Smaller farms and plots appear more productive per hectare than larger ones in most developing country data. Using unique plot-level panel data from Uganda, with agricultural plots matched over a decade using geospatial location, we estimate the size-productivity relationship using variation in plot size over time under plot fixed effects, as well as variation in plot size within a farm, under household-period fixed effects. Like many other authors, we find that the observed inverse relationship arises at the plot, not farm, level, and that observable plot characteristics such as soil quality cannot explain the relationship. However, the plot perimeter/area ratio explains most of the inverse plot size-productivity relationship, reflecting an edge effect widely acknowledged in the agronomy literature wherein productivity is highest around the periphery of plots. We present suggestive evidence consistent with behavioral and biophysical mechanisms underpinning the edge effect.