In collaboration with an artificial language network, neuroscientists at Massachusetts Institute of Technology (MIT) have revealed what type of sentences most significantly engage the brain’s primary language processing areas. The study indicates that sentences featuring unusual grammar or unexpected meaning trigger a heightened response in these language-oriented regions, as opposed to more straightforward phrases, which only mildly stimulate these areas. Nonsensical word sequences are largely unstimulating to these regions as well.
Led by Evelina Fedorenko, senior author of the study and Associate Professor of Neuroscience at MIT, the team primarily concentrated on the language-processing regions situated in the brain’s left hemisphere. This includes Broca’s area and other sections of the left frontal and temporal lobes. The researchers gathered 1,000 sentences from various linguistic sources, which were read by human participants while their brain activity was monitored via functional magnetic resonance imaging (fMRI).
In addition to human linguistic response, an artificial language model processed the sentences, tracking its own activation patterns. The team developed an “encoding model” that correlated the brain activity seen in the human participants with those observed in the AI model. In turn, the model predicted the human brain’s response to any original sentence, based on the artificial language network’s response to the initial 1,000 sentences.
The researchers used this encoding model to form 500 new phrases that would amplify activity in the human brain, as well as sentences that would reduce activity. When presented to new human participants, these sentences indeed drove and suppressed brain activity as predicted.
The researchers analysed sentences based on eleven different linguistic parameters such as grammaticality, ease of visualization, emotional valence, and a concept referred to as “surprisal”, indicating sentence uncommonness. They found that sentences with higher surprisal and linguistic complexity, reflecting both grammatical coherence and plausibility, stimulated the most significant responses in the brain.
Extremely simple sentences or highly complex, nonsensical ones provoked little activation in the brain, whereas more challenging but comprehensible sentences required mental exertion and resulted in greater brain response.
The team plans to explore whether these findings can be extended to other languages and to assess the stimuli that might trigger the right hemisphere’s language processing regions. The research was funded by the MIT-IBM Watson AI Lab, the National Institutes of Health, and the Department of Brain and Cognitive Sciences at MIT, among others.
