Search Abstracts | Symposia | Slide Sessions | Poster Sessions
Aspects of syntactic production represented in inferior frontal gyrus
Poster Session A - Sandbox Series, Thursday, October 24, 10:00 - 11:30 am, Great Hall 3 and 4
This poster is part of the Sandbox Series.
Tianhao Lei1, Prashanth Prakash2, Crispin Foli4, Joshua Glaser1, A. Bolu Ajiboye4,5, Marc Slutzky1,2,3; 1Department of Neurology, Northwestern University, IL, 2Department of Biomedical Engineering, Northwestern University, IL, 3Department of Physical Medicine & Rehabilitation, Northwestern University, IL, 4Department of Biomedical Engineering, School of Medicine, Case Western Reserve University, OH, 5FES Ctr of Excellence, Rehab. R&D Service, Louis Stokes Cleveland Dept of Veterans Affairs Medical Center,OH
Inferior frontal gyrus (IFG) is thought to be an important node in the human language network. Lesions in this area, especially neurodegenerative, can lead to expressive language deficits including omission of grammatical elements and simplified sentence structure (agrammatism). Grammatical syntax provides words with context that highlight differences in time (e.g., tense), grammatical number (singular vs. plural) or semantics. Some imaging studies have reported an association between damage to IFG and syntax comprehension. Evoked responses in IFG, mainly in pars triangularis, Brodmann’s area 45, showed some modulation with tense processing (Sahin et al, Science 2009). However, IFG’s (in particular area 44’s) role in grammatical processing remains unclear. Particularly unclear is the type of processing in which IFG participates and the amount of information it contains about syntax on the neuronal scale. Here, we recorded intracortical activity from a human participant as part of the Reconnecting the Hand and Arm to the Brain (ReHAB) clinical trial. We recorded broadband (30 kHz sampling rate) signals with a 64-channel microelectrode array placed on the anterior border of areas 44 and 6v. The participant was instructed to read a sentence that had one verb missing displayed on a monitor. This was followed by the display of a root verb that the participant was asked to conjugate, in agreement with the grammatical context of the sentence. We used two types of verb inflection: grammatical number (e.g. kicks vs. kick) and tense (e.g. kicked vs. kick). After a go cue, the participant spoke the conjugated word aloud. We extracted spike band power (300 Hz – 5 kHz) and binned it at 50ms and aligned all the trials to voice onset time for further analysis. We observed in trial-averaged data that some of the electrodes demonstrated early modulation 500 ms prior to voice onset, while some others showed activity after voice onset. For each type of syntax decoding, we built a supported vector machine decoder with a nonlinear kernel at each time step. We used 10 causal bins of history to construct the neural features, resulting in 640 features per time step. A small set of trials were reserved as a test set. To prevent overfitting, we used the first 10 principal components as features. We observed significantly above-chance accuracy starting approximately 450 ms prior to voice onset for grammatical number decoding, and 200 ms prior to voice onset for tense decoding. These results suggest that IFG may play a role in processing syntax for speech production.
Topic Areas: Language Production, Syntax and Combinatorial Semantics