CONCLUSIONS Analyses have been carried out to produce an estimate of the frequency of fires on ships of capacity greater than 500 grt of all types which would have been suitable to carry Type B(U) or Type B(M) packages. Data for fire frequency were selected on the basis of qualitatively defined "severe fires", or "qualifying incidents" of fires/explosions, having the potential to threaten package integrity and thus frequency data derived in these studies are very much upper-bound estimates of the frequency with which some loss of package integrity might occur and should not be read as frequencies with which loss of integrity would occur.