Unheard Surprises: Attention-Dependent Neocortical Dynamics Following Unexpected Omissions Revealed by Intracranial EEG
Vegard Volehaugen1 (firstname.lastname@example.org), Sabine Liliana Leske1, Ingrid Funderud2, Anais Llorens3, Vinicius Rezende Carvalho1, Tor Endestad1, Anne-Kristin Solbakk1, Alejandro Omar Blenkmann1; 1University of Oslo, 2Oslo University Hospital, 3FEMTO-ST Institute
Neocortex may encode and relay sensory input as Bayesian surprisal, or prediction error (PE). By this view, unexpected sensory absence results in PEs that mirror prior expectations. Predictive processing is, however, multi-layered, and knowledge on how attentional states influence hierarchical integration of endogenously generated PE is limited. To address this gap, we used intracerebral field potentials to investigate the interplay between expectation and attention following auditory omissions. Methods: Stereo-electroencephalography (SEEG) was recorded from 20 patients with drug-resistant epilepsy undergoing presurgical evaluation. Sound sequences containing predictable and surprising omissions were played during attentive listening and a distraction task. Population-activity (HFBA, 65 - 250 Hz) was extracted from channels in auditory cortex (AC) along with temporal, cingulo-opercular, and frontoparietal cortices, Results: Responses to unexpected omissions depended on attentional state. When unattended, responses were primarily limited to AC, although notable modulations occurred in frontal operculum (Fop). In the attended state, responses were broadly distributed, with short-latency responses in AC and cingulo-opercular network (CON), followed by long-latency, sustained activity in frontoparietal and somatomotor networks (FPN/SMN). Conclusions: Macroscale neural dynamics induced by PEs are strongly determined by selective attention. Early integration in AC is followed by prominent modulations in Fop; the latter more dependent on attentional state than the former. In contrast, responses in FPN and SMN are fully determined by attention. In line with previous work, the results suggests a three-stage model of sensory deviance processing where PEs are monitored by the CON which implements network transitions based on estimated task relevance.
Topic Area: ATTENTION: Auditory
CNS Account Login
April 13–16 | 2024