Psychometric Challenges in Measuring Visual Working Memory Capacity
In experimental psychology, a subject's performance on computerized cognitive tasks is often used to measure a latent trait such as working memory capacity. Responses to these tasks are influenced by a variety of factors such as the variable of interest (i.e., working memory capacity), attentional lapses, and random guessing behavior. In this talk, I describe my experiences with modeling response behavior on a specific cognitive task. First, I introduce a popular task that is designed to measure visual working memory capacity and attention. I then demonstrate a major flaw of the current model and discuss the consequences for modeling response behavior on this task. Finally, I discuss possible solutions to this problem, including model restrictions, alternate models, and changes to the task.
Leah Feuerstahler is a postdoctoral fellow with the Quantitative Measurement & Evaluation (QME) program at UC Berkeley. Her research interests include item response theory and latent variable models. This talk is based on research conducted during her graduate studies at the University of Minnesota.
Presentation slides available upon request: firstname.lastname@example.org