Historically, deterrence has been the principal approach for prevention of driving under the influence of alcohol. Legislatures have established blood alcohol concentration (BAC(1)) limits for driving and law enforcement agencies have enforced those laws. Scientists have contributed to the establishment of BAC limits with data from experimental and epidemiological studies to identify the alcohol levels which produce driving skills impairment and increased crash rates.

In the scientific literature, impairment refers to a statistically significant decrease in performance under alcohol treatment from the performance level exhibited under placebo treatment. To reach statistical significance, performance differences in subjects under the two treatments must be reliable and substantial in magnitude.

In a report published by the National Highway Traffic Safety Administration (NHTSA), Moskowitz and Robinson (1988) reviewed the experimental literature from the 1950's through 1985 on alcohol effects on driving related skills performance. The report summarized 177 studies which met the following criteria for inclusion: Placebo treatments, statistical significance, and the ability to determine the BAC at the time of testing. Overall, 21% of the studies reported performance impairment by 0.04 g/dl, 34% by 0.05 g/dl, 66% by 0.08 g/dl and nearly all by 0.10 g/dl. The skills performance measures were organized into 10 response categories. The BAC at which impairment was first measured and the percent of studies reporting impairment at the various BACs differed by response area. Divided attention, visual functions, and tracking were impaired in the 0.01-0.02 g/dl range, with the number of studies reporting impairment increasing rapidly with increasing BACs. On the other hand, the impairment of simple reaction time and psychomotor measures began at higher BACs. Simple reaction time in particular was found to be an insensitive and unreliable measure.

Moskowitz and Robinson criticized studies which examined performance at one BAC only since a report of impairment at a single BAC sheds no light on the question of whether lower BACs might also be impairing. They concluded that the data identified no threshold BAC below which impairment does not occur. They also concluded that the scientific evidence supported a reduction of the BAC limit for driving to 0.05 g/dl. Finally, to facilitate the classification of examined behaviors, they urged investigators to include fuller descriptions of methods and procedures in reporting future research over a wider range of BACs.

Subsequently Kruger, et al., (1990) in Germany, performed a literature review capturing the European non-English language literature (English summary in Kruger, 1993). The Kruger, et al., review encompassed studies of alcohol effects on subjective reactions such as mood, and social relations, as well as on skills performance. The studies were organized into categories including: subjective reactions, social behaviors, psycho-physical functions, automatic behaviors, controlled behaviors, and driving. The terms automatic and controlled are taken from cognitive psychology theory. Automatic behaviors refer to over-learned tasks which require little conscious mental activity (easy tracking, simple and choice reaction time, mental arithmetic, concentrated attention, etc.). Controlled behaviors involve a greater mental workload (difficult tracking, divided attention tasks, information processing, etc.). Studies of controlled behaviors report impairment beginning at 0.03 g/dl. Impairment appears at somewhat higher BACs for automatic behaviors. The authors concluded that social and controlled behaviors are impaired at 0.030-0.049 g/dl in actual traffic whereas automatic behaviors are not impaired before 0.05 g/dl.

Holloway (1994, 1995) used Kruger's schema in a review for the period 1985 - 1993. In agreement with Kruger, he found that subjective effects and controlled behaviors are affected by alcohol at lower BACs than psycho-physical or automatic behaviors. He reported that 70 to 80% of studies examining controlled behavior and subjective intoxication reported impairment by BACs at or below 0.04 g/dl.

Determining whether a behavior is automatic or controlled, however, can be difficult. Shinar, et al., (1998), for example, demonstrated that the frequent description of gear shifting in a manual transmission vehicle, as an "automatic" process was erroneous. Subjects showed decreased detection of roadside signs during manual shifting in comparison with automatic transmission cars. This is not to argue the potential value of the cognitive psychology model of driving behavior, but as Ranney's (1994) review of driving behavior models noted, accurate assignment of driving activities as controlled or automatic awaits further research. Moskowitz and Robinson (1988) noted the difficulty of assigning studies to behavioral areas given that reports of experimental methods and procedures often are quite limited. As an example, driving simulator studies vary greatly in the types of roads traversed, the degree of interacting traffic, the length of travel, and mental workload. Such variability, which is obvious in simulators, also exists in other measured behaviors. To expand the boundaries of response categories into theoretical groupings without empirical studies validating that placement would only inflate the problem. A fully satisfactory resolution awaits a better taxonomy of behavior which, coupled with better specification and execution of experimental studies, will permit organization of diverse studies into theoretical formats.

This review examines the 1981 to 1998 literature on driving related behavior primarily under low BACs. The behavioral response categories have been organized in a form slightly different from that used by Moskowitz and Robinson (1988). Two caveats apply to the 13 categories of driving-related behaviors in the categorization scheme used in this review. First, note that assignment of tasks to the behavioral categories is arbitrary to some degree and, in some cases, is different from the investigator's categorization. For example, results from an experiment which reported that subjects were required to perform two tracking tasks simultaneously appear under divided attention rather than under tracking. Second, even within each of the 13 behavioral categories, there is great variability in experimental tasks and corresponding demands on the subjects. An attempt to incorporate the studies into larger categories of a theoretical schema would serve to increase within-category variability and would result in a blurring of distinctions between categories.

1. Alcohol in blood is measured in terms of weight per volume. In the US, BACs are typically reported in grams of absolute alcohol per deciliter of whole blood (0.08 g/dl). The symbol "%" is frequently used to denote g/dl. Note, however, that the "%" symbol is not a true percentage since it is describing a measure of weight in a measure of volume.

In this paper, g/dl is the measure utilized. Th values remain the same if the symbol "%" is substituted to g/dl. Other countries frequently use mg/dl (e.g., 80 mg/dl) or g/l (e.g., 0.8 g/l) as their units of measure.