The impact of exposure to measles before 6 months of age has been investigated by comparing survival to 5 years of age for exposed children and controls in an urban (Bandim) and a rural (Quinhamel) area of Guinea-Bissau. In Bandim, cumulative mortality from time of exposure to age 5 years was 34.4% among exposed children and 9.3% among controls. In a matched pair analysis, exposed children had a mortality ratio (MR) of 3.80 (95% confidence interval [CI]: 1.42-10.18) compared with controls. In an unpaired analysis using Cox' regression model to standardize for background variables (sex, measles infection, age at exposure, exposure from own household, measles vaccination), there was little change in the MR (3.84, CI: 1.55-9.48). Even after 2 years of age, the exposed children tended to have higher mortality (MR = 7.96, CI: 0.98-64.74). In the rural area, the MR between exposed children and controls was 11.39 (CI: 1.42-91.51). Limited serological data suggest that at least some of the exposed had subclinical measles. In the urban area, where two studies of early exposure have been carried out, excess mortality among exposed children corresponds to 40% and 52%, respectively, of the acute measles mortality. Since these deaths would not be associated with measles in a study of protection against death after vaccination, measles immunization may have a much greater effect on childhood mortality than has previously been assumed.