Although the neonatal mortality rate (NMR) in Bangladesh remained steady between 1995-99 and 1999-2003 (41-42 deaths per 1,000 live births), evidence from the management information system (MIS) of a large nongovernmental organization (NGO) program indicates that the NMR declined by about 50 percent between 1996 and 2002 in the area served. This study aims to validate the recording of neonatal deaths among the cohort of children registered as born in 2003 and to assess the evidence of a decline in the NMR. It also measures the coverage of reproductive health outreach services, focusing on 12 of the 27 NGOs that have provided services in the same areas since 1996. Field-workers' registers, verbal autopsy reports, and immunization records were checked to confirm infants' survival. Interviews were conducted with 142 mothers of children who died within 28 days postpartum and with a random sample of 109 women with registered stillbirths. Out of 11,253 registered live births in 2003, 210 neonatal deaths were found, compared with 194 deaths that were reported in the MIS for 2003. The corrected NMR was 19 deaths per 1,000 live births, and it was in the range of 15-29 deaths per 1,000 live births in 11 of the NGO areas. Because underreporting of neonatal deaths was probably higher in 1996 when the MIS-reported NMR was 39 deaths per 1,000 live births, the decline in the NMR is likely to have been genuine.