Search Abstracts | Symposia | Slide Sessions | Poster Sessions
Now you see it… ? : Agreement sensitivity in ‘at-a-glance’ reading in Spanish
Poster Session D, Saturday, September 13, 5:00 - 6:30 pm, Field House
This poster is part of the Sandbox Series.
Subhekshya Shrestha1, Dustin A. Chacón1; 1University of California, Santa Cruz
[INTRODUCTION] Theories of language processing have focused on ‘word-by-word’ processing. How do we process language ‘at-a-glance’? Whole sentences displayed for 200–300 ms are identified faster than word lists, non-word letter strings, or scrambled sentences, and EEG/MEG responses diverge 150–400 ms post-sentence onset. Proposed mechanisms include parallel activation of lexico-syntactic features, detection of basic constituent structure, or ‘filtering’ the percept using top-down expectations. This neural ‘sentence superiority effect’ (SSE) is insensitive to agreement errors in English (nurses {clean/*s} wound), which are robustly detected in EEG/MEG. This suggests ‘at-a-glance’ readers build less detailed syntactic parses. However, English subject-verb agreement frequently requires detection of a visually non-salient morpheme (–s, -Ø), and is a complex dependency between two phrases, dependent on an integrated sentential parse. We present pilot EEG data from a two-word parallel vs. serial presentation experiment on Spanish noun-adjective concord, a local relation marked on both words. Many nouns in Spanish inflect for gender overtly (chic-o ‘boy’, chic-a ‘girl’) while others do not (hombre ‘man’), but all trigger concord on following adjectives (chic-o alt-o ‘tall boy’, hombre alt-o ‘tall man’). Evaluating gender concord in non-transparent cases requires retrieval of the lexeme's gender. Are agreement errors in Spanish noun-adjective pairs ‘noticed’ in ‘at-a-glance’ reading ? If so, does ‘noticing’ agreement errors occur only for transparently marked nouns (before lexical access) or also for opaque nouns (concurrent/subsequent to lexical access)? [METHODS] Spanish speakers (N = 8) read 44 sets of noun-adjective pairs; brain data recorded by 256ch MagStim EEG. We have a 3✕3 design. We manipulated Transparency on word one: TRANSPARENT nouns with canonical gender inflections (aut-o ‘car’ [MASC]); OPAQUE nouns without(coche ‘car’ [MASC]), and PSEUDOMARKED nouns ending in a vowel typically corresponding to the other gender (moto ‘motorcycle’ [FEM]). We manipulated Concord on word two: CONCORD, NO-CONCORD adjectives (auto rápido/*rápida ‘fast car’) or another noun for a LIST condition (auto camión ‘car, truck’). Stimuli were presented in parallel (300ms, 500ms ISI) and serially (300ms/word, 500ms ISI) in separate blocks. Participants judged whether a second word pair matched the first (50% mismatch). [RESULTS] Spatio-temporal cluster-based permutation tests were conducted on sensor data 100–300ms from onset of word pair (parallel presentation) or adjective (serial presentation), with p < 0.05 threshold. During the N100-N170 complex, anterior and midline sensors showed a non-significant positivity for TRANSPARENT/CONCORD trials compared to the two ungrammatical TRANSPARENT conditions in parallel trials 100–300ms (p = 0.07), and in serial trials 200–300ms (p = 0.3). Similarly, during the N100-N170 complex, posterior midline and left lateral sensors showed a non-significant positivity for OPAQUE/CONCORD compared to the two ungrammatical OPAQUE conditions 100–150ms in parallel trials (p = 0.2), and a significant positivity 100–200ms (p = 0.05) in serial trials. No effects were observed for PSEUDOMARKED nouns. [CONCLUSION] Spanish readers may notice noun-adjective concord errors ‘at-a-glance’ in both transparent and opaque nouns, suggesting this occurs after or concurrent with lexical access. Failure to observe gender concord errors in pseudomarked cases, however, suggests that form-based inflectional processes and lexical access can conflict.
Topic Areas: Syntax and Combinatorial Semantics, Morphology