Abstract
People are quick to detect examples of real-world object categories in natural scenes, with evidence of target presence emerging in frontal event-related potentials (ERPs) as early as 150 ms after stimulus onset. This is far faster than predicted by classical attention theories, which suggest that naturalistic search should require the serial deployment of attention so that features characterizing objects can be bound together. Two proposals have been offered to solve this puzzle. One is that repeated experience might render the visual system sensitive to combinations of low-level features so that their combined presence can be detected without the need for spatial attention (eg. Evans & Treisman, 2006). This predicts that evidence of target detection might emerge in non-visual brain areas even before spatially-localized, attention-related mechanisms act on visual representations. The alternative is that naturalistic search relies on spatial mechanisms, but that global scene information constrains the search space and reduces the number of objects that need be resolved (eg. Wolfe et al., 2011). This predicts that spatially-localized object processing should be required for target detection, but that this may occur very quickly. Here we use ERPs to differentiate between these possibilities. In Experiment 1 we find that discrimination of a naturalistic target generates an early ERP effect associated with the suppression of salient distractors. In Experiment 2 we show that this early distractor positivity (Pd) is larger when participants must confirm that the scene does not contain a target. Critically, sensitivity to target presence in this spatial mechanism emerges 39 ms before evidence of target presence appears in frontal ERPs. Naturalistic search for heterogenous targets therefore appears to rely on spatial mechanisms, as predicted by classic attention theory, but these mechanisms act very quickly. This is consistent with the idea that global scene characteristics constrain the search space.