Evaluating alternative hypotheses for the origin of eusociality in corbiculate bees

We use a likelihood-based statistical test to evaluate the extent to which the available molecular data sets can be used to falsify alternative phylogenetic hypotheses describing the inter-relationship among corbiculate bee tribes. Based on the results of this test, we explore three alternative models of behavioural character state evolution and evaluate the support each model has for single-origin versus dual-origin hypotheses for 'highly' eusocial behaviour. We show that only one of four data sets could statistically reject any of the 15 possible outgroup-rooted phylogenetic hypotheses. However, a cytochrome b data set rejected all but three alternative topologies. Using this information, a simple model of behavioural character state evolution, in which transitions between solitary/communal, 'primitively' eusocial, and 'highly' eusocial are unconstrained, supports single-origin hypotheses for 'highly' eusocial behaviour, in spite of phylogenetic uncertainty. By contrast, an ordered model, in which 'highly' eusocial is constrained to be an evolutionarily terminal state, supports a dual-origins hypothesis. Our results show that the molecular phylogenetic evidence favouring a dual-origins hypothesis for 'highly' eusocial behaviour is, at present, conditional on information from one gene (cyt b) and on specific, though likely realistic, assumptions regarding the nature of eusocial evolution.