It is well known to the public in Lahontan Valley in rural Nevada, USA, that local aquifers produce water with varied, but sometimes very high concentrations of arsenic (> 4??ppm). As a result, many residents of the area have installed household reverse-osmosis (RO) systems to produce drinking water. We examined performance of RO systems and factors associated with arsenic removal efficiency in 59 households in Lahontan Valley. The sampling results indicated that RO systems removed an average of 80.2% of arsenic from well water. In 18 of the 59 households, arsenic concentrations exceeded 10??ppb in treated water, with a maximum in treated water of 180??ppb. In 3 of the 59 households, RO treatment had little effect on specific conductance, indicating that the RO system was not working properly. Two main factors lead to arsenic levels in treated water exceeding drinking-water standards in the study area. First, arsenic concentrations were high enough in some Lahontan Valley wells that arsenic levels exceeded 10??ppb even though RO treatment removed more than 95% of the arsenic. Second, trivalent As+ 3 was the dominant arsenic species in approximately 15% of the wells, which significantly reduced treatment efficiency. Measurements of specific conductance indicated that efficiency in reducing arsenic levels did not always correlate with reductions in total dissolved solids. As a consequence, improvements in taste of the water or simple measurements of specific conductance made by technicians to test RO systems can mislead the public into assuming the water meets safety standards. Actual measurements of treated water are necessary to assure that household RO systems are reducing arsenic concentrations to safe levels, particularly in areas where groundwater has high arsenic concentrations or where As+ 3 is the dominant species. ?? 2007 Elsevier B.V. All rights reserved.