How Long Will My Relationship Last? What the Research Can (and Can’t) Predict

By the Kismatri team · November 4, 2026 · Sources

Nobody can tell you exactly how long your relationship will last, and anyone who gives you a date is guessing. What research can do is tell you which patterns tend to come before a breakup: contempt during fights, a fading sense that your partner is committed, and a slow drop in warmth. Those signals shift the odds. They don’t sign the papers.

So if you typed “how long will my relationship last” into a search bar at 1 a.m., here is the honest version: the base rates, the studies that actually predicted breakups, and what they can and can’t say about the two of you.

Start with the base rates

Before you look at your own relationship, it helps to know what “normal” looks like across millions of couples. The cleanest long-term numbers come from marriage, because the government tracks it.

The CDC’s National Center for Health Statistics reports a provisional 2023 divorce rate of 2.4 per 1,000 people, across the 45 states (plus D.C.) that report it, and a marriage rate of 6.1 per 1,000. That’s a snapshot of one year, so it can’t tell you the odds for any one marriage.

For odds, you need survival data. A 2012 NCHS report using the National Survey of Family Growth (2006 to 2010) estimated these chances that a first marriage lasts:

How long	Women	Men
At least 10 years	68%	70%
At least 20 years	52%	56%

The authors are careful to say these figures don’t describe the chance of a marriage lasting a lifetime. They also found differences by history: women who hadn’t lived with their husband before marrying had a 57% chance of reaching 20 years, compared with 46% for women who had cohabited while engaged and 45% for those who had cohabited without being engaged. That’s a correlation inside a big survey, so read it as a pattern across groups and leave your own move-in date out of it.

And if you’re not married?

Dating relationships end faster, which makes sense. Many of them are supposed to be trial runs. In a 2010 study of 1,184 unmarried adults aged 18 to 35, Galena Rhoades, Scott Stanley, and Howard Markman found that 26.3% of the couples had broken up within eight months. Roughly one in four, in less than a year.

There’s no single government figure for “the average relationship,” because nobody registers a first date with the county. If you want the longer breakdown of averages by relationship type, that’s what our post on how long the average relationship lasts is for.

The Gottman research: how couples fight

The most famous attempt to predict breakups came out of John Gottman’s lab. In a 14-year study published in 2000, Gottman and Robert Levenson followed 79 couples. By the end, 22 of them (27.8%) had divorced.

The interesting part is that early and late divorces looked different. Negative emotion during conflict predicted couples who divorced early. A lack of positive emotion, in both fights and ordinary “how was your day” conversations, predicted couples who divorced later. As the University of Washington described it, the early group was volatile and the later group was emotionally cool, with feelings pushed down.

A model built from marital satisfaction, thoughts about ending the marriage, and emotion in those conversations predicted divorce with 93% accuracy. That number gets quoted everywhere, usually without the fine print. It came from 79 couples, and the model was tested on the same couples it was built from, with no fresh group to check it against. Strong evidence that these patterns matter. Weak evidence that anyone can look at your couple and call it.

The four horsemen, and why contempt is the one to fear

The same paper used four negative behaviors Gottman calls the “Four Horsemen”: criticism, defensiveness, contempt, and stonewalling. Among the couples who divorced, those four codes from the conflict conversation sorted early divorcers from later ones with 95% accuracy, though the authors note the analysis had low statistical power because it only included the divorcing couples.

Here’s how The Gottman Institute defines them:

Criticism: attacking your partner’s character instead of complaining about a specific behavior. “You never think about anyone but yourself,” instead of “I was upset you didn’t call.”
Contempt: mockery, eye-rolling, name-calling, sarcasm from a position of superiority.
Defensiveness: excuses, playing the victim, flipping the blame back.
Stonewalling: shutting down and going silent, often because you’re physically flooded.

The Institute calls contempt “the single greatest predictor of divorce.” That’s a strong claim from the people who built the framework, so hold it with some care. Still, it matches what most of us know from life. A fight about the dishes is survivable. A fight about the dishes where one of you sneers “of course you forgot, you always do” is a different animal. Contempt says the problem is who you are. We go deeper on that in our post on contempt in relationships.

The good news is that each horseman has a known counter. The Institute’s antidotes are a gentle start-up for criticism, a culture of appreciation and respect for contempt, taking responsibility for defensiveness, and physiological self-soothing (a real break, at least twenty minutes) for stonewalling. None of them require a personality transplant.

The machine-learning study that humbled everyone

If any single paper answers “can my relationship be predicted,” it’s this one. In 2020, Samantha Joel, Paul Eastwick, and dozens of co-authors pooled 43 longitudinal datasets from 29 labs, covering 11,196 couples, published in PNAS. They used a machine-learning method called Random Forests to test which of hundreds of measured variables actually predicted relationship quality.

