Re: Gallows bit sequences

[Jorge Stolfi <stolfi@xxxxxxxxxxxxx>]

    > [Jim Gillogly:] Without having done any computation, it looks
    > like there'll be more long runs of gallows and non-gallows than
    > a coin-flipping model would suggest, and that they're more
    > likely in the first line of a "paragraph" than

Indeed. How could we explain them?

Perhaps we are seeing the result of some automaton-like code (like
those Dee tables that Jim Reeds wrote about recently.)  I will leave 
this line of speculation to the experts; let me instead suggest a 
linguistic explanation.

As far as I know, Turkish words usually consist of a stem (usually 1-3
syllables) followed a string of suffixes (usually 1 syllable each).
Turkish has a rich set of suffixes, which often translate as separate
words in Indo-European languages.

Sometimes the string of suffixes can be quite long. A recent thread
in sci.lang discussed the "Guinness record" Turkish word 

 Cekoslovakyalilastiramadiklarimizdanmiymissiniz

which, if I recall correctly, means "are you one of those whom we were
unable to Chekoslovakize?".

Moreover, Turkish has this peculiar rule of "vowel harmony". The
vowels are divided into two symmetric classes, "front" and "back".
Generally speaking, in every Turkish stem, all vowels have the same
class; every suffix has a "front" and "back" version; and one may
only use suffixes that belong to same vowel class as the stem.

So, suppose that Voynichese is Turkish, each VMS "word" is actually a
Turkish stem or a suffix, and the gallows letters are vowel markers
for the front/back quality. This theory could explain the long runs of
0's and 1's in the "gallows bit" strings.

Perhaps the same case could be made for Hungarian, which I believe has
similar rules for suffixing and vowel harmony (isn't it remotely
related to Turkish?). Then there are other Turkic languages in Asia
(Uzbek, Chechen, ... ?)

If I am not mistaken, in the 1400's Turkish was commonly written in the
Arabic script. That script is not as well suited to Turkish as it is
to Arabic (which is one of the reasons why the country switched to the
Roman alphabet early this century). So the VMS author would have had a
good excuse for inventing a new alphabet, rather than using the
standard one --- especially if he/she was a "cultural transplant"
(a Turk in Europe, or an European in Turkey).

By the way, I gather that, until quite recently, public baths were
quite popular in Turkey (for both sexes), at least as much as in
classical Rome.

One obvious check for this theory is to see whether the ratio of front
and back words in Turkish is close enough to the gallows/no-gallows
ratio of Voynichese.

Anyone knows the Turkish names for the planets? Or the main star in Pisces?
Or how one would say "left kidney" and "right kidney"?  ;-)

    > [John Grove:] I believe the labels are extremely heavy in favour
    > of using Gallows. Just a quick glance at a zodiac page shows a
    > significantly higher number of words with Gallows than without.

This is not that relevant now... but indeed, the ratio is about 3:2
for labels. However labels are few compared to the text, so they have
little effect on the overall ratio.

The token counts for all sections are

              gallows in word
        -----------------------------
            ?     0     1     2     3 |   tot    SD
        ----- ----- ----- ----- ----- | ----- -----
text     2257 17363 17439   323     3 | 37385    96
labels    149   386   590    29     0 |  1154    16
        ----- ----- ----- ----- ----- | ----- -----
both     2406 17749 18029   352     3   38539    98
            - 49.1% 49.9%  1.0%  0.0%

(The percentages are over the good words only.)

All the best,

--stolfi