The National Highway Traffic Safety Administration web site reports that rear-end collisions in the United States exceed 1.5 million per year, or approximately 23% of all vehicle crashes. Car following behavior and the decision-making habits of drivers seem fundamental to understanding how to avoid these rear-end crashes. The present research aimed to reveal factors that govern car following under conditions of reduced visibility. It employed a KQ-Vection high-fidelity driving simulator to measure the behavior of automobile drivers following a lead vehicle at 13.4 m/s (30 MPH) or 22.4 m/s (50 MPH) under three visibility conditions--clear or one of two densities of simulated fog. At the higher speed, fog conditions separated participants into a group that stayed within visible range of the lead car, even though the headway time violated the NHTSA recommendations for the speed involved, and another group that lagged beyond the visible range. Data were compared to the model of Van Winsum for car following (The human element in car following models. Transportation Research Part F 2, 1999). Contrast and image size measurements allowed comparison to a standard contrast sensitivity function and allowed estimation of the JND term in the Van Winsum model.