Tracking Syntactic Conflict between Languages over the Course of L2 Acquisition: A Cross-sectional Event-related Potential Study

Abstract

One challenge of learning a foreign language (L2) in adulthood is the mastery of syntactic structures that are implemented differently in L2 and one's native language (L1). Here, we asked how L2 speakers learn to process syntactic constructions that are in direct conflict between L1 and L2, in comparison to structures without such a conflict. To do so, we measured EEG during sentence reading in three groups of German learners of Dutch with different degrees of L2 experience (from 3 to more than 18 months of L2 immersion) as well as a control group of Dutch native speakers. They read grammatical and ungrammatical Dutch sentences that, in the conflict condition, contained a structure with opposing word orders in Dutch and German (sentence-final double infinitives) and, in the no-conflict condition, a structure for which word order is identical in Dutch and German (subordinate clause inversion). Results showed, first, that beginning learners showed N400-like signatures instead of the expected P600 for both types of violations, suggesting that, in the very early stages of learning, different neurocognitive processes are employed compared with native speakers, regardless of L1–L2 similarity. In contrast, both advanced and intermediate learners already showed native-like P600 signatures for the no-conflict sentences. However, their P600 signatures were significantly delayed in processing the conflicting structure, even though behavioral performance was on a native level for both these groups and structures. These findings suggest that L1–L2 word order conflicts clearly remain an obstacle to native-like processing, even for advanced L2 learners.

Examining variability in the processing of agreement in novice learners: Evidence from event-related potentials.

“…With respect to number, in contrast to the predictions of the cognitive theories (e.g., DeKeyser, 2007; Ullman, 2001, 2005, 2015), we did not observe discrete stages in which agreement dependencies were first processed as lexical associations (N400; Carrasco-Ortíz et al, 2017; McLaughlin et al, 2010; Mickan & Lemhöfer, 2020; Morgan-Short et al, 2010; Osterhout et al, 2006; Reichle et al, 2016; Tanner et al, 2013). Our results did not reveal a significant effect of Session, showing similar ERP amplitudes across the different testing sessions, which suggests that the sensitivity we observed for number was in place from the first session 6 .…”

“…A similar pattern was observed in Tanner et al (2013), a cross-sectional study of English-speaking learners of German which also focused on subject-verb agreement, a property which is similar in this L1-L2 pairing (see also Mickan & Lemhöfer, 2020). Whereas native speakers of German and more advanced learners of German showed P600 to agreement violations, novice first-year learners showed a tendency toward a biphasic response, yielding both N400 and P600.…”

“…In addition, Lemhöfer et al (in press) show, based on a study of gender agreement in Dutch, that P600 is more likely to emerge in learners when there is a reason to doubt the reliability of the input in the experiment, either because of the presence of outright grammatical violations or to the GJT task itself. In the present study, we included a GJT to make the experimental task comparable to the study we were building on (Osterhout et al, 2006), as well as many other L2 studies on agreement in the literature (Mickan & Lemhöfer, 2020; Morgan-Short et al, 2010; Tanner et al, 2013, 2014; White et al, 2012, see Caffarra et al, 2015 for a review). It is important to note that differences emerged between number and gender in our experiment, and thus the inclusion of the GJT task could not have been solely responsible for the significant effects that emerged.…”

“…Because the learners in the study are at the beginning stages of L2 acquisition, and rejecting incorrectly judged trials would significantly reduce the number of trials per condition and thus the signal-to-noise ratio, we decided to include all experimental trials in the analysis regardless of accuracy in the GJT; this procedure mirrors that followed in the Osterhout et al (2006) study which we build on here and has also been adopted in other recent research examining beginning learners (e.g., Alemán Bañón et al, 2018; Mickan & Lemhöfer, 2020). Moreover, as pointed out by VanRullen (2011), restricting analyses to only correct trials runs the risk of introducing artifactual differences in the EEG responses across conditions.…”

“…N400 is typically taken to reflect semantic incongruency or the violation of lexical expectations (Kutas & Federmeier, 2011; Lau et al, 2008). Although N400 is typically not elicited for agreement violations in group analyses of native speakers (but see Tanner & Van Hell, 2014), N400 has been observed in several studies examining the processing of agreement violations in L2 learners, especially those at lower levels of proficiency (Carrasco-Ortíz et al, 2017; McLaughlin et al, 2010; Mickan & Lemhöfer, 2020; Morgan-Short et al, 2010; Osterhout et al, 2006; Reichle et al, 2016; Tanner et al, 2013).…”