During the spring of 2003, the NOAA Storm Prediction Center and the National Severe Storms Laboratory evaulated different short-range ensemble (SREF) methodologies, in an effort to understand the use of SREF output for forecasting severe convection. From mid-April through early June, forecaster teams analyzed a 15-member ensemble produced by NCEP, and an experimental 32-member MM5 ensemble, produced through forecaster generated perturbations (via the MM5 adjoint) and computed using a supercomputer at the Oklahoma Supercomputer Center for Education and Research (OSCER). The data generated by both ensembles were evaluated using a myriad of statistical post-processing techniques and graphical displays, and were used to generate an experimental Day-2 severe weather outlook for the United States. This paper will describe the experiment design, the SREF analysis methods that were used, overall results from the experiment, and thoughts on future use of SREF data for forecasting severe convection.