Just uploaded Morten Christiansen’s excellent keynote, “Implicit-Statistical Learning: A Tale of Two Literatures”, Lancaster, June 23, 2016. Thanks again to Morten for his great keynote and to our TV Unit for producing the recording! Click here to see the keynote. Please see below for the abstract.
Morten Christiansen (Cornell University, Aarhus University, and Haskins Laboratories): Implicit-statistical learning: A tale of two literatures
Although implicit learning and statistical learning both can be construed as uncovering the structure of the input from its distributional properties, the relevant research has largely been published in separate literatures and with surprisingly little cross-polination between them. This has resulted in apparently opposing perspectives on the computations involved in learning, pitting chunk-based learning against probabilistic learning. In this paper, I argue for a unification of the two approaches to learning under the heading of “implicit statistical learning” and trace its long historical pedigree going back nearly a century. Building on basic insights from the memory literature, I will sketch a framework for statistically-based chunking that may provide a unified computational basis for understanding implicit statistical learning.