Air infiltration represents an important part of the heating and cooling load of homes and is an important parameter in indoor-outdoor air pollution relationships. However, air infiltration rate is difficult to quantify, because it is not only a function of building tightness and configuration but also of inside-outside temperature differences, wind velocity, direction, and possibly other factors. There are formulas to estimate air exchange rates, but they are at best rough approximations.The tracer dilution method has been used for a number of years to measure air infiltration rates. Helium was one of the first tracer gases used, but others have also been used, such as ethane and nitrogen dioxide. Since Gregory's observation that sulfur hexafluoride (SF6) could be detected in nanogram quantities by means of an electron capture detector, it has also been used as a meteorological tracer. More recently it has been applied to the prediction of smoke movement in fires and to the measurement of air infiltration in buildings. In addition to the fact that it can be measured in trace concentrations, it is not a normal background constituent of air and is odorless and nontoxic in the concentration used. For determination of its toxicity and as part of the quality control process, mice are subjected to 80% SF6/ 20% oxygen mixtures for 24 hours. For the batch of gas to be certified as nontoxic, the mice must not show signs of irritation or distress.Heating energy measurements were recently performed on an instrumented townhouse located in an environmental chamber for the purpose of validating a computer program. The temperature in the chamber, as well as the air temperature within the house was controlled and measured. This study also provided an opportunity to measure air infiltration rates under various inside-outside temperature differences and under conditions where wind velocity was negligible, thus eliminating o:ne of the major variables. This paper describes the results of such measurements using SF6as a tracer. It attempts to determine the effect on air exchange rates of such factors as inside-outside temperature difference and the sealing of doors and ducts. It also considers the repeatability and accuracy of the SF6method, using different sampling techniques and reviews some of the possible sources of error.