Studies of cognitive vulnerability to depression in young children have, in the main, relied on self-report questionnaires (e.g. of self-esteem, attributional style). They have failed to produce convincing evidence of a cognitive vulnerability to depression in children under 8 years. In the current study latent depressive cognitions were investigated in the 5-year-old children (N = 94) of depressed and well mothers in a situation of mild stress, that is, the threat of losing a card deal in a modified version of the competitive children's card game “Snap”. In the context of “losing”, but not “winning”, deals, children who had been exposed to maternal depression, either in the previous 12 months or at any other time during their lifetime, were more likely than nonexposed children to express depressive cognitions (hopelessness, pessimism, and low self-worth). The association between depressive cognitions and recent exposure to maternal depression was in part accounted for by current maternal hostility to the child. The results of this study stand in contrast to those of studies which have used questionnaire methods to assess vulnerability to depressive cognitions in this age group. They suggest that it might be important to employ ecologically realistic situations to access latent self-cognitions in young children; and they underscore the importance, increasingly evident in research with adults and older children, of employing methods that involve the induction of low mood in order to elicit cognitions relevant to depression.