The Voynich Ninja

Rene, thank you again for doing and posting this analysis. I appreciate it very much.

I still have some comments and observations:

Many of these words may simply be hapax legomena. The longer and less typical the word, the more likely that this could be the case. In this case, we may not find the alternate form with the [d] substitution, simply because the word type was not used again in the MS at all.

About [dd], it would be fairly simple to propose a rule whereby in cases of [d] substitution producing [dd], the second [d] was simply not written.

It is not just the number of cases of [p]/[f] ~ [d] alternation that is striking to me, it is also the examples of long and/or atypical words, which one would not expect to have a "rhyming" word with [d], which nevertheless show this alternation. For example:

[shckhefy] / [shckhedy]

[chepchefy] / [chepchedy]

[qopchcfhy] / [qopchdy]

[pcheocfhy] / [pcheody]

[ckhcfhhy] / [ckhdy]

It is the quality of the words, not just the quantity of them, that is striking to me. nablator's glyph substitution table posted above is valuable and relevant, certainly. But I would like to see such an analysis, for example, that only includes occurrences in which at least one of the two words in the substitution/alternation is a hapax legomenon. It should be much less likely for a hapax legomenon to have a "rhyming" word with a single glyph alternation purely by chance or coincidence, as opposed to more frequent words, for which the "rhyming" words should be much more likely.

In short, I think 21 out of 86 words with a particular alternation may be a significant amount, when the [p] / [f] forms of these words are mainly hapax legomena.

Geoffrey

21 out of 86 is a minority. It is not a significant amount. How hard can it be to see this?

To explain 21 words you already need 3 rules.

In this list:

Quote:[shckhefy] / [shckhedy]

[chepchefy] / [chepchedy]

[qopchcfhy] / [qopchdy]

[pcheocfhy] / [pcheody]

[ckhcfhhy] / [ckhdy]

I have no problem with the first four, but the last is again a different rule, and now not only 'f' and 'cfh' but also 'cfhh' would be equivalent with 'd'.

Let me see if I can insert the complete list:


Now don't get me wrong: I would have been quite pleased if the suggested correspondence really existed, because it would tell us something new about the system behind the text. However, it only works (more or less) for a small subset of the words.

Rene, let me propose a refinement of the hypothesis for which there may be a stronger case:

Let us consider only those word types in this list that end in [-fy], [-cfhy], [-cfhhy], and [-cfhey].

(1) I count 8 word types in this list that end in [-fy]. One of them has the weird ending [-cfy]. If we are going to discount my matches such as [qokeefcy] for [qokeedy], we should also remove endings such as [-cfy] from the data set as well.

This leaves 7 word types in the list that end in [-fy].

6 of the 7 have doublets that end in [-dy]. Yes, in 2 of them, [-eefy] and [-eoefy] are shortened to [-edy] and [-eody].

Is this not significant?

(2) I count 6 word types in this list that end in [-cfhy].

4 of the 6 have doublets that end in [-dy]. Is this not significant?

(3) The only word type in the list that ends in [-cfhhy], has a doublet that ends in [-dy].

(4) The only word type in the list that ends in [-cfhey], has a doublet that ends in [-dey].

In sum, 12 of the 15 word types in the list that end in these sequences, have doublets with [d].

Is this not significant?

Geoffrey

(23-04-2019, 02:20 PM)geoffreycaveney Wrote: You are not allowed to view links. Register or Login to view.(1) I count 8 word types in this list that end in [-fy]. One of them has the weird ending [-cfy]. If we are going to discount my matches such as [qokeefcy] for [qokeedy], we should also remove endings such as [-cfy] from the data set as well.

Is there a logic behind this? This seems more retaliation than a serious attempt to test a theory.

(23-04-2019, 02:20 PM)geoffreycaveney Wrote: You are not allowed to view links. Register or Login to view.In sum, 12 of the 15 word types in the list that end in these sequences, have doublets with [d].

Is this not significant?

No, statistics on a handful of carefully cherry-picked examples are not significant.

