Madsen (2008) questions our (Uller & Olsson 2008) conclusions regarding patterns in reptilian multiple paternity and the processes that drive them. We counter this criticism below but would first like to point out three things: (i) our review is not primarily about the evolution of polyandry — it is a review of empirical data showing that more than one male sires offspring within the same clutch in lizards, snakes and turtles, taxa that largely lack parental care and in which female precopulatory choice is rare (Tokarz 1995; Olsson & Madsen 1998; Uller & Olsson 2008). Multiple paternity depends on female promiscuity but the frequency of multiple paternity in a population, and the degree of paternity skews within clutches, is set by many factors of which female voluntary remating is but one. (ii) The focus of Madsen's comment is the evolution of polyandry from a perspective of indirect genetic benefits, either via sperm competition, as in Madsen et al. (1992), or assortative fertilization with respect to parental genotypes, as in Madsen et al. (2005). We will refer to the latter as cryptic female choice, and sometimes combine the two for simplicity into post-copulatory processes (PCP). We reflect on Madsen's arguments for polyandry evolution. (iii) Finally, Madsen suggests that the paucity of support for PCP-mediated genetic benefits is due to low sample sizes in outbred populations. We discuss this issue and how we believe the scientific rigor of research on indirect benefits and PCP in reptiles can be improved.
(i) Multiple paternity is the complex net outcome of pre- and postcopulatory mate choice, male–male competition, coercive matings, sperm storage and a number of less obvious processes, such as climatic constraints on mating opportunities in ectotherms, operating in a density and operational sex ratio-dependent manner (Uller & Olsson 2008). Our review aimed to rank factors contributing to multiple paternity, not only to multiple mating (polyandry), and showed that multiple paternity is (virtually) ubiquitous in lizards, snakes and turtles. Furthermore, multiple paternity is a double-edged sword for a polyandry crusader; yes — it demonstrates that a female mated more than once (perhaps forcibly), but also that she never completely biased paternity towards a ‘best’ male (otherwise we would see multiple matings but single paternity). Thus, although our review reveals widespread opportunity for evolution of processes such as cryptic female choice, it also demonstrates its lack of perfection. The degree to which we expect to see cryptic female choice effects on probability of paternity are, hence, expected to be context dependent (when less important to female fitness, we expect stronger male-driven selection, e.g. positive covariation between risk of sperm competition, sperm number and testis size).
That males benefit from multiple mating is uncontroversial. The most parsimonious explanation for the evolutionary origin of polyandry therefore seems to be some male-driven selection, with female benefits arising later (and only sometimes), hence our reluctance to accept indirect genetic benefits of PCP as an explanation to multiple mating and multiple paternity generally. When PCP-mediated genetic benefits have been rigorously demonstrated, we have fully embraced this (e.g. Olsson et al. 1994a, b, 1996; Olsson & Madsen 1998). This is also why we suggest that among-population differences in genetic architecture are likely to differently affect offspring viability and selection for pre- or postcopulatory mate choice. We are surprised that Madsen dismisses this, because experimental work in a different population of adders (Vipera berus) to Madsen et al.'s (1992) failed to replicate Madsen et al.'s correlative polyandry effects on offspring viability (Capula & Luiselli 1994). Finally, our statement ‘it is quite possible that a general tendency towards female promiscuity in these species has evolved in the absence of indirect female benefits via strong selection on males’ is phrased in the perspective of ubiquitous multiple paternity across all species of lizards and snakes, regardless of population size and evolutionary history.
(ii) Verifying ongoing selection for polyandry is not the same as identifying its evolutionary origin. In summary, Madsen suggests that polyandry evolves because of indirect genetic benefits via PCP, in particular cryptic female choice. This is an interesting hypothesis that requires precise reasoning. The evolutionary origin of female choice is a classic question in sexual selection theory, far from resolved, and with most theoretical attention devoted to pre- rather than postcopulatory mate choice (e.g. Andersson 1994; Arnqvist 2006; Postma et al. 2006; Qvarnström et al. 2006a, b). How, and under what conditions cryptic female choice would arise ‘spontaneously’, without prior, strong selection on multiple matings in males, has as far as we know never been researched. Madsen (2008) suggests that we would not have arrived at our ‘mistaken conclusion’ had we consulted the literature he cites. However, the primary authors of some of these reports were careful to mention alternative explanations or cite studies where no, or even negative effects, of polyandry on offspring viability have been reported in their models or related taxa (e.g. Hoogland 1998; Byrne & Roberts 1999; Kraaijevald-Smit 2002; Blouin-Demers et al. 2005; see also Jennions & Petrie 2000; Jennions et al. 2004, 2007). What is needed is a phylogenetically more complete and unbiased data set to assess if, how, and when polyandry evolves in response to opportunity for females to access good or compatible paternal alleles/genes for their offspring, what corresponding paternity biases towards different genotype males result from this process, and whether there are gains in offspring viability or reproductive success resulting from such paternity bias. Furthermore, we need to examine alternative hypotheses, such as whether females avoid costs of harassment from males by accepting copulations, and to what extent such costs are driven by population parameters such as density and operational sex ratio.
