Biologists have been studying mating systems for >2,300 years (Darwin, 1871; Aristotle, 1943; Trivers, 1972; Smith, 1984; Eberhard, 1996; Parker and Birkhead, 2013). Despite this longstanding interest, the descriptions of mating systems can be confusing and controversial, and subject to different interpretations depending on how they are quantified (Wickler and Seibt, 1983; Gowaty, 1985; Westneat et al., 1990; Klug, 2018). For example, behavioral ecologists used the term “polyandry” in the 1970s to describe an overall mating system in which the variance in female lifetime reproductive success (LRS) is greater than the variance in male LRS (Trivers, 1972; Emlen and Oring, 1977; Alexander et al., 1979). More recently, however, behavioral ecologists have used the term “polyandry” to indicate copulation with ≥2 males by a single female (Zeh and Zeh, 2003; Bretman and Tregenza, 2005; Firman and Simmons, 2008; Johnson and Brockmann, 2010; Gowaty, 2012; Zhao et al., 2016). In an attempt to reduce confusion, we use the terms defined in Table 1 throughout this report to describe a mating system.

An early way to estimate the number of sexual partners for males and females of a species was to observe the social organization. For an avian or mammalian species in which the typical pattern is for a single male and a single female to care for juveniles in an isolated nest or den, for example, a behavioral ecologist might hypothesize that females are behaviorally monandrous and males are behaviorally monogynous (Lack, 1968; Kleiman, 1977; Wittenberger and Tilson, 1980; Getz et al., 1990; Mock and Fujiok, 1990; Reichard and Boesch, 2003; Klug, 2018). For species for which the typical pattern is for several sexually mature females to rear offspring in separate nests in the same territory (i.e., the same physical area) with a single resident sexually mature male, we might hypothesize that females are behaviorally monandrous and males are behaviorally polygynous (Orians, 1969; Bartholomew, 1970; Downhower and Armitage, 1971; Emlen and Oring, 1977; Thornhill, 1981). For many species, however, a clear social organization that shows ≥1 specific female(s) living in the same territory with a specific male is not evident (e.g., one sex or the other frequently moves between territories) (Birkhead and Moller, 1992; Shuster and Wade, 2003; Danchin et al., 2008; Oliveria et al., 2008; Raveh et al., 2010; Clutton-Brock, 2016).

A better way to estimate the number of sexual partners for males and females for certain species is to observe copulations. If both males and females usually copulate with a single partner of the opposite sex, we might hypothesize that most females are behaviorally monandrous and most males are behaviorally monogynous (Kleiman, 1977; Black, 1996; Lukas and Clutton-Brock, 2013). If each female usually copulates with a single male and each male usually copulates with several females, however, we might hypothesize that most females are behaviorally monandrous and most males are behaviorally polygynous (Gladstone, 1979; Berger and Cunningham, 1991; Gibson et al., 1991; Alatalo et al., 1996). One problem with describing a mating season solely from observations of copulations is that one can almost never be certain that (s)he has detected every single mating for a male or female, for at least three reasons. First, for many species, copulations under natural conditions are often difficult to observe (Westneat et al., 1990; Birkhead and Moller, 1992, 1998; Byers, 1997). Second, a female's period of sexual receptivity extends over several days for many species (Dewsbury, 1975; Beach, 1976; Smuts et al., 1987; Birkhead and Moller, 1992, 1998), so that documenting every copulation is challenging. Third, a female of many species sometimes roams over long distances while she is sexually receptive (Westneat et al., 1990; Berger and Cunningham, 1991, 1994; Birkhead and Moller, 1992, 1998; Byers, 1997), so that tracking a receptive female and observing all her copulations are difficult.

A third way to define a mating system is to identify the possible sexual partners for males and females, and then use genetic markers to assign probability of maternity and paternity (Hanken and Sherman, 1981; Burke et al., 1989; Moore and Ball, 2002; Solomon et al., 2004; Weinman et al., 2015). If clutches or litters typically show single maternity and paternity, while also showing that males frequently sire juveniles of several clutches, then we can hypothesize that females are usually genetically monandrous and males are usually genetically polygynous (Ribble, 1991; Decker et al., 1993). Genetic assignments of paternity have the potential to reveal subtle nuances in a mating system that might not be apparent from behavioral observations alone (e.g., extra-pair paternity arising from undetected copulations). However, while genetic methods can be powerful for resolving paternity when genotypic diversity within a population is high, they suffer from reduced precision in situations when candidate males are close kin (e.g., two full brothers) and therefore genetically similar, or when the number of polymorphic loci and the number of alleles per locus are small (Blouin, 2003; Bonin et al., 2004; Coltman, 2005; Csilléry et al., 2006; Fernandez and Toro, 2006).

