Abstract
Rapid categorization of natural scenes is often thought to involve high-level presetting of the visual system for diagnostic features. But real world situations often require extracting and processing specific physical features of multiple objects often belonging to different superordinate categories. Is the human visual system able to perform parallel object categorization, or does it use serial processing? Extracting feature conjunctions has generally been studied within single objects. Here we tested the ability of the visual system to detect in parallel everyday objects spatially distributed within photographs and belonging to different categories.
Human subjects performed a go/no-go visual categorization of natural scenes flashed for 26 ms. In a series of four successive tasks, subjects had to respond manually as fast and accurately as possible to the presence of (1) an animal, (2) a vehicle, (3) an animal AND a vehicle, (4) an animal OR a vehicle (not exclusive). Distractor stimuli were also complex natural scenes containing other object categories and, according to the task, animals or vehicles.
Accuracy (% correct) and reaction times analysis were performed on photographs that contained both an animal and a vehicle because they are seen as target in all tasks. Performance in single animal/vehicle categorization tasks was similar (79/80% correct; median RT: 452/477ms). The “OR” task was performed with a higher accuracy (90%) at no temporal cost implying that the visual system can indeed pre-activate and detect in parallel multiple features diagnostic to different object categories. In contrast, longer reaction times were observed in the “AND” task (523ms) with no accuracy cost, suggesting the need for additional decision processing possibly involving “visual search-like” serial mechanisms.
Additional interference effects between animal and vehicle processing considering their respective sizes and location in depth are also discussed within a neuronal competition framework.
Fondation pour la Recherche Médicale.