Dr. Lea Duncker
Assistant Professor, Columbia University
Postdoctoral fellow, Stanford University
PhD, University College London
Dr. Lea Duncker spent much of her graduate career trying to understand an unexpected result. She wanted to build a computational model explaining how neural activity influences how animals move, moment by moment. She was working with data from collaborators where they used optogenetics to excite neurons in the primary motor cortex in monkeys, expecting the perturbation to disrupt ongoing movements. Yet the monkeys’ arm movements changed remarkably little in response to the optogenetic stimulation, despite dramatic changes in neural activity. This created a tantalizing puzzle: if the loudest signals in neural populations had so little effect on behavior, then what patterns of activity actually mattered?
Resolving that puzzle would occupy Lea for years, and in the process, transform how she thought about neural computation. Today, as an Assistant Professor of Neuroscience at Columbia University and Principal Investigator at the Zuckerman Institute's Center for Theoretical Neuroscience, Lea develops statistical methods and theoretical models to understand how neural populations support learning and behavioral flexibility, while working closely with experimentalists to test those ideas against biological reality.
Lea’s fascination with the brain began in high school after watching a documentary on autism and savant syndrome. She was captivated by the example of the autistic artist, Stephen Wiltshire, who could reproduce an entire cityscape from memory after a single helicopter ride. Curious about how abilities like that emerge from the brain, she chose to pursue an individualized degree in Natural Sciences at University College London, studying neuroscience alongside chemistry, physics, and math. A stint in the chemistry lab convinced her that experimental work was not a good fit, turning her instead toward mathematics and statistics—a fortunate detour, since that quantitative training would become central to her scientific career. While not yet committed to an academic career, Lea explored multiple research opportunities during undergrad, eventually finding her way to computational neuroscience. Upon advice from one of her mentors, she followed up with a masters in Machine Learning and Statistics at the Gatsby Unit in UCL, working with Dr. Maneesh Sahani on statistical methods for neuroscience. Still not sure of applying to graduate school, she worked as an RA at Princeton before finally returning to Maneesh’s lab for a PhD. More than anything, she says, she needed the repeated exposure to convince herself that research was genuinely what she wanted to do.
Lea began her PhD around the time that advances in recording technologies were transforming systems neuroscience. Researchers could now record the activity of hundreds or thousands of neurons simultaneously, shifting the field away from studying individual neurons toward distributed population dynamics. For her PhD, Lea developed new statistical methods to make sense of these increasingly complex datasets by extracting underlying structure from noisy, high-dimensional data. In particular, she was interested in identifying the dynamical structure, i.e. how the recurrent connectivity between neurons determines how neural population activity unfolds over time. She wanted to link these dynamics to task-relevant computations, such as transforming sensory evidence into action plans, or keeping different items in working memory.
When she initially started a collaboration with Prof. Krishna Shenoy’s lab at Stanford University, she was hoping to extract just such structure from neural recordings in the motor cortex. She wondered if the reason why optogenetic activation of motor cortical neurons had so little impact on the monkey's arm movements had to do with the structure of underlying population dynamics. But resolving the puzzle of those paradoxical null results required a different perspective from a purely statistical description of population activity. Typically, neuroscientists use dimensionality reduction techniques to find the largest axes of variation in population activity, the “loudest signals”. But the task-relevant activity patterns—those that directly influence movement kinematics—turned out to be distinct from the dominant patterns and were relatively robust to “random” network perturbations. She then built computational models to show that it was the balance between local excitatory and inhibitory neurons that shaped cortical network dynamics and enabled the population state to recover quickly from perturbations. Early in graduate school, Lea had viewed characterizing population structure as the central challenge of computational neuroscience. This project convinced her that biological properties, like distinct cell types, could be part of the explanation. Statistical models were a way to reveal structure in neural and behavioral data—and still remain an important bottleneck today for studying complex tasks—but understanding biological computation required asking why those patterns emerged in the first place.
Working closely with experimentalists and as part of an interdisciplinary team was a formative experience for Lea. But the project outcome was also shaped by the insistence of her doctoral advisor, Maneesh, who repeatedly challenged her with the same question, “But why?” The project took years longer than Lea initially expected, and there were moments of frustration when the temptation to settle for a partial explanation was strong. In retrospect, however, she credits that experience with teaching her the value of resisting the pressure to “move on”, and how important that was for making significant discoveries.
Toward the end of her PhD, Lea briefly considered finding a job in AI and tech regulation, where her technical expertise would be particularly useful and directly impactful. But as the Covid-19 pandemic caused major upheaval worldwide, she decided to do a postdoc at Stanford University with Scott Linderman and Krishna Shenoy, her previous collaborator. In the end, she felt drawn back to academia for its relative freedom to pursue fundamental questions that still needed answering and the ability to mentor students. She secured a transition grant from the Simons Foundation and a faculty offer from Columbia University much earlier than she anticipated, but balancing job applications and science in the midst of an ongoing pandemic had been tough. Lea decided to defer her position for a bit and find time to rediscover her joy in science after the stressful few years.
Today, her group studies how neural populations support learning and flexible behavior—how animals rapidly switch between different tasks, adapt previously learned knowledge to new situations, and acquire new behaviors through learning. Learning poses a particularly tricky challenge because both behavior and neural activity change from one trial to the next. This makes it difficult to average across repeated trials, which is a standard procedure for most traditional analyses. Developing statistical methods capable of extracting meaningful structure from noisy single-trial measurements continues to be a central focus of her lab. At the same time, those methods are developed alongside theoretical models of network computation and close collaborations with experimentalists, reflecting Lea's conviction that theory and data each illuminate different aspects of the same problem. After years of being the person “caring the most about neuroscience in a machine learning lab”, or “caring the most about methods in a systems neuroscience lab”, building a team of her own where everyone enjoys working at these intersections has been especially rewarding.
Looking back, Lea is grateful to the many people who created the positive scientific environments that shaped her career. Senior scientists took time to answer questions, teach unfamiliar techniques, and guide her toward new opportunities, creating a culture of mentorship that “trickled down” to other people in the lab. Those early experiences not only convinced her of her own commitment to research, but also shaped her understanding of how science is done. Again and again, she felt encouraged not to rush toward the next milestone, but to leave room for uncertainty—to stay longer with difficult scientific questions, to ask for advice when navigating unfamiliar stages of academia, and to rest after intense periods of work. These decisions, she says, remind her that creativity requires “room to breathe”. It is easy to become consumed by the next paper, the next grant, or the next deadline, until the curiosity that first drew one into science is gradually stifled by the pressure to produce. Scientific discovery depends on creating space for ideas to mature and by bringing together different ways of thinking. Some answers, she has learned, cannot be rushed. They simply require continuing to ask why.
Find out more about Lea and her lab’s research here.
Listen to Harsha’s full interview with Lea on April 20, 2026 below!
