The frequency of and risk factors for shaken baby syndrome remain poorly documented for several reasons: the real number of “benign” cases of shaken baby syndrome are not known; information sources used are diverse, there have been changes over time in the definition of this pathology and few population-based epidemiological studies are available. Estimate the frequency of fatal shaken baby syndrome and determine its risk factors through research carried out on fatal cases in three regions of France while comparing them to data from international publications. A retrospective epidemiological study of all cases of fatal shaken baby syndrome affecting infants younger than 1 year of age who were examined by the courts during a 5-year period in a defined geographical area. Shaken baby syndrome cases were compared with infanticide cases for risk factors and a comparison was made of family characteristics with those of the general population. Thirty-seven cases of shaken baby syndrome were recorded (a rate of 2.9 for 100,000 live births). As in other studies, we found a strong predominance of male victims (78%), young age (median age: 4 months) and a high rate of prematurity (22%). Conversely, results on family educational and socioeconomical levels differ from those reported in numerous studies. Parent perpetrators of shaken baby syndrome belong to higher social classes than those of other types of homicide and socially reflect the population they come from. Our study suggests 1) that epidemiological studies on shaken baby syndrome should include both medical and judicial information sources and 2) that primary prevention strategies (especially in maternity wards) should target all social classes.