Abstract
Previous research on peripheral recognition has largely concerned artificial stimuli (e.g., letters or gratings) shown in isolation or in an uninformative context (e.g. surrounded by flankers). Under these conditions, recognition abilities decline rapidly with eccentricity due to crowding and a decrease in visual acuity. Under typical viewing conditions, however, the larger contextual scene in which an object appears may carry high levels of information about the identity of the object, particularly if the environment is familiar. This contextual information may serve to improve recognition ability in the periphery. To examine this potential role of context and familiarity, we tested participants’ ability to identify pictures of objects across the visual field. A wide variety of objects were used as target objects. On each trial, participants maintained fixation on a point while a photographed object in the periphery was cued by a flashing dot. The task was to identify the cued object. Fixation was moved progressively closer to the target until it could be successfully recognized. Using this technique, we compared performance when the object was shown within its entire contextual scene vs. when the image of the object was isolated. In addition, we compared performance when the contextual scene was familiar to the participants vs. when it was unfamiliar. The same set of objects, shown at the same eccentricities, was used across the three conditions (Familiar Context, Unfamiliar Context, No Context). Results showed an effect of contextual condition, with successful recognition in the Familiar Context condition at dramatically higher eccentricities than in the Unfamiliar Context condition (~20 degrees), which, in turn yielded higher eccentricities than in the No Context condition. Overall, these findings demonstrate that contextual information—and particularly a familiar context—allows for recognition of objects even in the far periphery, despite the constraints imposed by crowding and lowered acuity.
Meeting abstract presented at VSS 2015