Extended Data Fig. 6: Beak length before and after hurricane for macropterous soapberry bugs.

Boxplot of spatial sorting’s effect on eroding a history of natural selection driven host-associated beak length differentiation in macropterous soapberry bugs. Divergent host-associated beak lengths (s ee Extended Data Fig. 5) are compared pre- (white) versus post-hurricane (purple). Insect host association is denoted by images of the insect feeding on each of their given host plants: Cardiospermum (green), Koelreuteria (red) and Sapindus (orange). Model selection using linear mixed models found the three-way interaction to be significant (LMM: F_{(1, 518.5)} = 22.88, P < 0.001; Supplementary Table 24), however, a post hoc Tukey’s test suggests that host-associated beak lengths where no longer significantly different post-hurricane at the flooded sites (Tukey: P > 0.1). All Sapindus-associated sites flooded during the hurricane, so none were available in controls. Plots are aligned vertically by site condition with flooded sites on the left and unflooded control sites on the right. Plots are aligned horizontally by sex with females on the top and males on the bottom. The upper and lower edge of box plots indicate the first and third quartile, the midline indicates the median value and the whiskers show the 95% confidences intervals with dots as outliers. Boxes labelled with different letters are significantly different (Tukey’s test: P < 0.05). Sample sizes in grey included in panels.