Investigating the Assumption of Normality in Diffusion Models

Dr. Matthew Zeigenfuse

Diffusion processes have been used to study a variety of psychological phenomena including classification and identification, lexical decision making, executive control and psychological measurement. These models conceptualize an individual's choices as the result of repeatedly sampling and encoding relevant information and accumulating this information until some decision criterion is met. In diffusion processes, the encoded values, often called evidence, cannot be directly observed by the experimenter and are often assumed to follow a normal distribution. Unfortunately, this assumption can make interpreting experimental results difficult, because the extent to which inferences depend on this distributional assumption is unclear. This Master's thesis will investigate through simulation the dependence of diffusion model inferences on the normality assumption in the common situation where the difficulty of an experimental task is thought to depend on the value of a manipulated factor.