In this paper, four models for binaural perception based on the Head Related Transfer Function (HTRF)developed in MIT [Martin and Gardner, 1994] are presented and compared: linear prediction code (LPC)model of the HRfF (Kayand Marple, 1981) delay of the spherical head model, delay of the HRfF and delay of the HRfF plus magnitude equalization. Spatial hearing files from a monophonic source are generated and evaluated from the point of view of naturalness of the final sound, perception of path and followup as well as complexity for implementation. Results showed that it is possible to get lateral perception with simple spatial hearing models but no one is reliable enough for simulation of elevation.