An experimental restoration in Wisconsin planted in 1986 tested the hypothesis that growing-season fire maintained richness of native herbaceous dicots (forbs). Replicated plantings were burned in May or July, or left unburned, every third year from 1989 to 2004 and monitored for differences in cover and richness through 2006. Native forb richness was higher in burned than unburned plots, with greatest richness following July burns. Two seasons after the 2004 fires, counts averaged 2 more native forb species in replicates burned in July than those burned in May and 4 more species in replicates burned in July than those left unburned. The strongest statistical response to fire season was higher richness of early-flowering species in replicates burned in July, largely attributable to early-flowering forbs planted in 1986 that persisted better after July burns than in other treatments. Spring fire increased cover of late-flowering C₄ grasses. As of 2006, C₄ grasses accounted for 76% cover after May fires, 52% after July fires, and 39% in unburned plots. Replicates burned in July held more alien species for the first 12 y, after which alien richness declined and differences among treatments disappeared. Summer fire best maintained richness of native, especially early-flowering, species.