Modern ecology often emphasizes the distinction between traditional theories of stable, environmentally structured communities and a new paradigm of disturbance‐driven, nonequilibrium dynamics. However, multiple hypotheses for observed vegetation patterns have seldom been explicitly tested. We used multivariate statistics and variation partitioning methods to assess the relative influences of environmental factors and disturbance history on riparian and hillslope forests. Our study area was the Cummins Creek Wilderness, located in the Oregon Coast Range. Most of the wilderness burned at least once between the mid‐19th and early 20th centuries, creating a mosaic of younger forests with a few old‐growth patches. Species composition on hillslopes varied primarily along a climatic gradient from moist maritime environments to drier inland climates but was relatively insensitive to forest age structure. The abundance of Tsuga heterophylla, a fire‐sensitive, late‐successional tree species, decreased with distance from old‐growth patches, suggesting possible seed‐source limitations following the historical fires. In contrast to species composition, hillslope forest structure was primarily related to fire history but was largely independent of environmental gradients. Old‐growth structure characteristics such as large dominant trees, large snags, high down‐wood volumes, and high tree size diversity increased with stand age and with the presence of remnant trees that survived the fires. Riparian forests had high shrub cover, abundant hardwoods, and high down‐wood volumes, while the conifer‐dominated hillslopes had high overstory density and basal area. Maritime climates and their associated plant species extended further inland in riparian areas than on hillslopes. Advance regeneration densities were higher in riparian forests within 5 km of the coast than in any other portion of the study area. Riparian forest structure and composition were related to both environmental and disturbance variables, with stream gradient and size integrating much of the fine‐scale variability in disturbance regimes. No single theoretical framework was sufficient to explain the vegetation patterns observed in these forested watersheds. Our findings suggest a conceptual model of forest landscapes in which the relative influences of environment and disturbance on vegetation patterns are contingent on the facet of vegetation considered (composition vs. structure) and the portion of the landscape examined (hillslope vs. riparian).