Hawaiian lowland dry forests have been reduced by >90% since first human contact. Restoration has focused on protection from fire and ungulates, and removal of invasive grasses as ways to stimulate native forest regeneration. Despite these efforts, natural regeneration of native plants has been infrequent. To assess effects of previous restoration treatments on natural regeneration, we monitored seed rain and dynamics of seedlings and juveniles for a period of 3 yr (2004–2007) within three restoration units within a remnant forest on the island of Hawai'i. All units had been protected from fire for many decades but differed in time since ungulate exclusion and grass removal. The units were as follows: (1) long-term restoration (fenced 1956, grass removal initiated 1995), (2) shortterm restoration (fenced and grass removal initiated 1997), and (3) unmanaged (fenced 1997, no grass removal). Overall juvenile plant abundance was highest in the short-term unit, but native abundance was highest in the long-term unit. Native woody seedlings established in all units, but recruitment into larger size classes was restricted to units with grass removal, primarily the long-term unit. Native seed rain explained much of the variation in native seedling abundance between units with grass removal. Nonnative grass seed rain was extensive but was reduced by an order of magnitude with grass removal. This study suggests that natural regeneration may enhance restoration actions in sites with native canopy, but this is likely only when restoration activities have been maintained for several years to coincide with favorable rainfall conditions that are highly unpredictable over time.