Shaking as a mechanism of severe brain injury in infants has been challenged on a theoretical basis as insufficient to explain the magnitude of observed injuries. Computed tomography and magnetic resonance imaging, developed since shaken baby syndrome was first described, are helpful in establishing whether external trauma occurred for infants thought to have been shaken. We compiled data from 24 infants initially diagnosed as having shaken baby syndrome, including physical examination, roentgenograms, computed tomography or magnetic resonance imaging, and autopsy (when applicable). Half of the patients showed no evidence of direct impact, and evidence of external trauma was not predictive of a fatal outcome. These findings indicate that shaking by itself is sufficient to cause severe or fatal intracranial injury and that the shaken baby syndrome reflects a spectrum of such child abuse injuries that may include direct trauma or only shaking.