Olympic records keep flattening as athletes approach a physiological ceiling

Florence Griffith Joyner ran 10.49 in Indianapolis in 1988. Nobody has run faster since. The men’s long jump record (Mike Powell, 8.95 metres) is from 1991. Jürgen Schult’s discus mark of 74.08 metres goes back to 1986. Not all of these records are clean, and the ones that are clean were thrown by athletes whose successors, despite better food, better recovery, and better data, cannot quite catch them.

A 2008 paper in PLOS ONE argued that this is not coincidence. Geoffroy Berthelot and colleagues at the Institut de Recherche Biomédicale et d’Épidémiologie du Sport in Paris fitted a single equation to 3,263 world records spanning 147 quantifiable Olympic events from 1896 to 2007. Their conclusion was that records had already reached 99 percent of an estimated asymptotic ceiling, and that “half of all world records will not be improved by more than 0.05 percent in 2027” if the present conditions held. A recent Washington Post analysis of progression curves in swimming and athletics walked the same road from a different start. Sixteen years of data have not contradicted Berthelot.

The paper, called “The Citius End,” fitted a piecewise exponential decay to record progression with a mean adjusted r² of 0.91 ± 0.08. Records sat at roughly 75 percent of their estimated physiological maximum in 1896. They sat at 99 percent in 2007. Most of what most people think of as modern sport, including the wave of professionalisation, sport science labs, and aero everything, happened during the last 1 percent.

What “asymptote” means in athletes

There is no wall. The slope just goes shallow enough that the next sliver of improvement gets buried inside the noise of measurement, weather, altitude, and whichever generation happens to produce a runner with the right tendons. In a 2010 follow-up paper, again in PLOS ONE, Berthelot’s group ran their model over 41,351 elite track-and-field and swimming performers across 70 events. They were looking past the records themselves at something they called “atypicity,” basically a measure of how exceptional any given year’s top performers were against the historical distribution.

That curve had four bumps over the modern era. The last bump was in 1988. After 1988, 64 percent of T&F events have not improved in any sense the model can distinguish from random fluctuation. The 1988 line is also where the doping bookkeeping gets messy, of course. The late 1980s sat near the high-water mark of the East German programme. Out-of-competition testing under WADA only arrived in 1999. Some of the records set in that window are explicit pharmaceutical artefacts. Several have not been touched since. Griffith Joyner’s 10.49 is one of them.

A pattern that holds across species

A 2015 review in Sports Medicine, this time by Berthelot, Sedeaud, Marck and a wider list of co-authors, took the question past humans. Greyhounds, thoroughbred horses, and laboratory frogs all show the same sigmoid performance curve when their measurable outputs are graphed over time. The reading the authors went with was that whatever is going on, it is biological rather than cultural. Bone, muscle, tendon, and the cardiovascular plumbing behave like a constrained optimisation problem. Athletes with the right phenotypes for a given event have already been selected, and the morphological niche is more or less full.

Their estimate of residual progress was around 1 percent per Olympic cycle, with each gain smaller than the last. Most of the bigger mid-century jumps came from population growth, professionalisation, doping, and equipment innovation, all of which have run out of headroom. Swimming briefly broke the pattern in 2008 and 2009, when polyurethane suits added an estimated 3 percent across distances. FINA banned the suits in 2010 and swimming progress reverted to the same flattening curve as everything else.

The 2024 update

Thorsten Emig and Guillaume Adam, writing in Frontiers in Physiology last year, extended the Berthelot framework to running records through the most recent Olympic cycle. Their mathematical fit again separates linear-looking trends from genuine asymptotic approaches. The conclusion was the same. Events that look like they are still progressing are mostly oscillating around a flat ceiling. A separate calculation cited inside the Washington Post analysis put the theoretical floor for the men’s 100 metre at roughly 9.44 seconds, which is 0.14 seconds below Usain Bolt’s 2009 mark of 9.58.

What the body actually runs out of

Three constraints sit underneath the flattening curve. The first is morphology. The optimal body for a given sport is a narrower target than it used to be, and once those proportions fill in, picking the next outlier from a deeper talent pool yields fractional gains. The second is metabolic. Elite VO2 max scores in the published literature have not budged for decades, and the energy delivery system bounded by mitochondrial density and capillary supply does not respond to training the way muscle volume does. The third is connective tissue. Tendons and ligaments do not adapt linearly to the loads modern training imposes, which is part of why elite injury rates rise as athletes push closer to the edge.

One place the framework leaves room for surprise sits inside the longevity research community’s wheelhouse. As lifespan extends and mid-life biology improves, the career envelope of an elite athlete is widening. Tom Brady playing American football into his mid-forties, or Roger Federer playing competitive tennis at 40, are not record-breaking events in the Olympic sense, but they do push back what is possible at a given age. Berthelot’s model does not account for that aging-related expansion. It will not look like a faster Bolt. It might look like a 45-year-old who still trains like a 30-year-old.

Bottom line

The graphs in the Washington Post feature look depressing if you watch sport for the records. They look like physics if you watch it for the biology. Berthelot’s group made a quantitative prediction in 2008 about where the asymptote sits and how fast records would slow as athletes approached it. The data since has not seriously contradicted them. Whether the next breakthroughs come from gene therapy, carbon-fibre exoskeletons, or new event categories, the peer-reviewed literature is clear that they will not come from another generation of training stress applied to the same body.

References

1. Berthelot G, Thibault V, Tafflet M, Escolano S, El Helou N, Jouven X, et al. The citius end: world records progression announces the completion of a brief ultra-physiological quest. PLOS ONE 3(2):e1552. 2008. https://doi.org/10.1371/journal.pone.0001552
2. Berthelot G, Tafflet M, El Helou N, et al. Athlete atypicity on the edge of human achievement: performances stagnate after the last peak, in 1988. PLOS ONE 5(1):e8800. 2010. https://doi.org/10.1371/journal.pone.0008800
3. Berthelot G, Sedeaud A, Marck A, et al. Has athletic performance reached its peak? Sports Medicine 45(9):1263-1271. 2015. https://doi.org/10.1007/s40279-015-0347-2
4. Emig T, Adam G. Modeling running record progressions and physiological limits. Frontiers in Physiology 15:1372092. 2024. https://doi.org/10.3389/fphys.2024.1372092