Similarities and dissimilarities among the cluster profiles are evident in the graphs in Figures 3-1a and 3-1b. For males, cluster 1 represents the age pattern that predominates among the urban population of Russia, and it differs from the other clusters considerably. Rural patterns fall into two groups: Clusters 5 and 6 are very similar and geographically represent rural areas in the central part of European Russia, whereas Clusters 2 and 3 represent two other types of rural patterns, which are not closely matched. Cluster 4 is not easily grouped with the other clusters; compared with the others, it is most similar to average Russian mortality. Thus we can conclude that there were five distinct male patterns of mortality in Russia in 1988-1989—two urban and three rural.
For females, clusters 1 and 2 are very similar and can be grouped to represent the main urban profile. The same is true with respect to clusters 3 and 4, which represent the rural profile. Grouping of clusters 5 and 6 is more problematic; indeed, under stringent criteria, female patterns fall into four categories. Cluster 5 can be considered an extreme of the urban pattern, whereas cluster 6 is a ''mixed" urban and rural cluster, and has unusual features.
In the following discussion, we examine features of the age patterns of the clusters and the contribution of selected causes of death to those patterns. The rural and urban classification of the clusters is then examined.
In general, health and social conditions influence the shape of the mortality curve through causes of death. It is known that violent deaths induce a "hump" on the male mortality curve, indicating excess younger adult mortality. A similar effect is observed in the shape of female mortality pattern as a result of maternal mortality.
To investigate the cause-of-death composition of the cluster mortality patterns, we used the following approach for presentation of the profiles. We analyzed each cluster pattern of mortality relative to average Russian mortality by applying a method of decomposition of differences in the level of life expectancy by age and cause of death, using the approach of Andreev (1982) and Arriaga (1984, 1988) (the same approach as that of Pressat  and Shkolnikov et al. [in this volume]). Prior to the decomposition, we shifted the levels of expectation of life of the clusters to the levels of the Russian average life tables (64.0 years for males and 74.0 years for females) to eliminate differences in the level of mortality.8
Figures 3-1a and 3-1b show in bold lines the age components of the difference in e(0) between the clusters and the Russian average (left y-axes). These lines almost mirror the logit scores of age probabilities of dying, which means the