Frequency-dependent amplification of seismic waves by near-surface low-velocity layers is a well-known phenomenon. This phenomenon was examined from both the analytic and experimental viewpoints for body waves (P, SV, SH). Groundmotion data, recorded in conjunction with the underground nuclear testing program at the Nevada Test Site, are used to provide experimental validation of the analytic models. Experimental amplification factors are derived from these data for a variety of recording-site near-surface geological configurations (alluvium, mine tailings, fill) and a wide dynamic range of ground-motion intensity (10−5 to 100 g). The variability in the mean amplification observed at a site for repeated measurements is described statistically. This analysis shows that, although the amplification at a given site varies on the average by a factor of about 1.4 across the frequency band of interest, from detonation to detonation, the frequency and magnitude of the dominant amplification are fairly consistent. The quality of the comparisons of the observed and calculated amplification indicates that the available linear analytic models are capable of describing the major features of the frequency-dependent amplification observed for this wide range of groundmotion intensity and recording-site geology.