To demonstrate the impact on perinatal mortality of inadequate treatment for maternal syphilis despite adequate screening. In 12 clinics providing antenatal care in Hlabisa, South Africa 1783 pregnant women were screened for syphilis at their first antenatal visit between June and October 1998. Pregnancy outcome was determined among those with syphilis. A total of 158 women were diagnosed with syphilis: prevalence 9% (95% CI 8-10%). Mean gestation at first antenatal visit was 24 weeks. Thirty women (19%) received no treatment and 96 (61%) received all three recommended doses of penicillin. Among those receiving at least one dose, mean delay to the first dose was 20 days. Among those fully treated mean delay to treatment completion was 34 days. Pregnancy outcome was known for 142 women (90%) and there were 17 perinatal deaths among 15 women (11%). Eleven of 43 women (26%) who received one or fewer doses of penicillin experienced a perinatal death whilst only four of 99 women (4%) who received two or more doses of penicillin did so (P = 0.0001). Protection from perinatal death increased with the number of doses of penicillin: linear modelling suggests that one dose reduced the risk by 41%, two doses by 65% and three doses by 79%, compared with no doses. A dose-specific, categorical model confirmed reduction in risk by 79% for all three doses. Despite effective screening, many pregnant women with syphilis remain inadequately treated, resulting in avoidable perinatal mortality. Delays in starting and finishing treatment, as well as incomplete treatment occur. Near-patient syphilis testing in the antenatal clinic with early treatment could improve treatment of syphilis and reduce perinatal mortality, and a randomized trial to test this is underway.