The Rohonc Codex: 448 pages nobody can read

The Rohonc Codex is a bound manuscript of 448 pages, written in a script that matches no known writing system, with around ninety illustrations. It came to the Hungarian Academy of Sciences in 1838, part of a library donated by Count Gusztáv Batthyány.

It is the Voynich manuscript's less famous relative and in some ways the harder problem.

What is in it

The script uses on the order of two hundred distinct symbols — far more than an alphabet, closer to a syllabary or a system with logographic elements. It runs right to left, and the pages are read bottom to top in at least part of the manuscript.

The illustrations are the most informative part. They show scenes that appear religious: crosses, crescents, a swastika-like symbol, figures in what read as devotional postures, and what look like episodes from a narrative.

The mixture is odd. Christian, Islamic and other imagery appear together, which does not correspond to any obvious devotional tradition.

The forgery question

It is taken seriously by serious people, which distinguishes this case from most.

Suspicion has attached to Sámuel Literáti Nemes, a nineteenth-century Hungarian antiquarian and a documented forger of manuscripts, who was active in exactly the period and the circles the codex emerged from.

Nothing has established the connection. It rests on opportunity and known behaviour rather than on evidence about this object.

What structural analysis can do without meaning

The Király and Tokai work is worth understanding as a method, because it applies to every undeciphered text.

Before asking what a script says, you can ask how it behaves. Which symbols cluster together. Which appear at the start of sections. Whether symbol frequency follows the distribution natural languages produce.

That analysis identified reading direction, recurring sequences that behave like proper names, and a set of symbols that function as numerals — which allowed sections to be matched to a structure resembling a religious narrative.

Where decipherment stands

A long list of attempts, mostly claiming a known language — Hungarian, Dacian, an early Romanian, Hindi, Brahmi-derived scripts. None has been accepted, and most fail the basic test: a proposed reading has to work across the whole manuscript, not on a selected page.

The most substantial recent work is by Levente Zoltán Király and Gábor Tokai, who from around 2018 published a structural analysis identifying the direction of reading, recurring elements, and what appear to be numbers and proper names, and argued the content is a religious text.

That is not a decipherment and they do not present it as one. It is the first work to establish structure rather than assert meaning, which is how these problems get solved when they get solved.