Abstract
This paper shows that procedures such as classification by accident severity will fail in their original purpose if (and since) accident severity is thought to be a continuous variable and not a step function- In the first place, accident severity within a severity class will be different between any two groups white basically have a different severity distribution: secondly, relation of severity on the risk of DF-injuries, say, will be distorted. General formulae are shown, as well as a set of example calculations.