Research on the effect of paternal mental health problems, particularly on young children, is based predominantly on clinical levels of depression. Furthermore, potential mediators such as marital discord have often been overlooked. This longitudinal community study assessed the association between paternal mental health symptoms in a community sample (N?=?705) assessed at 3?months postnatally (Edinburgh Postnatal Depression Scale) and 36?months (General Health Questionnaire) and children's socio-emotional and behavioural problems at 51?months (Strengths and Difficulties Questionnaire) as reported by mother, father and teacher. Controlling for socioeconomic status and maternal mental health symptoms at 3 and 36?months, paternal postnatal depressive symptoms predicted more father-reported child problems at 51?months but, in contrast to previous findings, not mother-reported problems. Paternal mental health symptoms at 36?months predicted both maternal and paternal reports of child problems at 51?months controlling for both paternal and maternal postnatal symptoms. Paternal mental health symptoms at 3 and 36?months were not significant predictors of teacher-reported child problems. Postnatal marital discord and paternal mental health problems at 36?months both mediated the relationship between paternal postnatal symptoms and later child emotional and behavioural problems. Child gender did not moderate the relationship. Implications for interventions are discussed.