7 of the 8 word types ending -heeky have doublets in -heedy. 

Is this significant?

(23-04-2019, 03:56 PM)MarcoP Wrote: You are not allowed to view links. Register or Login to view.

(23-04-2019, 02:20 PM)geoffreycaveney Wrote: You are not allowed to view links. Register or Login to view.(1) I count 8 word types in this list that end in [-fy]. One of them has the weird ending [-cfy]. If we are going to discount my matches such as [qokeefcy] for [qokeedy], we should also remove endings such as [-cfy] from the data set as well.

What is this the logic behind this? This seems more retaliation than a serious attempt to test a theory.

The logic is that we should either include such unusual forms entirely, or exclude them entirely. An extraneous [c] without an [h] looks like a scribal error, most likely, or a transcription error, or some such minor issue.

One may argue that we should simply exclude all such forms from the data set entirely, to be consistent.

Or one may argue that we should include them, and consider such near matches as [qokeefcy] / [qokeedy] as significant.

But to argue that we should include them in the data set, and then count [dy] for [fcy] as a non-match, as if it had no significance at all, that I do not consider an objective approach at all.

(23-04-2019, 03:56 PM)MarcoP Wrote: You are not allowed to view links. Register or Login to view.

(23-04-2019, 02:20 PM)geoffreycaveney Wrote: You are not allowed to view links. Register or Login to view.In sum, 12 of the 15 word types in the list that end in these sequences, have doublets with [d].

Is this not significant?

No, statistics on a handful of carefully cherry-picked examples are not significant.

7 of the 8 word types ending -heeky have doublets in -heedy. 

Is this significant?

There is a huge difference between the endings [-fy] and [-cfhy], and your ending [-heeky].

The endings [-fy] and [-cfhy] each consist of two natural Voynich characters.

The ending [-heeky] consists of four or five Voynich characters, or rather 3.5 or 4.5 characters, in which [h] is not even a character, but a final part of the Eva transcription of several different Voynich characters.

I understand if one wants to set aside the cases of [-cfhhy] and [-cfhey]. They are just a single word type each.

But in the cases of [-fy] and [-cfhy], we have 10 of 13 word types with doublets in [-dy]. 

They are not "cherry-picked", because this includes every word type in the MS ending in [-fy] or [-cfhy] with another gallows character in the word.

Geoffrey

(23-04-2019, 04:19 PM)geoffreycaveney Wrote: You are not allowed to view links. Register or Login to view.

(23-04-2019, 03:56 PM)MarcoP Wrote: You are not allowed to view links. Register or Login to view.

(23-04-2019, 02:20 PM)geoffreycaveney Wrote: You are not allowed to view links. Register or Login to view.(1) I count 8 word types in this list that end in [-fy]. One of them has the weird ending [-cfy]. If we are going to discount my matches such as [qokeefcy] for [qokeedy], we should also remove endings such as [-cfy] from the data set as well.

What is this the logic behind this? This seems more retaliation than a serious attempt to test a theory.

The logic is that we should either include such unusual forms entirely, or exclude them entirely. An extraneous [c] without an [h] looks like a scribal error, most likely, or a transcription error, or some such minor issue.

One may argue that we should simply exclude all such forms from the data set entirely, to be consistent.

Or one may argue that we should include them, and consider such near matches as [qokeefcy] / [qokeedy] as significant.

But to argue that we should include them in the data set, and then count [dy] for [fcy] as a non-match, as if it had no significance at all, that I do not consider an objective approach at all.

On the contrary: the approach you discard is much more objective than what you propose.
If you base your research on Takahashi's transcription, considering all the words he transcribed as valid and correct is quite objective. If two words were transcribed differently they are different. 

If you want to go down the "scribal error" alley, you should define exactly what you consider a potential scribal error. A good approach is to only consider words with at least N (e.g.  5) occurrences. 
One can then note that 20 of the 21 'fy' word types are hapax legomena (according to Stolfi's frequencies). The entire class of -fy words is suspiciously small and infrequent. Objectively, they could all be dismissed as "errors".

