Predators are generally under selective pressure to get better at foraging, leading to steeper functional responses and stronger predator-prey interactions. Yet strong interactions can de-stabilize food webs, and most interactions across ecological communities are thought to be weak. This conflict between evolutionary and community expectations for the strength of predator-prey interactions represents a fundamental gap in our understanding of how the evolution of foraging plays out in food webs. Here we help to resolve the conflict by showing analytically that the expectation for the evolution of steeper functional responses is relaxed in communities with diverse prey types. We simulate communities with varying prey richness and show that increasing prey richness can indeed constrain the adaptive potential of predator foraging traits, but that at low prey richness predators can evolve to have a stronger interaction with prey that have high net energy yields. Our results also indicate that handling time plays a role in determining whether predators may evolve to have a stronger interaction with abundant prey, suggesting that the evolution of keystone predator modules in food webs is most likely when handling times are negligible. Our results also provide a new mechanism predicting more diffuse interactions in diverse tropical communities relative to more species-poor communities at higher latitudes.

This set of Matlab scripts and functions will initiate and perform the GEM simulation. Start with "Call_multi_prey_GEM.m", which is where you will specify prey levels to use, type of competition, and number of replicates. This file then calls the primary GEM function in a loop. The primary GEM function saves the simulations each in turn as the script runs through the levels and replicates. The primary GEM function is "GEMv5_multi_prey.m". This is a function and needs to be called. It is set up to run in parallel, so careful attention must be paid to changes in the script structure. This function depends on other subfunctions: "V4_pick_individuals.m", "V5_medians_and_cis.m", "V4_initiate_populations.m", and "create_multi_prey_model.m" that handle the repeated tasks. Funding provided by: James S. McDonnell FoundationCrossref Funder Registry ID: http://dx.doi.org/10.13039/100000913Award Number:

The data uploaded here are abundances and traits through time generated in simulations. Each file is a replicate given a level of prey species richness (1, 2, 4, 8) and interspecific competition (0.5 or 1, with 1 reflecting neutrality). Each file is named according to the run, e.g. "GEM_output_alpha=0.5_2_1_prey_high_h" is the output for competition of 0.5 (complementary prey species) with 2 prey types and replicate 1. In this case, we are also specifying a high handling time. These data can be imported and analyzed however you want, but the code we used is provided in the file called "Plot_GEMv5_multi_prey_multi_run4.m". This is a Matlab file. It will import each of the simulation files and reproduce the figures in the manuscript. It requires the function "jbfill.m". The remaining files are code to run the simulations. See "Usage notes" for explanation.