In the Midwestern United States, oak savannas have been reduced in area by at least 99% since European settlement, mirroring global trends for savannas, grasslands, and shrublands. Most remaining patches are highly degraded following decades of fire suppression and other anthropogenic impacts, and subsequent tree and shrub encroachment. Yet, reintroducing fire alone may not be sufficient to restore these ecosystems on desired timelines and mechanical thinning may be an important step in the restoration process, to increase understory light and promote the ground layer community. However, it is unclear how plant community dynamics develop under burn-only compared to thin-and-burn restoration scenarios. We investigated the impacts of prescribed fire and mechanical tree thinning on ecosystem structure and plant community dynamics over eight years in an oak savanna restoration experiment in southern Michigan. Established in 2010, this experiment utilized 15 0.4-1.2-hectare treatment units receiving either repeated prescribed fire alone, a combination of repeated prescribed fire and mechanical thinning, or no management. We used this design to test how differences in management affect understory and overstory structure, specifically understory vegetative cover and light availability associated with canopy openness, and plant community dynamics, specifically ground layer plant species richness and composition. We found that, over eight years of restoration, the response of ecosystem structure and the plant community was greatest in units where mechanical thinning was combined with prescribed fire. Thinned and burned units had greater canopy openness, vegetative cover, and plant species richness. Plant species composition also diverged between managed and unmanaged units. Canopy openness increased rapidly, within two years of restoration, while vegetative cover increased more gradually, over five years, and increases in richness were less pronounced overall. Composition diverged initially between managed and unmanaged units and continued to shift throughout the study period. Some effects of management peaked after four years, but were transient by the end of the study. Additional management will be necessary to capitalize on the initial response to restoration toward an oak savanna ecosystem. We predict that additional thinning will further increase light availability and development of graminoid fuels and in combination with prescribed fires will continue to promote open canopy structure and a ground layer dominated by savanna-associated species.