Realized Heritability

Why do offspring resemble their parents, and how did humans transform wild species into the diverse domestic plants and animals we see today? The answer lies in the principle that traits are passed down, but only to a certain degree. This measurable "degree" is a cornerstone of genetics known as heritability. It provides a quantitative framework to move beyond simple observation and begin predicting the course of evolution. This article addresses the fundamental challenge of quantifying this inheritance to forecast the outcomes of selection.

This article will guide you through the core concepts of realized heritability. In the first chapter, ​​Principles and Mechanisms​​, we will unpack the elegant breeder's equation ($R = h^2S$), explore the crucial distinction between additive and non-additive genetic variance, and examine the real-world complexities that influence heritability estimates. Following this, the chapter on ​​Applications and Interdisciplinary Connections​​ will demonstrate how this powerful concept is applied, from transforming agriculture and studying natural evolution to providing insights into human medicine and the evolution of culture itself.

Have you ever wondered why children often resemble their parents, but are never perfect copies? Or how our ancestors managed to turn wild wolves into the hundreds of dog breeds we know today, from the towering Great Dane to the tiny Chihuahua? The simple observation is that "like begets like," but the more interesting and powerful truth is that it begets like to a certain degree. This "degree" is not just a vague notion; it's a measurable, predictable quantity that forms the very bedrock of evolutionary theory and modern breeding. This quantity is what we call ​​heritability​​.

Let's imagine you are a farmer with a herd of goats, and you want to increase their milk yield. You start with a herd that produces, on average, 3.20 liters a day. Being a shrewd farmer, you don't just let them breed randomly. Instead, you select only the best milk-producers—say, a group that averages 4.10 liters a day—to be the parents of the next generation. You've made a choice. You've imposed selection. The difference between the mean of your chosen parents (4.10 L) and the mean of your original herd (3.20 L) is a measure of how "picky" you were. This is called the ​​selection differential ($S$)​​. In this case, $S = 4.10 - 3.20 = 0.90$ liters.

Now, what do you expect from the offspring? Will they be superstars averaging 4.10 liters just like their selected parents? Or will they revert to the old herd average of 3.20 liters? The truth, as is often the case in biology, lies somewhere in the middle. Let's say you measure the next generation and find their average milk yield is 3.56 liters. They are better than the original herd, but not as good as the elite parent group. The improvement you achieved—the difference between the offspring's mean (3.56 L) and the original herd's mean (3.20 L)—is called the ​​response to selection ($R$)​​. Here, $R = 3.56 - 3.20 = 0.36$ liters.

So, you asked for an improvement of $S = 0.90$ liters, but you only got an improvement of $R = 0.36$ liters. The relationship between what you ask for and what you get is captured in one of the most elegant and powerful formulas in quantitative genetics: the ​​breeder's equation​​.

$R = h^{2} S$

This equation states that the response to selection is the product of the selection differential and a crucial factor, $h^2$, called heritability. In our goat scenario, we can calculate this value directly from our experiment. This is called ​​realized heritability​​ because it's the value we realize from the real-world results of selection.

$h^{2} = \frac{R}{S} = \frac{0.36}{0.90} = 0.40$

This dimensionless number, 0.40, tells us something profound. It says that for milk yield in this goat population, under these conditions, 40% of the parental superiority was "inherited" or passed on to the next generation. The same principle applies whether you are breeding for more milk in goats, greater thermal tolerance in corals to save our reefs, or larger snouts in dogs.

The beauty of the breeder's equation is that it's not just for looking back at what happened; it's a predictive tool. If you know the heritability of a trait, you can predict the outcome of a selection program before you even start.

Imagine a biotechnology lab aiming to create a super-yeast that can tolerate higher concentrations of ethanol, a key goal for producing biofuels. They know from previous studies that ethanol tolerance in their yeast population has a high heritability of $h^2 = 0.85$. Their current population poops out at about 12.0% ethanol. If they select a group of yeast that can tolerate 14.0% ethanol (a selection differential of $S = 14.0 - 12.0 = 2.0\%$), they can confidently predict the response:

$R = h^{2} S = 0.85 \times 2.0 = 1.7\%$

The next generation of yeast should, on average, tolerate $12.0\% + 1.7\% = 13.7\%$ ethanol. This predictive power turns genetics from a descriptive science into an engineering discipline. It also allows us to work backwards. If a farmer saw their cattle's milk yield increase by 50 liters ($R=50$) and they knew the heritability was $h^2 = 0.25$, they could calculate how much selection they had unknowingly applied: $S = R / h^2 = 50 / 0.25 = 200$ liters. This means their selected parent cows must have been 200 liters above the herd average.

