Based on adolescent mothers' reports, longitudinal patterns of involvement of young, unmarried biological fathers (n=77) in teenage-mother families using cluster analytic techniques were examined. Approximately one third of fathers maintained high levels of involvement over time, another third demonstrated low involvement at both time points, and the final third started out highly involved at Wave 1 but decreased to low levels of involvement by Wave 2. Multinomial logistic analyses suggest that mothers' positive relationships with both the father and his family predict a greater likelihood of initiated and sustained high father involvement. In contrast, stronger support from the maternal grandmother is related to decreasing father involvement over time, and coresidence with the grandmother is related to sustained low father involvement. Whereas a decreasing pattern of father involvement was significantly associated with increased maternal parenting stress over time, the patterns of father involvement were unrelated to changes in young mothers' levels of depressive symptoms and mastery.