To demonstrate validity, questionnaires should measure the same construct in different groups and across time. The Edinburgh Postnatal Depression Scale (EPDS) was designed as a unidimensional scale, but factor analyses of the EPDS have been equivocal, and demonstrate other structures: this may be because of sample characteristics and timing of administration. We aimed to examine the factor structure of the EPDS in pregnancy and postpartum at 4 time-points in a large population-based sample. We carried out exploratory and confirmatory factor analysis on the Avon Longitudinal Study of Parents and Children sample (n = 11,195–12,166) randomly split in 2. We used data from 18 and 32 weeks pregnancy gestation; and 8 weeks and 8 months postpartum. A 3-factor solution was optimal at all time-points, showing the clearest factor structure and best model fit: Depression (4 items) accounted for 43.5–47.2% of the variance; anhedonia (2 items) 10.5–11.1%; and anxiety (3 items) 8.3–9.4% of the variance. Internal reliability of subscales was good at all time points (Cronbach’s αs: .73–.78). The EPDS appears to measure 3 related factors of depression, anhedonia, and anxiety and has a stable structure in pregnancy and the first postnatal year.