Objective-To evaluate the impact on mortality of standard Schwarz measles immunisation before 9 months of age. Design-Children vaccinated in 1980-3 at 4-5, 6-8, and 9-11 months of age were followed to migration, death, or the age of 5 years. Setting-One urban district and nine villages in two rural areas of Guinea-Bissau. Subjects-307 children vaccinated at 4-8 months and 256 at 9-11 months. Main outcome measures-Mortality from 9 months to 5 years of age for children immunised at 4-5, 6-8, and 9-11 months. Results-Mortality was significantly lower in children vaccinated at 6-8 months than at 9-11 months (mortality ratio=0.63, (95% confidence interval 0.41 to 0.97), p=0.047). As vaccination was provided in semiannual or annual campaigns it is unlikely that age at vaccination reflected a selection bias. The trend was the same in all three study areas. Improved survival after early immunisation was not related to better protection against measles infection. With a Cox multivariate regression model to adjust for age, sex, season at risk, season at birth, measles infection, and region, children vaccinated at 4-8 months had a mortality ratio of 0.61 (0.40 to 0.92, p=0.020) compared with children vaccinated at 9-11 months. Reimmunised children tended to have lower mortality than children who received only one vaccine (0.59 (0.28 to 1.27, p=0.176)). Conclusion-Standard measles vaccination before 9 months is not associated with higher childhood mortality than is the currently recommended strategy of immunising from 9 months, and it may reduce mortality. This has implications for measles immunisation strategy in developing countries.