Voynich Reconsidered: the {M} conundrum

In a previous article on this platform, I proposed a strategy for discovering meaning in the text of the Voynich manuscript. I called it the {8am} strategy. My idea was to focus on the glyph string {8am}, which is the most frequent “word” in the v101 transliteration of the Voynich manuscript. My proposed process was to test some alternative transliterations of the manuscript, and some target languages, and to see whether in one combination or another, {8am} would map to a recognizable word.

This research is ongoing. So far, it has yielded encouraging results for several transliterations, and for medieval Italian, represented by the OVI corpus, as the precursor language.

In these tests, I am dealing only with mappings of the glyphs {8}, {a} and {m}, which are common glyphs in the Voynich manuscript.

In a parallel research effort, as I reported in other articles, I have looked at mappings of the three right-justified sequences of glyphs on page f1r, the first page of the manuscript as currently bound. The hypothesis to be tested is that one or more of these sequences might represent a proper name. The third sequence is {okoe 8aM}. The glyph {M} is infrequent, and therefore presents a conundrum (of which, more below).

I have in mind to attempt mappings of the sequence {okoe 8aM} from various transliterations to a selection of medieval languages. The target languages include Albanian, Arabic, Bohemian, English, Finnish, French, Galician-Portuguese, German, Latin, Persian, Turkish and Welsh. This is a massive manual exercise and requires some objective criteria whereby one might prioritise the languages.

One such criterion is related to the rarity of the glyph {M}. In the v101 transliteration, {M} occurs just 45 times; its frequency is 0.1 percent; it is the 35th most frequent glyph. In most of the natural languages that I have examined, if we ignore punctuation, make no distinction between upper and lower case, and count only the letters with a frequency of at least 0.1 percent, the alphabet has between 24 and 30 letters.

If {M} is to map to any letter in a natural language, we have at least three possible strategies.

Alphabets with accents

If we retain all the glyphs in the v101 transliteration, then the alphabet of the target language must contain at least 35 letters. Mapping of the glyph {M} seems to require a language which has several accented letters in addition to the normal letters of the alphabet, or has at least 35 normal letters. However, as shown in the table below, of the languages that I have examined, the only one that meets this criterion is Ottoman Turkish.


Counts of letters in the alphabets of selected medieval languages. Author's analysis.

Upper and lower case

The Voynich manuscript shows no sign of a distinction between upper and lower case. That is to say, there are no pairs of glyphs in which one appears like a larger version of the other. This is so even we allow for differences in the shapes of upper and lower case letters, such as we see in the Latin script where “A” is not a magnified “a”.

However, to my mind we cannot exclude the idea that the Voynich producer provided the scribes with source documents that contained both upper and lower case letters; that he or she instructed them to map upper and lower case differently; even that the scribes did not know the precursor languages, and therefore could only map the shapes that they saw.

In that case, for each of the target languages (except Arabic, Persian and Turkish which have no upper and lower case), we could significantly increase the number of letters in the alphabet. Then, as shown in the table above, in several European languages we could find a mapping for a glyph as infrequent as {M}.

Combining groups of glyphs

Alternatively, or additionally, we could look for ways to move {M} upwards in the frequency rankings. The higher the ranking of {M}, the better its chance of mapping to a letter, even if an infrequent one, in some natural language.

One way to do so is to combine groups of visually similar glyphs, as I have done in many of my variants of the v101 transliteration.

For example, we can merge the v101 glyphs {6}, {7} and {&} with {8}; {3}, {5} and several variants with {2}; {j} with {g}; {u} with {f}; and {(} with {9}. This has the effect of moving the glyph {M} up to about 30th place. We can also redefine {m} as {iiN}, {n} as {iN}, and {M} as {iiiN}: in which case the sequence to be mapped becomes {okoe 8aiiiN}, and in place of {M}, we try to map {i} and {N} which are more frequent than {M}.

To my mind, all of these possibilities are worth a try. I will report in another article.