It is the year 731, and the platform at Copán is still cold underfoot, the jungle already loud with things you cannot see. A priest stands where his teacher once stood, watching the sun break the horizon at the same point it broke a year ago.
He is not guessing. He is one link in an unbroken chain, taken morning after morning for generations, and what that chain has produced is a number: the length of the solar year, accurate to within seventeen seconds of the figure astronomers use today. No telescope, no clock, nothing a contemporary European would have called an instrument.
The number everyone quotes
Most versions of this story cite a Maya solar year of 365.2420 days. Our current best value for the mean tropical year is 365.24219 days. Line those two up and the disagreement comes to roughly seventeen seconds. For comparison, the Gregorian calendar drawn up in 1582 runs on 365.2425 days, a slightly larger miss in the opposite direction.
No inscription says “the year is this long”, though. The figure is a reconstruction, assembled from the arithmetic the Maya left behind on their monuments, first by a chemical engineer with an unlikely sideline.
Where the figure actually comes from
John Teeple won the Perkin Medal for industrial work on potash. He also spent six years reading Maya stelae across Honduras, Guatemala and Mexico, and in 1930 the Carnegie Institution published the result as Maya Astronomy, a monograph that Mayanists still cite.
What Teeple noticed at Copán was a running tally of error. The Maya civil year, the haab, was fixed at 365 days with no leap day ever inserted, which means the real seasons slide backwards through it at about one day every four years and take 1,507 years to complete a full lap. Copán’s scribes were tracking that slide. Stela A records the seasonal year as having drifted 930 days through the calendar in the 3,844 years since their mythic zero date, and later monuments update the count. Teeple’s verdict was that the Copán figures beat the Julian calendar Europe would still be using nine centuries later, and came out near-identical to the Gregorian one.
He found the same appetite for precision in their lunar arithmetic. Copán settled on 149 lunar months equalling 4,400 days, which works out to 29.53020 days per month. Palenque preferred 81 months to 2,392 days, or 29.53086. The true figure is 29.53059, so the worse of the two rival schools was wrong by about 34 seconds, and the two cities apparently fell out over the difference.
How you time a year with a building
So how do you get a figure that fine with no instrument at all? Generations of people watching the same sunrise from the same spot, and writing down where it came up.
The archaeologist Ivan Šprajc went and measured the spots. Reporting in PLOS One, he surveyed the alignments of 71 of the plaza complexes usually called E Groups, plus 79 other buildings across the central Maya lowlands. Their central axes point at sunrise and sunset on specific dates, separated by intervals that are multiples of 13 and 20 days. Those are the building blocks of the ritual calendar. A pyramid on one side of a courtyard, a long platform on the other, and the sun rising in the notch between them on the morning you have been waiting for.
Šprajc also argues, against a very popular idea, that these were not equinox observatories, because an equinox is a Greek geometrical concept and not something you can actually see happen on a horizon. One paper, and a contested one, but it is built on more measured alignments than anything before it.
The table that kept working for seven centuries
Modest words for a remarkable object: John Justeson described the table to Sky & Telescope as “an accurate predictive model”. He and Justin Lowry had just published a reanalysis in Science Advances of the eclipse table in the Dresden Codex, the bark-paper book now held by the Saxon State and University Library in Dresden.
Eight pages of that codex list 405 lunar months as 11,960 days, which is out by about a tenth of a day. For a century, scholars assumed the Maya simply started a fresh table each time the old one ran out, which would have let errors pile up. Justeson and Lowry catalogued 145 solar eclipses visible in Maya country between 350 and 1150 CE and concluded the tables were instead overlapped, reset early at 223 or 358 months, intervals we now call the saros and inex cycles. Reset at the right moment and the errors cancel. Their reconstruction stays accurate for more than 700 years.
Nobody ever added the leap day
That is the detail I keep coming back to. Every piece of evidence says Maya astronomers knew their 365-day year was short.
They never patched it.
The haab ran 365 days flat, forever, while the seasons drifted through it and a separate body of arithmetic kept score of exactly how far they had travelled.
You can see the same instinct in the Venus pages. Writing in Estudios de Cultura Maya, Margarita Juárez Nájera and Mariana Castellanos point out that the four intervals painted at the foot of each page, 236, 90, 250 and 8, add up to 584 days. That is the nearest whole number to Venus’s cycle. And the whole table runs 37,960 days for a reason: that figure divides cleanly by 584, by 260 and by 365. Everything had to mesh.
Our calendar buries its error in a spare day every February and trusts nobody to look. Theirs left the error in plain sight and kept a running total against it, which is more work and much better bookkeeping. Offered a choice between a calendar that is quietly wrong and one that tells you precisely how wrong it is, I know which set of books I would rather be handed.
