Ecologists strive to build a theory to predict reliably population and community dynamics across taxa and systems. This ability is often hampered by an incomplete empirical understanding of the functional way different components fit together to determine dynamics. We show here how to overcome this critical hurdle. We describe a research program whose aim is to understand the implications of body size-dependent responses of herbivores to predation risk for food web interactions. We illustrate how to introduce body size-dependent risk response rules into a modeling framework that can scale from individual detail to community-level phenomena. We then show how to use the model to predict the attendant effects of such size-dependent responses of herbivores to predation risk on trophic interactions and food web structure, and explain how to evaluate the net food web effects of variation in herbivore body size using field experimentation that emulates the modeling. Our focus is on effects of predation risk and resource quality on size-dependent risk responses of grasshoppers in a New England old-field ecosystem. We show qualitative congruence between results of the simulation and field experiments, specifically, that grasshopper behavior can have a profound effect on the character of trophic interactions and the plant composition of the food web. We find, however, that herbivore body size had no significant effect on the net abundance of different plant resources. We therefore conclude that it would be reasonable to abstract considerable complexity associated with size-dependent responses of grasshoppers to predation risk and represent behavior more simply as their mean risk response in order to predict reliably the strength of plant herbivore interactions in this food web.