In the field of driving, the use of simulation in virtual environment is based at present on considerations of road safety and usability of systems of assistance to the driving. Driving simulators therefore make it possible to study human behavior under different conditions fixed experimentally. A major advantage of the simulation is that it does not expose the subject to objective risk and that the test conditions are "easily" implemented and manipulated. The use of simulation nevertheless raises the question of the generalization and transposition of the behaviors observed in simulation to the actual driving situation. Indeed, it is necessary to ensure that the simulator accurately measures the object of study without modifying or influencing it. Evaluation of the degree of validity is therefore crucial in any simulator study that seeks to elicit realistic behaviors of drivers. Driving simulation is one of the components of virtual reality, the purpose of which is to allow a person (or several) a sensori-motor and cognitive activity in an artificial world, created digitally. The design of virtual environments, however, has for too long been based on considerations focused on optimizing a high degree of realism. This techno centric approach to immersion and interaction seems outdated and it is widely accepted the need to adopt an anthropocentric approach in order to allow a man placed in a virtual environment to perceive himself as a full-fledged user. The interaction of an individual in the virtual world is a transposition of the loop perception, cognition, action of human behavior in the real world. Immersion in a virtual world can not be the same as that in the real world, the user has learned to act naturally in a real and non-virtual world (without sensory-motor biases), some authors then speak of immersion natural pseudo. In the same way, a user perceives and acts physically with the entities and elements of the virtual world without this always translating into physical sensory motor activity. This fact prevents us from grasping virtual reality as a simple copy of the real world. Thus immersion (in the sense of the sensorimotor contingencies allowed by the simulator) is a necessary but not sufficient condition for the virtual expression of a performance representative of the real driving situation. Faced with this problem, a concept, emerging from the beginnings of virtual reality in the 1980s, seems to address the question of the "ecological" validity of behaviors observed on a simulator: the concept of presence. This multidimensional concept is seen as the ability of individuals to adopt a behavior similar to that of everyday life and therefore their propensity to react to the various stimuli as if they were real. The postulate of our work therefore rests on the idea that presence is a key concept to evaluate the representativeness of driving behaviors observed in simulation. However, several difficulties will have to be resolved. One of the most crucial is that presence is hardly measurable in real time since the very fact of trying to measure it can break the state of presence more or less existing. Thus obtaining behavioral indicators of presence constitutes a real stake to characterize the nature of a phenomenon which at the moment is more fantasized by means of post-experimentation measures than concretely demonstrated by the demonstration of reproducible specific behaviors . 9 The objective of our work is thus to be able to highlight the determinants of the presence in order to develop and calibrate measurement tools capable of evaluating the ecological validity of the behaviors observed in the driving simulators. The aim of this thesis work is to propose a set of concrete and operational recommendations for the development of experimental protocols aiming to optimize the presence in driving simulators