Similar to processing language and scenes, we are highly specialized in face processing. We found that viewing a person in a scene automatically leads to an increase in spontaneous gaze following (Zwickel & Võ, 2010). This relies on the intentionality attributed to the agent (Võ & Zwickel, in prep). A great bulk of face perception studies has used static images of faces. In a recent study (Võ, Smith, Mital, & Henderson, 2012), we monitored participants' eye movements while they watched pedestrians being interviewed. Contrary to previous findings using static displays, we observed no general preference to fixate eyes. Instead, gaze was dynamically directed to the eyes, nose, or mouth depending on the currently depicted social event, e.g. talking or making eye contact.
Võ, M. L.-H., Smith, T. J., Mital, P. K., & Henderson, J. M. (2012). Do the Eyes Really Have it? Dynamic Allocation of Attention when Viewing Moving Faces. Journal of Vision, 12(13):3, 1–14. pdf
Zwickel, J., & Võ, M. L.-H. (2010). How the Presence of Humans Biases Gaze Movements. Psychonomic Bulletin & Review, 17 (2), 257-262. pdf