Models

The layered word structure does not obviously match the word structure of Indo-European languages. Semitic languages such as Arabic, Hebrew, or Ethiopian could perhaps be transliterated into Voynichese, but not by any straightforward mapping.

- Stolfi

Let us assume there is a plaintext. There are scenarios in which there is no plaintext, but the safer assumption is that there is. Assumption: the plaintext was composed in some natural language and the Voynich text as we have it (the V-text) is some rendering mapped from that original plaintext. It follows: if we understood how the V-text works we can extract and restore the original plaintext in its original language.

The simplest model of the relationship between plaintext and V-text is a direct correspondence between the plaintext language and the Voynich script. In such a model, the plaintext is simply rendered into the Voynich script more or less on a one-to-one basis. If the plaintext is Latin, say, then the V-text is simply Latin written in the Voynich script.

 

This model doesn't work. It doesn't work for every natural language that has ever been tested. Still, undeterred, people keep trying it. They propose, for example, that the plaintext is in Hebrew and that Voynich script corresponds (more or less) letter-to-letter to Hebrew. Therefore: if we reverse the mapping we can restore the V-text to the Hebrew plaintext.

There will always been some words that seem to fit this process. A certain number of Hebrew or quasi-Hebrew words can be found. But they make no sense and have no grammar or syntax. The only 'plaintext' that emerges is a hotch-potch lexicon of unrelated words that are at best Hebrew-ish. Worse, the results are no better than if one had tried, say, Arabic or Greek or Coptic or Catalan.

Every imaginable language has been tried in this manner, and the ones that perform best are languages, such as Polynesean tongues, that are historically and geographically so unlikely as to be not worth considering.

The only other possibility is that the plaintext language is a lost or extinct tongue. But if so, what sort of strange tongue would it be? The Voynich text is, in fact, very unlike natural languages in important aspects. Its observable structures and behaviors do not resemble those of any recognizeable natural language in any obvious way. If it is a case of straight-forward mapping, then the plaintext language must have been very unnatural among natural languages.

There are at least two complicating factors to consider here, though. For a start, there may not have been a very good mapping between the plaintext and the Voynich script. This could produce strange linguistic distortions. The Voynich glyphs might need to stand for several or many letters in a more extensive plaintext, for instance. Or the content - rather than the language - of the plaintext may produce strange effects in the V-text. A text full of numbers and measurements, for example, might be very repetitious in any language.

In general, though, this simple model seems exhausted and unpromising. There are no prospects that if we keep looking we will find an obscure dialect of an obscure tongue somewhere that maps directly from plaintext to V-text. It can't be done.

This predicament necessitates model two:

If there is a plaintext but it hasn't mapped directly to the Voynich text, then we must assume there has been in intervening step between the two. The plaintext has been prepared in some way prior to being transposed into the Voynich script. This preparation amounts to a transformation of the plaintext.

A simple scenario would be as follows:

1. The plaintext is in Latin.
2. The plaintext is heavily abbreviated according to an unknown system.
3. The abbreviated text is rendered into Voynich glyphs (which are a corresponding set of abbreviation marks.)

This is far more promising. But there are any number of possible transformations that might have taken place, for any number of reasons. Ciphers are in this category:

1. The plaintext is in Latin.
2. The plaintext is encrypted with a substitution cipher.
3. The encrypted text is rendered into Voynich glyphs which further hide the plaintext.

The transformation step can be made more elaborate:

1. The plaintext is in Latin.
2. The plaintext is encrypted with a substitution cipher and a large number of nulls are added.
3. The encrypted text - a simple substitution code with lots of nulls - is rendered into Voynich glyphs which further hide the plaintext.

There is a known level of encryption practiced in the relevant period; substitution and transposition ciphers mainly. None are sufficiently complex or sophisticated to produce what we see in the Voynich. But if several methods of simple encryption were to be compounded, then the transformation of the plaintext could be profound.

Encryption is not the only possibility here, though. Textual transformations can have linguistic purposes meant to clarify rather than obscure. Particles, articles, conjunctions and superfluous words might be stripped away, for instance. English is perfectly understandable as: quick brown fox jumped over lazy dog. The motive is economy. This simple trimming might explain why there are so few short words in the Voynich text. Perhaps the plaintext has been streamlined, cleaned up?

We should also remember that medieval texts were far more aural than modern reading texts. They were made to be spoken loud and pronounced more than for silent reading. The plaintext might be transformed for phonological reasons, in order to clarify or emphasize sounds, or to make it more euphonic. Euphonic transformation is a possibility, and perhaps even more likely than encryption.

Phonological factors might be especially relevant if the plaintext consists of material in a language that does not have established writing conventions such as an oral language being brought into writing for the first time.

* * *

Finally, there are also procedural methods - methods by which text is generated or transformed using a set procedure such as a system of cardon grille or a system of wheels (volvelles). Such systems are not necessarily devices to create gibberish. (In that case, of course, there is no plaintext. The plaintext is a phantom.) But it is conceivable that an intelligible plaintext was either composed or transformed by such systems.

An example of this are the systems of Ramon Llull's Ars Magna. Here volvelle wheels are used to create strings of letters which might seem like gibberish but which are like an algebra with meanings that can be extracted by the practice of the "Art". Deploying Llull's methods or similar might produce a text with the sort of combinatorics we see on display in the V-text and other characteristics that suggest procedural generation of some kind.

Procedural generation, or procedural transformation, does not automatically mean hoax. The text can be artificial without being meaningless. In favor of procedural generation are the various letter wheels and lists in the Voynich manuscript itself.  Do they depict a system by which the plaintext was created or transformed?

An example of procedural transformation:

1. The plaintext is in Latin.
2. The plaintext is turned into formulae of symbolic logic using a meaningful arrangement of volvelle wheels.
3. The resulting text is rendered into Voynich glyphs which are designed to facilitate this.

This method would need to be done by an adept of the Art.

Take the English: quick brown fox jumped over lazy dog.
Each of these terms (ideas) can be expressed by the Ars Magna.
The idea of "quick" (quickness, speed) can be analysed and traced to its roots in the Divine Attributes.
Its pedigree among logical ideas is expressed as a series of letters located using the volvelles. (On the element wheel, for example, we find "quickness" associated with FIRE, indicated by the letter P... or whatever.)

Similarly, the quality of the colour brown. And so on. All created things and qualities and actions have their origins in God, as the Ars Magna is designed to show.

In its fullest application the Llullian Ars Magna becomes a type of artificial philosophical language (which is where Liebniz was to take it.) In practical terms, it is as opaque as a complex encryption to one who does not know the Art.

* * *

In any case, if there is a plaintext composed in a natural language it does not reach the V-text unmolested. It undergoes some type of transformation. The evidence suggests a profound transformation. There is a step between the composing of the plaintext and its being rendered in the Voynich script. What has been done at that step and why?

R. B.