Network-averaged Lg spectra for 106 underground Nevada Test Site (NTS) nuclear explosions (recorded at the four broadband seismic stations operated by Lawrence Livermore National Laboratory) are shown to be approximately self-similar and to scale with cavity radius and scaled-cavity radius. Composite spectra produced by superimposing the spectra of individual tests of nearly identical yields conducted in a single geologic unit (e.g., Tunnel Beds tuff, Yucca Flat, below the water table) have amplitude variability of about a factor of 2. If a master spectrum is chosen that resembles the shape of a particular subgroup of nearly identical NTS explosions, the spectra of all other tests in the database can be superimposed on the master spectrum through the application of a shot-specific amplitude shift and a shot-specific frequency shift. The amplitude shift basically measures the difference between the low-frequency asymptotes of the individual spectrum and the master spectrum, and the frequency shift measures the difference in corner frequencies. We show that, for explosions producing cavities with a narrow range of measured scaled-cavity radii, the amplitude shifts scale with cavity volume. Further, if cavity volume scaling is applied to the amplitude shifts, the residuals are shown to depend on scaled-cavity radius or gas-filled porosity. The residuals (a measure of seismic coupling) appear to show slightly reduced scatter when plotted versus scaled-cavity radius as opposed to gas-filled porosity. The frequency shifts scale approximately as cavity radius to the 3/4 power. Although the NTSLg spectra are approximately self-similar, differences in high-frequency decay and other shape parameters exist. For example, the rate of high-frequency decay is observed to progressively decrease as explosions approach the static water level from above. These differences do not have a significant effect on the connections established between the amplitude and frequency shifts with cavity radius and other source parameters but do suggest source differences of potential interest in discrimination studies.