Driver impairment is defined as a change in driving performance caused by the deleterious effects of alcohol, fatigue, drugs, or sudden illness (Fairclough, 1997a). In 1995, approximately 24% of traffic casualties in the United Kingdom were driving under the influence of an illegal level of alcohol - that is, [greater than] 0.08% blood alcohol content (BAC; Clayton, 1997). A recent survey of UK drivers indicated that driver fatigue was a contributory factor in 9% to 10% of accidents (Maycock, 1995).
The development of transport telematics systems has been proposed as a potential countermeasure to accidents induced by driver impairment. These systems encompass real-time monitoring, diagnosis, and feedback of driving impairment (e.g., Brookhuis, De Waard, & Bekiaris, 1997; Haworth & Vulcan, 1991; Mackie & Wylie, 1990; Wierwille, 1994). The goal of this technology is to provide predictive feedback concerning early symptoms of impaired driving. It is predicted that warning feedback may persuade the driver to break from the journey and, therefore, avoid those safety-critical episodes of impairment that increase the risk of accidents. Driver impairment monitoring technology relies on the sensitivity and validity of sensor apparatus in order to function effectively. The suitability of various sensor technologies may be assessed with reference to a number of measurement criteria (O'Donnell & Eggemeier, 1986): (a) that measures should be sufficiently sensitive to the earliest symptoms of impairment; (b) that measures are diagnostic - that is, capable of discriminating the influence of fatigue from other categories of impairment, such as alcoholic intoxication; and (c) that measures are selective and therefore able to distinguish the impairment "signal" against a dynamic and highly variable "noise" from the driving environment. It is argued that a combination of these criteria should be used to index the global accuracy of an impairment monitoring system.
Psychophysiological measures may be collected on a remote basis (i.e., there is no requirement to attach electrodes to participants); for instance, machine vision apparatus may be used to monitor eyelid activity (Tock & Craw, 1992). Although psychophysiology appears to be sufficiently sensitive, problems of selectivity are anticipated because of the inherently "noisy" environment inside the vehicle cockpit (e.g., fluctuations in temperature and lighting conditions). Behavioral measures are also potentially useful. However, it is postulated that certain symptoms, such as head-nodding when fatigued, occur at a relatively late stage of impairment and therefore may lack sufficient sensitivity. (Evidence for this view was provided by Haworth & Vulcan, 1991.)
The measurement of driving performance represents a valid strategy to index impairment (i.e., impairment is inferred directly on the basis of primary task performance rather than via a proxy measure). However, a number of problems are associated with the use of driving performance as a predictive and diagnostic source of impairment (Fairclough, 1997b). Specifically, primary task measures are deemed to be insufficiently sensitive to the presence of impairment; that is, task performance is protected from the influence of impairment by compensatory strategies (De Waard, 1996; Hockey, 1997). In addition, there is evidence of selectivity problems for performance measures. For example, in a study of simulated driving, Desmond and Matthews (1997) demonstrated that the presence of road curves was sufficient to suppress impaired vehicular control. Studies encompassing driver impairment caused by alcohol and fatigue have revealed a significant degree of overlap, with both showing effects of a reduction in the fidelity of responses to speed changes of a lead vehicle and an increase in the variability of lateral control (De Waard & Brookhuis, 1991). Therefore, it may be argued that driving measures are not sufficiently diagnostic to differentiate one category of impairment from the other. …