Search Abstracts | Symposia | Slide Sessions | Poster Sessions
Disentangling syntax and prosody in sentence comprehension: The case of cross-serial and nested dependencies
Poster A50 in Poster Session A - Sandbox Series, Thursday, October 24, 10:00 - 11:30 am, Great Hall 4
This poster is part of the Sandbox Series.
Catalina Torres1, Sebastian Sauppe, Chantal Oderbolz, Martin Meyer, Balthasar Bickel; 1University of Zurich, 2NCCR Evolving Language, 3ISLE Interdisciplinary Study of Language Evolution
Non-adjacent dependencies in syntax are challenging to process because they separate in time what belongs together in hierarchical structure. The challenge is expected to be stronger for crossed (a1b2a1b2) than for nested (a1b2b2a1) dependencies because more memory stacks are recruited, i.e., the challenge increases with the Chomsky hierarchy. Simultaneously, this challenge is expected to be mitigated by prosodic information that helps parse syntax during listening through the highlighting of syntactic boundaries. We report a planned EEG experiment on the comprehension of adjoined, nested, and crossed dependencies in Swiss German (SG), using a prosodic manipulation. SG provides a unique test case because it exhibits these three dependency types by varying word order in sentence-final verb phrases, keeping compositional semantics and information structure constant. We use sentence-final clusters of two verbs, each with their own object (in pseudo-English translation of ‘that Maria saw Manu marinate the lamb’, CROSSED: that Maria Manu1 the lamb2 saw1 marinate2; NESTED: that Maria Manu1 the lamb2 marinate2 saw1; ADJOINED: that Maria Manu1 saw1 the lamb2 marinate2). Stimulus sentences were normed for naturalness and variations were created, ensuring the lexical material was constant. A female speaker of SG was recorded, and the speech was acoustically manipulated by resynthesizing duration and fundamental frequency to reduce the acoustic strength of boundaries and prominences. Forty participants will listen to 180 unique critical sentences (50% prosodically manipulated) while EEG is being recorded from 64 electrodes. The combination of word-order variation and natural vs. manipulated speech allows us to disentangle the roles and neural underpinnings of syntax and prosody in language comprehension. The cross-linguistic rarity and complexity of crossed dependencies should lead to increased processing demands. We test this by analysing event-related synchronisation/desynchronisation (ERS/ERD) in individually-defined theta, alpha, and beta bands, time-locked to the presentation of the main and the auxiliary verbs. Theta ERS as domain-general error monitoring system should be sensitive to encountering unexpected word orders. Alpha ERD reflects the release of inhibition in the language network, allowing more intense syntactic processing for crossed over nested/adjoined dependencies. Beta ERD reflects the updating of sentence-level representations when an unexpected or demanding dependency is to be parsed. If the availability of prosodic information facilitates, and possibly foreshadows, syntactic parsing, the ERS/ERD effects should be modulated by our prosodic manipulation. Furthermore, in the unmanipulated prosodic condition, we expect increased neural synchronisation between brain activity and the speech signal, in the delta band. The regression-based analysis takes linguistic surprisal effects into account and further exploratory analysis includes functional connectivity and microstate analyses to shed light on the transient neural network dynamics supporting syntactic and prosodic language comprehension. This study contributes to our understanding of real-time syntactic processing by investigating the variable linearisation of a SG syntactic structure. Additionally, our experimental paradigm allows us to disentangle syntactic from prosodic parsing. Finally, our study allows us to discuss effort during syntactic processing in a meaningful way since all participants are confronted with the three linearised verb clusters, and we don’t compare across different speaker groups, as previously done.
Topic Areas: Prosody, Syntax and Combinatorial Semantics