The relationship between size of the infecting dose and severity of the resulting disease has been investigated for salmonella infections by reanalysis of data within epidemics for 32 outbreaks, and comparing data between outbreaks for 68 typhoid epidemics and 49 food-poisoning outbreaks due to salmonellas. Attack rate, incubation period, amount of infected food consumed and type of vehicle are used as proxy measures of infecting dose, while case fatality rates for typhoid and case hospitalization rates for food poisoning salmonellas were used to assess severity. Limitations of the data are discussed. Both unweighted and logit analysis models are used. There is no evidence for a dose-severity relationship for Salmonella typhi, but evidence of a correlation between dose and severity is available from within-epidemic or between-epidemic analysis, or both, for Salmonella typhimurium, S. enteritidis, S. infantis, S. newport, and S. thompson. The presence of such a relationship affects the way in which control interventions should be assessed.