Forest removal for livestock grazing is a striking example of human-caused state change leading to a stable, undesirable invasive grass system that is resistant to restoration efforts. Understanding which factors lead to resilience to the alternative grass state can greatly benefit managers when planning forest restoration. We address how thresholds of grass cover and seed rain might influence forest recovery in a restoration project on Hawaiʻi Island, USA. Since the 1980s, over 400,000 Acacia koa (koa) trees have been planted across degraded pasture, and invasive grasses still dominate the understory with no native woody-plant recruitment. Between this koa/grass matrix are remnant native Metrosideros polymorpha (ʻōhiʻa) trees beneath which native woody plants naturally recruit. We tested whether there were threshold levels of native woody understory that accelerate recruitment under both tree species by monitoring seed rain at 40 trees (20 koa and ʻōhiʻa) with a range of native woody understory basal area (BA). We found a positive relationship between total seed rain (but not bird-dispersed seed rain) and native woody BA and a negative relationship between native woody BA and grass cover, with no indication of threshold dynamics. We also experimentally combined grass removal levels with seed rain density (six levels) of two common understory species in plots under koa (n = 9) and remnant ʻōhiʻa (n = 9). Few seedlings emerged when no grass was removed despite adding seeds at densities two to 75 times higher than naturally occurring. However, seedling recruitment increased two to three times once at least 50% of grass was removed. Existing survey data of naturally occurring seedlings also supported a threshold of grass cover below which seedlings were able to establish. Thus, removal of all grasses is not necessary to achieve system responses: Even moderate reductions (~50%) can increase rates of native woody recruitment. The nonlinear thresholds found here highlight how incremental changes to an inhibitory factor lead to limited restoration success until a threshold is crossed. The resources needed to fully eradicate an invasive species may be unwarranted for state change, making understanding where thresholds lie of the utmost importance to prioritize resources.