Neural mechanisms of perceptual decision making

Abstract

Perceptual decision making is a fundamental component of goal-directed behavior. Even simple perceptual decisions require interactions among different neuronal populations within and across brain areas. To relate neuronal activity to cognition and behavior meaningfully, we must find ways to parse these complex circuits according to their basic computations. Cognitive models that explain behavior in terms of simple computational processes can guide this effort. Stochastic accumulator models propose that perceptual decisions are generated by two processes: (1) a representation of perceptual evidence for potential responses, and (2) accumulation of evidence to a response threshold. I recorded single-unit spiking, local field potentials, and extracranial event-related potentials from macaque monkeys performing a visual search task to test whether different neuronal populations implement these processes. I found that distinct subpopulations of neurons in the frontal eye field of lateral prefrontal cortex either represent perceptual evidence or accumulate the evidence to a response threshold. I show that a stochastic accumulator model can explain decision-making behavior through specific interactions between these subpopulations. I also found that event-related potentials recorded over posterior visual cortex modulate in a manner consistent with a representation of perceptual evidence, which may reflect feedback from frontal areas. In contrast, I found that neurons in the supplementary eye field of medial frontal cortex do not correspond to any model component. Instead, some neurons in this area appear to monitor the outcome of previous decisions by signaling the occurrence of errors. This work highlights the utility and limitations of using cognitive models as a framework for understanding neuronal function.