Presentation

Search Abstracts | Symposia | Slide Sessions | Poster Sessions

Whisps and whisp-ers in the Brain: A crossmodal investigation into amodal representations in English morphological processing using HD-EEG

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.

Tyson Jordan¹, Dustin Chacón^2,1, Donald Dunagan¹; ¹University of Georgia, ²University of California, Santa Cruz

[INTRODUCTION] How does the brain process novel combination of morphemes? One key idea is that complex word forms are 'decomposed' into their parts. MEG signals from visual word form area (VWFA) show greater activity for morphologically complex forms ('baker' = 'bake-er') compared to form-matched simple forms ('winter') (e.g., Zweig & Pylkkänen 2009). Other evidence shows that morphological decomposition overapplies: 'brother' decomposes into 'broth-er' (Gwilliams & Marantz 2018; Rastle et al. 2004), suggesting form-based decomposition. Our question is: If morphologically complex forms like 'baker' involve reaccessing 'bake', then can we detect shared patterns of activity for 'bake'-'baker', indicating a shared representation? What about pseudocomplex words, like 'hung'-'hunger'? We also ask: are these shared representations detectable across modalities (e.g. reading vs. speech)? We present preliminary, 'sand-box' findings in reading only. [METHODS] [Materials] We prepared 90 words in three key conditions: Complex ('baker'), Pseudocomplex:ExistingStem ('hunger'), and Pseudocomplex:NoStem ('winter'). We also included their 1-syllable (pseudo)stems ('bake', 'hung', 'wint'). Filler items consisted of two- and one-syllable (pseudo)words (50% grammatical, total: 360 words). Key stimuli were controlled for length, whole-word and stem frequency, and phonological and orthographic neighborhood density. Stimuli were prepared as text and audio+video recordings of a speaker producing the word. [Procedure] Participants (N = 4) conducted lexical decision tasks in 4 modalities: text, audio, audio+video, and silent video (order counterbalanced). [RESULTS] [Behavioral] Participants performed near ceiling, with no detectable difference in accuracy (ps > 0.50) or response time (ps > 0.50) in key conditions. [EEG] Spatio-temporal cluster-based permutation tests with 1-factor ANOVAs (Complex, Pseudocomplex:ExistingStem, Pseudocomplex:NoStem) identified a cluster from 267–362ms (p = 0.14) in left parietal and posterior sensors, showing distinct EEG activity for Complex, Pseudocomplex:ExistingStem, and Pseudocomplex:NoStem words. We then conducted representational similarity analysis (RSA) to determine whether there are shared representations for stem-complex word pairs ('sing-'singer'). We subsetted the data to the Complex and Pseudocomplex stimuli, and their monomorphemic (pseudo)stems. Then, we fit representational similarity matrices (RSMs), one treating Complex words and their stems as similar (Morphological Identity; F(bake, baker) = 1.0, F(hung, hunger) = 0.0), and one treating Complex and Pseudocomplex words as similar to their (pseudo)stems (Orthographic Identity; F'(bake, baker) = 1.0, F'(hung, hunger) = 1.0). We used a searchlight RSA procedure (50ms, 30mm) to compute correlations between EEG data and the RSMs per subject. We conducted group-level t-tests to determine whether these correlations differed from zero. We found a correlation between Morphological Identity in left frontal sensors 182–400ms (p=0.2), but no correlation for Orthographic Identity. [CONCLUSION] Models of morphological decomposition hold that morphologically complex forms ('baker') involve decomposition into and activation of the stem ('bake'). These processes may only apply to morphologically complex words, or they may overapply in pseudocomplex cases. Here, we presented preliminary RSA suggesting that pairs like 'bake'-'baker' involve accessing similar representations in reading ~200–400ms, but not pseudocomplex pairs like 'hung'-'hunger'. Conference presentation will also describe results across 4 modalities, plus preliminary source space equivalences.

Topic Areas: Morphology,