Agnostic may seem an odd word to describe research on the Voynich manuscript, but it is the name for the way I approach the manuscript’s text. It is my paradigm to help me understand what I should—and should not—do when seeking to ask and answer questions. It is about what kind of knowledge is useful, but also what kinds are not. Agnostic is descriptive of the kinds of knowledge I deliberately seek to leave out, often in contrast to other researchers who consider such knowledge essential.
I have thought about this agnostic paradigm for a long while, and have been deeply influenced by Pallottino’s combinatorial method approach to Etruscan. It is also informed by my understanding of how other ‘decipherments’ of unknown scripts have worked. Despite this, I do not believe my paradigm is fully worked through and without flaws. But I wish to present it in order to show how I work and counter some common criticisms of linguistic analysis as a whole and errors made by other linguistic researchers.
The spark which has caused me to write about the paradigm I use for research into the text of the Voynich manuscript was the characterization by Nick Pelling of linguistic research as, “extract[ing] a set of individual word cribs to identify the language’s family”. As somebody who believes in the possibility that the manuscript is written in the plain in an unknown language and in an unknown script (that is, neither enciphered or encoded), my research does not work in the way described. Or, at least, the process he describes makes up the final step, taken only once a great deal of groundwork has been laid.
However, I first must agree with Nick that the kind of linguistic method he describes is unlikely to work. Although it has been used successfully in many previous ‘decipherments’ of ancient or unknown scripts, it has done so in very different circumstance. Suggesting the possible meaning of individual words and working to fit them to the script succeeds only when certain aspects of the text can be controlled. To explain why they do not work on the Voynich manuscript—and show how my paradigm differs—I first must take a detour through how the reading of unknown scripts has worked in the past.
When confronted by a text written in an unknown script we tend to make several initial and very useful assumptions: the text is meaningful, it is linguistic, and there is some method to the way it has been written. The ‘linguistic’ assumption is true even if the text is a cipher. Early substitution ciphers were solved by character frequency which is, in essence, a linguistic feature. Poe’s The Gold–Bug gives a great example of a simple cipher solved by the knowledge the reader has about the English language.
The task becomes harder when we know neither the script nor the language. Hence why bilingual inscriptions are so useful: they supply the content of the message, allowing us to reconstruct the relationship between different parts of the message, and provide a check on a trial–and–error approach. The presumed similarity between the meaning of the two texts hints at the solution and will confirm or disprove it. The Palmyrene script was independently solved by two scholars a few days after a set of bilingual inscriptions were published.
Yet some of the most famous ‘decipherments’, such as Egyptian hieroglyphics, were made when the script, the language, and the content, were basically unknown. Although scholars knew that ancient Egyptians likely spoke an earlier version of the Coptic language, and the Rosetta stone had been discovered, these weren’t actually the key to working out the script. The key was guessing that cartouches must contain proper names—specifically of kings and queens—and assuming that those names would be recognizably the same as in Greek. With these the meaning of a word could be narrowed to just a few possibilities (Ptolemy, Cleopatra, or so on), making the language irrelevant with the focus on proper names. The script, then, was the only things in question, and must yield an outcome within very defined limits.
This process is more of less how Stephen Bax presumes to work on the Voynich manuscript (I choose Bax’s work because it is a current example of linguistic research, but others have the same problems). By using the pictures of plants on herbal pages he can guess the possible names and attempt to find those names in the text on the same page. But this is a fraught process because of the lack of certainty. Not only are many plants insecurely identified, but most have multiple possibilities from different languages. Furthermore, it is hard to know exactly where in the text the name of a plant will occur.
The method thus neither controls the content nor the language in a way which allows confident working in the script. A researcher will still have to guess which language to draw a name from and which word in the text to apply that name to. Even if you think you can read the first word of a page as ‘juniper’ in Hebrew, others could well read the second of the same page as ‘cypress’ in Greek, the third word as ‘marigold’ in Turkish, and so on. Whichever one you choose, sooner or later the results will begin to contradict one another or lead to a dead end where the only yield is a meaningless jumble of letters. And because the language and the words were guesses, without any kind of control, there is no way of knowing when or where mistakes were made.
So what about my paradigm? You may have already guessed that I have called it agnostic because it seeks to render, as far as possible, the researcher agnostic to the issues of content and language. The question of which word means what, and in which language, are left to the very end and asked only once the underlying language is better understood.
At the heart of the paradigm are two needful assumptions: the text is linguistic and written in the plain. These are not assumptions that every researcher would share, and so I must stress that the paradigm is only for linguistic research. If you believe that the Voynich text is a hoax or a cipher then this paradigm may profit you little. However, if it yields results—or fails to—that will ultimately be useful to proving or disproving linguistic theories as a whole.
With these two principle assumptions it is possible to state that the characters of the Voynich script should represent linguistic facts, and the relationships between the characters of the script should represent relationships between those linguistic facts. Also, due to our understanding of known writing systems, the size of the character set of the script suggests that those linguistic facts are sounds. That is, characters represent sounds.
All natural languages use sounds in specific ways, either universal principles or general tendencies. And within a language sounds are used in ways which, while not universal, are specific to that language. Thus in any linguistic text the sounds are patterned according to multiples rules. How characters are organized within words and in context to their neighbours, as well as throughout the text, should belie the principles and rules which have caused the patterns observed. Careful observation, along with sound linguistic reasoning, will let us slowly narrow down exactly what sounds a character may represent.
With the helps of the sounds identified, the researcher will at some point be able to do one of the following:
- point to a language or family of languages which the text is likely to be written in due to the peculiarity of the sounds;
- specify exact sounds for a few characters and thus confidently begin to read a handful of words; or
- narrow down the possible sounds for many characters and thus allow a more careful process of matching plants with names in the text.
I expect that 3 is most likely to happen, simply because this stage will be reached before the other two and the temptation will be too great to wait. But at whatever stage a researcher chooses to go back to the text and look for cribs they will be better weaponed. They will have been able to think, talk, argue, and consider the text of the Voynich manuscript without having first had reference to any outside influences. Theories as to language will have to fit the facts, rather than fitting the facts to the theory.