The impact of language switching on statistical word learning – Experiment 3

It is currently unclear whether language switching hurts, facilitates, or has no impact on word learning. In Experiment 1, we observed that while language switching hurt cross-situational word learning at the trial level, it facilitated the acquisition of 2:1 mappings (two words mapping on the same object) when each word belonged to a different language (Experiment 2, a replication, is ongoing). In Experiment 3, we will test whether this advantage generalizes to 1:1 mappings (one word maps to one object), allowing us to distinguish a general benefit of language mixing from an effect specific to learning translation-like word pairs. Participants will learn English-like and German-like pseudowords in three conditions: English-only, German-only, and mixed-language. In the mixed condition, each object has a single label, but labels are split across languages (half English-like, half German-like), resulting in trial-to-trial language switches; in pure conditions, all labels are from a single language. We make no directional prediction for overall mixed vs pure learning: a mixed advantage would support a general benefit of mixing/switching, whereas no difference or a pure-language advantage would suggest that mixing benefits arise primarily when it is directly relevant (e.g., translation-like learning). Within the mixed condition, we predict higher accuracy on repetition trials than on switch trials (reflecting local switch costs, as in Experiment 1). We include a remapping phase in which each object receives a new label (half English, half German) and test whether remapping depends on language match (same vs different). We predict that remapping will be easier when the two languages (for learning and remapping) differ.