Abstract: The method shown in the title was proposed based on the fact of high false positive rate of diphenylamine residue in milk powder by the national standard method of GB/T 23210-2008. Tests for confirmation with gas chromatography-tandem mass spectrometry (GC-MS/MS), recovery and matrix effect were made, to compare purification effects of C₁₈ cartridge, Florisil cartridge and Carb/NH₂ cartridge on the actual samples. Milk powder sample (3 g) was homogenized with 20 mL of acetonitrile and 5 g of sodium chloride for 2 min, and centrifuged for 5 min, and the extraction was repeated once again. The supernatants obtained were combined to a heart-shaped pendant (50 mL), and evaporated by rotation to 1 mL at 40 ℃. The above solution was passed through the Carb/NH₂ cartridge activated beforehand, and the effluent was collected. Acetonitrile-toluene mixed solution (15 mL) at volume ratio of 3∶1 was used to wash the heart-shaped pendant, and the rinsing solution was loaded on the Carb/NH₂ cartridge. Acetonitrile-toluene mixed solution (20 mL) at volume ratio of 3∶1 was taken, and passed through the Carb/NH₂ cartridge for elution. Throughout the purification process, the effluence flow was controlled at 1 mL·min^-1. The collected eluate was combined, evaporated by rotation to nearly dry at 40 ℃, and dried by nitrogen. The residue was dissolved with 1 mL of ethyl acetate, and the solution was analyzed by gas chromatography-mass spectrometry. It was shown that the positive rate of diphenylamine was 100% when the 10 commercial milk powder samples were purified with C₁₈ cartridge, and the corresponding substances at m/z 169, 167, 168 (quantitative and qualitative ions of diphenylamine) were interfering substances in the sample solution which was confirmed by MS/MS, and it was concluded that the results provided by the national standard method were false positive. When purified with Florisil cartridge, many interferents in the sample solution were detected, with high recovery and strong matrix effect. When purified with Carb/NH₂ cartridge, the purification effect was better, and matrix matching method was not needed for quantification. Linear relationship between values of mass concentration and peak area of diphenylamine was kept in the range of 0.01-1.00 mg·L^-1, with detection limit of 3.4 μg·kg^-1. Test for the spiked recovery was made at the 3 concentration levels with negative milk powder sample as matrix, giving recoveries in the range of 84.6%-93.1%, and RSDs (n=6) of the determined values ranged from 4.2% to 7.5%. This method was applied to the analysis of the 10 milk powder samples described above, and the residues of diphenylamine were not detected.