Events
Prediction During Language Comprehension Impacts Downstream Memory
Ryan Hubbard
University at Albany, SUNY
Tuesday, March 17, 2026
12:30 p.m.–2 p.m.
Humanities Center, Conference Room D
During language comprehension, individuals can use contextual information to generate predictions about upcoming words. Numerous studies have examined how prediction impacts processing in real-time, but fewer studies have examined the downstream consequences of prediction during reading. My work examines the impact of prediction on later memory for words and sentences that were read. Additionally, recent work has identified a robust phenomenon: when participants read sentences in which the prediction is violated (e.g., “Father carved the turkey with a smile”), and are later tested on their memory, they often falsely remember seeing the predictable word (“knife”). One hypothesized cause of this phenomenon is that features of the predictable word are pre-activated prior to its occurrence, leading to encoding of the word into memory. I will discuss recent work conducted to test this hypothesis, in which I used representational similarity analysis of EEG data to examine neural pre-activation of the predictable word during reading, and if this neural pre-activation is related to later false memory. This research provides compelling evidence that predictions impact downstream memory, and that false memory for predictable words is likely due to encoding of pre-activated features into long-term memory.