Complementary Semantic Systems
Seminar | February 3 | 11 a.m.-12:30 p.m. | 5101 Tolman Hall
Dan Mirman, University of Alabama at Birmingham
Representing concepts in terms of semantic features, like < has 4 legs > or < is yellow >, has driven major advances in understanding semantic cognition, integrating behavioral, neural, and computational research. Feature-based approaches are very good at capturing taxonomic relations such as DOG -- BEAR, but it is less clear how they could capture thematic relations such as DOG -- LEASH, which are based on co-occurrence in time and space. Behavioral, computational, and neuroscience evidence shows that such relations are an important part of semantic cognition and that they dissociate from taxonomic relations. I will review some key pieces of this evidence, focusing on differences in activation time course, individual differences in strength taxonomic vs. thematic relations, and distinct neural correlates. I will conclude with two computational principles that may drive the development of these complementary semantic systems.