An investigation has been carried out of the limiting oxygen index (LOI) test. A review of the literature shows that, although there is abundant information on the test, it is not clear that its results correlate well with those of any other test, or indeed with those of real fires. Theoretical considerations indicate that the test could be improved by using it with bottom ignition rather than with the standard top ignition. A number of materials were tested in the cone calorimeter and in the LOI, and various correlations were attempted. In general, correlations between some of the cone calorimeter properties measured and the inverse of the LO1 made sense. These correlations were not, however, sufficiently sensitive, even when investigating small effects on a single base polymer system, to justify using the LO1 as a proxy for the cone in any way. The LO1 is likely to continue to be used extensively. This work suggests that quality control and, possibly, mechanistic or other flame-retardant additive studies, are its only applications where the results can be justified.