Cogsci Colloq: Nicholas Turk-Browne (Princeton)

Title: Statistical learning in mind and brain

Abstract: The environment is highly stable over time. Not only do we repeatedly encounter the same people, places, and objects, but they also tend to appear in regular sequences and configurations. These regularities are extracted rapidly and often without effort or awareness, resulting in higher-order knowledge of words, events, and scenes. Based on behavioral studies, I will suggest that this kind of statistical learning has widespread consequences for the mind, including showing that it distorts the perception of ensemble features such as numerosity, and that it commands attention to locations and features with structure. Based on neuroimaging, patient, and computational studies, I will then suggest that the medial temporal lobe memory system helps implement this kind of statistical learning in the brain, including showing that the hippocampus learns structure by shaping the representational space for objects to mirror both simple and complex regularities. These studies reveal the deep and broad ways in which statistical learning interacts with other aspects of cognition.