Juvenile, stream-dwelling salmon are a model system for studies of density dependence. However, while salmon exhibit diverse mechanisms and patterns of density dependence, the factors that drive variation in density dependence are not clear. I used the unique experimental setting of an ongoing population restoration effort to identify key factors that limit survival and growth of juvenile Atlantic salmon (Salmo salar), then tested how these factors interact with density dependence. Using a large-scale salmon population density manipulation in natural streams, I showed that slimy sculpin (Cottus cognatus), a common predator of salmon, limited salmon survival and reversed the direction of density dependence. Sculpin dramatically suppressed salmon survival at low salmon population density, but had little effect on survival at high salmon density. Such predator-mediated inverse density dependence can eliminate low-density populations even if predators have little effect when prey are abundant. Individual growth of juvenile salmon was also density dependent. Fish stocked at low density grew faster than those stocked at high density, and mean growth increased even more when high mortality further reduced population density. Increased growth may help compensate for decreased population abundance. However, variation in prey availability across streams, not population density, explained most of the variation in individual growth. Thus, the demographic effects of density-dependent growth depend on stream productivity. I also used the salmon restoration program as a controlled setting to investigate another applied problem: accumulation of toxic mercury in fish. Conducting ecological research within the applied context of an intensively managed population provides unique opportunities for direct application of the results as well as insight into the fundamental factors that drive population dynamics.