2018 Friday Poster 6452

Friday, November 2, 2018 | Poster Session I, Metcalf Small | 3pm

Increased Beta-Band Activity in the Active Maintenance of Fragile L2 Representations
L. Dekydtspotter, K. Miller, M. Iverson, Y. Xiong, K. Swanson, C. Gilbert

Non-native processing may involve weakly activated, and thus fragile, representations (Dekydtspotter & Miller, 2013; Miller, 2014, 2015) and delay (Boxell & Felser, 2017). Discrete brain rhythms meet discrete parsing demands: gamma, prediction; theta, unification; and beta, active maintenance (Lewis et al., 2015). We examined EEG activity in native (NSs;n=24) and advanced L1-English non-native speakers (NNSs;n=22) of French as information accompanying wh-expressions passes through bridge dit que ‘say that’. In (1a-d), dit foreshadows subordinate-clause dependencies; que confirms them. NP-modifiers (le concernant; 1a,c) and N-complements (à propos de lui; 1b,d) are distinguished in recursion (Chomsky, 2005; Lebeaux, 1988). Pronouns can be syntactically bound if re-represented in recursion (1b), or discursively coreferential (1a). Without matching antecedents (1c,d), these processes are thwarted. Analysis of ERPs revealed delayed distinctions and beta-band activity in NNSs.

Participants completed a RSVP task (300ms/word, 250ms/ISI) with 25 quadruples crossing Structure (NP-modifier/N-complement) and Antecedent Gender (Match/Mismatch) (1a-d). EEG was recorded via a 64-electrode EGI system with 50kΩ maximal impedance, a Net Amps 300 amplifier, and 1000Hz sampling rate. Data were preprocessed with a .05-100.5Hz band- pass filter and cleaned of artefacts via epoch/channel rejections and Independent Component Analysis. 87% of NS and 86% of NNS trials were retained. Mixed-effects models of average amplitudes from 250-550ms with 50ms-baseline into critical words (Phillips et al., 2005) dit and que were compared over four regions (Fiebach et al., 2002; Figure1) for potential SAN modulations revealing distinct processing loads. In the ERP time-frequency analysis, 1.88- second epochs, with 750ms baseline prior to bridge, were convolved with a 7-cycle wavelet transform with range 4-40Hz (FieldTrip, Ostenveld et al., 2011). We ran non-parametric cluster-based permutation tests from 250-550ms into each word. One thousand Monte Carlo simulations were performed per subject. Main effects of modifier-complement differences and of group, and interactions, were examined.

Mixed-effect models revealed modifier-complement differences at dit in NSs (p=.02) but at que in NNSs (p=.027) (Figure2). Topographies revealed focal effects in NSs, and non-focal effects in NNSs (Figure3). Power analyses revealed a main group effect at 18-21Hz (beta- range) 250-400ms into dit in two anterior clusters, with greater power for NNSs than NSs (p=.005) (Figure4). Main-effect modifier-complement differences arose anteriorly during dit at 34-40Hz (gamma-range) with greater modifier-complement differences in mismatch than match (p=.003) and during que at 4-5Hz (theta-range) with greater power-differences in match than mismatch (p=.035) (Figure5). Interactions with group arose over anterior clusters for 30ms during dit at 36-40Hz (gamma-range) (p=.038) and crucially 50ms during que at 21-24Hz (beta-range) (p=.013) (Figure6). Furthermore, post-hoc GLMs of average power over these clusters revealed that the interaction at 36-40Hz resulted from greater modifier-complement difference in match for NSs (p=.001) and in mismatch for NNSs (p=.002). The interaction at 21-24Hz resulted from a modifier-complement difference in mismatch in NNSs (p=.0005), absent in NSs (p=.077).

Gamma- and theta-rhythm activity at dit and que matched prediction and unification in parsing. Additionally, beta-band activity distinguishing NNSs from NSs matches the active maintenance of fragile representations for NNSs. Support of prediction at dit and referential processes at que plausibly explains delay.