So far, we've treated heritability as a simple conversion factor. But what is it, biologically? To understand this, we need to peek under the hood at the engine of inheritance: genes and variance.

The traits we measure, the ​​phenotype​​, are a product of an organism's genes (​​genotype​​) and its environment. In terms of population variation:

$V_P = V_G + V_E$

where $V_P$ is the total phenotypic variance, $V_G$ is the genetic variance, and $V_E$ is the environmental variance.

But the story gets more interesting, because not all genetic variance is created equal. Think of an individual's genetic value as being composed of two parts. The first part is the sum of the average effects of all its alleles—these are the ​​additive effects​​. If you get an "up" allele, your trait value goes up a bit; get a "down" allele, it goes down. These effects are reliably passed from parent to offspring. The variance in these additive effects across the population is called the ​​additive genetic variance ($V_A$)​​.

The second part consists of ​​non-additive effects​​. These arise from interactions, like a magic password that only works when you have two specific keys together. One type is ​​dominance ($V_D$)​​, where the effect of one allele at a locus depends on the other allele (e.g., the allele for brown eyes masking the allele for blue eyes). Another is ​​epistasis ($V_I$)​​, where the effect of a gene depends on genes at other loci. These interactive effects contribute to making an individual special, but because sexual reproduction shuffles the genetic deck every generation, these specific winning combinations are broken up and not reliably passed on.

This distinction gives rise to two kinds of heritability:

1. ​​Narrow-sense heritability ($h^2 = \frac{V_A}{V_P}$)​​: The proportion of total phenotypic variance due to additive genetic effects. This is the component of variance that makes offspring predictably resemble their parents. It is this heritability that governs the response to selection in sexually reproducing populations. The breeder's equation, $R=h^2S$, is fundamentally about narrow-sense heritability.

2. ​​Broad-sense heritability ($H^2 = \frac{V_G}{V_P}$)​​: The proportion of total phenotypic variance due to all genetic effects ($V_A + V_D + V_I$). It tells us how much of the variation we see is due to genes in the broadest sense.

A brilliant experiment can reveal the difference. Imagine a plant that can reproduce both sexually (via seeds) and asexually (by taking a cutting, creating a clone). If you select the best plants and propagate them by cloning, you are passing on their entire genotype—additive and non-additive effects included. The response to this clonal selection would be predicted by the broad-sense heritability, $H^2$. But if you take those same selected plants and cross-breed them sexually, you are only reliably passing on the additive effects. The response would be predicted by the narrow-sense heritability, $h^2$. In one such hypothetical study, scientists found $H^2 \approx 0.80$ while $h^2 \approx 0.32$. This tells us that a huge chunk of the genetic variation (nearly half the total phenotypic variance!) in that population was due to non-additive effects that are "invisible" to sexual selection, even though they are very much part of the plant's genetic makeup.

At this point, you might be tempted to think of heritability as a fixed, universal constant for a given trait—the heritability of height, the heritability of intelligence, and so on. This is one of the most common and dangerous misconceptions. Heritability is not a constant of nature. It is a statistical description of a specific population in a specific environment at a specific time. Change any of those, and the heritability can change too.

​​1. The Environment is Everything​​ Imagine running a wheat breeding program in a perfectly irrigated, fertilized field (Environment 1). You select for the highest-yielding plants and find they respond beautifully to selection. You then take the seeds from these "superior" plants and grow them in a drought-prone, rain-fed field (Environment 2). You'll likely discover that much of their advantage disappears. The genes that made them winners in the cushy environment might not be the same genes needed to be a winner in a harsh environment. The response ($R$) you measure in Environment 2 will be much smaller for the same selection differential ($S$) applied back in Environment 1. Consequently, the ​​realized heritability​​ is lower. This is a classic example of ​​genotype-by-environment interaction ($G \times E$)​​, a fundamental concept showing that a gene's effect can depend on the world it finds itself in.

​​2. We're Not Perfect Observers​​ Our measurements are never perfect. Let's say you're measuring insecticide resistance in fruit flies, but your automated assay is a bit noisy, introducing random measurement error. This error adds to the total phenotypic variance ($V_P$) but does not change the underlying genetic variance ($V_A$). Since $h^2 = \frac{V_A}{V_P}$, this extra "noise" in the denominator artificially deflates your estimate of heritability. A realized heritability of 0.23 might sound low, but if you discover that 12% of your measured variance is just instrument error, correcting for it reveals the true biological heritability is higher, around 0.26. Understanding the sources of error is part of doing good science.

