Returning native species to habitats degraded by biological invasions is a critical conservation goal(1). A leading hypothesis poses that exotic plant dominance is self-reinforced by impacts on ecosystem processes, leading to persistent stable states(2-6). Invaders have been documented to modify fire regimes, alter soil nutrients or shift microbial communities in ways that feed back to benefit themselves over competitors(2,5-7). However, few studies have followed invasions through time to ask whether ecosystem impacts and feedbacks persist(8,9). Here we return to woodland sites in Hawai'i Volcanoes National Park that were invaded by exotic C-4 grasses in the 1960s, the ecosystem impacts of which were studied intensively in the 1990s(10-12). We show that positive feedbacks between exotic grasses and soil nitrogen cycling have broken down, but rather than facilitating native vegetation, the weakening feedbacks facilitate new exotic species. Data from the 1990s showed that exotic grasses increased nitrogen-mineralization rates by two-to fourfold, but were nitrogen-limited(10,12,13). Thus, the impacts of the invader created a positive feedback early in the invasion. We now show that annual net soil nitrogen mineralization has since dropped to pre-invasion levels. In addition, a seedling outplanting experiment that varied soil nitrogen and grass competition demonstrates that the changing impacts of grasses do not favour native species re-establishment. Instead, decreased nitrogen availability most benefits another aggressive invader, the nitrogen-fixing tree Morella faya. Long-term studies of invasions may reveal that ecosystem impacts and feedbacks shift over time, but that this may not benefit native species recovery.