Monitoring measles vaccine efficacy is an important form of quality control in immunization programs. Retrospective cohort studies of vaccine efficacy were conducted in Mozambique with the aim of adopting this method for regular use. The authors studied 1,215 and 790 children aged 12-35 months in the cities of Nampula and Beira, respectively. In Nampula (1985), vaccine efficacy was estimated to be 40%, and in Beira (1986), it was estimated to be 59%. To investigate the possibility of falsely low results due to poor specificity of maternal reporting of measles cases, a seroepidemiologic study of 600 children aged 12-47 months was conducted in Nampula in 1986. The specificity and sensitivity of a maternal history of measles in an unvaccinated child were estimated as 83% and 56%, respectively. These results were used to obtain an adjusted vaccine efficacy estimate of 66%, almost double the estimate of 37% obtained using data on history of vaccination and illness alone. The large bias introduced into the vaccine efficacy estimation by low specificity of disease diagnosis is a serious limitation to the use of retrospective cohort methods to assess efficacy.