Probability judgments about logical propositions have raised substantial doubts about human rationality. Here we explore the idea that people’s probability judgments often may not refer to the relative frequency of a set, but instead to the probability of an explanatory logical pattern given the data. This idea has been formalized by Bayesian logic (BL), predicting a system of frequency-based logical inclusion fallacies. The studies presented concentrate on comparing probability judgments about sentences logically relating two attributes of a class or an individual (humans, animals, artifacts). Although BL cannot model probabilities of individual predications directly, it can do so if one assumes that inferences are made about unknown individuals based on imagined samples. The results for general as well as individual predication show a high number of systematic inclusion fallacies in line with BL. Nevertheless, some deviations were found. In the General Discussion, a polycausal approach to inclusion fallacies is advocated.