Pregnancy and childbirth are among the strongest risk factors for depression but the neurocognitive mechanisms underlying this enhanced risk are unknown. This study investigated emotional and non-emotional cognition in 57 pregnant women with or without an affective disorder during their third trimester, and the association between cognitive biases and subsequent postpartum depression (PPD). Of the pregnant women, 22 had a diagnosis of unipolar disorder (UD) and seven of bipolar disorder (BD) in full or partial remission, while 28 had no history of affective disorder. We included a control group of 29 healthy non-pregnant women. First, participants were interviewed, completed non-emotional and emotional cognitive tests and lastly filled out questionnaires. The participants were assessed two times after birth: at a home visit shortly after birth, and with a telephone interview to assess PPD in the first six months after birth. Healthy pregnant women rated infant cries less negatively than non-pregnant women, possibly reflecting preparation for motherhood. Pregnant women with UD exhibited a negative bias in ratings of infant cries, whereas pregnant women with BD showed a positive bias in ratings of infant happy faces and recognition of adult facial expressions. Across all pregnant women, more negative ratings of infant cries were associated with enhanced risk of PPD. Negatively biased perception of infant cries during pregnancy may thus signal vulnerability toward PPD.