Abstract
How do we recognize objects? This simple everyday task has befuddled scientists for centuries. Research has led to a two-stage theory. First, visual features, such as color and orientation, are detected. Next, these features are combined. Here we ask which features are combined. Crowding is excessive feature combination that occurs when a target object is flanked by other objects, making the target impossible to identify. Exploiting the Gestalt laws, we manipulate proximity, similarity, closure, and regularity (a component of continuity) in a crowding paradigm. We find that decreasing proximity, decreasing similarity, breaking closure, and decreasing regularity each dramatically reduces crowding (or feature combination). The Gestalt laws are simple principles of visual organization that describe how objects group with each other. We find that they also specify which features are passed from the detection stage to the combination stage of object recognition. We conclude that this process is a distinct grouping stage. There are not two, but three stages in object recognition.