In 1906 Arthur Newsholme linked artificial feeding and fatal diarrhea in infants aged one year and younger on the basis of two independent sources of information: mortality registration and a three-year (1903-1905) census of infants from Brighton, United Kingdom. Artificial feeding was more common in the infants who had died (89.3%) than in those in the survey (22.3%). However, boldly assuming the two data sources were nested, Newsholme computed the risks of fatal diarrhea: these were 48 times greater for infants fed fresh cow's milk and 94 times greater for those fed condensed milk than for infants who were exclusively breastfed. This mode of computing risks and risk ratios before the invention of the cohort study design was more innovative than was the usual investigation techniques of his contemporary epidemiologists. Newsholme's conclusions were consistent with the current knowledge that breastfeeding protects against fatal diarrhea.