A criterion for brain injury threshold is proposed which, for the first time, endeavours to take into consideration the combined effects of both translational and rotational kinematics. The validity of the model is assessed by way of a review of all known head injury data bases in which translational and rotational accelerations have been monitored. Available data appears to support no more than a simple linear proportioning of the two types of motion though a squared weighting, as originally proposed (and which intuitively is more plausible) also seems valid.