Trauma bonds, and breaking free
If you have left, gone back, and cannot explain to anyone — including yourself — why you went back, nothing is wrong with your judgment. Attachment gets stronger, not weaker, when affection and harm come from the same person at unpredictable intervals. That is the finding, it is forty-five years old, and it is the whole of what people mean by a trauma bond.
The largest US study of intimate partner homicide found that separation raises the risk, and raises it most where the partner was highly controlling. If any part of you is afraid of what happens when you go, plan the leaving before you announce it, and talk to someone who does this for a living first. The National Domestic Violence Hotline: 800-799-7233, or text START to 88788, free and 24 hours. Abuse and the divorce case covers what the court can do.
Where the idea comes from
Two Canadian psychologists, Donald Dutton and Susan Painter, named it in 1981. They were trying to answer a question that kept embarrassing the field: why the strength of a woman’s attachment to a violent partner did not fall as the violence rose. Their answer was that two specific conditions, together, produce an unusually strong bond rather than a weak one.
- A power imbalance. One person is, or comes to feel, dominated by the other. The subordinate position itself does work here: it makes the dominant person the source of relief as well as the source of harm.
- Intermittency. The mistreatment comes and goes rather than running continuously. Periods of kindness are not the exception that proves the relationship is fine — in this model they are the mechanism.
They tested it twelve years later on formerly battered women, and the paper is still the citation everyone else is standing on.
Dutton, D.G. & Painter, S., “Traumatic Bonding: The Development of Emotional Attachments in Battered Women and Other Relationships of Intermittent Abuse”, Victimology 6(1–4), 139–155 (1981); and “Emotional Attachments in Abusive Relationships: A Test of Traumatic Bonding Theory”, Violence and Victims 8(2), 105–120 (1993) — abstract record at the US Department of Justice.
Why it holds: the part that is not about love
The engine is intermittency, and it is the same engine that makes a slot machine harder to walk away from than a vending machine. A reward that comes every time is easy to give up when it stops. A reward that comes unpredictably is not, because no single disappointment is evidence that it is over. Lenore Walker described the shape it takes in a violent relationship in 1979, and the description has survived because people recognize it immediately.
Tension
Something is coming. You do not know what, and you begin managing the house, the children and yourself around a mood.
The incident
The thing happens. It does not have to be a hit — a rage, a punishment, a withdrawal of money or speech.
Reconciliation
Remorse, tenderness, promises, sometimes the best treatment you have had in months. This is the phase doing the damage.
Calm
It is genuinely fine for a while, which retroactively makes the last incident look like an aberration rather than a pattern.
Walker, L.E., The Battered Woman, Harper & Row (1979). The four-phase description above is the standard summary of her model; we have not read the 1979 text itself, and say so rather than quote it.
Two consequences follow, and they are the ones that make people think they are losing their minds. The good phases are remembered as the real relationship and the bad phases as the exception, because that is what the sequence teaches. And the person who caused the pain is also the person who ends it, so relief and harm arrive from the same direction until they stop feeling like different things.
What it is, and what it is not
“Trauma bond” is not a diagnosis. No doctor can write it on a form, no insurer pays for it, and it is not a term clinicians use with each other. It is a useful description of a real pattern, and it is worth knowing that it carries no clinical authority, because people are sometimes told they are “just trauma bonded” as a way of not taking them seriously.
Neither is Stockholm syndrome, which is the other term people reach for. It has never appeared in a diagnostic manual, and a 2008 paper in a psychiatric journal asked in its title whether it was a diagnosis or an urban myth.
What is a diagnosis is the thing that prolonged, repeated abuse can actually cause. The World Health Organization’s classification has carried complex post-traumatic stress disorder (6B41) since ICD-11, defined around exposure to prolonged or repeated events from which escape is difficult or impossible — and its criteria include, alongside the ordinary PTSD symptoms, difficulty regulating emotion, a persistently damaged sense of oneself, and difficulty sustaining relationships. If you have wondered why the trouble did not end when the relationship did, that is the box it belongs in, and it is treatable.
WHO, ICD-11 for Mortality and Morbidity Statistics, 6B41 Complex post-traumatic stress disorder. Namnyak, M. et al., “‘Stockholm syndrome’: psychiatric diagnosis or urban myth?”, Acta Psychiatrica Scandinavica (2008), doi 10.1111/j.1600-0447.2007.01112.x.
