Why people stay in toxic relationships has an empirical answer, and it is not about weakness. Caryl Rusbult and John Martz interviewed women at a shelter for battered women and found that commitment tracked three things: how satisfying the relationship was, how much had been invested in it, and — decisively — what economic alternatives existed.
“Just leave” is advice about the first one. Staying is mostly decided by the other two.
The short version
- Commitment runs on satisfaction, alternatives and investment. Not on willpower.
- In shelter data, poorer economic alternatives predicted greater commitment.
- Commitment predicted who went back. It is the variable doing the work.
- “Just leave” is an intervention on one input out of three.
The Model Has Three Inputs
Rusbult’s investment model holds that commitment to a relationship is not the same thing as liking it. Commitment is a function of three separate variables: satisfaction with the relationship, the quality of the alternatives available, and the size of what has already been sunk into it.
That distinction sounds academic until you apply it. Two of the three have nothing to do with the partner. A person can be miserable, know it, and still be highly committed — provided the alternatives are poor and the investment is large. The model predicts precisely that.
It is not a theory about love. It is a theory about the arithmetic of leaving.
What the Shelter Data Showed
Rusbult and Martz tested it where it matters. Writing in Personality and Social Psychology Bulletin in 1995, they analysed intake interviews at a shelter for battered women.
Commitment was greater among women who had poorer-quality economic alternatives, who were more heavily invested — those who were married, for instance — and who reported lesser dissatisfaction, including less severe abuse.
Then the finding that carries the argument. Commitment was strongly linked to stay-or-leave behaviour: it significantly distinguished the women who returned to their partners immediately on leaving the shelter from the women who did not. Their causal modelling suggests commitment is the channel through which alternatives, investments and satisfaction all do their work.
Read the first variable again. Economic alternatives. Whether a woman went back was predicted, in part, by whether she had anywhere else to go.
Why “Just Leave” Fails as Advice
Put the sentence everybody says next to the model and it stops looking helpful.
“Just leave” is an intervention on satisfaction. It says: notice that this is bad. But in the shelter data, dissatisfaction was the variable already working — women reporting more severe abuse were less committed, not more. That message had landed long before anybody said it out loud.
What the sentence cannot touch is the other two inputs. It does not conjure an alternative, and it cannot un-invest eight years. Which is why repeating it louder changes nothing, and why the people repeating it run out of patience: they are pushing on a variable that was never the bottleneck.
What Stacks Inside the Model
The investment model gives the structure. Several documented mechanisms sit inside it.
Investment is not only money and years. It is sunk cost, and prior investment is exactly what makes the next step look reasonable — the entrapment problem. When that investment cannot be justified from the outside, the mind manufactures a justification from the inside, which is cognitive dissonance taking over.
Alternatives are not only economic either. Coercive control, in Evan Stark’s framing, is a liberty crime: it works by removing options, which is the same thing as degrading alternatives. That is the machinery of emotional control dynamics.
And satisfaction is not a steady reading. Delivered on a variable schedule it spikes unpredictably, which is intermittent reinforcement — and the wanting that produces is not the same thing as enjoying anything, because dopamine drives pursuit rather than pleasure.
What the Research Does Not Say
The limits are real. Rusbult and Martz studied women at one shelter, the work dates from 1995, and it measures reported commitment and immediate return rather than long-run outcomes. An investment model describes a decision structure; it is not a moral account of anybody in it.
Nor does any of it suggest the abuse was tolerable, or that staying was consent. Nonvoluntary dependence is the phrase in their title, and it is doing real work: the model describes constraint, not preference.
What it does establish is where the leverage sits. If commitment is driven by alternatives and investment, then anything that improves the alternatives — money, housing, work, one person willing to take someone in — moves the variable that actually predicts leaving. That is a different conversation from the one about why she does not simply go, and it sits alongside the cycle of abuse, which describes the timing rather than the arithmetic.
If this describes your situation, the National Domestic Violence Hotline is available 24/7 on 1-800-799-7233.
Frequently Asked Questions
Why do people stay in toxic relationships?
Because commitment is not the same as satisfaction. Rusbult’s investment model holds that it is driven by satisfaction, the quality of available alternatives, and accumulated investment. In shelter data, poor economic alternatives and heavy investment predicted greater commitment — and commitment predicted who returned.
Does staying mean the abuse was not that bad?
No. Rusbult and Martz titled their paper “nonvoluntary dependence” for a reason. The model describes constraint rather than choice, and women with the poorest economic alternatives were the most committed regardless of what they were enduring.
Why doesn’t telling someone to leave work?
Because it argues with the one input that is usually already settled. In the shelter data, women reporting more severe abuse were less committed, not more — they knew. What the advice cannot supply is an alternative or a refund on years already invested.
Sources
- Rusbult, C. E. & Martz, J. M. (1995). Remaining in an abusive relationship: An investment model analysis of nonvoluntary dependence. Personality and Social Psychology Bulletin, 21(6), 558–571. Retrieved 16 July 2026.
- Rusbult, C. E., Martz, J. M. & Agnew, C. R. (1998). The Investment Model Scale: Measuring commitment level, satisfaction level, quality of alternatives, and investment size. Personal Relationships, 5(4), 357–387. Retrieved 16 July 2026.
- Stark, E. (2007). Coercive Control: How Men Entrap Women in Personal Life. Oxford University Press. Retrieved 16 July 2026.
