Recently, it has been proposed that the one of the main determinants of complex societies in Hymenoptera is colony size, since the existence of large colonies reduces the direct reproductive success of an average individual, given a decreased chance of being part of the reproductive caste. In this study, we evaluate colony size evolution in corbiculate bees and their relationship with the sociality level shown by these bees. Specifically i) the correlation between colony size and level of sociality considering the phylogenetic relationship to evaluate a general evolutionary tendency, and ii) the hypothetical ancestral forms of several clades within a phylogeny of corbiculate bees, to address idiosyncratic process occurring at important nodes. We found that the level of social complexity in corbiculate bees is phylogenetically correlated with colony size. Additionally, another process is invoked to propose why colony size evolved concurrently with the level of social complexity. The study of this trait improves the understanding of the evolutionary transition from simple to complex societies, and highlights the importance of explicit probabilistic models to test the evolution of other important characters involved in the origin of eusociality.
Citation: Rodriguez-Serrano E, Inostroza-Michael O, Avaria-Llautureo J, Hernandez CE (2012) Colony Size Evolution and the Origin of Eusociality in Corbiculate Bees (Hymenoptera: Apinae). PLoS ONE 7(7): e40838. doi:10.1371/journal.pone.0040838
Editor: William Hughes, University of Leeds, United Kingdom
Received: December 21, 2011; Accepted: June 17, 2012; Published: July 13, 2012
Copyright: © 2012 Rodriguez-Serrano et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Funding: This study was supported in part by FONDECYT 11080110 from the Comisión Nacional de Investigación Científica y Tecnológica, Gobierno de Chile http://www.conicyt.cl. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. No additional external funding received for this study.
Competing interests: The authors have declared that no competing interests exist.
Bourke  proposed that colony size, along with kin structure, is one of the main determinants of complexity in insect societies. This hypothesis is based on the reproductive conflict between members of a colony: any change in colony size entails a direct consequence for the reproductive potential of a worker. When colony size increases, workers experience a loss of reproductive power, given a decreased chance to be part of the reproductive caste, or to replace the queen. The main mechanisms involved in lowering the reproductive potential of a worker, are the diluted by number process and worker policing . As colony size increases, the proportion of any worker’s eggs among worker-laid eggs falls because it becomes increasingly difficult for the worker to prevail over all others and the mutual prevention by workers of another’s reproduction increases . Bourke  has shown that these mechanisms hold for different queens’ mating frequencies and all ploidy systems, with a few exceptions requiring further explanations (see –). As a consequence, the behavioral/morphological divergence between workers and reproductives is the result of strong selection for specialized workers with low reproductive potential .
A simple contrast between the level of sociality and colony size of different groups of social insects provides a positive relationship between these variables (e.g. –). However, standard correlation statistical tests are not suitable for use with non-independent data, such as species with shared phylogenetic history , . Furthermore, application of “phylogenetic corrected” statistical tests (a correlation between variables to which have been removed the phylogenetic component of variance, such as Phylogenetic Independent Contrasts; ), only gives information about the actual relationship between these variables . However, other tools of the comparative phylogenetics method , allow us to explore at least two aspects of Bourke’s hypothesis in a explicit phylogenetic context: i) if the correlation between colony size and level of sociality is implemented considering the phylogenetic relationship, the associative pattern could shown a general evolutionary tendency, and ii) the hypothetical ancestral forms of several clades within a phylogeny of social insects could be informative about idiosyncratic processes occurring at important nodes, and possibly why colony size evolved concurrently with the level of social complexity (e.g. , ). The latter approach has been used to evaluate the origin of monogamy and the correlation of this trait with both, cooperative breeding and eusociality (in birds, ; and Hymenoptera,  respectively).
