The Voynich Ninja

A while ago I You are not allowed to view links. Register or Login to view. about my experiments with the software related with the paper You are not allowed to view links. Register or Login to view. by Mans Hulden. In many ways, his algorithm is an extension / generalization of Sukhotin's system.

According to these experiments and to what I have read of the research of others (Emma and Rene, but also Jacques Guy), I believe that, if the script is phonetic and it is not an abjad, there is little doubt that EVA: e, o, a, y correspond to vowels. This result is obtained using different transcriptions systems and different classification algorithms. But (depending on the transcription system used) there might be other potential vowels. For instance, using EVA, the sequences ain / aiin suggest that i (following a) is a consonant and hence n (following the consonant i) must be a vowel.
I attach one of the graphs from the post I linked above. It is based on Rene's CUVA system. I have added in blue the corresponding EVA. The top level split corresponds to the hypothetical vowels / consonants. The lower levels are less reliable, still they point out one of the features mentioned by Rene: similar symbols behave similarly. For instance, the four Gallows are clustered together at level 3 (in the rightmost node). The two benches are clustered together down to level 4.

The lower levels of the leftmost / vowels cluster point out that e and ee behave similarly. Also, y and a are classified as similar, somehow confirming what Emma wrote in You are not allowed to view links. Register or Login to view..

Marco, one of the most amazing things about that diagram is the inability to cope with [i]. I'm sure it's not a failing of the principle, but it's really stark how [i] occurs all over the place.

It's also great to see yet another way of testing the script which shows the same set [e, a, o, y] split at the base from other glyphs.

(08-04-2019, 06:23 PM)Emma May Smith Wrote: You are not allowed to view links. Register or Login to view.Marco, one of the most amazing things about that diagram is the inability to cope with [i]. I'm sure it's not a failing of the principle, but it's really stark how [i] occurs all over the place.

It's also great to see yet another way of testing the script which shows the same set [e, a, o, y] split at the base from other glyphs.

Hi Emma, 
the way in which e,a,o,y differ from the other glyphs appears to be the clearest "phonological" fact. I agree that it is reassuring that several different methods arrive at the same result in this case.

Part of the problem with 'i' was an error in my EVA mapping: CUVA G corresponds to EVA:g (not EVA:il, I got confused with Currier's transcription which appears in the same table as CUVA on You are not allowed to view links. Register or Login to view.). I apologize. I have now updated the graph.
It is still bizarre to see 'q' clustered with 'in', but at least all 'i's are still together at level 3.

Maybe it is an indication the are related to roman numerals or some other symbols?

(08-04-2019, 06:57 PM)MarcoP Wrote: You are not allowed to view links. Register or Login to view.Hi Emma, 
the way in which e,a,o,y differ from the other glyphs appears to be the clearest "phonological" fact. I agree that it is reassuring that several different methods arrive at the same result in this case.

Part of the problem with 'i' was an error in my EVA mapping: CUVA G corresponds to EVA:g (not EVA:il, I got confused with Currier's transcription which appears in the same table as CUVA on You are not allowed to view links. Register or Login to view.). I apologize. I have now updated the graph.
It is still bizarre to see 'q' clustered with 'in', but at least all 'i's are still together at level 3.

Ah, I see! But, as you say, it's interesting that not only [q] but also [ch, sh], all of which tend toward the front of a word, cluster with [i] sequences which come at the end. Is the principle marking them as being in complementary distribution?

(08-04-2019, 07:14 PM)Emma May Smith Wrote: You are not allowed to view links. Register or Login to view.

(08-04-2019, 06:57 PM)MarcoP Wrote: You are not allowed to view links. Register or Login to view.Hi Emma, 
the way in which e,a,o,y differ from the other glyphs appears to be the clearest "phonological" fact. I agree that it is reassuring that several different methods arrive at the same result in this case.

Part of the problem with 'i' was an error in my EVA mapping: CUVA G corresponds to EVA:g (not EVA:il, I got confused with Currier's transcription which appears in the same table as CUVA on You are not allowed to view links. Register or Login to view.). I apologize. I have now updated the graph.
It is still bizarre to see 'q' clustered with 'in', but at least all 'i's are still together at level 3.

Ah, I see! But, as you say, it's interesting that not only [q] but also [ch, sh], all of which tend toward the front of a word, cluster with [i] sequences which come at the end. Is the principle marking them as being in complementary distribution?

Hulden defines the obligatory contour principle (OCP) as "a universal cross-linguistic tendency to avoid redundancy or repetition of similar speech features within a word or morpheme, especially if the phonemes are adjacent to one another".
The principle is that characters that tend to occur consecutively must have different sounds, while characters with similar sounds never or very rarely appear next to each other. At the light of this, it seems clear why q- and -in end up being considered similar: they never appear in consecutive positions. Similarly, gallows do not appear consecutively, so they must be similar sounds. 
The positional constraints of Voynichese are so strong that they make several consecutive combinations impossible or very rare: this could contribute to make the lower splits in this tree unreliable (because they depend on ridiculously small numbers). On the other hand, some of those constraints may very well be a consequence of the OCP.

