Do couples look alike? What the research says

In 1987, Robert Zajonc, Pamela Adelmann, Sheila Murphy and Paula Niedenthal published "Convergence in the physical appearance of spouses" in the journal Motivation and Emotion (volume 11, issue 4, pages 335 to 346).

The study was small. It used photos of a small number of couples taken when they married and again about 25 years later. People who did not know the couples were asked to judge how similar the faces looked. The researchers reported that the spouses' faces were rated as more alike after 25 years than at the start.

Their suggested explanation was about emotion. Partners who share their lives may mimic each other's expressions for years. Over time, the theory went, using the same facial muscles in the same ways might shape faces in a similar direction.

The idea was memorable and spread widely. But a small sample and human judgement leave a lot of room for chance, which is why later researchers came back to it.

In 2020, Pin Pin Tea-makorn and Michal Kosinski published "Spouses' faces are similar but do not become more similar with time" in Scientific Reports (volume 10, article 17001). They set out to test the convergence idea with a much larger sample.

They collected photos of several hundred couples taken around the start of the relationship and again many years later. Similarity was measured in two ways: by human judges, and by a face recognition algorithm that compares faces using learned descriptors.

The finding is right there in the title. Spouses' faces were already more similar than random pairs at the start. But they did not become more similar over time. The authors found no evidence for convergence.

Their reading was that the similarity comes from who people choose in the first place. People tend to pair up with partners who are similar in various ways, sometimes called assortative mating, and that seems to include faces. The resemblance was there from the beginning, rather than growing over the years.

Based on the larger and more recent study, a fair summary is: somewhat, on average, from the start. Couples in the 2020 sample looked a little more alike than random pairs. There was no sign that years of marriage made faces grow more alike.

"On average" matters here. Plenty of happy couples look nothing alike. Plenty of strangers look like siblings. Research on groups tells you about patterns across many people. It does not tell you anything about one particular couple.

Science also moves. One large study is strong evidence, not the final word. If newer work changes the picture, the honest thing is to update.

The 2020 study used a face recognition algorithm to measure similarity. Our tool has a signal built on the same general kind of technology. The face resemblance signal compares two 128-number face descriptors from the face-api.js recognition model and turns the distance into a number from 0 to 100.

That is where the overlap ends. Our resemblance signal is only 25% of the fun score. The rest comes from mood match (40%), which compares expressions in the two photos, and proportion match (35%), which compares five face ratios. The final score is then scaled to a kind range of 55 to 100.

So the photo love test is a game for fun. It is not a replication of any study, it has not been validated as a measure of anything, and it does not say whether two people are a good match. If you want the full details, the how it works guide shows the exact formula.

If you are curious, you can play with the idea yourself. Run a couple photo from your first year together, then one from this year. Keep the angle and lighting as similar as you can. Then compare only the face resemblance number, not the overall score.

Do not read too much into it. Two photos are a tiny sample, cameras and lighting change, and the model has its own quirks and biases. But it can be a nice excuse to dig out old photos together, which is a good outcome on its own. The couple photo test page has more ideas for anniversaries and date nights.