Human Factors and Highway-accident Causation: Some Theoretical Considerations

Traditional correlational analyses of human skills and highway accidents have not been ver) productive. Some of the likely reasons for this are discussed. An alternative approach to accident-cautation research is outlined in which the importance of a skill with good face validity to driving is assessed in terms of its sensitivity to frequently occurring transient human states such as fatigue, stress and alcohol intoxication. It is almost an axiom in highway safety research that traffic accidents are caused primarily by human errors, as opposed to vehicular malfunctions or adverse environmental conditions. Indeed, a recent in-depth accident investigation [Treat et al., 19771 cited human factors as definite or probable causes in 93% of the 420 accidents under investigation. Thus, it is not surprising that a substantial amount of research has been devoted to the investigation of the correlations between accidents and performance of a variety of skills, in efforts to identify skills critical to driving. However, results of such studies (e.g. the visual-skills study of Henderson and Burg [I9741 and a variety of studies reviewed by Goldstein [I%411 generally yield disappointingly low correlations. Such findings indicate that the performance of each of a variety of skills can account for only a fraction of the variance exhibited in the accident record. This paper presents arguments that: (1) indicate the likely reasons why none of the logically important skills have been shown to account for a substantial portion of the variance of the accident data, and (2) suggest an alternative approach in accident-causation research. There are two principal problems in the approach of correlating accident records with human performance measures. First, as Goldstein has pointed out, I’. . . accident involvement is not a highly stable characteristic of individual drivers [Goldstein, 1963, p. 21,” since there is only a low correlation between accidents in one period and accidents in another period. Therefore, “. . . accident involvement of individuals is not highly predictable from measurements of [relatively] stable characteristics [Goldstein, 1963, p. 21,” such as, for example, visual skills. Second, current correlational studies attempt to relate events occurring in two different settings: The accidents occur in the “real-life” domain, while measures of human performance are usually obtained in the laboratory. The two settings probably differ considerably in the sets of transient human states present (e.g. fatigue, stress, alcohol intoxication). Moreover, it is well known that such transient human states affect a range of human skills, including some with good face validity to driving. For example, fatigue from prolonged (actual and simulated) driving has been shown to affect perception(Brown, 19671, decision-making[Brown. Tickner and Simmonds, 19701, and psychomotor performance[Dureman and Boden. 19721. Therefore, it is not known how well, for example, laboratory-measured dynamic visual acuity replicates actual “on-the-road” dynamic visual acuity, since the “on-the-road” acuity may differ considerably from that measured in the laboratory. These arguments suggest that even if, for example, the actual “on-the-road” dynamic visual acuity were critical to safe driving, correlational analyses of laboratory-measured dynamic visual acuity and accident involvement would not necessarily yield high correlations. Thus. not only have past studies not found high correlations, but future studies are also unlikely to do so: These studies are unlikely to identify the critical skills, which when deficient, cause accidents. What then are potentially more fruitful areas of accident-prevention research? The most direct implication of the foregoing is that any correlationdl study should obtain the performance measures in a real-life setting. Optimally, such measures would be gathered by stopping motorists and testing them at the side of the road by means of a portable tester, No claim is AA,’ Vol 13. No 2-A 61