Spectral editing of the GABA spin system is hampered by coediting of macromolecule (MM) coherences. To reduce contamination arising from MMs in spectra edited for GABA, the highest field strength currently available for human experimentation (7 Tesla) and MEGA-based editing were used. Despite judicious choice of experimental parameters, MM contamination was found to arise from field drifts. When the MM contribution was accounted for, [GABA] = 0.75 +/- 0.14 micromol/g (mean +/- SD, N = 16) relative to 8 micromol/g creatine (Cr), whereas without accounting for the MM signal [GABA*] = 0.88 +/- 0.23 micromol/g (mean +/- SD, N = 16). Incorporating the direct experimental assessment of MM contamination to the edited GABA signal substantially reduced the variance of the measurement, resulting in concentrations that were in excellent agreement with previous (13)C labeling experiments.