Enhancement of road delineation can reduce safety.

Background: New in-vehicle technologies often outpace the scientific support for their value. In lieu of valid and consistent scientific support, common wisdom is used, as in the assumption that enhanced roadway delineation improves driving safety.

Objective: To evaluate the effects of a Visibility Enhancement System that selectively improves lane markers' visibility on driving safety.

Method: A simulation experiment assessed the effects of an in-car lane Visibility Enhancement System (VES) that highlights the edges of the road ahead on driver's behavior and overall safety, under normal and reduced visibility conditions. Thirty drivers drove in a fix-based simulator through a winding rural road, while attempting to avoid un-enhanced and unexpected obstacles that appeared on the driving lane from time to time. The simulated VES highlighted the road edges up to a distance of 90 m with two alternative configurations: two continuous red lines or a series of red crosses. The effects of the two VES configurations on performance were measured during night and fog driving. Performance measures included speed, lane keeping behavior, eye scanning pattern, reaction time (RT) and collisions with the un-enhanced unexpected obstacles. Subjective measures included confidence and stress.

Results: With the VES, drivers were more confident, less stressed, and drove faster, but had almost twice as many collisions with the unexpected obstacles. Also, steering/braking RT to the obstacles was longer with the VES than without it by nearly 44 msec.

Conclusions: The results are consistent with Lebowitz's theory (1977). While the VES enhanced spatial orientation, it fooled the drivers into assuming that the visibility of obstacles on the road was also improved, and thus actually reduced safety.

Practical Applications: When visibility is an issue in nighttime crashes, the site-specific crashes should be investigated, in cases of collision with objects-on-the-road, improved delineation should be ruled out.