We analysed results from seven American and British studies that compared groups of mothers with and without clinically diagnosed depression, and assessed the attachment category of their infants (under 3 years) using the Strange Situation. The samples were predominantly middle-income and free of risk factors other than maternal depression. Meta-analysis using loglinear modelling and standardised residuals showed that the effect of depression on the distribution of infants' attachment was statistically heterogeneous. However, after removing one outlier study, the effect of depression was homogeneous across the remaining six studies. Infants of depressed mothers showed significantly reduced likelihood of secure (B) attachment and marginally raised likelihood of avoidant (A) and disorganised (D) attachment. The first two effects varied considerably in magnitude between studies, whereas the increase in disorganised attachment, from 17% to 28% on average, was consistent.