Alcohol involvement of drivers or non-occupants is the dominant risk factor for serious highway crashes. Despite this, crash cost studies have not carefully analyzed the costs of crashes involving alcohol. This paper focuses on alcohol-involved crashes. It examines how costs vary with blood alcohol concentration (BAC). To compute national costs, it adjusts the national incidence estimates derived from police reports to incorporate data from the medical care system about underreporting. It also examines the trend in the incidence of alcohol-involved crashes. Methodology involved in the study included estimating injury and property damage only crash incidence; estimating the fraction of incidents by severity involving alcohol; and costing the incidents. The incidence of alcohol-involved highway crashes (ones in which a driver or non-occupant had been drinking) was estimated from Federal databases. The estimates were adjusted for police underreporting of alcohol involvement. Results reveal that in 1990, 22 percent of motor vehicle crash victims--1.2 million--were injured in crashes involving alcohol. Over 22,000 of these victims were killed. The comprehensive cost of alcohol-involved crashes was $148 billion in 1990, including $46 billion in monetary costs and $102 billion in lost quality of life. This represents $1.09 per drink of alcohol consumed. Crashes in which BAC exceeded .10 percent accounted for 32 percent of comprehensive crash costs, and crashes with lower positive BAC accounted for another 8 percent. Excluding drunk drivers and drunk non-occupants, alcohol-involved crashes caused 8,500 deaths, and left 21,000 people permanently disabled and another 605,000 less seriously injured. Averaged across all drinks, other people collectively pay $0.63 in crash costs every time someone takes a drink. A combination of increased public awareness and strong legal sanctions has been effective in reducing the incidence of alcohol-involved driving. The proportion of injuries in crashes that police reported were alcohol-involved dropped by 37 percent between 1982-1984 and 1990. The authors note that although participants in these crashes had used alcohol, this does not necessarily mean that their alcohol use was the cause of the crash. Further research is needed to determine the role that alcohol played in these crashes. References, tables