The purpose of this study was to increase our understanding of the head impact level that will produce concussion in humans. The technique employed was that of accident restaging.The investigation reported here was composed of three parts:1. The Cornell accident records were reexamined to establish the frequency of brain concussion as a function of windshield damage.2. Tests were conducted with instrumented cadavers to determine the head accelerations achieved when the appropriate windshield damage levels were obtained.3. Head injury indexes were calculated from the measured accelerations, and their predictions were compared to the Cornell field data.The present reexamination of the Cornell accident data found that the percentage of victims who received a concussion involving known unconsciousness reduces to, at most, 11% for the case of radial crack with bulge. The percentage obtained for radial crack-no bulge was, at most, 2.8%. Thus, the former condition is approximately four times as likely to produce a concussion as the latter condition.A number of head injury indexes were calculated from the cadaver head accelerations obtained in the current study. It would be expected that these should be somewhat greater for the bulged windshield condition than for the nonbulged condition. This was not found to be the case. The calculated criteria were either approximately equal or else greater for the no-bulge condition than the bulged condition. The reason for this disparity is not known.Most injury indexes treated the accelerations produced by the cadaver/windshield impacts in a fundamentally different way than accelerations produced by equivalent dummy impacts. For the cadaver data, the indexes focused on the spike caused by the glass breakage, whereas for dummy data, the indexes focused on the broad pulse resulting from the interlayer bulging. This difference arose from the fact that the cadavers produced lower accelerations during the bulge event than did dummies.