A record-correction · Ground Truth

The Yawn That Was Never About Oxygen

Everyone “knows” you yawn because your blood is low on oxygen. Someone checked — in 1987 — and it simply isn’t true.

The oxygen story makes a prediction you can test: if a yawn tops up your blood gases, then flooding the blood with oxygen should stop the yawning, and loading it with carbon dioxide should set it off. Below is a schematic of that experiment. Move the dials to explore the reported contrast; the bar heights and intermediate settings are illustrative.

The blood-gas bench · illustrate the 1987 result

Two dials, two schematic responses. The study tested air, pure oxygen and 3% or 5% CO₂; exercise was tested separately.

21% · air100% · pure O₂
0.035% · 1987 air5% · ×143 the 1987 baseline

The oxygen story’s prediction is the dashed ghost. The solid bar is a schematic of no significant yawning change at the tested conditions. Its fixed height is not a measured rate. The breathing gauge includes the reported increase under pure oxygen; all bar heights and intermediate or combined settings are illustrative.

Here is the result, and it is the entire correction. Breathing responded to every push — carbon dioxide, pure oxygen, exercise — exactly as the chemistry says it must; carbon dioxide, not oxygen, is the gas your brainstem actually watches. Yawning sat it out. Two systems the folk theory calls one pulled apart in front of the experimenters. Whatever a yawn is for, it is not for the air.

Then why do we yawn?

This is where the honest page has to slow down and refuse to trade one confident story for another. The best-supported positive account is brain-cooling — a yawn’s deep inhalation and jaw-stretch draw cooler blood and air toward a brain that runs best in a narrow thermal band. Gallup & Gallup (2007) produced actual numbers. They are real measurements, and they are not the final word.

The cooling evidence · two measured contrasts (Gallup 2007)

Leading hypothesis; not settled. These are two separate experiments; the four displayed summaries come from seven groups: four breathing groups and three pack groups. The 48% and 41% summaries pool comparison groups.

Cool the incoming blood: breathe through the nose

48% caught a contagious yawn (mouth / undirected)

Cool the forehead: a pack on the brow

41% caught one (warm 46 °C / room-temp pack)

Cool the head; by either route; and the yawning thins out. That points at temperature, not air. It is coherent and supported; its causal interpretation is contested. This page states it as the leading candidate, not a fact.

The check — what’s true, what’s illustration

True and sourced. Provine, Tate & Geldmacher (1987) had people breathe 100% oxygen, then air at 3–5% carbon dioxide; up to ~143× the CO₂ in ordinary air (5% ÷ the 1987 ambient of ~349 ppm); then exercise hard enough to double their breathing. Breathing rate rose every time; yawning showed no significant change. That dissociation is the citable refutation. Study figures and dates are transcribed from the cited sources; selected CO₂ multiples, percentage differences and ratios, and year spans are calculated offline in research/the-gasp-that-isnt-for-air/verify-the-gasp-that-isnt-for-air.mjs.

What’s illustration, not data. The breathing gauge illustrates the reported increases under oxygen, carbon dioxide and exercise; its magnitudes are chosen for display. The dashed ghost illustrates the oxygen story’s prediction, with an invented magnitude. The fixed yawning bar schematizes no significant change at the tested conditions. It is not an exact measured equality. Intermediate concentrations and combined manipulations were not tested in this experiment.

The asymmetry — the whole correction. What’s certain is the negative: yawning is not a blood-gas reflex — a direct experiment with a clean dissociation. What’s not settled is the positive: brain-cooling is the leading candidate, still argued. The thing everyone is sure of — oxygen — is exactly the thing that was tested and failed. Nothing on this page computes the cooling theory true.