(08-04-2019, 06:57 PM)MarcoP Wrote: You are not allowed to view links. Register or Login to view.

(08-04-2019, 06:23 PM)Emma May Smith Wrote: You are not allowed to view links. Register or Login to view.It's also great to see yet another way of testing the script which shows the same set [e, a, o, y] split at the base from other glyphs.

Hi Emma, 
the way in which e,a,o,y differ from the other glyphs appears to be the clearest "phonological" fact. I agree that it is reassuring that several different methods arrive at the same result in this case.

I appreciate all of the contributions to this interesting discussion, and especially Marco's extensive illustrative chart.

But I have a simple observation and a question about the set [e, a, o, y] :

A simple search on voynichese.com for the string "*[glyph]*" reveals how frequently (in how many word tokens) a given glyph occurs in medial position, with another glyph, signified by "*", occurring both before it and after it. (It does not have to be the same glyph before and after it, just any glyph before it and any glyph after it.)

Here are the results of this string search for the glyphs in the set [e, a, o, y] :

medial [e]  ("*[e]*")  :  14209 matches
medial [a]  ("*[a]*")  :  11749 matches
medial [o]  ("*[o]*")  :  14411 matches
medial [y]  ("*[y]*")  :      471 matches

Such a strikingly low frequency of medial occurrences, in contrast to all the other "vowel-like" glyphs, is rather strange for a vowel glyph, is it not?

What type of letter might pattern like a vowel in initial and final position, but rarely occur where vowels do in medial position?

Perhaps /s/, which can occur frequently even after other consonants in final position in many languages, and which can occur in initial position even before other consonants in many languages in clusters such as /st/, /sk/, /sp/, /sl/, /sm/, and /sn/, but which would much less frequently occur between two consonants in medial position, unlike vowels?

Geoffrey

It might be worth reading my post You are not allowed to view links. Register or Login to view.

I think it's also bearing in mind the possibility You are not allowed to view links. Register or Login to view. (though the latter is uncertain).

(08-04-2019, 09:38 PM)Emma May Smith Wrote: You are not allowed to view links. Register or Login to view.It might be worth reading my post You are not allowed to view links. Register or Login to view.

I think it's also bearing in mind the possibility You are not allowed to view links. Register or Login to view. (though the latter is uncertain).

Yes, the observations about [a] and [y] are very interesting and significant, I agree. But there may be many various explanations of these phenomena. For example, about the alternation of final [-y] and [-aiin] as suffixes of the same roots: Many researchers have observed that the glyph [y] has the shape of the common medieval Latin ms abbreviation for the "-us" suffix. If [-y] is also being used here to represent a suffix "-Vs", then this alternation could represent for example [-y] = nominative "-Vs" and [-aiin] = accusative "-Vn". In this case, the vowel is written as [a] in the [aiin] suffix, but not written separately in the [y] suffix. Such an explanation could be valid for this phenomenon in a number of medieval European languages.

I will definitely research your data about [a] and [y] in more depth, and consider other implications of these observations.

Geoffrey

(08-04-2019, 10:15 PM)geoffreycaveney Wrote: You are not allowed to view links. Register or Login to view.about the alternation of final [-y] and [-aiin] as suffixes of the same roots: Many researchers have observed that the glyph [y] has the shape of the common medieval Latin ms abbreviation for the "-us" suffix. If [-y] is also being used here to represent a suffix "-Vs", then this alternation could represent for example [-y] = nominative "-Vs" and [-aiin] = accusative "-Vn". In this case, the vowel is written as [a] in the [aiin] suffix, but not written separately in the [y] suffix. Such an explanation could be valid for this phenomenon in a number of medieval European languages.

Hi Geoffrey,
that idea is discussed by Jacques Guy inYou are not allowed to view links. Register or Login to view. (1991). I think Guy's paper marked a significant step forward in Voynich studies: it is definitely worth reading. Guy observes that (in the part of the text he analyzed) -y 'accounts for 50% of word finals, an extraordinarily high proportion. Do folio 79v and 80r contain a text with every other word or so ending in "-us" or "-os", or even plain "-s"? Possible of course, but quite improbable and thus a poor working hypothesis'. Such a high frequency strongly suggests that y must be a vowel (if the script is phonetic and not an abjad).

Guy also mentions the well known fact that in Medieval Latin manuscripts the '9' abbreviation 'word-initially [represents] the prefix "con-" or its variants (com-, cun-, cum-)'.
Of course, if one thinks that -9 as a suffix works as in Latin manuscripts, an implication is that also 9- as a prefix does so. But evidence suggests that EVA:y represents the same sound word-initially and word-finally. In particular, words ending with -y are followed by words starting with y- less frequently than expected (see You are not allowed to view links. Register or Login to view.). This can be observed for all Voynichese symbols that can occur both word-initially and word-finally. In many languages, the consecutive repetition of the same sound across word-boundaries is avoided through alteration of one of the two words (e.g. in Italian word-ending vowels are often dropped before words starting with the same vowel). See also Emma's discussion of You are not allowed to view links. Register or Login to view..