The concept of refugee species provides a theoretical framework towards increasing the predictive power of the ‘declining population paradigm’ through identifying species which are expected to suffer from a declining population syndrome. Using a simple habitat model as a framework, refugee species are defined as those that can no longer access optimal habitat, but are confined to suboptimal habitats, with consequences of decreased fitness and density, and attendant conservation risks. Refugee species may be difficult to detect in the absence of information on prior habitat use and fitness and their observed ecology will be constrained by the habitat limits forced on them. Identification of refugee species, characterisation of pre‐refugee ecology and the restoration of such species to optimal habitat is critical to their successful conservation. The concept is showcased by addressing the conundrum of a large grazing bovid, the European bison Bison bonasus, being managed as a forest specialist, despite its evolutionary background, dental morphology, neonatal behaviour, diet and microhabitat selection being characteristic of a grazing species inhabiting open, grass‐rich habitats. It is hypothesized that a combination of increasing replacement of open steppe by forest cover after the last postglacial period and increasing human pressure forced bison into forests as a refuge habitat. This process was then reinforced through active management of bison in forests as managers committed themselves to the ‘bison as forest species’ paradigm. A research agenda to test this hypothesis using an experimental approach in the conservation management of European bison by introducing populations into diverse habitat types is suggested.