Abstract
Rapid object categorization is a crucial ability for animals to react and adapt their behavior to an unexpected situation. However a relevant context can help object recognition, particularly when objects are small or partially masked: several studies in both human and monkey showed that rapid visual categorization performances are improved by the categorical congruence between objects and their embedding contexts (Bar 2004, Fize et al. 2011, Joubert et al. 2007). The present study first aimed at determining whether contextual priming could also facilitate object recognition, and second whether this influence was better depicted in terms of background features accumulation or context categorical processing. Three rhesus monkeys already experts in rapid visual categorization performed an Animal vs Non-animal go/no-go task. Contextual primes consisted in Rapid Serial Visual Presentations (RSVP) of nine pictures of either natural or manmade real world scenes, randomly varying in proportion and position within the RSVP flow. These background contexts were followed, after a short 100ms delay by a flashed test stimulus containing either an object or animal isolated on a gray background. Performances were analyzed based on the 6800, 6100 and 5800 trials that monkeys respectively performed, in terms of reaction times, accuracy and d-prime as function of natural contexts within the RSVP flow. Monkeys succeeded in this task (mean accuracy: 94.1%; mean RT=329ms), and results showed that contextual priming using an RSVP flow significantly influenced object categorization. Natural primes improved significantly Animal detectability and increased response speed. Importantly, an increasing number of natural scenes in the RSVP flow improved more generally d-prime scores in a linear way, providing further evidence for features accumulation mechanisms that integrate background and object information.
Meeting abstract presented at VSS 2013