Infidelity is the subject everyone has an opinion about and almost no one reads the research on. Here is the honest version, built from the strongest recent studies: what predicts it, what does not, why the numbers are shakier than they look, and why the truest account refuses both the monster story and the it-was-the-marriage story.
Almost everything written about cheating is an opinion wearing the costume of a fact. The magazine piece that knows exactly why people stray, the thread that reduces it to one clean cause, the confident voice that says it is always about the marriage, or always about the culture, or always about a certain kind of broken person: each of these offers the same secret comfort, which is a single story simple enough to hold in one hand. The trouble is that the research, when you actually read it, does not hold in one hand. It is messier, humbler, and far more useful than any of the hot takes, and the mess is the point. This essay is an attempt to say plainly what the best current science does and does not know about infidelity, without the flattery and without the ideology, so that you can think about the hardest thing that ever happened to you with better tools than a slogan.
There are two comfortable lies to clear away before anything true can be said. The first is the monster story: that a betrayal happened because your partner was secretly a predator, a defective person, and the wound tells you nothing except that you were unlucky in whom you trusted. The second is the mirror image, the it-was-the-marriage story: that infidelity is simply what happens when a relationship fails to meet someone's needs, that the affair is a symptom and the couple is the patient, and so the betrayed person carries half the responsibility for having built a home the other person wanted to leave. Both stories are popular because both let someone off the hook. The evidence supports neither cleanly. What it supports is a convergence, and learning to see the convergence is the beginning of actually understanding this.
Before we can say how common infidelity is or what causes it, we have to admit a problem the field has only recently faced squarely. In 2024 a large systematic review and meta-analysis in the journal Personal Relationships pulled together 305 studies covering more than half a million people across 47 countries, and its most important finding was almost a confession: the research cannot agree on what the word means. Nearly a third of the studies simply asked people whether they had 'cheated' and left the term undefined, so that one respondent was confessing to intercourse and another to a lingering text thread, and both answers went into the same number. Some studies count only sex. Some count emotional attachment. Some count pornography, or flirting, or a secret second account on a dating app. The behaviors are not the same behavior, and yet they are constantly averaged together and reported as a single rate.
This matters more than it sounds, because it means that almost every confident statistic you have ever heard about cheating rests on sand. When someone tells you that a precise percentage of people are unfaithful, the honest reply is: unfaithful how, measured how, and defined by whom. The definitional chaos is not a footnote to the science. It is the first real thing the science has to teach you, and it is a good inoculation against everyone who speaks about this with more certainty than the data allow.
Within those limits, some patterns do survive across studies. Across the large samples, men report sexual infidelity at higher rates than women, roughly on the order of one in four men versus one in seven women admitting to it, though every one of those figures should be read with the caveats above. The gap shrinks, and in some measures nearly closes, when you move from sexual infidelity to emotional and digital forms, where women's reported rates come much closer to men's. And the gap is narrowing generation by generation. Among younger cohorts the old, wide difference between the sexes is smaller than it was, which quietly undermines the tidy evolutionary just-so story that men are simply built to stray and women to stay. Whatever the deep causes, they are clearly responsive to culture and circumstance, not fixed by biology alone.
The other thing the numbers show is that self-report almost certainly undercounts. People lie about lying; that is nearly a definition of the trait involved. So the true rates are probably higher than the admitted ones, in a direction and by an amount no one can precisely measure. Hold all of this loosely. The value of the prevalence research is not a magic number. It is the shape: common enough that it is not a freak event, unevenly distributed in ways that track real differences between people, and impossible to pin to a decimal point.
If you want the one finding that has held up best, that is closest to a genuine predictor rather than a vague correlate, it is this: the best guide to whether a person will betray a partner is whether they have done it before. In a well-known longitudinal study out of the University of Denver, researchers followed 484 people out of one serious relationship and into the next, and found that those who had cheated in the first relationship were about three times more likely to cheat in the second than those who had not. The old folk saying, once a cheater always a cheater, turns out to be an overstatement of something real. It is not a law; plenty of people betray once and never again. But it is a genuine elevation of risk, large enough to take seriously, and it is one of the cleaner results in a noisy field because it is behavioral rather than self-flattering. It does not ask people to describe their character. It watches what they actually do, twice.
