Identifying objects in conventional and contorted poses: contributions of hemisphere-specific mechanisms

Three experiments were designed to test the hypothesis that different mechanisms are used to encode objects seen in unfamiliar contortions than are used to encode objects seen in conventional poses. When a familiar non-rigid form (e.g. an animal) is seen in a contorted pose, we hypothesize that object identification may be achieved by (1) encoding the objectʹs parts separately, (2) encoding the spatial relations among the parts, and (3) matching these encodings to a stored structural description. However, once this form has become familiar, its global shape can be directly matched to information stored in memory. Based on the idea that `categoricalʹ spatial relations are encoded better by the left cerebral hemisphere and are used in structural descriptions, we predicted a left-hemisphere advantage when one first encodes contorted poses; in contrast, based on the idea that overall shapes are encoded better by the right hemisphere, we predicted a right-hemisphere advantage for encoding the same shapes after they are familiar. Three experiments confirmed these predictions, which supports the hypotheses that different visual mechanisms operate in the recognition of familiar and unfamiliar views of known non-rigid objects. Moreover, correlational analyses between visual-field differences in several perceptual tasks (matching whole pictures to names, body parts to the whole body, and judging categorical spatial relations) revealed that the degree and lateralization of categorical spatial encoding predicts the left hemisphereʹs initial advantage in the identification of contorted shapes.