IMPRS DOCTORAL DEFENCE: Filiz Tezcan Semerci

Understanding spoken language requires transforming continuous sound into structured linguistic representations, drawing on learned knowledge and cognitive resources. This thesis investigates how neural responses are modulated across levels of language processing from acoustic perception to syntactic structure using Temporal Response Function analysis of magnetoencephalography data. The findings reveal that linguistic structure enhances phoneme-level tracking, highlighting the brain's reliance on abstract linguistic representations. Even in uncomprehended languages that listeners have statistical exposure to, phoneme tracking is strengthened, suggesting that exposure alone can refine neural encoding. Conversely, hearing one's native language suppresses early acoustic edge tracking, indicating that prior knowledge influences even low-level sensory processing. The thesis further disentangles the roles of lexical-statistical patterns and syntactic structure in neural tracking using controlled stimuli with frequency tagging and computational modeling. Both information sources independently contribute to linguistic processing. Increasing complexity recruits broader neural networks, supporting the involvement of widespread coordination in hierarchical language processing. A novel cross-frequency coupled oscillator framework demonstrates that syntactic tracking may emerge from synchronized neural activity aligned with hierarchical linguistic units. Together, these findings support a hierarchical structure-building account of comprehension, in which the brain integrates statistical cues and syntactic rules to construct meaningful representations from speech.

Filiz Tezcan obtained a bachelor's degree in Chemical Engineering from Boğaziçi niversity, Turkey, in 2007 and a master's degree in Cognitive Science from Yeditepe University, Turkey, in 2019. She started her PhD in the Language and Computation in Neural systems research group at the Max Planck Institute for Psycholinguistics, where she examined the dynamic encoding of hierarchical linguistic structure using MEG, under the supervision of Andrea E. Martin. She is currently working as a postdoctoral researcher at the Temporal Dynamics Lab, Maastricht University, the Netherlands, where her research focuses on temporal information integration, oscillatory mechanisms in auditory and language processing, and the effects of transcranial alternating current stimulation on neural activity.