Abstract The projection of mortality rates requires inter alia close examination of the mortality experienceof a population overa long periodof time and will usually also involvethe analysis of mortality trends by cause of death. In two of the more important recent contributions, techniques were devised for explaining change in life expectancy in terms of mortality changes in particularage groups and by different causes of death. The approaches adopted by the authors differ, and the purpose of this article is to reconcile the two and tie the results in with those obtained by earlier writers. A new method for explaining the change in a life expectancy differential in terms of the observed changes in the mortality differentials and the observed change in overall mortality level is also described.