The 2024-2025 colloquium series concludes on Monday, April 14, with a talk by Yi Ting Huang (Maryland), taking place in Dwinelle 370 and on Zoom (passcode: lx-colloq) from 3:10-4:30pm. The title of her talk is "Learning language, fast and slow: How to overcome sparse data and signal degradation during real-time processing and development." The abstract is as follows:
Traditional approaches to language development focus on relationships between aggregate inputs (e.g., total words heard) and outcomes (e.g., vocabulary size), assuming that what parents say is what children learn. This, of course, ignores an obvious fact about acquisition – Children initially have no idea what parents are talking about. Instead, they must infer linguistic representations through iterative encounters with sentences. To better understand these processes, I will introduce three lines of research that describe developmental algorithms for sentence processing, and their variation with language experience and impairment status. First, I will examine how 4- to 6-year-olds determine who did what to whom in sentences, and show ways in which their strategies vary with properties of the communicative context. Second, I will take a closer look at SES language gaps, and show how systematic variation in language experiences preserve learning ability but alter sentence-processing strategies. Finally, we will turn to children with Developmental Language Disorder, who face profound difficulties producing, understanding, and learning from language in the school-aged years. We present preliminary findings from a randomized controlled trial to alter sentence processing and improve comprehension. We will close by considering causal pathways between chronometric and ontogenetic processes, and discuss their implications for how children recreate language from input, what a satisfying algorithmic-level description of language acquisition might look like, and why more cognitive scientists should be working at the intersection of basic and translational research.