Figure 1: Motor variability is regulated by recent reward history.

A. Schematic of the experimental task used to study the relationship between performance and motor variability. Rats learn to press the 2-dimensional joystick close to a target angle (red) to receive water reward.

B. Example performance of a rat on the motor task. Trials are rewarded whenever the joystick press-angle (black dots) falls within the reward boundary (red shading). Reward boundaries are updated to shape rats’ press-angles towards the target (red line). Once a rat’s mean press-angle matches the target, a new target is automatically selected. Colored arrows and dots indicate example sessions whose trajectories are shown in panel C.

C. Joystick press trajectories from three example sessions indicated by arrows in panel B. The press-angle is measured between a line of best fit (dashed black line) to the joystick trajectory (example represented by solid black line) and the vertical. See also Figures S1.

D. Zoom-in of panel C showing behavioral variables of interest for our analysis: reinforcement (top), press-angles (middle) and variance of press-angles (bottom) calculated in moving windows of 5 trials.

E. Calculating a reinforcement-triggered average of motor variability for an example rat. Average reward (top) and press-angle variability (bottom), conditioned on the probabilistic outcome – reward (red) or no reward (blue) – of a single trial (at trial 0). See also Figures S1 and S2.

F. Difference between average levels of variability in response to single unrewarded and rewarded trials, averaged across the population of rats (n=10). These plots were generated by combining the results from our matching analysis and probabilistic reward manipulations. Black line indicates mean and grey shading indicates SEM across rats. τ represents the time-constant of an exponential fit to the decay of the single-trial variability effect (after trial 0). See also Figures S1 and S2.