An intercomparison of four Fourier transform infrared (FTIR) spectrometers, operated side by side by Jet Propulsion Laboratory (JPL), National Center for Atmospheric Research, and National Physical Laboratory groups, using two different spectral fitting algorithms, was conducted at JPL's Table Mountain Facility (TMF) during November 1996. A blind comparison of retrieved vertical column amounts, of preselected trace gases in preselected microwindows (mw), and subsequent reanalysis of the results are described. The species analyzed are N2(3 mw), HF (1 mw), HCl (1 mw), CH4(1 mw), O3(2 mw), N2O(2 mw), HNO3(2 mw), and CO2(1 mw). The column agreements from the blind phase were within 0.5--2%, except that for HNO3, HF, and O3 the disagreement of the results was up to 10%, 5%, and 4%, respectively. It was found that several systematic effects were neglected in the blind phase analysis. Taking these into account in the postanalysis reduced the disagreements to 0.5--1.0% for most cases, and to less than 4%, 3%, and 1% for HNO3, HF, and O3 respectively. It was concluded that zero offsets caused by detector nonlinearity were the main cause of the large errors in HNO3 and other gases (i.e., CO2) retrieved from the HgCdTe spectra. At shorter wavelengths (i.e., HF) we conclude that incomplete modeling of the instrument line shapes (ILS) was the main cause of column differences larger than 1%. ¿ 1999 American Geophysical Union