The role of parts and spatial relations in object identification

An investigation of the role of parts and their spatial relations in object identification is reported. At the most general level, two important results were obtained. First, proper spatial relations among components of an object are critical for easy identification. When parts were scrambled on the page, naming times and error rates increased. And, second, the way an object is divided into parts (parsed) affects identification only under the most impoverished viewing conditions. When subjects had as little as 1 s (and sometimes as little as 200 ms) to view an object, the way objects were divided into parts had no effect on naming times or accuracy. There was no hint of an interaction between type of parse and how parts were arranged on the page. This pattern of effects supports theories that suggest that objects typically are recognized without being parsed into parts. The findings are in agreement with theories suggesting that object features (not specifically related to parts) are matched directly with such features stored in long-term memory, with the constraint that the features of a single object are seen from a single viewpoint.