Ainsworth distinguished between secure and anxious attachment. Some secure infants strongly seek physical contact, are reassured by it and return to play, while others warmly greet their attachment figure (smile broadly, show toys, vocalize). But what they all have in common is that they are active in initiating renewed engagement with the caregiver. By contrast, those with "anxious attachment" either actively avoid their caregivers upon reunion or fail to be comforted by them.
Some critics have questioned whether the Strange Situation measures attachment patterns or simply reflects differences in infant temperament. Couldn't it be that some children are simply more difficult to comfort than others? But when one examines how Ainsworth's assessments were conducted, it becomes obvious why, as found in dozens of studies, temperament doesn't predict attachment security or insecurity. According to her methodology, the amount an infant cries during separation (its proneness to distress) isn't relevant to determining whether the attachment is secure or insecure. What determines an infant's level of attachment security is its behavior when the caregiver returns. While some babies are "thoroughly distressed" by separation, their relationship with the caretaker will be classified as secure if, despite their distress, they effectively seek contact upon reunion and are comforted by it, later returning to play. Among those who cry at separation, it's only those who fail to be comforted on reunion (either being passive or angrily resisting attempts at comfort) who are classified as having insecure or anxious/resistant attachment. Conversely, it isn't the case that infants who don't cry at separation are all in relationships classified as insecure. Babies who show no separation distress, but actively greet and initiate interaction with the caregiver upon reunion are classified as having secure relationships. Only nondistressed infants who ignore or otherwise actively avoid parents upon reunion (demonstrating avoidant attachment) are considered insecurely attached.
Developmental changes in the child further corroborate that these are relationship assessments and not measures of infant traits. Many 12-month-olds cry during lab separations, and those who are securely attached seek and are comforted by contact upon reunion. At 18 months, few toddlers cry at separation, though play generally becomes subdued. On reunion, those 18-month-olds who are secure typically don't seek physical contact (they don't need it now), but they actively engage the parent. Thus the attachment relationship can be classified the same as at an earlier age, even though all of the behaviors may change as the child grows. In fact, the amount of crying, smiling, and seeking proximity demonstrated by different infants at 12 months is unrelated to the amount of those same behaviors they show at 18 months. It's only the organization of the behavior that remains constant. Thus, the overall dynamics of the mother–child relationship have greater predictive value than do the more easily measured, individual behaviors.
What Ainsworth observed about children with avoidant attachment patterns has especially important meaning for clinicians. She found that avoidant infants had experienced routine rebuffs, specifically when they needed tender care from the caregiver. In general, their mothers held them as much as other mothers held their babies, just not when they really needed it. Therefore, they cried more in the routine home observations and explored less than did the securely attached babies. Later, they were strikingly more dependent on their schoolteachers. Bowlby specifically predicted that infants whose normal needs for sensitive responsiveness and emotional closeness weren't met, "including those pushed toward early independence," would later be more dependent. Simple temperament explanations can't account for these findings.