The subfamily Apinae (also “corbiculate bees” because they possess a “corbicula”, structure in the metatibia that enables the transport of pollen) is a monophyletic group, presenting sociality levels that include solitary species, to communal (Euglossini), to eusocial species with diverse caste determination strategies. The tribe Bombini shows a behavioral worker caste and subtle morphological divergence between reproductives and workers. The Apini and Meliponini tribes have morphologically distinct reproductives and workers . Also, corbiculate bees show great variation in colony sizes, ranging from solitary or few individuals in Euglossini, to dozens of thousands in Apini . These bees therefore represent an excellent model for studying the correlation between colony size and complex societies through the use of explicit evolutionary models . Recently, Cardinal et al. , propose a novel, detailed, comprehensive molecular phylogenetic analysis of all 33 apid tribes based on seven nuclear molecular markers. In this work, the obtained phylogenetic tree shows that Meliponini and Apini are not sister clades. Thus, “multiple origins” of complex sociality is the most probable evolutionary history of this trait based on the current phylogenetic hypothesis. Under this phylogenetic context, this group of Hymenoptera allows the direct testing of Bourke’s hypothesis given that there is no single character transition from solitary to eusocial species and thus any evolutionary transition in social system should be a consequence of an increase in colony size. In this study we assess two critical points of this hypothesis: a) the phylogenetic correlation between colony size and social structure, and b) the evolution of colony size and the social structure of the corbiculate bees through the reconstruction of ancestral character state.
Phylogenetic Relationships of Corbiculate Bees
The phylogenetic reconstruction that best fit to the rates and patterns of molecular evolution of the data was that obtained with 2GTR+Γ matrices. The consensus tree topology showed corbiculate bees to be a completely sustained monophyletic group (Fig. 1, Node G), where each tribe corresponded to a highly supported natural group. The relationship between tribes was observed to have a higher posterior probability (Fig. 1, Node E and F). Our phylogenetic hypothesis has no differenced from that reported by Cardinal et al. , so is a good historical support for further comparative analysis.
The tree was obtained by means of a phylogenetic mixture model based on Bayesian approach. The numbers next to nodes indicate the posterior probability of occurrence of the clade. The letters above the nodes correspond to the hypothetical ancestors, whose most probable character state is shown in Figures 2 and 3.
Colony Size and Social Structure Evolution
The predominant evolutionary transition rates among the different colony size states were, q12, q21 (Table 1); that is, the transition from few to hundreds of individuals in both directions. In addition, the predominant transition rates in the evolution of sociality were q21, q12, (Table 1), that is, the transition from solitary to communal in both directions. The ancestral character state estimation for the phylogeny nodes indicated that the most probable common ancestor of the corbiculate bee tribes was a colony of few individuals (Fig. 2 Node G; probability (p) = 0.51; 1.7 individuals) with a communal sociality level (Fig. 3 Node G; p = 0.44). The ancestor of the Meliponini and Bombini has a higher probability of having tens/hundreds of individuals (Fig. 2 Node F; p = 0.51; 18.9 individuals) and eusocial with behavioral castes (Fig. 3 Node F; p = 0.68). On the other hand, the ancestor of Euglossini and Apini has more probability of having colonies with few individuals (Fig. 2, Node E; p = 0.50; 1.9 individuals) with a communal social structure (Fig. 3 Node E, p = 0.50).The development of morphological castes occurred independently in two lineages: In the Meliponini ancestor (Fig. 3 Node D, p = 0.99) and in the Apini ancestor (Fig. 3 Node A, p = 0.98). The ancestors of both tribes had large colonies (Fig. 2 Node A p = 0.54; 11592 individuals; and Node D p = 0.83; 2077 individuals).
The reconstruction was based on both topology and branch lengths of the Bayesian phylogenetic trees. In parenthesis is shown the mean value and standard error for the continuous character reconstruction.
The reconstruction was based on both the topology and branch lengths of the Bayesian phylogenetic trees.
Finally, the results of the phylogenetic logistic regression show that the effect of colony size on social structure is significantly different from zero if the evolutionary relationships of corbiculate bees are part of the model (Table 2). Larger values of colony size are strongly related to more complex social structure (i.e. eusocial with morphological castes; Table 2, b1 = 7.7475; p<0.001). Concordantly, smaller colony sizes are strongly related with the absence of complex societies (i.e. eusocial with behavioral/morphological castes; Table 2, b1 = −2.2793; p<0.001).