The record

  1. 1987The oxygen story is refuted. 100% O₂ and 3–5% CO₂ move breathing but not yawning; exercise doubles breathing, yawning holds flat. The two responses dissociate.Provine RR, Tate BC, Geldmacher LL. Behavioral and Neural Biology 48(3):382–393. PMID 3120687.
  2. 2007Brain-cooling is proposed, with numbers. Nasal breathing drops contagious yawning 48% → 0%; a cold (4 °C) forehead pack drops it 41% → 9% versus warm/room.Gallup AC, Gallup GG Jr. Evolutionary Psychology 5(1):92–101.
  3. 2012The cooling theory’s lead author and a co-author defend the model and rebut its critics. This page’s assessment is leading, not settled.Gallup AC, Eldakar OT. Frontiers in Neuroscience 6:188.
  4. 2026Thirty-nine years on, the refuted answer is still the one you’ll hear first. The correction never travelled as far as the error.Elapsed-time arithmetic, 1987 → 2026.

The oxygen idea never came from a measurement. It is a back-formation from one visible fact — a yawn is a big inhalation, and big inhalations are how we get oxygen, so a yawn must be about oxygen. The shape of the act mistaken for its purpose. It survived not because it was confirmed but because it was repeated, until the repetition itself felt like evidence. The one thing the record fully licenses is what a yawn is not.

Sources

Corrections

Corrected 2026-10-14: The page described brain-cooling replications as mixed. The cited review reports replications and disputes over interpretation; it does not establish mixed replication outcomes (found by the assay line, 2026-10-14).

Corrected 2026-10-14: The page called 0.035% CO₂ ordinary air. This is a rounded historical baseline: NOAA's 1987 Mauna Loa annual mean was 349.31 ppm, or 0.034931% (found by the assay line, 2026-10-14).

Corrected 2026-10-14: The page called 5% CO₂ about 140 times the air the reader is breathing now. That comparison belongs to the 1987 baseline; NOAA's September 2026 Mauna Loa monthly mean of 426.07 ppm gives about 117 times (found by the assay line, 2026-10-14).

Corrected 2026-10-14: The page said brain cooling explains fetal yawning. The cited review treats fetal yawning as neutral on the proposed postnatal function (found by the assay line, 2026-10-14).

Corrected 2026-10-14: The page said brain cooling explains yawning in fish. The cited review proposes a possible role but leaves that function unestablished (found by the assay line, 2026-10-14).

Corrected 2026-10-14: The bench called its fixed yawning bar a measured ×1.00 response, including at untested settings. Provine reported no significant yawning change at the tested conditions, not exact equality; the bar is now labeled schematic (found by the assay line, 2026-10-14).

Corrected 2026-10-14: The cooling panel said only four values were measured. Its four displayed summaries come from seven groups in two experiments; the 48% and 41% summaries pool comparison groups (found by the assay line, 2026-10-14).

Corrected 2026-10-14: The bench kept breathing at resting under pure oxygen. Provine reported increased breathing under pure oxygen too; the gauge now shows that direction with an explicitly illustrative magnitude (found by the assay line, 2026-10-14).

The claims record 166 claims re-read against their sources, 14 October 2026

Written 2026-06-27. Claims re-read against their sources on 2026-10-14: 166 checked, 117 confirmed, 47 wrong, 0 unverifiable, 2 true when written and overtaken since (dated, not corrected). By the assay line (research/assay-line/queue-1014c/), Codex fixer directed by claude-assaying-boyle-1014; source record by Codex, checked by Codex and a Claude agent.

This was the directed repair pass for the 49 items in research/assay-line/queue-1014c/the-gasp-that-isnt-for-air/triage.json, using 2026-10-14 as the requested pass date. The fixer re-read the evidence before changing the authored markdown, hand-authored interactive page and layer verifiers, and retained the checkers' supported confirmations with their original scope. Each triage item has its own claim and disposition; duplicate findings share correction lines. MINOR is included in wrong for the counts. Non-triaged disagreements between the checkers are not promoted to CONFIRMED. The generated placard date remains pending the shared landing rebuild, so those two STALE items remain open rather than being reported as dated on the page.

This record is long (166 claim lines and 49 findings), so it is not part of the page until you open it. It loads here now if your browser runs scripts; otherwise, or to keep it, read the whole record as plain text.

The assay office: what a record is, and every layer re-read so far.