Narrow-Sense Heritability

In the study of life, one of the most fundamental questions is why individuals within a species vary. From the height of a tree to the disease resistance of a crop, these differences are the raw material for evolution and the focus of selective breeding. Quantitative genetics provides the tools to dissect this variation, and at its heart lies the concept of heritability. However, simply asking "how much is genetic?" is not enough. To truly predict how a trait will change over generations, we must ask a more precise question: how much of that genetic influence is reliably passed down? This is the critical gap addressed by the concept of narrow-sense heritability.

This article unpacks this powerful idea. In the first chapter, ​​Principles and Mechanisms​​, we will deconstruct phenotypic variation into its environmental and genetic components, further breaking down genetic variance into its additive and non-additive parts to reveal why narrow-sense heritability is the engine of evolutionary change. Following that, the chapter on ​​Applications and Interdisciplinary Connections​​ will demonstrate how this theoretical ratio is a practical tool used by breeders, evolutionary biologists, and genomicists to sculpt life, read the story of inheritance, and frame the search for the genetic basis of complex traits.

Imagine you're walking through a forest. You notice that the pine trees, while all recognizably the same species, are not identical. Some are taller, some have denser needles, some have thicker trunks. What causes this variation? Is it because they have slightly different genes? Or is it because one grew in a sunny, fertile spot while another struggled in a rocky, shaded patch? The answer, of course, is both. And in asking this simple question, you've stumbled upon the central problem of quantitative genetics.

Our goal is to untangle this puzzle. We want to know how much of the observable difference—the ​​phenotypic variance​​ ($V_P$)—in a trait like tree height is due to genetic differences, and how much is due to the environment. But we're going to find that this leads us to a deeper, more subtle, and far more powerful idea.

Let's start with a thought experiment. Suppose a biotechnology company creates a field of wheat where every single plant is a perfect clone of every other. They are all genetically identical. As the wheat grows, you walk through the field and notice, just like with the pine trees, that the plants aren't all the same height. Some are a bit taller, some a bit shorter. Since they are all genetically identical, the total ​​genetic variance​​ ($V_G$) in this population is zero. Therefore, any variation you see must be due to tiny differences in their environment—subtle variations in soil nutrients, water, or sunlight. This is the ​​environmental variance​​ ($V_E$). In this special case, the total phenotypic variance is just the environmental variance: $V_P = V_E$.

This simple example reveals a profound truth: ​​heritability is a property of a population, not an individual​​. A single gene in an individual isn't "heritable"; rather, the variation for a trait within a group can be heritable. And as we've just seen, heritability requires genetic variation. If there's no genetic variation, heritability is zero, no matter how much the trait is "controlled by genes."

Now, let's flip the scenario. Imagine two different populations. In Population 1, everyone follows a strict, uniform diet. In Population 2, people eat whatever they want. A researcher studies blood pressure and finds that the environmental variance in Population 2 is much larger because of the diverse diets. Even if the underlying genetic diversity is the same in both groups, the heritability of blood pressure will be lower in Population 2. Why? Because the total phenotypic variance ($V_P = V_G + V_E$) is the denominator in the heritability equation. By increasing $V_E$, we've made the genetic contribution a smaller proportion of the total variation you see. Heritability isn't a fixed constant of nature; it's a context-dependent measure for a specific population in a specific environment.

So, we've split variation into two big buckets: genes ($V_G$) and environment ($V_E$). Now, we need to look inside the genetic bucket, because not all genetic effects are created equal. This is where the story gets really interesting.

Think of your genetic inheritance from your parents like a financial inheritance. Some of it comes in the form of a reliable trust fund. This is the ​​additive genetic variance​​ ($V_A$). These are the effects of alleles that are "well-behaved." Each "plus" allele you inherit adds a little bit to your height, and each "minus" allele subtracts a little bit. These effects are dependable and are passed on faithfully from parent to child. It's this component that causes the fundamental resemblance between relatives.

But there's another part of your genetic inheritance that's more like a one-time lottery win. These are the ​​non-additive genetic effects​​. They arise from specific combinations of alleles that might make you a big winner, but the winning ticket gets torn up and reassembled in your children. There are two main kinds:

1. ​​Dominance Variance ($V_D$)​​: This arises from interactions between alleles at the same locus. Let's imagine a gene for running speed in mice, with a "fast" allele (F) and a "slow" allele (s). Suppose F is completely dominant over s. This means mice with genotypes FF and Fs are equally fast, while ss mice are slow. Now, if a fast Fs mouse mates, it passes on either an F or an s allele, each with 0.5 probability. It can't pass on its "winning combination" of Fs. The superiority of the Fs genotype over the average of FF and ss genotypes creates variance, but it's not reliably transmitted. It's a fleeting fortune based on a combination that meiosis breaks apart.

