Figure 1.

The influence of sensory evidence on choice and confidence derived from a motion discrimination task. (A) A trial of the motion task. The motion stimulus was presented for 700 ms after which participants signaled the response with an eye-movement and made a confidence report on a continuous scale ranging from complete chance on the left to absolute certainty on the right. Motion coherence was set to maintain accuracy at 67%. On average, confidence reports on correct trials were higher than on incorrect trials (one-tailed paired t-test, p < 5.10⁻⁸, df = 18). (B) Influence of motion energy fluctuations on choice. Motion energy fluctuations were obtained by applying a filter to the random dot stimulus and subtracting the mean motion energy set by the trial's motion coherence and direction. Upward and downward fluctuations were derived independently, and then averaged for the selected (blue) and non-selected (orange) directions. On average, fluctuations made motion energy in the selected direction to be above the mean, and motion energy in the opposite direction to be below the mean. (C) Time course of the decision kernel, obtained as the difference between E_S(t) and E_N(t). (D) Motion energy fluctuations for the selected and non-selected directions split according to confidence ratings into high (solid line) and low (dotted line) confidence. (E) Time course of the confidence kernel, showing that motion information had an early effect on confidence. (F) Motion energy fluctuations for the selected and non-selected directions, after subtracting low confidence trials from high confidence trials (shown in panel D). In panels B–F, shaded regions indicate SEM. Motion energy signals were smoothed with a Gaussian kernel with standard deviation of 40 ms.