What Has Been Claimed, and What Became Of It A Sumerian prayer, a Vlach war chronicle, a Hindi gospel, and a Catholic prayer book. The four proposed readings do not agree about the content, the language, the century, or even which direction the writing runs.
Four substantial readings of the Rohonc Codex have been published. Three collapsed under examination. One is taken seriously and is still not settled. Set side by side they are the clearest lesson available in how a decipherment goes wrong, which is why they are given here in detail rather than dismissed in a sentence.
The single most damning fact about the first three is not in any of the refutations. It is that they disagree about which way the text runs.
A Sumero-Hungarian prayer refuted
Nyíri turned the pages upside down, identified what he took to be a Sumerian ligature, and matched the remaining symbols to Latin letters they resembled. On that basis he produced a reading whose opening runs:
Eljött az Istened. Száll az Úr. Ó. Vannak a szent angyalok. Azok. Ó. Your God has come. The Lord flies. Oh. There are the holy angels. Them. Oh.
Gyürk, who had done the structural work a quarter century earlier, made the objection that matters: a method this permissive can produce any output the reader wants. If symbols may be matched to letters by resemblance, and the page may be rotated, and the language may be a composite of two unrelated ones, then the constraint that would make a correct answer recognisable has been removed. The reading cannot be wrong, which is exactly the problem. It also cannot be right in any way anyone could check.
A Vlach war chronicle refuted
The most ambitious of the four by a distance. Enăchiuc, a Romanian philologist, read the text as a dialect of vulgar Latin spoken in Dacia, running right to left and bottom to top, and produced a full translation: a history of the Vlachs in the eleventh and twelfth centuries, at war with the Hungarians and the Pechenegs. Her text names places that can be found on a map, including Arad, Dridu, Olbia, Ineu, Rărău, the Dniester and the Tisza, and has a voivode named Vlad in diplomatic contact with three named Byzantine emperors.
Deteti lis vivit neglivlu iti iti itia niteren titius suonares imi urast ucen. In great numbers, in the fierce battle, without fear go, go as a hero.
Two objections sank it. The first is technical and fatal: she transliterated the same symbol as different letters in different places. Whatever patterns the original text contains were therefore destroyed in the transcription, and any regularity in the output is a property of her choices rather than of the manuscript. The second is that the 87 illustrations are overwhelmingly Christian devotional scenes, which have no evident connection to a chronicle of Vlach military campaigns.
Her work did leave one durable mark. The Romanian spelling Rohonczi spread through later writing on the subject because of it.
A Hindi infancy gospel refuted
Singh read the script as an undocumented variant of Brahmi, running left to right and top to bottom, encoding Hindi. His reading of the opening produces a devotional text that parallels the New Testament infancy narrative:
he bhagwan log bahoot garib yahan bimar aur bhookhe hai / inko itni sakti aur himmat do taki ye apne karmo ko pura kar sake Oh, my God! Here the people is very poor, ill and starving, therefore give them sufficient potency and power that they may satisfy their needs.
It was criticised in the same issue of the journal that carried it. The transliteration is not consistent, which is the same defect that sank Enăchiuc, and it appeared alongside its own refutation.
A Catholic prayer book taken seriously
Tokai is an art historian at the Hungarian National Gallery. Király is a computer scientist who began working on the manuscript privately in 2005. They arrived from opposite directions, Tokai through the pictures and Király through the text, found the same things independently, and then published together.
What they report:
- The writing is not a substitution cipher, which is what nearly every earlier attempt assumed. It behaves as a code, in which small groups of symbols stand for whole words and the internal structure of those words is not represented. That is why every frequency-based attack had failed.
- There is a number system with dedicated symbols for powers of ten. Numbers were the way in, because a number in a devotional book is very often a chapter, a verse or a date, and can be guessed from context.
- Structural features recovered before any content: coded INRI inscriptions on the drawn crosses, codes standing for the evangelists followed by numbers, a seven-page table of contents, and chapter headings carrying biblical references.
- Specific passages match identifiable text, including Matthew 25:15, the parable of the talents, and a chronology of the life of the Virgin in years counted from the creation of the world.
- The content as a whole is an ordinary Catholic reader or breviary, mostly paraphrase of the Gospels, with some non-biblical devotional material. A date in the text points to 1593.
They also record damage that anyone who has looked at the Voynich Manuscript will recognise: bifolios missing, leaves swapped and reordered, and a later foliation that misleads rather than helps.
Not because the answer is more plausible, though it is. Because of the order of operations. Tokai and Király found the table of contents, the chapter headings, the evangelist codes and the number system before they could read any of it. Each of those is a structural prediction that could have failed. The three refuted readings each began by choosing a language and then finding it. That is the whole difference.
Király and Tokai describe their own system as far from simple and clean. They give no pronunciation, no grammar, and no full account of the mechanics, and they do not identify the underlying language. The number system has been questioned in review on the grounds that it has no historical precedent and that the equivalences are suggestive rather than compelling. Their identification of the codex as Christian devotional content is widely accepted. The full reading is not, and the paper was announced as the first of three or four.
V. Four Readings, Four Directions
Set the four side by side on the one question that has nothing to do with language, and the state of the field becomes obvious.
| Reading | Direction claimed | Language | Century |
|---|---|---|---|
| Nyíri 1996 | Pages inverted | Sumerian and Hungarian | unstated |
| Enăchiuc 2002 | Right to left, bottom to top | Dacian vulgar Latin | 11th to 12th |
| Singh 2004 | Left to right, top to bottom | Hindi | unstated |
| Tokai and Király | Right to left, top to bottom | not identified | 16th |
Only the last agrees with the reading order that Gyürk established in 1970 and Locsmándi confirmed with a computer in the 1990s, from the justification of the lines against the right margin. Three of the four readings begin by contradicting the one thing about this manuscript that had been demonstrated without reference to any language at all.
A method flexible enough to produce a translation from any input has not produced evidence. If symbols can be matched by resemblance, the page can be rotated, the same sign can be read as different letters in different places, and the language may be any of several, then a fluent output is guaranteed and means nothing. The test of a proposed reading is not whether it produces sense. It is whether it could have failed.