How long have you been at Stanford?
This is my 2nd year on faculty (I started in Fall 2024).
What types of things are you working on at Stanford, and in general?
I run a research lab called the Laboratory for Computation and Language in Minds and Brains (CLiMB Lab), where broadly we’re trying to figure out how we comprehend language so quickly and accurately. My group’s research falls into three overlapping buckets: psycholinguistics, computational linguistics, and language neuroscience. Beyond this, I teach, mentor student researchers at all levels (undergrad, grad, postbac, and postdoc), write grants, and do various kinds of service to the university (e.g., committees) and research community (e.g., peer reviewing).
Related to the last question, are you also teaching? What types of classes?
I teach a running lab course throughout the year that combines student presentations and guest lectures, and I also teach graduate and undergraduate introductions to psycholinguistics. For next year, I’m prepping a new course for grads and advanced undergrads on computational psycholinguistics.
Can you tell us about your work at CLiMB Lab?
Our mandate is to figure out how language comprehension works, and our approach to this covers three main streams. The first stream is language neuroscience, where we primarily use functional magnetic resonance imaging (fMRI) to map brain networks at the level of individuals, study how those networks specialize, and examine their contributions to language processing. A “sub-stream” in this space that I’m particularly excited about is a new collaboration with Stanford researchers Laura Gwilliams, Frank Willett, and Jaimie Henderson to apply these mapping techniques to the brains of people with amyotrophic lateral sclerosis (ALS) who have lost the ability to speak due to paralysis. The imaging helps us pick surgical implantation sites for microelectrode arrays that are then used to decode what the person is attempting to say, in effect restoring speech. For us, this is a special opportunity both to translate what we’re learning about brain function into practical benefits for patients, and to learn more about the cortical microcircuitry for language based on the recordings from these arrays.
The second stream is computational psycholinguistics, where we mainly test theories about the moment-by-moment mental processes that support language comprehension against behavioral measures of human comprehension effort (e.g., word-by-word reading times). This work so far has revealed that human allocation of effort during comprehension is largely consistent with “optimal” efficient processing, but that there are interesting edge cases where optimality makes no predictions or incorrect predictions about humans, offering us a chance to unpack what might be going on under the hood.
The third stream is computational linguistics, where we’re primarily focused on understanding the hidden representations and processing algorithms that are learned and used by statistical language models (including AI models like ChatGPT). This is not straightforward, which may seem odd given that these systems are engineered by humans. But while we can fully read out the numbers produced by the models and how those numbers were computed, there’s no engineered relationship between these numbers and things like “part of speech” or “negative polarity condition”. The success of the models depends in part on the fact that they can through training learn to use their numerical computations to do more abstract linguistic and conceptual work, and we want to understand how this happens. To do this, we use a combination of causal interventions (can we change the weights or activations in the model to produce behaviors consistent with having a linguistic representation of interest?) and neuroscience-inspired connectivity analyses. Our interest in this stream comes from the fact that these models provide a new window into both the natural organization of human language and the potential role of statistical learning in shaping human language acquisition and processing.
How do you like working at Stanford and living there?
I feel very lucky to have landed here. The intellectual community at Stanford is wonderful, with vibrant world-class groups in all the disciplines of relevance to my lab (linguistics, natural language processing, neuroscience, and psychology). The students impress and inspire me, and they make mentorship my favorite part of my job. The campus is beautiful, and I love being able to jog/hike/bike with my kids year-round. I also enjoy being part of a large cohort of 4 (!) junior faculty all hired in linguistics in quick succession, starting with OSU’s own Nandi Sims, followed by me, Robert Hawkins, and Meg Cychosz. Linguistics is a small field and it’s not always a given that early career faculty hires will have any “friends” in the department at their life and career stage. I feel very lucky to have three.
Do you have anything that you like to do or be involved in when you are not working on linguistics?
Boring answer, but at this stage in my life it’s pretty much just work and family. I have 4 kids aged 6-15 and my amazing spouse Rachel is an emergency physician, so we have a lot going on as parents and we’re often pinch-hitting for each other. I dabble unseriously in various pastimes for fun when possible, including cooking, running, hiking, biking, and reading. I’m currently working my way very unsuccessfully through the video game Returnal.
Is there anything else you’d like to share?
I think that covers it! I’m delighted to have had this chance to reconnect with my alma mater. If you’d like a deeper dive into our work, check out the lab website and feel free to reach out to me directly. I’m a proud Buckeye and always eager to talk to folks from OSU Linguistics.