On April 26, 1920, in the Baird Auditorium of the U.S. National Museum in Washington, Harlow Shapley and Heber Curtis gave back to back 40 minute lectures on the distance scale of the universe to an audience of National Academy of Sciences members who had just sat through talks on glaciers and zoology. No transcript survives; historians reconstructed the event from Shapley’s annotated typescript, Curtis’s surviving slides, and letters. This episode explains what each man argued and why both were brilliantly right and spectacularly wrong.

Shapley used Cepheid variable stars in globular clusters as standard candles to argue the Milky Way was 300,000 light years across and was the entire universe, with the Sun far from center and spiral nebulae as nearby gas clouds; a separate Andromeda would have to sit an unthinkable 100 million light years away. Curtis countered with a 30,000 light year Milky Way, wrongly placed the Sun at its center, but argued the spirals were island universes, citing Kant, the zone of avoidance caused by dust, and redshift velocities too high for gravitational binding. In the 1921 papers, Shapley cited the 1885 nova that outshone all of Andromeda and Adriaan van Maanen’s observed rotation of the Pinwheel; Curtis rejected the rotation data and proposed a second, brighter class of nova. Edwin Hubble later found Cepheids in Andromeda and settled it, the 1885 event proved to be a supernova, and van Maanen’s measurements were simply wrong.

• How Cepheid variables work as standard candles and why Shapley’s galaxy was ten times bigger than Curtis’s
• The zone of avoidance and Doppler redshift as Curtis’s evidence for island universes
• Why a single 1885 explosion and a spinning Pinwheel galaxy looked like checkmate for Shapley
• Curtis’s bold guesses that turned out right: a brighter class of nova and untrustworthy rotation data
• Hubble’s resolution, today’s estimate of 200 billion to two trillion galaxies, and the anniversary debates in 1995, 1996, 1998, and 2020