A study of 440,000 people found that humans are increasingly choosing partners who share their genetics. The trend has intensified every decade for the past 60 years.

When people explain how they chose their partner, they talk about chemistry, shared humor, compatible values, timing, and luck. The idea that something at the genetic level might be quietly shaping those choices, beyond the physical traits that are obviously visible, tends to feel like an overreach. People pick partners because of who those partners are, not because of what their genome contains.

A new study published in Nature Human Behaviour has examined what is actually happening across nearly half a million people, and the picture it reveals is more structured than the chemistry and luck account would suggest.

Researchers led by Robin Hofmeister at the University of Lausanne developed a method for estimating the genetic similarity between partners without requiring the genotypes of both partners directly. Instead, they reconstructed the parental genomes of 245,884 UK Biobank participants and 194,325 Estonian Biobank participants, a combined total of 440,209 individuals, by working backwards from the participants’ own genomes. Because a person inherits roughly half their DNA from each parent, the maternally and paternally inherited halves of their genome preserve a genetic record of how similar their two parents were to each other. By comparing these reconstructed parental genomes across 69 traits, the researchers could estimate the degree of genetic assortative mating, the tendency to choose genetically similar partners, for each trait separately, and then track how that tendency had changed across birth cohorts spanning the 20th century.

The result was consistent across both countries and across most of the traits examined. Genetic similarity between partners has increased with every generation studied. People born later in the 20th century had parents who were more genetically similar to each other than people born earlier in the century. The assortment has intensified, decade by decade, for 60 years.

What assortative mating actually means

Assortative mating is the tendency for individuals to select partners who are similar to themselves on some trait. It is well documented in humans for observable characteristics: people tend to partner with others of similar height, similar body weight, similar educational background, similar income level, and similar political orientation. None of this is particularly surprising. People meet potential partners in environments shaped by these traits, and they are drawn to familiarity.

What the new study adds is a measurement of the genetic footprint of these choices. When people choose partners who are phenotypically similar, and when the traits they are selecting on have a genetic basis, the partners will also tend to be genetically similar. The strength of the genetic similarity between partners is therefore a measure of how strongly, and how consistently, people have been selecting on heritable traits when choosing who to partner with.

The method Hofmeister and colleagues developed is a significant technical advance because it allows this measurement to be made without genotyping couples directly. Prior research on genetic assortative mating required DNA from both members of a partnership, which severely limited sample sizes. By reconstructing parental genomes from the DNA of their children, the team could use the vast UK and Estonian Biobanks, which were not designed to study couples, to answer questions about partnership patterns across generations.

“We developed a method that detects the genetic footprint of partner choice without requiring the input of partners’ genotypes,” the researchers wrote. “Applied to nearly half a million people in the UK and Estonia, our method reveals partner similarity that has intensified for height and educational attainment over the past six decades.”

The trait driving the strongest genetic sorting

Among the 69 traits analyzed, two showed the clearest genetic assortative mating signals: height and educational attainment. Both have been consistently documented in prior research as traits on which people sort when choosing partners. What the new study revealed is not just that sorting occurs but that it has changed over time, and the changes are not equal across traits.

For height, genetic assortative mating was detectable across the full period studied but showed relatively stable or modestly increasing trends. People have long chosen partners of similar height, and that preference has persisted without dramatic amplification.

For educational attainment, the trend showed stronger intensification. Educational assortative mating has increased more sharply across generations than height-based sorting, and because educational attainment has a substantial genetic component, driven by the same genetic variants that influence cognitive ability, personality traits, and socioeconomic trajectories, this means the genetic sorting of the population by education-related variants has been accelerating.

This finding matters beyond the immediate question of partner choice. When people with similar genetic profiles for educational attainment partner together, their children inherit genetic influences from both parents pointing in the same direction. The genetic variance for education-related traits in the population increases. The genetic distance between high-education and low-education families widens. Over generations, assortative mating on education can amplify genetic inequality in ways that interact with and compound social and economic inequality.

