Spectral assignment is a crucial step in the analysis of chemical structure by NMR spectroscopy, but is often limited by signal overlap. If the multiplet structure caused by spinspin coupling is suppressed, spectral resolution can be greatly increased, and the process of assignment correspondingly simplified. Collapsing a multiplet to a single "pure shift" peak can be achieved using a variety of methods, as has been demonstrated in 1D NMR spectroscopy, and in one dimension of several different homonuclear 2D experiments. [1][2][3][4][5][6][7][8][9][10] Recently, it has been shown that applying covariance processing to a TOCSY dataset decoupled in one dimension gives a fully decoupled 2D spectrum. [11] Previous routes to fully decoupled 2D spectra have used pattern recognition postprocessing applied to conventional, fully coupled 2D data. [12][13][14] The considerably increased resolution in, and simplicity of, fully decoupled 2D spectra makes them particularly attractive for automated structure elucidation.Here we compare two different decoupling methods, illustrating the very general nature, and potential for resolution gain, of pure shift covariance NMR spectroscopy, by applying different single decoupling methods to the NOESY and nQF-COSY experiments, producing doubly pure shift 2D spectra by covariance processing. Using pure shift acquisition for one spectral dimension greatly improves resolution, collapsing multiplet structure to singlets, whereas the use of covariance processing gives a further gain in interpretability by condensing the correlation information into 2D singlets.Methods for achieving broadband homonuclear decoupling in 2D NMR spectroscopy include the Zangger-Sterk (ZS) experiment and its adaptations; [1][2][3][4] the Pell-Keeler (PK) 458 projection of phase-sensitive 2D-J data; [5] the "BIRD" (bilinear rotation decoupling) family of sequence elements; [6,7] and the "constant time" (CT) approach. [8][9][10] All such methods trade sensitivity for resolution, to a greater or lesser extent. The first three methods are appropriate for experiments that generate in-phase cross-peaks, whereas CT