1. Successional change following disturbance is a fundamental ecological process that remains central to understanding patterns in plant ecology. Although succession has been studied for well over a century, understanding of the patterns and processes of change is still inadequate, partly because of the dearth of long-term studies. 2. Here, we use, as a model system, a volcanic disturbance that is widespread and of global relevance. We examine succession in old-growth conifer forest understories following burial by tephra (aerially transported volcanic ejecta) in the 1980 eruption of Mount St. Helens. Using four sites with different initial conditions and amounts of disturbance (tephra burial), we sampled plant communities at years 1, 20 and 36 since the eruption using permanent plots representing two treatments: undisturbed tephra and control plots from which tephra was experimentally removed. By using this long-term data, we were able to gauge change through time and differences between tephra-disturbed and control plots. 3. In tephra-impacted plots, cover reached pre-eruption levels by year 36 except for bryophytes at three sites and shrubs at one site. Cover in control plots also increased significantly for all growth forms except bryophytes, remaining above that on tephra plots at the same site in most situations. Sites generally increased in species richness first, and then gradually increased in evenness and Shannon’s diversity; after 36 years, differences between control and tephra-impacted plots remained at only one site. Community composition was stable in one site and shifted gradually through time at the two other sites, and differences between plot types persisted at one site. Communities change over 36 years could relate to various endogenic and abiotic factors. 4. Synthesis: These data suggest that, while recovery to the pre-disturbance status occurs in some cases, recovery is site-specific. Understory communities can be dynamic, even in old-growth forests, and may still be responding to disturbance over three decades later. Long-term experimental studies are key to understanding succession patterns and generalizations about the importance of initial conditions, disturbance intensity, and the complexity of interactions that determine vegetation change.