BACKGROUND: There are limited data on the prevalence and approaches to screening for depression among pregnant women living in resource poor settings with high HIV burden. METHODS: We studied the reliability and accuracy of the Center for Epidemiologic Studies Depression (CES-D) scale in 123 (36 HIV-infected and 87 -uninfected) pregnant women receiving antenatal care at Gulu Regional Referral Hospital, Uganda. CES-D scores were compared to results from the psychiatrist-administered Mini-International Neuropsychiatric Interview (MINI) for current major depressive disorder (MDD), a "gold standard" for assessing depression. We employed measures of internal consistency (Cronbach's alpha), and criterion validity [Area Under the Receiver Operating Characteristic Curve (AUROC), sensitivity (Se), specificity (Sp), and positive predictive value (PPV)] to evaluate the reliability and validity of the CES-D scale. RESULTS: 35.8% of respondents were currently experiencing an MDD, as defined from outputs of the MINI-depression module. The CES-D had high internal consistency (Cronbach's alpha = 0.92) and good discriminatory ability in detecting MINI-defined current MDDs (AUROC = 0.82). The optimum CES-D cutoff score for the identification of probable MDD was between 16 and 17. A CES-D cutoff score of 17, corresponding to Se, Sp, and PPV values of 72.7%, 78.5%, and 76.5%, is proposed for adoption in this population and performs well for HIV-infected and -uninfected women. After adjusting for baseline differences between the HIV subgroups (maternal age and marital status), HIV-infected pregnant women scored 6.2 points higher on the CES-D than HIV-uninfected women (p = 0.032). CONCLUSIONS: The CES-D is a suitable instrument for screening for probable major depression among pregnant women of mixed HIV status attending antenatal services in northern Uganda.