Abstract
Previous research suggests that targets remain uncrowded when at least one feature distinguishes them from their flankers (Poder, 2006). While spatial properties of crowding have been studied in detail, the time course of crowding remains unknown. The present research studies the time course of crowding, using a new cueing paradigm. On each trial, observers were shown a ring of 20 Gabors oriented radially or tangentially to the ring; the ring radius was 10°. Each Gabor was flanked by 4 plaids containing both radial and tangential components. After a 1s preview, half the Gabors (randomly chosen) changed their orientation by 90° and one among those was then spatially cued with a variable delay (0 - 800 ms). Our observers' task was to report whether the cued target Gabor's new orientation was radial or tangential to the ring. Our results indicate that target identification slowly degrades with cue latency, suggesting that following its appearance the target Gabor is initially not crowded and that crowding onsets at approximately 1 d' / s. Such a slow onset is quite remarkable, and sets a constraint on possible mechanisms of crowding. For example, early vision suppression phenomena happen an order of magnitude faster (Petrov & McKee 2009). We hypothesize that while spatial attention alone has insufficient resolution to isolate the target, the target appearance creates a transient motion signal, which could be used by feature and spatial attention combined to isolate the target from its flankers. As the motion signal extinguishes, the enhanced attentional resolution due to motion popout is lost and crowding gradually sets in.
Meeting abstract presented at VSS 2014