Colony Size Evolution and the Origin of Eusociality
Our results, based on analyses that evaluated the phylogenetic uncertainty, and the reconstruction of ancestral states, show that the colony size of corbiculate bees coevolves with the level of sociality (Fig. 2, 3; Table 2). Colony size is a complex trait, determined by various factors that favor or restrict the number of individuals . Those which restrict colony size are associated with traits related to life history , periodicity of trophic resources and space for nesting , among others. These multiple factors have differential importance throughout the tribes of Apinae. For example, with respect to the physical space for nesting : the species in the Meliponini tribe prefer to use hollow trees for nesting, and there are also hypogeous species . There is a strong spatially limiting factor involved in the development of highly large colonies, such as those of the Apini. Other factors could be influencing colony size in the Bombini tribe, especially factors correlated with latitude, due to the annual cycle of the colonies of this tribe , . These ecological factors that determine the colony size could be responsible for the qualitative change from the most recent common ancestor (MRCA) of Apini + Euglossini (Fig. 2, node E: categorical 0 to 10; continuous 1.9 individuals) to the MRCA of Apini (Fig. 2, node A: categorical >10000; continuous 11592 individuals).
The lack of intermediate form between these nodes does not have a straightforward interpretation. Some insight can be obtained from allodapine bees (Apidae: Allodapini), in which the colony size and the level of sociality is a function of environmental and ecological variables in the facultative eusocial Exoneurella tridentata (i.e. a xeric habitat and durable nesting substrate; ). This bee is an exception within the tribe, given that it shows both a discrete morphological gap between queen and worker castes, and evolution of colony size (∼40 females) with accelerated tempo , . The colony size and eusociality evolution of E. tridentata thus represent a discontinuity in the evolutionary change of the Allopadini tribe, which could be associated with overcoming a strong selective barrier (sensu ). A barrier that could be important at the MRCA of Apini + Euglossini (Fig. 2; node E), can be inferred from a general exploration of the main ecological traits of Euglossini vs Apini. Euglossini is a tribe of highly specialized bees on a few plant families that inhabits almost exclusively the Neotropics, whereas Apini is a tribe with a wide range of both food resources and distribution , . For example, strong resource dependence has been detected in Euglossa nigropilosa, which preclude many communal nesting females . This could imply that few individuals in a communal society (Fig. 2; node E) can either take advantage of generalist traits and increase colony size, or hold the character state as consequence of specialization (Fig. 2; nodes A, B). In the same way, the evolution from the communal MRCA of Apini + Euglossini (Fig. 3, node E) to the eusocial with morphological castes MRCA of Apini (Fig. 3, node A) can be addressed by scarce evidence from the halictine sweat bee Halictus sexcinctus (Hymenoptera: Halictidae ). This species shows a communal/eusocial polymorphism in the same population, where the eusocial colonies are composed of distinct morphological castes, which also differ from the communal foundresses . Since the social polymorphism of H. sexcinctus is an extraordinary case within Halictidae, Richards et al. , argue that it may represent an unstable intermediate step in the evolution of social behavior. The consequences of this labile character state are either to promote some individuals to dominate reproduction or the founding of solitary colonies . However, these species with facultative eusociality (E. tridentata and H. sexcinctus) show small incremental changes in colony size between the communal and eusocial states. The differential of colony size and level of sociality character states found on the transition from Node E to Node A (Fig. 2 and 3) seem to be quite different. Then, explanations are needed beyond actual ecological forces for this.
