Search Abstracts | Symposia | Slide Sessions | Poster Sessions
Tracking neural trajectories of word combinations in a flexible word-order language
Poster Session E, Friday, October 2, 11:00 am - 1:00 pm, Wangari Maathai
This poster is part of the Sandbox Series.
Gyu-Hwan Lee1, Liina Pylkkänen1; 1New York University
INTRODUCTION Languages allow multiple ways of expressing the same meaning. Although fixed-order languages dominate the experimental literature on sentence processing, word-order flexibility is widespread cross-linguistically: 13.7% of sampled languages lacked a dominant order of subject, object, and verb (Dryer 2013). A typologically diverse set of languages, including Japanese, Hindi, and Turkish, permit scrambling, i.e., the reordering of arguments without changing the propositional meaning (Karimi 2003). While we have studies on the localization of scrambling and its effects on sentence processing (Fiebach et al. 2004; Tanaka et al. 2017; Schlesewsky et al. 2003), it is not currently understood how the brain extracts a shared meaning from sentences with different word orders. DESIGN We address this question using Korean noun-quantifier-adjective (NQA) constructions. Exploiting the extensive word-order permutations available in Korean (Ko, 2018), we constructed four word-order variations. The Canonical word order for this type of sentence is NQA, as in akinun motwu kwiyepta ‘babies all cute.’ Two other conditions involve scrambled word orders, where one is farther from the canonical word order than the other. The LessScrambled condition has the word order of QAN while the word order in the MoreScrambled condition is AQN. The fourth word order of QNA is Ungrammatical, due to a grammatical constraint for the type of quantifiers involved in the construction (Ko 2014). The four sentence conditions are each matched with a list condition, composed of syllable-count matched nominalized adjectives, serving as a baseline for behavioral and neural effects. Participants process the sentences under rapid parallel visual presentation (RPVP), where an entire 3-word string appears for 300ms, followed by a matching task. This paradigm eliminates serial order constraints, allowing the brain to set the natural order of computations. BEHAVIORAL RESULTS An online behavioral experiment (N=33) showed that the sentence superiority effect (SSE; Snell & Grainger 2017) was strongest in Canonical and LessScrambled conditions, attenuated in MoreScrambled, and absent in Ungrammatical, supporting the graded and categorical effects of scrambled word orders. MEG PLAN Evidence from recent RPVP studies involving the same simple matching task points toward a neural capacity for the rapid detection of grammatical patterns (Fallon & Pylkkänen, 2024) and the subsequent combination of words (Flower & Pylkkänen, 2026). The combination of word order permutations with RPVP in this study provides a unique opportunity to track the neural emergence of a shared meaning from strings of identical words in different orders, when they are all presented at once. Current work involves MEG to track the temporal dynamics of sentence integration. Data collection is currently underway and is expected to be completed by the conference. First, we will localize the neural SSE across conditions to identify brain areas sensitive to word-order displacements and ungrammaticality. To determine the sequence of word integration, we will train neural decoders on separate RPVP trials of constituent word dyads (e.g., N+Q, N+A). By tracking when these two-word representations are reinstated during the processing of full three-word sentences, we aim to characterize how the brain serializes integration in the absence of incremental input.
Topic Areas: Syntax and Combinatorial Semantics, Speech Perception