Fig. 3

Ancestral state reconstructions of mating type and chronogram of Madagascan Bulbophyllum clade C. a Reconstruction under an unconstrained Mk2 model using Bayestraits. b Maximum likelihood (ML) reconstruction under the best-fitting BiSSE model (see also Table 4). Pies at internal nodes indicate mean relative posterior probabilities assigned to alternative states based on 5,000 BI (a) or 1,000 Beast trees (b). Transitions in mating type are marked by an asterisk. c Branch lengths proportional to time (million years ago [Ma]) estimated with a relaxed clock approach (Beast) and summarized onto the topology of the majority-rule consensus of the Bayesian analysis. Divergence dates were inferred using both fossil-based and biogeographic (island age) calibration points (see [Additional files 7, 8, 9, 10, 11, 12 and 13]). Branches only weakly supported by Bayesian analysis (PP ≤ 0.95) are indicated through stippled lines. Horizontal bars indicate 95 % highest posterior density (HPD) for node ages. Also shown are ancestral state reconstructions of mating type under the model-based (BiSSE) reconstructions, with all internal nodes (small open dots) reconstructed as outcrossing. In all probability, selfing evolved independently at five times within clade C since about the mid-Pleistocene (ca. ≤ 1.3 Ma; see node marked by an asterisk). The inset shows a log-lineage-through-time (LTT) plot for Madagascan Bulbophyllum clade C, based on the chronogram (black line) and 1,000 trees sampled from the posterior distribution of the Beast analyses (gray lines)