Proceedings of the Annual Meeting of the Cognitive Science Society

Language learning is a sophisticated process as learners need to detect and extract rich regularities embedded in the continuous speech inputs. Children, compared to adults, appear to learn languages more effortlessly. Nevertheless, early studies in implicit statistical learning revealed little developmental differences between children and adults. Recent work has found the speed of statistical learning in adults is associated with their neural sensitivity to probabilistic information in speech. It is not well understood, however, whether children share similar or different underlying neural processes for probabilistic information compared to adults. Specifically, are children similar to faster or slower adult statistical learners, or neither of them? In the current study, children aged between 5 and 12 completed a passive auditory oddball task, where they listened to syllables at different local and global frequency of occurrence. We used two neurophysiological measures, auditory mismatch responses (MMR) and late discriminative negativity (LDN) to compare children’s sensitivity to distributional probabilities in speech with adults. We found that children were more sensitive to probabilistic information in speech inputs at both the local and the global level than both faster and slower adult statistical learners. Moreover, unlike adults who integrate probabilistic information across global and local hierarchies, children seem to process different levels of probabilistic information in parallel.