The same operations that define “integral stimuli” for adults might be the operations required to provide a positive specification of the undifferentiated-holistic mode, presumably characteristic of the young child, and those operations that define “separable stimuli” for adults might provide the specification for the differentiated-analytic mode. In a classification task, adults sort color chips differing in saturation and brightness into groupings dictated by overall similarity relationships rather than groupings determined by possession of a common value on either dimension. The hypothesis that undifferentiated perception is characteristic of young children predicts correctly that preschool children are much less likely to form classes based on the dimensional structure of the set of objects. The recommended interpretation of young children’s classifications as based on overall similarity gains support from recent studies specifically designed to test the revised hypothesis. If the holistic, undifferentiated state is difficult to contemplate in the abstract, an appeal to intuitions might be helpful.