Abstract
Selective attention is sensitive to statistical properties in the environment. For example, visual search is faster when targets or distractors appear more often in a particular color relative to less frequent colors. However, it is unknown whether these learned selection effects operate by proactively enhancing (or suppressing) specific features, or whether they rely on reactive processes during visual search itself. Here we tested this by having participants (N=200) perform a visual search task in which (unbeknownst to the participants) either the target (Exp. 1) or the distractor (Exp. 2) appeared more often in a particular color (the ‘rich’ color, 75% of trials). Participants searched for a target Landolt C (left or right gap) among seven distractor Cs (top or bottom gaps). On 18% of trials, instead of presenting the search array, a probe was briefly flashed at a random location (either in the frequent color, another search color, or neutral gray), and participants indicated whether they saw a colorful probe or not. In Exp. 1, participants responded faster to search targets in the high-probability, target-rich color (p<.001, ηp2=0.15); they also detected probes more accurately in the target-rich color than other colors (p=.02, Cohen’s d=0.24). In Exp. 2, participants were faster when the search distractors were in the distractor-rich color (p<.001, ηp2=.12) and less accurate at detecting the distractor-rich color during the probe task (p=.007, Cohen’s d=0.28). These results show that experience-driven attention can enhance and suppress feature representations prior to search onset, consistent with proactive mechanisms of selection. This contrasts with findings of goal-driven attention, where explicitly cued distractor features are not proactively suppressed (Addleman & Stoermer, 2021). Our results thus point to dissociable mechanisms of goal-driven and learned ignoring of distractor features.