This paper explores the effects of infant predominant state of arousal upon very early mother-infant interaction, through observation of a group of 20 healthy 3-day-old nursery roommates and their mothers during hospital feeding sessions. In an experimental procedure involving a modified time-sampling technique, two observers simultaneously recorded the occurrence of an array of variables on printed, precoded forms. In addition to infant initial and predominant state, behaviors observed included a variety of maternal and neonatal behaviors. Repeated measures analyses of variance were performed for the dependent variables, with significant differences obtained for specific infant sucking behaviors, maternal stimulation behaviors and social-visual behaviors. Infant predominant state, in the earliest interactions, appears to be a very strong predeterminant of maternal behavior and of neonate-mother interaction. The comparison of group differences in patterns of neonate-mother interaction reveals the predominance of social interaction for alert neonates and their mothers, while the feeding process itself was predominant for sleepy neonates and their mothers. It is suggested that effects of infant predominant state have profound implications for the development of the early mother-infant relationship, particularly when feeding is done under restrictions of hospital schedules. Possible advantages are seen in rooming-in accommodations for mother and neonate. (BF)