Anton Lukashevich¹ (), Sabrina Hansman-Roth¹; ¹University of Iceland

Previous Feature Distribution Learning (FDL) paradigms demonstrated that observers implicitly represent distractor feature distributions to optimize visual perception and behavior. These findings furthermore, reveal that internal representation are much richer than merely encoding summary statistics. However, the neural mechanisms underlying the encoding of feature probabilities remain unclear. Here, we investigated whether encoded distractor probabilities modulate the N2pc event-related potential (ERP) component, an index of attentional selection, focusing on amplitude differences in a modified FDL task. Observers (N=5 in pilot) viewed arrays of colored diamonds (36) and searched for an oddly-colored target, reporting the location of the target in lower or upper half of the set. During learning trials distractor distributions were kept constant to ensure sufficient learning of the distractor probabilities. Subsequent test trials introduced role reversals: previous distractor features became the target, with its probability varying. EEG data were recorded via a 32-channel system, with epochs time-locked to stimulus onset. Behavioral results validated that observers implicitly encoded the distractor probabilities: search times for targets closer to the mean of the previous distractor distribution were slower than search times for targets further away from the mean of the previous distractor distribution. Visual inspection of EEG data indicated N2pc amplitude differences tied to distractor probabilities: attenuated amplitudes for high-probability reversals contralaterally, implying stronger suppression, versus larger amplitudes for low-probability ones, suggesting reduced filtering. Latency effects were subtle, meriting expanded testing. These preliminary findings bolster probabilistic encoding in vision, linking the learning of feature probabilities to neural signatures. Ongoing work will quantify latency shifts and scale up sample of participants. Work is supported by the Icelandic Centre for Research (Rannís): 239774-052 to SHR.