Skip to content
RelationshipAdvices
BreakupsModerate evidence

How do you know when to end a relationship?

The short answer

No threshold in the research tells you to leave, and anyone offering one is not describing evidence. What it does establish is more useful than it sounds: people weighing this hold strong reasons on both sides at once, so ambivalence is not a sign you have not thought hard enough; what keeps people in unsatisfying relationships is often investment rather than satisfaction; and most people overestimate how long the distress of a breakup will last.

11 min readPublished Updated

What readers think

When you think about leaving, is the bigger obstacle how you feel or what you would lose?

Anonymous. One vote per reader.

Most people arrive at this question having already asked it many times, which is itself the thing worth explaining. The experience of turning the same decision over for months without it resolving is so common that it is better understood as a normal feature of how these decisions work than as evidence of your own indecision. The research does not produce a verdict. It does produce a fairly clear account of why no verdict arrives.

Ambivalence is the normal state, not a failure to decide

When researchers asked people who were actively considering a breakup to list their reasons for staying and their reasons for going, the striking result was that both lists were long, and that the same person often held strong motivations in both directions simultaneously[1]. Around half of those in relationships reported some desire to leave, and around half of those who had already decided to break up reported real reasons to stay. This matters because of how people interpret their own uncertainty. The common assumption is that if the relationship were genuinely wrong you would know, and that continued doubt is therefore evidence the relationship is fine, or alternatively evidence that you are too weak to act. The data supports neither reading. Wanting to stay and wanting to go are not two ends of one scale that must resolve; they are separate feelings that coexist, and waiting for one to disappear before acting is waiting for something that often does not happen.

What actually keeps people in place

The most durable model in this literature holds that commitment is not simply a product of how happy you are. It is a product of three things: satisfaction with the relationship, the quality of the alternatives you perceive, and the size of what you have invested, meaning shared time, money, friends, housing, children and the identity built around being part of a couple. A meta-analysis across many studies found each of these three independently predicted commitment[2]. The practical consequence is uncomfortable. Investment keeps people in relationships they are not satisfied with, and it does so with the same force as satisfaction keeps people in relationships they are. Someone who cannot understand why they are still here despite being unhappy is usually not being irrational; they are responding to a variable they have not named. Recognising this does not tell you what to do. It does tell you that the difficulty of leaving is not a measure of how good the relationship is.

You will probably cope better than you expect

One of the more robust findings here concerns prediction rather than the relationship itself. When researchers asked people in relationships to forecast how distressed they would be if it ended, and then followed the ones who did break up, the forecasts were systematically too pessimistic[3]. People recovered faster and to a higher level than they had predicted. This is an instance of a general pattern in affective forecasting: humans overestimate both the intensity and the duration of future emotional states, because when imagining the aftermath they picture the loss without also picturing everything else that will still be happening. The relevance is direct. If part of what holds a decision in place is a forecast of unbearable distress, the evidence suggests that forecast is likely to be wrong in a specific and predictable direction. That is not an argument for leaving. It is a correction to one input people weight heavily.

What predicts satisfaction, and what does not

A large machine learning study drawing on data from more than eleven thousand couples examined what actually predicts relationship satisfaction. The findings were consistent across datasets: features specific to the relationship, such as perceived partner commitment, appreciation, sexual satisfaction and conflict, predicted satisfaction reasonably well, while the individual characteristics of the partners predicted very little in comparison[4]. Someone's personality, attractiveness, values or background told the models remarkably little about how satisfying their relationship would be. This is worth knowing when you are trying to evaluate a partner as though the question were whether they are, in some general sense, a good one. The evidence suggests that is close to the wrong question. What carries the signal is how the relationship is going, which is exactly the thing you already have direct access to and do not need a framework to assess.

It depends, on what?

What changes the answer

There is a category of situation this entire framing does not apply to, and it should be separated out rather than folded in. Where there is violence, coercion, or fear, the question is not a weighing of satisfaction against investment, and the research described here was not conducted on those relationships. The deliberative model actively misleads in that context, because it invites someone to keep evaluating a situation that needs a safety plan instead. Separately, it is worth naming that people commonly want the research to authorise a decision they have already made, and that this is not a discreditable motive. Where the material above can help is in removing bad reasons: the belief that certainty should have arrived by now, that difficulty leaving means the relationship is good, or that the aftermath will be as unsurvivable as it currently looks. What remains after those are removed is a judgement, and judgement is not something a study can do for you.

