This research tested the hypothesis that information-processing deficits associated with hyperactivity implicate a self-regulatory dysfunction. Hyperactive and control children were compared on nonspeeded classification tasks. In Studies 1 and 2, children classified sets of either three (triads) or four (tetrads) stimuli. The stimuli varied simultaneously on size and brightness or on length and density. They could be classified analytically (separably) on the basis of identical values for one dimension and holistically (integrally) on the basis of overall similarity. Control children made an equal number of dimensional classifications for triads and for tetrads. As predicted, however, hyperactive children made fewer dimensional classifications for tetrads, suggesting that they apparently resorted to less effortful holistic responding under the increased processing load. In an unexpected finding in Study 1, older hyperactive children appeared to behave like younger controls, making more dimensional classifications when size, rather than brightness, was the shared dimension. Study 3 explored the possibility that hyperactive children lagged behind the controls in their responding to the shared dimensions. Three age groups of normal children (mean ages: 5, 8, and 11 years) and adults were asked to classify stimuli that varied on size and brightness. Consistent with the lag hypothesis, normal preschoolers behaved like the young hyperactive children in Study 1, classifying equally on the basis of size and brightness. The discussion focuses on the contribution of processing load and stimulus salience to the cognitive deficits of hyperactive children.