2. ​​Epistatic Variance ($V_I$)​​: This is an even more complex lottery, involving interactions between alleles at different loci. The effect of a gene for, say, a pigment enzyme might depend entirely on whether another gene for a precursor molecule is functional. Again, these specific, lucky combinations of genes across the genome are shuffled and broken up during reproduction.

So, our genetic variance bucket is actually made of three smaller parts: $V_G = V_A + V_D + V_I$. The additive part is the reliable, transmissible portion. The dominance and epistatic parts are due to genetic interactions that are not faithfully passed on.

This crucial distinction between additive "trust funds" and non-additive "lottery wins" forces us to define two different kinds of heritability.

​​Broad-sense heritability ($H^2$)​​ is the proportion of total phenotypic variance that is due to all genetic factors combined. $H^2 = \frac{V_G}{V_P} = \frac{V_A + V_D + V_I}{V_P}$ This tells us how much of the variation we see is due to genes in the broadest sense. If we were to make clones of all the individuals in a population, $H^2$ would tell us how much of the original population's variation is captured by the variation among the clone averages. It's the total genetic potential locked within the population's gene pool.

​​Narrow-sense heritability ($h^2$)​​, on the other hand, zooms in on only the additive component. $h^2 = \frac{V_A}{V_P}$ This tells us what fraction of the total variation is due to the transmissible, additive genetic effects. It's the part that makes offspring resemble their parents. Because variance components like $V_D$ and $V_I$ cannot be negative, the total genetic variance $V_G$ must always be greater than or equal to the additive variance $V_A$. Consequently, it's a mathematical certainty that $h^2 \le H^2$.

Why go through all this trouble to distinguish $h^2$ from $H^2$? Because narrow-sense heritability is nothing short of a crystal ball for evolution. It is the single most important parameter for predicting how a population will respond to selection, whether that selection is imposed by a farmer (artificial selection) or by nature (natural selection).

The magic lies in the ​​Breeder's Equation​​: $R = h^2 S$ Let's unpack this elegant little formula. Suppose we are breeding microalgae to produce more of a valuable chemical. We measure the whole population and find the average yield. Then we select only the very best producers to be the "parents" of the next generation. The difference between the average of our chosen parents and the average of the original population is the ​​selection differential ($S$)​​. It’s a measure of how picky we are. The ​​response to selection ($R$)​​ is the change we actually see in the next generation's average yield.

The Breeder's Equation tells us that the response we get is only a fraction of the selection we apply, and that fraction is precisely the narrow-sense heritability, $h^2$. If $h^2=0.5$, then the offspring generation will make up 50% of the gap between the original population and the super-producers we selected. Why only a fraction? Because only the additive variance ($V_A$) is reliably passed on. The superior performance of the parents due to lucky dominance ($V_D$) or epistatic ($V_I$) combinations is lost in the shuffle of meiosis.

This explains why a breeder could be deeply disappointed. Imagine they find that the broad-sense heritability for yield is very high, say $H^2 = 0.75$. They might think, "Great! 75% of the variation is genetic! Selection will be incredibly effective." But suppose that most of this genetic variance is non-additive, and the narrow-sense heritability is only $h^2 = 0.15$. A strong selection effort will yield only a meager 15% response—a huge disappointment. In an extreme (but theoretically possible) case of overdominance, where the heterozygote is superior but all additive variance has been eliminated, $h^2$ could be 0 even if $H^2$ is very high. In that case, a breeder could select the best individuals forever and see zero improvement in the next generation. It's a perfect illustration of the principle: selection acts on phenotypes, but evolutionary change depends on the transmission of additive genetic variance.

This brings us to our final, and perhaps most beautiful, point. Heritability is not just a static measurement; it is a dynamic quantity that tells a story about a trait's evolutionary past and future.

If a trait is under strong, sustained ​​directional selection​​—like bacteria in a lab being selected for faster metabolism—the selection process will favor certain alleles. Over many generations, these advantageous alleles will increase in frequency and eventually become "fixed" (reaching a frequency of 100%). As they become fixed, the variation at those genetic loci disappears. This means the additive genetic variance, $V_A$, is consumed by the process of selection itself. Consequently, both $V_A$ and $h^2$ for the trait are expected to decrease, eventually approaching zero (in the absence of new mutations). The population has climbed its adaptive peak, and the genetic variation that got it there has been used up.

