A Bayesian Race Model for Characterizing Recognition Memory Performance

• Activity Number: 230
• Type: Topic Contributed
• Date/Time: Monday, August 10, 2015 : 2:00 PM to 3:50 PM
• Sponsor: Section on Bayesian Statistical Science
• Author(s): Trisha Van Zandt* and Sungmin Kim and Kevin Potter and Peter F. Craigmile and Mario Peruggia
• Companies: The Ohio State University and Google and The Ohio State University and The Ohio State University and The Ohio State University
• Keywords: Bayesian modeling ; cognitive science ; behavioral data ; individual differences ; recognition memory ; hierarchical modeling

Abstract

Many psychological models make use of the idea of a trace, which is a change in a person's cognitive state that arises as a result of processing a given stimulus. Many of these models also assume that a trace is always laid down when a stimulus is processed. We investigated a Bayesian hierarchical model of data obtained from a difficult recognition memory experiment in which a large proportion of people performed poorly. The model includes a stochastic component that probabilistically determines whether a trace is laid down according to the weight of evidence presented. The data are modeled using a minimum gamma race model, with extra model components that allow for the effects of stimulus, sequential dependencies, and trend. Subject-specific effects, as well as ancillary effects due to processes such as perceptual encoding and guessing, are also captured in the hierarchy. This novel modeling approach allows us to explain not only the mechanisms by which memory traces are established, but also to determine how people succeed or fail at performing the memory task.