Figure 1: Continuous tracking task and corresponding state-space model for studying adaptation. (A) Screenshot showing the Gabor target (circled for clarity; the actual task does not show the circle) in a field of larger Gabor distractors. The task requires the participant to track the movement of the target using a mouse, with or without pre-exposure to the target to induce adaptation. (B)  A single representative tracking trace from one participant in one trial. (C) Average tracking error across different adaptation conditions and participants. There is a slight trend suggesting that adaptation leads to better tracking performance but it is not significant. (D) Statespace model to study adaptation. It is a hierarchical model where contextual dynamics affect the latent state dynamics. (E, F) Model observations across a contextual change. (G,H) Latent state and the estimate of an observer with optimal adaptation (latent and context dynamics). (I-L) Systematic analysis of the relative adaptation benefits as a function of different latent state dynamics and observation noise; relative compared to the intrinsic task difficulty (oracle risk).