Abstract
Some of the very serious problems associated with the use of anthropomorphic dummies in crash testing are discussed, which arise primarily from the non-repeatable behaviour of the dummy and from the unknown correlation between it s behaviour and that of the human.