This study investigated the cross-lagged relationship between father involvement and child problem behaviour across early-to-middle childhood, and tested whether temperament modulated any cross-lagged child behaviour effects on father involvement. It used data from the first four waves of the UK’s Millennium Cohort Study, when children (50.3 % male) were aged 9 months, and 3, 5 and 7 years. The sample was 8302 families where both biological parents were co-resident across the four waves. Father involvement (participation in play and physical and educational activities with the child) was measured at ages 3, 5 and 7, as was child problem behaviour (assessed with the Strengths and Difficulties Questionnaire). Key child and family covariates related to father involvement and child problem behaviour were controlled. Little evidence was found that more father involvement predicted less child problem behaviour two years later, with the exception of father involvement at child’s age 5 having a significant, but small, effect on peer problems at age 7. There were two child effects. More hyperactive children at age 3 had more involved fathers at age 5, and children with more conduct problems at age 3 had more involved fathers at age 5. Child temperament did not moderate any child behaviour effects on father involvement. Thus, in young, intact UK families, child adjustment appears to predict, rather than be predicted by, father involvement in early childhood. When children showed more problematic behaviours, fathers did not become less involved. In fact, early hyperactivity and conduct problems in children seemed to elicit more involvement from fathers. At school age, father involvement appeared to affect children’s social adjustment rather than vice versa.