How Do Bilingual Brains Navigate Between Languages? New Study Explains
How does the human brain seamlessly switch between multiple languages without constant internal confusion? For decades, neuroscientists have debated how the brain represents the same meaning across different languages while keeping them distinct. A newly published study uses direct brain recordings in humans to reveal a preserved geometric map of meaning within the hippocampus, with different languages accessing this shared structure through distinct neural readout axes.
Note: This article is intended for general information and educational purposes. It summarizes scientific research in accessible language for a broad audience and is not an official scientific press release.
Understanding how the human mind comprehends and expresses equivalent thoughts across multiple languages remains a central question in cognitive neuroscience. A study published on June 24, 2026, in the journal Cell, titled “Shared neural geometries for bilingual semantic representations in human hippocampal neurons,” offers fresh insights into this mechanism.
The research was conducted by a collaborative team of scientists, including Xinyuan Yan, Ana G. Chavez, Melissa Franch, Kalman A. Katlowitz, Ivy Gautam, Brian Kim, Aaditya Krishna, Aadit Shrivastava, Katie Van Arsdel, James Belanger, Assia Chericoni, Taha Ismail, Elizabeth A. Mickiewicz, Danika Paulo, Hanlin Zhu, Alica M. Goldman, Vaishnav Krishnan, Atul Maheshwari, Eleonora Bartoli, Nicole R. Provenza, Seng Bum Michael Yoo, Benjamin Y. Hayden, and Sameer A. Sheth.
The authors were affiliated with the Department of Neurosurgery and the Department of Neurology at Baylor College of Medicine, Houston, Texas, USA; the Neuroengineering Initiative, the Department of Electrical and Computer Engineering, and the Department of Bioengineering at Rice University, Houston, Texas, USA; and the Department of Biomedical Engineering, the Department of Intelligent Precision Health Convergence, and the Center for Neuroscience Imaging Research at Sungkyunkwan University, Suwon, South Korea.
By recording the activity of individual cells in the human brain, the researchers found evidence that the brain maps meaning using a language-independent internal model structured as a high-dimensional geometric space.
What the Researchers Investigated
The study aimed to resolve a fundamental mystery in bilingual language processing: how the brain successfully matches corresponding concepts across different languages while simultaneously keeping them separate enough to prevent speech interference.
The researchers tested the idea that the bilingual brain implements a “shared neural geometry” for representing meaning. Under this framework, the brain maintains a stable mathematical distance between pairs of words in a high-dimensional neural space, regardless of the language spoken. For example, if the concepts “cat” and “dog” elicit highly similar neural activation patterns while the concept “door” elicits a highly dissimilar pattern, their corresponding translation equivalents (“gato,” “perro,” and “puerta”) should preserve those identical relational distances.
The investigation focused specifically on the human hippocampus, a brain region that has often been overlooked in language research but increasingly recognized for its role in tracking word meanings and flexibly organizing relationships between abstract concepts.
How the Study Was Conducted
The researchers recorded single-neuron activity from four rare clinical participants who were fully balanced native bilinguals (English/Spanish) undergoing neurosurgery for treatment-resistant epilepsy. All participants were women who had acquired both languages at ages 4–5 and used them regularly throughout their lives.
To record hippocampal activity, three participants had temporary microelectrode arrays implanted for seizure monitoring. A fourth participant underwent an anterior temporal lobectomy, during which researchers inserted a high-density Neuropixels probe into the exposed hippocampus before the planned surgical removal of brain tissue.
Neural activity was examined across three everyday language situations:
- Study 1 (Passive Listening): Participants listened to translation-equivalent English and Spanish stories, podcasts, and audiobooks for approximately two hours in each language, providing thousands of matched words for analysis.
- Study 2 (Reading Aloud): Two awake participants read 99 matched short phrases (3–4 words each) aloud in both languages, presented in random order.
- Study 3 (Natural Conversation): Two participants engaged in spontaneous conversations lasting 32–99 minutes with native English- and Spanish-speaking partners.
To analyze how individual neurons represented meaning, the researchers used multilingual BERT (mBERT), an artificial intelligence language model that generates comparable semantic representations across languages. Word embeddings from both languages were projected into a shared semantic space using principal component analysis (PCA). Finally, ridge regression models were used to predict the firing rates of individual hippocampal neurons from these shared semantic features.
What Makes This Study New
Compared with previous studies using functional magnetic resonance imaging (fMRI) or electrocorticography (ECoG), which mainly showed that different languages activate similar brain regions or speech-related activity, the authors describe this as the first study to examine how individual neurons represent meaning across two languages in humans.
The authors explain that previous studies could show similarities between languages across large groups of brain cells, but they could not reveal how the brain organizes the meaning of individual words at the level of single neurons.
