About

Jack Gallant is a Professor of Neuroscience at the University of California, Berkeley. His research interests include identifying cortical maps to discover how the brain represents information about the world and its own mental states. He is involved in computational neuroscience and cognitive neuroscience topics, focusing on understanding the neural basis of perception, cognition, and mental processes. As a faculty member in the Neuroscience Department, he contributes to advancing knowledge in these areas through his research and academic activities.

Research topics

• Computer Science
• Neuroscience
• Political Science
• Social Science
• Sociology
• Psychology
• Engineering
• Law
• Engineering ethics
• Internet privacy
• Cognitive psychology
• Public relations

Selected publications

• Representations of semantic relations in the human cerebral cortex

  bioRxiv (Cold Spring Harbor Laboratory) · 2026-02-20

  articleOpen accessSenior authorCorresponding

  Abstract An essential aspect of human cognition is the ability to explicitly think about semantic relations between concepts. Neuroimaging studies have found that individual concepts are encoded by distributed patterns of cortical activity, but relatively little is known about how semantic relations between concepts are encoded in the brain. Some theoretical models suggest that relation representations are embedded within concept representations, while others suggest that relation representations are independent of any specific concept pair. We designed a study to compare how semantic relations and concepts are encoded across the cerebral cortex. To characterize how relations are encoded across cortex, fMRI was used to record brain activity while six participants each answered over one thousand questions about different semantic relations. We find that relations are encoded independently of the specific concepts that are connected in any particular instance of the relation. Our results further suggest that relations and concepts are represented in the same set of cortical regions, and that, within these regions, each location is preferentially selective for specific relations. Overall, these results suggest that in the human cerebral cortex, relations and concepts may have the same type of functional representation.

  PublisherOA PDFDOI

• Visual semantic tuning across the cortex shifts between tasks

  bioRxiv (Cold Spring Harbor Laboratory) · 2026-02-20

  articleOpen accessSenior authorCorresponding

  Abstract Attention is a powerful mechanism that dynamically optimizes brain representations to prioritize behaviorally-relevant information. While previous studies focusing on single tasks have demonstrated that attention shifts tuning towards attended targets, real-world behavior often requires humans to switch between tasks with different demands. How does visual semantic tuning in the human brain shift to support the behavioral demands of different naturalistic tasks? To answer this question, we used voxelwise encoding models to compare visual semantic tuning across the human cerebral cortex between two distinct naturalistic tasks, movie watching and navigation. Results show that visual semantic tuning in the cortex differed substantially between tasks. Principal component analysis reveals that during navigation, tuning shifts increase the representation of vehicles and traffic signs compared to movie watching, and that these shifts were localized to distinct functional networks. These tuning shifts reconfigure the human brain’s representations of object categories based on their behavioral relevance. These findings demonstrate that visual semantic tuning in the brain dynamically shifts towards task-relevant information across naturalistic tasks, optimizing functional representations to achieve diverse behavioral goals. Significance statement In real-world situations, we switch between many tasks with behaviorally distinct goals that require us to attend to different targets. Here we show that visual semantic tuning across the human cortex shifts between passive movie-watching and active navigation. These tuning shifts increase the representation of behaviorally-relevant objects, such as cars during navigation. Tuning shifts towards different object categories are distributed across distinct functional networks, suggesting that they engage different cognitive mechanisms. These results show that visual semantic tuning in the human brain is highly task-dependent and is optimized to represent behaviorally-relevant information.

  PublisherOA PDFDOI

• Bilingual language processing relies on shared semantic representations that are modulated by each language

  Proceedings of the National Academy of Sciences · 2026-02-24 · 1 citations

  articleOpen access

  Billions of people throughout the world are bilingual, and they can extract meaning from multiple languages. While some evidence suggests that there is a shared system in the human brain for processing semantic information from native and non-native languages, other evidence suggests that semantic processing is language specific. We conducted a study to determine how semantic information for different languages is represented in the brains of bilinguals. Functional magnetic resonance imaging (fMRI) was used to record brain responses while participants read several hours of natural narratives in their native (Chinese) and non-native (English) languages. These data were then used to compare semantic representations between the two languages. We find that semantic representations are largely shared between languages, while there are fine-grained differences in the representation of some semantic categories across languages. These results reconcile current competing theories of bilingual language processing.

