Surrogates are used widely in ecology to detect or monitor changes in the environment that are too difficult or costly to assess directly. Yet most work on surrogates to date has been correlative, with little work on their predictive capacity or the circumstances under which they work. Our suggestion is to revisit and learn from research in the clinical medical sciences, including the causal statistical frameworks available to validate relationships between treatments, surrogate variables, and the outcome of interest. We adapt this medical thinking to ecology by providing a new framework that involves specification of the surrogate model, statistical validation, and subsequent evaluation in a range of spatial and temporal contexts. An inter-disciplinary surrogate concept will allow for a more rigorous approach to validating and evaluating proxy variables, thus advancing the selection and application of surrogates in ecology. Synthesis We draw together ideas from the medical sciences to define an explicit surrogate concept that has not previously been used in ecology. We present a new framework for specifying surrogate models involving validation using a causal framework, and subsequent re-evaluation in different spatial and temporal contexts - an approach closely aligned with that used by researchers in the clinical medical sciences. This rigorous method can advance the science underpinning the application of surrogates in ecology by shifting the focus away from correlative understanding to one that focuses instead on causation and prediction. An improved use of surrogates is imperative if we are to meet the challenge of properly measuring and understanding the multifarious and complex problems in contemporary ecology.