Representational similarity analysis of EEG reveals multiple spatiotemporal dynamics of selective attention

Selective attention allows listeners to follow a target speaker in complex scenes, but less is known about the neural mechanisms that control this capacity. We collected electroencephalography (EEG) during an auditory attention task encompassing spatial attention, talker-based attention, and passive listening. Time-resolved representational similarity analysis (RSA) of broadband scalp voltage and alpha-band oscillations revealed distinct representational trajectories in the two measurements: scalp voltage produced brief peaks in task discriminability shortly after the target cue and around 300 ms after target onset, whereas alpha-band power emerged more slowly and remained elevated throughout the trial. These patterns distinguish attentive from passive conditions and spatial from talker attention. Furthermore, broadband activity and alpha power encode complementary information, while other frequency bands contribute little independent explanatory power. These results demonstrate that RSA allows a unified interpretation of evoked and induced EEG measures, revealing how distinct neural computations jointly encode the spatiotemporal dynamics of selective attention.