A generalisability study examined the hypotheses that avoidant attachment, reflecting the representation of others, should be more relationship-specific (vary across relationships more than across individuals), while attachment anxiety, reflecting self-representation, should be more generalisable across a person's relationships. College students responded to 6-item questionnaire measures of these variables for 5 relationships (mother, father, best same-gender friend, romantic partner or best opposite-gender friend, other close person), on 3 (N = 120) or 2 (N = 77) occasions separated by a few weeks. Results supported the hypotheses, with the person variance component being larger than the relationship-specific component for anxiety, and the opposite happening for avoidance. Anxiety therefore seems not to be as relationship-specific as previous research suggested. Possible reasons for discrepancies between the current and previous studies are discussed.