A numerical procedure is outlined that is appropriate for the design of multiple quantum filter sequences targeted for the strongly coupled, multiple spin systems that occur in metabolites present in brain. The procedure uses numerical methods of solution of the density matrix equations, first, to establish the most appropriate resonance to target with the filter; second, to provide contour plots of a performance index of the filter in terms of critical sequence parameters; and third, to produce the response signals of the target and the background metabolites to the optimized filter. The procedure is exemplified for the AMNPQ spin system of the amino acid glutamate at a field strength of 3 T. The 2.3 ppm peak of the PQ multiplet of glutamate was identified as the target resonance, and the performance of the filter so derived was evaluated experimentally on phantom solutions and in human brain. These experiments clearly demonstrate that a linewidth of <or=4 Hz is required for full resolution of glutamate from glutamine at 3 T using this double quantum filter. Nevertheless, even at a linewidth of approximately 7 Hz in vivo, the 2.3 ppm peak of glutamate dominates the filter response and thereby removes a significant cause of uncertainty in measuring changes in glutamate by eliminating most of the background observed in unedited spectra obtained using PRESS or STEAM.