Where people disagree

Both sides of it

Reasonable people land in different places on this. Here is the strongest version of each case, not a straw man of the one we disagree with.

The case for yes

  • Recognising that ambivalence is normal removes the belief that you should have arrived at certainty by now, which is itself a source of paralysis.
  • Naming investment as a separate force from satisfaction lets you see why leaving feels hard without concluding the relationship must therefore be good.
  • The evidence on affective forecasting suggests the aftermath is likely to be more survivable than it currently appears.
  • Relationship-specific experience predicts satisfaction far better than partner traits, so your own direct sense of how it is going is a better instrument than any checklist.

The case for no

  • None of this produces a decision, and people looking for a threshold will find the honest answer unsatisfying.
  • The finding that distress is overestimated could be misread as encouragement to leave, when it only corrects one input among many.
  • The research is largely conducted on dating relationships in Western samples, and the weight of investment differs enormously where children, immigration status, financial dependence or family and community consequences are involved.
  • A framework that describes ambivalence as normal can also become a reason to stay indefinitely in something that is quietly ending.

Evidence

What the research says

How we rate evidence strength
Strong evidence
Multiple independent studies, including replications or meta-analyses, point the same way.
Moderate evidence
Several studies support this, but samples are limited or findings vary by population.
Limited evidence
Early, small, or single-study evidence. Treat this as a reasonable hypothesis, not a settled fact.
Contested
Credible researchers disagree, or the evidence points in conflicting directions.

Comparative view

Different perspectives

Traditions and disciplines answer this differently. We describe what each one teaches. We are not telling you which is right, and no tradition speaks with a single voice.

What the Evidence Supports

The investment model is the best supported item here and has held up across many samples, though almost entirely in correlational designs. The ambivalence finding is robust in the sense of being clearly measured, but comes from a small number of studies on dating populations. The affective forecasting result is consistent with a large general literature on misprediction, which strengthens it, but the specific breakup study is small and drawn from students. The machine learning work is the most methodologically rigorous of the four and also the most easily overstated: it shows partner traits carry little predictive information, not that partners do not matter. None of this evidence bears on what any individual should do.

Therapist Perspective

Therapists working with this question commonly report that clients arrive asking to be told what to do and that the useful work begins when that request is declined. A pattern practitioners describe often is the person who has been deciding for years, where the deciding has itself become the stable state and is more tolerable than either outcome. Discernment counselling was developed specifically for couples where one partner is leaning out and the other is leaning in, on the observation that standard couples therapy assumes a shared goal that this situation lacks. Clinicians also frequently note that people underestimate how much of their uncertainty is about identity and logistics rather than about feeling.

Secular Perspective

A secular framing treats a relationship as an arrangement between consenting adults that either may end, and locates the ethical content not in whether you leave but in how. On this view no one is obliged to remain in a relationship that has stopped working, and the obligations that survive are honesty about where you stand, not leaving someone to discover it by inference, and proportionate care for dependants and for shared commitments already made. It also treats prolonged ambivalence as carrying a cost to the other person, who is living inside a decision they have not been told is being made.

Islamic Perspective

Mainstream Islamic teaching treats marriage as a covenant to be preserved where possible, with an emphasis on reconciliation and on arbitration involving representatives from both families, drawing on the verse in Surah An-Nisa addressing arbitration when a breach is feared. At the same time separation is permitted rather than forbidden, and is described in a well-known hadith as the most disliked of permitted things, a phrase scholars read as discouraging rather than prohibiting. Jurists differ considerably on the grounds and procedures available, particularly those available to a wife, which vary by school and by jurisdiction. Presented as one tradition's reasoning rather than as a standard readers are expected to adopt.

Not a settled view

Schools differ substantially on the mechanisms by which a wife may initiate separation and on what grounds, including khulʿ and judicial dissolution, and contemporary scholars disagree on how far classical procedures should be applied in jurisdictions with civil family law.

