This study analysed maternal deaths occurring in 1992 and 1993 in the city of Recife, Pernambuco State, Brazil, among all deaths of females aged 10 to 49 years, and to compare the results with vital statistics records. Information was obtained from 1013 death certificates and was supplemented through an analysis of medical records, anesthesia forms, nursing reports, results of autopsies performed by the Forensic Medicine Institute, necropsy reports from the Death Investigation Service, and interviews with both attending physicians and the women's relatives. Maternal mortality rates were calculated for the city of Recife for each year and for the total period studied. Twenty deaths were identified by Death Certificates as reported maternal deaths, and they represented 2% of cases among women in this age group. After the investigation proccess, 22 additional maternal deaths were identified for the entire period. According to vital statistics records, the maternal mortality rate was 37.0 per 100,000 live births; however, after adjustment, this figure increased to 77.7. This difference represents an under-recording rate of 52.4%. The study concluded that in the case of Recife, the 3.0 correction factor recommended by the Ministry of Health for the Northeast Region of the country was not applicable, since it would lead to an overestimation of the maternal mortality rate.