Search Abstracts | Symposia | Slide Sessions | Poster Sessions
Neural signatures of speech perception and production with and without affective prosody as revealed by intracranial sEEG
Poster Session C, Thursday, October 1, 10:45 am - 12:45 pm, Wangari Maathai
Christian A Kell1, Henrik Kabel1; 1Translational Neuroscience at Cooperative Brain Imaging Center, Goethe University, Frankfurt
Language processing engages distributed brain networks supporting both speech perception and production. However, speech is also shaped by paralinguistic phenomena like affective prosody which conveys emotional information via modulation of sensorimotor speech parameters. Prosody processing involves both parts of the core language regions and brain regions beyond this set of regions. The overlap between language and prosody processing regions suggests interactions between linguistic and paralinguistic phenomena that are only partly understood on the neural level. Stereotactic EEG offer the opportunity to examine cortical structure function relationships with high spatial and temporal resolution. Here we set out to investigate sentence-level prosody processing both during speech production and speech perception and to identify gradients of activation as defined by broadband gamma band responses. To this aim, we recorded stereotactic EEG (sEEG) data from 10 epilepsy patients requiring invasive mapping during an emotional prosody identification task (sentences intonated either neutrally, happily, sadly, or angrily) and an overt sentence reading task with different prosody (happy or neutral intonation). Broadband gamma activity (75-130 Hz) was used to identify task-related responses. Signals were bandpass filtered, and amplitude was extracted using the Hilbert transform; statistical significance was determined by using permutation-based cluster-level inference. We observed a distinct pattern of task-selective broadband gamma responses but also sites that activated both during sentence perception and production in the left and right temporal cortex. Shared perception- and production-related responses were found particularly in superior temporal electrodes in and near the auditory cortex, whereas task-selective responses were distributed along a dorsoventral axis in the temporal lobe: perception-selective electrodes were located more closely to the Sylvian fissue, while production-selective electrodes extended further into middle and inferior temporal cortex. During sentence production, a subset of production-responsive electrodes showed a stepwise increase in broadband gamma activity from covert reading to neutral overt speech and happy speech, suggesting increasing recruitment with higher sensorimotor and prosodic demands. In addition, response profiles during production followed a dorsal-to-ventral temporal organization, with superior temporal electrodes showing fast, high-amplitude gamma responses and middle-to-lower temporal sites showing broader and weaker responses. These results confirm that perception and production engage both distinct and shared temporal-lobe sites, arranged along a spatiotemporal dorsoventral gradient. The four different intonations during the perception task allowed us to decode affective prosody using a multivariate approach across electrodes and features. We examined which electrodes contributed most strongly to decoding performance. These electrodes only partially overlapped with electrodes showing significant broadband gamma responses during speech perception, confirming that univariate activation studies do not fully capture task-relevant information. Together, these results identify a graded organization of temporal cortex in which shared auditory–productive responses, task-selective sites, and emotion-informative signals only partially overlap. The superior spatial resolution of sEEG compared to fMRI or MEG revealed a more complex structure-function relationship, which suggests that integration of linguistic and paralinguistic information may not necessarily rely on individual cortical sites that activate both during linguistic and paralinguistic processing. Instead, such information integration may rather occur in networks of distributed brain regions.
Topic Areas: Prosody, Language Production