Objective coding of the arrest reports revealed a fairly cohesive pattern of evidence relating blood-alcohol readings to driving behavior before the stop, general behavior after the stop, and performance on the field sobriety tests. There were several differences in detection and arrest procedures among the three metropolitan areas, and there was scattered evidence that the configuration of BAC limits and per se laws of the States influenced the weighting of blood-alcohol evidence in determining case outcomes. In general, case dispositions reflected a rational pattern of decisionmaking, in which drunk-driving convictions were systematically influenced by the quantity and quality of evidence for guilt. 6 notes, 22 references. (Author abstract)