Quote:

(23-04-2019, 03:56 PM)MarcoP Wrote: You are not allowed to view links. Register or Login to view.

(23-04-2019, 02:20 PM)geoffreycaveney Wrote: You are not allowed to view links. Register or Login to view.In sum, 12 of the 15 word types in the list that end in these sequences, have doublets with [d].

Is this not significant?

No, statistics on a handful of carefully cherry-picked examples are not significant.

7 of the 8 word types ending -heeky have doublets in -heedy. 

Is this significant?

There is a huge difference between the endings [-fy] and [-cfhy], and your ending [-heeky].

The endings [-fy] and [-cfhy] each consist of two natural Voynich characters.

The ending [-heeky] consists of four or five Voynich characters, or rather 3.5 or 4.5 characters, in which [h] is not even a character, but a final part of the Eva transcription of several different Voynich characters.

I understand if one wants to set aside the cases of [-cfhhy] and [-cfhey]. They are just a single word type each.

But in the cases of [-fy] and [-cfhy], we have 10 of 13 word types with doublets in [-dy]. 

They are not "cherry-picked", because this includes every word type in the MS ending in [-fy] or [-cfhy] with another gallows character in the word.

Geoffrey

If you don't like the half-character, we can easily fix things: "considering every word type in the MS ending in sheeky or cheeky (bench-eeky), 7 out of 8 word types have doublets in -eedy".
Significant, uh?

In fact, the evidence becomes even more compelling when one simply considers all word types in the MS that end in [-fy]:

There are 21 word types that end in [-fy] in the MS.

18 of the 21 have a doublet that ends in [-dy].

In 3 cases, [-oefy] has been contracted to [-ody].

In 2 cases, [-eefy] has been contracted to [-edy].

In 1 case, [-aify] has been contracted to [-ady].

The only 3 word types in the MS ending in [-fy] that do not have a doublet ending in [-dy] are [dorshefy], [opcheolfy], and [ypchocfy].
Of course the latter parts of these words do have doublets ending in [-dy] : [shedy], [cheoldy], [chody].

Here is the list of the doublets in full:


[ofy] 2
[ody] 

[chepchefy]
[chepchedy]

[shckhefy]
[shckhedy]

[cholfy]
[choldy]

[oteofy]
[oteody]

[chefy]
[chedy]

[chofy]
[chody]

[sheefy]
[sheedy]

[olfy]
[oldy]

[cheefy]
[cheedy]

[shfy]
[shdy]

[cphhofy]
[cphody]

[qokeoefy]
[qokeody]

[shoefy]
[shody]

[choefy]
[chody]

[okcheefy]
[okchedy]

[shekeefy]
[shekedy]

[raify]
[rady]

Geoffrey

Moreover, out of the 16 word types in the MS that end in [-cfhy] :

11 of the 16 have a doublet that ends in [-dy].

So, out of a total of 37 word types in the MS that end in [-fy] or [-cfhy] :

We find that 29 of the 37 have a doublet that ends in [-dy].

For the record, the one example of non-final [cfhy] in the MS, [cfhyaiin], does have a doublet [dyaiin].

The 5 word types ending in [-cfhy] that do not have a doublet ending in [-dy] are [tdokchcfhy], [tcfhy], [olchocfhy], [shokocfhy], and [sochorcfhy].

Here is the list of the 11 final [-cfhy] / [-dy] doublets:

[cfhy] 6
[dy]

[shocfhy] 2
[shody]

[chcfhy] 2
[chdy]

[eeecfhy]
[eeedy]

[fchecfhy]
[fchedy]

[chocfhy]
[chody]

[fchcfhy]
[fchdy]

[checfhy]
[chedy]

[ykocfhy]
[ykody] 

[alcfhy]
[aldy]

[qopchcfhy]
[qopchdy]

Geoffrey

@geoffreycaveney, I'm intrigued. This is truly an innovative solution to the very real problem of the distribution of [f] and [p], which unless adequately explained, is pretty damning to most linguistic hypotheses of the VM.

