Although the study was not able to include the important component of breath sampling from an intoxicated subject, it considered other important elements such as instrument accuracy and precision, potential interfering substance bias, components of variance, data analysis, and interpretation of results. Future efforts can build on these initial efforts by incorporating additional instrumental and program features. These might include error detection mechanisms, instrument performance in a normal field environment, instrument performance over a longer time interval, and instrument compliance with local quality control standards. Eleven jurisdictions participated in the study, which evaluated 27 modern instruments. Five wet-bath simulator solutions with ethanol vapor concentrations that ranged from 0.0254 to 0.2659 g/210 L were sent to participating programs, with the instruction to perform 10 measurements on each solution with the same instrument. Four of the solutions contained ethanol only, and one contained ethanol mixed with acetone. The systematic errors for all instruments ranged from -11.3 percent to +11.4 percent; coefficient of variation ranged from 0 to 6.1 percent. A components-of-variance analysis revealed at least 79 percent of the total variance as being due to the between-instrument component for all concentrations. 2 tables, 4 figures, and 13 references