We set out to explore the meaning of the attachment categories in the Cassidy/Marvin strange situation procedure, as employed in the home, using data from a longitudinal study of children adopted into UK families up to the age of 42 months from Romanian institutions, and of adopted children without the experience of institutional care--both groups being assessed at 4 and 6 years of age. Inter-rater reliability was found to be good. Security (meaning the use of the parent as a secure base and no negative behavior on reunion) was the modal categorical rating in both the institution-reared and comparison groups, but the category of anomalous non-normative behavior (meaning a lack of any ordered attachment behavior as covered by the standard ratings), previously labeled "insecure-other", was more common in the institution-reared children. Because this was unassociated with the usual manifestations of insecurity (such as avoidance or resistance) and because it was largely evident in interactions with the stranger, it is concluded that the adjective "insecure" was not appropriate. Although this "insecure-other" category was associated with disinhibited attachment as reported by parents (meaning a lack of differentiation among adults, a willingness to go off with strangers and a lack of checking back with parents in anxiety-provoking situations), it did not prove to be a good index of disinhibited attachment because changes over time in the "insecure-other" category were not associated with changes in the rate of disinhibited attachment. Also, whereas the rate of "insecure-other" fell markedly in the institution-reared group between 4 and 6 years of age, it rose in the comparison group, raising queries over its meaning. (Contains 4 tables and 1 figure.)