Data taken from the University of Nevada (UNR) array of surface-located wideband, digital, displacement seismographs placed over the 1980 Long Valley earthquake sequence show very strong site effects, which control spectral parameters and hence inferred source parameters, up to ML5.0. Observations from 4,000 seismograms show that differences in site effects can lead to inferred seismic moments at least a factor of two lower than previously reported. Observed S-wave spectral corner frequencies show upper-bound cutoffs which are widely disparate at different sites (e.g., 15 Hz at MG1 and 6 Hz and WT1), thus fmax is quite variable by site. This high-frequency cutoff, clearly a site effect at WT1, could be misinterprted as an apparent depature from source similarity for seismic moments below 1021 dyn-cm. Estimates of seismic moment and source duration from the UNR ''rock'' sites may be more accurate than those previously published, both because the UNR instrumentation has greater resolution at longer periods (0.1 to 1.0 Hz) and because of less near-site signal distortion at high frequencies. ¿American Geophysical Union 1991