This paper describes the implementation of the sound system used to generate aural cues for use in the distributed virtual environment of NPSNET. This sound system is called the NPSNET-3D Sound Server. It is a MIDI-based loudspeaker system consisting of commercial, off-the-shelf equipment. This sound system currently generates 2-D aural cues via four speakers, but is designed for the potential generation of 3-D aural cues via a cube con®guration of eight speakers.