Randy Gallistel, Rutgers University
Thursday, July 14th
6:45-8:00 PM Reception, La Sala de Puerto Rico (W20-2nd floor)
8:00-9:30 PM Talk, Kresge Auditorium (W16-109)
Commentator: Kenneth Wexler
The Nature of Learning and the Functional Architecture of the Brain
Two different conceptions of learning currently contend, the cognitive science conception and the neuroscience conception. In the first, learning is the extraction from experience of information about the world, which is carried forward in memory to inform subsequent behavior. In the second, learning is a remolding of a plastic brain by experience to make it better adapted to the experienced world. These conceptions imply different functional architectures for the brain. In the first, the brain has the architecture of a Turing machine; in the second, it has the architecture of a switching network. The latter architecture lacks elements essential to the realization of computation as we currently understand it. The lecture reviews simple behavioral evidence drawn from the rat, mouse and pigeon timing literature and the literature on insect navigation implying that the brain must have a Turing architecture. In particular it must have a memory that carries information forward in time and makes it accessible to computations not foreseen at the time the information was laid down. For linguists, this will come as no surprise, because language is unimaginable in a device that lacks such a memory. Our current ignorance about how such a memory might be neurobiologically implemented is a major contributor to the conceptual chasm between linguistics and neuroscience.