Assortative Mating

Assortative mating is a form of non-random mating within population genetics and evolutionary biology in which sexually reproducing individuals pair with mates according to their degree of phenotypic (or underlying genotypic) similarity rather than panmictic chance. When individuals preferentially mate with partners exhibiting similar traits—such as comparable body size, flowering time, acoustic courtship song, coloration, or human stature and educational attainment—the regime is termed positive assortative mating (homogamy), producing a positive phenotypic correlation between mates (r > 0). Conversely, when individuals preferentially mate with partners displaying contrasting traits, the regime is termed negative assortative mating or disassortative mating (heterogamy, r < 0).

First analyzed mathematically in Ronald Fisher's foundational 1918 paper "The Correlation Between Relatives on the Supposition of Mendelian Inheritance" and expanded by Sewall Wright in 1921, assortative mating alone (without differential fertility or viability selection) does not change allele frequencies (p and q) in the gene pool. Instead, it reorganizes genotype frequencies away from Hardy-Weinberg principle proportions at trait-determining loci, alters heterozygosity in a locus-specific manner, and generates directional gametic phase disequilibrium across polygenic architectures—playing a pivotal role in quantitative genetics, reinforcement, and sympatric speciation.

Interactive Simulation: Compare how non-random mating systems and inbreeding coefficients alter genotype proportions without shifting allele frequencies in the Allele Drift Simulator.

Complete and Partial Positive Assortative Mating at a Single Locus

To see why positive assortative mating increases homozygosity without altering allele frequencies, consider a biallelic autosomal locus with alleles A₁ (frequency p) and A₂ (frequency q = 1 − p) where all three genotypes (A₁A₁, A₁A₂, A₂A₂) display distinguishable phenotypes (codominance or incomplete dominance) and mate exclusively with their own exact phenotype (complete positive genotypic assortment) with equal fecundity across all pairs.

Let P_t, H_t, and Q_t denote the frequencies of A₁A₁, A₁A₂, and A₂A₂ in generation t. Homozygote × homozygote matings (A₁A₁ × A₁A₁ and A₂A₂ × A₂A₂) breed true, producing 100% homozygous offspring, whereas heterozygote × heterozygote matings (A₁A₂ × A₁A₂, representing a fraction H_t of all matings) segregate according to Mendelian genetics into 1/4 A₁A₁, 1/2 A₁A₂, and 1/4 A₂A₂:

• P_t+1 = P_t + (1/4) · H_t
• H_t+1 = (1/2) · H_t
• Q_t+1 = Q_t + (1/4) · H_t

Checking the new frequency of allele A₁ confirms that p_t+1 = P_t+1 + (1/2)H_t+1 = P_t + (1/4)H_t + (1/4)H_t = P_t + (1/2)H_t = p_t: allele frequencies remain invariant. However, observed heterozygosity decays by 50% every generation (H_t = H₀ · (1/2)^t = H₀ · exp[−t · ln(2)]), converging as t → ∞ to complete homozygosity (P_∞ = p, H_∞ = 0, Q_∞ = q). If only a fraction r of the population mates assortatively while (1 − r) mates at random (partial assortative mating), heterozygosity converges to the non-zero equilibrium H_eq = 2pq · (1 − r) / [1 − (r/2)], corresponding to a locus-specific fixation index of F_eq = r / (2 − r). When A₁ is completely dominant over A₂ so that A₁A₁ and A₁A₂ share the same dominant phenotype, heterozygosity declines more slowly according to H_t+1 = 2pq(P_t + H_t/2) / (P_t + H_t).

Gametic Phase Disequilibrium vs Single-Locus Homozygosity

A fundamental paradox arises when passing from a single Mendelian locus to a continuous polygenic trait (such as human height, bird beak depth, or flowering date) controlled by L unlinked additive loci (where L is large, e.g., L = 100 to 10,000):

In a polygenic trait controlled by L loci, each individual locus contributes only a tiny fraction O(1/L) of total phenotypic variance. Consequently, as Ronald Fisher (1918) and Sewall Wright (1921) proved, phenotypic assortative mating with marital correlation r causes only a negligible O(r / L) increase in homozygosity at any single underlying locus (F_locus ≈ r · h² / [2L], which vanishes as L → ∞). Yet positive assortative mating exerts a massive macroscopic effect on the total additive genetic variance (V_A) of the polygenic trait!

