Effect of rewarded and nonrewarded incorrect trials on T maze learning.