch size was not always available in the original publication and is indicated by ‘—’ in Table 1. Statistical analyses were only conducted on studies in natural populations (i.e. excluding semi-natural enclosures).
Overall, there was a negative correlation between the proportion of clutches with multiple paternity and the (log-transformed) total number of clutches examined for both major groups for which we had data (testudines: r = –0.36, P = 0.10, N = 22; squamates: r = –0.38, P = 0.023, N = 35). This patterned disappeared when using means per species (testudines: r = –0.44, P = 0.20, N = 10; squamates: r = –0.16, P = 0.47, N = 23). We conducted three analyses of the data based on a priori expectations. First, we used logistic regression (proc genmod, SAS 9.12) with the number of clutches with multiple paternity as the numerator and the total number of examined clutches as the denominator, a binary distribution and a logit link function to test for a difference between testudines (the vast majority of which were marine turtles) and squamates. The rationale being that population densities (and hence mate-encounter rates) of the former are frequently assumed to be smaller than for the latter. Second, we used the same approach to test for a difference between pair bonding squamates (genera Egernia, Oligosoma, and Tiliqua) and other squamates, based on the prediction that social monogamy should lead to lower mate-encounter rates and possibly higher cost of multiple mating (e.g. via mate desertion; Chapple 2003). We conducted our analyses twice; both using all data from natural populations and reducing the data set to one observation per species. In the latter analyses, we consistently retained the study with the highest sample size. It should be noted, however, that despite the nonindependent replication of species in the former, the reduced analysis may actually be more misleading, as some of the species for which we had multiple estimates deliberately included populations that differ in population size or density. Consequently, the remaining multiple paternity data in the reduced data set could become biased towards high-density populations (since the number of clutches sampled should be higher). We also conducted correlation analyses between the proportion of clutches with multiple paternity and the mean clutch size of paternity-assigned offspring to evaluate to what extent clutch sizes may constrain paternity estimates at the interspecific level. We did this both using the full data set and using the average across studies for species with more than one study.
Because our main predictors show strong phylogenetic conservatism, the data does not allow a meaningful analysis using independent contrasts. Thus, all analyses are potentially confounded by phylogeny and some taxa are clearly over-represented (Table 1) and the results from the statistical analyses should be considered preliminary and only serve as a complement to the extensive narrative evaluation rather than provide conclusive evidence in favour of a specific hypothesis. More detailed quantitative comparative analyses of covariation between multiple paternity and traits important in pre- and postcopulatory sexual selection should await further accumulation of data from natural populations (see Uller et al. forthcoming for detailed discussion).