It is remarkable that drivers (on average) can safely navigate through dense traffic at high speeds—conditions in which the time headways between vehicles are in the same order of magnitude as human reaction times. One explanation for this is the ability of drivers to anticipate on the traffic conditions in their surroundings. In this paper, we study, through simulation, the effects of reaction times, errors in perception and anticipation on the probability of accidents on freeways. To this end we extend an existing model for car following and lane changing with a perception and anticipation model inspired by Ensley’s three levels of situational awareness (perception, understanding and projection). By systematically varying driving behavior with different reaction times over a range of perception errors, and anticipation strategies, we compute efficiency effects (capacity and total time spent) and safety effects (the probability density of accidents happening as a function of these different contributing factors and errors). The results provide some evidence that safe driving is robust with respect to perception errors under simple anticipation strategies and small reaction times. When reaction times grow larger, more advanced anticipation strategies are needed to guarantee safe driving.