Member speaker: Stevan Harnad
Date: Thursday, January 18th, 1:00 pm
Location: Rabinovitch House, 3640 de la Montagne, room 101
Title: Perceptual Grounding of the Lexicon
Abstract: I will describe a hypothesis about how the words in dictionaries get connected to their referents, based on experimental evidence that learning new visual and auditory categories changes perception and neural activity in human participants so as to make category members “pop out” by perceptually separating members from non-members and compressi
ng members (“categorical perception”). Similar effects occur in a neural net model for category learning. But not every word needs to be directly grounded perceptually in this way; in most do not. Once a minimal set of words is grounded directly, the rest can be grounded indirectly via verbal definitions and descriptions. This is the unique and universal power of the subject/predicate propositions of natural language.