Metacognitive Resolution: From Prediction to Causation via Meta-Analysis and Empirical Examination

ABSTRACT Metacognitive Resolution reflects the extent to which confidence judgments discriminate between correct and incorrect responses. We rely on these judgments to allocate time and effort across items and tasks, making it essential to distinguish between what we know and what we do not. While resolution tends to be stable within tasks, it varies across studies. Despite extensive research, the factors that influence resolution remain unclear. The current research addressed this gap through a two-part investigation. First, a meta-analysis (202 effect sizes from 79 articles) identified predictors of resolution. In particular, response format, multiple-choice (MC) versus open-ended (OE), and the number of response options were found to predict resolution. Prior work proposed that MC formats, but not OE, allow correct guessing despite low confidence, thereby reducing the accuracy-confidence relationship underlying resolution. In contrast, we hypothesized that as MC options increase, participants abandon elimination as their main strategy in favor of inner answer production, like in OE tasks, and this process improves resolution. To test this hypothesis, we conducted an experiment (N = 349) with four MC conditions (2 to 8 response options) and one OE condition. We replicated prior findings that OE resolution is stronger than in 2-options MC. However, supporting our predictions, resolution in the 8-option MC condition was equivalent to the OE condition. These findings suggest that more response options prompt deeper answering strategies, like OE tasks. Together, the meta-analysis and the experiment identify key factors that affect resolution and highlight response format as a manipulable factor supporting improvement.