Two things everyone says that we could not stand behind
Both of these appear on almost every page written about trauma bonding. We went looking for the source of each of them and did not find one. They are here because you will be told them, and you should know what they rest on.
“It takes seven attempts to leave”
We could not trace this figure to any study. Every version of it we followed led to another page repeating it without a citation. The real research on leaving describes it as a process rather than a single event — a series of shifts in how a person thinks about the relationship, punctuated by returns — but it does not produce a number, and a number that cannot be sourced should not be used to tell anyone how many more times they are expected to fail.
The review we did read: Anderson, D.K. & Saunders, D.G., “Leaving an Abusive Partner”, Trauma, Violence, & Abuse 4, 163–191 (2003).
“Oxytocin and cortisol chemically bond you to your abuser”
This is stated everywhere as established neuroscience. The peer-reviewed paper most often cited for it is about something else: how social bonding and the oxytocin system help people recover from trauma in general. We found nothing supporting the specific claim that reconciliation chemistry bonds a victim to an abuser.
It may yet turn out to be true. Today it is a hypothesis wearing the clothes of a finding, and you do not need it: intermittent reinforcement explains the behavior without any chemistry at all.
Coercive control, and where the law has caught up
The pattern underneath a trauma bond usually has a name in the research literature: coercive control — monitoring, isolation, control of money, transport, food, sleep and contact with other people. Evan Stark’s 2007 book made the case that this, rather than any individual assault, is the actual harm, and it has been changing how legislatures write their statutes ever since.
It matters practically for one reason. In a state whose law reaches coercive control, you do not have to have been hit to qualify for an order. In a state whose law does not, a pattern of surveillance, isolation and financial control that has terrified someone for years may not meet the statutory definition of abuse at all.
Every row below was read at the state’s own legislature, judiciary or official court site, or it is marked as not read. We are not repeating the secondary tables that circulate on this subject, because when we checked them against the statutes they were wrong in both directions — six states listed as having this law do not have it, and three that have it were missing.
| State | Does the law reach it | What it does, and where it says so |
|---|---|---|
| California | Yes, by name | A restraining order may be granted for disturbing the peace, which the statute says includes coercive control. Cal. Fam. Code § 6320 |
| Colorado | Yes, by name | A finding of coercive control alone triggers a rebuttable presumption against awarding custody. The only state where it reaches custody this way. Colo. Rev. Stat. § 14-10-103 |
| Connecticut | Yes, by name | Defines domestic violence for the whole of title 46b to include coercive control, so it reaches restraining orders and family matters together — the broadest placement of any state here. Conn. Gen. Stat. § 46b-1(b) |
| Hawaii | Yes, by name | A protective order may be granted on coercive control alone, and it is also part of the criminal abuse offense. Haw. Rev. Stat. § 586-1 |
| Massachusetts | Yes, by name | A 209A restraining order may be granted on a pattern of coercive control with no violence and no threat of violence. Mass. Gen. Laws ch. 209A, § 1 |
| Vermont | Yes, by name | Calls it coercive controlling behavior, and makes it a freestanding second branch of abuse for a relief-from-abuse order. 15 V.S.A. § 1101 |
| Washington | Yes, by name | Defines it, with examples running from isolation and financial exploitation to vexatious litigation. Wash. Rev. Code § 7.105.010 |
| Arkansas | Yes, other words | Calls it a course of control, defined almost identically, and it is enough on its own for an order of protection. Ark. Code Ann. § 9-15-219 |
| Delaware | Yes, other words | Reaches financial control, alarming or distressing conduct, and coercion through a companion animal. 10 Del. C. § 1041 |
| Illinois | Yes, other words | Reaches it as interference with personal liberty and as harassment, both of which stand alone. 750 ILCS 60/103 |
| Maine | Yes, other words | Compelling someone by coercion or intimidation is abuse. The word coercion was added in 2025, after a wider coercive-control draft was narrowed. Me. Rev. Stat. tit. 19-A, § 4102 |
| Nebraska | Yes, other words | Not in the protection-order statute but in the Parenting Act: a pattern of intimidation, isolation or economic abuse counts in a custody decision. Neb. Rev. Stat. § 43-2922 |
