We examined vegetation dynamics from 2004 through 2010 in replicate 90 × 90-m hemlock-dominated plots in which we experimentally simulated hemlock woolly adelgid (Adelges tsugae Annand) infestation (by girdling trees) and preemptive logging of Eastern hemlock [Tsuga canadensis (L.) Carrière]. Using Chao-Sørensen abundance-based similarity indices, we assessed compositional similarities of forest vegetation (trees, shrubs, forbs, and graminoids) between the seed bank (Fig. 1), seed rain, and standing vegetation over time and among treatments. Post-treatment seed rain was a reliable proxy for canopy tree composition and was similar among replicate plots. Per-plot species richness in the seed bank was nearly identical in 2004 (24 species) and 2010 (23 species). After seven years, standing vegetation in the hemlock controls remained dissimilar from the seed bank, reflecting suppressed germination (Fig. 2). In contrast, the seed bank and standing vegetation were statistically identical in the girdled (Fig. 3) and logged plots (Fig. 4) in 2010. Understories of the logged and girdled plots differed in 2004 and 2010 and provided predictors of future forest composition. Our experimental findings accord with long-term observations of forests following hemlock decline. Our approach uniquely integrates multiple life-stages and compartments through time and provides a model for future studies of forest dynamics following loss of foundation tree species.