The frequency of canopy disturbance over the past 150 years was reconstructed on a 5-ha study area dominated by a patchy mosaic of old-growth sugar maple (Acer saccharum Marsh.) and eastern hemlock (Tsuga canadensis (L.) Carr.) forest in the Sylvania Wilderness Area in western Upper Michigan. The study area was divided into a 10-m grid system and one tree was cored near the center of each grid cell so that the spatial patterns of tree cohorts could be examined. The canopy turnover rate, averaged over all species and 150 years was 5.4% per decade, with a corresponding canopy residence time of 186 years. Canopy-residence times do not vary much between sugar maple (170 years) and hemlock (167 years), but yellow birch has a much longer canopy-residence time (232 years). Canopy-residence times calculated for individual decades over the last 150 years varied from 81 to 556 years. The spatial pattern of gaps of various ages is caused by disturbances in light intensity (2-12% canopy removal) that occur nearly every decade, each of which creates several to many small gaps scattered across the study area. As a result, the study area has a fine-grained random spatial mixture of age-classes at all distance classes from 5 m to >100 m. This mixture is stable throughout the mesic forest in the study area. None of the cohorts resulting from disturbance correspond spatially to patches dominated by either hemlock or sugar maple. Apparently, the dynamics of patch formation by gap-creating disturbances operate independently from the dynamics of the much larger monodominant patches. In forests such as the northern hardwood - hemlock type, where several tree lifetimes pass between any two large-scale catastrophic disturbances, spatial and temporal stability of the patch-dynamic processes (quasi equilibrium) may exist for periods of several decades in areas of <1 ha, and several thousand years for landscapes >10 000 ha in size.