During the revision of Standard ISO 9612, the method employed to estimate measurement uncertainty led to long discussions within the ISO working group, although the main objective was clear: to provide an estimation method in conformity with the Guide to the expression of uncertainty in measurement. To clarify the nature and the origin of the choices of the ISO working group, this article firstly describes the methodological difficulties which arise when the uncertainty related to noise sampling must be estimated. Two types of difficulties occur: i) the non linearity of the mean energy criterion, which defines the basic equation used to calculate the mean sound level from a sample of N measured values of equivalent A-weighted sound pressure levels; ii) the need to provide a method both easy to use and reliable, even when the samples are of reduced size. Thus, the method finally selected by the ISO working group is presented by giving further information on its statistical bases. The estimator selected is without statistical bias and it is exact, if the assumption of a normal distribution of the measured values is valid; it is in conformity with the mean energy criterion.