An understanding of trajectories in long-term forest development is essential for examining several fundamental issues in forest ecology and management. Long-term forest development has important implications for issues as diverse as forest productivity, carbon storage, biological diversity, and ecological forestry methods. While much has been learned from studies of old-growth forests, a limitation is that these have necessarily been brief snapshots of species composition and forest structure at one point in time from which accurate trends are difficult to infer. The overarching themes of this dissertation are the long-term trajectory of forest development and the impact of the historic natural disturbance regime on that trajectory. Specific objectives were: (1) to quantify long-term trends in above-ground tree biomass and stand structure in northern hardwood forests, and effects of the natural disturbance regime, (2) to estimate the frequency of structural stand stages, along with residence times and transition rates among stages, under several disturbance regimes, and (3) to develop a quantitative approach to evaluating demographic sustainability of tree populations under several conditions, including those of restricted recruitment. Computer simulation with the CANOPY model was used to clarify long-term trends. However, to the extent possible, these model predictions were verified against available field data, including an extensive field survey and a 30-year record of change on permanent plots in the Porcupine Mountains Wilderness State Park. Major conclusions from this work were: (1) The Bormann-Likens hypothesis of a peak in biomass followed by a decline to a lower level was largely supported by both simulation and field data. The asymptotic biomass trend documented in recent chronosequences may be a consequence of using maximum tree age to estimate stand age. (2) The majority of landscapes of old-growth northern hardwoods is occupied by non-equilibrium stands, but nearly 1/3 was also structurally indistinguishable from steady state. Individual stands were highly dynamic, spending only a few decades in any structural stage. (3) Minimum sustainable tree populations had a variable slope in their size distribution. Enforcing a shallow fixed slope in understory trees typical of those used in uneven-aged management for larger trees rendered the stand unsustainable.