The pneumatic impact test has long been used to screen nonmetallic materials for oxygen service. This method is also used to rank batches or lots of particular materials. A study has been made to evaluate the sources of variability in the test and to determine the mechanism of ignition. The results of the study show that much of the variability in test data observed in the past can be eliminated by better control of system cleanliness, tightening the specifications for the opening time of the high-speed valve, and conducting tests over a wide range of pressures and using the logistic statistical model to analyze the data. Though the mechanism of ignition was not determined conclusively, it is probably heating of the oxygen initially in the test chamber by adiabatic compression; heating of gases trapped in void spaces within the material could also be important. Further research is recommended to determine if batch or lot control has any value and if specimen preparation and conditioning significantly affect the test results.