  PublisherDOI

• autoflatten: automatically create cortical flatmaps from FreeSurfer surfaces

  Zenodo (CERN European Organization for Nuclear Research) · 2025-12-19

  otherOpen accessSenior author

  Python pipeline for automatically creating flattened versions of FreeSurfer cortical surfaces. Features JAX-accelerated optimization with vectorized gradient descent, template-based cut mapping via FreeSurfer's surface registration, and FreeSurfer-compatible patch file output.

  PublisherDOI

• A map of the cortical functional network mediating naturalistic navigation

  bioRxiv (Cold Spring Harbor Laboratory) · 2025-12-17 · 1 citations

  articleOpen accessSenior authorCorresponding

  Navigation through the real world requires close coordination of perception, planning, and motor actions. Prior neuroimaging studies that used controlled stimuli and tasks have suggested that navigation-related information is represented broadly across the human cerebral cortex. While three regions in the anterior visual cortex have been well-studied, the extent and functional properties of regions in the parietal and prefrontal cortices are not as well-characterized. To map and characterize the full cortical network that underpins active navigation in the real world, we used functional magnetic resonance imaging to record brain activity from participants performing a naturalistic navigation task. Banded ridge regression was used to fit high-dimensional encoding models for 28,134 features to this data. Results show that naturalistic navigation is supported by a network of 11 functionally distinct cortical regions: five prefrontal and three parietal regions, along with three regions in the visual cortex that had been identified and characterized in previous studies. Analysis of encoding model weights shows that these 11 regions transform perceptual inputs through decision-making processes to produce action outputs, and are organized along distributed cortical functional gradients. These results provide a unified description of the functional properties and organization of the cortical network that mediates naturalistic navigation. We anticipate that these maps will provide rich targets to inform more targeted future studies of human navigation.

  PublisherOA PDFDOI

• autoflatten: automatically create cortical flatmaps from FreeSurfer surfaces

  Zenodo (CERN European Organization for Nuclear Research) · 2025-12-20

  otherOpen accessSenior author

  Python pipeline for automatically creating flattened versions of FreeSurfer cortical surfaces. Features JAX-accelerated optimization with vectorized gradient descent, template-based cut mapping via FreeSurfer's surface registration, and FreeSurfer-compatible patch file output.

  PublisherDOI

• autoflatten: automatically create cortical flatmaps from FreeSurfer surfaces

  Zenodo (CERN European Organization for Nuclear Research) · 2025-12-21

  otherOpen accessSenior author

  Python pipeline for automatically creating flattened versions of FreeSurfer cortical surfaces. Features JAX-accelerated optimization with vectorized gradient descent, template-based cut mapping via FreeSurfer's surface registration, and FreeSurfer-compatible patch file output.

  PublisherDOI

• The cortical navigation network is organized into distributed functional gradients

  Journal of Vision · 2025-07-15

  articleOpen accessSenior author

  Actively navigating through the natural world requires close coordination of perception, planning, and motor actions. Multiple regions in the human cerebral cortex have been implicated in representing navigation-related information. Rodent studies suggest that navigation-related information is represented in multiple overlapping spatial gradients that extend across functional regions (Minderer et al., 2019, Tseng et al., 2022). Is this organizational principle shared by the navigation network in the human cerebral cortex? How is the navigation network related to other networks, such as those that process visual inputs and produce motor outputs? To address these questions, we used fMRI to record brain activity from six subjects performing a taxi-driver task in a naturalistic virtual world (110-180 minutes of data per subject). We extracted 28,161 features across 38 visual-, motor-, and navigation-related feature spaces from the experiment recordings, and used banded ridge regression to fit encoding models for all feature spaces simultaneously. To identify the network of cortical regions that represent navigation-related information, model connectivity (MC) was used to hierarchically cluster the voxel weight vectors. MC identified a network of 11 regions in the visual, parietal, and prefrontal cortices that each represent a distinct combination of navigation-related features. To determine its organizational principles, UMAP was used to recover a functional space for the navigation network. In this space, the navigation regions are organized into a continuous functional distribution, and this distribution maps to continuous spatial gradients on the cortical surface. Finally, to relate the navigation network to other networks, UMAP was used to recover a functional space for the whole cortex. In this space, the navigation network appears to be positioned in the middle of a broad gradient extending from visual to motor networks. These results provide a detailed characterization of the functional gradients underlying the cortical network that mediate active, naturalistic navigation.

