There is a kind of couple who introduce themselves as opposites. You have met them. He is the spreadsheet, she is the chaos, and they deliver the line at dinner like a party trick they have polished for years. I used to nod along, because it sounded romantic and made the arithmetic of love feel generous, as though the universe hands you the exact thing you are missing.

Then I got curious about why my own longest relationships had nothing to do with contrast, and started reading. “Opposites attract” turns out to be more than a soft saying. It is one of the most heavily tested claims in relationship psychology, and it keeps losing.

The whole idea rests on 25 couples

The phrase has a birthday. In 1955, a sociologist called Robert Francis Winch published a paper in the American Sociological Review arguing that people pick partners who fill their gaps: the dominant one finds someone passive, the introvert finds an extrovert. He named it complementary needs and built the theory on a sample of 25 young married couples, most of them students.

Lovely theory. It walked straight into the culture and never left, turning up in rom-coms, wedding speeches and the way we explain our own partners to friends. What it never managed was surviving contact with other researchers. Replication attempts started failing almost immediately, and reviewers looking back decades later were cataloguing the failures rather than the support.

Then somebody looked at eight million couples

The proper answer landed in 2023, when a team at the University of Colorado Boulder ran the largest study of partner similarity ever attempted. They pulled 480 partner correlations across 22 traits from 199 published papers going back as far as 1903, covering data on roughly 8.5 million people, then added a fresh analysis of 133 traits in nearly 80,000 couples from the UK Biobank.

Almost every correlation came back positive. Political values landed at .58. Birth year was .87. Education, drinking habits and social attitudes all clustered high. Even the personality traits Winch’s idea specifically predicts should run negative came back positive, just faintly. Extraversion, the trait the whole introvert-meets-extrovert story leans on, sat at .08. Positive, but barely. Time put it more simply: couples match on up to 89% of everything measured.

The one real exception is a sleep schedule

Out of 133 traits, exactly one produced a meaningful negative correlation: chronotype, at -.18, meaning night owls pair off with early birds slightly more than chance would predict.

Which is quietly hilarious.

Nobody has ever described that arrangement as chemistry. Ask a couple living it and you will hear about logistics, in weary detail. The single place where genuine opposites turned up in the data is also the thing couples bicker about on a Tuesday night.

Feeling similar beats being similar

Here is the part that changed how I think about it. A 2008 meta-analysis crunched 460 effect sizes from 313 studies and split similarity in two: actual similarity, meaning you are measurably alike, and perceived similarity, meaning you believe you are. Both predicted attraction strongly early on. But once researchers looked at established relationships, actual similarity stopped predicting anything significant. Perceived similarity kept working at every stage.

Read that twice, because it does something strange to the dinner-party couple. Partners who feel deeply aligned are often not especially alike, and they do fine anyway. The pair announcing they are total opposites may simply be describing texture, while quietly agreeing on every question that matters.

Sort your differences into two piles

Back when I was running restaurants, my business partner and I were the kind of temperamental opposites people describe approvingly. He was front of house, charm on tap, remembered every regular’s grandchild. I lived in the kitchen and the spreadsheets. It worked for years, and at the time I credited the contrast.

I was wrong. It worked because we wanted the same restaurant. The moment we stopped agreeing on where the thing was heading, all that lovely contrast curdled into a slow, expensive argument.

So the useful move is sorting. Differences in pace, taste, tidiness, volume and how much small talk you can bear at a party are fine, occasionally even handy. Differences in direction are another species: money, children, honesty, ambition, what you owe your family, where you want to be living in ten years. Nobody complements you on those. You negotiate, forever.

The trap is that early on both piles feel identical, because novelty is novelty and your brain has not sorted it yet.

What I would actually do with this

Stop auditing dates for contrast, and resist the opposite reflex of auditing for sameness, which produces its own dull kind of relationship. Ask about direction. Where does this person think their life is heading, and does that description have any room in it for you?

Then notice whether you feel understood, because the perceived-similarity finding suggests that feeling is doing real structural work, not just being pleasant. Treat “we’re nothing alike” as a note on surface texture. It tells you almost nothing about whether the thing will hold.

One honest caveat. The Boulder researchers were careful to say their correlations are modest and should not be inflated into a rule about who belongs with whom, partly because this field has an ugly history of being borrowed by eugenicists. Fair enough. Modest still counts for something, and after 120 years of data, it still tilts one way.

Winch noticed something real. He just misread it. The spark you feel across a difference is genuine, and it is also the least reliable information you will ever collect about a person.