We have compared different spectral methods (the standard windowed Fourier transform, the multitaper method (MTM), and the maximum entropy method (MEM)) to determine the coherence between Bouguer gravity and topography and estimate the variations of the effective elastic thickness (Te) of the lithosphere. These methods differ significantly in spatial resolution (1000 to 2000 km) determined by the width of the windows required to resolve the wavelength of transition from low to high coherence. The tests with synthetic data show that the standard deviation on elastic thickness estimates is high with the three methods but that the mean value obtained by the maximum entropy is close to the input value, while the other methods underestimate the elastic thickness. In the Canadian Shield, the effective elastic thickness values vary from ≈40 km to >100 km over distances less than the width of the moving window. There is no clear geographical or geological pattern in these variations as high and low elastic thickness values are found near the edge as well as in the center of the continent, and in all provinces regardless of their age. Correlation between Te and the heat flow field is weak. The values of Te being much higher beneath Hudson Bay than the Williston Basin are consistent with models of basin evolution and seismic estimates of mantle temperatures.