Does viewpoint dependency effect influence scene consistency effect?
16.446, Friday, May 10, 5:30 - 8:00 pm, Orchid Ballroom
Kazuhiko Yokosawa1, Gergo Sastyin1, Ryosuke Niimi1; 1The University of Tokyo
We examined the effect of object viewpoint on perception of natural scenes comprising one object against a background. The consistency of objects and backgrounds greatly influences object recognition (Palmer, 1975; Hollingworth, & Henderson, 1998). Davenport and Potter (2004) also reported that recognition of an object facilitates perceptual processing of its background and vice versa: this is the scene consistency effect. This effect holds even for brief scene exposures. Based on these results, Davenport and Potter presented an interactive recognition model for simultaneous object and background processing. Object recognition is also strongly influenced by the viewpoint from which an object is observed. We hypothesized that the scene consistency effect does not occur for objects observed from an accidental viewpoint; instead, the latter should decrease efficiency of object recognition. We tested the scene consistency effect using canonical and accidental views of objects (i.e., canonical/accidental conditions). The semantic consistency of objects with backgrounds was also manipulated (consistent/inconsistent conditions). Participants viewed briefly presented scenes and named either objects (Experiment 1) or backgrounds (Experiment 2). Results of Experiment 1 showed a significant scene consistency effect, i.e., object naming accuracy was higher in the consistent condition. We also found a significant viewpoint effect (lower accuracy for accidental views). However, no interaction was observed between these two factors; regardless of object viewpoint, a scene consistency effect was always present during the object recognition task. Results of Experiment 2 replicated the scene consistency effect for the background naming task. However, the difference due to object viewpoints completely disappeared; object viewpoint showed no effect on the background naming accuracy. These asymmetrical results in object and background processing imply that scene perception is not always mutually interactive as Davenport and Potter proposed.