Community ecologists use functional groups based on the rarely tested assumption that within-group responses to ecological processes are similar and thus members are functionally equivalent. However, recent research suggests that functional equivalency may break down with human impacts. We tested the equivalency assumption and model predictions of responses to simulated human alterations in nutrients and large herbivores for two models of coral reef algae, the Relative Dominance Model (RDM) and the Functional Group Model (FGM). Results of both mesocosm and field experiments using assembled communities were compared to model predictions, and within- and between-group variability were assessed. Both models' predictions of group response to herbivory matched experimental outcomes, but only the RDM predicted response to nutrients. However, within-group variability was dramatic, because the RDM grouped species with opposite responses to herbivory and the FGM grouped species with unique responses to nutrients. These heterogeneous responses resulted in loss of information and masked strong interactions between herbivory and nutrients that were not included in the models. As humans continue to impact major ecological processes in ecosystems globally, we postulate that functional-group models may need to be reformulated to account for shifting baselines.