Dynamic neural representations of auditory selective attention
Abigail Noyce1 (email@example.com), Wenkang An2, Weizhe Guo1,3, Yuhang Li1, Barbara Shinn-Cunningham1; 1Carnegie Mellon University, 2Boston Children's Hospital, 3Johns Hopkins University
Representational similarity analysis (RSA) has been used to characterize where and when different stimulus features or categories are represented in the brain. We extended this approach to investigate executive control processes - namely, auditory selective attention - by analyzing representations within electroencephalography (EEG) and functional magnetic resonance imaging (fMRI) data. Young adults listened for a target syllable among distractor syllables while we recorded EEG (N=30) or fMRI (N=19). The rich (21 conditions) design included spatial attention, nonspatial attention, and passive listening. In EEG, we extracted representational features at each time point during a trial using the topography of either the raw voltage timecourse or cognitively-relevant frequency bands (e.g. alpha). In fMRI, we extracted representational features using a searchlight approach. In both cases, we estimated dissimilarity between each pair of conditions at one timepoint or one region via the validated classification accuracy of a linear support vector machine (trained separately for each subject, with leave-one-trial-out cross-validation) . Within EEG, we observed differences in the dynamics of neural representation between features. For example, when subjects were cued to use spatial versus non-spatial attention, we observed a transient representation of attention type in the raw voltage, but a sustained representation in alpha band power. Within fMRI, we also observed region-specific representations. For example, posterior superior temporal gyrus encoded the difference between active and passive listening, but not attention type, while regions along parietal sulcus robustly encoded the type of attention. Representational analysis allows new investigations of neural processes underlying executive control.
Topic Area: METHODS: Neuroimaging
CNS Account Login
April 13–16 | 2024