If related species share enemies, variation in the damage experienced by species within a community may be predictable based on phylogeny. We examined the hypothesis that plant species more closely related to other community members experience greater herbivory by assessing leaf damage to native and exotic plants in two North American communities: an Eastern hardwood forest and a Rocky Mountain montane community. Pairwise phylogenetic distances between focal species and the hundreds of other native species in each community were calculated. We examined the influence of four measures of relatedness within each community: NND (phylogenetic distance to the nearest native neighbor), MPD (mean phylogenetic distance to the native species in the community), and two new metrics, MIPD (mean inverse phylogenetic distance) and INND (inverse nearest neighbor distance). These new metrics assume a nonlinear increase in interaction strength with relatedness; in the context of natural enemies, they posit that the sharing of enemies between any two species increases nonlinearly with their relatedness. Using regression models, we found that herbivore damage decreased with decreasing phylogenetic similarity of focal species to native species (as measured by MIPD) in both sites, although the pattern was significant only for native focal species in the montane community and exotic focal species in the hardwood forest. Similar decreases in herbivory with decreasing relatedness were detected using INND (montane natives) and MPD (hardwood forest exotics). There was no significant relationship between NND and herbivory for any of the four site by focal plant origin combinations. Our results are the first to support the hypothesis that native species can escape attack as a function of their phylogenetic dissimilarity to the larger community of native species, and to demonstrate that exotic species show these patterns in the wild (as opposed to in common gardens). We suggest that phylogenetic distance metrics assuming a nonlinear increase in interaction strength with relatedness show promise for broader application.