Coastal and inland waters are continuing to decline in many parts of the world despite major efforts made to restore them. This is due in part to the inadequate role that ecological science has played in shaping restoration efforts. A significant amount of fundamental ecological knowledge dealing with issues such as system dynamics, state changes, context-dependency of ecological response, and diversity is both under-used by managers and practitioners and under-developed by ecologists for use in realworld applications. Some of the science that is being 'used' has not been adequately tested. Thus, restoration ecology as a science and ecological restoration as a practice are in need of reform. I identify five ways in which our ecological knowledge should be influencing restoration to a far greater extent than at present including a need to: shift the focus to restoration of process and identification of the limiting factors instead of structures and single species, add ecological insurance to all projects, identify a probabilistic range of possible outcomes instead of a reference condition, expand the spatial scale of efforts, and apply hierarchical approaches to prioritization. Prominent examples of restoration methods or approaches that are commonly used despite little evidence to support their efficacy are highlighted such as the use of only structural enhancements to restore biodiversity. There are also major gaps in scientific knowledge that are of immediate need to policy makers, managers, and restoration practitioners including: predictive frameworks to guide the restoration of ecological processes, identification of social-ecological feedbacks that constrain ecosystem recovery and data to support decisions of where and how to implement restoration projects to achieve the largest gains. I encourage ecologists to respond to the demand for their scientific input so that restoration can shift from an engineering-driven process to a more sustainable enterprise that fully integrates ecological processes and social science methods.