Abstract
The technique which University College, London, is using to investigate accidents is described. It is a compromise between statistical methods which loss detail and in-depth investigations which do not allow general conclusions. It involves classifying accidents and type. Classification is also carried out according to severity of injury.