Figure 1.

(A) Schematic representation of an experimental trial with deviation angle $\theta$. The dotted line represents the participant’s hand movement and the continuous line represents the rotated movement observed on the screen. (B) Experimental protocol. The continuous line represents the deviation angles $\theta$ imposed during one experimental cycle, where trials 1 to 25 constitute the forward process and trials 34 to 58 constitute the backward process. The dotted line represents the beginning of the next cycle. (C) Illustration of the equilibrium distributions (2) with $b,{\theta }_n=0$ resulting from the exponential quadratic error (1) and, respectively, $\beta =1,1.5,2$. The shaded area represents the target, which tolerates, at most, an error of $2^{\circ }$. (D) Comparison between the equilibrium distributions that we fit using the initial 100 trials (before participants experience any perturbation) and participants’ performance in the washout plateaus between cycles (the sequence of trials with $\theta =0$ that separate forward and backward protocol), to check whether participants equilibrate between cycles, as required by the fluctuation theorem. Red shows the normalized error histogram for the in-between plateaus exemplarily for participant 7, green shows the histogram of the fitted equilibrium distribution for the initial block of 100 trials of the same participant. The comparison for all other participants can be found in Fig. 7.