Two five-year cohorts, 1967-1971 and 1977-1981, of the Medical Birth Registry of Norway, were utilized to analyze secular trends in early neonatal mortality rates, controlling simultaneously for birth weight, parity, maternal age, and sex. Improvement in the crude early neonatal mortality rate, from 6.5 to 2.9 per 1,000, was partitioned into one portion (18.4%) attributable to changes in the distribution of the birth weight or the other independent variables and another portion (81.6%) attributable to improved survival within each specific subgroup in the multivariate table. The early neonatal mortality rate was found to decrease in all subgroups, the low birth weight groups showing a significantly greater decline than the normal birth weight groups. Women with diabetes, epilepsy, or blood group antibodies were studied in separate analyses. For women with diabetes, the odds ratio of the early neonatal mortality rate between the first and the second periods was found to be three times the odds ratio for women without diabetes, and 2.9% of the total improvement of the crude early neonatal mortality rate was attributable to an improved survival in these infants.