Driving simulators are routinely used by researchers to study driver behavior. However, (limited) previous research has shown that the user gender may modify the perception, or even usability of these devices. For example, [Mourant 00] shows that men and women have different sensitivity to simulator sickness. These differences are liable to introduce biases in behavior studies and other virtual reality applications such as driver training. The validity of driving simulators depends on a number of factors, such as platform dynamics or 3D-database realism. In our research, we focus on simulated traffic behavior. Although there is a huge literature on traffic simulation, its main purpose has been to study capacity, security or recently environmental impact of traffic. Hence, the use of traffic simulation models in virtual reality devices, where the entities interact with users, has not been validated. In this paper, we study how users perceive virtual entities, focusing on the role of gender. An experiment has been developed to evaluate the users perception of the driving simulator and of the simulated drivers, using questionnaires and user behavior data. We use the well-known presence concept - developed by [Witmer 98] - to assess the overall validity, and an ad hoc questionnaire to assess the virtual entities behavior credibility. We propose a methodology including a set of tools and data processing units enabling inter-human and virtual agents behavior comparison. We have found very few significant correlation between driving traces, the perception of the virtual environment and background questionnaires, including gender, driving behavior (see [Reason 90]), and immersion tendency, which seems to indicate that the Archisim driving simulator does not induce gender bias, and that the virtual drivers replicate in the same way men and women behavior. This study calls for further evaluation of the impact of age on the use of driving simulators.