The top relationship-specific predictors were:

Perceived-partner commitment (how committed you think your partner is)
Appreciation
Sexual satisfaction
Perceived-partner satisfaction (how happy you think your partner is)
Conflict

The top individual traits were life satisfaction, negative mood, depression, and attachment avoidance and anxiety.

Now the humbling part. Your own views of the relationship explained up to 45% of the variance in how satisfied you were at the start of each study, but only up to 18% by the end. Your partner’s reports didn’t add anything beyond yours. Neither did either person’s personality. And the line that matters most for anyone wanting a forecast: change in relationship quality over time, whether it got better or worse, was “largely unpredictable from any combination of self-report variables.”

Read that again if you’re about to take a quiz. Eleven thousand couples, every questionnaire the field had, and the researchers still couldn’t say whose relationship would improve and whose would slide.

What that means for you

Notice what made the list. Almost everything at the top is something you feel about the other person right now. Do you think they’re in this? Do you feel grateful for them? Is the sex good? When you fight, how bad does it get? These are things you already know better than any test does.

Also notice what fell away. Once the researchers knew how people saw their own relationship, personality traits added nothing, and neither did the partner’s side of the story. The things that predicted quality were the ones that live between two people, as seen by the one answering.

Rusbult’s investment model: why people stay (or don’t)

Gottman studied how couples fight. Caryl Rusbult studied why people stay. Her investment model, as Rhoades and colleagues summarize it, says commitment rests on three things: how satisfied you are, how good your alternatives look, and how much you’ve invested (time, shared friends, a lease, a dog).

In her 1983 longitudinal study, Rusbult followed 34 college students in new dating relationships for seven months, with a questionnaire every 17 days. About 29% of the relationships ended. For the people who stayed, rewards rose, satisfaction grew, alternatives looked worse, investments grew, and commitment climbed. For the leavers, the reverse.

The finding that sticks with me is the one about people whose partners left them. Rusbult described them as showing “entrapment”: their satisfaction wasn’t growing much, but their alternatives looked worse and they kept investing heavily anyway. If that sounds like the friend who keeps planning trips with someone who’s half out the door, it should.

The sample was small, so treat it as a classic starting point. The Rhoades study, with more than a thousand people, found that dedication and several kinds of constraint each predicted who stayed together over eight months. Together, though, they explained only about 16% of the variance. Commitment matters. It still leaves most of the story unwritten.

So, how long will yours last?

Put the research together and you get something more useful than a date. You get a short list of signals that tilt the odds:

Contempt is showing up. Eye-rolls, mocking, “you’re pathetic” energy during fights.
You doubt their commitment. In the Joel study, how committed you believe your partner is ranked at the top.
Appreciation has dried up. You can’t remember the last time either of you said thanks for something small.
The warmth went quiet. Fewer fights, but also fewer laughs. Gottman’s later divorcers looked like this.
You’re staying for the lease. Investments are high, satisfaction is low, and the alternative feels scary.

If several of these fit, that’s worth a real conversation, maybe with a couples therapist. Gottman himself said both early and later divorcing types “can benefit from marital therapy and don’t necessarily have to end.” If none fit, you’re probably overthinking a text left on read. For the full warning list, see signs your relationship is ending.

Are “how long will my relationship last” quizzes accurate?

No quiz can know. The best research tools available, given to thousands of couples, couldn’t predict who would improve and who would decline. A ten-question quiz built for clicks won’t do better.

That doesn’t make quizzes useless. A good one can get two people talking about things they’d usually avoid, like whether they feel appreciated or what they’d do if this ended. Take the result as a prompt for a conversation, then have the conversation. The number at the end is a party trick.

Curious where you two land anyway? Kismatri is a $2.99 chemistry test you take with someone you already know, and its forecast of how long you’d last is for entertainment. Try it at kismatri.com, laugh at the result, then go talk to the actual person.

FAQ

Can you predict how long a relationship will last?

Not for one specific couple. Research can identify patterns linked to breakups, like contempt and low perceived commitment, but a 2020 study of 11,196 couples found that changes in relationship quality over time were largely unpredictable from self-report questionnaires.

How long does the average relationship last?

There’s no single official number for dating relationships. For first marriages in the U.S., NCHS estimated a 68% (women) to 70% (men) chance of lasting at least 10 years, and 52% to 56% for 20 years. In one study of unmarried adults, about a quarter of couples broke up within eight months.

What is the biggest predictor of a breakup?

The Gottman Institute names contempt as the single greatest predictor of divorce. In the large 2020 machine-learning study, the strongest predictor of relationship quality was how committed people believed their partner to be.

Are relationship length quizzes accurate?

No. They can’t account for what researchers themselves couldn’t predict. Use them as a fun way to start a conversation with your partner, never as a verdict.

Can a relationship with warning signs still last?

Yes. Each of Gottman’s four horsemen has a learnable antidote, such as gentle start-ups and building appreciation, and Gottman has said troubled couples can benefit from therapy rather than having to end.