The beaver, exterminated in Sweden in the 19th century, was reintroduced in the 1920s. The population is still increasing its range and individuals at the expansion front can still move into optimal habitats. Therefore, there is no need for beavers to occupy marginal habitats, and preferred habitats are not yet altered by the beavers themselves. Salient features of occupied and unoccupied habitats were analysed by partial least square projections to latent structures (PLS) in order to identify the important factors. The tortuosity of streams, and occurrence of grasses and forbs in the ground layer, had the strongest positive effects. Availability of deciduous forest differed univariately between used and unused sites, but added little to the model. It is suggested that variables describing geomorphological stages in stream development will increase both the general applicability and reliability of studies that attempt to assess the quality of beaver habitat.