MIT neuroscientists discovered that sentences that are more complex, either because of unusual grammar or unexpected meaning, generate stronger responses in the brain’s key language processing centers. This discovery was made possible with the help of an artificial language network. Conversely, straightforward sentences barely engaged these centers, and nonsensical sequences of words produced minimal responses.
The researchers found that the brain network was most active when processing unfamiliar and puzzling sentences, but went quiet when reading something very straightforward. In other words, a certain level of language complexity is required to effectively engage the brain’s language processing centers.
In the study, the scientists focused on language-processing regions found in the left hemisphere of the brain. They compiled a set of 1,000 sentences from diverse sources and had human test subjects read them while the language network activity was measured with functional magnetic resonance imaging (fMRI). The same sentences were fed into an artificial language model, and its activation patterns were recorded.
A mapping model, or an “encoding model,” was subsequently developed to relate the human brain activation patterns to those observed in the artificial language model. Then, the encoding model was used to predict how the human language network would respond to any new sentence. Also, the model helped identify sentences that would generate maximal or minimal activity in the human brain’s language centers.
The researchers tested this approach on new human participants, revealing that these selected sentences indeed drive and suppress brain activity as predicted. This represented a novel approach to modulating brain activity during language processing and marked the first demonstration of such a method in brain sectors associated with higher cognition, such as the language-processing network.
Moreover, the researchers also discovered that sentences with high “surprisal,” or uncommonness, generated higher responses in the brain, and readers found such sentences more challenging to process. Linguistic complexity, assessed by adherence to English grammar rules and sentence plausibility, also influenced the language network’s responses. Sentences that required effort to understand but still made sense evoked the strongest responses.
In the future, the researchers aim to see if these findings apply to speakers of languages other than English and to explore stimuli that may activate language processing centers in the brain’s right hemisphere. The study was funded by a variety of educational, scientific, and research institutions and organizations.