| Mississippi | Only in part | Uses the phrase, but only in the statute that makes what you tell an advocate confidential — not in the test for getting an order. Miss. Code Ann. § 93-21-125 |
| Idaho | No | Idaho Code § 39-6303 |
| Iowa | No | Iowa Code § 236.2 — a bill is introduced, not law |
| Kansas | No | Kan. Stat. Ann. § 60-3102 |
| Kentucky | No | KRS 403.720 / KRS 456.010 — 2025 HB 96 would have added “coercive control” definitions here but did not pass — a 2025 bill died after its second reading |
| Louisiana | No | La. Rev. Stat. § 46:2132 — the cited section is the elder-abuse statute |
| Maryland | No | None enacted; pending HB 1586 — a 2026 bill died in committee |
| Minnesota | No | None enacted; pending HF 3946 / SF companion — a bill is pending |
| Missouri | No | Mo. Rev. Stat. § 455.010 — coercion there means force or threat of force |
| Montana | No | Mont. Code Ann. § 45-5-203 — the cited section is criminal intimidation |
| Nevada | No | Nev. Rev. Stat. § 33.018 — coercion there means force or threat of force |
| New Hampshire | No | N.H. Rev. Stat. Ann. § 173-B:1 — abuse is an enumerated list of violent acts |
| North Carolina | No | N.C. Gen. Stat. Chapter 50B |
| North Dakota | No | N.D. Cent. Code § 14-07.1-01 |
| Ohio | No | Ohio Rev. Code § 3113.31 |
| Oregon | No | Or. Rev. Stat. § 107.705 — confirmed absent by the judiciary’s own benchbook |
| Pennsylvania | No | 23 Pa. Cons. Stat. § 6102 |
| Rhode Island | No | R.I. Gen. Laws § 12-29-2 |
| South Carolina | No | S.C. Code § 20-4-20 — a bill is pending |
| Texas | No | Tex. Fam. Code § 71.004 |
| Utah | No | Utah Code § 78B-7-102 |
| Virginia | No | Va. Code § 16.1-228 |
| West Virginia | No | W. Va. Code § 48-27-202 — a 2026 bill died in committee |
| Wisconsin | No | Wis. Stat. § 813.12 — the custody statute uses the word coercive without defining it |
| Alabama | Not confirmed | Not read at an official source |
| Alaska | Not confirmed | Not read at an official source |
| Arizona | Not confirmed | Not read at an official source |
| Florida | Not confirmed | Not read at an official source |
| Georgia | Not confirmed | Not read at an official source |
| Indiana | Not confirmed | Not read at an official source — the statute site is an application that renders nothing to a reader |
| Michigan | Not confirmed | Not read at an official source — no legislature host would answer |
| New Jersey | Not confirmed | Not read at an official source — a bill exists; whether it passed could not be confirmed |
| New Mexico | Not confirmed | Not read at an official source — the official code database is an application that renders nothing |
| New York | Not confirmed | Not read at an official source — a bill uses the phrase; it is not law, and the statute itself was not read |
| Oklahoma | Not confirmed | Not read at an official source — the official PDFs are too large to read and the courts site refuses readers |
| South Dakota | Not confirmed | Not read at an official source — the statute site is an application that renders nothing |
| Tennessee | Not confirmed | Not read at an official source |
| Wyoming | Not confirmed | Not read at an official source |
The five things in that table worth knowing
- Connecticut and Colorado put it somewhere different from everyone else. Connecticut’s definition sits in § 46b-1, which defines domestic violence for the whole of title 46b rather than for one remedy, so it reaches restraining orders and family-court matters together. And Colorado did something no other state has done. Everywhere else, coercive control gets you a protection order. Colorado’s HB26-1309, in force since 27 May 2026, put it in the custody statutes instead: a finding of coercive control alone triggers a rebuttable presumption against giving that parent parental responsibilities, with no physical violence and no criminal act required.
- The list is growing, and quickly. Vermont in 2023, Massachusetts in 2024, Colorado in 2026. Bills were live this year in Iowa, Minnesota, South Carolina and New Jersey, and died this year in Kentucky, Maryland and West Virginia. If your state is a no, check again.
- The words matter less than you would think. Illinois has no such phrase anywhere in its Domestic Violence Act and gets to the same place through interference with personal liberty. Arkansas calls it a course of control. Maine added the single word coercion in 2025 after a wider draft was cut down in committee. Ask what conduct the statute reaches, not whether it uses the fashionable term.
- A citation in a table is not a law. Montana appears on the standard matrix with a section number that turns out to be the criminal intimidation statute. Oregon appears on it too, and the Oregon judiciary’s own benchbook says the section does not mention coercive control. Nevada, New Hampshire and Missouri all define coercion as force or the threat of force, which is the opposite of the point.