However, I'm not satisfied with your solution's answer to the question of what [cfh] and [cph] should become when they're not on the first line of a paragraph or the first character of a line. If the distinction between [p] and [cph] is important in Voynichese, and there are minimal pairs that hinge on this distinction, it does not logically follow that the scribe would reduce both of these to [d]. For what it's worth, I'll abide [f] and [p] both mapping to [d], since it's not at all clear that [f] and [p] are not variations of the same glyph. But both of them and their benched varieties all mapping to [d] when not in the first line stretches credibility a bit.

Now, what I would be open to is [cfh] and [cph] regularly mapping to [*d] or [d*] or [*d*], where [*] is a wildcard representing another repurposed Voynichese glyph, whose identity should be discernible via statistical analysis, if your theory holds up.

(04-10-2019, 08:19 PM)RenegadeHealer Wrote: You are not allowed to view links. Register or Login to view.@geoffreycaveney, I'm intrigued. This is truly an innovative solution to the very real problem of the distribution of [f] and [p], which unless adequately explained, is pretty damning to most linguistic hypotheses of the VM.

However, I'm not satisfied with your solution's answer to the question of what [cfh] and [cph] should become when they're not on the first line of a paragraph or the first character of a line. If the distinction between [p] and [cph] is important in Voynichese, and there are minimal pairs that hinge on this distinction, it does not logically follow that the scribe would reduce both of these to [d]. For what it's worth, I'll abide [f] and [p] both mapping to [d], since it's not at all clear that [f] and [p] are not variations of the same glyph. But both of them and their benched varieties all mapping to [d] when not in the first line stretches credibility a bit.

Now, what I would be open to is [cfh] and [cph] regularly mapping to [*d] or [d*] or [*d*], where
[*]is a wildcard representing another repurposed Voynichese glyph, whose identity should be discernible via statistical analysis, if your theory holds up.

Thank you for the complimentary comment and for your thoughts on the topic. I appreciate it very much.

To reply to your argument about [cfh] and [cph], I guess I am not convinced that the distinction between [p] and [cph] is necessarily important in Voynichese. If we think of this script as a cryptographic cipher in modern terms, with a necessarily precise enciphering process requiring a precise deciphering process to read it, then yes, in that case we would need to suppose a clear and precise reason for the distinction between writing [p] in some places and [cph] in other places.

But if the script is simply the writing system of a natural language, the spelling could be just as illogical in some respects as most other natural languages' writing systems. Many languages' scripts have silent letters with no phonetic meaning in the contemporary language, and many languages' scripts have two different symbols or combinations of symbols to write the same sound.

If you read through my posts about my Judaeo-Greek theory of the Voynich ms, you will see that I basically identify Voynich [p], [f], and [d] as representing the Greek sounds "ph", "p", and "b". But the Judaeo-Greek script itself was ambiguous about its expression of these Greek sounds: It did not consistently distinguish Greek "ph" and "p", using Hebrew "pay" for both. It used both "bet" and "vav" for Greek "b", and it also used "vav" for Greek "o" and "u". (Greek "b" at this time was really pronounced "v", and it is natural to use "vav" for both "v" and "u", an overlap that occurs in many languages and writing systems.) Judaeo-Greek used "he" for the "h" sound that was typically not written in Greek or only marked with a "rough breathing" diacritic.

I basically identify Voynich [p] with Judaeo-Greek "pay" in this system, Voynich [f] with Judaeo-Greek "bet", Voynich [ch] with Judaeo-Greek "he", and Voynich [d] with Judaeo-Greek "vav". Sometimes the author just wrote [p] for Judaeo-Greek "pay" meaning Greek "p" or "ph"; sometimes the author wrote [cph] to specifically indicate "pay+he" for Greek "ph". Both of these, as well as [f] for "bet" for Greek "b"/"v", could be substituted with [d] for "vav". I admit this is a more natural explanation as a substitution of [d] for [f] ("vav" for "bet") and a less natural explanation as a substitution of [d] for [p] ("vav" for "pay"). But this is what I have been able to come up with.

Geoffrey