P-wave attenuation (expressed in terms of t* = T/Q, the ratio of travel time to the P-wave quality factor Qα) was measured for a large number of teleseismic paths using spectral ratios for nuclear explosions in the short-period band 0.5 to 4 Hz. If these new t* values along with other data in the literature are subdivided into two groups, one involving wave paths which descend and emerge entirely under shields or stable platforms and the other involving paths with either end of the wave path under the Western United States (WUS), a definite difference can be found between the average t* values of the two groups. The difference in attenuation, which gives a t* difference greater than 0.2, is sufficient to explain the fact that teleseismic events appear to be, on the average, about 0.26 units lower in magnitude when measured at WUS stations as compared to Eastern United States (EUS) stations, assuming that the observed dominant frequency of teleseismic P waves is about 1 Hz. A similar difference exists in the spectra of teleseismic short-period S waves observed in WUS relative to those seen in EUS.
The observations indicate that Qα is high Qα > 1500) throughout the upper mantle under shield areas while the WUS (with the possible exception of the Colorado Plateau) is underlain by a low-Q layer. If this layer coincides in vertical extent with the upper-mantle low-velocity layer, the Qα there should be about one-half that of the value given by Archambeau et al. (1969). The t* values for short-period P waves are considerably lower than those derived by the other authors from body waves in the long-period band, indicating that Qα is frequency dependent and increases with frequency.