- Fourteen states are blank because we could not read them today, not because the answer is no. Several of them publish their statutes only through a viewer that will not answer a reader at all. Connecticut was the fifteenth until today: its own server presents a certificate chain no automated reader can get past, so it was read by hand in a browser instead, which is why it is now in the table. We will not print a definition we have not read.
If your state is a no, or a blank, that is not the end of it. Ask the court’s self-help office, or your state coalition against domestic violence, what the standard actually is where you live — and read the definition of abuse in your own protection-order statute, because words like harassment, intimidation and interference with personal liberty are doing this work in several states that never use the term. How protective orders work.
Stark, E., Coercive Control: How Men Entrap Women in Personal Life, Oxford University Press (2007). Fifty-state statutory review carried out on 5 September 2026; every entry marked yes or no was read at the state’s own legislature, judiciary or official court site.
Is this a trauma bond, or just a bad breakup that hurts?
Worth being straight with you: no clinician has published a test for this, because it is not a clinical term. Anyone offering you a ten-question quiz made it up. What the research does support is that the features below are what separate a coercive relationship from an ordinary unhappy one — and none of them is about how much it hurts, because ordinary breakups hurt enormously.
- The relief comes from the same person as the harm. In an ordinary bad relationship, comfort comes from elsewhere: friends, family, a sibling. Here the only reliable source of relief is the person causing the injury.
- Your account of your own life has been argued with. Not disagreed with — corrected. You find yourself checking your memory of events against theirs and losing.
- The reach extends past the relationship. Money, transport, phone, friendships, work, sleep, what you eat. A bad marriage is unhappy; a controlling one is administered.
- Leaving frightens you in a specific rather than a general way. Ordinary breakups frighten people about the future. This is fear of a person.
- It did not end when the relationship did. Difficulty regulating emotion, a damaged sense of who you are, and trouble being close to anyone are the criteria for complex PTSD, not the aftermath of a normal separation.
Breaking free
Two things are true at once, and the honest version has to hold both. Leaving is a process and going back is part of it for most people, so a return is not a failure and not the end of the attempt. And the period around leaving is the most dangerous period, so the process needs planning rather than a moment of courage.
- Write it down while it is happening. Not for a lawyer — for you. The single most effective counter to being talked out of your own memory is a dated, boring, contemporaneous record. Keep it somewhere the other person cannot reach.
- Rebuild one outside relationship before you need it. Isolation is not an accident of these relationships; it is the load-bearing wall. One person who knows the truth changes what is possible.
- Get money in your own name and know what exists. Credit in your own name and know what you own are the two pages that matter most here, and they are worth doing quietly and early.
- Plan the leaving with someone who does it professionally. A hotline advocate or a local program will build a safety plan with you in one conversation, free, and will know what your local court actually grants. That is a different skill from legal advice and you want both.
- Expect the withdrawal, and name it when it arrives. The pull to go back tends to peak after the danger has passed, not during it, and it will arrive dressed as a reasonable thought — that you overreacted, that they have changed, that it was not that bad. It is the intermittency finishing its work.
- Ask for the treatments that have trials behind them. The US Department of Veterans Affairs’ National Center for PTSD names cognitive processing therapy, prolonged exposure and EMDR as evidence-based for PTSD including in survivors of partner violence, alongside two protocols written specifically for this population — cognitive trauma therapy for battered women, and HOPE. Say those names out loud to a therapist and you will find out quickly whether they do this work.
Campbell, J.C. et al., “Risk Factors for Femicide in Abusive Relationships: Results From a Multisite Case Control Study”, American Journal of Public Health 93(7), 1089–1097 (2003) — the source for separation raising risk, and raising it furthest where the partner was highly controlling. Treatment list from the VA National Center for PTSD.
National Domestic Violence Hotline — 800-799-7233, text START to 88788, or chat at thehotline.org. Free, confidential, 24 hours, and you do not have to have decided anything to call. If you are in immediate danger, 911.
Keep reading
- Abuse and the divorce case — what a court can actually do, and how it interacts with the divorce
- Protective orders: how they work — the mechanics, and what to bring
- Moving out, and the house — and why leaving does not cost you your share
- Digital safety — phones, location sharing and shared accounts
- Therapy, and how to find it
- If you cannot afford a lawyer