Broad-Sense Heritability

The staggering diversity of life raises a fundamental question: what makes individuals different? For centuries, this has been framed as a simple "nature versus nurture" debate. However, modern biology offers a more nuanced and powerful approach through the field of quantitative genetics. Instead of asking what part of a single individual's trait is genetic, we ask a more meaningful question: within an entire population, what proportion of the observable differences can be attributed to genetic variation, and what proportion to environmental variation? This shift in perspective, from the individual to the population, is the key to scientifically untangling the complex interplay of genes and environment.

This article delves into the core concept used to answer this question: heritability. It breaks down the total observable variation of a trait into its fundamental components, providing a statistical framework to quantify the influence of genetics. In the following chapters, we will first explore the foundational "Principles and Mechanisms" of broad-sense heritability, defining it and distinguishing it from its crucial counterpart, narrow-sense heritability. Then, in "Applications and Interdisciplinary Connections," we will journey through its practical uses, from the selective breeding programs that feed the world to the twin studies that illuminate the human condition, revealing how this elegant concept is a vital tool across the life sciences.

Why are individuals in a population different? Walk through a field of wildflowers, and you'll see a tapestry of varying heights, colors, and bloom times. Look at the people in a crowded room, and the diversity is just as striking. For centuries, we've debated the roots of these differences, often framing it as a battle between "nature" and "nurture." But science gives us a more powerful way to think about this. Instead of asking what fraction of a single flower's height is due to its genes versus its soil—a question that is ultimately meaningless—we can ask something much more interesting: of all the variation we see across the entire field of flowers, what proportion can be attributed to the differences in their genetic blueprints, and what proportion to the differences in their environments?

This is the central question of quantitative genetics. It shifts the focus from the individual to the population, and from absolute causes to the sources of variation. The total observable variation for a trait in a population is called the ​​phenotypic variance ($V_P$)​​. Think of it as a statistical measure of all the differences we can see and measure. The beautiful insight is that we can partition this total variance into two main sources. The part due to genetic differences among individuals is the ​​genetic variance ($V_G$)​​, and the part due to differing environmental conditions is the ​​environmental variance ($V_E$)​​. The simplest version of this grand equation is:

This little equation is our primary tool for untangling the contributions of nature and nurture to the diversity of life.

With our variance pie sliced into genetic and environmental pieces, we can define a wonderfully useful concept. The ​​broad-sense heritability​​, denoted as $H^2$, is simply the proportion of the total phenotypic variance that is genetic.

Imagine a team of agricultural scientists studying potato tuber weight. They find that the total variance ($V_P$) in weight is $12.0 \text{ g}^2$, and they estimate through pedigree analysis that the genetic variance ($V_G$) is $7.8 \text{ g}^2$. The broad-sense heritability is then straightforward to calculate: $H^2 = 7.8 / 12.0 = 0.65$. This number doesn't mean that 65% of any single potato's weight is "genetic." It means that 65% of the differences in weight we observe among all the potatoes in that specific population are due to the fact that they have different genes.

To truly grasp this, let's consider the extremes. What would it mean if $H^2 = 1.0$? This was explored by biologists studying feather length in a captive population of finches. A value of $H^2 = 1.0$ implies that $V_G = V_P$. Plugging this into our main equation ($V_P = V_G + V_E$), we are forced to conclude that $V_E = 0$. Does this mean the environment is irrelevant to growing feathers? Of course not! A finch still needs air, food, and water. What it means is that in the specific, controlled environment of the study, all the birds experienced identical conditions. Because there were no environmental differences, the environment could not be a source of variation among the birds. Therefore, any and all differences in feather length observed in that group had to be due to their genetic differences.

This leads to a clever experimental method for measuring heritability. How can we possibly isolate the environmental variance? By eliminating the genetic variance! Imagine you have a population of clones, like a special isogenic line of ants where every colony is genetically identical. Since they are all genetically the same, $V_G = 0$. Any variance you observe in a trait like foraging efficiency must be purely environmental ($V_P = V_E$). In the experiment, the variance among these identical ant colonies was 29.8 units². This gives a direct measurement of $V_E$. When the researchers then studied a genetically diverse population under the same range of conditions, they found a total variance ($V_P$) of 86.4 units². The rest is simple arithmetic: the genetic variance must be the total variance minus the environmental part, $V_G = V_P - V_E = 86.4 - 29.8 = 56.6$. The broad-sense heritability is then $H^2 = V_G / V_P = 56.6 / 86.4 \approx 0.655$. This is a beautiful example of how a simple, elegant experimental design can allow us to partition the sources of variation.

Here we arrive at one of the most critical and often misunderstood aspects of heritability. An $H^2$ value is not a fixed, universal constant for a trait. It is a property of a specific ​​population​​ in a specific ​​environment​​ at a specific ​​time​​. Change the environment, and the heritability will almost certainly change.

