An interactive showing · Ground truth

The Hand He Cracked for Fifty Years

Your mother said cracking your knuckles gives you arthritis. One doctor cracked his left hand at least twice a day for fifty years — ≥36,500 cracks — and left his right hand alone as a control. No arthritis in either. His n is one; the larger studies report no increased arthritis association. Below, use the 1990 study's group sizes to see how often a hypothetical effect would be detected.

Someone told you, probably before you were ten, that cracking your knuckles would give you arthritis. It is one of the most durable warnings in the household canon — repeated with such confidence, by people who loved you, that it acquired the texture of a fact. It is not one. And the most charming refutation in the medical literature is a single doctor who decided, as a boy, to test his mother on it, and then kept testing for half a century.

His name was Donald L. Unger, a physician in Thousand Oaks, California. As a child he was told by relatives that cracking his knuckles would lead to arthritis. So he ran the only experiment a stubborn child can run on his own hands: he cracked the knuckles of his left hand at least twice a day, every day, for about fifty years — and cracked his right hand rarely, keeping it as a control.¹²

The count he kept

The number is the thing, so let us build it in the open. Two cracks a day, three hundred sixty-five days a year, fifty years — 2 × 365 × 50 = 36,500. The phrase at least is doing honest work: "twice a day" is a floor, and the round 36,500 leaves out leap days. For illustration, 1 January 1948 to 1 January 1998 holds 13 leap days: 18,263 days, or 36,526 cracks at two a day. Those are not Unger's documented start and end dates; his stated minimum is 36,500.

After fifty years, Unger examined both hands. His finding, in full: "there was no arthritis in either hand, and no apparent differences between the two hands." His conclusion: "there is no apparent relationship between knuckle cracking and the subsequent development of arthritis of the fingers."¹ He published it in 1998 as a one-page letter in Arthritis & Rheumatism, the journal of the American College of Rheumatology. In 2009 it won him the Ig Nobel Prize in Medicine, for achievements that "first make people laugh, and then make them think."³ In his acceptance, he reportedly addressed his mother.

Why one hand can't settle it — and what does

Here is where the clean story needs a brake. Unger's experiment is n = 1. One man, one hand against the other, unblinded, with the experimenter and the subject the same delighted person. This one unblinded end-point comparison supplies no population-level test of arthritis risk. Other single-subject designs can support statistical tests; this recorded comparison cannot, alone, establish a general safety claim.

It is not the only evidence. In 1990, Castellanos and Axelrod examined 300 consecutive patients aged 45 and over — 74 habitual knuckle crackers and 226 non-crackers — and looked for arthritis of the hand. They found no increased arthritis among the crackers.⁴ The anecdote and the cohort point the same way — which is exactly the configuration in which you are allowed to believe an anecdote. But "no difference found" is a slippery result. A study can miss a real effect even with good power: 80% power means detection in 80% of repeated studies under the assumed effect, and a miss in 20%. So one useful question is: if cracking really did raise arthritis risk, how often would a study with these group sizes catch it?

Answer it yourself. Below is the real study geometry — 74 crackers, 226 non-crackers. Suppose a baseline arthritis rate, then suppose cracking adds some extra risk on top. The panel draws independent binary outcomes for hypothetical patients, runs a two-proportion test, and reports how often it detects the effect. That detection probability is power; this is a model, not a reanalysis of the original observations.

Push the effect up and the curve climbs. Simulated detection frequencies fluctuate near the approximate curve, with both sampling and approximation error. Two readings matter. First, at the far left — set the extra risk to zero, so cracking does nothing — the study still "finds" a difference about 5% of the time. That is the model's approximate false-alarm rate at the nominal 0.05 threshold. Second, the curve is shallow near the bottom: a small, few-point effect is nearly invisible to a study this size. The number that matters is where the curve crosses 80% — the smallest effect the study could reliably catch.

At a hypothetical 30% baseline, this model with 74 crackers and 226 non-crackers has about 80% power at an extra +18 points, and lower power for smaller effects. That describes detection across repeated hypothetical studies. It does not tell us what the actual study's null result rules out: that needs analysis of the observed data, such as an interval for the difference. Low power for a few-point effect at this baseline is not a claim that every possible study of this size must miss it, or that no observed result could exclude it. Nor does this observational comparison establish causal absence.

The honest caveat the cohort found

The careful version of "knuckle cracking is harmless" is not "knuckle cracking does nothing." Castellanos and Axelrod did find something — just not arthritis. The habitual crackers in their study were more likely to have hand swelling and lower grip strength, and cracking was associated with manual labor, nail-biting, smoking, and drinking.⁴ The authors went so far as to conclude that habitual cracking "results in functional hand impairment." That is a real finding and it belongs in the answer. The defensible claim is narrow: no evidence that cracking your knuckles causes arthritis — the thing the myth actually asserts — with a documented hint that habitual cracking may track with weaker, puffier hands. Not "proven harmless." Just: not the disease you were warned about.

What the pop actually is

If the cartilage is not being ground down, what makes the sound? For decades the textbook answer was a collapsing gas bubble. In 2015, a team led by Greg Kawchuk watched it directly: they slid a volunteer's finger into an MRI and pulled until it cracked, filming the joint in real time. The study is titled "Real-Time Visualization of Joint Cavitation," better known as the "Pull My Finger" study.⁵⁶ The film showed the pop coincides with a gas-filled cavity forming in the synovial fluid as the joint surfaces suddenly separate, within one imaging frame of 310 milliseconds. The authors interpreted cavity formation, or tribonucleation, as the source of the sound. Its source remained debated: a 2018 model supported partial bubble collapse. The MRI observation does not establish absence of harm.

The record

What is documented, what is standard, and what is genuinely assumed — so you can weigh each claim at its true strength.