Gerd Gigerenzer and colleagues propose a comprehensive theory of confidence in knowledge, called the theory of probabilistic mental models (PMM theory). This theory explains two key effects in confidence studies: the overconfidence effect (mean confidence exceeds percentage of correct answers) and the hard-easy effect (overconfidence increases with question difficulty). It also predicts a new phenomenon, the confidence-frequency effect, which is a systematic difference between confidence in a single event and the frequency of correct answers in the long run. Two experiments support PMM theory by confirming these predictions and explaining anomalies in previous studies. PMM theory suggests that people are good judges of the reliability of their knowledge if it is representatively sampled from a reference class. It integrates the overconfidence and hard-easy effects and specifies conditions under which these effects appear, disappear, or invert. The theory is based on the idea that confidence judgments are influenced by the structure of the task and the use of probabilistic information from the environment. The theory distinguishes between two types of mental models: local mental models (LMMs) and probabilistic mental models (PMMs). LMMs are based on direct memory and logical operations, while PMMs use probabilistic information from a natural environment. PMMs are used when LMMs fail and involve inductive inference based on a reference class and probability cues. PMM theory predicts that confidence judgments are influenced by cue validity, which is the probability that a cue correctly predicts the target variable. It also predicts that confidence and frequency judgments refer to different reference classes, leading to systematic differences between them. The theory also predicts that overconfidence disappears when questions are randomly sampled from a natural environment, but frequency judgments may show underestimation. The theory is tested in two experiments. In the first experiment, subjects were asked to answer general-knowledge questions and provide confidence judgments. The results showed that overconfidence was present in the selected set of questions but not in the representative set. Frequency judgments were accurate in the representative set but showed underestimation in the selected set. In the second experiment, the results confirmed these predictions, showing that overconfidence disappeared when questions were randomly sampled from a natural environment. The results support PMM theory and show that confidence and frequency judgments are influenced by different reference classes. The theory provides a comprehensive framework for understanding confidence in knowledge and explains the overconfidence and hard-easy effects. It also predicts the confidence-frequency effect, which is a systematic difference between confidence in a single event and the frequency of correct answers in the long run. The theory is supported by the experimental results and provides a new understanding of how people judge their knowledge.Gerd Gigerenzer and colleagues propose a comprehensive theory of confidence in knowledge, called the theory of probabilistic mental models (PMM theory). This theory explains two key effects in confidence studies: the overconfidence effect (mean confidence exceeds percentage of correct answers) and the hard-easy effect (overconfidence increases with question difficulty). It also predicts a new phenomenon, the confidence-frequency effect, which is a systematic difference between confidence in a single event and the frequency of correct answers in the long run. Two experiments support PMM theory by confirming these predictions and explaining anomalies in previous studies. PMM theory suggests that people are good judges of the reliability of their knowledge if it is representatively sampled from a reference class. It integrates the overconfidence and hard-easy effects and specifies conditions under which these effects appear, disappear, or invert. The theory is based on the idea that confidence judgments are influenced by the structure of the task and the use of probabilistic information from the environment. The theory distinguishes between two types of mental models: local mental models (LMMs) and probabilistic mental models (PMMs). LMMs are based on direct memory and logical operations, while PMMs use probabilistic information from a natural environment. PMMs are used when LMMs fail and involve inductive inference based on a reference class and probability cues. PMM theory predicts that confidence judgments are influenced by cue validity, which is the probability that a cue correctly predicts the target variable. It also predicts that confidence and frequency judgments refer to different reference classes, leading to systematic differences between them. The theory also predicts that overconfidence disappears when questions are randomly sampled from a natural environment, but frequency judgments may show underestimation. The theory is tested in two experiments. In the first experiment, subjects were asked to answer general-knowledge questions and provide confidence judgments. The results showed that overconfidence was present in the selected set of questions but not in the representative set. Frequency judgments were accurate in the representative set but showed underestimation in the selected set. In the second experiment, the results confirmed these predictions, showing that overconfidence disappeared when questions were randomly sampled from a natural environment. The results support PMM theory and show that confidence and frequency judgments are influenced by different reference classes. The theory provides a comprehensive framework for understanding confidence in knowledge and explains the overconfidence and hard-easy effects. It also predicts the confidence-frequency effect, which is a systematic difference between confidence in a single event and the frequency of correct answers in the long run. The theory is supported by the experimental results and provides a new understanding of how people judge their knowledge.