Search Abstracts | Symposia | Slide Sessions | Poster Sessions
Psychosis Proneness Associates with Faulty Predictions across Linguistic Levels: a Naturalistic EEG Study
Poster Session C, Thursday, October 1, 10:45 am - 12:45 pm, Wangari Maathai
Suong Welp1,2, Yifei He3, Anna Tumanyants3, Lars Meyer1,2; 1Clinic for Phoniatrics and Pediatric Audiology, University Hospital Münster, DE, 2Language Cycles Research Group, Max Planck Institute for Human Cognitive and Brain Sciences, DE, 3Translational Neuroimaging Group, Philipps-Universität Marburg, DE
It has been suggested that language deficits in Schizophrenia arise from an imbalance between top-down semantic predictions and bottom-up acoustic-phonetic prediction errors. Here we present electroencephalography (EEG) evidence in favor of this hypothesis. Our naturalistic EEG study investigated whether psychosis-proneness in healthy subjects modulate the interaction of top-down semantic predictions and bottom-up acoustic-phonetic prediction error; we also present preliminary clinical data. Methods. Twenty-eight healthy participants listened to a 15-minute German story. EEG was analyzed using temporal response functions (TRFs). Semantic predictability was modeled through word entropy whereas acoustic-phonetic prediction errors where modeled through phoneme surprisal. TRF amplitudes were extracted in theoretically motivated time windows (N400: 300–500ms; beta-band power: 300–800ms) from frontocentral and left-temporal electrodes. Psychosis-proneness was indexed by the Peters Delusions Inventory (PDI) and Launay–Slade Hallucination Scale (LSHS). Results. At the group level, the interaction of word entropy and phoneme surprisal revealed an N400-like TRF (peak around 304ms, frontocentral distribution), reflecting modulation of phoneme-level prediction errors by prior semantic certainty, alongside a beta-band component (peak 600ms, left-temporal distribution). N400 amplitudes correlated positively with PDI (r=.411, p<.05), such that higher delusion-proneness was associated with reduced N400 attenuation under low entropy and high phoneme surprisal. This tentative result is consistent with relatively stronger top-down weighting that overrides sensory evidence. A trend into the same direction was observed for LSHS (r=.245, p=.124). Beta-band amplitude correlated with both PDI (r=.500, p<.01) and LSHS (r=.421, p<.05). Neither word entropy nor phoneme surprisal alone predicted PDI or LSHS, suggesting psychosis-proneness associates with the cross-level interaction specifically. Pilot data from 11 patients with schizophrenia spectrum disorders reveal more heterogeneous TRF patterns, supporting the hypothesis that cross-level prediction dynamics differ across clinical subgroups defined by hallucination status. Ongoing data collection will allow formal subgroup comparisons. Conclusion. The neural interaction of semantic predictions and phonetic prediction errors, indexed by naturalistic TRF components in the N400 and beta bands, is associated with subclinical psychosis-proneness and shows preliminary divergence in schizophrenia, supporting further investigation of cross-level predictive coding mechanisms in psychosis.
Topic Areas: Speech Perception,