The use of generalized correlation analysis (Noda, I. Appl. Spectrosc. 1993, 47, 1329-1336) for processing two-dimensional arrays of NMR data is described. This analysis produces complex two-dimensional spectra whose cross-peak intensities are related to correlations in the responses of pairs of signals to systematically incremented perturbations. The technique extends and generalizes the applicability of two-dimensional NMR by allowing model-independent analysis of nonperiodic signals as well as model-dependent analysis of such signals. When applied to diffusion-ordered NMR data, the processing scheme produces two-dimensional output spectra having two frequency axes. Relative diffusion coefficients are encoded in the signs and intensities of the cross-peaks. Key proper-des of the resulting spectra are model-independent, so the approach provides an alternative to traditional DOSY processing and offers advantages for data sets that do not provide pure exponential or Gaussian response curves. When data do conform well to a known response function, the technique provides a method for extracting descriptors in a two-dimensional plot having one axis corresponding to the descriptor and the other axis corresponding to the usual chemical shift scale. Finally, the technique may be used to identify differences in the response functions of closely related samples, generating a one-dimensional spectrum with signals at frequencies whose response functions differ between two samples.