Frequency-tagged fMRI reveals dual mechanisms of attentional enhancement and suppression

Selective attention biases competition between concurrent visual inputs, yet its computational implementation in the human cortex remains unclear. Using frequency-tagged fMRI (ft-fMRI), we resolve how attention shapes the spatiotemporal dynamics of competing stimulus-driven responses at fine spatial scales. Distinct cortical populations exhibit either in-phase synchronization (target enhancement) or anti-phase synchronization (surround suppression) with attended stimuli. Population receptive field mapping reveals a target–surround organization, with in-phase populations biased toward the attended target and anti-phase populations biased toward the surrounding visual field. Directed attention uniformly amplifies surround suppression across the visual hierarchy, consistent with a distributed suppressive field, whereas modulation of target-related activity increases hierarchically from V1 to hV4. Critically, modulation of surround suppression, but not target enhancement, predicts behavioral performance. These findings dissociate the mechanisms of target enhancement and surround suppression and reveal the computational architecture underlying biased competition in the human cortex, converging with recent primate electrophysiology research. Rafeh et al. use frequency-tagged fMRI to demonstrate that selective attention relies on dual, independent neural mechanisms for target enhancement and surround suppression, elaborating on traditional unitary models of human visual attention.