Fig. 1

Visualization of trial sequence and cues. a Each trial began with a random noise mask for 250 ms, and consisted of a video that lasted between 2034 and 5034 ms. In a given video, the hazard could appear between 1034 and 4000 ms after the video began. The portion of the video containing the hazard was always 1000 ms long, and was followed by a 250 ms random noise mask. Participants were instructed to use the left and right arrow keys to indicate the lateral location of the hazard, and could respond any time after the video started. b The video contained either no cue or one of three cues: a temporal cue (red bar at the bottom of the video), a spatiotemporal valid cue (expanding ring superimposed on the hazard, represented by the solid red circle and the larger dashed red circle) and a spatiotemporal invalid cue (expanding ring superimposed on a nonhazardous object in the scene). The no-cue, temporal cue, and spatiotemporal conditions were blocked, and block order was randomized across participants. Within the spatiotemporal cue block, valid and invalid spatiotemporal cues were randomly interleaved and appeared with equal frequency (50% cue validity). The expanding ring spatiotemporal cue was chosen using the procedure described in the “Cue selection” section