Research

Here are a few topics the lab is pursuing currently:

Cortical representations of intermediate-level visual features

Visual cortex performs hierarchical computations that span from simple features, like edges and colors, to complex features, like object and scene parts. What are the features and representations that bridge between these levels of complexity? We’re developing new computational approaches to understand how intermediate visual features are computed in the brain and how they contribute to behavior. We are also aiming to determine how the statistical distribution of features in the environment is reflected in the tuning properties and organization of visual cortex.

Related past work:

• Henderson, M.M., Tarr, M.J., & Wehbe, L. (2023). A texture statistics encoding model reveals hierarchical feature selectivity across human visual cortex. Journal of Neuroscience. (pdf)

• Henderson, M.M., Tarr, M.J., & Wehbe, L. (2023). Low-level tuning biases in higher visual cortex reflect the semantic informativeness of visual features. Journal of Vision. (pdf)

Natural image statistics & category learning

Our visual environment is highly structured in its feature statistics, with certain features being encountered more often than others. Morevoer, the statistics of our visual inputs may evolve over time during development, due to changes in visual sensitivity as well as visuomotor behavior. How do these properties of the environment provide a curriculum that enables efficient category learning? How is the structure of visual representations constrained by the structure of visual inputs? We’re working to address these questions using simulated learning experiments performed in deep neural network models.

Related past work:

• Jinsi, O.* , Henderson, M.M.*, & Tarr, M.J. (2023). Early experience with low-pass filtered images facilitates visual category learning in a neural network model. PLOS ONE. (pdf)

• Henderson, M.M., & Serences, J.T. (2021). Biased orientation representations can be explained by experience with non-uniform training set statistics. Journal of Vision. (pdf)

Flexible processing in higher visual cortex

To support behavioral goals, visual cortex circuits may adaptively adjust stimulus representations according to task-relevance. This idea of flexible neural coding has often been studied using simple tasks with controlled stimuli, leading to many open questions about how flexible neural codes might contribute to more naturalistic, high-level visual tasks. How do we search for a target category of objects within a cluttered natural scene? How is the “object space” represented in higher visual cortex warped by changes in behavioral priority? We’re addressing these questions using behavior and fMRI studies.

Related past work:

• Henderson, M.M., Serences, J.T., & Rungratsameetaweemana, N. (2023). Dynamic categorization rules alter representations in human visual cortex. bioRxiv; under review.

• Henderson, M.M. & Serences, J.T. (2019). Human frontoparietal cortex represents behaviorally relevant target status based on abstract object features. Journal of Neurophysiology. (pdf)