Label Encoding Method Update

The other day I discovered a mistake that prevented the smooth encoding of labels from the Astro section– I had transposed two letters in the Latin letter frequency arrangement. Once that was fixed in the encoding rectangles, many of the problems disappeared.

There are three sets of letter values as reported by other researchers, and three encoding rectangles. Not all of the values have been determined.

I have uploaded the modified method file to the MediaFire folder (link to the side).

Of the 22 encoded labels presented, only five have a misplaced letter error, for a success rate of ~80%.

f68r1 Paragraph Update

Due to family problems, my research has had to take a back seat for some time now, and I’m hoping to get back on track soon.

But I have discovered this: the 3rd word in Line 3, EVA <Shey> always comes out as “the”, no matter what other words I use as a crib. The “Sh” and the “y” swap the values “T” and “E”, but the “e” is almost always “H”.

By Philip Neal’s letter substitution rules (see below), “Sh” and “e” have two values, and “y” has one. Currently the other value of “Sh” is “C”, but I have not found the 2nd one for “e”.

My current crib words are “COMET” and “MOON”. I’ve found words that fit: a five letter word with five different letters, and a four letter word with three, one of them doubled. The letters in common allow the translations, and it’s a matter of determining (read: guessing) which other values go with which other letter. Then spread those values to the rest of the paragraph.

I’ve just started this version, but another word found is the 2nd word of Line 2, EVA <otChl>, which comes out as “OMEN”.

Research Update: f68r1

As I’ve mentioned, the letter value sets don’t work on the text paragraphs.

Since I have an interpretation for f68r1, I’ve been using likely words to find a crib, and I’m assuming it’s written in English. To this point, here is what I’ve got:

Line 1: COMET word word word word word MOON word MOTION
Line 2: word word word word word word word word BIT
Line 3: ME ON THE word word word word word word
Line 4: word word

Yes, it strikes me as weird too. I’ve probably gone wrong somewhere, but I’m going to keep working with this. Who knows?

Letter Columns on Folio f1r

I’ve studied the three letter columns on f1r off and on for some time, and here is a table of everything I’ve managed to tease out.

I note that the value of EVA d matches the one I found for Value Set 1, and that EVA e and EVA r matches their alternate values.

If anyone has found more and would like to work with me to add to this data, please feel free to contact me.

“Degrees of Freedom”

One of the most common objections to a proposed translation method is to invoke “degrees of freedom”. This multiplies the letters and gives the total of possible combinations. If anagrams or dropped letters are allowed, this increases the total.

So it’s “Whoops, your method has too many degrees of freedom, so the results can be almost anything.” The ones objecting stop here, aka ‘hit brick wall, end of story’. But that is NOT the end of the story. As raw math, “degrees of freedom” is valid, but it ignores the rules imposed that reduce the number of possible answers.

It’s exactly the same as an argument I’ve seen advanced by Creationists: “There are so many elements that the odds against certain ones combining to make life are astronomical. Therefore a Creator is necessary.” That too is raw math, and ignores the rules of chemistry that make the odds of the right elements combining a near certainty.

As an idealized example I am using the first label I cracked, by the large star in the upper left pie slice of f68r3. It is laid out in what I call a breakdown box. This is NOT the VMs decoding method, but a workaround I use in my label research. These are Set 1 letter values, and the word is complete (on the folio the 2nd A and R were dropped).

There is a length of 9 letters, 6 different, and 5 have alternate values with 2 connected. If I understood correctly, there are 1134 possible strings. If that was all there was to the method, there would indeed be too many possibilities.

However, imposing rules changes that:

1. Discard all nonsense strings.
2. Discard words of other than 9 letters.

The program WordFind has a lexicon of ~150,000 words, and running the values returns 14,  reducing the number of possible strings by 98.7%:


These results are filtered through a third rule: in any Voynich word, a given letter may have only one value. (That is, EVA Ch is either E or S, and any word containing both is invalid.)

That leaves one valid result: Aldebaran.

The lesson here is that one should not be so quick to reject a method based solely on a lowest-level mathematical calculation.