On the frontlines of speech science with Liberty Hamilton and Alex Huth

April 20, 2026

by Alexander Rony

Last summer, UC Berkeley hired not one but two computational neuroscientists, Liberty Hamilton and Alex Huth, who study how the brain processes language. The pair has a lot more in common. Both received their Ph.D. at Berkeley before teaching at the University of Texas at Austin. Both now have joint appointments in the neuroscience and statistics departments. They’re also married.

UC Berkeley writer Alexander Rony spoke with Hamilton and Huth about their research and their return to campus.

How would you describe your research?

Alex Huth: I study how our brains process language. We do this by having people listen to stories, then we use cutting-edge AI tools to build computational models of how different parts of the brain understand language. We map out which parts of the brain do different kinds of processing and which parts represent different kinds of information, then we read out information from the brain.

For example, can we decode words that you’re hearing from your brain activity? That is something that could be really useful for people who have communication disorders like aphasia, where they’re unable to take their ideas and express them in words.

Liberty Hamilton: I am also interested in how our brains are able to turn sound into meaning — from the pitch of someone's voice to who is speaking to the words they’re saying — and how that develops from an early age through young adulthood. We are also trying to understand how brain activity is different when people are speaking versus listening and how the brain tracks what someone is paying attention to.

Separate headshots of a woman and a man smiling.

LIBERTY HAMILTON AND ALEX HUTH STARTED FACULTY POSITIONS AT UC BERKELEY IN JULY 2025

I primarily work with patients with severe epilepsy. They are unable to treat their seizures with medication or diet, so they go in for brain surgery, where electrodes are placed into their brain for approximately a week. Clinical teams look for areas of the brain they might be able to remove to treat seizures without affecting the person’s ability to speak, so we’re mapping out those areas that are important to maintain.

What motivates your patient-volunteers? It sounds like fairly involved research.

Hamilton: Patients and families are really generous with their time. Many are motivated by the thought that they may help someone else in the future.

Since our research happens while these patients are in the hospital, we need to be sensitive to that. It’s really about meeting them where they’re at and finding things that are fun for them to break up the time in the hospital. I make sure that people on my team have high emotional intelligence and know how to read the room. We come in, watch a few movies, play a game, and develop a rapport. With adults, you can do listening experiments, and people will do it because they know it’s for research. Adults are used to doing boring things. Kids don’t want to do that. We have found that if we use movie clips from Disney or Pixar movies, those are much more enjoyable, and it doesn't feel like doing a research task.

Could your research help those who have deafness or other auditory impairments?

Hamilton: One of the things that people complain about, especially in aging, is the inability to understand speech in noisy environments. Even if someone has “normal” hearing, if they go into a restaurant, it’s difficult to follow a conversation. My lab has been looking into the neural processes and where those break down. We are seeing that certain brain areas seem to have a preference for speech that gets stronger as you get older, at least up through young adulthood. I imagine what may be happening is that the ability to separate speech from background sounds gets worse as people age.

Huth: We’ve been thinking about situations where people might not be able to produce speech, even though they have intact cognitive representations. There has been a lot of work recently on brain-computer interfaces. A bunch of people are trying to read out from the motor cortex how you would move your mouth, but that only works if you can actually activate your motor cortex correctly, and that’s pretty rare. We’re interested in stroke aphasia, but we’re also starting to investigate other populations of patients who have similar problems producing language, and one that’s really interesting is nonverbal autism.

A visualized view of a brain with linked data points and purple and green nodes

A SPATIOTEMPORAL MAP OF A BRAIN FROM THE HAMILTON LAB. COLORFUL DOTS REPRESENT ELECTRODES THAT WERE MORE REACTIVE DURING SPEAKING (PURPLE) AND LISTENING (GREEN). (CREDIT: KURTEFF ET AL.)

What drew you to your area of focus?

Hamilton: I’m a musician — I play the piano — so I’m really interested in how our brains process sound and what happens when you’re learning a piece. I also was really interested in foreign languages; for a while, I thought about going into translation. It turned out that neuroscience is a really nice intersection of all of the things I’m interested in and has real-world impacts.

Huth: I started out being interested in AI in the early 2000s at a time when AI was not a hot research area. I was interested in what it means to think and be conscious. I got into neuroscience to study these questions and learn how humans do these things in order to someday build better machines that can do them. Now, the world has turned upside down, where we can’t really understand how humans do these things, but we can build machines that do them and use those machines to understand human brains, which is not the future that I envisioned as a college student, but it's been pretty cool.

Berkeley is not just about asking big fundamental questions, but also about having a positive social impact. Would you mind sharing your thoughts on how to keep your research from being implemented in ways that can harm people?

