Repeatability and reproducibility are usually defined by the relationR=k√2sR, whereRstands for repeatability or reproducibility, andsRis the corresponding standard deviation. It is then stated that the probability is C that a difference between two test results will lie between (−R) and (+R). ForC= 0.95, which is the usual choice, the values that have been proposed for the multiplierkare 1.96, 2, or Student'st.However, C is actually a random variable with a highly skewed distribution. It is shown that regardless of the above choice ofk, the probability that C will lie in the“vicinity of 0.95,”such as for example 0.92 to 0.97, is very small, unless the number of participating laboratories is large (30 or more). Nevertheless, for any given interval defining a“vicinity of 0.95,”a value ofkexists that maximizes the probability that C lies in that interval. For a number of reasonable choices for“vicinity of 0.95”the optimalkis close to 1.96.