It has been reported that Zung's Self-Rating Depression Scale (SDS) loses its validity in predicting cases of depression when used repeatedly. The validity of SDS was tested against the subject's Research Diagnostic Criteria (RDC) diagnoses of major/minor depressive disorders in 120 pregnant women four times throughout the perinatal period. Different sets of predictive SDS items were found at different time points. We developed an 'RDC-like' algorithm from SDS items. Though varying in sensitivity, we found that this formula yielded low, but stable, positive predictive values and constantly high negative predictive values. We suggest that the RDC-like algorithm is a better alternative for screening depression among perinatal women.