Consider a study on fruit weight in strawberries. In a uniform greenhouse, the genetic variance ($V_G$) was $40.0 \text{ g}^2$ and the environmental variance ($V_E$) was low, say $10.0 \text{ g}^2$. The total variance is $V_P = 40 + 10 = 50$, and the heritability is $H^2 = 40/50 = 0.8$. Now, take those same plants (with the same $V_G$) and grow them outdoors in a variable field, where differences in sun and soil quadruple the environmental variance to $V_E = 40.0 \text{ g}^2$. The genetic variance hasn't changed, but the total phenotypic variance has shot up to $V_P = 40 + 40 = 80$. The new heritability is now $H^2 = 40/80 = 0.5$. The heritability of the exact same trait in the exact same population dropped from 0.8 to 0.5 simply because the environment became more variable.

The same principle works in reverse. A wild plant population might show low heritability for flower size in its native mountain habitat because the harsh, variable conditions create a large $V_E$. Bring those same plants into a controlled greenhouse, and you drastically reduce $V_E$. Since $V_P = V_G + V_E$ gets smaller while $V_G$ stays the same, the heritability $H^2 = V_G / V_P$ necessarily increases. This is a profound point. The high heritability of human height in well-nourished populations doesn't mean nutrition is unimportant; on the contrary, it means nutrition is so uniformly good that it's no longer a major source of variation among people, allowing the underlying genetic differences to become more prominent in the statistics.

So far, we've treated genetic variance, $V_G$, as a single, monolithic entity. But the way genes produce traits is far more complex and interesting than that. The effects of genes are not always simple and cumulative. This is where we must peek inside the "black box" of $V_G$. Genetic variance itself can be partitioned into several key components:

• ​​Additive Genetic Variance ($V_A$)​​: This is the "well-behaved" component of genetic variance. It arises from the average effects of alleles that are added together to produce the phenotype. This is the primary reason offspring tend to resemble their parents.

• ​​Dominance Variance ($V_D$)​​: This component arises from interactions between alleles at the same locus. In a heterozygous individual, one allele might mask the effect of the other (dominance). The phenotype of the heterozygote is not simply the average of the two homozygotes. This is a non-additive effect.

• ​​Epistatic Variance ($V_I$)​​: This is variance from interactions between alleles at different loci. The effect of a gene for, say, pigment production might be modified by an entirely separate gene that controls pigment deposition. These inter-locus interactions are another form of non-additive effect.

This deeper understanding allows us to define a second, critically important type of heritability. ​​Narrow-sense heritability ($h^2$)​​ is the proportion of total phenotypic variance due solely to the additive genetic variance:

Because $V_A$ is just one part of $V_G$, it is a mathematical necessity that narrow-sense heritability can never be greater than broad-sense heritability ($h^2 \le H^2$). The difference between them, $H^2 - h^2 = (V_D + V_I) / V_P$, represents the proportion of variance due to those non-additive genetic interactions.

Why make this distinction? Because it has profound practical consequences. Imagine you're a plant breeder trying to select for larger leaves. You measure the broad-sense heritability and find it's very high, say $H^2 = 0.85$. Great! This tells you that most of the variation in leaf size is genetic. You confidently select and cross the plants with the biggest leaves, expecting their offspring to inherit the trait. But to your surprise, the next generation shows almost no improvement. What happened?.

The answer is that you've run into the breeder's dilemma: you have high broad-sense heritability, but very low narrow-sense heritability. While most of the variation is genetic ($H^2$ is high), almost none of it is additive ($h^2$ is low). This means the large leaves you selected for were likely caused by specific, fortunate combinations of genes—favorable dominance or epistatic effects. But these lucky combinations are broken apart during sexual reproduction and are not reliably passed on to the offspring. The response to selection depends on the predictable, additive effects, which are captured by $h^2$.

We can see this principle in action with a simple, concrete example. Consider a trait controlled by a single gene with a dominant allele $A_1$ and a recessive allele $A_2$. Let the phenotypes be $A_1A_1 = 10$, $A_1A_2 = 10$, and $A_2A_2 = 2$. If we assume there is no environmental variation ($V_E = 0$), then all variance is genetic, so $H^2 = 1.0$. However, because of the complete dominance, the heterozygote's phenotype is not intermediate. This non-additive effect creates dominance variance ($V_D$). A careful calculation for a population with specific allele frequencies shows that the additive variance ($V_A$) is only a fraction of the total genetic variance. In this specific case, the narrow-sense heritability turns out to be only $h^2 \approx 0.571$. The trait is 100% "heritable" in the broad sense, but its response to selection is governed by the much lower narrow-sense heritability.

