Reward magnitudes of 1, 2, or 4, food pellets on one side of a T maze were combined factorially with rewards of 0, 1, or 2 pellets on the other. Groups given a choice between unequal incentives eventually learned to choose the larger; rate of acquisition was affected by both incentive values. A quantitative description of the relationship between learning-rate and reward amount was proposed. Extinction rate, both in terms of choices and running speeds, was essentially independent of acquisition incentive conditions.