Currently, angular acceleration is believed to be more damaging to the brain than linear acceleration, even though both are present in any head impact. In a recent experiment, it was found that a helmeted head sustained the same degree of angular acceleration as the unhelmeted head for the same impact, but its linear acceleration was decreased significantly. So, if angular acceleration is the cause of brain injury, then how is the brain protected by the helmet? This paper proposes a new hypothesis of brain injury and suggests that input acceleration limits should be replaced by response variables.