The effect of natural enemies on the coexistence of competing species - an empirical test using Bayesian modern coexistence theory

O_LIThe role of natural enemies in promoting coexistence of competing species has generated substantial debate. Modern coexistence theory provides a detailed framework to investigate this topic, but there have been remarkably few empirical applications to the impact of natural enemies. C_LIO_LIWe tested experimentally the capacity for a generalist enemy to promote coexistence of competing insect species, and the extent to which any impact can be predicted by trade-offs between reproductive rate and susceptibility to natural enemies. C_LIO_LIWe used experimental mesocosms to conduct a fully-factorial pairwise competition experiment for six rainforest Drosophila species, with and without a generalist pupal parasitoid. We then parameterised models of competition and examined the coexistence of each pair of Drosophila species within the framework of modern coexistence theory. C_LIO_LIWe found idiosyncratic impacts of parasitism on pairwise coexistence, mediated through changes in fitness differences, not niche differences. There was no evidence of an overall reproductive rate - susceptibility trade-off. Pairwise reproductive rate - susceptibility relationships were not useful shortcuts for predicting the impact of parasitism on coexistence. C_LIO_LIOur results exemplify the value of modern coexistence theory in multi-trophic contexts and the importance of contextualising the impact of natural enemies. In the set of species investigated, competition was affected by the higher trophic level, but the overall impact on coexistence cannot be easily predicted just from knowledge of relative susceptibility. Methodologically, our Bayesian approach highlights issues with the separability of model parameters within modern coexistence theory and shows how using the full posterior parameter distribution improves inferences. This method should be widely applicable for understanding species coexistence in a range of systems. C_LI