A simple and selective (19)F NMR method has been validated for the quantitation of fluoxetine (FLX) and fluvoxamine (FLV) in methanol solutions and in human plasma and urine. The regression equations for FLX and FLV showed a good linearity in the range of 1.4-620 microg mL(-1) (3.3 x 10(-6)-1.8 x 10(-3) mol L(-1)) with a limit of detection of approximately 0.5 microg mL(-1) (1.3 x 10(-6) mol L(-1)) and a limit of quantification of approximately 2 microg mL(-1) (4.6 x 10(-6) mol L(-1)). The precision of the assay depends on the concentrations and is comprised between 1.5 and 9.5% for a range of concentrations between 1.2 x 10(-3) and 3.2 x 10(-6) mol L(-1). The accuracy evaluated through recovery studies ranged from approximately 96 to 103% in methanol solutions and in urine, and from approximately 93 to 104% in plasma, with standard deviations <7.5%. The low sensitivity of the method precludes its use for the assay of these antidepressants in biofluids at least at therapeutic doses as the ranges of FLX and FLV plasma levels are 0.15-0.5 microg mL(-1) and 0.15-0.25 microg mL(-1), respectively. The method was applied successfully to the determination of FLX and FLV contents in pharmaceutical samples (brand-named and generic) purchased in several countries or via the Internet. All the commercial formulations contain the active ingredient in the range 94-103% of stated concentration. A "signature" of the formulations (solid and liquid) was obtained with 2D Diffusion-Ordered SpectroscopY (DOSY) (1)H NMR which allowed the characterisation of the active ingredient and excipients present in the formulations studied. Finally, the DOSY separation of FLX and FLV whose molecular weights are very close was obtained by using beta-cyclodextrin as host-guest complexing agent.