  PublisherDOI

• The Voxelwise Encoding Model framework: a tutorial introduction to fitting encoding models to fMRI data

  2025-04-02

  preprintOpen accessSenior author

  The Voxelwise Encoding Model framework (VEM) is a powerful approach for functional brain mapping. In the VEM framework, features are extracted from the stimulus (or task) and used in an encoding model to predict brain activity. If the encoding model is able to predict brain activity in some part of the brain, then one may conclude that some information represented in the features is also encoded in the brain. In VEM, a separate encoding model is fitted on each spatial sample (i.e., each voxel). VEM has many benefits compared to other methods for analyzing and modeling neuroimaging data. Most importantly, VEM can use large numbers of features simultaneously, which enables the analysis of complex naturalistic stimuli and tasks. Therefore, VEM can produce high-dimensional functional maps that reflect the selectivity of each voxel to large numbers of features. Moreover, because model performance is estimated on a separate test dataset not used during fitting, VEM minimizes overfitting and inflated Type I error confounds that plague other approaches, and the results of VEM generalize to new subjects and new stimuli. Despite these benefits, VEM is still not widely used in neuroimaging, partly because no tutorials on this method are available currently. To demystify the VEM framework and ease its dissemination, this paper presents a series of hands-on tutorials accessible to novice practitioners. The VEM tutorials are based on free open-source tools and public datasets, and reproduce the analysis presented in previously published work.

  PublisherDOI

• Selective attention warps the representation of space throughout cortical visual networks

  Journal of Vision · 2025-07-15

  articleOpen accessSenior author

  Visual attention systems in the brain selectively enhance visual information that is relevant to the target or goal (Cukur et al., 2013). Most object-based visual attention studies have focused on how attention affects the representations of high-level semantic categories or objects. However, high-level semantic category representations rely on the integration of low-level visual representations from earlier stages of visual processing. Given that the visual system is densely interconnected, it is important to understand how object-based attention modulates low-level visual representations to facilitate downstream object-based perception. Here we examine whether object-based attention affects spatial representations across cortical regions of interest (ROIs) during a naturalistic movie-watching task. Six participants watched compilations of movie clips either while fixating passively or while covertly searching for “humans” or “vehicles” in the movies (Cukur et al., 2013). High-level semantic features and low-level visual features were recovered from the movies, and these were used to fit voxelwise encoding models to predict brain responses in each participant, under each attention condition (Dupré la Tour et al., 2022). Task-specific functional networks were recovered using model connectivity (Meschke et al., 2022), and were used as ROIs in the analyses. To assess semantic representation changes across attention conditions, the semantic model weights were correlated with ‘human’ and ‘vehicle’ semantic templates (as in Cukur et al., 2013). To assess spatial representation changes across attention conditions, the visual model weights were correlated with a spatial template that partitioned the visual field into ‘center’ and ‘periphery’. These analyses revealed that selective attention to either semantic target shifts spatial representations towards the periphery in the majority of networks. Furthermore, in some networks, the magnitude of the spatial representation shift was correlated with the magnitude of the semantic representation shift. These results suggest that object-based selective attention is supported by changes in low-level spatial representations.

  PublisherDOI

Recent grants

• CRCNS: Cortical representation of phonetic, syntactic and semantic information during speech perception and language comprehension

  NSF · $889k · 2012–2017

• Representation of information across the human visual cortex

  NIH · $2.9M · 2010–2020

• NIH Grant R01EY012241

  NIH · $2.5M · 2010

• Attentional Modulation of Brain Representations

  NIH · $1.6M · 2013–2018

• CRCNS US-German Research Proposal: Language representations in bilinguals

  NSF · $920k · 2019–2024

Frequent coauthors

• Jeffrey A. Herron

  University of Washington

  34 shared

• Gabrielle Strandquist

  University of Washington

  31 shared

• Raphael Bechtold

  University of Washington

  31 shared

• Daryl Lawrence

  31 shared

• Shravanan Ravi

  University of California, San Francisco

  30 shared

• Simon Little

  30 shared

• Alicia Zeng

  Amgen (United States)

  30 shared

• Tomasz Frączek

  30 shared