We present evidence of significant variations in time-domain amplitude and spectral estimates of local earthquake recorded by small-aperture arrays. We examine data from three arrays: two arrays of different scales deployed at Piñon Flat, California, in 1990 and 1991 and an array deployed in the summer of 1993 in the former Soviet republic of Turkmenistan. We find consistent evidence in all cases for significant variations in measured amplitudes over scale lengths comparable to feasible wavelengths of incident wave fields. This phenomenon, however, is strongly frequency dependent. At the Piñon Flat site, variations in power spectral estimates exceed a factor of 100 at frequencies over 4 Hz. Analysis of teleseismic signals, however, demonstrates that these variations diminish rapidly at lower frequencies and falls to negligible levels below 0.6 Hz. The Turkmenistan site shows similar overall characteristics; high-frequency variations are less dramatic. Variations comparable to the Piñon Flat site do not occur below 20 Hz. Analysis of teleseismic signals yield results similar to Piñon Flat, although the transition to negligible variability seems to occur at a slightly lower frequency of 0.3 Hz. The 1990 Piñon Flat experiment utilized simultaneous recording in two boreholes directly beneath the array. Comparison of spectral estimates from these boreholes to the surface sensors strongly suggests that the deviations in high-frequency spectral estimates we observe across the array are due to interaction of the wave field with the near-surface, weathered layer. We suggest the differences in high-frequency variations of signals recorded at Piñon Flat compared with that of the Turkmenistan site can be explained by differences in near-surface conditions at the two sites. The low-frequency transition to negligible signal variation seen at both sites occurs when the array aperture becomes small compared with the wavelength of surface waves, suggesting that signal variations in intermediate frequencies may be influenced by body-wave to surface-wave conversions caused by crustal velocity variations and topography.