Four years after a study of paternal involvement among intact, middle-class families with an infant, a follow-up was conducted of 26 of the still intact families to determine the stability of paternal involvement and the psychological predictors of fathers' behavior at the time. Paternal involvement was assessed at both times in terms of care, play, and overall involvement. For purposes of comparison, scores at each time were converted to standard scores. In the first study, parents completed Spence and Helmreich's (1978) PAQ and AWS. Data were analyzed through regressions and repeated measures ANOVAs. Fathers' involvement at the time of the first study was fairly conventional. The average father provided 32 percent of his infant's parenting. On the basis of Radin and Goldsmith's (1985) report, it was anticipated that fathers' involvement would be stable over the period examined, and that fathers would be less involved with daughters than with sons at the time of the second study. The first hypothesis was supported, but no sex of child differences in fathering were found. In the first study, the father's attitude toward women (AWS), but not his personality (PAQ), predicted his caring for the baby. That relationship persisted over time. Fathers' attitudes at the time of the first study were the best predictors of their involvement when the child was five. Implications of the findings and their generality are discussed. (Author/RH)