A 1996 study of Danish twins helped make a careful, memorable claim about human longevity: genes mattered, but not as much as many people assumed. In that study, DNA explained roughly a quarter of the differences in lifespan between people.

Nearly 30 years later, a new paper argues that the old estimate may have been too low, not because the Danish study was sloppy, but because lifespan is a difficult trait to measure. Some deaths reflect age-related decline more than others. Some are caused by infection, accident, violence, or other outside events that can obscure the biological signal a twin study is trying to detect.

None of us are doctors, geneticists, or clinical researchers. This is a reading of the research and its limits, not medical advice.

The older paper, led by Anne Maria Herskind and published in Human Genetics, examined 2,872 like-sex Danish twin pairs born between 1870 and 1900. It included twin pairs with known zygosity in which both members had survived to age 15, and it used Danish population and death records to estimate how much of the variation in age at death could be attributed to genetic factors.

The answer was modest: 0.26 for men and 0.23 for women. In plain terms, the model estimated that about a quarter of lifespan differences in that cohort were explained by genetic factors, with most of the variation left to individual-specific environmental factors and other non-shared influences.

What heritability does and does not mean

The first thing to avoid is the common misreading of heritability. A heritability estimate is not a prediction for an individual person. It does not mean that 25% of your lifespan is fixed by DNA and 75% is available for lifestyle, luck, medicine, income, geography, or anything else.

Heritability is a population statistic. It asks how much of the variation in a trait, within a particular population at a particular time, can be statistically associated with genetic differences. Change the population, the era, the causes of death, or the model, and the estimate can change.

That matters especially for lifespan. Height is measured while people are alive. Blood pressure can be measured repeatedly. Lifespan is measured only once, at death, and death can arrive for reasons that have little to do with the slow biology of aging. A person hit by a car at 35 and a person who dies from dementia at 96 both contribute an age at death to the dataset, but those numbers do not carry the same kind of biological information.

This is one study, not settled consensus, and the newer work should also be read that way.

The newer argument: outside causes can dilute the signal

In 2026, Ben Shenhar, Uri Alon, and colleagues published a paper in Science arguing that earlier estimates of lifespan heritability had been confounded by what they call extrinsic mortality. Their paper, “Heritability of intrinsic human life span is about 50% when confounding factors are addressed”, used mathematical modelling and twin-cohort analyses to separate deaths driven by intrinsic aging from deaths caused by outside forces such as accidents and infections.

The authors note that previous twin studies often placed lifespan heritability at 20 to 25%, while some large pedigree studies put it far lower. Their claim is that those numbers mix together two things: the inherited biology that influences vulnerability to age-related decline, and the external hazards that can end a life before that biology is fully expressed.

If many people in a cohort die early from infection, unsafe work, poor sanitation, violence, or accidents, the logic goes, then family resemblance in lifespan will be weakened. Genes cannot show their influence on late-life disease if a person never reaches late life.

That is especially relevant to the Danish cohort born from 1870 to 1900. Those twins lived through a world with different infectious-disease risks, medical treatments, workplace hazards, and public-health conditions from those faced by many people today. The 1996 Herskind study was valuable partly because the records were unusually complete, but completeness does not remove the historical conditions that shaped the causes of death.

Why the Danish result still mattered

The Danish-twin paper was not saying that genes do not matter. It was saying the measurable genetic contribution to lifespan, in that cohort and under that model, was moderate. That was an important correction to a simple family-story view of longevity, where a 95-year-old grandparent is treated as evidence that long life is mostly inherited.

It also separated family environment from individual environment. Herskind and colleagues reported no evidence for a strong shared family-environment effect. In other words, growing up in the same family did not explain much of the lifespan resemblance once the model accounted for twin type. Most of the non-genetic variation appeared to be non-shared.

That finding remains useful because it makes lifespan look less like a single trait and more like a final score assembled from many risks. Two people can share genes and childhood conditions, then diverge through infections, occupations, injuries, pregnancies, smoking exposure, medical care, income, stress, diet, social position, and random events. Some of those factors are partly chosen, some are imposed, and some are simply unlucky.

Why the newer 50% estimate also needs care

The 2026 Science paper does not turn longevity into genetic destiny. It argues that if the question is narrowed to intrinsic lifespan, after correcting for outside causes of death, heritability may be above 50%. That is a different question from asking what determines the actual age at which a real person dies in the world as it is.

There is also already debate about the model. A 2026 bioRxiv preprint by Sergey Kornilov, “Omitted familial extrinsic risk inflates inferred intrinsic lifespan heritability”, argues that some familial external risks could be absorbed into the intrinsic component if they are not modelled properly. That critique is a preprint, not peer-reviewed final evidence, but it is a useful warning: separating intrinsic and extrinsic mortality is conceptually clean and statistically difficult.

Another complication is that different older studies have pushed in different directions. A 2018 paper in Genetics by J. Graham Ruby and colleagues, using large public family-tree data, concluded that estimates of longevity heritability had often been inflated by assortative mating, meaning people tend to partner with others who share social, cultural, or health-linked traits. That paper estimated true heritability well below 10% for the birth cohorts it studied.

So the field is not moving from one simple answer to another simple answer. It is asking a more precise question: heritability of what, in which population, under which mortality conditions, and with which assumptions about family resemblance?

The better takeaway

The Danish twin study remains a landmark because it put a number on a question that is easy to mythologize. Its estimate, about a quarter, made longevity look only moderately heritable. The newer Science paper argues that this may underestimate the inherited component of intrinsic aging because deaths from accidents, infections, and other external causes dilute the comparison.

Both things can be true. Genes may matter more for the biological pace of aging than the older all-cause lifespan estimates suggested, while actual lifespan remains deeply shaped by environment, medicine, social conditions, behaviour, and chance.

For readers, the useful point is not that DNA writes the ending in advance. It is that lifespan is not one clean measurement. It is a collision between biology and history. The hard part for longevity science is working out which part of the record belongs to aging itself, and which part belongs to everything that interrupted it.