Jean-Rémi King

Hierarchical dynamic coding coordinates speech comprehension in the brain

This study proposes and tests the Hierarchical Dynamic Coding (HDC) hypothesis, showing how the brain maintains and updates a language hierarchy during speech comprehension.

Modality-Specific and Amodal Language Processing by Single Neurons

The paper identifies modality-specific and amodal language processing by single neurons using neural recordings from patients during sentence comprehension.

Tracking the behavioral and neural dynamics of semantic representations through negation

The paper investigates how negation affects neural representations of adjectives using behavioral and MEG data, finding that negation suppresses rather than inverts meaning.

Algebraic structures emerge from the self-supervised learning of natural sounds.

This paper shows deep learning models can spontaneously detect algebraic structures in sounds through self-supervised learning, with performance depending on exposure to natural sounds and pretraining on music or environmental sounds.

Dimensionality and ramping: Signatures of sentence integration in the dynamics of brains and deep language models

The paper investigates how neural representations of sentences evolve, finding that meaningful sentences have higher dimensionality and specific dynamic patterns in both human brains and deep language models.

The best of two worlds: Decoding and source-reconstructing M/EEG oscillatory activity with a unique model

This paper introduces a method combining CSP and beamformer source reconstruction for decoding M/EEG oscillatory activity, achieving performance comparable to source space decoding with lower computational cost.

Top-down information shapes lexical processing when listening to continuous speech

The paper investigates how the brain resolves lexical ambiguity in speech using context, showing top-down processing dominates over bottom-up cues in early neural responses.

GPT-2's activations predict the degree of semantic comprehension in the human brain

Across 101 listeners, how well GPT-2 layer-8 activations predict fMRI responses to spoken stories tracks story comprehension, with a correlation of R = 0.50 across subject-story pairs.

Spontaneous neural oscillations influence behavior and sensory representations by suppressing neuronal excitability

The paper investigates how spontaneous neural oscillations affect behavior and sensory processing by modulating neuronal excitability during an auditory task.

Language processing in brains and deep neural networks: computational convergence and its limits

In 3,600 transformer checkpoints compared with fMRI and MEG of 102 readers, middle layers best predict brain activity and grow more brain-like as word-prediction accuracy improves; input embeddings diverge.

Neural dynamics of phoneme sequences: Position-invariant code for content and order

The human brain represents speech phoneme sequences through dynamic coding that tracks content and elapsed processing time, revealing how it combines phonemes to access the mental lexicon.

Discriminating the Influence of Correlated Factors from Multivariate Observations: the Back-to-Back Regression

The paper introduces Back-to-Back regression to disentangle correlated factors from multivariate observations, showing it outperforms existing methods in causal identification.

The Human Brain encodes a Chronicle of Visual Events at each Instant of Time

This study decodes and tracks multiple rapidly-changing visual images from human brain activity, showing the brain simultaneously represents multiple images at each time instant.

Recurrent Processes Emulate a Cascade of Hierarchical Decisions: Evidence from Spatio-Temporal Decoding of Human Brain Activity

The paper shows that recurrent neural processes generate a cascade of hierarchical decisions in the human brain during visual perception and motor response.

Slow cortico-cortical connectivity (2-5Hz) is a new robust signature of conscious states

The paper demonstrates that slow cortico-cortical connectivity (2-5Hz) is a robust marker distinguishing conscious from unconscious states across multiple experimental conditions.

Differential brain mechanisms of selection and maintenance of information during working memory

This study investigates the neural mechanisms of information selection and maintenance in working memory using MEG, identifying distinct patterns of oscillatory activity for selection and reactivation of memory content.

Riemannian Geometry Boosts Representational Similarity Analyses of Dense Neural Time Series

This paper improves representational similarity analysis of neural time series using Riemannian geometry, achieving higher decoding accuracy than standard methods.

Tracking the Canonical Computations of Perceptual Processing in a Quasi-Continuous Sensory Flow

This study tracks canonical computations like neural adaptation and prediction-error signals in continuous visual stimulation using EEG and neuronal simulations, showing they can encode multiple stimuli sequentially.

Performance-optimized hierarchical models only partially predict neural responses during perceptual decision making

The paper compares performance-optimized models with neural responses during perceptual decision making, finding partial alignment with some CNNs but discrepancies in others.

A theory of working memory without consciousness or sustained activity

The paper demonstrates non-conscious working memory through visual masking and magnetoencephalography, showing it lacks sustained neural activity but retains decodable representations.

Selective maintenance mechanisms of seen and unseen sensory features in the human brain

This study shows that relevant sensory features are maintained in the brain over time, while irrelevant features are transient, even for unseen stimuli.