Now consider a different kind of selection: ​​stabilizing selection​​. This acts on traits that are critical for survival, where there is a "just right" optimum, and any deviation is bad. Think of the number of legs on a beetle, or the number of photophores on a deep-sea creature that uses them for camouflage. For millions of years, nature has ruthlessly eliminated individuals with too many or too few. This long, intense process has also scoured the population of most of its additive genetic variance for the trait. What remains is often non-additive variance. So, counterintuitively, many traits most essential for survival have very low narrow-sense heritability. It's not because they aren't "genetic"—they are profoundly genetic! It's because natural selection has already done its job so well that the raw material for further change ($V_A$) has been exhausted. The low $h^2$ is itself a footprint of a long and successful evolutionary history.

And so, from a simple question about pine trees, we've journeyed to the heart of evolutionary theory. Narrow-sense heritability is more than just a fraction. It is the engine of short-term evolution, the predictor of a breeder's success, and a living record of the ageless dance between variation and selection.

What if I told you there was a number, a single value, that could tell a farmer how much fatter his pigs will get, a biologist how quickly a finch's beak might change, and a geneticist why it is so difficult to find the "gene for" height? This number is not a piece of arcane trivia; it is a tool of immense practical and intellectual power. It is the narrow-sense heritability, $h^2$. Having journeyed through its theoretical foundations, we now arrive at the real playground: the world of its applications. Here, we will see how this simple ratio of additive genetic variance to total phenotypic variance, $h^2 = V_A / V_P$, acts as a bridge connecting genetics to evolution, agriculture, medicine, and even the study of behavior.

For millennia, humans have been shaping the organisms around them. We have selected the most docile wolves, the most fruitful grains, and the fastest horses, slowly molding them to our needs. This process, artificial selection, was an art long before it was a science. Narrow-sense heritability is the formula that demystified the art.

The magic lies in a wonderfully simple and powerful relationship known as the ​​Breeder's Equation​​: $R = h^2 S$ Here, $S$ represents the selection differential—a measure of our effort. If the average height of a plant population is 100 cm and we choose only plants that are, on average, 103 cm tall to be parents, our selection differential $S$ is 3 cm. $R$ is the response to selection—the actual evolutionary change we observe in the next generation's average height. The equation tells us something profound: the outcome ($R$) is not equal to our effort ($S$). Instead, the offspring generation only recaptures a fraction of the parents' advantage. That fraction is precisely the narrow-sense heritability, $h^2$.

If $h^2 = 0.5$, then half of the selected parents' superiority is passed on. The other half was due to luck—favorable environmental conditions, or clever combinations of genes that get broken up during reproduction. Thus, $h^2$ is the currency of selection; it dictates the rate of evolutionary change. A plant breeder armed with this knowledge can predict the outcome of a selection program before it even begins, saving years of effort.

This principle extends far beyond traditional farming. In the cutting-edge field of bioengineering, scientists use "directed evolution" to create microbes with novel abilities. Imagine we want to evolve bacteria that can degrade plastic waste. We can measure the degradation rate of different bacterial colonies, select the most efficient ones, and let them reproduce. The Breeder's Equation, perhaps in its more general form using a selection gradient ($\beta$), allows us to predict how much faster the next generation of bacteria will be at their task. From breeding taller corn to engineering plastic-eating microbes, $h^2$ is the quantitative key to sculpting life.

But how do we know the heritability of a trait in the first place? We cannot always conduct a multi-generation selection experiment. Often, we must be detectives, inferring the secrets of inheritance from the patterns we see in nature. The most direct clues lie in the resemblance between relatives.

You have probably noticed that tall parents tend to have tall children. Quantitative genetics puts a number on this observation. By plotting the trait values of offspring against the trait values of their parents, we can find the best-fit line through the data. The slope of this line is a direct window into heritability.

In the simplest case, if we plot the average trait of the offspring against the average trait of their two parents (the "mid-parent" value), the slope of that line is exactly equal to the narrow-sense heritability, $h^2$. If the slope is 0.63, it means that for every unit of deviation the parents show from the population average, their offspring are expected to deviate, on average, by 0.63 units.

More often, we might only have data from one parent, for example, fathers and sons. Because an offspring shares only half of its genes with any single parent, the covariance is halved, and the slope of the regression line turns out to be $h^2/2$. This technique is powerful because it can be applied to any quantifiable trait, including behaviors. By measuring the complexity of songs in father and son warblers, ornithologists can estimate the heritability of this learned, yet genetically influenced, behavior. The simple act of observing families becomes a scientific measurement of evolutionary potential.

