The long view

Are we running out of life expectancy? What the slowdown actually shows

In the longest-lived countries, the pace of improvement in life expectancy has slowed since the 1990s. The observation is rarely disputed; what it implies is disputed sharply.

SnapLongevity9 September 20265 min read

Lives this long have always existed. The scientific argument is about how many people can be brought to them, and whether anything lies beyond.

The disagreement in this field is not mainly about the data. In the countries that already live longest, life expectancy at birth has continued to rise, but more slowly than it did through most of the twentieth century; that deceleration is what the published analyses of the recent record report. What they disagree about is whether it is the approach of a biological ceiling or a plateau that future medicine could break.

That distinction matters more than any prediction, because the two readings are compatible with exactly the same history. Nothing in an observed slowdown, by itself, distinguishes a limit from a pause.

The shape of the disagreement

Conceptual drawing only. No data, country or year is plotted, and the axes carry no values: the point is that the same observed record is compatible with more than one future.

Four things that keep getting confused

In plain English

Life expectancy: An average produced by a life table from the death rates observed in a population, usually in a single year.Why it matters: It is a population statistic. It can rise or fall without anyone's biology changing.
Lifespan: How long an individual actually lives.Why it matters: Individual lifespans vary enormously within any population, including populations with low life expectancy.
Longevity record: The longest individual lifespan for which documentation has been verified. Verification at extreme ages is demanding, and claims are assessed case by case.Why it matters: A record is an observation about one person, not a measurement of what is possible.
Maximum lifespan or biological limit: A hypothesised upper bound on human life imposed by biology, which no dataset has directly observed.Why it matters: Whether such a bound exists, and where, is the actual scientific question.

The demographic argument for a slowdown

The most cited recent statement of the case is Olshansky and colleagues in Nature Aging in 2024. They examined life expectancy in the populations with the longest lives — a set of high-income countries plus Hong Kong, along with the United States — over the period from 1990 onward. They report that the annual rate of improvement in these populations has decelerated compared with earlier decades, and argue that continued radical extension would require reductions in mortality at older ages far larger than anything observed.

Their conclusion is deliberately conditional: on the trajectory observed so far, they consider it unlikely that most people alive in these countries today will reach 100, and they treat the twentieth-century rate of gain as a period effect that has largely run its course rather than a permanent feature. It is an argument about extrapolation, not about biology.

What the analysis cannot establish is what an unobserved intervention would do. A study of past mortality can show where the historical gains came from, and can show that early-life mortality in these countries is now so low that further reductions there have little room to move the average. It cannot rule out a future source of improvement. No demographic method can.

The argument on the other side

Oeppen and Vaupel made the classic counter-case in Science in 2002. Looking at 'best-practice' life expectancy — the highest value recorded in any country in each year — they found it had risen almost linearly for 160 years, at roughly three months per year, with no sign of deceleration. They also documented a long record of published ceilings that were subsequently broken, sometimes within a few years of being proposed. That history is the strongest reason to be cautious about any new ceiling claim.

On individual lifespan rather than population averages, Dong, Milholland and Vijg argued in Nature in 2016 that the maximum reported age at death had plateaued since the 1990s, suggesting a natural limit around 115 years. The paper attracted several published technical rebuttals in the same journal, disputing the statistical treatment of the data and the inference drawn from it. It is not a settled result.

Barbi and colleagues, studying Italian data covering people aged 105 and over in Science in 2018, reported that mortality appeared to level off after about age 105 rather than continuing to climb. If a mortality plateau at extreme ages is real, it weakens the case for a hard wall at a specific age — though the number of people alive at those ages is small, and the finding has been contested on data-quality grounds elsewhere.

Swipe the table sideways to see all columns.

The competing positions and what each rests on
Position	What it rests on	What it cannot establish
Gains are decelerating towards a limit	Slower annual improvement since 1990 in the longest-lived populations, where early-life mortality is already very low and leaves little room for further gains from that source.	That no future intervention will change old-age mortality.
No limit is in sight	Best-practice life expectancy rising almost linearly for 160 years; a long record of proposed ceilings later exceeded.	That a linear historical trend will continue; past extrapolation is not a mechanism.
Individual lifespan has a ceiling near 115	Reported plateau in maximum age at death since the 1990s.	A robust conclusion — the statistical inference drew published rebuttals in the same journal.
Mortality plateaus at extreme old age	Italian data on people aged 105 and over showing levelling mortality rates.	Much about the general population; the number of survivors at those ages is small.

Summary of the sources listed at the foot of this article. The positions are not equally weighted here and are not presented as a balanced pair: they differ in what they measure — population averages in the first two rows, individual maximum lifespan in the last two.

Why life expectancy is the hardest of these to move now

There is an arithmetic reason the twentieth-century pace is difficult to repeat. Life expectancy at birth is an average over deaths at all ages, and a death in infancy removes far more expected years than a death at 85. In a population where early-life mortality is already very low, there is much less scope for large additional gains from reducing it further, so larger future gains would have to come from improvements at older ages, where each one adds fewer years to the average. How the life table produces that average.

This is why 'the slowdown' is not, in itself, evidence about biology. It is what you would expect from any population that has finished the easy part of the problem, whether or not a biological ceiling exists behind it.

What evidence would change the debate

Sustained reductions in mortality at ages above 80 in a large population, which have not so far been observed at the scale required.
An intervention shown to slow ageing itself in humans, measured by survival rather than by biomarkers or animal models.
Better verification of extreme ages, since much of the argument about a ceiling depends on a small number of contested individual records.