This study examines the influence of mother-child and family interactions on the development of child attachment representations in middle childhood for a sample of 49 families. Mother-child interactions were observed during a snacktime in a lab setting (Moss, Rousseau, Parent, St-Laurent, & Saintonge, 1998) when children were 5-6 years old. Three years later, children's attachment representations were assessed using a doll play narrative procedure (Solomon, George, & DeJong, 1995) in the lab setting. Within 6 months of the second lab visit, family interactions were filmed during mealtime and coded using the Mealtime Interaction Coding System (MICS; Dickstein, Hayden, Schiller, Seifer, & San Antonio, 1994). Results showed clear differences between attachment groups on quality of mother-child and family interaction with the secure/confident group showing highest and the disorganized/frightened group showing lowest quality interactions. Family interactions predicted children's attachment representations, after controlling variance explained by prior mother-child interactions.