UC San Diego

Perceptual decision making (PDM) involves choosing one option among several on the basis of sensory evidence and is a highly adaptive mechanism for organisms to successfully interact with their environments. Such a choice requires integrating and interpreting sensory information for the purpose of guiding subsequent behavior (e.g., seeing a ball move rightward and veering accordingly to catch it). Typical single-unit recording studies examining PDM utilize simple sensorimotor tasks (e.g., a macaque views a noisy array of dots moving in one of two possible directions and deploys a saccade in the chosen - and presumably, perceived - direction) in order to parse various aspects of PDM. With the aid of mathematical models, these experiments have found that the activity of individual neurons involved in motor response generation comprises perceptual decisions, and that PDM can be formalized as an accumulation of sensory evidence towards a particular choice (as represented by an increase in neuronal firing rate) until some threshold is reached. Explaining the mechanisms of PDM at the level of neural populations and linking ensemble patterns of neural activity to perception, however, still remains unclear. With a combination of visual psychophysics, neuroimaging, and modeling, I present a set of studies that examines the neural correlates subserving PDM in human cortex (Experiment 1), clarifies the relationship between sensory representations in visual cortex and perceptual performance (Experiment 2), and tests the behavioral predictions derived from single-cell recordings (Experiment 3). These findings both challenge and confirm some of the previous neurophysiological work: Experiment 1 provides evidence of a neural mechanism of PDM not based purely on oculomotor regions, Experiment 2 shows that the optimality of activation patterns in visual cortex predicts task performance, and Experiment 3 illustrates that attentional manipulations influence perception in a manner consistent with the enhancement and suppression of distinct neural populations predicted from single-unit recordings. Furthermore, these studies demonstrate the utility of model-based cognitive neuroscience in quantifying psychological processes of interest for each individual and relating between-subject differences with corresponding brain measurements