Table 3 Recommendations for metacognition researchers

Experimental design

1. Use relatively easy tasks to avoid instability related to low d’ values

2. Whenever possible, use designs with a single difficulty level

3. Collect at least 100 trials per participant

4. For individual differences research, ideally collect at least 400 trials per participant

Analysis

1. Use several measures of metacognition

2. M-Ratio continues to be a good default measure of metacognitive ability

3. If results could plausibly depend on task performance or metacognitive bias, then confirm that results remain the same when using meta-noise or meta-uncertainty

4. Do not use difference measures to correct for differences in task performance

5. If multiple conditions are present, use meta-noise or meta-uncertainty (custom modeling necessary)

Interpretation

1. Do not automatically assume that M-Ratio < 1 indicates signal loss from the decision to the metacognitive system

2. Do not automatically assume that M-Ratio > 1 indicates signal gain from the decision to the metacognitive system