According to the Self-consistency model of subjective confidence (Koriat, 2012), confidence judgments for two-alternative forced-choice questions are related to the consensuality of the answer (the proportion of participants who choose the answer) rather than to its accuracy. Jackson (2016) extended the Self-consistency model to open-ended questions and proposed that higher response cardinality (the number of unique response options) should reduce confidence. We tested the assumptions of the Self-consistency model using the syllogistic problems. Participants solved 64 syllogistic problems. All logically possible pairs of premises regarding mood of premises and syllogism figure were included. Participants were asked to generate a conclusion to each problem and to make confidence judgments. The relationships between accuracy, confidence judgments, consensuality and cardinality were analysed. The obtained results strongly supported the predictions of the Self-consistency model. First, confidence was not related to response accuracy. Second, confidence was negatively correlated with response cardinality. Problems with higher number of unique responses were associated with lower confidence ratings. Third, confidence was positively correlated with response consensuality. Answers endorsed by large proportion of participants were given with high confidence, regardless of their accuracy. The obtained results clearly point to the conclusion that confidence judgments do not monitor the actual reasoning performance, but that they rely on consistency (indicated by the consensuality of the response) and response cardinality. |