​​3. The Meddling Observer​​ Sometimes, our experimental design itself can trick us. In an animal breeding study, it's common for parents with "good genes" that lead to higher body mass to also be better at providing care, securing more food, or having better territories for their offspring. This creates a ​​positive environmental covariance​​: better genes are correlated with better environments. This inflates the resemblance between parent and offspring, making it look like the genes are doing more work than they really are, thus leading to an overestimation of $h^2$. Similarly, if the diet of a selected line steadily improves over generations, the response to selection ($R$) will be artificially inflated, not just by genetics but by better nutrition, again leading to an inflated realized heritability. Careful experimental design, such as cross-fostering, is needed to disentangle these effects.

​​4. Evolution's Moving Goalposts: The Bulmer Effect​​ This is perhaps the most subtle and beautiful point. The very act of selection can change the parameter you are using to predict it. Imagine a trait is controlled by many genes. When you select for the highest trait values, you are preferentially picking individuals who happen to have more "up" alleles. But you are also creating non-random associations between alleles at different loci—a phenomenon called ​​linkage disequilibrium​​. Specifically, directional selection builds up a negative association where "up" alleles at some genes become statistically linked to "down" alleles at other genes. This negative covariance subtracts from the total additive genetic variance ($V_A$).

The result is the ​​Bulmer effect​​: as you select on a population, the additive genetic variance (and thus the narrow-sense heritability) actually decreases for the first few generations! It's like trying to run up a sand dune: each step you take displaces some sand, making your next step a little harder. In a selection experiment, the realized heritability in the first generation might be 0.50, but by the third generation, it might have dropped to 0.35, not because of lost genes, but because of these temporary negative associations. Eventually, the force of recombination, which breaks down these associations, will balance the force of selection creating them, and the heritability will settle into a new, lower equilibrium. This reveals heritability not as a static property, but as a dynamic parameter in an elegant evolutionary feedback loop.

Understanding heritability, then, is a journey. We start with a simple, powerful equation, but soon discover layers of complexity and nuance. It teaches us that to predict evolution, we must understand not only the genes themselves, but how their effects are inherited, how they interact with their environment, and how the very process of selection reshapes the genetic landscape on which it acts.

Now that we have built the elegant machinery of the breeder's equation, what can we do with it? Is it merely a neat theoretical toy, or can we take it out into the world and put it to work? It is here, in its application, that the true power and beauty of realized heritability come to life. It is not just an abstract ratio; it is a predictive tool, a diagnostic lens, and a conceptual bridge that connects seemingly disparate fields of science. Our journey through its applications will begin in the farmer’s meticulously managed fields and end in the wild, unpredictable landscape of human culture.

For millennia, humans have been shaping the plants and animals around them through selective breeding. We chose the plumpest grains, the most docile dogs, the cows that gave the most milk. This was an art, guided by patience and intuition. Realized heritability, however, transforms this art into a quantitative science. It answers the crucial question: if I select the "best" parents, how much "better" will their children be?

Imagine a team of agricultural scientists aiming to fortify a wheat crop against increasingly common droughts. They develop a metric for water efficiency and find the average score in their population. They then select only the most resilient plants—say, those with a significantly higher score—to be the parents of the next generation. Will the offspring be just as resilient as the selected parents? Almost never. Will they be any better than the original population? Almost always. The breeder's equation, using the realized heritability, tells us precisely how much improvement to expect. If the realized heritability ($h^2$) for the trait is $0.5$, we predict that the offspring will gain half of the advantage exhibited by their selected parents. This predictive power is revolutionary. It allows breeders to forecast the success of a program, to estimate the number of generations needed to reach a target, and to decide which traits are worth pursuing. The same logic applies to improving the protein content in lentils or increasing the growth rate of farmed fish.

Geneticists have even devised clever experimental designs to get a cleaner, more reliable estimate of this crucial parameter. In a "divergent selection" experiment, for instance, one might establish two breeding lines from the same starting population. In one line, you select for the heaviest seeds, and in the other, for the lightest. By measuring the difference between the two offspring lines after a generation, you can largely cancel out the confounding effects of year-to-year environmental changes, which affect both lines equally. This gives a more robust estimate of the heritable raw material you have to work with.

The principles of artificial selection are but a special case of a much grander process: natural evolution. The breeder is simply a focused and impatient form of natural selection. For an evolutionary biologist, realized heritability is the key that unlocks the relationship between selection acting today and the evolutionary change seen tomorrow.

