Poster D13, Friday, August 17, 4:45 – 6:30 pm, Room 2000AB

Neural dynamics of repetition-based learning of language comprehension parallels perceptual learning

Ayelet Gertsovski1, Olga Aizenberg1, Merav Ahissar1;1Hebrew University of Jerusalem

Detection of task-relevant regularities facilitates learning and allows the acquisition of expertise in perceptual tasks. It provides reliable predictions regarding upcoming stimuli, and reduces the required ad-hoc computations. The neural dynamics associated with formation of reliable predictions was previously studied for simple tone discrimination. It was found to be accompanied by a shift from fronto-parietal activity, associated with higher working-memory demands, to posterior superior temporal activity, related to successful detection of regularity introduced by a repeated reference tone in consecutive trials (Daikhin & Ahissar, 2015). We now asked whether similar neural dynamics is evident in learning from task-relevant repetition in a high-level language comprehension task. We conducted two fMRI experiments with the same behavioral protocol, but with slightly different acquisition parameters. In study 1 (n=17, TR of 1 s and 42 acquired slices of 3 mm) our scanning did not fully cover the cerebellum. In study 2 we modified our scanning protocol (n=20, TR of 1.2 s in order to collect 51 slices) to fully cover the cerebellum. Participants heard a sentence with a novel semantic content, then saw an illustrative cartoon, and had to determine whether it matches the sentence. To reveal the areas activated during an effortful on-line task, we compared two syntactic structures known to differ in processing difficulty: object-extracted relative clauses (ORs, e.g. “Mr. Rectangle is the creature that Mr. Circle pushes”) and subject-extracted relative clauses (SRs, e.g. “Mr. Circle is the creature that pushes Mr. Rectangle”). To reveal the areas sensitive to regularity (repeating either the doer or the doee of the action in consecutive SR sentences), we compared SR trials with and without repetition. As expected, behavioral results in both studies showed that participants performed worse in ORs compared with SRs, and benefited from trials with repetition. They also benefited more from repetition of the doer compared with repetition of the doee. The contrast between ORs and SRs revealed known language areas in the left hemisphere, and was specifically associated with a strong left frontal activity. The contrast between non-repeated and repeated SR sentences revealed activation of posterior superior temporal areas (and not frontal areas). The cerebellum was activated both in the OR versus SR contrast and in the doer versus doee repetition contrast. These results suggest that fast improvement in sentence comprehension is similar both behaviorally and in brain distribution to fast learning in simple auditory discriminations. It is facilitated by item-specific repetition and is associated with changes in activity in posterior stimuli-specific brain areas, presumably encoding the context and forming stimuli-specific predictions. This suggests that the functional division of labor between frontal and posterior language areas reflects the level of sentence-specific expertise rather than the domain of linguistic analysis. Our results further imply that the cerebellum is involved both in verbal working memory processes, as was previously shown (e.g. in Marvel & Desmond, 2010), and in fast learning from repetition.

Topic Area: Phonology and Phonological Working Memory