The study also shows something unexpected: although individual neurons do not all respond to words in the same way, together they still preserve the same overall map of meaning across both languages. This suggests that the brain can maintain a shared representation of concepts without requiring every neuron to behave identically.
Key Findings from the Study
The researchers reported several important findings about how the bilingual brain organizes meaning across languages:
- A Small Group of Neurons Responded Across Both Languages: The researchers identified a small but statistically significant group of neurons (5.2%–19.4% during listening tasks, and up to 83.6% during isolated word reading) that responded similarly to translation-equivalent words. For example, if a neuron responded strongly to the English word “friends,” it also tended to respond strongly to its Spanish translation, “amigos.”
- Individual Neurons Still Behaved Differently: Although some neurons responded to the same concepts in both languages, the overall pattern of which words each neuron preferred differed between English and Spanish. In other words, individual neurons did not use exactly the same “language map.”
- The Brain Preserved the Same Overall Map of Meaning: When the activity of many neurons was analyzed together, the overall organization of meaning remained remarkably similar across both languages. Concepts that were closely related (such as “cat” and “dog”) were represented close together, while unrelated concepts (such as “human” and “galaxy”) remained far apart in both English and Spanish.
- The Brain Could Predict Meaning Across Languages: Using the neural activity recorded for English words, the researchers were able to accurately predict how the brain would respond to the corresponding Spanish words, even for words that had not been used to build the model.
- The Shared Map Did Not Depend on a Small Group of Specialized Neurons: Perhaps the most surprising finding was that removing the specialized “cross-language neurons” did not disrupt the brain’s shared semantic map. The overall organization remained almost unchanged, suggesting that bilingual meaning is distributed across many neurons rather than relying on a small, specialized group.
Authors’ Conclusions
Based on their findings, the authors propose that the human hippocampus stores meaning in a shared, language-independent way. Rather than assigning identical roles to individual neurons in every language, the brain appears to use the same underlying network of neurons while reading out that information slightly differently depending on the language being used. According to the researchers, this may help explain how bilingual people can switch smoothly between languages without confusing words from one language with another.
The researchers also acknowledged several limitations. Most importantly, the study included only four participants because balanced bilingual patients undergoing this type of neurosurgery with hippocampal microelectrode recordings are extremely rare. In addition, all participants had epilepsy and were undergoing neurosurgical treatment, which should be considered when interpreting the findings. The study also examined only English and Spanish, so it remains unknown whether the same principles apply to more distantly related languages. Finally, the researchers did not investigate how this shared representation of meaning develops during language learning. They suggest that future studies follow people as they learn a new language to better understand how these shared neural representations emerge over time.
Understanding the Broader Context
These findings add to growing evidence that the brain represents meaning through the combined activity of large groups of neurons rather than relying on individual “concept cells” dedicated to specific ideas.
For many years, scientists proposed that bilingual brains either stored each language separately or constantly searched for translations between them. This study supports a different view: both languages may share the same underlying representation of meaning, while each language accesses that shared representation in its own way. In other words, translation may not require switching between two separate mental dictionaries, but instead involves expressing the same concepts through different linguistic systems.
Conclusion
In summary, the findings suggest that the bilingual human brain maintains a shared organization of meaning within the hippocampus during both listening and speaking. While individual neurons respond differently depending on the language context, the overall pattern of relationships between word meanings appears to remain stable across languages. What remains unknown is how this shared structure forms and becomes aligned across languages. Future studies evaluating non-cognate, structurally distinct language families will be necessary to expand our understanding of these internal models of meaning. In simple terms, this study asked how the brain can understand the same idea in two languages without mixing them up. The findings suggest that the brain may preserve a shared structure for meaning across both languages while still keeping them distinct.
The information in this article is provided for informational purposes only and is not medical advice. For medical advice, please consult your doctor.
References
- Yan, X., Chavez, A. G., Franch, M., Katlowitz, K. A., Gautam, I., Kim, B., Krishna, A., Shrivastava, A., Van Arsdel, K., Belanger, J., Chericoni, A., Ismail, T., Mickiewicz, E. A., Paulo, D., Zhu, H., Goldman, A. M., Krishnan, V., Maheshwari, A., Bartoli, E., Provenza, N. R., Yoo, S. B. M., Hayden, B. Y., & Sheth, S. A. (2026). Shared neural geometries for bilingual semantic representations in human hippocampal neurons. Cell, 189, 1–16. https://doi.org/10.1016/j.cell.2026.05.020
- Rice University. (2026, June 24). How do bilingual brains navigate between languages? Scientists discover geometric neural code. https://news.rice.edu/news/2026/how-do-bilingual-brains-navigate-between-languages-scientists-discover-geometric-neural
