In 1981 Donald Dutton and Susan Painter proposed an explanation for something that everyone working with abused women had observed and nobody could account for: attachment to the abuser was often not weakened by the abuse, and was sometimes strengthened by it. They named the phenomenon traumatic bonding and argued that it required exactly two conditions, neither of which is severity.
The first is a marked power imbalance. The second is intermittency — periods of abuse alternating with periods of warmth, relief, or apparent remorse.
Their 1993 study of battered women tested this directly and found that the strength of attachment to the abuser was predicted by the power differential in the relationship and by the intermittency of the abuse, rather than by how severe the abuse had been. Women in relationships with worse violence were not necessarily more attached. Women in relationships with more unpredictable violence were.
The mechanism follows from the previous chapter with one addition. Fear produces physiological arousal. When the source of the fear then provides relief — the apology, the tenderness, the return of normality — that relief is powerfully reinforcing precisely because of how bad the preceding state was, and it becomes conditioned to the person who caused both. Bowlby's attachment system compounds this: attachment behavior is activated by threat, and it directs the frightened person toward their attachment figure. When the threat and the attachment figure are the same person, the system runs in a loop it was never designed to resolve. The person's own biology is delivering them back to the source of the danger.
This is the same structure that produces the phenomenon popularly called Stockholm syndrome, a term whose clinical status is genuinely contested — the original 1973 Norrmalmstorg case has been substantially reinterpreted, and it is not a recognized diagnosis. Traumatic bonding is the better-evidenced concept and does not depend on the hostage literature.
Three consequences follow, and they matter for anyone trying to help.
Loyalty intensifies after incidents rather than before them, which is why an outsider who intervenes immediately after an episode often meets the fiercest defense of the abuser. Second, the bond is not a judgment about the relationship and cannot be argued with as though it were. Presenting evidence of mistreatment to a traumatically bonded person is presenting evidence that is not in dispute. Third, and most dangerously, the bond does not dissolve on separation. Withdrawal from an intermittent reinforcement schedule is a real phenomenon with a real course, and the period immediately after leaving is when returning feels most compelling — which coincides with the period that homicide statistics identify as the most dangerous.
Because the mechanism is conditioning rather than reasoning, the countermeasures that work are behavioral and structural rather than argumentative: contact interruption long enough for extinction to begin, a written record made contemporaneously so that memory of the bad periods is not overwritten by the good ones, and other people who function as an external reference for reality. Chapters 311 and 341 through 345 in this guide cover the clinical picture, and they are the ones to read if this describes someone you know.
The single most useful thing an outsider can understand is that just leave is not advice. It is a request to override a conditioned attachment with an argument, and arguments do not reach the system that is holding the person in place.
Defense: The Bond That Hurts: Recognizing and Breaking Traumatic Attachment