A study of the incidence of diabetes mellitus in Oslo for the period 1925-54 (Westlund, 1966), based on a comprehensive compilation of hospital documents, has been extended to 1956-65. The incidence rates for the two periods have been compared, by taking the year of diagnosis and age at diagnosis in each case. The incidence rates for 1925-54 were revised by adding cases for this period discovered during the 1956-65 study, and showed no major changes. In particular, few new cases were found that had been diagnosed during the second world war, so that the dramatic reduction in incidence shown in Westlund's paper is not affected. The trend is fairly constant, with two exceptions: there was an increase in incidence for boys and girls younger than 10 years and a decrease for women aged between 40 and 59 years. After considering the various sources of error in such a comparison it is concluded that since 1950 there may have been a real increase in rates of diabetes in children. The decrease in rates for middle-aged women is not supported by mortality data for which statistics are given for associated as well as for underlying causes. Mortality data are however difficult to interpret, and by relying on hospital documents and death certificates considerably higher incidence rates have been presented than would have been the case had the incidence estimations been based on death certificates alone. The incidence rates in 1956-64 for children tally fairly well with those reported for Great Britain and Ireland (Bloom et al., 1975) and for northern Norway (Bratlid, 1976).