Practical steps

What you can actually do

  1. Write both lists rather than one. Reasons to stay and reasons to go are separate and both real, and seeing them side by side is closer to the actual state you are in than trying to reach a single verdict.

  2. Separate the question of whether you want to be here from the question of whether you could survive leaving. They get answered together and they are different questions.

  3. Name what you have invested explicitly, including the things that feel embarrassing to count. Unnamed investment operates on the decision anyway.

  4. Judge the relationship on how it goes rather than on whether your partner is a good person in the abstract, since the first predicts satisfaction and the second barely does.

  5. Consider discernment counselling rather than couples counselling if one of you is leaning out. It is designed for that specific situation and ordinary couples work is not.

Worth unlearning

Common misconceptions

  • That you would know if it were really wrong. Research on people actively considering a breakup finds strong reasons on both sides held at once, in roughly half of those surveyed.

  • That finding it hard to leave means the relationship must be worth keeping. Investment predicts commitment independently of satisfaction, so difficulty leaving is not a measure of quality.

  • That the pain of a breakup will be as bad as you imagine. Forecasts made before a breakup are systematically more pessimistic than what people actually experience.

  • That you should assess whether your partner is a good partner in general. Individual traits predict satisfaction poorly; relationship-specific experience predicts it far better.

  • That therapy will tell you whether to stay. Individual therapy helps clarify what you want, and discernment counselling exists for the leaning-out case, but neither issues a verdict.

The short version

Key takeaways

  • People considering a breakup commonly hold strong reasons to stay and to go at the same time, so ambivalence is not diagnostic.

    Research

  • Commitment is predicted by satisfaction, perceived alternatives and investment size independently, which is why unhappy people stay.

    Research

  • Predictions of post-breakup distress are systematically more pessimistic than the distress people actually experience.

    Research

  • Relationship-specific experience predicts satisfaction substantially better than the individual characteristics of either partner.

    Research

  • Where there is violence or coercion, this deliberative framing does not apply and safety planning takes precedence.

    Clinical consensus

Worth saying

When to get professional help

If there has been violence, threats, or a pattern of control, a domestic abuse service is the right contact rather than a counsellor, and safety planning matters more than deliberation; leaving is also the period of highest risk in abusive relationships, which is a reason to seek specialist support rather than to act suddenly and alone. Individual therapy is generally more useful than couples counselling when you are trying to work out what you want, since a joint session is a poor setting for deciding whether to stay and a good setting for working on a relationship both people intend to keep. Couples counselling is worth considering when both of you want to try, and discernment counselling exists specifically for the case where one wants to leave and the other wants to stay, which ordinary couples work handles badly.

Find support services near you

Still wondering

Related questions people ask

It is common rather than diagnostic. Prolonged ambivalence is the normal shape of this decision for many people, and the research finds strong reasons on both sides held simultaneously. What is worth examining is whether the deciding has become a stable state that is easier than either outcome.

Community

What other people say

These are readers describing their own relationships, not professionals. Take them as experience, not advice, and please do not name anyone.

0/4000

    No one has shared their experience yet. If you have been here, you would be the first, and probably the most useful thing on this page.

    Check our work

    Sources

    1. [1]
      Wanting to stay and wanting to go: Unpacking the content and structure of relationship stay/leave decision processes(opens in a new tab)

      Joel, S., MacDonald, G., & Page-Gould, E. · Social Psychological and Personality Science · 2018

    2. [2]
    3. [3]
      Mispredicting distress following romantic breakup: Revealing the time course of the affective forecasting error(opens in a new tab)

      Eastwick, P. W., Finkel, E. J., Krishnamurti, T., & Loewenstein, G. · Journal of Experimental Social Psychology · 2008

    4. [4]
      Machine learning uncovers the most robust self-report predictors of relationship quality across 43 longitudinal couples studies(opens in a new tab)

      Joel, S., Eastwick, P. W., Allison, C. J., and colleagues · Proceedings of the National Academy of Sciences · 2020

    Researched and written by the RelationshipAdvices Desk, reviewed before publication, and updated as the evidence changes. We are not licensed therapists. How we work.

    breakupscommitmentambivalenceinvestmentdecision making

    People who read this also asked