INTRODUCTION

Separation into castes is one of the cornerstones of the evolution of social insects. In some species, reproductive “queens” and nonreproductive “workers” differ only in behavior, but in other species such behaviors are accompanied by physiological and morphological specializations and body-size variation (Oster and Wilson, 1978). The degree to which castes are distinct often serves as part of the definition of the society itself, with greater caste differentiation indicating “higher” sociality (Bourke, 1999). In some groups, such as swarm-founding neotropical wasps, the species are described as highly social, although the distinction between queens and workers is considered slight and often confusing (Richards and Richards, 1951; Richards, 1978, cited as Polybiini, junior synonym of Epiponini, see Carpenter, 1993, 1997).

Wasps (Hymenoptera: Vespidae) have frequently been used to test evolutionary models for the origin of social behavior because of their different levels of sociality—from solitary to eusocial (West-Eberhard, 1978, 1996; Wilson, 1985; Itô, 1986; Spradbery, 1991). General theory holds that primitive systems of dominance rely on physical aggression, with the winner becoming the egglayer. More derived systems require some kind of caste determination that may occur during the immature stages of an individual's development (preimaginal determination), with an unambiguous physiological or morphological result that cannot be reversed. Absence of morphological castes is a plesiomorphic (“primitive”) condition in the three subfamilies of social wasps, so that morphological castes were attained independently in different lineages (Carpenter, 1991). Most distant to the question at hand, females of Stenogastrinae are barely distinguished as queens or as workers (Turillazzi, 1991). Closer, and sister to the subfamily examined here, queens of Vespinae are larger than workers (Spradbery, 1991), but such a trait may be an adaptation for a queen's solitary winter survival in most species (West-Eberhard, 1978). In the Polistinae (studied here), the most basal genera (Polistes and Mischocyttarus), initiate their nests solitarily, with egglayers only slightly larger than non-egglayers. However, polistines present a large diversity that can be arranged along a spectrum from taxa in which queens and workers are externally similar, to others with fairly distinct caste attributes (reviewed in O'Donnell, 1998; Shima et al., 1998). The Paleotropical tribe Ropalidini (perhaps the sister to the Neotropical tribe Epiponini examined here) shows diversity that is hard to interpret without more study. For instance, some species of Ropalidia present distinct morphological castes (Yamane et al., 1983), including those of small colony size (Wenzel, 1992), but others show morphologically similar castes despite larger colony size (Gadagkar and Joshi, 1982, 1983; Yamane and Itô, 1987; Gadagkar, 1987). Parapolybia has castes (Yamane and Maeta, 1985) but their behavior is poorly known. Belonogaster is also poorly known (Pardi and Piccioli, 1981) but is thought to show great morphological variation (Keeping, 2000, 2002). In Polybioides, morphological castes are known in P. tabidus (Turillazzi et al., 1994), which builds large colonies and reproduces by swarms. However, despite the repeated evolution of castes in social wasps as a whole, the evolution of castes is largely interpreted through our understanding of Epiponini.

Epiponini is a tribe of Polistinae, with about 20 well-marked genera and at least 229 species, plus another 27 subspecies, all of which are Neotropical. All Epiponini are polygynic (meaning that many queens commonly reproduce simultaneously on the same nest), and all reproduce by swarms, which is taken as a sign of sophisticated, obligatory, high sociality. Initially, Richards (1978) documented three forms of caste discrimination, and recent research builds upon these: (1) Queens larger than workers. In some species caste differences are so pronounced that they must be determined by nutritional differences during larval development (Evans and West-Eberhard, 1970; Jeanne and Fagen, 1974; Sakagami et al., 1996; Noll et al., 1997), with important consequences for fertility and reproductive asymmetries (West-Eberhard, 1978; O'Donnell, 1998). The classical landmark is a case of different, nonallometric castes of Agelaia areata, where queens and workers have different shapes as well as different sizes (Jeanne and Fagen, 1974). It seems that these differences must represent preimaginal bias in nutrition. (2) Allometric differences, with queens smaller than workers in some body parts, but larger in others, which may result from ontogenetic reprogramming in growth parameters (Jeanne et al., 1995), according to the idea that the body plans represent fundamentally different end products, not different points along a continuum. Sometimes such castes are barely detectable (Jeanne, 1996). (3) Slight or indistinct morphological differences, with castes absent, or manifested weakly, based on only a few traits (Richards, 1978; Jeanne, 1980). One great problem is that integration of the empirical studies is not straightforward because few studies are exactly comparable in measurement methodology or phase of colony cycle studied, so generalization is difficult. A rough summary of species known to fall into these three conditions can be found in table 1. It has been suggested that preimaginal determination may be basal for Epiponini (Jeanne et al., 1995; Jeanne, 1996; Hunt et al., 1996; O'Donnell, 1998), especially because such a trait is present in Apoica (Shima et al., 1994), the most basal genus in Epiponini. However, Apoica is peculiar and highly apomorphic in many ways (general morphology, nest architecture, and nocturnal habit, see Richards, 1978; Wenzel, 1992; and Pickett, 2003) and so it may not be a good model of basal synapomorphies of Epiponini. It has not been demonstrated that preimaginal determination is the basal condition for the whole clade.

