Yasemin Gokcen

My research interests are at the crossroads of neuroscience, linguistics, and computational methodology. I am most interested in studying neural mechanisms of cognition and take inspiration from neurobiology and artificial intelligence techniques.

Prediction is fundamental to language, as humans use preceding context to anticipate upcoming words (Levy, 2008; Ryskin & Nieuwland, 2023). This process requires a memory system that is both selective and adaptive. We draw from prefrontal cortex models, where biologically plausible gating mechanisms actively maintain and update task-relevant information, improving cognitive flexibility and working memory (O’Reilly & Frank, 2006; Kriete et al., 2013). Here, we investigate how such gating mechanisms support real-time language prediction. Using EEG data from a naturalistic story-listening task, we first replicate findings that unpredictable words (high surprisal) elicit larger N400 effects (Frank et al., 2015). To examine gating’s role in prediction, we compare language models with and without gating, showing that this difference reflects word-by-word working memory demand. Our results reveal that this gating metric correlates with EEG amplitude in later time windows after word onset. This suggests that gating mechanisms may play a key role in dynamically managing context for prediction during real-time language comprehension.