Fig. 4.

Phylogenetic independence of cyclone mortality: representative randomization test for serial independence of cyclone mortality data, given the hypothesized phylogeny (Fig. 3). The observed mean C-statistic for the given series (20·6110) does not fall far to the right of the distribution of mean C-statistics for these 1000 randomized replicates (P = 0·1590). Cyclone tolerance is not confounded with phylogeny; lack of evidence of phylogenetic autocorrelation for cyclone tolerance leaves open the possibility of selection for higher cyclone tolerance dependent on geographic distribution.