This is worth dwelling on because it cuts directly against the it-was-the-marriage story. If infidelity were purely a symptom of a particular unhappy relationship, then a new partner and a fresh start should reset the odds. The finding says otherwise. Something travels with the person across the boundary between one relationship and the next, something dispositional, and that something is the thread the rest of the research keeps trying to name.
The clearest name the research has for that something is sociosexuality, a slightly clinical word for how comfortable a person is with sex outside of love and commitment. People high in what psychologists call unrestricted sociosexuality are more willing to have sex without emotional closeness, more drawn to novelty and variety, more easily aroused by the idea of a new partner. It is measured with a standard questionnaire, and it predicts infidelity across studies and across both sexes. A 2025 study went further and showed that it is often the mechanism through which other things convert into betrayal: when someone is dissatisfied in a relationship, whether that dissatisfaction actually turns into cheating depends heavily on their sociosexuality. The unhappy partner low in it tends to stay faithful and either repair or leave; the unhappy partner high in it is the one for whom dissatisfaction becomes an affair. The orientation is the amplifier.
This reframes a common confusion. People often ask whether affairs are caused by unhappiness or by character, as though it had to be one or the other. The better answer is that unhappiness supplies the motive and disposition supplies the willingness, and you need both for the act. A happy person high in sociosexuality may never stray because there is no pull; an unhappy person low in it may ache and still never cross the line. It is the combination that predicts, which is exactly why single-cause stories keep failing.
Sociosexuality is about desire. The other dispositional thread is about honesty, and here the newer personality science is sharper than the familiar Big Five. The HEXACO model of personality adds a sixth factor that the Big Five leaves out, called Honesty-Humility: the disposition toward sincerity, fairness, and an unwillingness to exploit others even when you could get away with it. Low Honesty-Humility is the personality signal most specifically tied to betrayal of all kinds, and a 2024 experimental study found that its power to predict cheating grows stronger the more certain a person is that they will not be caught. That is a chilling and precise result. It says the trait is not really about temptation; it is about what a person does with the freedom that secrecy provides. Give a low Honesty-Humility person the certainty of impunity and the behavior follows.
Alongside it sit the darker traits. Across the research on the Dark Triad, narcissism, Machiavellianism, and psychopathy, it is psychopathy, the callous, impulsive, low-empathy strand, that most consistently predicts infidelity, because it combines the appetite for novelty with an indifference to the partner's pain. And in ordinary Big Five terms, careful dyadic studies that measured both members of a couple found that low Agreeableness and low Conscientiousness, in either partner, predicted a person's own straying: the disagreeable are more willing to hurt, and the unconscientious less able to restrain an impulse they have already decided is wrong. None of these traits is destiny, and the effect sizes are modest, but together they describe a real dispositional terrain. Some people carry more risk into every relationship they enter, not because they are monsters, but because the ordinary dials of their character sit where they sit. This is the grain of truth the monster story distorts. The distortion is in imagining a separate species; the truth is a distribution, and everyone is somewhere on it.
So far this could read as though betrayal were entirely a matter of who a person is. It is not, and the most sophisticated study in the field is the one that proves it. A team used interpretable machine learning, an approach that lets the data rank its own predictors rather than having the researcher decide in advance, on two large samples, and asked the blunt question: of everything we can measure, what actually forecasts infidelity? The answer that came back was humbling for everyone with a pet theory. There was no single dominant cause. And the strongest predictors, consistently, across both samples and both sexes, were not personality or demographics but the qualities of the relationship itself: relationship satisfaction, sexual satisfaction, sexual desire, romantic love, and how long the couple had been together. Age, gender, education, and the other demographic variables that fill magazine articles carried almost no predictive weight at all.
