The observation that, for unstressed plants, light use efficiency of a plant canopy, defined as the ratio of net primary productivity (NPP) to absorbed photosynthetically active radiation (APAR), is approximately constant with respect to changes in APAR, implies that NPP can be modeled using a linear relationship with APAR. However, such a linear relationship is counter-intuitive because the relationship between leaf photosynthesis and absorbed light is strongly nonlinear. Three arguments have been advanced to explain the observed linear relationship between NPP and APAR. In this paper, a detailed, physiologically based model of canopy radiation absorption and photosynthesis (MAESTRO) was used to analyze these arguments. The first argument is that the canopy is structured so that radiation is distributed throughout the canopy such that most leaves are exposed to non-saturating quantum flux density, resulting in a linear response of canopy photosynthesis to APAR. Simulations of MAESTRO indicated that this explanation is inadequate, because daily values of canopy photosynthetic light use efficiency calculated with MAESTRO were highly variable regardless of canopy structure. The second argument is that variability in light use efficiency decreases with increasing time scale. The simulations showed that this is true to some extent, although simulated annual canopy photosynthetic light use efficiency still varies across sites with different LAI or light climate. The third argument is that changes in canopy nitrogen content act both to maximize net canopy photosynthesis and to keep light use efficiency constant. This argument could not be tested with the model, but the failure of the first two explanations suggests that this third explanation deserves closer attention.