A fundamental challenge in language acquisition is to associate words to their referents in the environment and identify syntactic structures in which these words occur. One proposed mechanism underpinning this process is known as cross-situational learning, during which learners track co-occurrence statistics across situations to build up word-referent mappings and grammar knowledge. While young children acquire their first language (L1) rapidly and seemingly effortlessly, usually without substantial explicit instruction, the process of additional language (L2) learning becomes more complex as they enter middle childhood. This transition, marked by increasing cognitive maturity and the onset of formal education, introduces variation in both internal factors (e.g., changes in chronological age, individual differences in memory capacity) and external factors (e.g., openness to formal instruction) to shape the language acquisition process and outcome. Despite the critical cognitive and metalinguistic developments occurring from middle childhood through adolescence, this population remains underrepresented in L2 acquisition research more broadly, and in cross-situational learning research specifically. Across three experiments involving 220 children, this thesis investigates the effects of age, individual differences in memory, and explicit instruction on L2 learning through cross-situational learning with a particular focus on eight- to thirteen-year-old learners. Study 1 investigated whether 8- to 9-year-old children could simultaneously acquire novel vocabulary and grammar via a cross-situational learning paradigm adapted from Rebuschat et al. (2021). During the task, learners were exposed to ambiguous sentence-scene mappings without feedback or instruction, and they could only develop word-referent mappings as well as grammar knowledge (i.e., sentence word order knowledge) through tracking the input statistics. Results indicated that children at this age acquired grammar (i.e., word order) effectively, but no evidence of vocabulary learning. This suggested that the sentence word order information may become available earlier and serve as a scaffold for subsequent lexical learning. This study also suggested a child-adult difference in cross-situational learning: whereas previous adult studies showed effective learning of both vocabulary and grammar from similar input, children only presented successful grammar learning. Study 2 expanded the age range and introduced an explicit instruction condition to examine when and to what extent instruction affects child L2 learning. This study observed a qualitative change at around the age of 12, at which point children became more responsive to explicit instruction. Results also indicated age and metalinguistic awareness as moderating factors on the effect of explicit instruction on L2 learning. Study 3 examined whether individual differences in working memory, procedural memory and declarative memory explained additional variance in L2 learning under explicit instruction, beyond the effect of age. It was observed that age, memory abilities and their interactions accounted for the variance in early stages of L2 development. Overall, these findings demonstrate that the early stages of L2 learning between the ages of 8 and 13 are shaped by the interplay of external (explicit instruction) and internal (age and memory capacities ) factors. Rather than assuming implicit learning works equally well for children, our results highlight significant room for metalinguistic knowledge and awareness to facilitate learning. In turn, explicit instruction is not uniformly effective but requires developmental readiness. These findings have practical implications for educators, curriculum developers, and AI-based language learning tools, supporting the design of age- and cognitively- appropriate instructional approaches.
Wensi Zhang (Thu,) studied this question.