The technique of latent class analysis has not previously been applied to demographic data. The present paper analyzes 3709 admissions to a hospital for the retarded on the basis of five variables, and shows how a latent class analysis reduces the population to two groups, each of which tend to feature within group independence among the five variables. A mortality analysis of each group shows great differences in the mortality rate when length of stay in the hospital is held constant. It is concluded that latent class analysis has possibilities as an alternative standardization process for demographic analysis.