Ambient in situ measurements of the important sulfur gas carbonyl sulfide (OCS) very frequently rely on permeation devices based on weight loss measurements (gravimetry) as the method of calibration. Both time series studies as well as measurements of OCS sources and sinks typically employ this approach. Although this calibration approach is based upon fundamental quantities, its accuracy directly depends upon the purity of the permeating effluent. The present study describes a systematic investigation of four permeation wafers and their time history based on gravimetry, direct absorption spectroscopy employing a tunable diode laser absorption spectrometer (TDLAS), and two different ratio approaches. Gravimetric determinations for two of the devices, which employed fluorinated ethylene propylene (FEP) Teflon wafers as the permeating medium, indicated mass loss rates which decayed exponentially with time. This is in contrast to the direct absorption and ratio measurements, which were significantly lower at the start (37% in one case and 19% in the other) and displayed a small linear drop with time. We present evidence for the presence of a CO2 impurity in FEP permeation devices, either from the starting material or the breakdown of OCS. This impurity copermeates along with OCS but at a much faster rate, giving rise to an apparent exponential decay in the OCS rate. Calculations based on fundamental permeability parameters for CO2 in FEP Teflon and typical levels of CO2 impurity (0.1 to 2%) in the starting material yield results that are consistent with the present interpretation. Two other permeation wafers, which employed polytetrafluoroethylene (PTFE) Teflon membranes, exhibited completely different behavior; all the measurement methods indicated similar results and showed a linearly decreasing permeation rate with time. The CO2 impurity problem prevalent in FEP wafers is insignificant in PTFE wafers because the permeability of PTFE is at least 21 times higher. It is possible that the high pressure inside the membrane (POCS=17.7 atm at 40 ¿C) alters the crystalline PTFE structure with time, resulting in a real decrease in permeation rate. Although more studies are needed to further confirm and extend these findings, the effects observed here suggest that ambient OCS measurements could in some circumstances by systematically high by as much as 37%. Such errors can produce erroneous temporal trends and/or inconsistencies between ambient measurements carried out by different laboratories. ¿ 1998 American Geophysical Union