Despite the apparent advantages of adaptive plasticity, it is not common. We examined the effects of variation and uncertainty on selection for plasticity using an individual-based computer simulation model. In the model, the environment consisted of a linear gradient of 50 demes with dispersal occurring either before or after selection. Individuals consisted of multiple loci whose phenotypic expression either are affected (plastic) or are not affected (nonplastic) by the environment. Typically, evolution occurred first as genetic differentiation, which was then replaced by the evolution of adaptive plasticity, opposite to the evolutionary trend that is often assumed. Increasing dispersal rates selected for plasticity, if selection occurred before dispersal. If selection occurred after dispersal, the highest plasticity was at intermediate dispersal rates. Temporal variation in the environment occurring after development, but before selection, favored the evolution of plasticity. With dispersal before selection, such temporal variation resulted in hyperplasticity, with a reaction norm much steeper than the optimum. This effect was enhanced with negative temporal autocorrelation and can be interpreted as representing a form of bet hedging. As the number of nonplastic loci increased, plasticity was disfavored due to an increase in the uncertainty of the genomic environment. This effect was reversed with temporal variation. Thus, variation and uncertainty affect whether or not plasticity is favored with different sources of variation-arising from the amount and timing of dispersal, from temporal variation, and even from the genetic architecture underlying the phenotype-having contrasting, interacting, and at times unexpected effects.