Search Abstracts | Symposia | Slide Sessions | Poster Sessions
Neurophysiological mechanisms of phrase compression at the phrase boundary
Poster Session B, Friday, September 12, 4:30 - 6:00 pm, Field House
Sophie Inchyna1, Christophe Smith, Jacob Odeh, Jenna Hooper, Bryant Barrentine, Nicole Bentley, Matthew Nelson; 1University of Alabama at Birmingham
In previous work, we developed the node-tracking framework of how the activation of neuronal populations in the brain might be construct tree-structure representations during sentence comprehension. We derived this from observations that while reading a sentence, population-level neural activity measured by high gamma power (70-150 Hz) in the left Middle Temporal Gyrus (MTG) builds up word-by-word within a phrase but declines sharply after the phrase boundary. Concomitantly, left Inferior Frontal Gyrus (IFG) activity increases transiently only at phrase boundaries, proportionate to the amount of decline in MTG activity. Here we refer to this activity pattern as the phrase compression activity profile. This framework has been highly impactful, but as of yet , the evidence behind it is merely correlational. Self-paced reading tasks have long been employed in the field as a way of measuring the dynamic processing requirements at each word in a sentence. In these tasks, longer Response Times (RTs) on a given word are interpreted as reflecting more processing required for that word. However, these tasks have seldom been paired with simultaneously recorded neural activity of any kind, and never with intracranial neural activity. We hypothesize that when doing so, participants will show longer per-word RTs at phrase boundaries, reflecting processing related to the phrase compression activity profile. This leads to a prediction that pauses at phrase boundaries will co-occur with the phrase compression activity profile. To test this, we recorded intracranial activity (sEEG and ECoG) in 43 epilepsy patient volunteers while they performed a self-paced reading task. Participants read 3 types of sentences: Object-relative sentences, Subject-relative sentences, and control sentences. After each sentence, participants saw an array of 4 pictures, and they were tasked with choosing the picture that matches the sentence they just read. Across patients, the neural activity in the left IFG around the time of the phrase boundary showed a trial-by-trial correlation with per-word RT’s at the phrase boundary that flipped it’s sign between before versus after presentation of the word. Higher IFG activity before the phrase boundary word onset correlated with faster response times at the phrase boundary word, suggesting that this anticipatory activity on these trials lead to shorter word RTs. However, higher IFG activity after the word onset correlated with slower word RT’s, suggesting that the lack of preparatory activity on these trials created a need for this area to “catch up” and perform these processes after the word onset, leading to a longer word RT. This effect was specific to the object-relative sentences, suggesting a particular role for the IFG for comprehension of these sentences, consistent with many other findings in the field. Neural activity in the left STS also correlated with longer pauses at the phrase boundary of object-relative sentences. Moreover, we also observed Ramping-up neural activity through-out the first phrase of the sentence, which is explained by and consistent with the node-tracking framework. Altogether, this shows a causal link between phrase comprehension measured by self-paced reading pauses at the phrase boundary, and the phrase compression neural activity profile.
Topic Areas: Syntax and Combinatorial Semantics,