Stochastic Dynamic Models (Choice, Response, and Time)

Abstract

Stochastic dynamic models are models of decision making in simple perceptual and cognitive tasks, which assume that decisions are based on the accrual in continuous time of noisy, time-varying stimulus information. The dynamics of information accrual in such models are represented mathematically by stochastic differential equations. Accrual processes of this kind can form the basis of dynamic signal detection models, in which accrual is terminated at the end of a fixed sampling interval, or of sequential-sampling models, in which information is accrued to a criterion. Functionally, the main difference between these two kinds of model is that sequential-sampling models predict choice probabilities and response time (RT) simultaneously, whereas signal detection models predict choice probabilities only. Usually, also, the mathematical analysis of sequential sampling models is more complex than that of the corresponding signal detection model. This entry describes the methods that are used to obtain predicted RT distributions and response probabilities for two main classes of dynamic sequential-sampling model: diffusion process models and Poisson parallel counter models. RT distributions for diffusion process models are obtained using numerical integral equation methods. Those for counter models are obtained by analyzing the accrual of counts from a pair of independent time-inhomogeneous Poisson processes.