A method to compute the variations in lithospheric elastic thickness (Te) has been developed, using the wavelet transform. The technique, which uses a superposition of two-dimensional Morlet wavelets in a geometry named a 'fan' wavelet, is designed to yield isotropic yet complex wavelet coefficients for the co- and cross-spectra of gravity and topography data. These are then used to compute a spatially-varying, isostatic coherence, from which both global and local estimates may be obtained. We applied the method to synthetic gravity and topography generated for a thin elastic plate of uniform thickness 20 km, yielding an apparent, spatially variable Te of 24.5 ¿ 3.5 km. The estimated global coherence for this model appears to fit the theoretical prediction as well as Fourier transform-based estimates, and is smoother than these. We also computed the wavelet coherence, and hence spatially-varying Te, for a plate of non-uniform thickness, yielding a difference with the model of -2.0 ¿ 1.7 km.