Avoidable mortality is defined as deaths that should not occur given current medical knowledge and technology. Numerous different lists of causes of death and the ages at which they should be considered avoidable have been used to measure avoidable mortality. In this analysis of the importance of definition we compare the two most commonly used approaches using a data set including all 11.8 million deaths that occurred in Britain in 1981-1998. These mortality data, disaggregated by age and sex, are analysed within a multilevel statistical framework, which allows analysis at a number of geographical scales simultaneously. A substantial difference in both the average trends and spatial patterns of the two definitions of avoidable mortality is found, indicating that the causes of death chosen have a considerable impact on the results found. Indeed, one particular cause of death was found to be largely responsible for the differences between the definitions. In addition, the spatial pattern of the two types of avoidable mortality is very different at the larger geographical scale while the pattern at the smaller scale is very similar. The findings illustrate the importance of considering the goals of any study before deciding on the definition of avoidable mortality to use.