This distinction is crucial for everything from agriculture to conservation biology. A clonally propagated crop, like many fruit trees, can take full advantage of high broad-sense heritability. Since you're just making copies, those lucky non-additive gene combinations are preserved perfectly. But for a sexually reproducing population that needs to adapt and respond to selection, it is the narrow-sense heritability that holds the key to the future. It tells us how much raw material for evolution is available, how faithfully traits are passed down, and how effectively nature—or the breeder—can shape the generations to come.

Now that we have grappled with the principles of heritability, we might be tempted to put it on a shelf as a neat, but abstract, statistical concept. Nothing could be further from the truth. The idea of partitioning variance into its genetic and environmental components is one of the most powerful and practical tools in all of biology. It is the bridge that connects the theoretical world of genetics to the tangible challenges of improving our crops, understanding human health, and conserving the rich biodiversity of our planet. It allows us to move from mere observation to informed prediction and action. Let us now embark on a journey to see how this single idea finds its expression in a remarkable variety of fields.

Perhaps the most direct and historic application of heritability lies in agriculture and aquaculture, where for millennia we have sought to mold organisms to our needs. Imagine you are an aquaculturist trying to breed faster-growing tilapia. You observe that some fish grow much larger than others. The crucial question is: how much of this variation is due to their genes, which you can select for, versus their specific corner of the tank, which you cannot?

To answer this, a geneticist can perform a clever experiment. They can raise two populations under identical conditions: one that is genetically diverse, and another that is a highly inbred, isogenic line where every fish is essentially a genetic clone of the others. In the isogenic line, any variation in growth rate must be due to subtle, unavoidable differences in the environment, since their genes are the same. This gives us a direct measurement of the environmental variance, $V_E$. Once we know $V_E$, we can subtract it from the total phenotypic variance, $V_P$, of our diverse population to find the holy grail: the genetic variance, $V_G$. The broad-sense heritability, $H^2 = V_G / V_P$, then tells us what fraction of the total variation in growth rate is due to genes. A similar strategy is used in plant breeding, where one might compare a genetically variable F2 generation of corn with the genetically uniform F1 generation it came from to estimate the heritability of a trait like plant height.

These methods give breeders a vital first look at a trait's "genetic potential." A high $H^2$ suggests that a selective breeding program could be very effective. However, the real world of animal breeding often requires more sophisticated tools. By analyzing detailed family pedigrees—for example, by comparing groups of half-siblings (sharing a sire) to groups of full-siblings (sharing a sire and a dam) in a cattle herd—quantitative geneticists can use statistical methods like the Analysis of Variance (ANOVA) to tease apart not just $V_G$ and $V_E$, but the different flavors of genetic variance itself. This leads us to a crucial and subtle distinction.

If you are a breeder trying to improve a trait, you will soon discover that broad-sense heritability, $H^2$, can be a seductive but misleading guide. It tells you the total impact of genes, but not all genetic influence is created equal when it comes to sexual reproduction.

The total genetic variance, $V_G$, is actually a composite of several effects. The simplest is the ​​additive genetic variance ($V_A$)​​, which represents the sum of the average effects of individual alleles. But there is also ​​dominance variance ($V_D$)​​, which arises from interactions between alleles at the same locus (like a recessive allele's effect being masked by a dominant one), and ​​epistatic variance ($V_I$)​​, which comes from interactions between alleles at different loci.

Broad-sense heritability, $H^2 = (V_A + V_D + V_I) / V_P$, includes all these components. But when parents reproduce sexually, they don't pass down their entire genotype as a block. Meiosis and recombination shuffle the deck. A parent might have a fantastic combination of genes that produces a winning phenotype, but they only pass on a random half of their alleles to each offspring. The specific, favorable combinations that give rise to dominance and epistatic effects are broken apart. Think of it like a winning hand in poker; you can't pass the hand to your child, only individual cards from your half of the deck.

The only component that is reliably transmitted from parent to offspring and creates a predictable resemblance between them is the additive variance, $V_A$. For this reason, the truly predictive measure for a breeder is ​​narrow-sense heritability ($h^2$)​​, defined as $h^2 = V_A / V_P$. The famous ​​breeder's equation​​, which predicts the response to selection ($R$) from the selection differential ($S$), depends exclusively on $h^2$:

$R = h^2 S$

Imagine trying to breed fish for more aggressive behavior. You select the most aggressive individuals, whose average aggression is far above the population mean. How much will the next generation's average aggression increase? If you naively used a high $H^2$ of, say, $0.80$, you'd expect a dramatic response. But if the narrow-sense heritability is much lower, perhaps $h^2 = 0.35$, because most of the genetic variance is non-additive, the actual response to your selection will be far more modest. This distinction is the difference between a successful breeding program and a costly failure.

