The signal used in the original acoustic rhinometers was an impulse of short duration and wide energy spectrum. A rhinometer utilizing a new signal of a similarly broad energy but different time course has been developed. We investigated the accuracy of area reconstructions computed using this signal by comparing them with the original areas of simple models. This study demonstrates that, at present, acoustic rhinometry using a pulse train signal is subject to significant systematic errors. Some of these are intrinsic to the rhinometer. Others are caused by the geometry of the models. Areas reconstructions beyond narrow constrictions are particularly inaccurate. The findings are significant because the nasal valve can act as such a constriction. The errors can cause the parameters used in the clinical application of acoustic rhinometry to deviate substantially from the true values.