Search Abstracts | Symposia | Slide Sessions | Poster Sessions
An explicit comparison of word-internal and word-external semantic violations: An HD-EEG study in English
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.
Dustin A. Chacón1, Tyson Jordan2, Donald G. Dunagan2; 1UC Santa Cruz, 2University of Georgia
[INTRODUCTION] Cognitive scientists have long stressed the brain's capacity for syntax as the fundamental sine qua non of language. However, it is difficult to disentangle syntactic and semantic processing (Pylkkänen 2019; Shain et al. 2024), and compositional processes may be shared between words and sentences (e.g., Fruchter & Marantz 2015; Gwilliams 2020). But, semantic anomalies appear to engender different responses within and between words: Words in unexpected contexts yield neural responses around 200-400ms ('N400'; 'M350'), localizing to left superior temporal lobe (LSTL) (Lau et al. 2008). Conversely, Neophytou et al. (2018, et sub) found that MEG responses to semantic violations internal to a word, e.g., 're-laugh', result in greater orbitofrontal cortex (OFC) activity, ~400–500ms. Our question is: Could similar brain areas subserve detection of semantic anomalies internal to words and between words, when explicitly compared? We suggest 'yes.' [METHODS] [Materials] The suffix -able requires a verb with an internal argument, and the resulting adjective must combine with a noun that can serve this role. The suffix -ing combines with any verb, but the resulting adjective normally prefers a noun that satisfies the external argument. We produced 60 pairs of +Grammatical/-Ing and –Grammatical/-Able stimuli ('the glowing diamond'; '*the glowable diamond'), and 60 pairs of +Grammatical/-Able and –Grammatical/-Ing stimuli ('the stealable jewel, '*the stealing jewel). Grammatical stimuli were controlled for verb-noun semantic association, noun surprisal, verb–adjective transition probability (p > 0.10). There were also 120 filler trials, 50% grammatical. [Participants] N = 6 participants saw the stimuli for 300ms + 500ms of blank screen. They then judged whether the phrase was grammatical using a keyboard. EEG signals were recorded using a 128ch BrainVIsion actiChamp+ system. [RESULTS] [Sensor Space] Sensor-space spatio-temporal cluster-based permutation-tests (Maris & Oostenveld 2007) were conducted using 2 × 2 ANOVAs, crossing ±Grammatical × Suffix, 200–600ms. A cluster was identified over right posterior sensors 348–600ms showing greater negativity for –Grammatical trials over +Grammatical trials (p = 0.01). A separate cluster was identified over left fronto-parietal sensors 276–544ms, showing greater negativity for -Able stimuli compared to -Ing stimuli (p < 0.01). [Source Space] High-density EEG arrays can produce source localization results similar to MEG (e.g., Klamer et al. 2014). Sensor positions were reassigned from default positions to positions identified using a CapTrak digitizer, then coregistered with fsaverage template. Forward solutions were computed, estimating different conductivity for outer-skull, inner-skull, and scalp layers. Inverse solutions were computed using sLORETA (Pascual-Marqui 2002). Spatio-temporal cluster-based permutation tests were conducted over left posterior and middle temporal lobes (LPTL+MTL), and bilateral OFC and constrained to the 348–600ms ±Grammatical cluster. A significant ±Grammatical cluster was identified in LPTL+MTL, showing greater activity in LPTL 360–449ms (p = 0.01) for +Grammatical vs. –Grammatical conditions. A significant cluster of Suffix was identified in OFC showing greater activity in left OFC 348–457ms for -Able vs. -Ing stimuli. [CONCLUSION] Preliminary data shows that LSTG responds to semantic anomalies regardless of whether they arise from illicit stem-suffix relations within a word or adjective-noun combinations between words.
Topic Areas: Syntax and Combinatorial Semantics, Morphology