From a sample of middle-class mothers and their 3-year-old children, a selected group of 36 mothers were divided into 2 groups according to the quality of their responses to the Adult Attachment Interview as a Questionnaire (Crandell, Fitzgerald, & Whipple, 1997). Twenty mothers provided coherent accounts of their early parent–child relationships (secure) and 16 mothers provided idealised, entangled, or otherwise incoherent accounts of their early parent–child relationships (insecure). The mothers were administered an abbreviated version of the WAIS-R and the children were given an abbreviated version of the Stanford-Binet Intelligence Scale. The quality of mother–child interactions was assessed by videotaping a 20-minute play episode and clean-up period, and by rating the degree of synchrony according to a modified version of the Belsky Parent–Child Interaction System (Whipple, Denburg, & Davies, 1993). The results were that children of secure mothers scored 19 points higher on the Stanford-Binet test compared to children of insecure mothers. (The adjusted mean difference was 12 points when maternal IQ, education, and family SES were taken into account.) The group difference in the children's IQ remained significant when comparisons were made between a subgroup of 12 secure and 12 insecure mothers who were matched for maternal IQ. Finally we examined the subgroup of 16 cases where child IQ scores were either 10 points higher or lower than maternal IQ. In all 6 cases where child IQ was at least 10 points below maternal IQ, the child had a mother who was insecure; in contrast, only 4 of the 10 children who had IQ scores 10 points higher than maternal IQ had an insecure mother. In terms of parent–child interaction patterns, there was suggestive evidence that the degree of parent–child synchrony was also related to child IQ. The results suggest important social-developmental contributions to young children's performance on standardised tests of intellectual ability.