The development of shared syntactic representations in late L2-learners: Evidence from structural priming in an artificial language

Abstract

• How do shared syntactic representations develop in late L2 learners? • Structural priming as tool to investigate shared syntactic representations. • Priming assessed during 5 sessions in which an artificial language is learned. • Cross-linguistic priming from Day 1 for transitives and Day 2 for ditransitives. • • Syntactic representations first shared within and later between languages. Two longitudinal studies investigated the development of syntactic representations in late second language (L2) learners by means of structural priming in an artificial language (AL) paradigm. Several studies found cross-linguistic structural priming in L2 learners, suggesting that they have shared syntactic representations across languages. But how are these shared representations established? Hartsuiker and Bernolet’s (2017) account claims that 1) L2 syntactic representations evolve gradually from being item-specific to more abstract, and that 2) over time these representations are integrated with available native language (L1) representations. We tested predictions of this theory with native Dutch speakers, who acquired the AL in the lab during five sessions by means of a battery of tasks, the last of which was a sentence priming task. The AL syntax resembled Dutch syntax. We manipulated the relation between prime and target to investigate whether structural priming occurred in conditions with meaning overlap (item-specific) and without overlap (abstract). In Experiment 1, participants responded only in the AL, but in Experiment 2, where the AL was more difficult, the target sentence could also be in Dutch. In both studies, there was an effect of within-language priming and AL-Dutch priming in transitives and ditransitives, but no effect of Dutch-AL priming in ditransitives on Day 1. On Days 2–5, however, priming emerged in most cross-linguistic priming conditions. These findings partly confirm Hartsuiker and Bernolet’s (2017) predictions and suggest that at least for structures that are very similar between languages, shared syntactic representations can be established very early during language learning.