Habitat loss and alteration is the leading cause of species' declines world-wide, therefore habitat restoration and protection is a prominent conservation strategy. Despite obvious connections between habitat and threatened or endangered species, conservationists have been hard pressed explicitly to link abundance or population health with habitat attributes. Given that habitat relationships with species are often characterized at a spatial scale that does not account for the functional relationships between habitat and populations, it is not surprising that the habitat-population conundrum persists. In order to explore the influence of spatial scale on the apparent relationship between habitat and populations, we examined the relationship between GIS-based habitat data and spring/summer chinook salmon (Oncorhynchus tshawytscha) redd (spawning nests built by females) densities in the Salmon River basin, Idaho, at two very different spatial scales: stream reach and watershed. Redd density was strongly correlated with climate, geology, wetlands and terrain. However, our stream-reach scale models provided poor predictive power compared with the watershed scale models. Based on these results, we conclude that our perception of which habitat attributes were important was clearly a function of our scale of observation, and that restoration efforts should focus on conditions at the watershed or landscape scale when attempting to do local or reach scale restoration projects.