The current study used data from the Early Childhood Longitudinal Study (Birth Cohort) to explore determinants of father involvement. Belsky's (1984) process model of parenting and Lamb and colleagues' (1987) model for measuring fathering involvement were used as foundations for the development of a theoretical framework examining specific constructs of caregiving and play. Father, mother, and child characteristics were measured at three time points (9-months, 2 years, and 4 years of age) for approximately 2,900 biologically related cohabitating families. Specifically, parental socioemotional functioning, the marital relationship, father role identity, and maternal involvement, as well as child gender and temperament were examined in relation to father caregiving and play over time. In addition, demographic characteristics of the family were included as covariates. Latent dual change score models were examined for father caregiving and play and determinants of each behavior were included in a final model. Results indicated that caregiving and play are differentially determined, with only child gender and father role identity relating to both behaviors. Fathers engaged in more caregiving when mothers reported higher depression. Fathers' role identity was related to increased caregiving and play behaviors, as well as faster rates of change in play when fathers more strongly identified with their role as a father. Mother play and father play were related, and fathers increased in play over time at a faster rate with girls than boys. In addition, temperament and infant gender interacted to predict changes in caregiving over time. A theoretical framework for father involvement is presented.