I Population dynamics of Liriodendron tulipifera throughout the successional sere were investigated with a spatial simulator of forest community dynamics. The simulations were complemented by data from Appalachian cove forests at different stages of development. Gap size was a key consideration in simulation of Liriodendron recruitment; seed regeneration of this shade-intolerant species was unlikely except in large gaps (> 0.04 ha). 2 In simulations of long-term forest development (800 years) without exogenous disturbance, Liriodendron was a dominant species in early succession (50-150 years), but it was absent or present at low levels in stands > 500 years of age. The simulated basal area levels of Liriodendron approximated actual levels in cove forests having a disturbance regime characterized by a very low frequency of severe, large-scale events. 3 The idea that canopy gap size distributions change over the course of forest development was supported by the simulations. Given that intolerant species require large gaps for successful establishment, it was clear that temporal patterns of large gap formation affected the dynamics of such species. Intolerants were unable to regenerate during the stem exclusion phase, but they could regenerate during the subsequent understorey reinitiation and old-growth phases having large gaps. 4 Certain simulated disturbances imposed on old forests elevated Liriodendron basal area. Components of the disturbance regime included patch size, return interval, severity and spatial dispersion. The largest increases in Liriodendron basal area resulted from disturbance patches > 0.04 ha. Gaps, 0.1 ha in size, formed at intervals < 100 years by the mortality of c. 25% of the canopy trees in a stand, were sufficient to maintain the high levels of Liriodendron observed in some primeval forests of eastern North America.