A collaborative study was conducted to determine multiple pesticide residues in apple, green bean, and carrot by using supercritical fluid extraction (SFE) and gas chromatography/mass spectrometry (GC/MS). Seventeen laboratories from 7 countries participated in the final study, and a variety of different instruments was used by collaborators. The procedure simply entails 3 steps: (1) mix 1.1 g drying agent (Hydromatrix) per 1 g frozen precomminuted sample, and load 4-5.5 g of this mixture into a 7-10 mL extraction vessel; (2) perform SFE for 20-30 min with a 1-2 mL/min flow rate of carbon dioxide at 0.85 g/mL density (320 atm, 60 degrees C); and (3) inject the extract, which was collected on a solid-phase or in a liquid trap, into the gas chromatograph/mass spectrometer, using either an ion-trap instrument in full-scan mode or a quadrupole-type instrument in selected-ion monitoring mode. The ability of GC/MS to simultaneously quantitate and confirm the identity of the semivolatile analytes at trace concentrations is a strong feature of the approach. The selectivity of SFE and GC/MS avoids the need for post-extraction cleanup steps, and the conversion of the CO2 solvent to a gas after SFE eliminates the solvent evaporation step common in traditional methods. The approach has several advantages, but its main drawback is the lower recoveries for the most polar analytes, such as methamidophos and acephate, and the most nonpolar analytes, such as pyrethroids. Recoveries for most pesticides are >75%, and recoveries of nonpolar analytes are still >50%. The (within-laboratory) repeatability relative standard deviation (RSDr) values of the recoveries are generally <15%. More specifically, the average results from the 9-14 laboratories in the final analysis of 6 blind duplicates at 3 concentrations for each pesticide are as follows: carbofuran in apple (75-500 ng/g), 89% recovery, 7% RSDr, 9% reproducibility relative standard deviation (RSDR); diazinon in apple (60-400 ng/g), 83% recovery, 13% RSDr, 17% RSDR; vinclozolin in apple (6-400 ng/g), 97% recovery, 13% RSDr, 18% RSDR; chlorpyrifos in apple (50-300 ng/g), 105% recovery, 11% RSDr, 13% RSDR; endosulfan sulfate in apple (150-1000 ng/g), 95% recovery, 15% RSDr, 17% RSDR; trifluralin in green bean (30-200 ng/g), 58% recovery, 11% RSDr, 27% RSDR; dacthal in green bean (60-400 ng/g), 88% recovery, 11% RSDr, 17% RSDR; quintozene in green bean (60-400 ng/g), 79% recovery, 13% RSDr, 18% RSDR; chlorpyrifos in green bean (50-300 ng/g), 84% recovery, 11% RSDr, 17% RSDR; p,p'-DDE in green bean (45-300 ng/g), 64% recovery, 14% RSDr, 27% RSDR; atrazine in carrot (75-500 ng/g), 90% recovery, 11% RSDr, 15% RSDR; metalaxyl in carrot (75-500 ng/g), 89% recovery, 8% RSDr, 12% RSDR; parathion-methyl in carrot (75-500 ng/g), 84% recovery, 14% RSDr, 15% RSDR; chlorpyrifos in carrot (50-300 ng/g), 77% recovery, 13% RSDr, 19% RSDR; and bifenthrin in carrot (90-600 ng/g), 63% recovery, 12% RSDr, and 25% RSDR. All analytes except for the nonpolar compounds trifluralin, p,p'-DDE, and bifenthrin gave average Horwitz ratios of <1.0 when AOAC criteria were used. These 3 analytes had high RSDr values but lower RSDR values, which indicated that certain SFE instruments gave consistently lower recoveries for nonpolar compounds. The collaborative study results demonstrate that the method meets the purpose of many monitoring programs for pesticide residue analysis, and the Study Director recommends that it be adopted Official First Action.