Proceedings of the Annual Meeting of the Cognitive Science Society

The interception of moving targets is a fundamental sensorimotor task involving perception and action. For this task, the dominant approach has been to model the behavioral dynamics using online control laws such as the constant bearing angle strategy, which explain behavior without assuming internal models. Here, we derive a Bayesian model-based optimal control model of an interception task and compare it against the constant bearing angle strategy. First, we show that both models equivalently capture average trajectories, suggesting that observing the interception trajectories in an experiment cannot adjudicate between the two models. However, including realistic levels of perceptual uncertainty, motor variability, and sensorimotor delays leads online control without an internal model to quickly deteriorate in interception performance. We conclude that the empirically observed robustness of the constant bearing angle strategy speaks against a direct coupling of environmental variables and behavior, but instead implies some form of internal model.