Many change agents in western forests, such as fire, wind, drought, and beetle outbreaks, are episodic, and although they can occur on decadal frequencies – a long time horizon for most research – forest stands that are centuries old have likely experienced these disturbances many times. Each of these episodic disturbances affects different species and different diameter classes with patterns that can be random, highly aggregated, or overdispersed. The observed variation in both the distribution and sizes of forest trees in primary forests therefore represents the integration of many occurrences of these episodic disturbances over tree lifespans. Attempts to restore forest composition and structure are more difficult when the range of structural heterogeneity is poorly understood, and if heterogeneity is underestimated, ecosystem function in restored stands may be limited. Better understanding of forest structure and structural development requires large data sets that can capture the variation present within forest stands. Data sets must include sufficient spatial representation for elucidation of characteristic patterns, and also sufficient temporal depth to understand the range of variation of demographic processes.