How does assortative mating inflate polygenic variance without making individual loci homozygous? It does so by generating positive directional gametic phase disequilibrium (inter-locus genetic linkage disequilibrium across both homologous and non-homologous chromosomes) between trait-increasing ("plus", +) alleles at all L loci:

V_A = ∑_i=1^L 2p_iq_iα_i² + ∑_i≠j 2 · Cov(g_i, g_j)

When tall individuals mate with tall individuals and short individuals mate with short individuals, offspring inherit gametes enriched for "+" alleles simultaneously across loci i, j, k... or enriched for "−" alleles simultaneously across all loci. Even though unlinked loci recombine at rate c = 0.5 during meiosis, each generation of assortative mating replenishes the positive covariance Cov(g_i, g_j) > 0 across all L(L − 1) pairs of loci. At equilibrium in the infinitesimal limit (L → ∞), the equilibrium additive genetic variance (V_A,eq) is inflated above the random-mating panmictic variance (V_A,0) according to Fisher's formula:

V_A,eq = V_A,0 / (1 − r · h_eq²) = V_A,0 / (1 − r_G)

where h_eq² = V_A,eq / V_P,eq is the equilibrium narrow-sense heritability and r_G = r · h_eq² is the genetic correlation between mates. For example, if a trait has baseline heritability h₀² = 0.80 and marital correlation r = 0.30, positive assortative mating inflates additive genetic variance in the population by ~33% and expands the frequency of extreme phenotypes in both tails of the bell curve. Crucially, if the population returns to random panmixia (r = 0), free recombination decays this cross-locus gametic phase disequilibrium by 50% per generation, restoring V_A back to V_A,0 within ~5–6 generations.

Heterozygote Excess and Rare-Morph Mating Advantage

Under negative assortative (disassortative) mating, individuals reject mates bearing their own phenotype and pair preferentially with unlike phenotypes (r < 0). Disassortative mating produces two immediate population-genetic consequences:

1. Excess Observed Heterozygosity (H_o > H_e, F < 0): Because matings occur preferentially between different genotypes (e.g., A₁A₁ × A₂A₂ → 100% A₁A₂), the proportion of heterozygous offspring exceeds Hardy-Weinberg expectations.
2. Induced Negative Frequency-Dependent Sexual Selection: Unless population morph frequencies sit at the exact symmetric balance point, individuals belonging to the rarer phenotype have a much larger pool of acceptable (unlike) mates than individuals belonging to the common phenotype, generating negative frequency-dependent selection that stabilizes genetic polymorphism.

Empirical Examples of Disassortative Mating

• Distyly and Tristyly in Flowering Plants (Primula and Lythrum): First elucidated by Charles Darwin (1877), heterostylous primroses (Primula vulgaris) exist in two floral morphs: Pin (long style, low anthers, recessive ss) and Thrum (short style, high anthers, heterozygous Ss controlled by a hemizygous S-locus supergene). Because insects transfer pollen accurately only between anthers and stigmas positioned at the same floral height (and physiological self-incompatibility blocks intra-morph fertilization), nearly 100% of successful matings are disassortative crosses (Pin ss × Thrum Ss), producing a perpetual 50% Ss : 50% ss backcross ratio in every generation.
• White-Throated Sparrow (Zonotrichia albicollis) Plumage Supergene: North American white-throated sparrows exhibit two plumage and behavioral morphs: white-striped (WS) birds, which are territorial and aggressive but invest less in parental provisioning, and tan-striped (TS) birds, which are less aggressive and invest heavily in nestling care. Controlled by a large pericentric chromosomal inversion on chromosome 2 (ZAL2^m/ZAL2 in WS versus ZAL2/ZAL2 in TS), over 96% of breeding pairs in the wild consist of a disassortative WS × TS pairing, combining territorial defense with parental care and maintaining both inversion arrangements at ~50:50 morph frequencies (Tuttle et al., 2016).

Sympatric Speciation and Reinforcement of Reproductive Isolation

In evolutionary divergence, positive assortative mating is the primary behavioral bridge linking disruptive ecological selection to reproductive isolation (speciation). When disruptive natural selection favors two ecological extremes (such as large-beaked versus small-beaked finches, or benthic versus limnetic sticklebacks) while intermediate hybrids suffer reduced fitness, random mating continuously breaks apart divergently adapted gene complexes and regenerates unfit intermediates.

