Lateral cascade of indirect effects in food webs with different types of adaptive behavior

It is widely recognized that indirect effects due to adaptive behaviors can have important effects on food webs. One consequence may be to change how readily perturbations propagate through the web, because species' behaviors as well as densities may respond to perturbations. It is not well understood which types of behavior are more likely to facilitate versus inhibit propagation of disturbances through a food web, or how this might be affected by the shape of a food web or the patterns of interaction strengths within it. We model two simple, laterally expanded food webs (one with three trophic levels and one with four), and compare how various adaptive behaviors affect the potential for a newly introduced predator to change the equilibrium densities of distant species. Patterns of changes in response to the introduction were qualitatively similar across most models, as were the ways in which patterns of direct interaction strengths affected those responses. Depending on both the web structure and the specific adaptive behavior, the potential for density changes to propagate through the web could be either increased or diminished relative to the no-behavior model. Two behaviors allowed density changes to propagate through a four-level web that precluded such propagation in the no-behavior model, and each of these two behaviors led to qualitatively different patterns of density changes. In the one model (diet choice) in which density changes were able to propagate in both web structures, patterns of density changes differed qualitatively between webs. Some of our results flowed from the fact that behaviors did not interact directly in the systems we considered, so that indirect effects on distant species had to be at least partly density-mediated. Our models highlight this as an inherent limitation of considering in isolation behaviors that are strictly foraging-related or strictly defense-related, making a case for the value of simultaneously considering multiple interacting types of behavior in the same model.