Statistical learning and consolidation in language and non-language patterns

About This Grant

The aim of this research is to understand how children and adults learn language from the speech that they hear. The research also studies how children and adults learn visual patterns. The aim of these studies is to ask whether language learning and visual pattern learning show similar principles and timing. This information is important for the basic science of understanding learning. It is also important for supporting children and adults who show problems in learning language, in school or in everyday tasks that are important for real-world success. The specific focus of this research is on a process in memory called ‘consolidation.’ Consolidation is critical because it transforms memories from a fragile state into more stable, long-term memories, which allows people to retain memories without being disrupted by new information. Other investigators have shown that immediate memory for events is very sparse, but, over time, memories of specific events are integrated and enriched to form a more complete memory for significant life episodes. The studies in this project ask whether consolidation also happens for specific spoken language sequences or specific visual patterns, and whether consolidation changes specific sequence memories into learning the grammatical patterns of a language. Understanding consolidation is important for addressing a major challenge in artificial intelligence such that the system can learn new information without losing previously learned information. Additionally, this project involves STEM research training and mentoring for undergraduates and high school students. To examine complex learning in a short time, researchers design miniature artificial languages and then, in a single ~20-minute exposure, present learners with auditory sequences (like ‘sentences’) from the language, or sequences of visual symbols that have similar patterns. Learning is assessed by asking participants to choose sequences like those they have heard (or seen), versus sequences that have errors and are not like what they have heard or seen. Some learners are tested immediately after exposure, while other learners are tested one day or one week later, with no further exposure to the language during that time. Changes in learning after a delay, without further exposure, indicates consolidation. Across studies, this research asks about when and how consolidation occurs; what types of input undergo consolidation; and what activities can enhance or disrupt these consolidation processes. Answering these questions is important for understanding and supporting optimal learning in children and adults, and developing artificial intelligence systems that are less susceptible to catastrophic forgetting. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.