An important corollary of this finding is that attachment studies which do not control for a family's social background--and many do not--may produce results which, for example, seem to prove levels of security in attachment classification that have a strong positive correlation with a factor like maternal sensitivity, when, in fact, levels of security and sensitivity are both caused by other unstudied factors from their social background. This means that the results of correlational studies which seem to prove the long-term effects of an infant's attachment classification cannot be taken at face value if those studies were uncontrolled: the strength of correlations is likely to have been inflated by one or more of the unstudied background variables.

...

The best-known longitudinal studies to have researched the long-term effects/correlates of infant-caregiver bonds on grown-up functions have all failed to produce the kinds of results attachment theory predicts. This is especially true in studies that have controlled for the continuity and discontinuity of background variables experienced by the child while she or he was developing. One longitudinal study in Minnesota showed that, if background variables were properly taken into account, attachment security in infancy only accounted for 5% of the variability in social competence when the study-children had reached the age of nineteen. Two well-known German studies also failed to find any significant correlation between security of attachment (as measured in the SSP) in toddlers and a variety of measures of social relating after age ten. Long-term Israeli research also found that continuity in attachment representation correlated strongly with the stability of the caregiver's environment, and discontinuity with instability.