Frequency-dependent spatial attenuation of Pn waves, incorporating the combined effects of geometrical spreading and anelastic dissipation, is investigated in eastern Canada between 3 and 15 Hz. Measurement of this total attenuation, instead of the two separate effects, circumvents the usual need for making a priori assumptions regarding the Pn geometrical spreading rate. The data consist of 77 Pn spectra, all amplitude-normalized using a set of previously measured spectral ratios of eastern Canadian earthquake sources. The normalization procedure eliminates the need for explicit source spectral assumptions, another common source of error in Pn attenuation measurements. An unconventional technique is introduced to account for differences in observed Pn spectra, which arise from geological site effects and instrument response error, both of which are frequency-dependent phenomena. We conclude that, at each frequency, the Pn amplitude falls off with distance according to: Δ−n, where Δ is epicentral distance. The exponent n is weakly frequency-dependent and takes the form: 2.2 + 0.02f. A unique interpretation of the behavior of the Pn spectral amplitude decay is unattainable, owing to the difficulty in separating the effects due to anelastic dissipation and the velocity structure in the uppermost mantle. It is interesting to note, however, that the spatial decay of pure elastic head waves follows the classical Δ−1/2L−3/2 relation, where L is the distance the waves travel in the mantle refractor. For the distance range of our interest (260-1087 km), this amounts to Δ−2.2. This indicates that towards low frequencies our n approaches a value which is consistent with the spatial decay rate expected of pure elastic head waves.