Rapid and accurate processing of multiple objects in briefly presented scenes

Abstract

Humans can detect multiple objects in briefly presented natural visual scenes, but the mechanisms through which the objects are segmented from the background and consciously accessed remain open. By asking participants to report how many humans natural photos presented for 50 ms contain, we show that up to three items can be rapidly enumerated from natural scenes without compromising speed or accuracy. In contrast to standard parallel and serial models of object selection, our results revealed that the participants were fastest in enumerating two objects; even enumerating one single item required additional processing time. Also enumeration accuracy slightly increased in the subitizing range as number increased. Our results suggest that the visual system is tuned to process multiple items, which may underlie spatial and numerical cognition, and be beneficial in real-world situations that often require dealing with more than one object at a time.