The scattered tree layer that defines savannas is important for structuring the understory community and determining patterns of overstory recruitment. However, encroachment by woody plants has altered overstory tree densities and regeneration dynamics. We characterized seedling success of the savanna-forming species Quercus alba within Midwestern (USA) oak savannas that had been degraded by encroachment (control; n = 4) or experimentally restored by removal of encroaching woody vegetation (treatment; n = 4). In early 2004, 981 seedlings were transplanted along transects radiating from tree boles of overstory Q. alba trees to inter-canopy gaps and monitored for three growing seasons. Seedlings in restored sites had greater survival (>2×), height growth (by >50%), and basal diameter growth (by >20%). In general, seedling survival and growth parameters increased with distance from overstory trees and were greatest in inter-canopy gaps of restored sites. By the final growing season (2006), the seedling survival-by-distance from tree correlation was stronger in control ( r 2 = 0.25) than treatment sites ( r 2 = 0.18), due to relatively uniform (and greater) survival at all distances from trees in treatment sites. In 2006, growth parameters (seedling height, diameter, Δ height, Δ diameter, and # leaves) were significantly (and more strongly) positively correlated with distance from trees in treatment sites. However, seedling herbivory was also greater after treatment and increased with distance from overstory trees. To understand seedling/microenvironment relationships, we created logistic (survival) and linear regression models (Δ height, Δ basal diameter, # leaves in 2006). Control seedling models had consistently greater predictive power and included more variables, suggesting that savanna restoration may decouple seedlings from their microenvironments, potentially by decreasing competition for limiting resources. Encroachment of the savannas in this study is limiting regeneration of Q. alba, suggesting substantially altered regeneration dynamics from those under which these savannas originally formed. Initial responses from our test of restoration, however, were promising and mechanical encroachment removal may be a means to promote overstory regeneration of this species. Finally, the savannas in this study appear inherently unstable and a scattered canopy tree configuration is unlikely to persist without regular disturbance, even in the restoration sites. Repeated mechanical thinning treatments with selected retention of recruiting Q. alba individuals or reintroduction of understory fire or grazing animals may be potential mechanisms for promoting long-term persistence of savannas at these sites.