CAMBRIDGE, Mass. — How do our brains process language? It turns out that some sentences we read are simply more exciting for our brains than others.
Neuroscientists from Massachusetts Institute of Technology (MIT) are using artificial intelligence to discover how complex sentences stimulate the human brain. Employing an artificial language network, the team found that complex or unexpected sentences are more likely to activate the brain’s key language processing centers.
The study shows that straightforward sentences barely engage these regions, while nonsensical word sequences have little effect as well. For example, the sentence “Buy sell signals remains a particular” from the C4 language dataset elicited a strong response, while a simple sentence like “We were sitting on the couch” did not.
“The input has to be language-like enough to engage the system,” says study senior author Evelina Fedorenko, an associate professor of neuroscience at MIT and a member of MIT’s McGovern Institute for Brain Research, in a media release. “And then within that space, if things are really easy to process, then you don’t have much of a response. But if things get difficult, or surprising, if there’s an unusual construction or an unusual set of words that you’re maybe not very familiar with, then the network has to work harder.”
This research focused on language-processing regions in the left hemisphere of the brain, including Broca’s area and parts of the left frontal and temporal lobes. Greta Tuckute, the study’s lead author and an MIT graduate student, aimed to discover what types of sentences drive the left hemisphere language network.
“This language network is highly selective to language, but it’s been harder to actually figure out what is going on in these language regions,” says Tuckute. “We wanted to discover what kinds of sentences, what kinds of linguistic input, drive the left hemisphere language network.”
The team compiled 1,000 sentences from various sources and had five human participants read them while undergoing functional magnetic resonance imaging (fMRI), a technique that measures and maps brain activity. These sentences were also fed into a large language model similar to ChatGPT. Researchers then developed an “encoding model” that could predict how the human language network would respond to any new sentence based on the artificial language network’s response.
To test their model, researchers identified new sentences that would either maximize or minimize activity in the brain’s language network. These predictions were confirmed in tests with three new human participants.
The study also analyzed sentences based on several linguistic properties, like grammaticality, plausibility, emotional valence, and the ease of visualizing the sentence content. This analysis showed that sentences with higher “surprisal” – or uncommonness compared to other sentences – generated stronger brain responses. This aligns with previous findings that sentences with higher surprisal are more challenging to process.
The largest responses were elicited by sentences that somewhat make sense but require effort to understand, like “Jiffy Lube of — of therapies, yes,” from the Corpus of Contemporary American English dataset.
“We found that the sentences that elicit the highest brain response have a weird grammatical thing and/or a weird meaning,” notes Fedorenko. “There’s something slightly unusual about these sentences.”
The MIT team now plans to extend these findings to speakers of other languages and explore stimuli that may activate language processing regions in the brain’s right hemisphere.
The study is published in the journal Nature Human Behaviour.
You might also be interested in:
- The more languages you speak, the easier it is for the brain to learn more
- 3 in 5 people ‘articulate’ themselves with big words to sound smarter — even if they don’t know the meanings
- Ancient grammatical puzzle solved after 2,500 years by PhD student