Boomsma ,  proposed that lifetime monogamous species are more likely to evolve obligate eusociality (i.e. eusocial with morphological castes). This hypothesis argues that in lifetime monogamous species, relatedness is maximized throughout the lives of helper cohorts, promoting the evolution of sterile caste . The previous examples of allodapine and halictid bees only show an advanced form of cooperative breeding . Although this proposal has been supported in a phylogenetic context, including the estimated character state at MRCA of corbiculate bees , further questions arise. For instance, once lifetime monogamy is established, how do sterile castes then evolve? In this context, the explanation for the transition between the MRCA of Apini + Euglossini to the MRCA of Apini, could be addressed from the reproductive abilities of a worker “by its own choice” . The direct reproductive benefits tend to decrease, as well as the indirect benefits increase, while the colony grows in size. This is because the worker loses opportunities to replace the queen in a colony where more and more workers are trying to take over the colony , and, concurrently, this crowded colony represents a full-sibling system. Then, the benefits may outweigh the costs simply by increasing the number of individuals in a colony. Bourke  used an extended argument that incorporates the mutual inhibition between workers of their reproduction potential (i.e. worker policing, ). This behavior leads to the selection of highly specialized workers accompanied by morphological skew between castes, which at the same time incurs positive feedback, permitting the existence of larger colonies. If this mechanism operates, then it is not necessary that the ancestor of eusocial forms with morphological castes be eusocial with behavioral castes, only that it have a monogamous reproductive system. Based on our results, we suggest that it is possible for eusociality with morphological castes to evolve either from a gradual increase in colony size and complexity (Fig. 2 from node F to node D), or from a threshold of colony size (Fig. 2 from node E to node A). Once it has reached this character state, the species can’t revert, consistent with previous studies  (Fig. 2 Node B, C).
Finally, our study, based on a Bayesian probabilistic framework, provides strong support for Bourke’s proposal ; since we have observed that the one of the main determinants in the evolution of morphological castes in complex societies of corbiculate bees is colony size. The study of this trait improves our knowledge regarding the evolutionary transition from simple to complex societies, and highlights the importance of explicit probabilistic models to test the evolution of eusociality and other important characters in social species.
Material and Methods
We compiled a datebase of the colonial size and social structure of corbiculate bees from an exhaustive search in the literature (Table S1). Colony size data were treated both as categories following the ranges described by Bourke , and continuous values (i.e. as mean value reported in the literature (Table S1). For the discrete coding, four colony size categories were used: 1 = 0 to10 individuals; 2 = 10 to 100 individuals; 3 = between 1,000–10,000 individuals; and 4 = 10,000 or more Meliponini and Apini individuals. The continuous and discrete coding uses of colony size are proposed as complementary approaches given the missing categories (100–1000) in the data. For the case of the social system variable, we assign categorical variables to level of sociality as proposed by Kukuk , where 1 = solitary; 2 = communal; 3 = eusocial with behavioral castes; 4 = eusocial with morphological castes.
BMCMC Molecular Phylogeny
Although the phylogeny of corbiculate bees is well resolved (see ), we reconstruct the phylogenetic relationships of the species belonging to this groups to obtain a sample of phylogenetic trees. With this sample of topologies and branch lengths, we can incorporate the phylogenetic uncertainty in the estimation of ancestral character states. We used DNA aligned sequence data from Cardinal et al. , specifically the Long Wavelength Rhodopsin (LWRh) and Arginine kinase (Argk) nuclear genes from the 33 species of corbiculate bees and three species of the sister group (Centris cockerelli) and from more distantly related families as outgroup (Epiclopus gayi, and Xylocopa amamensis; sensu Cardinal et al. ; Table S2). The alignment was performed using Clustal W . These sequences were concatenated to reconstruct a phylogenetic hypothesis using the Markov Chain Monte Carlo (MCMC) method within Bayesian framework (BMCMC) to estimate the posterior probability of phylogenetic trees. Due to the markers selected comimg from different genomic locations and, as has been shown, their patterns and rates of nucleotide substitution being heterogeneous , we applied a general likelihood-based mixture model (MM) of gen-sequence evolution as describes by Pagel and Meade ,  based on General Time Reversible model (GTR, ). The MM model accommodates cases in which different sites in the alignment evolved in qualitatively distinct ways, but does not require prior knowledge of these patterns or partitioning of the data. The Reversible-Jump Markov Chain Monte Carlo (RJMCMC) procedure ,  was used for integrating the results of all patterns, and produces an MM that summarizes the sequence evolution. This approach enables researchers to explore the variety of possible models and parameters, converging towards the model that best fits the data in the posterior tree sample. These analyses were made in BayesPhylogenies 1.1 software (http://www.evolution.rdg.ac.uk/BayesPhy.html). We ran four independent BMCMC analysis using 35,707,000 iterations of the phylogenetic trees, sampled every 1000th to assure that successive samples were independent. We used the chains which had the highest marginal likelihood. From this sample of trees we removed the first 500 trees of the sample to avoid including samples before the convergence of the Markov Chain. Then we re-sampled every 46 trees to obtain a final sample of 770 trees with an autocorrelation of −0.008678 in ln-likelihood. This final sample was used for the comparative analysis.
