No-take marine reserves are an increasingly popular conservation and management tool. Assessing reserve performance in an adaptive management framework ideally involves predicting the response of populations and communities to reserves (typically in the design process) and testing predicted outcomes against observations. Here we compare existing modeling and empirical studies on no-take marine reserves, and provide a prospectus for their future integration. Numerical models of ecological responses to reserves typically project longterm, steady-state interactions over the relatively broad spatial scales of larval dispersal, reserve configuration, fishing effort, and fish movement. Existing empirical studies focus on short-term outcomes over small scales, typically aggregated over many explanatory factors. Linking models and empirical data together for the adaptive management of marine reserves requires adjusting the spatial and temporal scales of models to match empirically feasible tests, and adjusting the metrics and scale of empirical studies to account for the interacting biological and human factors affecting reserve outcomes. © The Ecological Society of America.