Abstract
Objects often appear in predictable locations in the environment. Theoretically, this contextual information can facilitate recognition when the object is unrecognizable based on its own features, for example, under conditions of poor illumination, occlusion or peripheral viewing. Across three experiments, we assessed how location information contributes to the identification of an object whose image has been degraded (blurred), making its identity ambiguous. We also considered how location information interacts with an ‘internal’ object feature (color) during identification. First, subjects performed a visual search task, in which they had to determine whether a novel target object was present in a rendered 3D bedroom scene. This served to train subjects on the features and locations of the target objects. Then, subjects were briefly shown a blurred image of the scene with a single ambiguous target object within it. The subjects’ task was to pick which object was in the scene from a lineup of the target objects. In Experiment 1, some of the target objects had fixed locations within the scene. We found that subjects used this location information during search and later to identify the blurred target objects. In Experiment 2, both the location and color of each object was variable but statistically predictive of the object’s identity; i.e. each object appeared most often (but not always) in a particular location and was most frequently (but not always) a particular color. We found that subjects used both sources of information—color and location— equally when identifying the blurred image of the object. In Experiment 3, one property (location or color) was fixed while the other was variable. We found that fixed location was given higher priority than fixed color in identifying the object. Overall, these findings show that people use location information to identify objects when the objects' intrinsic features are ambiguous.
Meeting abstract presented at VSS 2012