Single paternity of a clutch or litter (i.e., genetic monandry) can result even when a female copulates with ≥2 males (i.e., behavioral polyandry), for at least two reasons. First, via “sperm competition,” sperm from one male are sometimes better than sperm from other males at fertilizing eggs (Parker, 1970, 1984; Dewsbury, 1984; Moller and Birkhead, 1989; Wigby and Chapman, 2004; Firman and Simmons, 2008). Second, via “cryptic female choice,” a female that copulates with ≥2 males might selectively use sperm from only one male to fertilize all her eggs (Eberhard, 1996; Reeder, 2003).

Combining information from several approaches can lead to a richer understanding of animal mating systems, and reveal behavioral complexity that might not be apparent from relying on just one approach. Here we describe the results of research that compares behavioral and genetic estimates of polyandry and polygyny from social organization, observations of copulations, and assignments of paternity from microsatellites for white-tailed prairie dogs (Cynomys leucurus, hereafter “WTPDs”) living under natural conditions over 7 consecutive years. Many previous short-term studies with other species have yielded estimates of behavioral or genetic polyandry and polygyny from one or two of these methods (e.g., Lack, 1968; Orians, 1969; Westneat et al., 1990; Birkhead and Moller, 1992, 1998; Davies, 1992; Alatalo et al., 1996; Goossens et al., 1998; Dugdale et al., 2007; Raveh et al., 2011; Gowaty, 2012), but none has integrated large sample sizes for all three approaches in a wild population over many years.

Studies of mating systems offer intriguing examples of how social organization and copulations are often related to, but are not necessarily fully compatible with, the effective pattern of parentage revealed by genetics (Westneat et al., 1990; Birkhead and Moller, 1992; Solomon and Keane, 2007; Lukas and Clutton-Brock, 2013; Klug, 2018). Because each female remains in or near (within 20 m) her home territory and is sexually receptive for only a single day each year during a short mating season, WTPDs are exemplary animals for quantifying frequencies of behavioral polyandry and behavioral polygyny. In theory, careful watching of a WTPD female on her single day of estrus should reveal all males with whom she might have copulated. Further, live-trapping of all WTPDs within a colony to obtain tissue-samples is practical, so that estimates of genetic polyandry and genetic polygyny are also feasible. Our long-term research allowed us to compare estimates of polyandry and polygyny from social organization, observations of estrus and copulations, and assignments of paternity from genetics. Our combination of these different approaches offers unique and complementary insights into the mating system of WTPDs living under natural conditions. Our results show that the highly structured WTPD social organization (with multiple females commonly living in the same territory with a single male) influences the number and identity of copulatory partners. However, the frequencies of behavioral polyandry and behavioral polygyny are not strictly translatable to the frequencies of genetic polyandry (multiple paternity) and genetic polygyny, because some multiply mated females produce litters of single-paternity, and others produce no litters at all. Further, females observed to copulate with only a single male sometimes produced litters with multiple paternity, indicating the occurrence of both cryptic, undetected copulations, and sperm competition.

Our research indicates that 83% (226/272) of WTPD females appeared to have only one “social mate” because they lived in a discrete clan with only one male, or shared a male with females living in an adjacent clan. Similar designation of social mates is possible for some species, but not for many other species for which discrete social groupings are not obvious (Birkhead and Moller, 1992; Shuster and Wade, 2003; Danchin et al., 2008; Oliveria et al., 2008; Raveh et al., 2011; Clutton-Brock, 2016). Compositions of clans also suggested that 3% of females had ≥2 social mates, and that 14% of females had no social mate, because they lived in a clan with no male (Figure 5). For males, compositions of clans suggested that 41% had only one social mate, because they lived in a clan with only one female; compositions suggested that the other 59% of males had ≥2 social mates, because they lived in a clan with ≥2 females (Figure 10).

When compared to the frequencies estimated from social organization (i.e., compositions of clans), observations of copulations indicated a higher frequency for both behavioral polyandry and behavioral polygyny. These observations show that 27% of all females (N = 190 for which we were able to document estrus and all putative sexual partners per year) copulated with ≥2 males during the mating season, much higher than the 3% for which social organization (i.e., clans with ≥2 males) indicated polyandry (Figure 9). Our observations of copulations also showed that 84% of males (N = 77 for which we could record all putative sexual partners per year) copulated with ≥2 females during the mating season, substantially higher than the 59% inferred from social organization alone (Figure 10).