“Assortative mating plays an important role in shaping human behaviour, social structure and genetic architecture of human populations,” the researchers wrote. “In social sciences, assortative mating and homogamy have long been studied as a mechanism contributing to the closure of social groups and the emergence of growing inequality over generations.”

Why the trend has intensified

The study documents the trend but cannot fully explain it. Several mechanisms have been proposed in the broader literature, and more than one is likely operating simultaneously.

Educational expansion is one candidate. As more people attend university and postgraduate education, they spend more of their young adult years in environments that are strongly sorted by education-related traits. Meeting potential partners in those environments increases the probability of matching with someone genetically similar in the relevant dimensions. The more education becomes the dominant social environment in which partnerships form, the stronger the educational assortment becomes.

Changing marriage markets are another factor. The shift away from neighborhood and workplace partnerships, which were geographically constrained and therefore exposed people to more diverse potential partners, toward online dating and other selection-based platforms that allow people to filter explicitly on educational and socioeconomic characteristics may be amplifying assortative tendencies. When you can screen explicitly for someone who went to the same type of university, the genetic sorting that results from that screen is more powerful.

Delayed partnership formation may also be contributing. People who partner later in life tend to do so in contexts where educational and career sorting is more complete, meaning they are choosing from a pool that is already more genetically similar to themselves.

None of these explanations can be directly tested with the current data. What the data establish is the endpoint: genetic similarity between partners has increased, and the increase is robust across two different national populations studied independently with the same method.

What this means for genetic inequality across generations

The social implications of intensifying genetic assortative mating extend well beyond individual relationships. When like increasingly partners with like at the genetic level, the distribution of genetic variants in the population does not remain stable. Variants that influence education, income, and social mobility become more concentrated in certain lineages. The children of two genetically high-education parents carry more education-associated genetic variants, on average, than the children of mixed partnerships. Their advantages compound.

This genetic stratification operates alongside and interacts with the economic and social stratification that sociologists have documented in rising income inequality, declining social mobility, and the growing separation of educated and non-educated communities. The two processes are not independent. Genetic assortative mating on education-related traits is partly a product of social stratification, because educational sorting determines who people meet. And it is partly a driver of future stratification, because the resulting genetic distribution in the next generation shapes the distribution of outcomes.

“In social sciences, assortative mating and homogamy have long been studied as a mechanism contributing to the closure of social groups and the emergence of growing inequality over generations, and fragmentation in modern societies,” the researchers noted. “Conversely, intermarriage is seen as a driver of cultural integration, reducing distinctions between groups over generations.”

The finding that genetic assortative mating has intensified over 60 years, rather than remaining stable or declining, suggests the first dynamic has been winning over the second.

What the study cannot establish

The analysis relies on a methodological assumption: that the genetic similarity estimated between reconstructed parental genomes accurately reflects the genetic similarity between actual partners. The validation the researchers conducted against mate-pair-based estimates, where both partners were directly genotyped, showed their method performed comparably to or better than existing approaches, which is reassuring but not a perfect confirmation.

The study is also limited to UK and Estonian populations with access to biobank-level genetic data. Whether the same generational trend appears in other countries, particularly those with different educational systems, different urbanization patterns, or different histories of social mobility, is unknown. The Estonian and UK results were consistent with each other, which strengthens confidence in the finding, but two countries are not the world.

The mechanism connecting the measured genetic trend to specific behaviors and social changes also remains incompletely understood. The study demonstrates the genetic footprint of increasing assortative mating. It cannot directly observe the choices, social structures, and environmental changes that produced it.

What it establishes with the strongest evidence yet assembled for this question, across 440,209 people and six decades of partnership data, is that the genetic sorting of human populations by partner choice is not static. It has been changing, and the direction of change, toward greater similarity between partners, has been consistent and measurable across generations.

The study, “Parental haplotype reconstruction in up to 440,209 UK and Estonian individuals reveals recent assortative mating dynamics”, was authored by Robin J. Hofmeister, Davide Marnetto, and Zoltán Kutalik at the University of Lausanne and collaborating institutions, and published August 17, 2026 in Nature Human Behaviour.

Source: University of Lausanne. DOI: 10.1038/s41562-026-02556-8