One of Bowlby’s students, Mary Ainsworth studied the bonding between mothers and children in a variety of cultures. She devised what’s called the strange situation paradigm, and with it noticed that mother-child relationships fell into three general types. In these experiments the child, usually about 18 months old, would be with the mother in a room. When mother was nearby, the child would explore objects in the surroundings, occasionally check that mother was nearby, then explore some more. If mother left, babies often reacted with distress and exploration behavior ceased. When mother returned, Ainsworth noted several sorts of reactions on the part of the baby. Differences among these reactions formed the basis of Ainsworth’s three attachment styles.
The majority of babies (55 to 60%) exhibited what Ainsworth called a Secure attachment style. They were more active and curious, they interacted well with strangers. They became very distressed when they mother left, but were very happy when she returned. Another group of babies exhibited what Ainsworth called an Avoidant style (about 25%). These interacted more coolly with their mother, exhibited less distress at her departure and were cool to her upon her return. Finally, Ainsworth’s Anxious-Ambivalent style (15-20% of babies), were distressed at mother’s leaving and angrier with her on her return.
Bowlby had suggested that attachment was a "cradle to grave" feature of relationships and working models of relationships are formed in infancy that influence relationships throughout adulthood. But no one really investigated attachment styles beyond childhood until the seminal contribution of Hazan and Shaver (87). Their contribution was to apply attachment theory to adult relationships. They devised a simple one-item measure of each attachment style. When they administered it to adults, they found proportions of each style nearly identical to those found by Ainsworth with babies.
There have been criticisms of the Hazan and Shaver approach. One criticism was that it was not true to the original Bowlby theory about attachment "working models." In response to this, Bartholomew and Horowitz attempted a revision of the typology, changing it from three to four categories. They argue that their conceptualization is closer to Bowlby’s original intent. They argue that the working models are based on two dimensions: feelings towards self and feelings towards others; and these feelings may be positive or negative. These dimensions yield the secure type (positive of self and other), dismissing type (positive to self, negative to other), preoccupied type (positive to other, negative to self), and fearful (negative to both).
Other objections have also surfaced. The Hendricks, for example, question aspects of the Ainsworth paradigm and its applicability to adult relationships. Pediatric researchers have found little evidence of long-term stability of infant attachment styles.
Baldwin, Fehr, their colleagues and other theorists have begun to question the tacit assumption of some early adult attachment work, that attachment styles are trait-like variables. A potentially more productive approach is to return to Bowlby’s working model notion, cast in cognitive terms, i.e., as relationship schemas. A series of studies by Baldwin et al. showed the feasibility of this approach. They found, for example, that people might be classified using the Hazan and Shaver typology, yet report relationship histories of other types of relationships. If attachment style were a trait, one might expect more consistency across situations. For example, Avoidants did not report that most of their relationships had been of this type. Rather they, and all three types, reported that most of their relationships had been of the Secure type. In other studies, Baldwin et al. showed that most people have cognitive access to all three relationship types. Other experiments showed that these types, like other schemas, could be primed or made more accessible, which exerted some influence on information processing. Other recent studies have supported the cognitive interpretation of attachment styles.