A methodology is presented for determining the yield of underground nuclear explosions from Lg wave amplitudes. The methodology is applied to Nevada Test Site (NTS) explosions, for which the data from short-period, vertical component analog seismographs at three stations are used to develop calibration curves for unsaturated material and water-saturated rock source conditions. The latter curves are found to provide reasonably accurate estimates of the yields of explosions in other areas of the United States and in the French Sahara, suggesting that they may be applicable to all continental areas. If so, they also can provide an estimate of the bias of mb(P) magnitudes between different continental sites. For example, the Lg data from NTS explosions indicate a 0.31±0.02 magnitude unit bias between NTS and eastern North America, similar to the approximately 0.33 unit bias found between western and eastern North America previously by use of earthquake data.