Proceedings of the Annual Meeting of the Cognitive Science Society

The fan effect is a well-known phenomenon in the study of associative memory. It refers to the finding that as the number of associations linked to a concept increases, retrieval becomes slower and less accurate. Modeling accounts have largely focused on reaction times or accuracy in isolation, and have primarily modeled testing-phase performance. In the present study, we examine whether fan effects already arise during learning and propose a unified account of reaction times and accuracy at the trial level using a Bayesian hierarchical log-normal race model. We report data from a Dutch fan experiment showing that fan effects are already observable during the learning phase and increase across repetitions. The hierarchical Bayesian framework allows us to capture both reaction times and accuracy simultaneously while accounting for individual differences. Our model successfully captures slower incorrect responses, consistent with predictions from activation-based theories such as ACT-R. Our results demonstrate that race models can reproduce key empirical patterns associated with fan effects during memory acquisition.