A right-hemispheric language network at single-neuron resolution www.biorxiv.org/content/10.64898/2026.03...

Huge team effort. This work reflects a close collaboration across clinicians, neurosurgeons, and research teams. It would not have been possible without the commitment and trust of our participant.

Comments welcome!

Together, our results suggest that language in the brain, not just speech, is structured in a way that we can begin to access at the level of single neurons.

This opens an avenue for mechanistically specific neurorehabilitation such as brain-computer interfaces for language decoding.

Frontal regions were dominated by time-dependent structure, while SMG emphasized stimulus-by-time interactions (dPCA). Representational geometry reflected semantic and phonological structure (RSA), supporting the dissociation seen on the unit level.

We used LLM-based embeddings to separate semantic and phonological contributions. Both predicted spiking activity, with a regional dissociation: prefrontal units (MFG, IFG) showed a stable semantic preference, while SMG flexibly shifted feature preference with task demands.

Item identity could be decoded above chance in all regions and tasks with linear SVMs. Decoding in prefrontal regions generalized across time and across tasks, consistent with abstract lexical representations in association cortex.

This breaks at the level of selectivity. Units were selective for individual items in all regions, strongest in naming. Like activity clusters, selectivity clusters were intermingled within arrays. Yet, in contrast to activity, selectivity was task-specific and not correlated across tasks.

Activity patterns formed clusters that were stable across the 10-month recording period and distributed across all four arrays. Crucially, overall activity was correlated across tasks.

All recorded regions (MFG, IFG, SMG, ANG) showed task engagement. Single-unit activity tracked task events and was modulated by both task and items (image or audio recording to be processed).

We studied an individual with chronic non-fluent aphasia after extensive left-hemispheric stroke. Across 54 sessions over 10 months, we tested naming, repetition, and comprehension. Performance matched the syndrome: comprehension and repetition relatively preserved, naming impaired.

🚨PREPRINT ALERT🚨

Do single neurons in the right hemisphere encode language after stroke?

We recorded single neurons for 10 months from right-hemispheric association cortex in aphasia.

We find structured language signals, not diffuse compensation. A thread.

www.biorxiv.org/content/10.6...

Activity patterns formed clusters that were stable across the 10-month recording period and distributed across all four arrays. Crucially, overall activity was correlated across tasks.

#SfN25 is drawing to a close. The pace of progress is mind-boggling! I am deeply grateful for the many encounters with friends and colleagues, old and new, and the excitement we shared. A conference is only as good as the people attending. We had a blast! 💥 See you next year! #neuroskyence

Human Single Neuron Meeting at Caltech coming to an end, and my insula is fully convinced it was a high-value reward signal: happiness overflow. Now gearing up for 5 more days with the entire neuroscience community in San Diego.⚡️🤯
Come chat at our poster on Sunday afternoon at MM10.

The team will be at #SfN25 in San Diego. Come for a chat at our posters. We have #human #single #units, #microstimulation, #miniscopes and much more! 🧠🧪 #neuroskyence

Our #BCI work was featured in #Tagesthemen, Germany’s most important newscast. Check it out 👇🏻 www.tagesschau.de/tagesthemen/...

Filming in the lab today for a feature on our #BCI work. Scorching hot day! Well done @lauraschiffl.bsky.social and all! ☀️☀️☀️😎🎥

Today, we celebrated the laboratory’s 10th anniversary. 🎉 You know what? I’ve been in science long enough. It’s not the papers or the grants that make me come to the lab every day. It’s the people I have the pleasure of working with. Thanks for all the fun!! 🧠 🧪