Even though each WTPD female's period of sexual receptivity was limited to several hours on a single day each year, 27% of females copulated with two or more males. Polyandrous WTPD females produced more offspring that lived for at least 9 months after weaning (our best estimate of female reproductive success) than monandrous females, and this advantage resulted mainly because polyandrous females were more likely to conceive and get pregnant (Hoogland, 2013b). With this clear advantage to fitness, why were not more females polyandrous? In proximate terms, mate-guarding by the first male to copulate with a female sometimes precluded copulations by additional males. The first-copulating male chased away other males, and tried to sequester the female in a burrow so that she was unable to search for additional males (Figures 2, 3). In ultimate terms, polyandrous WTPD females were less likely than monandrous females to survive until the next mating season (Hoogland, 2013b). This scenario indicates a life history tradeoff between current-year and future reproduction.

With an understanding that other factors are also important (Trivers, 1972; Emlen and Oring, 1977; Kleiman, 1977; Smith, 1984; Clutton-Brock, 2016), we predicted that frequencies of polyandry and polygyny would be lower for WTPDs than for many other species of animals with longer periods of sexual receptivity. Support for this prediction is minimal. Our estimated frequencies for behavioral polyandry (27%) and behavioral polygyny (84%) are similar to estimates of the frequency of polyandry and polygyny for animals for which females have longer periods of sexual receptivity that sometimes last for several days (Westneat et al., 1990; Birkhead and Moller, 1992, 1998; Davies, 1992; Alatalo et al., 1996; Dugdale et al., 2007; Raveh et al., 2011; Gowaty, 2012). These WTPD frequencies are also similar to estimates of polyandry and polygyny for other species of ground-dwelling squirrels ((Hanken and Sherman, 1981; Lacey et al., 1997; Goossens et al., 1998; Hoogland, 2007, 2013b; Raveh et al., 2010); Raveh et al., 2011).

Because of the diagnostic aboveground behaviors we observed just before and just after a WTPD underground consortship (anal sniffing of the female by males, mating calls by the male, self-licking of the genitals by both sexes, and late final submergence at the end of the day by the female), we are confident that we usually were able to determine which male(s) consorted with a female on the single day when she was sexually receptive. Fifty-six percent of females copulated exclusively underground, however, so we could not be 100% certain that every underground consortship involved a copulation. Further, some females were especially secretive with their estrus and copulations, so that we could not accurately record their sexual partners. Consequently, litters showing multiple paternity sometimes resulted for mothers that appeared to be behaviorally monandrous. By contrast, however, litters that showed single paternity almost always resulted from females that were behaviorally monandrous. Thus, while genetic paternity assignments usually agreed with expectations from social organization and copulations (Figures 12, 13), overall about 11% of juveniles (78/718) were assigned paternity to a male outside a mother's home territory and for which we observed no copulations with the mother.

What does it mean when a WTPD male from an adjacent territory had the highest LOP for a juvenile but evidently had no copulations with that juvenile's mother on her day of estrus? At least three explanations are possible. The first is that the copulating male and the female had an underground consortship that was exceedingly subtle, so that we did not detect it—even though we were watching every day from dawn to dusk during the entire mating season, we focused on males and unmated females, and we especially concentrated on females that showed one or more signs of estrus. Support for this first explanation is our finding that some WTPD females were more covert than others with estrus and copulations. Some estrous females, for example, had only 1 underground consortship, elicited no mating calls, and had <5 interactions with only 1 male; documenting copulations for these secretive females was challenging. Estrous females at the other end of the spectrum, by contrast, had as many as 21 underground consortships with as many as 5 males, elicited as many as 26 mating calls, and had as many as 84 interactions with several males; documenting copulations for these conspicuous females was much easier.

The second possible explanation for differences between behavioral and genetic data concerns statistical errors (i.e., false positives or false negatives) in our assignments of paternity. When amplifying and scoring microsatellite loci, we included a large number of sample duplicates (N = 27 individuals with duplicate tissue samples) that confirmed we had a low genotyping error rate (0.5% of scored genotypes), and Cervus 3.03 considered this error rate during calculations of LOPs. We are therefore confident that scoring errors were highly unlikely to lead to false positives in assignments of paternity.

