This paper evaluates the theoretical potential of the notion of 'constraints' for attempts to understand infant object knowledge, by considering some important senses in which constraints have been proposed and following through the implications of attempting to apply them to infant perceptual-cognitive development. Marr's explanatory framework helps to clarify the interpretation of data that suggest that domain-specific innate constraints on infant object perception. Once perception is appropriately contextualized within action, however, neither Marr's idea of a universal computational theory nor Keil's proposal for constraints in the form of abstract, general properties shared by initial and mature cognitive structures are readily applicable to development. These approaches face problems of 'substitutability', 'scale', and 'restructuring', and they mask significant qualitative changes in the computational mechanisms underlying infants' understanding.
This paper is not available online