Human Inferences about Sequences: A Minimal Transition Probability Model

The brain constantly infers the causes of the inputs it receives and uses these inferences to generate statistical expectations about future observations. Experimental evidence for these expectations and their violations include explicit reports, sequential effects on reaction times, and mismatch or surprise signals recorded in electrophysiology and functional MRI. Here, we explore the hypothesis that the brain acts as a near-optimal inference device that constantly attempts to infer the time-varying matrix of transition probabilities between the stimuli it receives, even when those stimuli are in fact fully unpredictable. This parsimonious Bayesian model, with a single free parameter, accounts for a broad range of findings on surprise signals, sequential effects and the perception of randomness. Notably, it explains the pervasive asymmetry between repetitions and alternations encountered in those studies. Our analysis suggests that a neural machinery for inferring transition probabilities lies at the core of human sequence knowledge.Author Summary: We explore the possibility that the computation of time-varying transition probabilities may be a core building block of sequence knowledge in humans. Humans may then use these estimates to predict future observations. Expectations derived from such a model should conform to several properties. We list six such properties and we test them successfully against various experimental findings reported in distinct fields of the literature over the past century. We focus on five representative studies by other groups. Such findings include the “sequential effects” evidenced in many behavioral tasks, i.e. the pervasive fluctuations in performance induced by the recent history of observations. We also consider the “surprise-like” signals recorded in electrophysiology and even functional MRI, that are elicited by a random stream of observations. These signals are reportedly modulated in a quantitative manner by both the local and global statistics of observations. Last, we consider the notoriously biased subjective perception of randomness, i.e. whether humans think that a given sequence of observations has been generated randomly or not. Our model therefore unifies many previous findings and suggests that a neural machinery for inferring transition probabilities must lie at the core of human sequence knowledge.