Summary: New intracranial recordings reveal how the brain groups words into phrases during reading.
Source: Cornell University
How does the brain move from reading individual words to understanding full sentences?
An international research team, including cognitive scientists from Cornell, presents direct physiological evidence showing how the human brain constructs phrase structure in real time as people read. The findings offer a rare window into the neural dynamics that underlie sentence comprehension and support long-standing theories in psycholinguistics about hierarchical phrase structure.
“Phrase structure indeed appears as a major determinant of the dynamic profile of brain activity in language areas,” the authors report, summarizing results published in the April 17 issue of the Proceedings of the National Academy of Sciences (PNAS).
Co-author John T. Hale, an associate professor of linguistics at Cornell, said the study provides unusually direct evidence linking neural signals to phrase structure during language understanding. The results also reinforce the idea that readers anticipate upcoming words or phrases based on context, but they show that brain activity in language areas cannot be explained solely by predictability.
To track language-related brain activity with high temporal and spatial precision, the researchers relied on intracranial recordings from patients who had electrodes implanted for clinical reasons. A dozen volunteers with epilepsy read simple sentences presented one word at a time in their native language (English or French) while researchers recorded activity across language-sensitive regions of the left hemisphere.
Sentences were short—three to ten words long—and included straightforward examples such as “Bill Gates slept in Paris.” As participants processed each word, the team measured high-gamma power and other neural signals to determine when and where activity rose or fell as the sentence unfolded.
The central pattern that emerged was consistent across multiple left-hemisphere language regions: neural activity increased with each incoming word but dropped sharply whenever a grammatically coherent phrase could be formed by merging the words read so far. In other words, the brain appeared to accumulate individual words and sustain activity while retaining them, and then compress those stored elements into a single phrase, producing a sudden reduction in activity once the merge occurred.
Hale explained the interpretation: maintaining words in short-term memory while waiting to combine them into meaningful units requires ongoing neural resources. When those words can be merged into a single phrase, the memory burden is reduced and neural activity declines. This pattern supports models in which a merge operation builds nested phrase structure incrementally during comprehension.
The researchers compared their intracranial recordings to several computational models of word-by-word parsing. Models that explicitly included an operation that merges words into larger multiword phrases produced the best fit to the observed neural data across superior temporal and inferior frontal language areas. By contrast, shallower models based solely on sequential transition probabilities over lexical or syntactic categories captured activity only in more posterior temporal regions.
Formal model comparisons indicated that multiword phrase construction provided a better account of activity at most sampled sites than purely probability-based alternatives. The overall pattern of results is compatible with implementations such as a bottom-up or left-corner parser that compresses syntactically well-formed sequences into a hierarchical representation as language input arrives.
The research team included Matthew Nelson, Christophe Pallier and Stanislas Dehaene of the NeuroSpin Center in France; Imen El Karoui, Laurent Cohen and Lionel Naccache of the French National Center for Medical Research; Kristof Giber and Sydney Cash of Harvard Medical School; Xiaofang Yang of Stanford University; Hilda Koopman of the University of California, Los Angeles; and John T. Hale of Cornell University.
Source: Susan Kelley – Cornell University
Image source: NeuroscienceNews.com image is in the public domain.
Original research: The study is reported as “Neurophysiological dynamics of phrase-structure building during sentence processing” and was published online April 17, 2017 in PNAS (doi:10.1073/pnas.1701590114).
Abstract
Neurophysiological dynamics of phrase-structure building during sentence processing
Although sentences unfold sequentially, one word at a time, most linguistic theories propose that their underlying syntactic structure involves a tree of nested phrases rather than a linear sequence of words. Whether and how the brain builds such structures, however, remains largely unknown. Here, we used human intracranial recordings and visual word-by-word presentation of sentences and word lists to investigate how left-hemispheric brain activity varies during the formation of phrase structures. In a broad set of language-related areas, comprising multiple superior temporal and inferior frontal sites, high-gamma power increased with each successive word in a sentence but decreased suddenly whenever words could be merged into a phrase. Regression analyses showed that each additional word or multiword phrase contributed a similar amount of additional brain activity, providing evidence for a merge operation that applies equally to linguistic objects of arbitrary complexity. More superficial models of language, based solely on sequential transition probability over lexical and syntactic categories, only captured activity in the posterior middle temporal gyrus. Formal model comparison indicated that the model of multiword phrase construction provided a better fit than probability-based models at most sites in superior temporal and inferior frontal cortices. Activity in those regions was consistent with a neural implementation of a bottom-up or left-corner parser of the incoming language stream. Our results provide initial intracranial evidence for the neurophysiological reality of the merge operation postulated by linguists and suggest that the brain compresses syntactically well-formed sequences of words into a hierarchy of nested phrases.
“Neurophysiological dynamics of phrase-structure building during sentence processing” by Matthew J. Nelson, Imen El Karoui, Kristof Giber, Xiaofang Yang, Laurent Cohen, Hilda Koopman, Sydney S. Cash, Lionel Naccache, John T. Hale, Christophe Pallier, and Stanislas Dehaene in PNAS. Published online April 17, 2017. doi:10.1073/pnas.1701590114