This technique has produced convincing information with practical consequences in many areas, such as proving that followup calls might not improve survey response rates and that arrested persons with community ties may be safely released before trial without bail. However, an effective experimental approach using comparison groups contains certain essential elements. First, there must be two or more alternate ways of proceeding, and these procedures must be definitively described. The classes of subjects to whom the treatments are applicable also must be clearly defined, and the relevant outcomes and fair measurement methods agreed upon in advance. The alternative procedures should be applied simultaneously, with experimental and control groups determined by mechanically random choice. Using these principles as guidelines, the Secret Service could assess its tasks and procedures. The first experiment should focus on a small problem with a high likelihood of success because continuous vigilance is necessary to prevent any deviations from the essentials of controlled experimentation.