Passive capillary samplers (PCAPS), which sample water from the vadose zone via a hanging water column in a fiberglass wick, have shown potential to provide superior estimates of soil water flux compared to alternative methods. The objectives of this study were to evaluate the performance of PCAPS under natural rain-fed conditions concerning (1) their operational characteristics and (2) their ability to estimate soil water flux. Forty-two PCAPS were installed in 21 commercial agricultural fields in Lane County, Oregon. Monthly measurements of soil water flux and precipitation were recorded at each site for the 4-year project duration. Of the 42 installed PCAPS, 12 samples at six sites were inoperable or did not operate efficiently: 10 samplers were consistently below the water table, which overflowed the collection vessels rendering the samplers inoperable. Only two of the PCAPS exhibited technical failure resulting in unusually low collection efficiencies, thought to be due to a collapse of the collection vessel from oversuction during sample retrieval. On average, the 30 remaining PCAPS measured soil water flux 25% greater than that obtained from a water balance estimate. This discrepancy represents ~8% of the total annual precipitation and irrigation each site received. PCAPS collection efficiency was found to be significantly correlated (R2=0.75) to the water balance yearly estimated recharge. The difference between PCAPS measured and water balance estimated percolation could be the result of inaccuracy in water balance evapotranspiration estimates and/or oversampling in the presence of high water tables. To estimate the mean yearly recharge at each site with a 30% bound on the mean at the 0.05 confidence level, eight PCAPS are required. This number corresponds closely to the results of Brandi-Dohrn et al. <1996a> and is thought to be due to intrinsic variability of percolation. ¿ 2000 American Geophysical Union