The advent of high-speed computing has facilitated a revolution in modeling ecological systems. Increasingly, models of ecological populations and communities simulate the complex interplay between the local environment and the individual. This complexity is also the major liability of these models since it becomes increasingly difficult to understand which details about individual interactions control emergent behavior of the model. I identify the fine-scale interactions controlling broad-scale community behavior in SORTIE, a mechanistic, individual-based, spatially explicit simulation model of forests in the northeastern United States by simplifying aspects of the local interactions among model trees. SORTIE predicts robust features of forest landscapes such as old growth forest composition and patterns of forest succession through time. Forests simulated without local spatial structure exhibit significantly reduced total biomass, accelerated successional dynamics, and often predict the wrong dominant species relative to the detailed model. Two processes, competition for light and dispersal of offspring contribute to local spatial structure (covariance) in SORTIE. These two processes control different aspects of forest structure in SORTIE. Local competition for light controls forest biomass, while dispersal controls the rate of succession. I believe that experimentation with a detailed, data-defined model is a powerful method for addressing the related questions of relevant detail, emergent properties and scale.