The likelihood that degraded watersheds can recover after elimination of an anthropogenic stressor and the time required for these systems to return to pre-disturbance conditions (defined here as ecological resilience) are critical questions in applied ecology. Despite significant investments in restoration projects globally, quantifying ecological resilience and predicting the success of these programs remains challenging. Inadequate study designs, ineffective pre- and post-restoration monitoring and failure to consider ecological theory significantly impede our ability to learn from these management “experiments.” Ecological resilience of streams is determined by numerous factors, including the specific nature of the disturbance (e.g., chemical versus physical), watershed characteristics, natural community variation and the proximity of sources of recolonization. Compound perturbations and the establishment of novel communities in disturbed ecosystems often impede recovery, even after complete elimination of a stressor. A better understanding of factors that influence ecological resilience is essential for predicting the success of watershed restoration projects. Using long-term (20-37 years) data from several western streams previously impacted by historic mining, I examined the relative importance of biotic and abiotic factors that affected ecological resilience. The distance from upstream sources of recolonization and the establishment of novel, metal-tolerant communities were especially important for determining the rate of recovery. Long-term shifts in community composition resulting from alterations in stream hydrology, water temperature and other physicochemical characteristics associated with climate change also affected recovery trajectories. Finally, because responses to restoration are endpoint-specific, our definition of what constitutes recovery is often influenced by conflicting ecological and socioeconomic considerations. Although full recovery of some watersheds is elusive, even highly damaged systems can eventually approach reference conditions following the removal of a stressor. Maintaining public support for ecological restoration requires that we objectively document the success of these projects but not overstate the potential for a complete return to pre-disturbance conditions.