Although widely used to compare the performances of capillary columns the separation number is a poor index of the quality of a column. It is shown that it is a rubber ruler whose length is a function of temperature: the separation number of a column is a function not only of the effective plate number, which itself varies largely with the retention as measured by the capacity factor k, but also of the relative retention of then-alkanes used for the measurement. This relative retention is strongly dependent upon the temperature for a given stationary phase. Unless the temperature at which the separation number is measured be normalized as well as the k range in which the second alkane of the pair used is eluted, the separation numbers are meaningless for the appreciation of the performance of a column or comparison of different columns. Using alkanes eluted in the same k range to measure the separation number of a column we obtained a value almost twice as large at 80°C as at 180°C. In spite of their imperfections the plate number and the coating efficiency are easier to use and less misleading.