(1) The STORM simulation was developed to predict the response of regional stand age structure to a given disturbance regime. The outcome of STORM depends upon rotation periods of canopy disturbance, diameter growth rates of trees, and the relative susceptibility of pole, mature and large trees to windthrow. (2) The simulation was run using data on disturbance frequencies from remnants of primary forest in western Upper Michigan. Rotation periods ranged from 69 years for > 10% canopy removal to 3734 years for > 70% canopy removal. Stands were classified into eight structural types, ranging from even-aged sapling stands to steadystate stands, which reflect the state of recovery from recent disturbances. (3) The model suggests that most (87-4%) stands in the primeval hardwood forest were multi-aged, with several major and many minor age classes. Quasi-even-aged stands resulting from catastrophic disturbance occupied 9-2% of the landscape, and steady-state stands about 3-6%. The low frequency of steady-state stands on the landscape appears to be determined primarily by the high frequency of disturbances removing part of the canopy, rather than the interval between catastrophes. (4) Sensitivity analyses indicate that changes in disturbance rates result in disproportionately large changes in the proportion of even-aged stands on the landscape, while mean canopy residence time for trees shows a damped effect.