Hold that honest surprise next to the previous section and you have the whole shape. Disposition travels with the person and raises their baseline risk. But whether that risk is realized depends heavily on the living conditions of the particular bond: whether there is still desire, still love, still satisfaction, still the felt sense of being chosen. This is why neither comfortable lie is adequate. The monster story ignores that good relationships protect and starved ones expose. The it-was-the-marriage story ignores that two people in the identical unhappy marriage will do different things depending on what they carry. Betrayal lives at the meeting point of a person and a relationship, and the machine learning, refusing to pick a side, is the most trustworthy voice in the room precisely because it declines to give you the single answer you wanted.
One more strand deserves naming, because it connects character and circumstance. How people bond under stress, what psychologists call attachment, shows up reliably in the infidelity research. Both of the insecure patterns raise the risk, by different routes. The anxiously attached, hungry for reassurance and quick to feel unchosen, sometimes seek outside a relationship the validation they cannot feel secure about inside it. The avoidantly attached, uncomfortable with closeness and prone to keeping an exit in view, sometimes use an affair as distance, a way of not being fully captured by anyone. This is not an excuse dressed as a diagnosis. It is a description of two well-worn paths to the same betrayal, and it explains why the same event can spring from opposite emotional needs, one person reaching for more closeness and the other for less.
Now the part that separates this from the content it is trying to correct. Every finding above is real, and every finding above is modest. The effect sizes are small to moderate, which means they describe tendencies across large groups and predict very little about any single human being. The samples lean heavily toward wealthy, educated, Western populations, so the confident global claims are on thinner ice than they pretend. The measures rest on self-report about behavior people are motivated to hide. And correlation, as always, is not cause; a trait that travels with betrayal is not proven to produce it. A person who tells you they can predict whether your specific partner will cheat is selling something. The science predicts populations, not persons.
This is not a weakness to apologize for. It is the finding. The truest thing the research has to say about infidelity is that it is overdetermined and underpredictable: caused by many small things acting together, and therefore impossible to reduce to the clean law that both the cynics and the romantics keep reaching for. Anyone who hands you a single cause, whether it is bad men, or bad marriages, or a broken culture, or a hidden disorder, is trading the actual evidence for the comfort of certainty. Refusing that trade is not fence-sitting. It is respect for how the thing actually works.
So what do you do with a body of knowledge that refuses to give you a formula. More than it first appears.
You can stop torturing yourself for having missed a sign that, statistically, was probably not there to see. Most betrayal is not preceded by a flashing warning the vigilant would have caught. The predictors are dispositional and quiet, and they hide well, especially in the early glow when everyone is on their best behavior.
You can take the one behavioral signal seriously without turning it into a superstition. A history of betrayal genuinely raises the odds, threefold in the best study, and it is reasonable to weigh that in whom you trust with your life, while still holding open the real possibility of a person who changed. The question of whether a cheater can actually change is its own long subject, and the honest answer is that it is possible but rare and demands more than remorse.
You can watch disposition over declaration. Words are cheap and people high in the relevant traits are often fluent in exactly the words you want to hear. What you are looking for over time is not a speech but a pattern: whether a person is honest when honesty costs them something, whether they treat your reality as real, whether their care survives the loss of an audience. Much of what the research calls low Honesty-Humility and the dark traits shows up, in ordinary life, as the person who quietly cares less and lets you carry the relationship.
And you can hold your own patterns to the same light, not to blame yourself for what was done to you, but because knowing the shape you keep repeating is the part of this you actually control. The research on who betrays is only half the story; the research on who we keep choosing, and why, is the half that can change your next relationship instead of just explaining your last one.
