How this works, and what it cannot do
This page explains exactly how the score is built, where the items come from, why they are weighted the way they are, and where the method is weak. If you came here because you are deciding whether to trust this thing, that is the right instinct. Read this first.
What this is
RedInside is a structured behavioral checklist. You mark the behaviors you have personally observed, and each marked behavior contributes a fixed number of points to a total. That total maps to one of four levels.
That is the entire mechanism. There is no model, no algorithm learning from your answers, and nothing analyzing your writing. The same answers always produce the same result, and you could reproduce the arithmetic by hand.
This is a weighted checklist informed by established frameworks. It is not a validated instrument.
It has not been through a validation study. It has no published reliability figures, no norming sample, and no outcome data behind it. The frameworks it draws on have that kind of evidence base. This checklist does not inherit it, and we are not going to imply that it does.
Where the items come from
Three bodies of work shape this checklist. None of them were built for it, and none of their authors have any connection to it.
The Duluth Power and Control Wheel
Developed in the mid 1980s by the Domestic Abuse Intervention Project in Duluth, Minnesota, from focus groups with women who had experienced battering. It is the most widely used description of how control operates inside an abusive relationship, and it organizes the tactics into eight categories: intimidation, emotional abuse, isolation, minimizing and denying and blaming, using children, using male privilege, economic abuse, and coercion and threats.
Two of our seven areas map onto it directly. Control, intimidation and safety covers the wheel's intimidation, isolation, and coercion and threats segments. Accountability and conflict response is essentially the wheel's minimizing, denying and blaming segment.
We should also say what we left out. This checklist does not cover three of the eight segments: economic abuse, using children, and male privilege. That is a real gap, not an oversight of the framework, and it is the first thing we would add.
The Duluth model is also not universally accepted. It has been criticized, particularly over how well its intervention programs work and over how it handles violence that is not one directional. We are citing it as the best available map of controlling behavior, which is what it is, not as settled science.
Coercive control as a legal concept
The idea that a pattern of control is itself the harm, rather than a series of separate incidents, is the organizing idea behind this whole checklist. It is most associated with the sociologist Evan Stark and his book Coercive Control: How Men Entrap Women in Personal Life.
It is now recognized in law, and the shape of that recognition differs sharply on either side of the Atlantic. In England, Wales, Scotland and Ireland, coercive control is a criminal offense. A pattern of controlling behavior is a crime in itself, separate from any assault inside it.
In the United States it is overwhelmingly civil, working through protective orders and family law rather than the criminal code. California, Connecticut, Colorado and Washington have all brought coercive control into that framework, along with roughly ten more states and the District of Columbia. Hawaii is the single exception, having placed it in its criminal code.
We are being precise about this because it is easy to get wrong in a way that flatters the argument. In the US, for most women reading this, coercive control affects what a court can grant you. It is not something a prosecutor can generally charge.
Sources for the US picture: the Battered Women's Justice Project maintains a codification matrix tracking which states have acted and how, and the American Bar Association's Family Law Quarterly covers the same ground.
The point for this tool is simpler than any of that. Legislatures looked at patterns of control and concluded the pattern is the harm. That is why control weighs more here than lying does.
Campbell's Danger Assessment
Developed by Jacquelyn Campbell of the Johns Hopkins School of Nursing, the Danger Assessment is a validated instrument for estimating risk of severe or fatal violence in abusive relationships. It is the source of the single most important thing on this site.
Its research found that separation is a period of elevated danger, particularly where a partner is highly controlling. That finding is why the safety note on every result says that leaving is the highest risk window and to plan with support rather than confrontation. We did not arrive at that ourselves and we are not the authority for it.
The Danger Assessment is the kind of instrument this is not. It was built for risk assessment, it has been through validation, and it is administered in clinical and advocacy settings. If you are assessing physical danger, that is the tool, not this one.
Why behavior and not feelings
Every item is written so you can answer it from memory of something that happened. Not what he intended, not what he is like, not how it made you feel.
The first reason is that feelings are the thing under attack. If someone has spent months telling you that you are insecure and imagining things, a questionnaire that asks how you feel hands the argument straight back to him. A questionnaire that asks what happened does not.
The second is that behavior is the part that is actually checkable. You either found accounts you were not told about or you did not. Reasonable people can disagree about what a behavior means. They cannot disagree about whether you saw it.
How the weighting works
Behaviors are worth 1, 2, 3, or 4 points depending on which area they sit in. The weight reflects how much a behavior raises concern on its own, not how upsetting it is.
| Area | Points per behavior |
|---|---|
| Control, intimidation and safety | 4 |
| Emotional manipulation | 3 |
| Sexual health and intimacy honesty | 3 |
| Honesty and deception | 2 |
| Secrecy and digital behavior | 2 |
| Accountability and conflict response | 2 |
| Pattern of investment | 1 |
What we do not publish is how many behaviors sit in each area, or what each area can total. The weighting is here so you can argue with it. The rest of the instrument stays private.
