Comprehension is rarely passive. In real life, we listen because we intend to respond, to answer a question, continue a conversation, or signal understanding. Yet the overwhelming majority of studies ask participants to do precisely nothing: sit still, listen to a story, answer a few questions at the end. Whether this unnatural setup distorts the neural signature of comprehension is almost entirely unknown.
The comprehension-by-production framework (Pickering & Gambi, 2018) holds that understanding language and preparing to produce it are not sequential steps but deeply intertwined processes. They claim that the motor system is active during listening, and anticipating a response shapes how incoming speech is encoded, and that the coupling between perception and production is part of what makes comprehension efficient. If that is true, then a listener who knows they will have to speak their answer should track speech differently from one who will type it, and both should differ from one who will do nothing at all.
This project will build a new dataset in which participants listen to continuous speech across experimental conditions that systematically vary what is asked of them afterwards: e.g., passive listening with no response, listening followed by typing a fixed motor response (e.g., always same word or last word listened) or a comprehension answers, and listening followed by spoken verbal answers (or repeated last word). You will help develop a design to cleanly separate motoric preparation from linguistic production effects on the incoming tracking signal.
You will join an interdisciplinary research team composed of the principal investigator, three postdoctoral researchers, six PhD students, and two research assistants. The team meets weekly to discuss exciting ongoing work from within and outside the team, which is also an opportunity for Master’s students to experience research life and learn about the latest advances in brain and generative AI research (especially speech, language, and music processing).
Lab website: https://diliberg.net