To investigate, through simulations and in vivo experiments, the extent to which B₀ field instability impacts MM-suppressed measurements of GABA. J-difference editing of GABA relies on selective manipulation of the coupling between the GABA signals at 1.9 ppm and 3 ppm. However, there is a similar coupling between macromolecule (MM) signals at 1.7 ppm and 3 ppm, and editing pulses applied (to GABA spins) at 1.9 ppm partially invert MM signals (at 1.7 ppm), resulting in co-editing of MM in the GABA-edited spectrum. This most widely used edited experiment therefore measures a signal referred to as ‘GABA+’ which has a ~50% MM contribution. It is possible to use a symmetrical editing scheme to suppress the MM contribution (1), applying editing pulses at 1.9 ppm in ON experiments and 1.5 ppm in OFF experiments. Thus, MM signals at 1.7 ppm are inverted to equal degrees in the two halves of the experiment and are removed on subtraction. There is great interest in using the MM-suppressed GABA experiment, as it avoids the confound of co-edited MM signal. However, there are also concerns over the extent to which scanner field drift and subject motion may adversely impact symmetrical MM suppression. GABA+ editing simulations. Bloch equation simulations were used to define the inversion envelope of a 14-ms sinc-Gaussian (sG) editing pulse (commonly used for GABA+ experiments). This function defines the change in editing efficiency of GABA as the B0 field offset changes. The editing efficiency of MM is represented by the same function offset by 0.2 ppm , and the net editing efficiency of GABA+MM in the GABA+ experiment is the sum of these functions (shown in Figure 1A). MM-suppressed GABA simulations. The editing efficiency of the MM-suppressed experiment (which uses 20 ms editing pulses for improved selectivity) is the difference between the inversion profiles of two 20-ms sG pulses offset by 0.4 ppm. The editing efficiency of MM is represented by the same function offset by 0.2 ppm, which has a zero-crossing (as required for suppression) for zero frequency offset. As before, the net editing efficiency of the MM-suppressed experiment is the sum of these two functions (shown in Figure 1B). 16 MM-suppressed GABA-edited spectra were acquired in 7 healthy adults with the following parameters: TR 2s TE 80 ms (2), 320 averages of 2048 datapoints sampled at 2 kHz; (36 mm)3 midline parietal voxel; VAPOR water suppression; 20 ms sG editing pulses applied at 1.9 ppm and 1.5 ppm. Data were acquired under a range of conditions of B0 stability, some deliberately preceded by imaging to generate a range of gradient heating and associated B0 field drift. GABA levels were quantified relative to the unsuppressed water signal from the same volume, using the ‘Gannet’ program (3).As can be seen from the simulations in Figure 1C, the slope of the MM–suppressed experiment is substantially steeper than that of the GABA+ experiment. MM-suppressed GABA measurements are almost 8 times more susceptible to B₀ offsets (which naturally arise from scanner heating/cooling and/or subject motion), such that even a 1 Hz offset changes the measured signal by 5%. In vivo spectra shown in Figure 2, acquired with zero offset (A) and extreme negative (B) and positive (C) offsets show very different 3-ppm 'GABA' signals. As predicted by simulation, in vivo results show a linear relationship between measured GABA levels and the mean B₀ field offset during the measurement, as shown in Figure 3. Subtraction artifacts (and their post-processing corrections) related to field offsets have received much attention in the literature (4,5). The second impact of field changes, that editing pulses are no longer applied directly to the intended spins, has received less attention, but this effect has a serious impact on MM-suppressed GABA editing. Small field offsets rapidly generate positive or negative MM signal; therefore it is not advised to perform MM-suppressed measurements of GABA unless accurate prospective frequency correction methods are available. Additional work is required to establish prospective field-frequency locking.