These weights are editorial judgment. They were not derived from data.
No study produced the number 4. What produced it is a reading of the frameworks above, which consistently treat control and intimidation as the categories that precede serious harm. A different careful person could weight these differently and would not be wrong. Here is our reasoning, so you can disagree with it specifically:
- Control and intimidation carries the most weight, four times that of the lightest area, because it is the category the risk literature ties to physical danger. This is the part about your safety rather than your feelings.
- Emotional manipulation and sexual health honesty carry 3 points because both do real damage. One erodes your ability to trust your own perception. The other can affect your body without your knowledge or consent.
- Lying and secrecy carry 2 points. Serious, and the most common reason people arrive here, but on their own less tied to danger than control is.
- Pattern of investment carries 1 point. A man can be a disappointing partner without being a dangerous one, and the scoring should not confuse the two.
The Critical Item Rule
Six items, all in the control and intimidation area, are marked critical. Check any one of them and the result is raised to at least High regardless of the total. Check two or more and it is Severe.
This exists because a purely additive score can hide the thing that matters most. Someone whose partner monitors her phone and isolates her from her family, but who scores low everywhere else, would land in Moderate on points alone. That would be a failure of the tool. Safety is not something a total can average away.
The rule only ever raises a result. It never lowers one.
The levels
Bands are applied to the raw total, then the Critical Item Rule is applied on top.
| Level | Raw total |
|---|---|
| Low | 0 to 15 |
| Moderate | 16 to 35 |
| High | 36 to 60 |
| Severe | 61 and above |
The levels do not predict anything.
A level describes how much of this pattern you reported. It is not a probability, a forecast, or a risk score. Nobody can honestly tell you that a total of 44 means a specific chance of a specific outcome, because the data that would support that claim does not exist for this checklist. Where the boundaries sit is a judgment about when a pattern has become dense enough to name, nothing more.
If you want an instrument that estimates risk, that is the Danger Assessment, and it should be done with an advocate.
What this deliberately does not do
- It does not identify anyone.
- No name is required. The optional nickname exists so your own saved results are readable to you. Nothing about him is stored as an identity.
- It does not determine sexuality.
- A lot of women arrive asking whether a partner is secretly living a double life. We understand why, and it is a fair question. But no behavioral checklist can determine anyone’s sexual orientation, and one that claimed to would be doing something both dishonest and harmful. Secrecy, deception, and unprotected sexual risk are scored here because they endanger you, whatever is behind them. That is the honest answer and it is also the more useful one.
- It does not diagnose.
- Not him, not you. Terms like narcissist and sociopath do not appear in the items on purpose.
- It does not detect lying.
- It scores lies you already caught.
- It does not tell you what to do.
- No result recommends leaving, staying, or confronting anyone. The safety information appears at every level because it should be available, not because a score decided something about your life.
Where this method is weak
Every measurement approach has failure modes. Here are ours.
- It is one person’s account.
- This measures what you report. That is the correct input for a tool meant to help you think, and it is also a real limitation. It is not evidence and would not be evidence in any proceeding.
- The weights are judgment, not measurement.
- Stated above and repeated here because it is the most important limitation on the page.
- It covers less than the frameworks it draws on.
- Economic abuse, using children, and male privilege are all in the Duluth wheel and none of them are scored here.
- Context is flattened.
- A checkbox cannot tell the difference between one bad month during a bereavement and four years of the same behavior. Frequency and duration are not scored, which is the single largest gap in the method.
- Recall is uneven.
- Distress changes what you remember and how you weight it. Someone in a bad week may check more items than the same person would a month later. This is why saved history exists: the same assessment taken twice, months apart, tells you more than either alone.
- A low score does not mean everything is fine.
- The list is not exhaustive. People get hurt in ways no checklist covers. If your score is Low and something still feels wrong, this tool has told you nothing about your situation, and you should believe yourself over it.
- It is written for one situation.
- A woman assessing a male partner, where the concerning behavior is secrecy and control. Other relationships and configurations are not what these items were chosen for.
Who this is not for
If you are in immediate danger, this is not the right tool and not the right moment. National Domestic Violence Hotline, 1-800-799-7233 · thehotline.org
If you are looking for something to hand a partner as proof, this will not do that, and using a score that way is likely to make an unsafe situation worse.
If you are asking these questions about your own behavior, that is a real question, but this instrument was not built to answer it.
How this gets corrected
The rubric is versioned. The current version is v1, and every saved assessment records the version it was scored under, so a change to the items never silently rewrites an old result.
We expect to be wrong about some of these weights. The likeliest changes are the frequency and duration gap, the missing Duluth categories, and the balance between the deception areas and the control area. When items or weights change, the version number changes with them and this page is updated in the same release.
If you work in this field and think something here is wrong, we would rather hear it than not. There is a contact form, and it stores nothing.