Research
Research
My curiosity for neurolinguisitcs starts from vision, the very first system neuroscience lesson: we seem to understand a lot about how light becomes visual perception, step by step. Can we achieve a similarly explicit account of how language is implemented in the brain? I am deeply driven by how the hierarchically organized and distributed components of the language network encode and integrate information, and how these processes give rise to our rich linguistic experience.
At Stanford, I am involved in two complementary lines of projects addressing this problem. One line draws on linguistic theory and models of sentence processing to investigate how linguitic information is represented spatially and temporally in observable behavior and neural activity. Another line explores how linguistic representations relate to the neural processes involved in speech planning and motor control.
I also enjoy exploreing differnt analysis methods that transform noisy data into intepretable results, building reusable pipelines, and building thoughtful tasks that can elicit meaningful behavioral and neural responses. This is what I keep learning from lab experiences across differnt domines.
01
Online naturalistic sentence processing
Language arrives sequentially, yet listeners and readers rapidly build interpretations from ongoing input. I study this process using real-time behavior from self-paced reading and time-resolved microelectrode array recordings during naturalistic sentence processing. My work draws on surprisal, noisy-channel inference, and continuous-time modeling to ask how linguistic features shape processing as a sentence unfolds.
This work contributes to the incremental-processing literature by linking continuously varying linguistic information to moment-by-moment behavioral and neural responses. I am especially interested in when lexical, syntactic, and semantic representations emerge, how long they persist, and how information propagates between brain regions rather than appearing only as an average response within one region.
02
Mapping the language–speech interface
My fMRI work examines how distributed language representations interface with the sensory and motor systems that support speech. In the INT project and related collaboration with Stanford’s Neural Prosthetics Translational Lab, I work with intensive deep-scan fMRI data and language–speech interface localizers to characterize this organization at the individual level.
I use representational similarity analysis to test what information multivariate activity patterns encode and generalized psychophysiological interaction analysis to examine how coupling among regions changes with task context. Together, RSA and gPPI move the question beyond where activity occurs: they ask what a neural pattern represents and how that representation is coordinated across a network. This framework can inform both basic theories of language organization and the design of speech-restoration systems.
03
Experience, proficiency, and sensory grounding
My emerging research asks how language representations change with experience. I am motivated to examine how proficiency and bilingual language use reshape the timing, geometry, and network organization of comprehension. A second asks how vision, audition, action, and other sensory signals constrain linguistic predictions and ground meaning in the physical world.
Questions guiding my work
- How does neural activity become linguistic representations? I want to connect population-level dynamics with the representations experienced as words, sentences, and meaning.
- How are temporally partial representations maintained and updated? Natural language requires the brain to preserve earlier information while continually integrating what arrives next.
- How does language interact with perception, motor plans, and thought? I am interested in the interfaces that connect abstract linguistic knowledge with sensory evidence, motor plans, and world knowledge.
- How does experience reorganize the system? Bilingualism and proficiency provide a powerful lens on which aspects of language processing are stable and which remain plastic.