Virtual reality experiments with virtual crowds are necessary to study human behavior under panic or stressful situations that cannot be evaluated in the real world (i.e., building evacuation due to fire). In order to carry out those experiments it is necessary to use a crowd simulation model in which a real person is seamlessly immersed and experiences a high sense of presence when interacting with such a crowd. This paper studies several crowd simulation models in order to determine which could best enhance presence for a user within a virtual environment. Egocentric features that affect presence are considered in the evaluation. Once we have a realistic simulation, we could use it to study human behavior and obtain relevant data. That data could then be used to update agent behaviors in the simulation system to further improve the overall realism of large groups of autonomous agents.