Abstract
Observers could categorize objects in natural scenes in 30 ms (Schyns & Oliva, 1994, Psychological Science). Yet, Davenport and Potter (2004, Psychological Science) had found that participants were more accurate in identifying objects without scene background than with background after viewing the scenes for 80 ms. We postulated that such discrepancy might be a result of a difference in the spatial frequency content of the background image. We investigated how object identification performance changed with spatial frequency spectrum of the image and viewing time.
Photos of 24 possible target objects were presented either alone or on one of 13 possible background of natural scenes. The targets and the background were thematically consistent. The spatial frequency of objects and scene backgrounds were processed by either low-pass or high-pass filters with six different cutoff frequencies (1, 3, 5, 7, 9, and 11cpd). The viewing duration was either 36 ms, 100 ms or until response. The task of the observer was to name the target object. The identification accuracy increased with information content in all background and viewing durations. The presence of scene background decreased the identification accuracy for all cut-off spatial frequencies with short viewing durations but increased accuracy for low cut-off frequencies with long viewing durations. This result is consistent with the hypothesis that the background effect on target identification is influenced by the richness of information in the images.
Supported by NSC 96-2815-c-002-072-H to CFC, NSC 96-2413-H-002-006-MY3 to CCC and NSC 96-2413-H-002-007-My3 to YYY.