Fig. 6: Experiment 2 results.
From: Prior information differentially affects discrimination decisions and subjective confidence reports
a Decision level model results against data.The left panel shows the observed accuracies per posterior level and condition (N = 25 participants). The right panel shows the predicted accuracies generated from sampling the fit posterior group mean parameter distributions 1000 times, and simulating 720 trials per participant for each of those sampled parameters, also using each participant’s staircased coherences, internal noise and decision bias. Error bars capture standard deviation (SD) of accuracies across participants. b Confidence level model results against data. On the left are the data, showing confidence following correct and incorrect decisions per posterior level and condition (N = 25 participants). On the right is mean confidence generated from sampling the fit posterior parameter distributions 1000 times and simulating 720 complete trials per participant based on those parameters, also using each participant’s staircased coherences, internal noise and decision bias. Note that here, unlike in Fig. 4b, we show results from the hierarchical model for which, due to the necessary simplification, the underestimated confidence is expected. However, this does not interfere with our interpretation of the differences between conditions or wconf. Error bars capture standard deviation (SD) of mean confidence across participants. c Posterior group difference distribution of wchoice − wconf. The posterior distribution for the difference in the group mean parameters wchoice and wconf. The blue shaded region shows the 89% credible interval and the vertical black dashed line reflects no difference in the two parameters (wchoice − wconf = 0). 0 is excluded from the 89% credible interval, suggesting wchoice and wconf to be credibly different from one another. Source data are provided as a Source Data file.
