How long can humans live? The question has sparked one of the most contentious debates in longevity science. In a correspondence published in Nature on 7 July 2026, the original authors of a landmark 2016 paper have fired back at the latest wave of criticism, insisting that their evidence for a maximum human lifespan of roughly 125 years remains intact.
Brandon Milholland (now at IQVIA, Durham, NC), Xiao Dong (University of Minnesota), and Jan Vijg (Albert Einstein College of Medicine) were responding to a Nature interview in which longevity researcher Saul Newman claimed their 2016 paper, “Evidence for a limit to human lifespan” (Nature 538, 257-259), rested on “basic mathematical errors.”
“Both statements are incorrect,” the trio wrote. They reject the characterization that they argued for a “hard limit to human survival.” Their 2016 analysis, they say, pointed to a natural limit, the age at which survival probability becomes vanishingly small, not an absolute, impassable ceiling.
The original argument
Dong, Milholland and Vijg’s 2016 paper examined global demographic data from the Human Mortality Database and found two patterns. First, improvements in survival with age decline after 100, the rate of progress in extending life slows dramatically at extreme ages. Second, the maximum reported age at death (MRAD) has not increased since the 1990s, when Jeanne Calment died at 122. No one has surpassed her since.
Using log-linear regressions based on the Gompertz mortality model, the authors estimated a maximum human lifespan of roughly 115 years as a central estimate, extended to an outer bound of approximately 125. A 2017 follow-up reply to de Beer et al. was titled: “Maximum human lifespan may increase to 125 years.”
The critique
Newman’s critique, originally published in F1000Research in 2017 with co-author Simon Easteal, identifies three specific mathematical issues. First, variable sample sizes: Dong et al. pooled data across populations with very different sizes without proper normalization. Second, age-biased rounding errors: the Human Mortality Database survival rates (l_x) are rounded to the nearest integer. Above age 90, more than half of these values round to zero, masking actual variation. Third, and most critically, log(0) instead of log(1): when these rounded-to-zero values were used in log-linear regressions, Newman argues that Dong et al. treated log(0) as if it were log(1) (i.e., as zero), since log(1) = 0 but log(0) is mathematically undefined. This, he claims, critically biased the fitted regression lines.
“If you adjust for the straightforward mathematical errors that they made,” Newman told Nature, “the limit disappears.”
A broader debate
The 2016 paper has attracted no fewer than five formal rebuttals published in Nature itself in 2017 (by Brown et al., Hughes & Hekimi, Rozing et al., Lenart & Vaupel, and de Beer et al.), each with replies from Dong et al. The critiques fall into two broad camps.
The statistical camp argues that mortality after age 110 plateaus at approximately 47-50 percent per year, meaning the rate does not continue increasing as the Gompertz model would predict. Rootzén & Zholud (Extremes, 2017), using extreme value theory, concluded flatly: “Data does not support that there is a finite upper limit to the human lifespan.” However, even they found that the chance of anyone living past 128 in the next 25 years is negligible, a de facto limit, if not a mathematical one.
The data-quality camp, led by Newman, argues that many supercentenarian records are unreliable. Newman tracked down approximately 80 percent of people recorded as aged 110-plus and found that “almost none have a birth certificate.” Extreme-age records show clustering on multiples of 5 and 10, a basic demographic data-quality red flag. Newman’s earlier work, which won an Ig Nobel Prize in 2024, showed that purported longevity hotspots (Okinawa, Sardinia, and others) coincide with regions of poor record-keeping and pension fraud.
Where the debate stands
The 2026 Correspondence makes clear that Milholland, Dong and Vijg reject the “hard limit” framing. The trio has previously described their limit as an “effective limit”, the age at which the chance of survival becomes vanishingly small (arXiv:1803.04024, 2018), rather than a mathematical ceiling.
In practice, both sides of the debate agree on more than they disagree about. Whether the limit is framed as an impassable ceiling (Dong et al.’s critics) or an effective practical bound (Dong et al.’s own later phrasing), the empirical reality is the same: no human has lived past 122 in verified records, and the demographic data suggests that exceeding roughly 125 years remains extraordinarily improbable under current biology. The disagreement is whether that improbability reflects a fundamental biological constraint or merely the limitations of current data and sample sizes.
The Correspondence, published in Nature Vol. 655, p. 540 (2026), is short, just two paragraphs, but it reasserts a position its authors have defended for a decade. “All available evidence,” they write, “supports a maximum human lifespan of approximately 125 years.”
Sources
1. Nature, “How long can humans live? All evidence points to a maximum of 125 years” (Correspondence, 7 July 2026). DOI: 10.1038/d41586-026-02111-5
2. Dong, X., Milholland, B. & Vijg, J., “Evidence for a limit to human lifespan”, Nature 538, 257-259 (2016). DOI: 10.1038/nature19793
3. Newman, S. & Easteal, S., “The dynamic upper limit of human lifespan”, F1000Research (2017). DOI: 10.12688/f1000research.11438.2
4. de Beer, J., Bardoutsos, A. & Janssen, F., “Maximum human lifespan may increase to 125 years”, Nature 546, E16-E17 (2017). DOI: 10.1038/nature22792
5. Rootzén, H. & Zholud, D., “Human life is unlimited, but short”, Extremes (2017). DOI: 10.1007/s10687-017-0305-5