Augmented Reality (AR), the overlay of virtual images onto the real world, is an increasingly popular technique for developing new human-computer interfaces. As human navigation and orientation in different environments depend on both visual and auditory information, sound plays a very important role in AR applications. In this paper we explore users' capability to localize a spatial sound (registered with a virtual object) in an AR environment, under different spatial configurations of the virtual scene. The results not only confirm several previous findings on sound localization, but also point out some important new visual-audio cues which should be taken into consideration for effective localization and orientation in AR environment. Finally, this paper provides tentative guidelines for adding spatial sound to AR environments.