Unfolding suppression: statistical learning drives suppression through dynamic attentional states

How does statistical learning drive distractor suppression? A central debate in current research concerns whether the underlying mechanism operates proactively or reactively. To address this issue, we examined the temporal dynamics of distractor suppression across four visual search experiments. Participants were exposed to a probabilistic electric shock when either a neutral distractor (Experiments 1a, 1b) or a shock-associated distractor (Experiments 2a, 2b) appeared at a high-probability location with shock. Trial-averaged results revealed classic signatures of learned distractor suppression. In distractor-present trials, response was faster and the first saccade was less likely to land on the distractor when it appeared at the high-probability location. Conversely, in distractor-absent trials, response was slower and the first saccades were less likely to be directed toward the target when it appeared at the previously high-probability distractor location. Furthermore, SMART analyses showed that attentional suppression dynamically shifted among distractor-driven capture, unguided search, and target-directed guidance as a function of spatial probability rather than threat association. The low-probability location showed a transition from distractor-driven capture toward unguided search, whereas the high-probability location exhibited a gradual shift toward target-directed guidance as learning accumulated. Overall, our findings reveal that learned distractor suppression is a dynamic process in which statistical learning progressively reshapes attentional guidance over time, providing new insights into how experience-dependent regularities modify attentional priority.