98.6

Carl Reinhold August Wunderlich published Das Verhalten der Eigenwarme in Krankheiten in Leipzig in 1868. He drew on roughly 25,000 patients and more than a million readings taken in the armpit. His conclusion was 37 degrees Celsius. Converted into Fahrenheit, that becomes 98.6, and the conversion invented a decimal place out of nothing.

For over a century the number went unchallenged. Then in 1992 Philip Mackowiak, Steven Wasserman and Myron Levine measured 148 healthy adults and found a mean oral temperature of 36.8 degrees Celsius, with an upper limit of normal at 37.7. They wrote that 37 should be abandoned as a concept relevant to clinical thermometry. Nothing was abandoned.

The comfortable explanation arrived next. Philip Mackowiak, who led the 1992 study, went back through Wunderlich's own thermometry in a 1994 history written with Gretchen Worden. The verdict that stuck was that the old instruments and the way they were read ran high. Nineteenth-century equipment was the culprit, and modern bodies had not changed at all. That account held for a quarter century.

Then Myroslava Protsiv and colleagues compared three cohorts totaling about 189,000 people and found temperature falling 0.03 degrees Celsius per birth decade. Among men born between 1800 and 1997 the drop was 0.59 degrees. The decline appeared within the Union Army cohort as well as between cohorts, so no single instrument change explains it. Bodies cooled.

Two separate findings, and the number survived both. The measurement site moved from armpit to mouth, which should have pushed readings up rather than down. Thirty years after a paper told the profession to drop it, 98.6 is still on the packaging, a Celsius approximation from 1868 wearing a decimal that arithmetic gave it by accident.