A third possible explanation for the discrepancy between behavioral and genetic data is the low variability across microsatellite loci at our study colony. Despite extensive initial screening efforts, we were able to find only seven microsatellite loci that were polymorphic, and the combined non-exclusion probability across all seven loci was relatively low at 0.075. Greater resolution for assigning genetic paternity might be possible in the future via development of a large number of diagnostic genome-wide markers such as single nucleotide polymorphisms (SNPs). Other studies have shown the utility of SNPs for genetic paternity assignment in recently bottlenecked or kin-structured populations (e.g., Tokarska et al., 2009; Weinman et al., 2015).

The low frequency of microsatellite polymorphisms we observed has been noted previously (Sackett et al., 2009), and likely reflects a biological characteristic of WTPDs in general. Perhaps this low frequency results from a historically low or bottlenecked effective population size that might have occurred at some point (Seglund et al., 2006). Nevertheless, the low levels of microsatellite variation at our study-colony limited our statistical power to assign paternity with a high confidence threshold (95%), and prompted us to use a less stringent confidence threshold (80%) when assigning paternity based on LOP. Thus, even though 164/718 (23%; see Figure 12) of juveniles were assigned paternity from LOP to a male from an adjacent territory when that male was never observed to consort underground or copulate aboveground with the mother when she was in estrus, this frequency is about what we would expect by chance alone at an 80% confidence threshold. The frequency of paternity assigned to males who were never observed to copulate drops to 11% when we instead consider the criteria based on non-exclusion (Figure 13). In these cases, when male-A copulated with female-B and could not be excluded from paternity, we think that he is a more likely sire than a male with a higher LOP (based on 80% threshold) who was not observed to copulate with the mother. We consider 11% to be our most accurate estimate of the true rate of cryptic genetic paternity at our WTPD study-colony over our 7 years of research there.

We predicted that our three approaches would yield similar estimates of polyandry and polygyny, but our results provide only minimal support for this prediction. Our results over seven consecutive years from both copulations and genetic assignments of paternity indicate an average rate of 27% for both behavioral and genetic polyandry (Figure 9). However, the year-to-year variability in these estimates was substantial (Figure 9). Further, annual variation in our estimates of behavioral and genetic polygyny was also large (Figure 10). A single year of observations of copulations and genetic assignments of paternity would have given a misleading impression of the overall frequencies of behavioral and genetic estimates of polyandry and polygyny for WTPDs. Several other studies have documented large annual variation in the frequencies of behavioral and genetic polyandry and polygyny as reported here for WTPDs (e.g., Schwagmeyer and Brown, 1983; Birkhead and Moller, 1992, 1998; Hare et al., 2004; Isvaran and Clutton-Brock, 2006; Bergeron et al., 2011; Wells et al., 2017). Possible reasons for so much annual variation include yearly differences in factors such as weather (Bergeron et al., 2011; Martin et al., 2014), density of females (Emlen and Oring, 1977; Wells et al., 2017), female age and female body mass (Cotton et al., 2006; Jones et al., 2012; Hoogland, 2013b), reproductive synchrony among females (Emlen and Oring, 1977; Hoogland, 2007), and the operational sex ratio (i.e., ratio of the number of sexually mature females ready to copulate to the number of sexually mature males ready to copulate; Emlen and Oring, 1977; Michener and McLean, 1996). With so many variables and only 7 years of information, we made no attempt to make a quantitative dissection of the underlying causes of the extreme annual variation in the mating system of the WTPDs at our study-colony.

For 86 of the 135 juveniles (64%) weaned by behaviorally polyandrous females, the 1st-copulating male had a higher LOP than 2nd- and later- copulating males (for similar results, see Hanken and Sherman, 1981; Schwagmeyer and Foltz, 1990; Hoogland, 1995; Lacey et al., 1997; Raveh et al., 2010). By contrast, when a female was behaviorally monandrous, she showed single paternity of her litter at a frequency of 80% (99/123 litters) (Table 2). Rates of genetic polygyny estimated from non-exclusion and rates of behavioral polygyny from social organization were always lower than rates of behavioral polygyny from observations of copulations, which indicated that a mean of 84% of males copulated with ≥2 females (Figure 10).

