The L2 Impact on Learning L3 Dutch: The L2 Distance Effect

Abstract

Cross-classified random effect models (CCREMs) are often used for partitioning variation in both experimental and observational linguistic data. How- ever, crossed random effects may have more complex interrelationships than is generally assumed. This becomes clear when comparing first language (L1) and second language (L2) influences on proficiency in Dutch as a third language (L3). Using a large database of L3 speaking proficiency scores, we assessed the mutual dependency between the crossed random effects of the L1 and the L2. The results suggest independent and robust linguistic distance effects of the L1 and the L2: the smaller the linguistic distance to the L3, the higher the L3 proficiency, with the L2 effect being weaker than the L1 effect. Although a model that incorporates an additional L1-by-L2 random interaction effect fits the data best, this model still stipulates the relative importance of an independent L2 distance effect. We found that the L1 distance effect is robust against the L2 distance effect and that the L2 distance effect is robust against interactive effects. We discuss possible explana- tions for interactions between the L1 and the L2. Overall, the data support inde- pendent linguistic distance effects of both the L1 and the L2, besides L1‒L2 inter- actions. We recommend that researchers compare the fit of their crossed random effects models with the fit of models that also include the respective interaction ef- fects.