UC Berkeley researchers create semantic map of brain

Jack Gallant/Courtesy

Related Posts

A group of UC Berkeley researchers has mapped a semantic atlas of a relatively uncharted region: the human brain.

In a study published Wednesday in the journal Nature, campus scientists mapped regions of the brain that each represent information about the meaning of language. The research suggests that related concepts are clustered into semantic domains found at multiple locations in the brain, according to Jack Gallant, campus psychology professor and a researcher on the study.

“Each location represents a constellation of related semantic concepts,” Gallant said.

The research has clinical implications for people affected by brain injuries, such as strokes, that affect language-related areas, according to the researchers. One of those researchers, campus psychology professor Frederic Theunissen, said the thoughts of those who are paralyzed or cannot speak could potentially be reproduced on a screen.

The research — funded by the National Science Foundation — was selected from a “competitive” pool in which not more than 15 percent of applicants receive funding, according to Ken Whang, the program director of computational neuroscience at the foundation.

Apart from charting the semantic maps, the research also showed that the organization of these maps was similar across different people, according to Theunissen. He added that a surprising result of the research was the number of brain areas involved in deriving the meaning of words.

The “bilateral” representation of language indicated in the research differs from traditional perceptions of language as being highly concentrated, or “lateralized,” in specific areas of the brain.

Gallant said past studies that supported the lateralization of language were largely focused on language production, not language comprehension.

In the study, seven participants were made to listen to stories from The Moth Radio Hour as their neural activity was tracked by an fMRI machine, which measures changes in brain activity using fluctuations in blood flow.

“One aspect of what stood out is these investigators were using a really novel, intensely data-driven strategy to get at a difficult analysis problem and also a very complex brain phenomenon,” said Whang of the study.

Whang noted that mapping out how semantic information is represented in the brain based on fMRI signals is “a hugely complex problem” because of the sizable vocabulary, structure and complexity present in language.

Campus linguistics professor George Lakoff said in an email that the results of the study should be interpreted narrowly and would mostly be applicable in medical contexts such as stroke victims.

According to Gallant, the researchers were planning to create models for other aspects of language, such as phonemes and syntax and even study different languages in the future.

“A globe is a round atlas, (and) if I give you a globe, you can ask any question,” Gallant said. “Our semantic atlas is the same way: You can ask any question you want.”

Contact Shradha Ganapathy at [email protected] and follow her on Twitter at @sganapathy_dc.