Learning Patterns in Noise: Environmental Statistics explain the Sequential Effect

Subjects display sensitivity to local patterns in stimulus history (sequential effects) in a variety of decision-making, perceptual, and motor tasks. Sequential effects are typically cast as a prime example of human irrationality. We propose a Bayesian model that explains sequential effects as the natural consequence of one incorrect assumption: instead of assuming a stable world, subjects assume change. We test and confirm one specific prediction of our model in a 2alternative forced-choice reaction time task. We manipulated participants’ beliefs about the comparative likelihoods of binary events and then showed that these beliefs determine biases in sequential effects. We conclude that the origin of sequential effects is belief in a changing world.