Learning and optimization of decision boundaries in animals

Establish the mechanisms by which animals learn and optimize decision boundaries used in evidence-integration models during perceptual decision making under uncertainty.

Background

Evidence-integration (accumulate-to-bound) models successfully account for a wide range of behavioral and neural findings in perceptual decision making. However, while these models explain how decisions are formed given a boundary, they do not specify how animals initially acquire and tune that boundary through experience. The paper motivates a model-free reinforcement learning approach precisely because the learning and optimization of this boundary is not settled by existing theories.