The Working Model of the Child Interview (WMCI) was used to investigate mothers' internal representations of their 3-month-old infants in a clinical sample of eight mothers all of whom had a prenatal history of Major Depressive Disorder. Their representations of their infants were less than half as likely to be classified as balanced (balanced 25%, disengaged 25%, distorted 50%) compared to normative data obtained from non-clinical populations (balanced 52-65%, disengaged 3-24%, distorted 24-32%). This effect was entirely due to the WMCI representations of the six mothers who were depressed after delivery, none of whom had balanced representations of their infants. While these findings need replication in a larger sample, there are important clinical implications given the strong link between mothers' WMCI classifications and infant attachment behaviour at 1 year. The WMCI may be of value as a screening instrument and in the assessment of therapeutic interventions.