The world, however, is a messy place, and our elegant models must be applied with wisdom and caution. The resemblance between relatives is not always purely genetic.

Imagine estimating the heritability of aggressive behavior in fish by comparing mothers and their offspring. You might find a strong correlation, suggesting high heritability. But what if aggressive mothers also provide a more stressful early environment for their young, making them aggressive for non-genetic reasons? This shared environment ($V_{Ec}$) can masquerade as genetics, inflating our heritability estimate. A careful scientist might set up a "common garden" experiment, where offspring are raised separately from their parents in a standardized environment. Often, the "realized heritability" measured from the actual response to selection in such an experiment is lower than the initial estimate from a simple parent-offspring regression. The difference between the two reveals the magnitude of the "ghost" of the common environment.

This brings us to a fundamentally important distinction. The total genetic variance ($V_G$) includes not just the additive effects ($V_A$) that are reliably passed down, but also non-additive effects like dominance (where one allele masks another) and epistasis (where genes interact in complex ways). The broad-sense heritability, $H^2 = V_G/V_P$, tells us what proportion of variation is due to genes in any form. But for predicting selection, it can be misleading. A trait can be highly genetic (high $H^2$) but have very low narrow-sense heritability (low $h^2$) if most of the genetic variance is non-additive. These non-additive gene combinations are like winning lottery tickets—they produce an exceptional result in one individual but are not passed on intact to the next generation. It is $h^2$, and only $h^2$, that measures the reliably heritable variation that fuels evolution by selection.

It is tempting to think of heritability as a fixed, immutable property of a trait. But this is a dangerous simplification. Heritability is a property not of a trait alone, but of a trait in a particular population in a particular environment at a particular time.

Consider a farmer battling a fungal disease in her wheat. She finds that the heritability of disease resistance is very low, say $h^2 = 0.1$. The Breeder's Equation tells her that even with strong selection, the gains from breeding will be painfully slow. Meanwhile, an agronomist finds that changing irrigation practices can produce a much larger and more immediate improvement in crop health. In this context, focusing on improving the environment is the more effective strategy. If the population were moved to a different environment, or if new genetic variants were introduced, the heritability itself could change.

Heritability can even change over an organism's lifetime. The genes that influence an individual's size as a juvenile may be different from those that influence its final adult size. It is entirely possible for early growth to be governed by genes with complex, non-additive interactions (leading to low $h^2$), while later growth is controlled by a different set of genes that act in a more additive fashion (leading to a higher $h^2$). Observing heritability increase as an organism ages gives us a fascinating glimpse into developmental biology—the genetic program unfolding through time.

For most of the 20th century, genetic variance was a statistical abstraction, a "black box." We could measure its effects, but we couldn't see inside. The genomic revolution has changed everything, allowing us to connect the classical concept of heritability to the underlying DNA itself.

We can now search for the specific genomic regions—the Quantitative Trait Loci (QTLs)—that contribute to the additive genetic variance. This endeavor, however, has revealed a stunning truth about the genetic architecture of most complex traits. When narrow-sense heritability is low-to-moderate, and we know that the trait is influenced by many genes, the Breeder's Equation implies a sobering reality: the effect of any single gene must be tiny. A hypothetical QTL for growth rate in salmon might explain only a fraction of a percent of the total variation in the population. This is why "gene hunting" for complex traits is so challenging; we are searching for a multitude of needles in a haystack.

This challenge has led to one of the great puzzles in modern human genetics: the "missing heritability" problem. For many human traits, like height, we can estimate a reasonably high $h^2$ from twin and family studies. Yet, when we add up the effects of all the genetic variants we can identify with our current technology (using genome-wide SNP arrays), the sum often falls frustratingly short of the predicted heritability. The solution to this puzzle lies in the fine print of our technology and the nature of genetic variation. Our standard tools are good at detecting common genetic variants, but they are less effective at capturing the effects of rare variants. If a significant portion of the additive genetic variance is due to many different rare mutations, our current methods will systematically underestimate the heritability captured by the genome, thus explaining the "missing" fraction.

From the farmer's field to the human genome, narrow-sense heritability is more than just a concept; it is a lens. It allows us to see the potential for change, to understand the constraints on evolution, to make practical decisions, and to frame the deepest questions about the connection between our genes and our lives. It is a testament to the power of a simple idea to unify a vast and complex biological world.