Search Abstracts | Symposia | Slide Sessions | Poster Sessions
Intracranial signatures of syntax in the absence of conceptual semantics are isolated from non-linguistic recursive representations in frontotemporal language sites
Poster Session A, Friday, September 12, 11:00 am - 12:30 pm, Field House
Elliot Murphy1, Oscar Woolnough2, Nitin Tandon3; 1University of Texas Health Science Center
How the human brain constructs syntactic structures remains perhaps the most challenging topic in the neuroscience of language. We present an intracranial investigation of the spatiotemporal dynamics of basic syntactic structure processing in a cohort of 23 epilepsy patients. Data were acquired from subdural grid electrodes or stereotactically placed depth electrodes (sEEG). We had generous electrode coverage over frontotemporal language sites, in addition to parietal and ventrotemporal sites. Three patients had sEEG probes in thalamus (pulvinar, anterior nucleus, centromedian nucleus). Four patients were subject to an awake intraoperative surgery and conducted the task in the OR. We utilized a novel phrase reading paradigm involving Function/Content word pairs to manipulate conceptual content. We mixed combinations of these words, followed by a third phrase-completing word to remove wrap-up effects, analyzing effects at the onset of the structure-marking second word. Patients were instructed to engage metalinguistic judgements concerning whether the phrase was acceptable (2AFC). Stimuli were presented in RSVP (500ms per word). 50% of trials included a grammatical or semantic violation, either at the second or third word. Behavioral performance was high (>95%) and only correct trials were analyzed. We contrasted grammatical Function-Function trials (‘to the car’) with ungrammatical (‘they the car’) and asemantic (‘ti the car’) trials to focus on logico-semantic and syntactic information. We also contrasted phrases with and without conceptual content. After isolating structure-sensitive electrodes in high gamma power (70-150Hz) (>10% amplitude increase from baseline, then a two-sided paired t-test: alpha-level 0.01), we applied a more stringent test by isolating electrodes that were also sensitive to the contrast between Content-Function (‘swim the ocean’) versus unacceptable Function-Function trials. We isolated sites in left posterior temporal cortex and inferior frontal cortex that exhibited increased high gamma activity initiating between the 200-400ms window after the onset of the second word for grammatical structures, with effects lasting around 200ms. Greater activity for unacceptable trials was found in parietal cortex. A number of ventrotemporal sites were sensitive to semantic but not syntactic composition. We also discovered thalamocortical low-frequency phase-locking during the composition of phrase structure. Next, in three of these patients we used a separate paradigm to isolate visual recursive processing using a series of increasingly complex geometric shapes via RSVP that either violated or adhered to a recursive rule at the final shape. We used this to determine if the syntax-sensitive electrodes were also recruited for non-linguistic hierarchical structure. We discovered that most of the cortical sites sensitive to visual recursion were immediately neighboring but still distinct from the linguistic sites in the temporal lobe, but in frontal cortex there was a greater degree of overlap. These results indicate the rapid sensitivity of frontotemporal cortex to minimal syntactic structure building in the absence of conceptual semantics, with a potential role for thalamocortical control. This research thus provides neurobiologically plausible signatures of natural language syntax recoverable from isolated fragments of grammatical structure, helping to mitigate for common confounds (such as working memory, world knowledge and situation model construction) found in sentence-level and naturalistic paradigms.
Topic Areas: Syntax and Combinatorial Semantics, Reading