Fig. 6: Continuous model of criterion updating.

a Behavior of the model for the same 10-trial sequence illustrated in Fig. 4a. Green trajectories are the model’s computed decision criterion, with τ = 30 s. For comparison with the continuous model, the discrete model computation is shown by the dotted line. b Gray traces give the observed trial n psychometric function midpoints, in the manner of Fig. 3a, as the ITI from n − 1 to n grows, smoothed with a moving average window of 8 bins. Traces are sorted according to Δspeed of stimulus n − 1 (see gray scale). Green traces give the model’s predicted trial n psychometric function midpoints, as the ITI from n − 1 to n grows. c Comparative performance of the continuous and discretized history models. The same GLM of Fig. 3f (here shown in gray) was fitted to artificial choices simulated from trial to trial with the discrete and the continuous model. d Optimized time constants, τ, are estimated in each cross-validation partition and are shown for both discrete and continuous models in six rats. The size of the points corresponds to the average performance of the rat in the selected subset of sessions. Source data are provided as a Source Data file.