Objective: To determine whether published subtypes of attachment disorder can be reliably identified by trained clinicians reviewing data from high-risk populations and to investigate the relationship between disorder classification and standardized measures of attachment behavior. Method: Twenty or more children aged 18 to 48 months and their primary caregivers were recruited from three sites: a treatment team for maltreated young children (n = 20), a homeless shelter (n = 25), and Head Start centers (n = 24). All dyads completed a semistructured clinical assessment and laboratory and home-based attachment measures. Results: All but one type of attachment disorder could be identified reliably by clinician raters ([kappa] range = 0.62-0.74, depending on subtype). Children from the maltreatment sample were significantly more likely to meet criteria for one or more attachment disorders than children from the other groups (p < .001). As predicted, children without an attachment disorder were more likely to be classified as securely attached than those with an attachment disorder (p = .03); however, children classified as having disorganized attachment were not more likely to receive an attachment disorder diagnosis. Conclusions: Attachment disorders can be reliably diagnosed in young children, though research on refining disorder criteria should precede intervention trials. J. Am. Acad. Child Adolesc. Psychiatry, 2004;43(5):568-577. Key Words: reactive attachment disorder, diagnosis, infancy, early childhood.