Fig. 2: Plant–herbivore interactions of native and exotic plants and their composite communities in the mesocosm experiment.
From: Exotic plants accumulate and share herbivores yet dominate communities via rapid growth
A Herbivore species were twice as likely to interact with exotic (orange, n = 2876 potential plant–herbivore interactions, with 961 realised) than native (blue, n = 2652 potential plant–herbivore interactions, with 646 realised) plant species within their fundamental host range (F = 5.93, P = 0.015). B Mean herbivore species biomass was 72% higher on exotic (n = 1333 plant–herbivore interactions) than native (n = 809 plant–herbivore interactions) plant species (F1,41 = 24.71, P = 0.0001). C Exotic herbivore total biomass per mesocosm (log-transformed; orange circles) increased with the proportion of exotic species planted into mesocosm communities (slope = 1.78, t = 4.29, P = 0.00009), whereas no relationship was observed for native herbivores (blue triangles, slope = 0.28, t = 0.68, P = 0.501; plant provenance × herbivore provenance interaction: F1,134 = 43.67, P = 8.4e−10; n =  80 mesocosms per herbivore provenance). D Mean herbivore species richness (quantified as normalised degree, the proportion of interactions observed out of all possible interactions) did not differ between native (n = 193) and exotic (n = 242) plants (F1,48 = 1.35, P = 0.251). E Herbivore species richness of mesocosm communities (n = 80) increased with the proportion of exotic plant species planted (slope = 0.41, F1,18 = 9.65, P = 0.002). F Mean percent chewing and scraping damage to leaf tissue from invertebrate herbivores did not significantly differ between native (n = 320) and exotic (n = 320) plants (F = 12.76, P = 0.062). Different lowercase letters indicate significant differences (P < 0.05) between back-transformed estimated marginal means (±SEM) from (generalised) linear mixed models. Scatterplot linetypes indicate slopes that significantly differ from zero (solid lines, P < 0.05) or do not (dashed lines). A small amount of jitter has been added to separate overlapping points on the x-axis. Corresponding violin plots showing the distribution of raw data are presented in Supplementary Fig. 15.