Estimation of Ancestral Character States, Evolutionary Transition Rates and Phylogenetic Logistic Correlation
We used the ancestral state reconstruction package available in BayesTraits 1.0 software , to evaluate the evolution of social traits and the colony size of corbiculate bees. For this we used both the topology as well as the branch lengths obtained from the sample of phylogenies in the BMCMC molecular phylogenetic analysis (in this case we used 770 trees). Using these data we were able to relate the evolution of eusociality, with estimated colony size ancestral states. For reconstruction of the ancestral characters states we used the continuous-time Markov k-state model ,  estimating states under Bayesian MCMC framework. We run MCMC analysis with a uniform prior distribution on the rate coefficients with a range from 0 to 100. Several values of the Rate Deviation parameter were alternated to avoiding spurious acceptance between successive MC steps (i.e. appropriated acceptance values are in the range 0.28 to 0.32; Dr. Andrew Meade, personal communication). We ran 50000000 MCMC iterations, discarding the first 50000 iterations to avoid results outside the likelihood convergence. This analysis was run in the Multistate module of BayesTraits software . We determined all instantaneous transition rates defined as qxy, where x is the initial character state at the beginning of a branch, and y the final state at the tip of the same branch . These rates were estimated among the discrete states of the colony size character, for both forward and backward rates, but without establishing restrictions in the trait transitions, that is, we allowed this character to vary from a few to a large number of individuals and from a large number to a few, going through all the possible intermediate values in the different interior nodes of the phylogenies (Table 1). Concordantly, we estimated the ancestral character states of the continuous values of colony size (i.e. mean of colony size reported for each species incorporated in this study; Table S1). This analysis was run in the Continuous module of the BayesTraits software . For this, we ran 50000000 MCMC iterations and discarding the first 50000 iterations. In this approach the Data Deviation parameter was treated in the same way that Rate Deviation parameter to achieve a 28–32% acceptance. In the same way we determined some transition rates for the level of society character evolution, with important restrictions, that is, we allowed this character to vary between immediate states (e.g. solitary to communal an vice-versa: q12 and q21). The transition rates from eusocial with morphological castes to any other state (i.e. q43, q42, q41) were set to 0, given that the eusociality with morphological castes is an evolutionary endpoint without reversion . Also, and given the extremely unlikely scenario with evolution of eusociality with morphological castes from a solitary ancestor, we restrict the rate q14 to 0.
Finally, we perform phylogenetic logistic regressions with Firth correction to evaluate the relationships between colony size and social structure through the evolutionary history of this group. Briefly, this method allows the analysis of cases in which the dependent variable is binary (0 or 1) and the values are non-independent among species, with phylogenetically related species tending to have the same value of the dependent variable. The method is based on an evolutionary model of binary traits in which trait values switch between 0 and 1 as species evolve up a phylogenetic tree . So, to evaluate if higher colony size was correlated with higher level of sociality (i.e. morphological castes), we categorized the presence of morphological castes as 1 and its absence as 0 in the species under study. In addition, in order to evaluate the Bourke’s  prediction of that small colonies should never be complex (e.g. show morphological/behavioral castes) we categorized the presence of the lower social level (i.e solitary and communal social level) as 1 and its absence as 0. These regressions were performing using a variance-covariance matrix of the species trait constructed using the consensus of tree samples (previous section) in the module PDAP  of Mesquite software version 2.6 . The phylogenetic logistic regression was run using the PloGReg.m function written by Ives and Garland . This function simultaneously tests for phylogenetic signal while estimating the parameters of the regressions. The independent variable was standardized to have mean equal to zero and standard deviation equal to one; this makes the regression coefficients represent effect sizes of the independent variables whose magnitudes reflect the size of effect of the variable . A bootstrapping procedure involving 2000 simulations (as proposed by Ives and Garland ) was used to generate the confidence intervals and test for statistical significance of the slope of the regression model and phylogenetic signal of the discrete data. Convergence of model parameters was achieved in all cases after these simulations.