Selection therefore favors the evolution of positive assortative mating (reinforcement, Dobzhansky 1940) to prevent maladaptive hybridization. As Joseph Felsenstein (1981) demonstrated, reinforcement is hindered when recombination breaks up the association between ecological adaptation loci and mate-choice loci ("two-allele models"), but proceeds rapidly via one-allele mechanisms or "magic traits" (Gavrilets, 2004)—traits that are simultaneously the direct target of divergent ecological selection and the cue used for positive assortative mate choice. Classic magic traits include host-plant shifts in apple maggot flies (Rhagoletis pomonella), where divergent fruit ripening times automatically sort hawthorn- and apple-feeding flies by mating season (allochronic isolation), beak size and vocal resonance in Darwin's finches (Geospiza), and depth-adapted visual opsin sensitivity and nuptial coloration in African Lake Victoria cichlids (Pundamilia).

Assortative Mating in Human Populations and GWAS Confounding

In human populations (Homo sapiens), extensive biometric and genomic studies document moderate to strong positive assortative mating for stature (r ≈ 0.20–0.25), educational attainment (r ≈ 0.40–0.60), cognitive test scores (r ≈ 0.35–0.45), and psychiatric/behavioral traits. In modern genome-wide association studies (GWAS) and Mendelian randomization, cross-chromosome gametic phase disequilibrium induced by assortative mating—including cross-trait assortative mating (where individuals high in Trait A mate preferentially with individuals high in Trait B)—inflates polygenic score correlations, upwardly biases SNP heritability estimates, and generates spurious genetic correlations between biologically independent traits unless controlled using within-family sibling comparison designs (Border et al., 2022).

How does positive assortative mating differ from inbreeding?

Although both positive assortative mating and consanguineous inbreeding increase homozygosity relative to Hardy-Weinberg expectations, they differ fundamentally in genomic scope and mechanism. Inbreeding is mating between genealogical relatives (e.g., cousins or siblings) who share identical DNA segments across their entire genome by descent from common ancestors; therefore, inbreeding reduces heterozygosity by the same proportion (1 − F) at every locus across the entire genome. Positive assortative mating is mating based on phenotypic similarity for a specific trait; it increases homozygosity (and cross-locus gametic phase disequilibrium) strictly at the loci that control the chosen phenotype and their tightly linked chromosomal neighbors, while leaving unlinked neutral markers across the rest of the genome in Hardy-Weinberg equilibrium.

Why doesn't positive assortative mating change allele frequencies by itself?

If every individual in the population has the same expected number of surviving offspring regardless of their phenotype (no viability or fertility selection), then assortative mating merely changes which gametes pair together into diploid zygotes without changing how many copies of each allele are passed into the next generation's gamete pool. Every A₁A₂ × A₁A₂ mating still transmits exactly 50% A₁ and 50% A₂ alleles, just as A₁A₁ × A₁A₁ transmits 100% A₁. However, if assortative mating is imperfect such that rare phenotypes have difficulty finding a similar mate (inducing sexual selection against rare types), or if emerging homozygotes suffer inbreeding depression, then allele frequencies will shift due to the accompanying selection.

Does the Major Histocompatibility Complex (MHC) drive disassortative mate choice in vertebrates?

Following Claus Wedekind's (1995) famous "sweaty T-shirt" olfactory experiments in humans and laboratory studies in house mice (Mus musculus), researchers proposed that vertebrates use MHC-associated odorant peptides in urine and sweat to select mates with dissimilar MHC genotypes (MHC-disassortative mating), thereby maximizing offspring MHC heterozygosity and avoiding consanguineous inbreeding. Subsequent large-scale genomic evaluations across wild vertebrates and human cohorts indicate that MHC-based disassortative mate choice occurs strongly in certain salmonid fishes, seabirds, and rodents, while in humans its effect size varies across populations and is often secondary to socio-cultural and phenotypic positive assortative mating.

1. Fisher, R. A. (1918). "The correlation between relatives on the supposition of Mendelian inheritance." Transactions of the Royal Society of Edinburgh, 52(2), 399–433. doi:10.1017/S0080456800012163
2. Wright, S. (1921). "Systems of mating. III. Assortative mating based on somatic resemblance." Genetics, 6(2), 144–161. doi:10.1093/genetics/6.2.144
3. Jiang, Y., Bolnick, D. I., & Kirkpatrick, M. (2013). "Assortative mating in animals." The American Naturalist, 181(6), E125–E138. doi:10.1086/670160
4. Border, R., Athanasiadis, G., Buil, A., Schork, A. J., Cai, N., Young, A. I., Werge, T., Flint, J., Kendler, K. S., Sankararaman, S., Dahl, A. W., & Zaitlen, N. A. (2022). "Cross-trait assortative mating is widespread and inflates genetic correlation estimates." Science, 378(6621), 754–761. doi:10.1126/science.abo2059