Abstract
Representing the orientation of objects in the visual field is important for a variety of reasons. Perceiving the direction in which predators are facing is a life-or-death matter for many creatures, and a person reaching for an object must apprehend how that object is oriented in order to position her hand appropriately. Orientation is also important in processing visual symbols, such as arrows on street signs, or alphabetic letters (e.g., b, d). Despite the significance of orientation information, little attention has been directed toward questions concerning how the orientation of objects is represented in the visual system. We present a theoretical perspective that conceives of orientation as a relationship between reference frames. According to this theory, orientation representations relate an object-centered reference frame to a frame of reference external to the object, and this external reference frame may in turn be related to additional external frames. The theory also offers specific assumptions about how relationships between reference frames are represented, proposing that these representations consist of several components that together specify a mapping between coordinate systems. Given these assumptions, the theory makes predictions about the types of errors that should occur in tasks involving perception and memory for object orientation. For example, the assumption that object-centered frames are implicated in orientation representation predicts errors in the form of reflections across an object axis. We report two experiments designed to test the theory's predictions. Pictures of objects and artificial shapes were presented briefly, and participants reported the orientation of the stimuli. We argue that the observed error patterns support the theory's assumptions about the componential structure of orientation representations, as well as the assumption that object-centered reference frames play a critical role in orientation representation. We also consider alternative hypotheses, suggesting that these hypotheses cannot readily account for our results.