Predicting drowsiness accidents from personal attributes, eye blinks and ongoing driving behaviour

26 participants drove at night for 135 min on a simulated two lane rural road with light traffic and filled out a battery of questionnaires. Six drivers left the road entirely and ten others left the pavement with one or two wheels. Drivers scoring high on an “extraversion-boredom” personality cluster were more likely to depart from the road due to falling asleep. Drivers scoring high on a “disinhibition-honesty” cluster were more likely to cross solid lane markings but did not seem to fall asleep. The best predicting measures for poor driving were the frequency of eye-closures exceeding 1 s and the number of times that time-to-line crossings were below 0.5 s. The participants’ own judgements on susceptibility to drowsiness was a poor predictor. Dissociation of physiological and subjective measures was observed and explained by a two level information processing model.