2018 Friday Session C 1130
Friday, November 2, 2018 | Session C, Terrace Lounge | 11:30am
Processing of Remention Biases in Korean Learners of English
H. Kim, T. Grüter
This study investigates how L2 learners use remention (or implicit causality) bias for referential processing in real time. Some interpersonal verbs create biases to remention either its subject or object in causal dependent clauses (“Tom surprised Bill because he ” vs “Tom hated Bill because he ”)[1][2]. Several studies have provided evidence that remention bias affects L1 speakers’ reference processing before the sentence is otherwise disambiguated[3][4]. Only two studies have explored L2 learners’ use of remention bias in online processing, both with highly proficient speakers[5][6]. Little is known about the role of proficiency, a factor argued to modulate L2 listeners’ engagement in predictive processing[7]. Here we present evidence from a visual-world eye-tracking experiment investigating how L2 learners with varying English proficiency use remention bias in reference processing while listening to English sentences in real time.
Methods: 52 Korean-speaking learners (NNS) and 52 native speakers (NS) of English participated in an eye-tracking experiment comprised of sentences with NP1-biased (n=24), and NP2-biased (n=24) verbs, and 48 fillers. Each item was comprised of three sentences (see (1)). Half of the critical sentences had a bias-consistent, half a bias-inconsistent ending. Visual scenes showing the two arguments (Fig1) were displayed for the duration of the context, critical and question sentences. Participants answered the question by mouseclicking on the scene. NNS additionally completed two proficiency tasks (LexTALE, cloze).
Results: We analyzed participants’ preference for fixating NP2 vs NP1 in three successive temporal windows (offset by 200ms; Fig2): onset of because to pronoun offset (Conn+Pro), 500ms after pronoun offset, and 500ms-1000ms after pronoun offset. For each window, linear mixed-effects regression was conducted including group (NS, NNS) and verb-type (NP1-biased, NP2-biased) as fixed effects. In the first two windows, interactions between verb-type and group emerged (Conn+Pro: b=3.0, p=.019; 500ms: b=2.5, p=.045), driven by effects of verb-type among NS (Conn+Pro: b=2.3, p=.013; 500ms: b=3.5, p<.001) but not NNS (b=-0.7, p=.583; b=1.1, p=.342), indicating NS’ but not NNS’ looking patterns were guided by remention bias. In the third window, a main effect of verb-type (b=3.6, p<.001) emerged, not modulated by group; this effect also reached significance among NS (b=4.2, p<.001) and NNS (b=2.9, p=.017) separately. These results indicate early sensitivity to remention biases among NSs, consistent with focusing accounts of referential processing[8], and delayed sensitivity among NNS, potentially consistent with integration accounts proposed for NS[9].
To probe for effects of proficiency, additional modeling of the NNS data included proficiency as a continuous predictor. No interactions between proficiency and verb-type reached significance, although visual comparison of looking patterns in higher- vs lower-proficiency groups (median split, Fig3) suggest the (late) effect of verb-type is greater for higher than lower-proficiency learners.
Overall, these findings suggest that L2 learners can use remention bias information for referential processing during online comprehension, but the effect of this information is delayed compared to L1 processing. This delay may be associated with learners’ reduced ability to access and retrieve lexical representations in their use of remention bias information during real-time sentence processing.