Recently it has been demonstrated that dramatic spectral simplification can be obtained by editing a regular ‘H NMR spectrum into a set of subspectra by means of the one-dimensional HOHAHA experiment (I). As pointed out earlier, scalar couplings are difficult to measure from such a spectrum because of phase and intensity distortions of the muhiplet components in the various subspectra. Here, it is demonstrated that pure phase subspectra can be obtained readily by combination of the 1D HOHAHA experiment with the z-filtering technique (2). Guidelines for minimizing the number of T averages, needed for the z filter, are also presented. If the Zeeman term from the Hamiltonian is eliminated, the spin system will evolve solely under the influence of scalar couplings. Braunschweiler and Ernst (3) lirst utilized this concept in isotropic liquids for generating total coherence transfer (TOCSY) 2D spectra of spin systems of coupled protons. A number of more efficient mixing schemes based on the same mechanism have since been reported (1, 4, 5). We prefer to use the so-called MLEV- 17 sequence (4) for mixing because it covers a substantial bandwidth (kO.3~) and because it is relatively insensitive to exact phase shifting or pulse width adjustment. Moreover, this mixing scheme lengthens the apparent Tip (4, 5) so that less signal is lost during this mixing period relative to other mixing schemes. To record pure phase spectra, Braunschweiler and Ernst proposed cc&ding a number of spectra, recorded with different mixing times. This procedure works satisfactorily when few spins are involved in a J-coupled network. For larger spin systems however, this procedure is not perfect because the magnetization transfer appears to lose its oscillatory character, i.e., the oscillation is overdamped. Consequently, the spectra obtained are not 100% absorptive and measurement of scalar couplings may lead to errors. Moreover, in a large network of coupled spins (alkaloids, steroids, etc.) averaging over a number of mixing times, some of which would have to be quite long, gives a subspectrum of nearly the entire molecule and therefore defeats the purpose of the experiment. As demonstrated here, the z-filtering technique, proposed by Slarensen