Neural and Computational Mechanisms of Perceptual Decisions Between Multiple Alternatives Based on Multiple Sources of Evidence.

摘要

The goal of the project was to elucidate computational and neural mechanisms underlying perceptual decisions between multiple alternatives. Behavioral experiments in humans as well as combined behavioral and neurophysiological experiments in non-human primates were performed, and computational modeling techniques were used for addressing the computational mechanisms underlying the perceptual decisions. A multi-alternative version of the random dot motion direction discrimination task, using a multi-component stimulus, was found to provide a rich behavioral dataset spanning accuracy levels from chance to perfect and a wide range of mean response times while providing simultaneous experimental control over how much sensory evidence is provided for each of the alternatives. The behavior was found to be consistent with a number of integration-to-threshold mechanisms. While these different mechanisms produce virtually indistinguishable decision behavior, the internal dynamics arc quite different which can be addressed by performing high- resolution recordings of brain activity. Therefore, non-human primates were trained to perform the same decision task and multi-electrode recordings of decision-related activity were made in parietal cortex.