Modeling the range of performance on the serial subtraction task

We present a model of serial subtraction, a task where subjects repeatedly subtract a 1- or 2-digit number from a 4-digit number. The model performs 4 min. blocks of these subtractions like subjects do. The current model replicates part of the pace and % correct for group data. Because performance on this task varies widely between subjects, we explore what it means to match the data distribution. We find that our model represents individual subjects better than group means. We can start to model a distribution of performance and illustrate some of what this approach will entail.