A paper by R. MICHEL in this journal (J. Radioanal. Nucl. Chem., 245 (2000) 137) describes, among other things, a numerical example: the calculation of characteristic limits in an analysis of 129I via radiochemical NAA. A key problem in Bayesian characteristic limits is a proper formulation of the expression for the uncertainty associated with the measurand if its true value is equal to zero, ũ(0). We show that MICHEL's expression for ũ2(0) is not correct and propose a new one. Some numerical errors found in the paper are also indicated.