Figure 6.

The Bayesian Model Predicts Belief Reports

(A) Evolution of beliefs for six example blocks. The colored thick lines are the data. The thin lines represent predictions from ten simulations of the Bayesian model using the same sequence of trials (motion strength, duration, and direction) as seen by the subjects. They differ because the evidence is sampled randomly on each trial $\left(\left\{e,t_e\right\}\right)$, which may lead to differences in choice and confidence.

(B) Evolution of belief as a function of the trial number within the block. Lines are predictions of the model, and the shaded areas show the SE of the data (as in Figure 2D).

(C) Predicted belief for the last trial of each block plotted against the data from the same block. The colored circles identify the six blocks shown in (A). The predictions (abcissa) are from one simulation of the full experiment; the average R² indicated in the panel is an average over 200 simulations.

(D) Same comparison as in (C) but averaged by actual base rate. Error bars are SE (usually smaller than the data points); dashed line is identity. See also Figures S4 and S5.