The traits that can be studied are not limited to agricultural yields. Consider the bioluminescence of a lanternfish in the deep sea. A brighter light might help it find mates but also attract predators. By conducting selection experiments in a controlled setting, biologists can estimate the realized heritability of brightness. This value helps them build models to understand why some populations are brighter than others and to predict how they might evolve under different ecological pressures.

The concept extends seamlessly to one of the most complex and fascinating classes of traits: behavior. Experiments to select for, say, reduced aggression in mice by consistently breeding the most placid individuals can reveal the realized heritability of this behavioral trait. Often, the heritability for complex behaviors turns out to be quite low. A value of $h^2 = 0.03$, for example, tells us that even with very strong selection, the population's average temperament will change at a glacial pace. This is a profound lesson: heritability is not destiny, but it is the governor on the engine of evolution.

Realized heritability also helps us understand the limits of evolution. Imagine a long-term experiment to increase the cellulose content in switchgrass for biofuels. For dozens of generations, scientists select the best plants, and each generation, the cellulose content dutifully climbs. But then, the progress slows, and eventually, it may grind to a halt, reaching a "selection plateau." Why? The breeder's equation gives us the answer. Sustained directional selection uses up the very fuel it runs on: additive genetic variance. By fixing the best alleles at all the relevant genes, selection depletes the heritable variation until there is none left for it to act upon. Realized heritability, which depends on this variance, dwindles towards zero, and evolution stops.

Furthermore, the process is not always smooth. In vast, well-mixed populations, selection can act efficiently. But in a small, isolated population—perhaps founded by a few individuals on an island—another force comes into play: genetic drift, the random hand of chance. An experiment comparing selection in large versus small populations of houseflies might show a steady, predictable response in the large group but an erratic, sputtering response in the small one. In the small population, valuable alleles can be lost by sheer bad luck, lowering the realized heritability and hampering the population's ability to adapt.

Perhaps the most exciting application of realized heritability is when we push its boundaries, forcing us to ask: what does "heritable" even mean? The breeder's equation, $R = h^2 S$, is beautifully agnostic. It doesn't ask how parents resemble their offspring, only that they do. This opens the door to a world of interdisciplinary connections.

Let's start with medicine. Many diseases, like certain heart conditions or autoimmune disorders, are not "on/off" traits. They are binary—you either have the diagnosis or you don't. How can we speak of the heritability of such a trait? Quantitative geneticists solve this by postulating an unobserved, continuous variable called "liability" or "risk." Every individual has a liability score, determined by a mix of genetic and environmental factors. If your liability crosses a certain threshold, you express the disease. By studying the incidence of the disease in a population and among the relatives of affected individuals, it's possible to apply the logic of the breeder's equation to this underlying liability scale. This allows epidemiologists to estimate the "heritability of liability," a crucial step in understanding the genetic architecture of common diseases.

This broader view also helps us navigate tricky methodological waters. If we estimate heritability by comparing parents and offspring who live together, we might get an inflated value. Why? Because they share not only genes but also a common environment—the same diet, the same stressors, the same home. This "common environment effect" ($V_{EC}$) creates non-genetic resemblance. A carefully designed selection experiment, where offspring are raised in a randomized environment, measures the realized heritability. The discrepancy between this value and a simple parent-offspring regression can be used to quantify just how much of the family resemblance is due to nurture, not nature.

The journey gets even more fascinating when we consider that information can be transmitted across generations by means other than DNA. Your body is an ecosystem, home to trillions of microbes that influence your health, metabolism, and even behavior. If you acquire your initial microbiome from your mother at birth, this represents a non-genetic channel of inheritance. The evolutionary response of a population to selection will then depend on the total transmissible variance, which includes both the standard additive genetic variance ($V_A$) and any portion of the microbial variance ($V_M$) that is reliably passed down. The breeder's equation still works perfectly; we just have to expand our definition of "heritable" to include these microscopic passengers.

The final leap takes us out of biology altogether and into the realm of human culture. Can we talk about the heritability of a belief, a skill, or a dialect? Dual Inheritance Theory does just that. Imagine a cultural trait, say, a particular foraging technique. A child might learn it from their parent ("vertical transmission"), which is analogous to genetic inheritance. Or, they might learn it from another skilled adult in the community ("oblique transmission"). The overall "cultural heritability" of this trait—how faithfully it is passed down and how it responds to social pressures—can be modeled as a function of the rates and fidelities of these different learning pathways. This framework uses the core logic of quantitative genetics to understand how cultures evolve.

From predicting the yield of corn to modeling the evolution of ideas, realized heritability proves to be an astonishingly versatile and powerful concept. It is a testament to the unity of science, showing how a single, elegant principle can illuminate the mechanisms of change in any system where the past leaves its mark on the future.