Search Abstracts | Symposia | Slide Sessions | Poster Sessions
Orthographic decomposition vs. Lexical access in the VWFA
Poster D59 in Poster Session D, Thursday, October 1, 4:30 - 6:30 pm, Wangari Maathai and Lise Girardin
This poster is part of the Sandbox Series.
Stefan Pophristic1, Alec Marantz1; 1New York University
Introduction. The visual word form area (VWFA) is one of the first brain regions responsible for language-specific processing of visually presented words—yet there is still no clear model of what computations are done in this area. In this study, we use a novel variable of conditional well-formedness to test: 1) whether the VWFA engages in purely orthographic morphological decomposition or early form-based lexical access; and 2) whether the same computations apply to real vs. nonce/pseudo-words. We focus on infinitival Spanish verbs, which obligatorily occur with one of three suffixes (-ar/-er/-ir). Using a phonotactic learner (HayesWilson2008), we calculate the probability distribution of letters in stems conditioned on the suffix which the stem occurs with (e.g. phonological/orthographic generalizations for all stems that take -er suffix). We test whether deviance from this distribution modulates VWFA activity for verbs in a visual lexical decision task (LDT). Question 1. Studies using Magnetoencephalography (MEG) argue that the VWFA engages in the decomposition stage of lexical processing whereby a word-form is decomposed into its morphological constituents (farmer to farm+er), indexed through the M170 component (WrayEtAl2022). These studies show that the VWFA must access some linguistic knowledge, as unique-stem words like “vulner-able” are decomposed (despite the stem “vulner” appearing nowhere else in the language) (GwilliamsEtAl2018), whereas pseudo-affixed words like “winter” (where “wint” isn’t a morpheme but “er” is) are not (ZweigEtAl2009). It’s unclear whether these results reflect 1) top-down linguistic knowledge on an otherwise orthographic process or 2) form-based lexical access. Testing how conditional well-formedness modulates the M170 can help tease these two possibilities apart. If the VWFA engages in purely orthographic decomposition, then conditionally well-formed stems (for their suffix), as a measure of orthographic regularity, should facilitate decomposition and result in decreased M170 amplitudes, as shown with bigram/trigram variables (SolomyakMarantz2009). Otherwise, if the VWFA engages in form-based lexical access, then conditional well-formedness should reflect lexical competition effects (CarreirasEtAl1997) and result in increased M170 amplitudes. Question 2. While studies find identical VWFA amplitudes for real vs. nonce mono-morphemic words (VinckierEtAl2007), since “winter” is not decomposed (presumably due to the non-existence of the stem “wint”), it is unclear whether the VWFA decomposes nonce words like “wuger” (nonce stem “wug”). If the VWFA decomposes nonce words, we expect nonce word well-formedness values to have the same effect as for real words. Otherwise, if the VWFA doesn’t decompose nonce words, the well-formedness values should have no effect on M170 amplitude. Methods. 30 participants will complete an LDT of 504 real/nonce words and a functional localizer task (WhiteEtAl2023) with simultaneous MEG recording. To establish an fROI, we will conduct spatio-temporal permutation clustering over the 100-250ms time window in the occipital/temporal lobes (GwilliamsEtAl2018). The dependent variable(s) for linear mixed-effects models are dSPM values averaged by trial/participant over the spatial/temporal extent of the group-level localizer cluster. We test: Q1) whether the well-formedness of real-words positively or negatively predicts the dependent variable; Q2) whether well-formedness of nonce-words is significant in same direction as for the real-word model. Data collection will begin in a few weeks.
Topic Areas: Morphology, Phonology