Inbreeding depression is one of the major selective forces driving the evolution of mating systems. Previous theories predict that long-lived plants will show a negative correlation between inbreeding depression and the level of inbreeding (as determined by an inbreeding coefficient) at maturity, but the extent of this correlation may vary among life stages because of variation in the genetic basis for inbreeding depression at different stages. To test this prediction, I used electrophoretic allozyme analysis and pollination experiments to examine the fixation index (F is) at maturity and inbreeding depression in the early and late life stages of two populations with different outcrossing rates of a highly self-fertilizing tree, Magnolia obovata. The magnitude of inbreeding depression for early survival (δ e) in an outcrossing population (t m = 0.51; F is = −0.015) was higher (δ e = 0.97) than that in an inbreeding population (t m = 0.18; F is = 0.15; δ e = 0.38). From these results, I estimated that both populations exhibited high inbreeding depression for late survival (δ l) (0.94 in the outcrossing population and 0.93 in the inbreeding one) and lifetime survival (δ t) (0.99 and 0.96, respectively). My results and previously published data demonstrate the predicted relationship between inbreeding depression and the level of inbreeding for early survival, but not for late survival. This suggests that there is a differential genetic basis for inbreeding depression at different life stages. The inbreeding depression for late survival appears to play a central role in the maintenance of reproductive traits that promote outcrossing in M. obovata.