Performance of reaction time (RT) tasks was investigated in young children and adults to test the hypothesis that age-related differences in processing speed supersede a "global" mechanism and are a function of specific differences in task demands and processing requirements. The sample consisted of 54 4-year-olds, 53 5-year-olds, 59 6-year-olds, and 35 adults from Russia. Using the regression approach pioneered by Brinley and the transformation method proposed by Madden and colleagues and Ridderinkhoff and van der Molen, age-related differences in processing speed differed among RT tasks with varying demands. In particular, RTs differed between children and adults on tasks that required response suppression, discrimination of color or spatial orientation, reversal of contingencies of previously learned stimulus-response rules, and greater stimulus-response complexity. Relative costs of these RT task differences were larger than predicted by the global difference hypothesis except for response suppression. Among young children, age-related differences larger than predicted by the global difference hypothesis were evident when tasks required color or spatial orientation discrimination and stimulus-response rule complexity, but not for response suppression or reversal of stimulus-response contingencies. Process-specific, age-related differences in processing speed that support heterochronicity of brain development during childhood were revealed. (Contains 2 tables and 2 figures.)