In using inverse-time relays to give discriminative protection on an electricity supply system, it is general practice to calculate the settings of the relays, using an arbitrary discriminative time (generally 0.5 sec) between the operating times of successive relays at the maximum calculated fault current. This time interval has not always been found satisfactory, and an investigation of the components that make up the discriminative time is described herein.The component due to calibration and testing tolerances on the relays is dealt with statistically, and statistical control of testing is discussed. The effects of ageing of relays and of variation of characteristics on distorted waveforms are also examined.A method of measurement of the overswing characteristics of these relays is described, and some results are given for different designs. It is shown that the overswing has a marked effect on the discriminative time, and varies considerably with different designs of relay.The investigation of switch-clearing time is considered to be outside the scope of the paper, but, taking suitable figures for this, comparisons are made between discriminative times arrived at by the methods described in the paper, and the previously accepted arbitrary time of 0.5 sec.