Bilingual people use largely the same brain system to understand meaning in both their languages, but each language subtly reshapes how that system processes different categories of words, according to new research from UC Berkeley.
The study, published Monday in the Proceedings of the National Academy of Sciences, used functional MRI to record the brain activity of Chinese-English bilinguals as they read hours of natural stories in each language. The results settle a long-standing debate in neuroscience over whether bilingual brains maintain one shared meaning system or two separate ones: the answer, the researchers found, is essentially both.
Same brain regions, same word responses — mostly
The researchers, led by Catherine Chen and Fatma Deniz, scanned six fluent Chinese-English bilinguals across four separate three-hour sessions as participants read narratives word by word. Each story was presented in both languages.
The same brain regions — spanning the temporal, parietal, and prefrontal cortex — were active during reading in both Chinese and English. More strikingly, 81 per cent of the brain locations that responded to meaning showed identical semantic tuning across both languages. A brain area tuned to family-related words in Chinese, for instance, was also tuned to family-related words in English.
“Semantic representations are largely shared between languages, while there are fine-grained differences in the representation of some semantic categories across languages,” the study states.
Each language tunes the system differently
Despite the broadly shared system, the researchers identified a consistent pattern of language-specific modulation. Using a statistical technique called principal component analysis, they found a primary dimension along which semantic processing shifts between languages.
In some brain regions, reading in English emphasized words related to actions and human relationships — words like “leave,” “boyfriend,” and “family” — while reading in Chinese in those same regions emphasized words related to numbers and collections, such as “three,” “both,” and “few.” Other brain regions showed the reverse pattern.
These shifts were consistent across all six participants and held up across multiple control analyses, including tests using different computational models of word meaning and comparisons with native English speakers reading the same texts.
Native language drives stronger brain responses
The study also found that brain responses to semantic information were significantly stronger when participants read in Chinese, their native language, compared to English across every region of the semantic system. The difference was not caused by variations in brain signal quality between languages — it reflected genuinely deeper semantic processing in the native tongue.
Resolving a decades-old debate
The findings reconcile two competing theories that have long divided the field. Some behavioural studies had shown that bilinguals’ languages interfere with each other — a phenomenon suggesting a shared processing system. Others showed that emotional intensity, memory recall, and conceptual descriptions differ between a bilingual person’s two languages, suggesting separate systems.
The Berkeley study offers an explanation for both observations: the semantic system is shared, which explains cross-language interference, but each language modulates how meaning is encoded within that system, which explains language-specific behavioural differences.
The study was limited to six participants from a single population — Chinese-speaking university students who learned English as a second language — and the authors caution that results may differ for other language pairs, learning histories, or cultural contexts. They have made their full dataset and analysis code publicly available for other researchers to build on.
The research was funded by the National Science Foundation, the German Federal Ministry of Education and Research, and UC Berkeley.
Source: Chen, C. et al. (2026). “Bilingual language processing relies on shared semantic representations that are modulated by each language.” Proceedings of the National Academy of Sciences, 123(9). DOI: 10.1073/pnas.2503721123
