A recent study from MIT has uncovered that the human brain’s principal language processing centers are most activated while reading complex, unusual sentences. The artificial language network assisted study revealed that the more intricate a sentence was, either through unconventional grammar or unexpected meaning, the more these language processing centers were activated. In contrast, simple phrases yielded barely any engagement, just as nonsensical sequences of words had minimal effect. For the study, sentences from a large variety of sources were gathered and analyzed using Functional Magnetic Resonance Imaging alongside an artificial language model. The results were mapped onto an “encoding model” which was then utilized to predict how the human brain would respond to new sentences.
Lead author, Greta Tuckute, and the team used this model to determine which sentences would produce maximum (“drive” sentences) vs. minimum (“suppress” sentences) brain activity. The researchers found that these new sentences had the desired effect on a group of new participants. They then analyzed these sentences based on 11 linguistic properties to identify what contributes to a sentence’s ability to elicit activity in the brain. The findings indicated that sentences with higher ‘surprisal’ or uncommonness produced greater responses in the brain. Linguistic complexity, a measure of how well a sentence abides by English grammar rules and the sense it makes, was another contributing factor. Those sentences that were either incredibly simple or overly complex, to the point of not making any sense, resulted in almost no activation in the language network.
Senior author Evelina Fedorenko pointed out that sentences prompting the highest brain responses displayed unusual grammar or meanings. The researchers hope to extend these findings to speakers of other languages and also explore what form of stimuli may activate other language processing regions in the brain. The study is expected to advance our understanding of how the human brain processes language and how this could be potentially applied in areas such as language learning and recovery from stroke or injury. The research was financially supported by an Amazon Fellowship from the Science Hub, an International Doctoral Fellowship from the American Association of University Women, the MIT-IBM Watson AI Lab, the National Institutes of Health, and others. The study’s findings were published in the peer-reviewed journal, Nature Human Behavior.
