International Graduate School of Neuroscience

Research Interests

Our research focuses on identifying the neural mechanisms that integrate sensory signals and prior expectations in cognition. We aim to understand how the human brain combines prior expectations with incoming sensory information, adapts to changing contexts, learns from experiences, and generalizes prior knowledge to new sensory inputs under uncertainty. To explore these questions, we combine computational modeling based on deep neural nets and large language models with behavioral and neuroimaging techniques, including eye tracking, pupillometry, EEG, and structural and functional MRI (s/fMRI).

Predictive processing, particularly predictive coding, offers a potential framework for understanding cognition in context. By testing these theories against alternative models in areas like learning, perception, and memory, my research seeks to uncover the neuro-computational mechanisms underlying human cognition, specifically in speech and face perception. Additionally, our work explores individual differences in the weighting of prior expectations versus new sensory information, with the goal of applying our findings to clinical populations.

Overall, our research supports the notion that the brain actively learns and applies prior knowledge from experience and context to facilitate successful perception under uncertainty in social interactions.

Current Research Projects

Serial dependence

Our perception of sounds is shaped not only by what we hear in the moment, but also by what we’ve heard before. In vision science, this influence of past experiences is known as serial dependence, where previous stimuli subtly bias our perception of new ones. Our research examines how this phenomenon plays out in hearing. By altering the voice and speech characteristics in sound sequences with varying consistency, we uncover two coexisting effects: a repulsive influence from previous stimuli and an attractive influence from previous decisions—each shaping how we interpret incoming auditory information.

How do priors shape the perception of speech?

Speech recognition isn’t just about the clarity of what we hear—it’s also shaped by what we expect to hear. In noisy environments, different listeners can interpret the exact same speech in completely different ways. At our lab, we explore how expectations—shaped by factors like our knowledge of the speaker—affect how speech is processed and represented at various levels of the brain’s speech processing system.

How is face perception influenced by context?

Our ability to recognize faces is significantly influenced by context information. Predictive processing theories suggest that our brain uses context to guide where we look for sensory evidence. By using eye tracking we can show that expectations shape face processing by guiding early eye movements to anticipated locations, supporting predictive processing theories. To understand the brain mechanisms behind this, we combine deep neural network modeling with fMRI analysis. Our work explores how the brain integrates prior expectations with sensory input—either by responding strongly to the unexpected (prediction error), or by enhancing expected features through sharpened neural representations.