Resource selection studies are common in the wildlife ecology literature and typically rely on the comparison of locations used by wildlife and locations assumed to be available for use but where use was not observed. While standard use-availability designs are helpful for establishing general patterns of species occurrence, they are limited in their ability to help researchers understand the underlying behavioural mechanisms that lead to observed space-use patterns. Based on spatially-explicit behavioural observations from coyotes Canis latrans in Yellowstone National Park, Wyoming, we estimated resource selection for specific behaviours (i.e. predatory, laying and travelling) and for all used locations irrespective of behaviour, to test whether resource selection is behaviour-specific and not generalizable across behaviours. Behaviour-specific models differed significantly from the model not partitioned by behaviour. In particular, the predatory model identified selection for mesic-meadows which have previously been documented to have high small-mammal abundance. The non-partitioned model, however, showed avoidance of this vegetation type. Our results show that resource selection differs between behaviours and suggest that standard techniques for estimation of resource selection might be of limited use for understanding the underlying behavioural mechanisms of space use. Future research should continue to improve on methods for partitioning fine-scale movement data obtained from telemetry collars into discrete movement bouts representative of different behaviours.