Ester types (2, 4-dichlorophenoxyacetic acid (2, 4-D) ethyl, 2, 4-D butyl and MCPA-thioethyl) and free acid types (2, 4-D, 2, 4, 5-trichlorophenoxyacetic acid (2, 4, 5-T), 2-methyl-4-chlorophenoxyacetic acid (MCPA), m-chloroperbenzoic acid (MCPB), mecoprop and dichlorprop) of phenoxy acid herbicides in agricultural products were simultaneously extracted from samples with acetone for vegetables/fruits, and with acetonitrile for cereals/beans, and then re-extracted with ethyl acetate. Three ester types of herbicides extracted from the samples were cleaned up by Sep-Pak^[○!R] Florisil, and then determined by ECD-GC and GC-MS (SIM). Recoveries of the ester types of herbicides determined by GC and GC-MS spiked at 1.25 ppm to asparagus, taro, brown rice and corn were 61.4-76.9% and 50.8-63.9%, respectively. Six free acid types of herbicides were esterified by pentafluorobenzyl bromide or diazomethane and determined by GC-MS (SIM) without cleaning up by solid phase extraction. Recoveries of the free acid types of herbicides esterified by pentafluorobenzylation and methylation spiked at 1.25 ppm to the samples were 68.5-98.5% and 53.7-88.1%, respectively. It was clarified that the concentration of pentafluorobenzyl bromide in acetone should be 5-10% to have sufficient recoveries for the esterification. Methylation was superior to pentafluorobenzylation for the simplicity of manipulation, but was inferior for the sensitivity of GC and GC-MS.