Stem exclusion and understory reinitiation are commonly described, but poorly understood, stages of forest development. It is assumed that overstory trees exert strong controls on understory herbs and shrubs during the transition from open- to closed-canopy forests, but long-term observations of this process are rare. We use long-term data from 188 plots to explore patterns and correlates of variation in understory richness and abundance 15–45 years after clear-cut logging and burning of two experimental watersheds in western Oregon, USA. We test whether variation in the temporal dynamics of plots can be explained by topoedaphic factors that influence resource availability (insolation and soil moisture), variation in the pace and intensity of overstory development, or characteristics of the vegetation prior to canopy closure. Changes in forest structure were substantial over the study period; canopy cover increased fourfold, stem density by 75%, and bole biomass by two orders of magnitude, although trends were highly variable among individual plots. In contrast, understory richness, foliar cover, and biomass declined only 30–40%, driven by loss of early-seral colonists, not residual forest species. Canopy closure occurred earlier on north aspects but declines in understory biomass, reflecting loss of colonizing shrubs (without concomitant increases in forest shrubs), were limited to south aspects. In contrast, variation in effective soil moisture had little influence on the pace of decline. Temporal trends were highly asynchronous among plots: nearly 50% of plots experienced some form of decline, but >35% showed no discernible trend. Declines were more likely in plots with greater tree influence before or at peak overstory development, but also in plots with greater understory development prior to canopy closure. Quantile regression models indicated weak relationships between understory biomass and overstory structure at most points in time. Our long-term data support a model of understory dynamics in which characteristics of the pre-closure vegetation are as important as overstory structure in determining the timing and nature of decline. Long-term studies are critical for elucidating patterns and processes that cannot be inferred from short-term experiments or space-for-time substitutions.