Integrating studies of caste in Epiponini is difficult because studies to date have been limited in scope, with most publications covering one species at a time. Measurements are not necessarily comparable across studies. There are sometimes doubts about the phase of colony cycle that was measured, particularly when the colony was collected opportunistically by visitors on brief fieldtrips. In addition, when reproductive status does not relate closely to morphological variation, there is a tendency to rescue a concept of caste by categorizing problematic females as “laying workers” or “replacement queens” on an ad hoc basis. Such axiomatic descriptions may serve only to obscure the original question, protecting an inadequate concept of caste itself from refutation by empirical data. We address these issues by presenting here the most complete set of comparable values regarding caste determination. A total of more than 13,500 measures encompassing nine dimensions of 13 species are offered for the first time. These are integrated with comparable data from other studies to include 68 species. We establish the basal condition for Epiponini, providing evidence for both preimaginal and postimaginal caste determination. In addition to morphometric variation in caste, we stress the existence of various types of social regulation.

MATERIALS AND METHODS

Workers and queens were taken from mature colonies (i.e., post worker emergence) of 13 different species of Epiponini collected in three different localities. Angiopolybia pallens (Lepeletier), Asteloeca ujhelyii (Ducke), Charterginus fulvus Fox, Clypearia sulcata (de Saussure), and Leipomeles dorsata (Fabricius) were collected in Iquitos, Peru. Nectarinella championi (Dover) was collected in Costa Rica, near Atenas. For these two localities, the samples were collected by John W. Wenzel and James Carpenter, killed in cyanide and preserved promptly in 70% ethanol, and deposited in the AMNH. Polybia (Polybia) liliacea (Fabricius), Polybia (Formicicola) rejecta (Fabricius), Chartergus metanotalis Richards, Epipona tatua (Cuvier), Metapolybia aztecoides Richards, Polybia (Pedothoeca) spinifex Richards, and Synoeca surinama (Linnaeus) were collected in Nova Xavantina, Mato Grosso, Brazil by Fernando B. Noll and Sidnei Mateus. Vouchers are deposited at the Museu de Zoologia (MZUSP), São Paulo, Brazil, and the AMNH. They were taken by being put into a plastic bag provided with ether-moistened cotton balls or with open cyanide tubes. Populations were fixed in Dietrich's solution and then kept in 70% ethanol for dissection.

From each of these colonies 100 workers and all queens were measured and dissected. When castes could be discriminated by size or color, all possible queens were selected, and from the remaining individuals 100 workers were arbitrarily selected. In small colonies where castes were not obvious, all individuals were measured and dissected. In large colonies where castes were hard to identify, 100 individuals were chosen randomly. After identifying some morphological characters to discriminate queens, such as abdominal size, all individuals having such a pattern were chosen and used in the analysis. The following characters were measured under a binocular microscope (smallest unit = 0.01 mm): head width (HW), minimum interorbital distance (IDm), gena width (GW), width of mesoscutum (MSW), alitrunk length (AL), length of gastral tergite I (T₁L), basal height of T₁ (T₁BH), basal widths of tergite II (T₂BW), and partial length of the forewing (WL) (fig. 1, for more details about the characters see Shima et al., 1994). Ovary condition was determined under a stereomicroscope. In order to analyze insemination, the spermatheca was removed and put on a slide in a 1:1 solution of glycerin and ethanol (70%). The presence of sperm cells was confirmed by microscope.

Before statistical analysis, data were converted to millimeters and later converted by log transformation in order to avoid problems of variance. Means and standard deviations were calculated from the nine morphological measurements. Statistical characterization of caste was explored in two ways. One-way ANOVA was used for comparisons of means to demonstrate whether castes (as determined by dissection) differ in some body dimensions. Because multiple measures that relate to shape are not always captured well in single comparisons, stepwise discriminant function analysis was used to see if combinations of variables could be useful in predicting caste. In this method, variables are successively added to the model based on the higher F to enter values, adding no more variables when the F-ratio is no longer significant. However, variables that enter the model early may be completely subsumed by separate variables that enter later. The significance of Wilks' lambda (an estimate of common variance between sets of variates) was used to determine the degree to which separate measures contributed to the final model. This is an alternative to using an F to remove at each step. Variables that appear in the final model but do not have significant F-ratios represent variance components that are explained by some combination of the other variables also in the model, and therefore no longer contribute to the discrimination itself. Wilks' lambda varies from zero to one: the lower the value, the greater the significance.

