3-156 - Real-time Processes of Statistical Language Learning

Session Type: Paper Symposium

Integrative Statement

A large body of literature shows that young learners rely on statistical learning to break the barrier of early language acquisition. They are skilled at detecting statistical regularities to accomplish various learning tasks, such as segmenting continuous speech into individual words and mapping words to referents. However, a deeper understanding of the phenomena requires further investigations on how statistical learning works in real time, as the outcomes of statistical learning rely on online cognitive processes to selectively attend to relevant information in a noisy environment, to process selected information moment by moment, to aggregate and integrate information over time, and to store and retrieve acquired language knowledge from memory.

The theme of this symposium is on real-time information processing in statistical language learning. The first talk examines the time course of word segmentation by studying how learning of two types of regularities (within-and across-word regularities) unfolds over time. The second talk investigates how people learn words by aggregating information across naturalistic learning moments with different levels of ambiguity. The third work focuses on how timing influences the informativity of naming instances during real-world interactions and how these temporal characteristics further influence children’s data selection processes. The final talk uses a dynamic, autonomous model of cross-situational learning to understand how children’s attention, visual memory and word learning systems co-evolve over development.

Taken together, evidence from empirical studies of children and adults and computational modeling will provide insight into not only real-time statistical learning mechanisms but also domain-general cognitive and learning processes.

Sub Unit

• Language, Communication

Chairs

• Yayun Zhang, Indiana University

• Chen Yu, Indiana University

Individual Presentations

• The Time Course of Tracking Multiple Regularities In Speech - Presenting Author: Viridiana L. Benitez, University of Wisconsin-Madison; Jenny Saffran, University of Wisconsin - Madison

• Real-Time Cross-Situational Learning from the Child’s View - Presenting Author: Yayun Zhang, Indiana University; Chen Yu, Indiana University

• Referential Timing in Early Word Learning - Presenting Author: Timothy Dawson, University of Pennsylvania; Lucia Pozzan, University of Pennsylvania; Lila Gleitman, University of Pennsylvania; John Trueswell, University of Pennsylvania

• Word-Object Learning via Visual Exploration in Space (WOLVES): A Dynamic Neural Field Model of Statistical Learning - Presenting Author: Laura Colosimo, University of East Anglia; Larissa Samuelson, University of East Anglia; John Spencer, University of East Anglia