Purpose The Edinburgh Postnatal Depression Scale (EPDS) aims at detecting postpartum depression. It has been hypothesized that a subscale (items 3, 4, 5) may detect anxiety. The aim of this study is to assess whether this EPDS anxiety subscale is present in a community-based dataset, and if so, to assess its validity and stability during the first six months postpartum. Methods We obtained EPDS data of a community sample of 1612 women at 1 month, with follow-up at 3 and 6 months, postpartum (Post-Up study). We performed an exploratory factor analysis on the EPDS forcing two- and three-factor solutions. We assessed the correlations of the extracted factor subscales and the total EPDS with the short-form of the STAI (STAI-6). We examined the stability of the identified factors by means of a confirmatory factor analysis (CFA), using the EPDS data collected at 3 and 6 months postpartum. Results Both the two- and three-factor solutions contained a hypothesized anxiety subscale of items 3,4,5,10, and fitted well with the 3- and 6-months EPDS data, with CFI and TLI values >.99 and RMSEA and SRMR values < .035 and < .45. The subscale’s Pearson correlations with the STAI-6 were moderate: .516, compared to .643 for the total EPDS. Conclusions The factor structure of the EPDS is stable across the first six months postpartum, and includes the subscale assumed to represent anxiety. However, this subscale as well as the total EPDS correlate only moderately with anxiety criteria. Using the EPDS thus does not imply adequate screening for anxiety.