The Illusion of Heritability: How Redefining Concepts Can Transform Scientific Understanding

When we ask how heritable a trait is, we're usually looking for a simple percentage. But as this provocative analysis reveals, that percentage can change dramatically depending on how we define the very concept of heritability itself.

The Redefinition Game

The article begins with a clever thought experiment about hair color. We know that identical twins almost always share the same hair color, but the age at which they go gray can differ due to lifestyle factors and random biological variation. However, if we redefine "hair color" to exclude factors like graying and dyeing, suddenly hair color becomes 100% heritable. This isn't wrong per se—heritability is indeed an arbitrary concept—but it reveals how easily we can manipulate our understanding by changing definitions.

The same logic applies to IQ. In our world, children face vastly different environments: some receive private tutoring and organic food while others contend with lead paint and limited resources. These environmental differences affect IQ scores. But if we redefine "IQ" to mean "IQ in a hypothetical world where every child received identical education, nutrition, and parenting," the heritability of IQ would necessarily increase. The article's central insight is that heritability isn't a fixed property of a trait but rather a ratio that depends entirely on the context we choose to measure it in.

The Lifespan Study: A Case Study in Redefinition

The paper under scrutiny, published in Science, attempts to estimate the heritability of human lifespan using a novel approach. Traditional twin studies suggest lifespan is 23-35% heritable, based on correlations between identical and fraternal twins. The researchers built a mathematical model to simulate lifespans in a hypothetical world where no one dies from non-aging-related causes—no car accidents, murders, drug overdoses, or infectious diseases.

In this simulated world, heritability jumps to 46-57%. The paper's title and framing suggest this reveals the "true" heritability of lifespan, but the article argues this is misleading. What the study actually shows is that in a hypothetical world without extrinsic mortality, heritability would be around 50%. This is tautological: if you eliminate all non-genetic sources of variation in lifespan, what remains must be genetic.

The Mathematical Model

The researchers created an equation modeling the probability of death at any given age, with parameters varying between individuals to reflect genetic differences. They assumed these parameters follow a Gaussian distribution that, when simulated, produces death ages matching historical data. The model includes an "extrinsic mortality" parameter representing deaths from accidents, violence, and disease.

By adjusting this parameter, they could simulate different worlds. Lower extrinsic mortality means less non-genetic randomness in lifespan, so heritability increases. The clever aspect is that this allows estimation of how heritability might differ in modern times versus historical periods, since non-aging-related deaths have decreased due to medical advances.

The Science Publishing Problem

The article criticizes Science's publishing style, which prioritizes accessibility over technical detail. The paper contains almost no equations, and the appendix provides disconnected explanations rather than a systematic walkthrough. This makes reproduction nearly impossible and wastes readers' time trying to decipher what was actually done.

The Real Takeaway

The article's main objection isn't to the methodology but to the framing. Pretending that car accidents and murders are "confounding factors" that need to be "adjusted for" misunderstands what heritability is. These deaths aren't statistical noise to be removed—they're real parts of how lifespan varies in our world. There is no Platonic "true" heritability independent of our actual circumstances.

However, the study does have value. It provides a way to estimate how heritability might differ across time periods or hypothetical scenarios. The article suggests a more accurate title would be: "Heritability of intrinsic human lifespan is about 40% when extrinsic mortality is adjusted to modern levels, according to our simulation."

Broader Implications

This analysis reveals several important points:

First, heritability is not an intrinsic property of a trait but a ratio that depends on the population and environment being studied. Second, scientific conclusions can be dramatically affected by how we choose to define and measure concepts. Third, the prestige publishing system sometimes prioritizes headline-grabbing claims over clear, reproducible methodology.

Most importantly, this reminds us that when we talk about the "heritability" of any trait, we're making implicit assumptions about what we're including and excluding from our definition. The number we get depends entirely on those choices—and different choices can lead to very different answers.

The article concludes that while the mathematical model is clever and the question is interesting, the presentation obscures more than it reveals. In a world where scientific findings are often reduced to clickbait headlines, understanding how these numbers are generated—and what assumptions they rest on—becomes crucial for meaningful scientific literacy.