The Stability of Individual Differences in Infant Attachment: Comments on the Thompson, Lamb, and Estes Contribution

In a recent article, Thompson, Lamb, and Estes reported that only 53% of middle-class infants seen in the Ainsworth Strange Situation at 12.5 and 19.5 months received the same attachment classification (avoidant, secure, or resistant) at both ages. Thompson et al. suggest that differences between their results and very high levels of stability reported by Waters can be attributed to subject selection that minimized the incidence of family stress and maximized the possibility of finding stability in the earlier study. In fact, Waters did not employ any procedure for maximizing stability beyond the use of a middle-class sample. Stability of individual differences in infant attachment is a well-replicated finding. Nonreplication in a single small sample is difficult to evaluate. Among the possible explanations, sample differences are weak and necessarily post hoc as explanations for the low stability reported by Thompson et al. It is suggested that when populations are heterogeneous with respect to significant influences on development, quasi-experimental comparisons among well-defined elements of the population and relevant control groups can inform developmental theory; correlational data from small unselected samples cannot.