Colony size and level of social complexity for the species used in phylogenetic comparative analyses. Also are provided the reference for each data.
GenBank Access Number for the nuclear genes of each species used in this study.
The authors are very grateful to Paula E. Neill, Matthew R. Lee, William O. H. Hughes and two anonymous reviewers (especially to reviewer 1) for comments and suggestions that greatly improved the manuscript.
Conceived and designed the experiments: ERS OIM. Performed the experiments: OIM. Analyzed the data: ERS OIM. Contributed reagents/materials/analysis tools: ERS OIM JAL CEH. Wrote the paper: ERS OIM.
- 1. Bourke AFG (1999) Colony size, social complexity and reproductive conflict in social insects. J Evol Biol 12: 245–257.
- 2. Ratnieks FLW (1988) Reproductive harmony via mutual policing by workers in eusocial Hymenoptera. Am Nat 132: 217–236.
- 3. Ratnieks FLW (1990) The evolution of polyandry by queens in social Hymenoptera: the significance of the timing of removal of diploid males. Behav Ecol Sociobiol 26: 343–348.
- 4. Pamilo P (1991) Evolution of colony characteristics in social insect. II. Number of reproductive individuals. Am Nat 138: 412–433.
- 5. Alexander RD, Nooman KM, Crespi BJ (1991) The evolution of eusociality. In: Sherman PW, Jarvis JUM, Alexander RD, editors. pp. 3–44. New Jersey.
- 6. Breed MD (1976) The evolution of social behavior in primitively social bees: A multivariate analysis. Evolution 30: 234–240.
- 7. Jeanne RL (1980) Evolution of social behavior in the Vespidae. Annu Rev Entomol 25: 371–96.
- 8. Villet M (1990) Qualitative relations of egg size, egg production and colony size in some ponerine ants (Hymenoptera: Formicidae). J Nat Hist 24: 1321–1331.
- 9. Karsai I, Wenzel JW (1998) Productivity, individual-level and colony-level flexibility, and organization of work as consequences of colony size. Proc Natl Acad Sci USA 95: 8665–8669.
- 10. Holbrook CT, Barden PM, Fewell JH (2011) Division of labor increases with colony size in the harvester ant Pogonomyrmex californicus. Behav Ecol 22: 960–966.
- 11. Felsenstein J (1985) Phylogenies and the comparative method. Am Nat 125: 1–15.
- 12. Harvey PH, Pagel MD (1991) The Comparative Method in Evolutionary Biology. Oxford Series in Ecology and Evolution, Oxford University Press, Oxford.
- 13. Thompon GJ, Kitade O, Lo N, Crozier RH (2000) Phylogenetic evidence for single, ancestral origin of a ‘true’ worker caste in termites. J Evol Biol 13: 869–881.
- 14. Pagel M, Meade A (2006) Bayesian analysis of correlated evolution of discrete characters by reversible-jump Markov chain Monte Carlo. Am Nat 167: 808–825.
- 15. Cornwallis CK, West SA, Davis KE, Griffin AS (2010) Promiscuity and the evolutionary transition to complex societies. Nature 466: 969–991.
- 16. Hughes WOH, Oldroyd BP, Beekman M, Ratnieks FLW (2008) Ancestral monogamy shows kin selection is key to the evolution of eusociality. Science 320: 1213–1216.
- 17. Cameron SA, Mardulyn P (2001) Multiple molecular data sets suggest independent origins of highly eusocial behavior in bees (Hymenoptera; Apinae). Syst Biol 50: 194–214.
- 18. Michener CD (2000) The Bees of the World. Johns Hopkins University Press, Baltimore, MD.
- 19. Pagel M, Meade A, Barker D (2004) Bayesian estimation of ancestral character states on phylogenies. Syst Biol 53: 673–684.
- 20. Cardinal S, Straka J, Danforth BN (2010) Comprehensive phylogeny of apid bees reveals the evolutionary origins and antiquity of cleptoparasitism. Proc Natl Acad Sci USA 107: 16207–16211.
- 21. Oster GF, Wilson EO (1978) Caste and Ecology in the Social Insects. Princeton University Press, Princeton, New Jersey.
