Experimentally-determined flow velocities in the 6 mm upstream from the sampling cone of an inductively coupled plasma mass spectrometer were compared with velocities determined from a computer simulation of the flow and those calculated from a modified hemispherical sink model. The measured values and those from the simulation agreed within experimental error, but differed from the values calculated from the modified hemispherical sink model by as much as 30%. An empirical alternative to the modified hemispherical sink model is presented that allows for accurate calculation of flow properties upstream from the sampling cone under a range of plasma conditions.