Perceptual development.

Our selective review of perceptual development has attempted to highlight three different structural levels: sensory primitives, perceptual representations, and higher-order operations. Each of these levels appears to undergo considerable development. Moreover, both within- and between-level developments appear to influence the emergence and form of perceptual abilities. The distinction between levels is often elusive, defined more by task than by logic. As a result, some distinctions seem clearer than others. First, the emergence of sensory primitives, particularly in the visual modality, may severely constrain the quality of perceptual representations. Despite these constraints, however, there is strong support for the hypothesis that distal stimulation is primary at the level of perceptual representation. Second, perceptual representations, particularly in the domain of speech sounds, appear to undergo considerable reorganization that cannot be attributed to the emergence of sensory primitives. Although some of this development in perceptual representations can be attributed to representational changes per se, much of this reorganization at the level of perceptual representations appears to be mediated by emerging cognitive operations and linguistic processes. Unfortunately, an evaluation of reorganizational factors (e.g. familiarity) that are potentially independent of higher-order cognitive and language influences is confounded by the co-occurrence of postnatal experience with the emergence of more mature cognitive and language abilities. Obviously, to disentangle the effects of experience from the effects of higher-order processes on perceptual development, it would be necessary either to eliminate these higher-order processes or to accelerate their potential influence. Controlled experiments of this sort are difficult or impossible to conduct with humans, although suitable animal models of some aspects of perceptual development may be possible. Current and future research on perceptual development appears to be directed toward three general goals. First, the ecological and psychophysical perspectives appear headed for a marriage that will enhance both the richness and the utility of these two approaches. Second, computational techniques appear to offer a new level of sophistication that should force developmentalists to sharpen their hypotheses and to make point predictions rather than simple directional predictions. Third, the search will continue for methods to assess the role of higher-order processes.