A source scaling analysis is presented based on network-averaged, teleseismic P-wave spectra determined from short-period data recorded from a sample of 20 Pahute Mesa explosions. These explosions, which were all detonated below the water table in saturated tuff/rhyolite emplacement media, cover a range of announced yields from 155 to 1300 kt. The spectra were analyzed using a simple set of source and propagation models consisting of a Mueller/Murphy source coupling model, a conventional, frequency-independent t* model of anelastic attenuation and a “quasi-linear” description of the surface-reflected pP phase. It is demonstrated that these models can account for virtually all the observed spectral variability over the frequency band extending from 0.5 to 2.0 Hz, down to a level which is close to that associated with measurement uncertainty. In particular, the use of network averaged model parameters of t* = 0.75 sec, an average pP/P-amplitude ratio of about 0.4 and an average source medium velocity of 3.5 km/sec reduces the spectral amplitude data from these explosions to an essentially frequency-independent constant value with an associated standard error of estimate which averages to only about 20 per cent over this frequency band.