Crowd simulation models are currently lacking a commonly accepted validation method. In this paper, we propose level of presence achieved by a human in a virtual environment (VE) as a metric for virtual crowd behavior. Using experimental evidence from the presence literature and the results of a pilot experiment that we ran, we explore the egocentric features that a crowd simulation model should have in order to achieve high levels of presence and thus be used as a framework for validation of simulated crowd behavior. We implemented four crowd models for our pilot experiment: social forces, rule based, cellular automata and HiDAC. Participants interacted with the crowd members of each model in an immersive virtual environment for the purpose of studying presence in virtual crowds, with the goal of establishing the basis for a future validation method.