Neuroscientists at MIT, with the aid of an artificial language network, have determined the type of sentences that most likely activate the brain’s main language processing centers. The recently published study demonstrates that sentences which are more complex, either due to exceptional grammar or unexpected meanings, stimulate stronger responses in these regions. On the other hand, highly direct sentences barely engage these areas just as nonsensical word sequences.
The researchers found that when reading unusual sentences such as “Buy sell signals remains a particular,” this particular brain network was most active. However, it was less responsive when reading straightforward sentences. According to Evelina Fedorenko, the Associate Professor of Neuroscience at MIT, language has to be complex enough to engage the system, and the more challenging the processing requirement, the harder the network works.
The focus of the study was the language-processing regions found in the left hemisphere of the brain. To understand the kind of linguistic input that drives these regions, the researchers analyzed 1,000 sentences from various sources. These sentences were read by five participants while the team measured their language network activity using functional magnetic resonance imaging (fMRI). The same sentences were then input into a large language model, with the activation patterns being measured in response to each sentence.
An “encoding model” was then developed, which links the activation patterns seen in the human brain with those observed in the artificial language model. This model was used to predict how the human language network would respond to a new sentence. Following this, the researchers identified 500 new sentences that would generate maximum activity in the human brain. These sentences were successfully used in a group of three new participants to effectively drive and suppress brain activity.
Further analysis of these sentences revealed that higher “surprisal” sentences generated greater responses in the brain. The researchers found that the most significant responses came from sentences that made some sense but required work to understand. Fedorenko commented that these sentences have an element of weirdness in terms of grammar or meaning.
The researchers plan to test these findings in speakers of other languages and explore stimuli that may activate language processing regions in the brain’s right hemisphere. This research was funded by various organizations, including an Amazon Fellowship from the Science Hub and the MIT-IBM Watson AI Lab.