Analysis of covariance (ANCOVA) was used, with alitrunk length (AL) as a covariate, in order to identify different growth rates in the body parts measured. AL was chosen based on the effectiveness in previous publications (Jeanne et al., 1995; Jeanne, 1996; Hunt et al., 1996).

RESULTS

DISCUSSION

A distinct physiological separation of females into those that are reproductive and those that are nonreproductive is possible based on ovarian condition and insemination in some of the species reported here. While complete absence of ovarian development in workers (i.e., the absence of both types B and C) was not found in the species studied here, such a syndrome has been reported in Apoica flavissima, Agelaia vicina (Sakagami et al., 1996), Ag. pallipes, and Ag. multipicta (Noll et al., 1997). In our present study, queens' ovaries (type D) were much longer than workers' ovaries in Chartergus metanotalis, Epipona tatua, Polybia liliacea, Po. rejecta, and Po. spinifex. This situation is also known in Apoica flavissima (Shima et al., 1994), Protonectarina sylveirae (Shima et al., 1996a), Polybia dimidiata (Shima et al., 1996b), Po. scutellaris, Po. paulista and Po. occidentalis (Noll and Zucchi, 2000), Agelaia vicina (Sakagami et al., 1996), and Ag. pallipes and Ag. multipicta (Noll et al., 1997). Similarly, though less discrete, intermediate type C was not found in Asteloeca ujhelyii, Chartergus metanotalis, Clypearia sulcata, Epipona tatua, Metapolybia aztecoides, Polybia liliacea, Po. rejecta, Po. spinifex, and Synoeca surinama (table 2). Other species previously reported to have similar distinction are Protonectarina sylveirae (Shima et al., 1996a), Polybia dimidiata (Shima et al., 1996b), Po. scutellaris and Po. paulista (Noll and Zucchi, 2000), and Chartergus globiventris (Noll and Zucchi, 2002). Some species are found to have individuals representing all four stages, implying a near continuum (but that insemination is always associated with type D ovaries). These species include Angiopolybia pallens, Charterginus fulvus, Leipomeles dorsata, and Nectarinella championi (table 2). The same pattern, implying a lack of physiological caste in any meaningful way, was previously found in other species using the same procedure, such as Chartergellus communis (Mateus et al., 1999), Pseudopolybia vespiceps (Shima et al., 1998), Parachartergus smithii (Mateus et al., 1997), and Brachygastra lecheguana (Shima et al., 2000).

Ovarian development presents a confusing array of conditions, yet previous researchers have tried to find ways to capture the totality of this variation with some kind of overarching hypothesis of caste that would be applied to all species of Epiponini equally. For example, the concept of cyclical oligogyny invokes demography to provide a process by which individuals may become queens at swarm initiation of colonies, but not later in the colony cycle (West Eberhard, 1978). Colony cycle is known to relate to the degree of morphological differentiation in species that do have marked differences, with differences growing as colonies age (e.g., Noll and Zucchi, 2000, 2002), and comparing mature colonies is now required if one is to argue that the animals that are sampled actually represent the species' breadth of morphological diversity. Nonetheless, the search for a single explanation of caste (its presence or its ambiguity) in Epiponini is still lacking. We offer here a novel view that relies on returning to the fundamental question of whether caste exists at all. We then integrate the answers to those questions to provide the most coherent view of caste to date, one that does not seek to explain all Epiponini through a single fundamental process.

CONCLUSIONS

This analysis indicates that caste has evolved in different ways in different lineages of Epiponini, starting from casteless societies. Our decision to call the basal condition “casteless” relies on two independent lines of argument. The first argument is that females do not differ meaningfully either in morphology or physiology. The second is that the reproductives do not represent dominant animals, as they do in more primitive societies such as those of Polistes and Mischocyttarus. As discussed above, workers are known to police the egglayers in Apoica, Synoeca, and Metapolybia, whereas the egglayer herself polices the brood comb in Polistes. Thus, the egglayers of the genera represented by the groups of Angiopolybia through Nectarinella, and Synoeca through Asteloeca do not represent the same social role as egglayers in Polistes; instead, they are simply individuals that lay eggs in a society of females that are not otherwise distinguished. This may not have been recognized before because, despite universal acceptance of Hamiltonian perspectives regarding inclusive fitness, the community of vespid researchers still imagines wasps engaged in a struggle for direct fitness rather than joined in a society of inclusive fitness. It seems likely that different types of socio-regulation exist in epiponines as suggested by Noll and Zucchi (2002). It is possible that some confusion results in the literature from a desire to find a single syndrome to be representative of the entire diversity of swarming wasps, whereas the actual evolution of caste has perhaps been multifaceted.