The Typo That Hid for 354 Years
Someone in 1672 made a mistake. It survived in some copies of a book and not others, sitting on page 168, waiting. Scholars knew something was off. Proving it took a browser tab and some computer vision.
Spot the Difference
Here's a puzzle. Two words: callgation and caligation. One appeared on page 168 of some copies of Thomas Browne's Pseudodoxia, printed in 1672. The other appeared on page 168 of other copies. Same edition. Same print run. Different letters.
One is almost certainly a compositor's mistake. The other is almost certainly what Browne wrote. Word game players, you know exactly what to do with this information.
This kind of single-letter divergence hid in plain sight across hundreds of surviving copies for over three centuries. If you've ever stared at a Wordle board convinced the yellow square is lying to you, you understand why this matters.
Six Editions, Thousands of Copies, Infinite Patience Required
Browne first published Pseudodoxia in 1646. He kept revising. By 1672 he had released his final edition, with six editions published during his lifetime in total. Each one represents thousands of chances for a compositor to swap a letter, drop a syllable, or introduce a ghost word into the text.
The first edition alone had an estimated print run of 5,000 copies. At 17th-century printing speeds, that would have taken several days just to complete. Multiply that across six editions and you start to understand the scale of comparing them all against each other.
The traditional tool for this job is an optical collator: a device using pairs of plane mirrors to quasi-superimpose images of two copies so your eyes can catch the differences. It works. It also requires both physical copies in the same room, access to a specialized machine, and the patience of someone who genuinely enjoys finding tiny things wrong.
The New Way
In September 2026, the Oxford Traherne project released version 3.0 of the Traherne Digital Collator, built with the Visual Geometry Group at the University of Oxford. It runs in a browser. No installation. You open it, load your images, and computer vision automatically resizes and overlays pages from different copies, correcting for differences in scale, perspective, and page curvature.
Visualization options include transparent overlay, two-colour channels, and a toggle. You can flip back and forth between two pages printed 354 years apart and watch letters diverge in real time. You can use it in a library with live images of physical volumes, or remotely with digitized image sets from home.
The software is freely available to scholars from other editing projects. Harriet Phillips, working on The Complete Works of Sir Thomas Browne, wrote about using it publicly. Dr Austen Saunders and Professor Rob Whalen gave a seminar on it to the Traherne Online Research Network.
Why Word People Should Pay Attention
Spotting a letter swap on page 168 of a 1672 book is the same skill you use every time you play Scrabble, Wordle, or any game where precision matters and a single letter changes everything. The collation work is just applied at a scale of 5,000 copies over 354 years instead of five rows and six guesses.
Words either mean something or they don't. Scholars chase textual variants because one wrong letter in one copy changes what an author actually said. That's not pedantry. That's love of language taken to its logical conclusion.
Next time you catch a one-letter difference that everyone else missed, know that the same instinct is driving some very serious people in very serious libraries. Their board is just 400 pages and four centuries old.
Source: Languagehat