Ten police officers administered the tests in the laboratory to 297 drinking volunteers with blood alcohol concentrations (BACs) ranging from 0 to 0.18%. The officers were able to classify 81% of these volunteers, on the basis of their test scores, with respect to whether their BACs of people they tested differed by 0.03% on the average from the actual BAC. Interrater and test-retest reliabilities for the test battery ranged from 0.60 to 0.80. In a limited field evaluation police officers filled out 3128 data forms, each represented a driver stopped during a three month period. Police officers, after training on the administration and scoring procedures for the test battery, tended to increase their arrest rates and appeared to be more effective in estimating BACs of stopped drivers than they were before training. Anonymous breath testing of released drivers who were stopped indicated that many of the drinking drivers were never given a sobriety test. (NTIS abstract)