The aim of this study was to investigate the development of mother–infant interaction patterns from 3 to 12 months among three groups of mother–baby pairs recruited during pregnancy: one group from residential substance abuse treatment (n=28), a second group from psychiatric outpatient treatment (n=22), and a third group from well-baby clinics (n=30). The mother–infant interaction at 3 and 12 months was assessed by the Parent–Child Early Relational Assessment (PCERA), which consists of maternal, child and dyadic subscales (Clark, 2006). Linear mixed effects models were used to analyze group differences and the changes in mother–infant interaction from 3 to 12 months.At 3 months, pairwise comparisons showed that the group with psychiatric problems had significantly more difficulties in the mother–infant interaction than the two other groups. The group with substance abuse problems was not significantly different from the two other groups. At 12 months, the mother–infant pairs in the substance abuse group showed significantly more relational disturbances than the non-clinical pairs, as well as a poorer affective quality of interaction than the dyads in the group with psychiatric problems. Analysis of change from 3 to 12 months showed that difficulties in the interaction increased among the mother–baby pairs in the substance abuse group, while improvements were displayed in the two other groups. These results underline that mother–infant pairs at double risk due to maternal substance abuse and other non-optimal factors, are in need for long-term follow up in order to prevent the development of negative interactional patterns.