Questions Have predictable relationships between environmental variables and vegetation developed in primary succession following a volcanic eruption? Has the rate of succession changed? Have vegetation trajectories converged or diverged? Location The Abraham Plain of Mount St. Helens, Washington, USA (46 degrees 12'42 ' N, 122 degrees 08'27 ' W, elevation 1360 m), was sterilized in 1980 by a blast, scoured by lahars and buried by pumice. Method We monitored 400 100 m2 contiguous permanent plots annually (1988-2008), and classified each plot from every year into ten community types (CTs). We characterized the terrain by topography and surface features. Redundancy analysis assessed relationships between vegetation and possible explanatory variables, which included sample location. We used detrended correspondence analysis (DCA) to assess successional rates and trends. Results Relationships between species composition and explanatory variables were only significant after 1996, when position and presence of rills became significant. By 2006, explained variation remained low (13%) but significant. Species accumulated slowly, restricted by stress and isolation. Changes in mean DCA position slowed. Composition shifted from pioneer to persistent species and vegetation became more stable with time. Species accumulated for two decades and then stabilized, while cover has continued to increase. Diversity increased and then declined slightly as dominance developed and pioneer species became less common. Conclusions We demonstrate weak but increasingly predictable trends in species composition using environmental variables. The rate of succession slowed and trajectories formed a reticulate network of transitions dominated by divergence. Convergence was not evident because vegetation responded distinctively to minor topographic features that allowed alternative stable communities to develop.