The comfortable lie is that infidelity has a cause, singular, locatable, and safely outside yourself, so that once you have named it, the monster or the marriage or the culture, you are done and clean. The truer thing is that it is a convergence: a person's disposition, sociosexuality and honesty and the dark end of temperament, meeting the living conditions of a particular bond, its desire and satisfaction and felt security, at a moment when an opportunity appears. No single element is the answer, which is why the field's most honest instrument, the one that let the data speak for itself, refused to name one. That refusal is not the research failing to find the truth. It is the research finding it. And the better thing that grows from accepting it is not a formula for spotting the guilty in advance, which no one honestly has, but a steadier way of seeing people: over time, by their patterns rather than their words, with the humility to know you are watching probabilities and not certainties, and the clarity, hard-won, to leave when the pattern is plain. The subject resists the single answer for the same reason people do. It is made of more than one thing, and telling the truth about it means saying so.
- Warach, B., & Josephs, L. (2024). The current state of affairs in infidelity research: A systematic review and meta-analysis of romantic infidelity prevalence and its moderators. Personal Relationships.
- Vowels, L. M., Vowels, M. J., & Mark, K. P. (2022). Is Infidelity Predictable? Using Explainable Machine Learning to Identify the Most Important Predictors of Infidelity. Journal of Sex Research.
- Knopp, K., Scott, S., Ritchie, L., Rhoades, G. K., Markman, H. J., & Stanley, S. M. (2017). Once a Cheater, Always a Cheater? Serial Infidelity Across Subsequent Relationships. Archives of Sexual Behavior (University of Denver).
- Sociosexuality, relationship satisfaction, and intentions toward infidelity: a moderated mediation model (2025). Sexual and Relationship Therapy.
- Certainty improves the predictive validity of Honesty-Humility and Dark Triad traits on cheating behavior (2024). Journal of Experimental Social Psychology.
- Altgelt, E. E., Reyes, M. A., French, J. E., Meltzer, A. L., & McNulty, J. K. (2018). Who is sexually faithful? Own and partner personality traits as predictors of infidelity. Journal of Social and Personal Relationships.
- The interplay of attachment styles and marital infidelity: A systematic review and meta-analysis. Heliyon.
What is the single biggest predictor of infidelity?
There is no single dominant cause, which is itself the most robust finding. The closest thing to a strong individual predictor is a prior history of cheating: a large longitudinal study found people who were unfaithful in one relationship were about three times more likely to be unfaithful in the next. Beyond that, the strongest predictors are qualities of the relationship itself, satisfaction, sexual desire, and love, rather than demographics like age or gender.
Do men cheat more than women?
On average, men report higher rates of sexual infidelity, roughly one in four versus one in seven in large samples, though these numbers vary widely with how the studies define cheating. The gap narrows sharply for emotional and digital infidelity, and it has been shrinking generation by generation, which suggests the difference is shaped by culture and circumstance and is not simply fixed by biology.
Is 'once a cheater, always a cheater' true?
It is an overstatement of something real. A history of infidelity roughly triples the odds of future infidelity, so it is a genuine and significant risk factor, not a superstition. But it is a tendency across groups, not a law about individuals; many people betray once and never again. It is reasonable to weigh a person's history seriously while still allowing for the rarer reality of someone who has genuinely changed.
Does an affair mean the relationship was bad?
Not necessarily, and this is a common and painful misconception. Relationship dissatisfaction does raise the risk, and the quality of a bond is among the strongest predictors, but whether dissatisfaction becomes an actual affair depends heavily on the straying person's disposition, especially their sociosexuality. Two people in the same unhappy relationship will do different things depending on what they carry. The relationship contributes, but it is rarely the whole cause, and the betrayed partner does not own half the blame by default.
Can you tell in advance if someone will cheat?
No, not reliably, and anyone who claims otherwise is overselling. The real predictors are dispositional and quiet, personality traits like low Honesty-Humility, sociosexuality, and the darker traits, and they hide well, especially early in a relationship. The science predicts tendencies across populations, not the behavior of any single person. The best you can do is watch patterns over time rather than words, and pay attention to whether someone is honest when honesty costs them something.