Selection vs. integration task demands shape the similarity of information neural coding

Attention is a function that enables selection and integration of multiple sources of information. However, how these demands influence neural coding of information is not well understood. In this study we used EEG to examine how the selection vs. integration of stimuli shapes the content and geometry reflected on neural patterns, during both preparation and target processing. Participants performed a size judgement task in a cue-target paradigm that, depending on the block, required judging either the size of a selected item and ignoring the additional stimulus or integrating both items to respond. Decoding analyses showed that under selection demands, categorical templates of the cued stimulus were activated during preparation and target coding, contrasting with integration, where the cued category was active only during preparation. Notably, RSA suggested a specific exemplar encoding during its processing, that was sustained also across the post-stimulus window during selection, yet not under integration contexts. Our results also suggest that attentional demands shape the similarity between stimulus categories, by increasing the distance between selected stimuli and distractors or by increasing the similarity between to-be-integrated stimuli. Overall, this study uncovers the dynamics of stimulus encoding under selection and integration demands, offering crucial advances to understand how top-down processes shape information representation in the human brain.