In the insect order Hymenoptera, fertilized eggs usually develop into diploid females and unfertilized eggs develop into haploid males. However, because of the single-locus complementary sex determination mechanism ancestral in this order (Crozier 1977; Periquet et al. 1993; Crozier & Pamilo 1996), diploid males are occasionally produced via homozygosity at the sex-determining locus. If a female's mate shares one of her alleles at this locus, their diploid offspring will be 50% female and 50% male.

Diploid males are considered ‘reproductive dead-ends’ for at least two reasons. First, they are a drain on colony resources because they perform no labour and are produced at the expense of female workers. Second, diploid males are often functionally sterile, producing unreduced diploid sperm that result in sterile triploid offspring if fertilization occurs (Whiting 1961; Naito & Suzuki 1991; Ayabe et al. 2004; but see Cowan & Stahlhut 2004).

Diploid males have been reported in over 39 hymenopteran species (Crozier & Pamilo 1996), but rarely in social wasps. This is surprising given that social wasps, especially Polistes species, serve as model systems for social evolution research (Reeve 1991; Gamboa 1996; Queller et al. 2000). Diploid males have been observed directly in only one population of a single Polistes species, P. chinensis antennalis (Tsuchida et al. 2002, 2004). The presence of triploid females recently reported in three North American Polistes species suggests only indirectly that diploid males are more common than previously thought and are capable of successful mating (Liebert et al. 2004). Two critical pieces of evidence would confirm this speculation: direct observation of diploid males and diploid male paternity of triploid offspring.

Additionally, we lack information regarding the behaviour and reproductive success of diploid males and triploid females. In temperate climates, the Polistes annual colony cycle consists of four phases (Reeve 1991). In the spring, mated females (‘foundresses’) build nests alone or in small groups and perform all colony tasks until female workers emerge in early summer. Workers maintain the nest and care for reproductive-destined male and female offspring that emerge in late summer/early autumn. These offspring eventually leave the nest to mate and mated females enter winter diapause until spring, when the cycle repeats. Triploid females have been collected among spring foundresses (Liebert et al. 2004), which demonstrates that they are able to survive winter diapause. However, we do not know whether triploids mate or lay eggs. If triploids are sterile, their best nesting option would be to aid relatives.

Here we report the first direct observation, to our knowledge, of diploid males in Polistes dominulus. Using a combination of genetic and morphological data, we also present evidence that diploid males mate and produce triploid female offspring, and that despite their probable sterility, these triploid females mate, overwinter and join spring nesting associations.

Table 2 Morphological and offspring production data for adult females on three nests collected from Carlisle, MA, USA in 2003. (Bold type indicates triploid status. All others are diploid.)

Triploid offspring genotypes were consistent with being produced by one collected foundress mated to a diploid male (figure 1). At each locus, all offspring genotypes contained: (i) one allele that was also present in the foundress, and (ii) the same one or two additional alleles presumably contributed by the male's diploid sperm. The two foundresses on this nest were unrelated, allowing unambiguous maternity assignment.

Both triploid foundresses were mated and their ovaries contained yellow bodies, indicating previously laid eggs. However, neither foundress possessed more than one mature oocyte and one foundress had only four instead of six ovarioles. Both triploid foundresses' genotypes were incompatible with having produced offspring in their respective colonies or with being related to their cofoundresses.

We have presented the first direct evidence, to our knowledge, for diploid males in P. dominulus and the first evidence that these males are capable of mating and producing triploid offspring via unreduced diploid sperm. Our data also show that a female's failure to recognize a diploid mate leads to a cascade of costly recognition errors. For example, a triploid-producing foundress was able to monopolize reproduction on a nest with an unrelated cofoundress, and the presence of triploid pupae shows that neither foundress recognized and removed triploid brood. Both triploid foundresses had mated; thus their mates also may have been unable to recognize the females' triploid status. Although helping kin would have provided these foundresses with their only possible fitness benefit, they joined non-relatives. At this point, the ‘dead-end’ originating with the diploid male is finally reached 2 years later.

These results have special meaning because P. dominulus is a very successful species in both its native and introduced ranges and has been actively studied in Europe since the 1940s (e.g. Pardi 1948). Despite its recent introduction (1978), P. dominulus has been the subject of two decades of behavioural research in North America. However, our data may prove problematic for current methodology, in which behaviour studies are conducted without regard to population genetics. Such studies are common in Polistes wasps, which are often used to develop and test theories of social behaviour such as sex investment ratios and reproductive skew models (e.g. Nonacs 2002). Without genetic data to identify factors such as diploid male and triploid female production, we risk interpreting these studies erroneously.

Investigation of sex investment ratios is complicated by the consideration of diploid males (Vargo 1996). For example, previous studies have suggested that early males served an adaptive function as potential mates for early emerging female offspring (Kasuya 1981; Strassmann 1981; Suzuki 1985, 1997; Page et al. 1989). If early males are diploid, they may not be able to mate or will produce sterile offspring. Previous authors found 95% diploidy among early males in one population of P. chinensis antennalis (Tsuchida et al. 2002, 2004) and concluded that the reproductive success of early males could be ignored in that species. Our results demonstrate that some diploid male P. dominulus may also be an unintended consequence of the sex-determining mechanism; however, two out of the five nests in this study contained haploid males. Consequently, we cannot make assumptions about the ploidy of early males in this species without collecting relevant genetic data.

Another example demonstrating the necessity of testing assumptions relates to reproductive skew models, which make predictions about the division of reproduction within a group (reviewed in Johnstone 2000). These models assume that all group members are capable of independent reproduction, but triploid females clearly do not have the same options for solitary nesting as their diploid cofoundresses and, therefore, cannot be equal partners in a ‘social contract’ (Reeve & Nonacs 1992). Our finding of triploid foundresses among 11.1% of cooperative nesting associations suggests that this possibility cannot be ignored.

Genetic data are particularly important for studies of introduced or fragmented populations, which are likely to have fewer alleles at the sex-determining locus and thus relatively high rates of diploid male production. Indeed, this phenomenon has been suggested as an indicator of pollinator decline (Zayed et al. 2003). High diploid male frequency may, in turn, result in strong selective pressure for increased fertility of diploid males as recently described for a solitary vespid wasp (Cowan & Stahlhut 2004). Introduced species provide a tremendous opportunity to study such evolutionary processes in action (Lee 2002) and this can best be accomplished by integrating population genetics with behavioural and life-history data.

Although high rates of diploid male production are expected in bottlenecked populations, our evidence suggests that this phenomenon also occurs when population genetic diversity is not severely reduced. Our data were collected near the site of the oldest known P. dominulus introduction in Cambridge, MA in the late 1970s (Eickwort 1978; Hathaway 1981). Recent evidence suggests that this population has a high level of genetic diversity and has not experienced a genetic bottleneck (Johnson & Starks 2004). It is possible that allelic diversity has been reduced at the sex locus but not at the tested microsatellite loci, but the successful establishment of this species across the USA suggests that a high level of inbreeding is unlikely. In combination with the previous report of 4.7% triploidy among foundresses in a native population of Polistes fuscatus (Liebert et al. 2004), our results demonstrate that this phenomenon is not limited to introduced or bottlenecked populations. Integration of genetic data with behavioural research on Polistes wasps will help us to examine the impact of diploid males and triploid females in this model genus.

A. Berry, C. Blackie, D. Steele and G. Switz helped collect nests. G.S. also helped with genotyping of foundresses. We thank the REU programme of the NSF for support of A.B. and Tufts University for a FRAC grant to PTS.