Young, recently burned forests are increasingly widespread throughout western North America, but forest development after large wildfires is not fully understood, especially regarding effects of variable burn severity, environmental heterogeneity, and changes in drivers over time. We followed development of subalpine forests after the 1988 Yellowstone fires by periodically resampling permanent plots established soon after the fires. We asked two questions about patterns and processes over the past 25 years: (1) Are plant species richness and community composition converging or diverging across variation in elevation, soils, burn severity, and post-fire lodgepole pine (Pinus contorta var. latifolia) density? (2) What are the major controls on post-fire species composition, and has the relative importance of controls changed over time? For question 1, we sampled 10-m2 plots (n = 552) distributed among three geographic areas that differ in elevation and substrate; plots spanned the spectrum of fire severities and were resampled periodically from 1991 to 2013. For question 2, we sampled 0.25-ha plots (n = 72), broadly distributed across areas that burned as stand-replacing fire, in 1999 and 2012. Richness and species composition diverged early on between infertile low-elevation areas (lower richness) and more fertile high-elevation areas (greater richness). Richness increased rapidly for the first 5 yr post-fire, then leveled off or increased only slowly thereafter. Only 6% of 227 recorded species were nonnative. Some annuals and species with heat-stimulated soil seed banks were associated with severely burned sites. However, most post-fire species had been present before the fire; many survived as roots or rhizomes and regenerated rapidly by sprouting. Among the 72 plots, substrate, temperature, and precipitation (the abiotic template) were consistently important drivers of community composition in 1999 and 2012. Post-fire lodgepole pine abundance was not significant in 1999 but was the most important driving variable by 2012, with a negative effect on presence of most understory species, especially annuals and shade-intolerant herbs. Burn severity was significant in 1999 but not in 2012, and distance to unburned forest had no influence in either year. The 1988 fires did not fundamentally alter subalpine forest community assemblages in Yellowstone, and ecological memory conferred resilience to high-severity fire.