The NCEP Regional Spectral Model (RSM) is used to generate ensemble forecasts over the Southwest United States during the 151 days of 1 November 2002 to 31 March 2003. RSM forecasts to 24 h on a 12 km grid are produced from 0000 and 1200 UTC initial conditions. Eleven ensemble members are run each forecast cycle from the NCEP GFS ensemble analyses (one control and 5 pairs of bred modes) and forecast lateral boundary conditions. The model domain covers two NOAA River Forecast Centers: the California Nevada River Forecast Center (CNRFC) and the Colorado Basin River Forecast Center (CBRFC). Ensemble performance is evaluated for probabilistic forecasts of 6 h and 24 h accumulated precipitation in terms of Brier skill scores relative to sample frequencies, reliability diagrams, Relative Operating Characteristic curves, potential economic value curves, and ranked histograms. Uncertainties in several NCEP precipitation analyses are assessed, and the NCEP Stage IV analyses are selected to represent truth for most verification metrics. Forecast quality and value depend highly on the verification dataset, geographic region, and precipitation threshold. In general, the RSM ensemble forecasts are skillful for 24-h accumulations over the CNRFC region for thresholds between 1-75 mm, but lack skill over the CBRFC region. The model exhibits a wet bias for all thresholds that is relatively stronger over Nevada and the CBRFC than over California. Deficiencies in the RSM cumulus parameterization scheme, insufficient representation of topography at 12 km and large representiveness errors in the verifying analyses are hypothesized to be important contributing factors to the bias. Removal of such bias over the interior Southwest will pose challenges to modeling community in view of the uncertainties inherent in the verifying analyses.