Conceptual content in images triggers rapid shifts of covert attention

Abstract

The visual system can use conceptual information to search for targets even in the absence of clear featural signifiers1, and visual saccades are often directed at target objects defined by conceptual content2. These abilities are a core component of our facility with the visual world. Here, we evaluate whether contingent mechanisms of visual attention, known to trigger in response to target features such as motion, color or luminance3, are also triggered by visual patterns that match conceptually specified categories. These pre-registered experiments provide convergent behavioral and electrophysiological support that covert spatial attention is rapidly triggered by natural image exemplars from superordinate conceptually described target sets such as dinner food or four-legged animal, even when each target was viewed only once. In the behavioral experiment when two targets were presented with onsets separated by only 167ms, subjects reported the second target more often when it was in the same spatial location as the first. In the EEG experiment, images elicited clear N2pc and P3 components only when they matched the conceptually specified target set. The latency of the N2pc peaked at roughly 250ms, which is comparable to that commonly found in other N2pc studies for simpler stimulus types. These results suggest that vision quickly decodes conceptual information from natural images and selectively deploys spatial attention to locations containing information that matches current search goals.