Abstract
Many psychological theories imply the existence of intratask change, that is, change that occurs as a task is being performed, but few statistical models incorporate this concept. Intratask change usually involves three complicating issues: items are not repeated, outcomes are categorical, often dichotomous, and subjects differ in the amount they change. In this research, I developed a family of item response models applicable to the study of intratask change. I used these models, which I call intratask change item response models (ICIRMs), to test a psychological theory involving intratask change: the hypothesis that the age-related decrease in working memory span is at least in part caused by an age-related increase in the effects of proactive interference. Proactive interference is an intratask change concept, as it accumulates throughout a working memory span task and leads to decreased performance. Previous research has ignored the dynamic aspects of PI, so an ICIRM is needed as a direct test of the theory. In order to be identified, ICIRMs require randomized item presentation order; therefore, I collected data on a working memory span task over the internet. A series of ICIRMs were fit to these data, and an exponential ICIRM was found to provide optimal fit. Contrary to expectations, however, average intratask change was positive, indicating increased performance over the task. Age-related differences in intratask change did not account for the age-related decline in working memory span. Simulations suggested that the conclusions on the shape and direction of intratask change were valid, but the conclusions on relations to age may not be reliable. I interpret the results as suggesting that strategy production may be a more important source of individual differences in working memory span than proactive interference.
Note: Abstract extracted from PDF text
