Colton Casto

The paper identifies 17 language-selective brain areas outside the core fronto-temporal network using fMRI data from 772 participants, showing they respond to both auditory and written language and are distinct from general task demands.

The paper maps human cortical connectivity using es-fMRI and finds that functional similarity, not distance, primarily determines anatomical connections.

This study identifies four cerebellar regions involved in language processing using fMRI, with one region showing selectivity for language and functional connections to the neocortical language system.

The paper identifies distinct brain areas that process prosody in speech across multiple experiments and participants.

This study shows that high-performing ANNs and brains converge on similar representations, and identifies stimuli that modulate model-to-brain alignment.

This study shows that speakers' brains generate information-rich words through extra neural computation, evidenced by stronger neural activity before articulation, and that LLMs similarly require deeper computation for improbable words.

The paper identifies three distinct temporal receptive window sizes in neural populations of the language network using intracranial recordings.

The paper uses fMRI to show that language processing involves a distributed network sensitive to both structure and meaning, challenging theories of distinct brain areas for each.

In nine epilepsy patients, cortical activity carries next-word information up to 800 ms before onset, tracks GPT-2 surprise after onset, and fits GPT-2 contextual embeddings better than static ones.