This study examined the impact of various temporal patterns of maternal interactive behavior with their infant on the infant's quality of attachment. The sample consisted of 52 dyads from the Berlin Longitudinal Study of Early Adaptation. Quality of attachment was assessed at 21 months with the Strange Situation Procedure. Nine infants were classified as securely attached: they sought and enjoyed psychological intimacy; 16 as insecure-defended: they avoided intimacy; and 19 as insecure-coercive: they displayed alternately angry and disarming behavior, thus maintaining their mothers' attention; 8 could not be classified. Maternal interaction quality was assessed from videotaped interactions at 3, 6, 9, 12, and 18 months, and the following components of maternal interaction were weighted: (1) sensitive; (2) covertly controlling; (3) hostile controlling; and (4) unresponsive. Prediction analysis was used to determine the association of maternal interaction patterns with later quality of attachment. Results suggested that a recent experience of maternal sensitivity seemed to be of more impact than a more continuous experience. Sensitive mothers seemed to be able to provide continuously "good enough" care but also improved in their behavior over time. Mothers of defended children were primarily characterized by covertly controlling behavior displayed continuously, but with improvement. Mothers of insecure-coercively attached infants had a strong component of unresponsive and/or hostile behavior as a nearly continuous pattern over the first 18 months. Unresponsive and/or hostile behavior, especially at the beginning of the mother-infant relationship, seemed to be of additional impact for later insecure-coercive attachment quality. (KDFB)