Previous studies documenting both behavioral and genetic polyandry and polygyny for a diverse array of other animals typically have involved smaller sample sizes over a shorter time span (Hanken and Sherman, 1981; Burke et al., 1989; Moore and Ball, 2002; Solomon et al., 2004). Some previous researchers have used the term “polygynandry” to describe a species in which behavioral and genetic estimates of polyandry and polygyny are both high (Davies, 1992; Baker et al., 2004; Dugdale et al., 2007; Munroe and Koprowski, 2011; Raveh et al., 2011; Schwanz et al., 2016). Perhaps polygynandry is also a good term to describe WTPDs, for which females commonly are both monandrous and polyandrous (behaviorally and genetically) and males commonly are both monogynous and polygynous (behaviorally and genetically).

Male and female WTPDs appear to have a conflict of interest regarding the optimal number of sexual partners per year for each female. Because polyandry yields more offspring that live for at least 9 months after weaning, a female maximizes annual reproductive success by mating with more than one male. A male, by contrast, maximizes annual reproductive success by monopolizing, and then being the only male to copulate with, as many females as possible (Figures 2, 3). Neither sex is completely “winning” this intersexual conflict of interest regarding the optimal number of sexual partners for females (Figure 9), and the opposing sexual selection pressures on males and females evidently have produced the polygynandry that we observed.

The overall frequencies of polyandry from observations of copulations vs. polyandry from genetics were identical (27%; Figure 9). This suggests that cryptic (post-copulatory) female choice is minimal, but pre-copulatory female choice might still be possible. Consistent with this latter notion, we observed that a WTPD female commonly did not copulate with all the available males that resided in, or temporarily visited, her home territory on her single day of estrus (J. L. Hoogland, personal observation). These insights emphasize the benefits and insights that are possible from combining both behavioral and genetic data measured at different points in the processes of copulation through the weaning of offspring.

Each of our three methods for documenting polyandry and polygyny among WTPDs has advantages and disadvantages (see also Hanken and Sherman, 1981; Gowaty, 1985, 2012; Westneat et al., 1990; Hunter et al., 1992; Kempenaers et al., 1995; DeWoody et al., 2000; Raveh et al., 2010; Mooring and Penedo, 2014). Estimating behavioral polyandry and behavioral polygyny from social organization requires neither observations during the mating season nor collection and analysis of DNA samples, for example, but this approach is only tangentially related to male and female annual reproductive success. Indeed, a study of social organization alone is the only approach that does not provide unequivocal evidence for either polyandry (behavioral or genetic) or polygyny (behavioral or genetic). Observations of copulations offer the best way to document behavioral polyandry and behavioral polygyny. Behavioral polyandry does not always produce multiple paternity, however, and the observation of copulation(s) with a single male does not necessarily rule out cryptic behavioral polyandry via undetected copulations with other males. Further, direct observations during the mating season require an inordinate number of person-hours of watching to generate large sample sizes. Estimates of paternity from genetics provide a powerful complement in these cases, and sometimes can provide the ultimate, unequivocal documentation of which, and how many, juveniles a male actually sires (Burke et al., 1989; Moore and Ball, 2002; Solomon et al., 2004; Weinman et al., 2015). Genetic analyses of paternity sometimes have little statistical power when genetic variation is low (as occurred at our study-colony), however. The choice of method(s) for estimating the frequencies of behavioral and genetic polyandry and polygyny will depend on the biology of the study-animal (e.g., degree of coloniality, ease of livetrapping, visibility of copulations, length of the mating season, and genetic variation) and research objectives. The combination of behavioral and genetic methods not only capitalizes on the contrasting strengths and weaknesses of each approach, but also reveals information about dynamic mating systems that would not be possible from using only one method. Information from WTPD social organization, for example, helped us to know which males were likely to copulate with an estrous female, and information from both social organization and copulations helped us to know which males to consider as possible candidate sires for each offspring. Our research was an extension of previous research on other species of ground-dwelling squirrels (e.g., Hanken and Sherman, 1981; Lacey et al., 1997; Goossens et al., 1998; Raveh et al., 2011; Wells et al., 2017), and should be informative for other social species as well. Our results underscore the valuable insights that can result from integrating diverse perspectives on the mating system gathered from long-term behavioral and genetic studies of natural populations.

Behavioral and genetic data analyzed in this study are in the tables and figures or have been deposited into Dryad (doi: 10.5061/dryad.jn365c2).

With help from research assistants, JH was responsible for collecting all the information on social organization and copulations. SK and RT were responsible for all the genetic analyses of paternity. JH and SK wrote the manuscript.