26-02-2026, 01:26 PM
(This post was last modified: 26-02-2026, 01:39 PM by Jorge_Stolfi.)
In previous posts I had proposed that the elements of the Voynichese script were
Together with the seven-slot ("crust-mantle-core",CMC) model, this model for
the elements implies many restrictions on the words, such as "the 'ed' digraph cannot occur at the start of a word" and "the i glyph must be followed by i, n, m, or r".
I had included {iiir} and {iiim} for symmetry with {iiin}, but had already observed that they practically do not occur:
113.500000 0.00091 {n} 868.250000 0.00697 {m}
1665.500000 0.01336 {in} 40.000000 0.00032 {im}
3779.000000 0.03032 {iin} 15.000000 0.00012 {iim}
159.000000 0.00128 {iiin} 1.0 0.00001 {iiim}
487.750000 0.00391 {ir}
130.500000 0.00105 {iir}
1.0 0.00001 {iiir}
(The numbers are total counts in running text, and frequencies relative to all elements.)
It is also worth noting that {iir} is less common than {ir}, even though the pattern of {in} and {iin} is the opposite.
I was unhappy with those {ir} and {iir} elements anyway because I could not let {r} alone be a coda element, like {n} and {m} were. Making {r} a coda element would have created a serious ambiguity in the parsing, since the final r in words like otar could be parsed either as a dealer (slot 6) or a coda (slot 7), and the choice would affect the limit on the number of dealers of the CMC model. Excluding {r} from the codas would avoid that ambiguity but made the set of elements less symmetrical.
But now I have become convinced that many, if not all, of the hundreds of ir endings are in fact scribal errors for iin. These errors could have been caused by the Author's sloppy "cursive" handwriting on the draft, where an in could easily be confused with an r:
Thus I have decided to remove the elements {ir}, {iir}, and {iiir} endings from my model. When parsing words into elements, I may label any word that ends in ir as invalid, or "auto-correct" it by replacing the final ir by iin, as appropriate.
If ir is in fact a valid ending for Voynichese words, the auto-correction of ir to iin will introduce some errors. But I am convinced that it will also fix many errors, thus it is unlikely to make the puzzle harder to crack.
However, that decision now implies an ambiguity about the r without preceding i at the end of words like otar. The "sloppy draft" argument should apply here too: presumably many of those final rs are actually ins. Should I declare them illegal too? That would exclude a lot of common words. Should I auto-correct them too to in? That may be a sensible choice, but (unlike the case of final ir) I don't have any evidence that those are mostly errors. I will let them stand for now...
All the best, --stolfi
- "O" the /circles/ {a} {o} {y},
- "Q" just {q}.
- "D" the /dealers/ {d} {l} {r} {s}.
- "X" the /benches/ {ch} {sh} {ee} with an optional 'e' suffix.
- "G" the /simple gallows/ {k} {t} {p} {f} with optional 'e' suffix.
- "H" the /platform gallows/ {cth} {ckh} {cph} {cfh} with optional 'e' or 'h' suffix.
- "N" the /codas/ {n}, {in}, {iin}, {iiin}, {m} {im} {iim} {iiim}, {ir}, {iir}, {iiir}.
Together with the seven-slot ("crust-mantle-core",CMC) model, this model for
the elements implies many restrictions on the words, such as "the 'ed' digraph cannot occur at the start of a word" and "the i glyph must be followed by i, n, m, or r".
I had included {iiir} and {iiim} for symmetry with {iiin}, but had already observed that they practically do not occur:
113.500000 0.00091 {n} 868.250000 0.00697 {m}
1665.500000 0.01336 {in} 40.000000 0.00032 {im}
3779.000000 0.03032 {iin} 15.000000 0.00012 {iim}
159.000000 0.00128 {iiin} 1.0 0.00001 {iiim}
487.750000 0.00391 {ir}
130.500000 0.00105 {iir}
1.0 0.00001 {iiir}
(The numbers are total counts in running text, and frequencies relative to all elements.)
It is also worth noting that {iir} is less common than {ir}, even though the pattern of {in} and {iin} is the opposite.
I was unhappy with those {ir} and {iir} elements anyway because I could not let {r} alone be a coda element, like {n} and {m} were. Making {r} a coda element would have created a serious ambiguity in the parsing, since the final r in words like otar could be parsed either as a dealer (slot 6) or a coda (slot 7), and the choice would affect the limit on the number of dealers of the CMC model. Excluding {r} from the codas would avoid that ambiguity but made the set of elements less symmetrical.
But now I have become convinced that many, if not all, of the hundreds of ir endings are in fact scribal errors for iin. These errors could have been caused by the Author's sloppy "cursive" handwriting on the draft, where an in could easily be confused with an r:
Thus I have decided to remove the elements {ir}, {iir}, and {iiir} endings from my model. When parsing words into elements, I may label any word that ends in ir as invalid, or "auto-correct" it by replacing the final ir by iin, as appropriate.
If ir is in fact a valid ending for Voynichese words, the auto-correction of ir to iin will introduce some errors. But I am convinced that it will also fix many errors, thus it is unlikely to make the puzzle harder to crack.
However, that decision now implies an ambiguity about the r without preceding i at the end of words like otar. The "sloppy draft" argument should apply here too: presumably many of those final rs are actually ins. Should I declare them illegal too? That would exclude a lot of common words. Should I auto-correct them too to in? That may be a sensible choice, but (unlike the case of final ir) I don't have any evidence that those are mostly errors. I will let them stand for now...
All the best, --stolfi