Martin N. Hebart

The paper identifies two dominant axes organizing high-dimensional object representations in the human ventral temporal cortex using fMRI and factor analysis.

The paper investigates coarse-scale functional organization in occipitotemporal cortex using fMRI data, revealing complexities in animacy-size organization and alternative dimensions.

The paper aligns human and macaque inferotemporal cortex representations of objects using multivariate cross-species alignment, revealing shared high-dimensional spaces and species-specific differences.

The paper shows that early visual cortex in humans and non-human primates rapidly computes perceived beauty of objects without explicit judgments.

This study identifies multidimensional tuning in human visual cortex, showing that category-selective areas encode both subcategory features and cross-category distinctions, supporting both local and global coding.

The paper uses encoding models to reveal fine-grained feature selectivity in category-selective areas of the occipitotemporal cortex for bodies, hands, and tools.

The paper investigates how visual ambiguity is resolved by analyzing a dataset of ambiguous images and human ratings, showing that higher-level visual features drive subjective identification.

The paper investigates how multidimensional object properties are represented in the human brain over time using MEG signals and behavioral data.

The paper demonstrates image-specific foveal feedback during saccade preparation using fMRI, showing decoding of peripheral targets from foveal areas and a u-shaped decoding curve.

This paper introduces a deep learning tool to predict functional topography of the human visual cortex from anatomy, demonstrating its utility across diverse conditions and large-scale datasets.

This study uses MEG and fMRI to show that numerosity is encoded rapidly in the brain, with early linear and later curved representations reflecting different coding schemes.

This study investigates semantic tuning of concept cells in the human medial temporal lobe, finding graded responses and region-specific categorical profiles.

The paper identifies and characterizes brain representations of scenes relevant for categorization behavior across different tasks using fMRI and deep neural networks.

This study investigates fine-grained visual information in the ventral visual pathway using fMRI, finding that representations are dominated by image-specific effects but also contain subtle category-specific information.

The paper shows that neural responses during natural vision are better explained by saccade-related events than by fixation onsets, challenging traditional stimulus onset paradigms.

This paper introduces a computational method to predict perceived similarity of natural objects using neural network activations, achieving high accuracy with triplet judgments.

The paper maps behavior-derived object dimensions onto the human brain, showing distributed representations that outperform categories in predicting brain responses.

The paper investigates what makes images memorable by analyzing a large dataset of object images and finds that semantic features have a stronger influence than perceptual features on memorability, and that memorability is not simply tied to typicality.

THINGS-data provides multimodal neuroimaging and behavioral datasets for studying object representations, enabling hypothesis testing and cross-modal analysis.

The paper investigates how the brain processes natural images and line drawings, finding shared neural mechanisms across different levels of visual abstraction.

The paper identifies low-dimensional, interpretable dimensions organizing human action similarity judgments using data-driven methods.

The paper investigates how the hippocampal formation represents different types of relational information by analyzing fMRI data to show distinct semantic and statistical maps.

This paper introduces feature-reweighted RSA (FR-RSA) to improve the alignment between computational models, brain activity, and behavior by reweighting features, showing it enhances model fit and selection.

The paper introduces THINGS-EEG, a dataset of EEG responses to 1,854 object concepts, to study visual object processing in the human brain.

THINGSvision is a Python tool that simplifies extracting DNN activations across layers, enabling easier analysis of neural networks' relationship to brain activity and behavior.

This study examines de-differentiated topographic maps in somatosensory cortex of younger and older adults using fMRI and behavioral tests, finding that reduced representational similarity and altered cortical distances correlate with specific behavioral changes.

The paper introduces the THINGS database, containing 1,854 object concepts and 26,107 naturalistic images, validated through semantic embeddings and neural network analysis.

This paper identifies and discusses confusions in the interpretation of multivariate decoding results in neuroscience, highlighting differences from classical univariate analysis and their impact on neuroimaging data interpretation.

The paper estimates the temporal emergence of conceptual object representations by analyzing MEG data and comparing them to behavioral and semantic models.

This study investigates how task context influences object processing in humans using MEG and fMRI, revealing distinct temporal dynamics and interactions between task and object representations.