To study whether exercise during pregnancy reduces the risk of postnatal depression. Design. Randomized controlled trial. Setting. Trondheim and Stavanger University Hospitals, Norway. Population and Sample. Eight hundred and fifty-five pregnant women were randomized to intervention or control groups. Methods. The intervention was a 12week exercise program, including aerobic and strengthening exercises, conducted between week 20 and 36 of pregnancy. One weekly group session was led by physiotherapists, and home exercises were encouraged twice a week. Control women received regular antenatal care. Main Outcome Measures. Edinburgh Postnatal Depression Scale (EPDS) completed three months after birth. Scores of 10 or more and 13 or more suggested probable minor and major depression, respectively. Results. Fourteen of 379 (3.7%) women in the intervention group and 17 of 340 (5.0%) in the control group had an EPDS score of ≥10 (p=0.46), and four of 379 (1.2%) women in the intervention group and eight of 340 (2.4%) in the control group had an EPDS score of ≥13 (p=0.25). Among women who did not exercise prior to pregnancy, two of 100 (2.0%) women in the intervention group and nine of 95 (9.5%) in the control group had an EPDS score of ≥10 (p=0.03). Conclusions. We did not find a lower prevalence of high EPDS scores among women randomized to regular exercise during pregnancy compared with the control group. However, a subgroup of women in the intervention group who did not exercise regularly prior to pregnancy had a reduced risk of postnatal depression.