For estimating a bounded normal mean with known variance, we exhibit situations of pronounced discrepancy between the credibility of Bayes credible regions and frequentist coverage. Analogously, frequentist confidence intervals are shown to have credibility one in some cases.