Huth: I think that’s hugely important, especially with work that can take brain signals and interpret them. This has clear implications for mental privacy. I’ve been involved in a study group with the Uniform Law Commission where they’re deciding whether they want to draft model laws that states could enact. I’ve talked to a lot of people who are very active in the world of neuro-rights and mental privacy advocacy.

There’s a spectrum of possible experiments that range from helping people with disabilities to getting something out of somebody’s brain when they don’t want you to. We’ve steered toward helping people, but when I talked to ethicists and lawyers, they pushed me to do more of the dubious experiments, which surprised me. Their rationale was that somebody’s going to, and people like them need to know what is possible and what are the risks to understand what legal limits should be in place. I kind of resonate with this idea, but we haven’t really pushed that angle, because it still freaks me out. It’s something I take seriously, though.

Hamilton: Well, you and Jerry [Tang, Huth’s postdoc] did the —

Huth: Yeah, we did some experiments asking if a person can control what’s being decoded from their brain. Can they decide to shut it off? And the answer is yes, but it’s pretty effortful.

You both earned your doctorates here. What originally attracted you to UC Berkeley?

Hamilton: We actually applied to grad school together. I personally was looking for a place with excellent auditory neuroscience and people I was excited to work with. I ended up working with Shaowen Bao (now at University of Arizona) on optogenetics in the auditory system, which at the time was a relatively new technique. It was so exciting to be able to see in real-time how different cell types could contribute to perception of sounds.

Huth: For me, what made me come to Berkeley was Jack Gallant, who ended up being my Ph.D. advisor. His office is down the hall. I was at Caltech for my undergrad and master's, and I had seen him give a revelatory talk.

You’re now back as faculty. How has the move been?

Language decoding from functional MRI

A VIDEO BY JERRY TANG ABOUT HIS RESEARCH WITH ALEX HUTH ON RECONSTRUCTING LANGUAGE FROM BRAIN RECORDINGS.

Hamilton: It’s been great. It’s been really fun to come back, reconnect with people, and meet brand new people. Now, we are in the neuroscience and statistics departments, and we’re having really interesting interactions across those departments and even in different departments, so that spirit of collaboration continues.

Our labs are up and running. Our space is really nice. I like that our labs are interacting a lot. We both moved with some grad students and postdocs who came with us from UT Austin. That was a big move for everybody, but we’ve tried to make it as smooth as we can.

Bob Knight has been super helpful in getting my lab up and running. He and I both do intracranial recordings. He’s helped me get collaborations established with Bay Area folks, and I've been in contact with his postdocs. It’s been really nice to be back and have that support here.

Huth: Yeah, it's been really fun. I lived in California all my life before moving to Texas a few years ago, so it feels like coming home. The lab is doing the work we want to do. I'm excited about new postdocs, grad students, and collaboration opportunities.

I’m jazzed to use this sophisticated, high-resolution instrument, the NextGen 7T MRI scanner, to see how the brain works, is organized, and represents information, which we couldn’t see using existing scanners. A machine as ridiculously complicated as that one is not that useful without a big set of people who know everything about how to use it, so we’re interacting heavily with David Feinberg, Joseph Vu, and their groups.

Hamilton: There are a number of groups tackling the area of speech and language from different angles, whether it’s the hardware, algorithms, non-invasive versus invasive, translational applications, or even animal models. Having all these people together is really helpful to build up these bigger initiatives.

How can philanthropists help advance your labs’ research?

Hamilton: For your typical NIH grant cycle, it’s difficult to make things line up to have both short-term and long-term projects at scale across multiple places.

Patient work is very distributed. I am collaborating with a hospital site at UCSF and two sites in Texas. I have some people who are working full-time for me in Houston, I have a graduate student who is in Austin, and I have my group here at UCSF. It’s difficult to have enough money at each site to support the staff that you need, as well as the overhead in administering grants. There’s a lot of logistics.

The other thing that is very helpful for philanthropy is for these long-term, developmental projects that are not guaranteed to produce a particular result. My lab is working with rare patient populations. We would like to get data about patients at various ages, but we don’t see that many of them per year, so it takes a long time. In addition, some of our work is with patients with specific causes for their epilepsy, such as tuberous sclerosis complex (TSC), so I hope to engage more with organizations who support people with TSC specifically.

Huth: Since what we do overlaps heavily with AI, our work evolves at a quick pace that is incompatible with many grants. We’re building on newly released computational models that didn't exist three months ago, and in six months, there will be something entirely new. That is incompatible with NIH award timescales, so that’s where philanthropy could be really helpful.

Topics

Biological Sciences topic page, Faculty topic page, Research & Innovation topic page