Inter-judge agreement in the evaluation of attachment style was examined across different levels of acquaintanceship. A self-rating scale on attachment style was administered to 242 students. Then, 72 participants-25 avoidant, 25 secure, and 22 ambivalent-were invited to the laboratory in unacquainted same-sex pairs, videotaped during a five-minute conversation, and asked to rate the attachment style of their conversation partner. In addition, each participant asked four friends to rate his/her attachment style and the videotaped interactions were shown to a separate sample of strangers who rated participants' attachment styles. The results showed high correlations between participants' self-reported attachment styles and the manner in which they were perceived by their friends and conversation partners. In contrast, correlations with ratings by strangers based on videotaped interactions were low. This pattern of findings was observed in both discrete and continuous measures of attachment. The findings suggest that attachment style can be considered an observable interpersonal trait.