Fig. 7: Summary of results.

The figure lists the values obtained for each measure of metacognition for various criteria. Precision is the measure developed in this paper and the values listed are the average of the values in Fig. 1b, c. Higher precision values are better. For dependence of task performance and metacognitive bias, the figure lists the average Cohen’s d values reported in the paper. For dependence on response bias, the figure lists the average correlation between each measure of metacognition and the absolute value of response bias (\(\left|c\right|\)). Lower absolute value of these dependencies is better. The reported split-half reliability is the average value across datasets obtained for a bin size of 100, whereas the reported test-retest reliability (ICC) is the average value obtained for a bin size of 400. Higher reliability values are better. Color coding is meant as a general indicator but should be interpreted with caution. Green indicates very good properties, yellow indicates good properties, orange indicates problematic properties, and red indicates bad properties. Colors were assigned based on the following thresholds: 0.5 for precision, 0.3 and 1 for Cohen’s d, 0.5 for test–retest reliability. Green was not used in any of the columns regarding dependence on nuisance variables as to not give the impression that any measure is certainly independent of any of the nuisance variables. The figure also lists several unique advantages and disadvantages of each measure discussed in the main text.