Around 50% of a person’s lifespan is determined by genetics, a new study suggests, more than doubling previous estimates of the heritability of longevity.

The new research, published in the journal Science, used a carefully designed mathematical model to reach this conclusion. With the model, the team behind the work could account for external causes of death, such as accidents or infections, eliminating these environmental factors from their heritability estimates.

The heritability of different human traits is usually determined using twin studies, which enable scientists to compare individuals who share either nearly 100% or 50% of their DNA. Identical, or “monozygotic,” twins share nearly all of their DNA, while fraternal, or “dizygotic,” twins share only 50%.

The researchers looked at the correlation of lifespan and genetics in individual sets of twins, and then compared how well those metrics matched across many sets of twins. “If a trait is very genetically determined, then the correlation in the monozygotic twins will be much higher than the correlation in the dizygotic twins,” said study co-author Joris Deelen, a geneticist at Leiden University in the Netherlands.

Previous estimates from such studies have placed the heritability of human lifespan between just 6% and 25%, which suggested genetics have a limited influence on how long people live. Those estimates are substantially lower than those for other complex human traits, such as psychiatric disorders, or the heritability of life span observed in other mammals, which are both typically placed at around 50%.

However, observations of long-lived families and the genetic risk associated with age-related diseases, such as heart disease, suggested to Deelen and colleagues that longevity likely has a far larger genetic contribution than scientists once thought.

The difficulty lies in separating drivers of death with strong genetic components — such as the risk of age-related diseases or the speed of physical decline — from external factors, such as accidents and infections. Deelen did note that the divide between these genetic and external factors is not always clear cut; but in the case of infections, for instance, they focused on diseases that are generally very treatable, such as scarlet fever.

“Previously, when we studied lifespan and predictors, we tended to use all-cause mortality, where we’re just looking at what age people died and not really considering what the causes are — cause of death is often missing [from those records],” said Luke Pilling, a geneticist at the University of Exeter in the U.K. who wasn’t involved in the work.