The purpose of this paper is to discuss what rationality in probability judgment is. We argued that rationality should be discussed by axioms which rational persons should obey. Coherent rules can be derived from the axioms and they can be used to check whether each judgment is coherent or not. It seems to us that only Bayesian approach has had an adequate axiomatic development, although the same kind of axiomization for other fuzzy measures seem to be possible. Next, we examined ordinary and intuitive probability judgments, especially in terms of the results obtained from three evaluation problems, which are called;Alarm set and Theft;Two Urns, and;Three Prisoners Problems. Our conclusion is that, even when the subjects responses appear to be very different from the numerical value calculated by Bayes theorem and the like, it can be explained why the subjects modified coherent probability judgment. We believe that by being educated suitably, they can be more coherent evaluator of probabilities.