The genetic benefit of mate choice is a hotly debated topic, both with regards to ‘good genes’ and ‘compatible genes’. In brief, genetic mate choice evolution in postcopulatory cases should be subject to similar constraints to those that apply to precopulatory cases. For example, the lek paradox (why and to what extent mate choice is maintained despite depleting additive genetic variation) has been examined exclusively from a perspective of precopulatory mate choice. Thus, we do not know to what extent the precision in mate choice at a haploid compared to diploid level results in depletion of genetic variation at a faster or slower rate than in precopulatory mate choice. This is a blank page in postcopulatory sexual selection and needs to be theoretically examined in order to predict whether female postcopulatory genetic mate choice is likely to occur widely. Recent theoretical advances show that nonadditive genetic effects can be translated into heritable genetic variation, potentially contributing to a resolution of the lek paradox and similar phenomena (Lehman et al. 2008; Neff & Pitcher 2008). Still, to what extent depletion of genetic variation poses a constraint on the evolution of PCP and polyandry is completely unstudied.
(iii) Experimental design, statistical power and data interpretation. Madsen suggests that researchers have failed to detect genetic benefits from polyandry in outbred populations because of low sample sizes. We think the argument that widespread lack of power should explain rare reporting of genetically compromised offspring is relatively weak and, more importantly, misses important factors relating to experimental design and data interpretation. First, smaller benefits of polyandry in outbred populations also mean weaker selection on polyandry in such populations. The available evidence suggests that multiple paternity occurs across all populations of the same species, but that the degree of paternity skew may vary among populations, and that rates of multiple paternity are negatively correlated with population density (Jensen et al. 2006; Uller & Olsson 2008). This is strong evidence that selection scenarios varies among taxa and populations. Second, the reference provided (Madsen & Shine 1998) contains no information on polyandry (it demonstrates skewed hatching and recaptures success from some, but not other, clutches in the water python, Liasis fuscus). Madsen et al. (2005) however, show that the longer wild-caught python females were kept in captivity (awaiting oviposition), the fewer male microsatellite alleles were detected in their broods, and the poorer was the corresponding hatching success. Although this reference contains interesting correlative evidence, it is hard to reconcile increased genetic diversity within a clutch with the evolution of cryptic female choice of paternal alleles. Selection acts on the individual sibling and any fitness benefits to the mother from genetic diversity within a clutch lends more support to a genetic bet-hedging mechanism (i.e. with females creating genetically different rather than genetically ‘better’ young, a hypothesis fraught with its own problems, and an unlikely general explanation for the evolution of polyandry; Yasui 1998; Uller & Olsson 2008). A more precise test of a ‘shopping for paternal alleles’ hypothesis would have been to compare half-siblings of known paternity and genetic architecture with respect to components of viability and fitness. Furthermore, Kingsolver et al.'s (2001) summary of > 2500 estimates of selection across 62 species and 63 studies showed that the strongest selection coefficients were from studies of low sample sizes (i.e. most likely inflating the strength of ongoing selection). Madsen et al.'s (2005) study, demonstrating the link between paternal genetic diversity and hatching success, was based on 14 clutches. One of the largest field studies ever conducted shows heritabilities of fitness of about 4% or less, based on > 8500 ringed collared fly catchers (Ficedula albicollis) studied over 24 years. Again, this suggests among-population differences in the detection of additive genetic components of fitness, even with sample sizes far greater than usually achieved (Qvarnström et al. 2006a).
In conclusion, correlative evidence from natural populations, preferably in combination with formal selection analyses (sensu Lande & Arnold 1983 and followers), can prove extraordinarily powerful for directing research on mating system evolution (e.g. Qvarnström et al. 2006a). However, such work has to be followed by carefully controlled experimental analysis of causal relations and mechanisms within taxa, and comparative analysis across taxa. The rigorous experimental protocols and careful evaluation of alternative explanations that are employed in the study of multiple mating in some taxa (e.g. Jennions et al. 2004, 2007; Fisher et al. 2006) have yet to be fully embraced in studies of reptiles. The gold standard of polyandry studies (Fisher et al. 2006) experimentally verified that polyandry effects via sperm competition increased offspring survival threefold in a marsupial (even in this case, we do not know whether polyandry first evolved to enhance offspring viability). For many squamate reptile species, Fisher et al.'s approach can easily be applied, used to generate large numbers of half-siblings that can be molecularly characterized at loci of interest, for which maternal effects are controlled, and for which fitness can then be estimated in the wild.
Madsen et al. (1992) led the way by empirically lending support to Walker's (1980) proposition that females may benefit from polyandrous matings. The time is now ripe to experimentally investigate proximate and selective causes of multiple paternity and the ubiquity of offspring viability effects from multiple matings in the wild. Short-lived squamate reptiles would lend themselves very well to this enterprise.