- 22. Wilson EO (1959) Some ecological characteristics of ants in New Guinea rain forests.Ecology 40: 437–447.
- 23. Wille A (1983) Biology of the stingless bees. Annu Rev Entomol 28: 41–64.
- 24. Heinrich B (1979) Bumblebee Economics. Harvard University Press.
- 25. Williams PH (2007) The distribution of bumblebee colour patterns world-wide: possible significance for thermoregulation, crypsis, and warning mimicry. Biol J Linn Soc 92: 97–118.
- 26. Dew RM, Rehan SM, Tierney SM, Chenoweth LB, Schwarz MP (2012) A single origin of large colony size in allodapine bees suggests a threshold event among 50 million years of evolutionary tinkering. Insect Soc 59: 207–214.
- 27. Houston TF (1976) New Australian allodapine bees (subgenus Exoneurella Michener) and their immatures (Hymenoptera:Anthophoridae).Trans R Soc S Aust 100: 15–28.
- 28. Dressler RL (1982) Biology of the orchid bees (Euglossini). Annu Rev Ecol Syst 13: 373–394.
- 29. Otero JT, Ulloa-Chacón P, Silverstone-Sopkin P, Giray T (2008) Group nesting and individual variation in behavior and physiology in the orchid bee Euglossa nigropilosa Moure (Hymenoptera, Apidae). Insect Soc 55: 320–328.
- 30. Richards MH, von Wettberg EJ, Rutgers AC (2003) A novel social polymorphism in a primitively eusocial bee. Proc Natl Acad Sci USA 100: 7175–7180.
- 31. Boomsma JJ (2007) Kin selection versus sexual selection: Why the ends do not meet. Curr Biol 17: R673–R683.
- 32. Boomsma JJ (2009) Life-time monogamy and the evolution of eusociality. Phil Trans R Soc 364: 3191–3207.
- 33. Kukuk PF (1994) Replacing the terms “primitive” and “advanced”: new modifiers for the term “eusocial”. Anim Behav 47: 1475–1478.
- 34. Thompson JD, Higgins DG, Gibson TJ (1994) CLUSTAL W: improving the sensitivity of progressive sequence alignment through sequence eighting, position-specific gap penalties and weight matrix choice. Nucleic Acids Res 22: 4673–4680.
- 35. Pagel M, Meade A (2004) A Phylogenetic mixture model for detecting pattern heterogeneity in gene sequence or character-state data. Syst Biol 53: 571–581.
- 36. Pagel M, Meade A (2005) Mixture models in phylogenetic inference. In: Gascuel O, editor. pp. 121–142. Mathematics of Evolution and Phylogeny Oxford University Press, U.S.A.
- 37. Rodríguez F, Oliver JF, Marín A, Medina JR (1990) The general stochastic model of nucleotide substitution. J Theor Biol 142: 485–501.
- 38. Green PJ (1995) Reversible jump Markov chain Monte Carlo computation and Bayesian model determination. Biometrika 82: 711–732.
- 39. Pagel M, Meade A (2008) Modelling heterotachy in phylogenetic inference by reversible-jump Markov chain Monte Carlo. Phil Trans R Soc 363: 3955–3964.
- 40. Pagel M, Meade A (2007) BayesTraits version 1.0 computer package. Software and Manual. Available: http://www.evolution.reading.ac.uk/BayesTraits.html.
- 41. Pagel M (1994) Detecting correlated evolution on phylogenies: a general method forcomparative analysis of discrete characters. Proc R Soc B 225: 37–45.
- 42. Lewis PO (2001) A likelihood approach to estimating phylogeny from discrete morphological character data. Syst Biol 50: 913–925.
- 43. Ives AR, Garland Tjr (2010) Phylogenetic Logistic Regression for Binary Dependent Variables. Syst Biol. 59: 9–26.
- 44. Midford PE, Garland T Jr, Maddison WP (2005) PDAP Package of Mesquite. Version 1.07.
- 45. Maddison WP, Maddison DR (2009) Mesquite: a modular system for evolutionary analysis. Version 2.6. http://mesquiteproject.org.