There is, as always in biology, a fascinating exception that proves the rule. What if you are breeding a plant, like a potato, that is propagated asexually by cloning? In this case, the parent's entire genotype—including all the wonderful and complex dominance and epistatic interactions—is passed on intact. The "winning hand" is not reshuffled. Therefore, for predicting the response to selection in clonal organisms, the broad-sense heritability, $H^2$, becomes the relevant measure once again. This beautiful contrast underscores the profound impact of the mode of reproduction on the evolutionary process.

While we can run controlled breeding experiments on corn and cattle, we cannot and should not do so with humans. Yet, we have a deep-seated desire to understand the roots of our own traits, from our risk for diseases like schizophrenia to our cognitive abilities. How can we apply the logic of heritability here? We look for "natural experiments," and the most powerful of these is the study of twins.

Identical (monozygotic, or MZ) twins originate from a single fertilized egg and share essentially 100% of their genes. Fraternal (dizygotic, or DZ) twins develop from two separate eggs and are genetically as similar as any other pair of siblings, sharing on average 50% of their segregating genes. This simple fact provides an extraordinary research tool.

The most elegant form of this is the study of identical twins who were separated at birth and raised in different environments. Because they have the same genes but different environments, any correlation in a trait between them provides a direct and stunningly simple estimate of the broad-sense heritability, $H^2$. The similarity between these pairs of genetic duplicates raised in uncorrelated worlds is a powerful testament to the influence of the genome. By also studying identical twins raised together, we can begin to untangle the effects of the shared family environment from the genetic effects.

Even when studying twins raised together, the comparison between identical and fraternal pairs is incredibly revealing. If, for a given trait or disease, identical twins are significantly more likely to be concordant (both have the trait) than fraternal twins, it provides strong evidence for a genetic contribution. This logic has been fundamental in establishing the genetic underpinnings of countless human conditions, from heart disease to psychological disorders.

The concept of heritability is also vital for understanding populations in their natural habitats, especially in the context of conservation biology. Here, however, we must be extra cautious. Heritability is not a fixed biological constant like the speed of light; it is a property of a specific population in a specific environment at a specific time.

Consider a conservation program aiming to save an endangered bird by creating a captive population from two genetically distinct wild populations, say one from the highlands and one from the lowlands. If these two populations have diverged genetically, resulting in different average beak depths, mixing them creates a new kind of variance. The total genetic variance in the mixed group is not just the average of the variances within each source population; it also includes the variance between the means of the two groups. This between-group component is entirely genetic, and it can artificially inflate the calculated broad-sense heritability for the combined population. This phenomenon, a form of the Wahlund effect, is a critical warning: an ecologist who measures a high $H^2$ in such a mixed group might mistakenly conclude there is vast potential for selection, when in fact much of the variance is due to the historical baggage of population structure.

Our journey ends at the cutting edge of modern genomics, where heritability remains a central—and challenging—concept. For decades, twin and family studies have suggested high broad-sense heritabilities for many human traits, such as height (around $H^2=0.8$). This means about 80% of the variation in height among people is due to genetic differences.

In the 21st century, Genome-Wide Association Studies (GWAS) have allowed us to scan the genomes of hundreds of thousands of individuals, identifying specific genetic variants (SNPs) associated with traits. A natural idea was to add up the effects of all these height-associated SNPs to see if they explained the 80% heritability seen in twin studies. They didn't. The sum, often called "SNP heritability," was much lower. This gap became famously known as the "missing heritability" problem.

Where did it go? The framework of variance components we have been exploring holds the key. The puzzle can be solved by recognizing what each method measures.

1. ​​Non-Additive Effects:​​ The simple GWAS approach primarily captures additive effects ($V_A$). The broad-sense heritability from twin studies, however, also includes dominance and epistatic effects ($V_D$ and $V_I$), which are "missing" from the simple SNP-based sum.
2. ​​The Impact of Rare Variants:​​ Initial GWAS technologies were best at detecting common genetic variants. Much of the heritability, however, may be due to a vast number of rare variants, each with a tiny effect. These were largely invisible to early studies.
3. ​​Imperfect Tagging:​​ A SNP chip does not measure every one of our three billion DNA letters. It samples key "tag" variants. The actual causal variant might be near a tag, but not perfectly correlated with it. This imperfect correlation means the GWAS only captures a fraction of the true additive variance.

A sophisticated analysis reveals that the "missing" heritability isn't really missing at all. It was hiding in non-additive effects, rare variants, and the statistical noise of imperfect measurement. Solving this puzzle shows how a concept developed nearly a century ago to aid cattle breeders remains indispensable for making sense of the human genome and paving the way for personalized medicine. From a simple ratio, a world of application and insight unfolds.