Fig. 2: Stability and precision of the perceptual decisions made in the two tasks.

a Top: percentage of agreement across the two tasks (computed as the percentage of trials in which stimuli were classified similarly: voices classified as certain and honest versus doubting and lying correspond to an agreement). White asterisks show the significance of the result of the two-sided t test comparing the percentage of agreement between tasks with chance level (50%) and reported in the main text, with *** corresponding to p < 0.001. Bottom: normalized (z-scored) confidence ratings averaged separately for agreements and disagreements. Black asterisk shows the result of the two-sided t test comparing confidence for agreements versus disagreements reported in the main text, with *** corresponding to p < 0.001. Data are presented as mean values with error bars showing the 95% confidence interval. Dots show individual data. b Top: percentage of agreement within each task, computed as the percentage of double-pass trials in which stimuli were classified similarly. White asterisks show the significance of the result of the two-sided t test comparing the percentage of agreement within each task with chance level (50%) reported in the main text, with *** corresponding to p < 0.001. The black asterisk shows the results of the two-sided t test comparing the two tasks reported in the main text; *p = 0.02. Bottom: confidence ratings depending on agreement in the honesty (green) and certainty (blue) tasks. Green (honesty task) and blue (certainty task) asterisks show the result of the two-sided t test comparing confidence for agreements versus disagreements within each task, with *** corresponding to p < 0.001. Data are presented as mean values with error bars showing the 95% confidence interval. Dots show individual data. c Probability of responding that the first voice (p(choose S1)) sounds more certain (left, blue) or honest (right, green) as a function of the area under the curve computed by subtracting sensory evidence for the first minus the second stimuli, summed for the three acoustic dimensions. Darker lines correspond to high confidence trials (above the median) and lighter lines to low confidence trials (below the median). Circles show mean values and error bars the 95% confidence interval. d Average confidence, sensitivity, metacognitive sensitivity, and efficiency in the honesty and certainty tasks. Data represent mean values with error bars showing the 95% confidence interval, and dots show the individual data; black asterisks show the result of the two-sided tests comparing the two tasks, and white asterisks show the results of two-sided tests against chance level; t tests were used for confidence (normally distributed data), and Wilcoxon signed-rank tests for sensitivity, metacognitive sensitivity, and efficiency (non-normal data); *p < 0.05; **p < 0.01; ***p < 0.001; confidence: p values for the comparison between tasks p = 0.037; sensitivity: p values testing the difference with chance level, for certainty p = 0.0011/honesty, p = 0.012; comparison between tasks, p = 0.01; metacognitive sensitivity (0.0004/0.034/0.026); metacognitive efficiency (0.0004/0.01/0.72). Source data are provided as a Source data file.