Notes on the Voynich Manuscript - Part 5 [1992 January 3] ---------------------------------------- Comments on previous decipherments. Chapter 5 of D'Imperio's monograph is called "Major Claims of Decipherment", and one thing I did over the break was read this chapter and check back to the major references, most of which are reprinted in Brumbaugh's collection. This is a summary of what I found, with some tentative deductions. 5.1 Newbold I agree with Manly and D'Imperio that the decipherment is incredible. The proposed means of encoding is by micro symbols visible only under the microscope. The proposed decipherment contains sufficient leeway that almost anything can be read into any text. Nuff said. 5.2 Feely He at least realised that, if the plain text is Latin, it would be mediaeval Latin. He seems to have started out with conventional "gold bug" analysis - letter and word counts, both on the Voynich MS and on Roger Bacon's works. As far as I can tell, he got nowhere. His next attempt was based on guessing the individual words that labelled the drawings. This is not easy, for it implies guessing first the language, then the meaning of the drawing, and finally the word labelling it. So, even if the text is Latin, and the drawing is of a poppy, is the text "poppy" or "to induce sleep" or "for headaches"? In my view, his guesses were pretty much wrong, and the proof is that the keys they provide don't unlock the text with any degree of credibility. 5.3 Strong This suffers, in my view, from the same defects as Newbold: the cypher is too complex, and the recovered text is not credible. 5.4 Brumbaugh This is a critical claim, for if it is true, the MS is indeed largely a forgery. Brumbaugh's cypher is pretty simple, but be warned that I am depending on D'Imperio's reconstruction, since Brumbaugh's own articles don't give full information. In brief, the decoding process is this: each Voynich symbol stands for one of the digits 1 to 9, and there are about three alternatives for each digit, to obfuscate. Each digit, in turn, stands for one of three possible letters of the roman alphabet, take your pick. The result, when you pick wisely, is intelligible dog-latin, ie Latin much simplified and with the inflections largely replaced by the common ending -US. The encoding process, of course, is the reverse: collapse the plain text into a sequence of digits, and write the sequence in Voynich, with appropriate alternation rules to ensure that each symbol gets used a decent number of times. Here is an example of a decoded word 12127339 ABABGCCI JKJKPLLZ VRVRYWWus showing the digit sequence and the three possible choices for each digit. Note: this is given on p37 of D'Imperio in a form that cannot be reconciled with the tables in her Fig 26. I have changed it to make it consistent. Note also that -us is encoded by one symbol, the digit 9 (one of whose Voynich symbols is also "9"). Simple mathematics tells us that there are 3^8 or 6561 possible decipherments of this word. Brumbaugh chooses "ARABYCCUS", which is passable dog-latin for "arabiclike". And, indeed, there are very few choices that lead to pronounceable words. Brumbaugh's critical argument is this: that, in spite of this ambiguity, one who understands what the plain text is about can indeed read the cypher text and comprehend it. I propose to test that by an experiment. Here is a familiar Latin phrase in the Brumbaugh code, with the possible decodes beneath it: 5619 6246 45336284 EFAI FBDF DECCFBHD NOJZ OKMO MNLLOKQM XTVus TRST SXWWTRUS Did you read the phrase? If so, Brumbaugh's argument is at least partially confirmed. Now, here is the entire encode/decode box, from D'Imperio: 1 2 3 4 5 6 7 8 9 A B C D E F G H I J K L M N O P Q Z V R W S X T Y U us My first comment, is that if this is meant as a device to encode Latin plain text, it's a pretty bad one. Maybe we need to distinguish I and J, though most authors didn't, but we surely don't need both U and V. After all, whoever wrote the signature numbers didn't bother to distinguish. Then, why have one symbol for -us but not one for qu-? And, if the purpose is to encode Latin, why have K and W at all? Finally, assigning the Voynich symbol "9" to -us is really stupid, since that is already a well-known shorthand symbol for the same ending. And given that this supposed -us is one of the most common features of the text, the choice of "9" as the cypher symbol makes no more sense that choosing a substitution cypher for English that encodes 'e' as "E". Doctor Dee's double bluff? Far-fetched, I think. Secondly, while it is probable that the inflection structure of any underlying language has been simplified, I doubt very much that any educated person would "simplify" Latin by making every noun end in -us. Latin has been deliberately simplified at least twice, and has simplified itself many more times in the historical evolution of the romance languages, and as far as I know in every case the simplification process discarded the -us ending, usually for something derived from the dative or ablative case. [Note: The best-known artificially simplified Latin is Peano's "latino sine flexione", which, as you can see, does not use -US] And my final thought was this: Brumbaugh in his decipherment is violating his own key assumption, for he is a reader who does NOT know what the underlying text is about. That led me to another experiment, where I tried to apply his decoding rules myself. Here is the digit sequence and the decode 676517 14665 28379 851 581 84 46 FGFEAG ADFFE BHCGI HEA EHA HD DG OPONJP JMOON KQLPZ QNJ NQJ QM MO TYTXVY VSTTX RUVWusUXV XUV US ST Which I read, allowing myself similar latitude as did Brumbaugh in respect of contractions and simplifications: OP[um] TE AG[e]; AMO TE BULGUS; UNA EQU[i] US[um] DO "Do ye the work; I love you, ye Bogomil; I give to one [woman] the use of a horse!" Which proves, perhaps, that the Voynich MS is indeed part of a secret kinky Cathar cult? Not quite. You see, the original digit sequence is not taken from the MS; it is taken from the rightmost column of a random page of my table of seven-figure logarithms, using spaces and zeros to divide the groups. Which proves rather, I think, that the Brumbaugh method does not recover the underlying text; the meaning is inserted by the ingenuity of the decipherer, and any text whatever may be so "deciphered". My tentative conclusion, folks, is that we have here pitfalls to be avoided. Robert [Note: the test phrase in the text is, of course, "novus ordo seclorum", and several members of Team Voynich got it.]