Using the example of France from 1925 to 1993, the authors examine the problems posed in the analysis of causes of death over long periods of time by changes in the way the data are classified. They note that there are three major problems to be resolved: changes in the International Classification of Diseases [ICD], changes in the coding practices employed by government agencies, and inconsistencies among individual physicians in diagnosing the cause of death. They present methods for overcoming each of these difficulties and urge the implementation of a transitional period using both old and new classification for future revisions of the ICD.