Search Abstracts | Symposia | Slide Sessions | Poster Sessions
Processing of Russian multiple wh-questions challenges the presumed syntactic nature of established ERP effects
Poster Session E, Sunday, September 14, 11:00 am - 12:30 pm, Field House
Jenna Baranov1,2, Robert Kluender1, Ellen Lau2; 1UCSD, 2UMD
Long-distance dependencies are known to elicit sustained anterior negativity (SAN) and P600 effects, but the exact interpretation of these effects is still up for debate. We used Russian to conduct the first EEG study of multiple long-distance dependencies in order to disambiguate the functional interpretation of the following: (1) SAN: SAN between a displaced constituent and its gap is often interpreted as an index of syntactic working memory load. Recent studies have questioned this functional interpretation, either failing to replicate the same effect or reporting a decrease in amplitude of the effect due to contextual support. (2) P600: Positive-going voltage deflection at the gap site has been interpreted as an index of syntactic integration difficulty reflecting the number of syntactic operations involved in closing the dependency. (3) Phasic LAN: Several studies have reported transient LAN effects at the post-gap position indexing a retrieval process, rendering the syntactic integration interpretation of the P600 at the preceding position problematic. Previous ERP studies compare the presence of one long-distance dependency to its absence, making it hard to distinguish between different possible interpretations of known ERP effects. However, Russian allows multiple wh-elements to be displaced to the left periphery of the sentence, creating the possibility of a three-way distinction: no dependency vs. a single dependency vs. multiple dependencies of the same kind. As the complementizer ‘that’ (čto) is syncretic with ‘what’ in Russian, we used why-questions — arguably forming no dependency — as controls. 120 stimulus and 120 filler sentences were presented via RSVP to native Russian speakers (n=27). We ran repeated-measures by-subjects ANOVAs to test how the number of dependencies affected the amplitude of (1-3): (1') Multiple long-distance dependencies elicited a left-lateralized SAN effect (p < 0.05) compared to single and no-dependency conditions, which did not differ from each other. (2') At the verb, multiple and single dependencies did not differ from each other, but each elicited a P600 effect relative to the no-dependency condition (p < 0.05). (3') At the post-gap position, multiple and single dependencies again did not differ from each other, but each elicited a phasic LAN effect compared to the no-dependency condition (p < 0.05). (1’) suggests that SAN is not simply a response to the formation of a syntactic dependency, and is thus not purely syntactic in nature. In contrast, the lack of a P600 difference between multiple and single wh-dependencies in (2’) at the end of the dependency casts doubt on interpreting the P600 as indexing syntactic integration difficulty: there is more syntactic material to integrate in a multiple wh-dependency. We suggest the P600 may instead reflect an all-or-nothing gap identification mechanism insensitive to the amount of associated syntactic information. This is also more consistent with a subsequent LAN effect in (3’) indexing retrieval at the post-gap position in multiple and single dependencies. The lack of a difference between multiple and single dependencies here suggests that, if LAN indeed indexes retrieval, the nature of the retrieved information must be independent of the number of completed dependencies.
Topic Areas: Syntax and Combinatorial Semantics,