Why “Similars” Choose “Similars”: Simple Genetic Mechanics That Generate Associative Marriages

The idea that couples are more likely to resemble each other than random people has long been confirmed - from height and education to political views and habits. But "why is that?" has remained controversial: whether we are brought together by common environments and social circles, or whether we consciously select partners based on matching traits, or whether similarities "accumulate" after the start of a relationship. A new preprint on PsyArXiv offers a surprisingly parsimonious explanation: if both preferences and preferred traits are heritable, then genetics itself eventually links "having a trait" with "liking that trait in a partner," and the similarity between partners arises without additional assumptions.

Background of the study

In human couples, similarity is the norm rather than the exception: partners are statistically similar in education, attitudes, and a number of behavioral habits, while the similarity is weaker in most personality traits. Modern reviews and meta-analysts record positive correlations between spouses on dozens of traits and emphasize that the observed “similarity” can arise both from deliberate selection “for oneself” and from social sorting (we meet in the same circles) and convergence after the beginning of a relationship (a common life makes us a little more similar).

The classic question is why couples turn out to be similar. In addition to social explanations, behavioral genetics has long discussed the role of heritability: if both the traits themselves and the "tastes" for a partner are partly determined by genes, then preferences can "pull" people with similar genetic predispositions. There are empirical signals in this direction: twin studies find non-zero heritability of preferences (for example, for height or similarity in attitudes), and genomic studies show that long-term associative mating leaves a recognizable "mark" in the genome of populations.

The new preprint proposes a parsimony mechanics: even in a “sterile” model without social filters, it is enough to assume that variations of a trait and preferences for this trait are inherited. Then, over time, a genetic correlation is formed “I have a trait ↔ I prefer the same trait in a partner,” and from this, associative selection of pairs naturally grows. This does not cancel out other mechanisms, but provides a minimal, general biological explanation that can work together with social factors. Interest in the work is also fueled by the first popular retellings: the idea has already been picked up by specialized sites, and the article itself is available on PsyArXiv with materials.

This has practical implications for interpreting large cohorts and genetic studies: associative marriages change the variance of traits, amplify extreme combinations, and may distort estimates of heritability and correlations if marriages are assumed to be “random” by default. Therefore, when modeling risks and social outcomes, it is prudent to consider both social sorting and genetically mediated selection, keeping in mind that the contribution of each mechanism may differ across traits (attitudes, education, habits) and cultures.

What hypothesis was tested?

The authors put forward a minimalist hypothesis: heritable variations in (a) traits and (b) preferences for these traits generate genetic correlations of the type "have a trait → prefer the same trait in a partner." If so, then even without "social filters" and without a special "search for similarities" mechanism, the population will slowly self-tune to associative selection of pairs (homogamy). To test this, they created an agent-based model: 100 generations of partner selection, each agent has 40 loci for a trait and 40 loci for a preference for this trait; pairs are formed according to the strength of preferences, then inheritance, and so on in a circle.

What happened in the simulations

Already on this “toy” model, after several dozen generations, two key effects appeared:

• Correlations between the characteristics of partners (the same “similar with similar”) - that is, associative marriages as a result of the natural course of generation after generation.
• Genetic links between a "trait" and a "preference for the trait" - for example, carriers of alleles that increase the level of X are more likely to also carry alleles that "love" X in a partner.
  The authors did not introduce any "magic" mechanisms: heritability and selection by preference are sufficient. This makes the explanation parsimonic and universal - applicable to different traits and species.

How does this fit into what we already know from big data?

Large meta-analyses and databases like the UK Biobank have long shown that partner similarity is widespread and covers dozens of traits: from social attitudes and education to some behavioral habits; it is less noticeable for trait personality. The new preprint does not dispute these facts, but provides a minimalist mechanics that can generate them even without a social “sieve”. In other words, “similar to similar” may not only be a consequence of the environment or living together, but also a natural consequence of the heritability of preferences and traits.

What exactly was done in the model (important details)

The authors asked:

• Genomics "on the fingers" - each agent has a set of loci for the trait itself and a separate set for the preference for this trait in the partner.
• The selection rule is that pairs are formed to maximize preference satisfaction (without complicating conditions such as geography or the marriage market).
• Vertical transmission - offspring inherit both traits and preferences from both parents, which causes genetic correlations "trait ↔ preference" to grow over time.
  The result: even in such a sterile "clean" environment, a positive correlation between partners by trait grows - that is, associative mating as a "by-product" of heritability and choice by taste.

How does this differ from classical explanations?

There are at least three popular lines of explanation for associative marriages with which the new mechanics are compatible but distinct:

• Social stratification / "social filters": people meet and converge within the same classes and circles.
• Phenotype matching: We like things that are similar to us.
• Similarity after the start of a relationship (convergence): couples become similar by living together.
  The model shows that even without all this, the similarity of partners will increase if traits and preferences are inherited. And in the real world, all mechanisms can add up - hence the strong effects regularly recorded by meta-analyses.

Why this matters for genetics and population dynamics

Associative marriages change the dispersion of traits and can amplify extreme values - this has long been insisted upon by population genetics. If the described mechanism is correct, it helps to explain why and how quickly such shifts occur even without obvious "social barriers". Practical consequences:

• Adjustments in study designs (e.g. twin and GWAS estimates may be biased if random marriages are assumed).
• Demography and epidemiology: similarities in education, behavioral risks, etc. are reflected in the health of offspring and social mobility.
• Evolution of preferences: genetic linkage of preference ↔ trait helps us understand how tastes are fixed and why “like-like attraction” is stable across cultures.

Limitations (the authors emphasize them themselves)

This is a "toy model" and not a realistic marriage market simulator:

• no geography, age, partial monogamy/polygamy, economic framework, etc.;
• the "strength" of the proposed mechanism was not compared with alternatives in the same system;
• the robustness of genetic correlations to drift, migration, and selection by other traits was not tested.
  And yet, in scientific logic, this is a strong demonstration of the principle: the heritability of preferences and traits is capable of itself generating similarity between partners, and then in the real world social and cultural factors will be “added” to it.

What to check next (roadmap)

• Implement a partner market: add spatial constraints, “search cost”, age, gender imbalance, network effects.
• Merge with data: Estimate the heritability of preferences across broad cohorts and see if it predicts increases in partner similarity across generations.
• Multi-trait scenarios: when several traits (appearance, education, attitudes) are important at once, how do cross-correlations between traits behave in spouses?
• Compare the mechanisms head-on: how much of the observed similarity is explained by the “genetic linked” mechanism alone, and how much by social stratification or convergence?

What the reader should take away right now

• "Similar with similar" is not necessarily a consequence of "vicious circles" - part of the effect may be natural genetic dynamics.
• Big data does not contradict this idea: meta-analyses show a wide range of similarities, especially in social attitudes and education.
• For policy and science, this is a signal to interpret correlations in pairs more carefully and to take associative marriages into account when modeling risks, heritability, and social outcomes.

Source(s): Harper KT, Zietsch BP Assortative Mating is a Natural Consequence of Heritable Mate Preferences and Preferred Traits. https://doi.org/10.31234/osf.io/ycaxz_v1