Analysis of accelerograms recorded at the 20 m range from a buried 5-lb detonation in alluvium revealed wide (as large as 20 dB in the amplitude modulus of the Fourier transform) variations in response for frequencies above 35 Hz. Additional experiments were performed which ruled out source asymmetry or instrumental irregularity as the cause of these variations. The observations suggest that scattering by geologic inhomogeneity is responsible for the frequency-dependent spatial variability in ground motion. A thorough understanding of the physical processes responsible for this variability requires that a quantitative relationship be established between the subsurface material property information and the observed ground motion characteristics. An attempt was made to do this using available standard penetration test data from the test bed where the initial experiment was detonated. Autocorrelations of the blow count data from the standard penetration test were computed and compared with theoretical exponential and Gaussian distributions. The exponential distribution with a scale length between 2.0 and 3.0 m best matches the data. Assuming the Born approximation, a scale length of 2.0 to 3.0 m implies that significant scattering should occur above 10 Hz. The recorded ground motions are, however, coherent out to about 35 Hz, suggesting a scale length on the order of 0.5 to 1.0 m, which is beyond the resolution of the standard penetration test technique. This scale length is, however, not unreasonable in light of the general geologic characteristics of the test area.