Currently used criteria based on functions of spinal acceleration obtained from crash test dummies are shown to be invalid indicators of chest injuries in blunt frontal impacts. Cadaver impact data are analyzed; and injury is found to be a statistically significant function of chest deflection, chest depth, and cadaver age at death. Based on the resulting regression equations, injury-limiting chest deflections are recommended for various size test dummies. The recommendations apply only to test dummies that have significant thoracic biofidelity for blunt frontal impact. They are valid for environments which include signigicant blunt frontal impact. Their extension to other environments has not been validated.