Acquisition of Long-Term Visual Representations: Psychological and Neural Mechanisms

How do we so rapidly achieve an organized, coherent visual percept of our superficially chaotic world? One way of reducing the complexity of the input is to take advantage of the statistical regularities and regular co-occurrences between aspects of objects and between objects and their spatial locations. In this chapter, converging data obtained from normal and brain-damaged individuals, as well as from single unit recording studies in monkeys, are presented, all of which address the psychological and neural mechanisms associated with statistical learning. The first section deals with learning regularities associated with particular spatial locations, presumably a function of the dorsal ‘where’ stream and data from normal individuals and from patients with hemispatial neglect are presented. The second section reports the findings from human and monkey studies, which show how statistical contingencies of the visual environment are reflected in behavior and how neurons in monkey inferotemporal cortex, the ventral “what” stream, appear to mediate these statistical effects. Taken together, using data from a variety of methodologies, this work attests to the flexibility and robustness of the visual system and sheds light on the way in which perceptual organization operates to convert raw input into long-term visual representations.