Slide Slam B17
An investigation of the neuronal signature of word order effects in Russian
Anastasia Stoops1, Jeff Green2; 1University of Illinois at Urbana-Champaign, 2Brigham Young University
Restrictive context can predict a specific word (e.g. Kuperberg & Jaeger, 2016), but is infrequent in naturalistic conditions. A graded prediction of upcoming word features is a more robust mechanism (Luke & Christianson, 2016). Electrophysiological studies have identified a positive event-related (ERP) component with a peak between 500-700ms (P600) that reflects the need to update the mental representation of an entire utterance when the upcoming input renders the previous mental representation untenable (Brouwer, et al., 2012). Posterior P600 was reported for structural and semantic revisions: grammatical (Frederici et al., 1996) and semantic violations (sometimes preceded by N400; see Kuperberg, 2013 for a review) or unexpected garden-path sentences (Qian et al., 2018). Frontal P600 is associated with syntactically complex grammatical sentences without the semantic revision (Jackson et al., 2020; Kaan & Swab, 2003). However, prior neurological studies used strict word order Germanic languages where the ability to control the target length or the word count and length of the preceding context is difficult. Moreover, the ERP is an aggregate signal that pulls on different frequency bands. Thus, time-frequency measures pinpoint what frequency oscillation at what topography contributes to the ERP components. This project investigated the role of the preceding word order on the target word, controlling for word length and the preceding context. We examined neuronal brain responses evoked by violations of canonicity of word order and grammaticality in Russian. Method: Russian's free word order and rich inflectional system allowed us to examine separate contributions of canonicity and grammaticality. Electrophysiological responses were recorded from 12 Native Russian speakers (6 female; mean age=25) during word-by-word (500 SOA) reading. Grammaticality and canonicity were crossed in a 2x2 Latin square design (SVO[baseline]/SVS/OVS/OVO; a total of 160 4-sentence units). Semantic plausibility and predictability of the arguments was controlled through norming. Word order frequency: SVO(60%)/OVS(30%) (Lobanova, 2011). ERPs were baseline corrected to the first noun (-100:0). Time-frequency analyses used a Slepian sequence (multitaper filter with the Hanning taper) with the adaptive window in Fieldtrip open-source software (Oostenveld et al., 2011). ERP analyses revealed grammaticality effects in the P600 window with frontocentral topography, but only for the canonical word order (SVS>SVO, p<.01). Evoked power (EP) and intertrial coherence (ITC) phase-locked to the target word revealed analogous interaction between grammaticality and canonicity in the delta band (2-3hz) correlated to the P600 time window with elevated delta level for ungrammatical endings in the canonical word order SVS > SVO > OVS > OVO (p<.01). First, grammaticality elicited an effect only in the canonical SVO with the ungrammatical second subject (SVS) but not the second object (OVO) in the noncanonical OVS. Second, our study enriches the literature on P600 and delta oscillations by demonstrating that evoked and ITC delta power is sensitive to morphosyntactic information. Finally, our study contributes to the growing research on graded prediction by showing that preceding context can induce the expectations regarding word order and morphosyntactic case inflections as reflected in the electrophysiology of neuronal oscillations.