The long-term implications of sexual abuse in childhood or adolescence (CSA) have been relatively well documented regarding attachment (disorganized attachment in childhood, unresolved trauma in adulthood), stress reactions (altered patterns of stress reactivity under experimental conditions), and psychopathology. Attachment has been shown to mediate the implications of CSA, namely on psychopathology. The implication of attachment on stress responses of abused persons has not been documented. Twenty-seven 20-46 years old women who had experienced episodes of CSA, and 17 controls have been interviewed using the Adult Attachment Interview. Sixty-three percent of abused women presented an unresolved trauma (12% for the controls). Thirty-six women (14 controls and 22 abused) came again to the laboratory for a session involving an experimental stress challenge (TSST). Subjects provided repeated appreciations of perceived stress on visual analogue scales and saliva samples were collected to assay cortisol levels. Whereas abused women with unresolved trauma showed the highest levels of perceived stress, they simultaneously presented the most suppressed cortisol reactions (there were significant post hoc differences between "unresolved abused" and controls on the increase of perceived stress and on cortisol recovery after the acute stress). It is suggested that important stressful experiences (such as CSA), especially when they have not been psychologically assimilated, may cause a disconnection, during subsequent mildly stressful circumstances, between the perception of stress and natural defensive body reactions.