Fig. 4

Proportion of lineage splitting events characterized by reconstructed host shifts (i.e. shifts in either host species or host tissue used) and by reconstructed dispersal events within different time periods. Host shift and region are not exclusive: a split may be associated with both host shift and allopatry, one of these, or neither. a Proportions based on parsimony reconstruction on the majority rule BEAST tree (Figs. 2 and 3), with states represented by > 5% of total specimens considered. Tips are recorded as different if they do not overlap in character state; DELTRAN optimization was used to resolve reconstruction ambiguity at internal nodes. Dashed lines indicate proportions of splits > 1 Ma (truncated tree). The number of splits in each category is listed above each bar. b Boxplots summarizing distribution of proportions found in 2500 stochastic character mappings simulated on 250 randomly chosen untruncated BEAST trees (see Additional file 3: Figure S3 for results based on truncated tree). Observed frequencies of each character state among identified specimens were input as Bayesian priors for each simulation