Study of 3242 full first order synthetic seismograms of isolated normal mode multiplets shows that constructive and destructive interference of the singlets comprising a multiplet can account for the noise-corrected variance of free oscillation attenuation measurements in the frequency band 3.0-4.0 mHz. This interference gives rise to a distribution of synthetic Q measurements which is sensitive to the short wavelength velocity features of a model. While individual attenuation measurements depend also on the moment tensor, the overall variance of Q fluctuations is insensitive to the characteristics of the source employed and depends little on the length of the time series processed. The Woodhouse-Dziewonski model M84A of upper mantle velocity structure causes a scatter in the synthetics large enough to interfere significantly with any signal from lateral Q structure. Correlation of individual synthetics with data is poor, however, which suggests that either there are remaining uncertainties in the velocity model, that mode-mode coupling occurs in the data in a way that is poorly understood, or that lateral attenuative structure contributes to the observed scatter. Resolution of any such 0 structure will be difficult given the present data set.