Human temporal bone specimens are used in experiments measuring the sound transfer of the middle ear, which is the standard method used in the development of active and passive middle ear implants. Statistical analyses of these experiments usually require that the TB samples are representative of the population of non-pathological middle ears. Specifically, this means that the specimens must be mechanically well-characterized. We present an in-depth statistical analysis of 478 data sets of middle ear transfer functions (METFs) from different laboratories. The data sets are preprocessed and various contributions to the variance of the data are evaluated. We then derive a statistical range as a reference against which individual METF measurements may be validated. The range is calculated as the two-sided 95% tolerance interval at audiological frequencies. In addition, the mean and 95% confidence interval of the mean are given as references for assessing the validity of a sample group. Finally, we provide a suggested procedure for measuring METFs using the methods described herein.