The role of input variability in generalizing phrasal constructions featuring non-adjacentdependencies

Abstract Controlled experiments demonstrate that increased input variability in the intervener-slot of non-adjacent dependencies (NADs) improves children’s inductive generalization of the dependency ( Gómez, 2002 ). Experiments targeting different linguistic structures show that variability benefits not only children’s inductive generalization but also their productive extension of those generalizations to novel contexts ( Wonnacott et al., 2012 ). Combined, these findings motivated our investigation into applying variability benefits to NAD learning in classrooms, considering their ubiquity in natural language. We present a two-week quasi-experimental teaching intervention with two British Year 2 German as a foreign language classes (age 6; 20 students/class). This proof-of-concept trial, comprising a high and low input variability condition, focused on three sets of NADs, realized by German subordinate clauses (“Subj [intervener prepositional phrase] Verb”), featuring 30 or 5 ‘interveners’, respectively. Post-tests showed ambiguous results regarding the effect of increased input variability on the generalization of NADs and the ability to extend the generalization to novel contexts (i.e., unknown interveners). Nonetheless, the findings across conditions demonstrate a positive role for targeted input flooding of particular constructions in instructed FL development suggesting young learners can pick up large linguistic units after minimal exposure. We discuss the ambiguous results and highlight methodological and pedagogical implications.