A study in the journal Cell provides evidence that the human brain uses a shared organizational map to represent word meanings across languages. By recording individual brain cells in bilingual people, scientists found that each language produced distinct activity patterns, yet the relationships between meanings stayed consistent. The brain may therefore keep a language-independent internal model of meaning.
Earlier brain imaging showed that bilingual speakers use overlapping regions for both languages. But broad scans could not reveal how the brain matches equivalent concepts while keeping the languages separate. Researchers from Baylor College of Medicine, Rice University and Sungkyunkwan University examined this question at the level of single cells. They tested whether the bilingual brain uses a shared neural geometry — the mathematical distances between words in a high-dimensional space.
The team focused on the hippocampus, a deep brain region important for memory and for linking concepts. The scientists studied four people who were equally fluent in English and Spanish and recorded single-neuron activity with high-density microelectrodes.
All four participants spent about 120 minutes listening to matched stories and podcasts in both languages. Two also read 99 matched short phrases aloud and held conversations with native speakers of each language. The scientists then aligned the audio with the neural recordings word by word.
First, the team searched for cells that responded in the same way to translated pairs such as "earth" and "tierra." A small number existed, but there were too few to explain fluent language switching. This suggests that translation does not depend on special dictionary-like cells. Instead, meaning emerges from coordinated activity across many neurons.
At the level of individual cells, the two languages looked different. Most neurons did not show matching responses to English and Spanish meanings. A cell might respond strongly to the English word "dog" but remain quiet for the Spanish word "perro."
Across the larger population of cells, however, the same geometric organization appeared in both languages. The distances among neural responses to meanings in English closely matched those in Spanish. The researchers compare this to viewing a three-dimensional object from two angles: its profile changes, but its underlying shape remains the same. The brain may use the same neurons while reading their activity along different axes for each language.
This structure could also predict neural responses. Starting with a group of related English words, the scientists mathematically rotated the data and predicted how the brain would respond to a Spanish word that had been left out of the analysis. They did this without finding a one-to-one match between individual neurons.
The researchers also compared the neural data with a multilingual artificial intelligence model. The model's organization of meaning resembled the geometry found in the hippocampus. This does not mean artificial intelligence works exactly like the brain, but the similarity may point to shared principles for organizing knowledge.
The study has clear limits. It involved only four highly fluent bilinguals who learned both languages at about age four or five. It also examined only English and Spanish, which share many roots. Future studies could test unrelated language pairs and follow people learning a new language to see how a shared map of meaning develops.