The Voynich Ninja

(29-11-2021, 03:49 PM)Koen G Wrote: You are not allowed to view links. Register or Login to view.I'm wondering if you would see anything like this at all in regular manuscripts. Maybe in the first position due to capitalization, and maybe in the final position due to increased incentive for using abbreviation-related glyphs. For anything outside of that, I can't think of any good examples.

Having a suitable data structure on hand, I pfeasterized the more frequent words in the King James Bible (admittedly not a manuscript) and looked for gross asymmetries.  One sample was the first five books of the Old Testament, with chapters parsed as "paragraphs," and grammatically complete sentences as "lines" (which leaves some capital letters stranded at intra-sentence verse breaks).  Representative density maps are shown below on a discrete 20 x 20 grid, as justified by the typical lengths of chapters and sentences.  Fractional populations of specific words are computed relative to the total number of words in each bin.  Greater density is represented as darker gray value.

Most word distributions are flat.  Some show understandable biases toward or against the beginning or end of lines, for example "And" and "and"

[attachment=6077]

Other word distributions have chapter-internal structure.  Two conspicuous examples from further down the frequency table are "Moses" and "said"

[attachment=6078]

I am not convinced by inspection that there are significant chapter-downwardness trends in any of the 40 most-frequent Pentateuch words.  The Gospels offer other candidates, "said" "me"

[attachment=6079]

None of the distributions in these samples are convincingly peaked at intermediate downwardness.

Thank you, @obelus for confirming suspicions.  As this is a very visual way of representing high level non-language trends in the VM this is not a surprising result but it is good to have the data.  I guess the next step is to investigate possible “whys” - but given the likely effort to pre-scramble the input l want to think on possible tests.  

At the very least the stark differences between Currier A and B are food for thought.

(29-11-2021, 04:19 PM)MarcoP Wrote: You are not allowed to view links. Register or Login to view.Given that aiin is so frequent in B, it seems possible that (in B) daiin is a line-initial variant of aiin, one of Emma's You are not allowed to view links. Register or Login to view.. As I discussed with Emma (and possibly on the forum) years ago, it does not seem that daiin in A can be equivalent to aiin in B, since daiin often appears reduplicated as daiin.daiin while aiin never does.

I tried plotting [daiin] in blue versus [aiin] in green for Currier B to see how their distributions compare with each other.  

Alas, it's tricky.  This is one of those cases where the treatment of uncertain spaces can make a big difference.  The break before [aiin] often falls into what I think of as an "ambivalent" category, with the combination of [aiin] and the preceding vord also appearing written together as a single vord without an intervening space, as well as with ambiguous spacing, e.g., with [raiin], [r.aiin] and [r,aiin] all being attested.

With that in mind, on the left below I show the result ignoring all comma breaks, and on the right I show the result treating all comma breaks as real vord divisions.

[attachment=6080]

On the left, it looks as though [aiin] peaks in second position, with additional peaks at mid-line and near-end-of-line (but not in actual last position).  On the right, the second-position peak seems to spill over into the next position, and the near-end-of-line peak is much stronger, with two weaker [aiin] peaks appearing between the other peaks.  In both cases (but especially on the left), there also seem to be mid-line areas where [daiin] is more common than [aiin], though these are dark areas overall in which both vords are underrepresented.  It seems almost as though regions favoring [daiin] or [aiin] are alternating with each other, or else as though mid-line [aiin] is concentrated in a few peak positions while mid-line [daiin] is more evenly distributed as a kind of background.

For what it's worth, here's a similarly generated plot for [daiin] in blue versus [chol] in green in Currier A:

[attachment=6081]

That's not quite the same pattern, but it provides another example where a difference between first and second line positions seems to be "echoed" through a kind of mid-line alternation or vertical banding.  I'm trying to decide whether this is an artifact of processing or not.  I've tried limiting analysis to lines of specific lengths, and the banding seems to persist -- that is, to cite just one example, [daiin] turns up a fair amount as the third and fifth vords of ten-vord lines in B, but not as the fourth vord.  Still, there's less data available when we limit our scope in that way, so it can be hard to tell whether there's a meaningful pattern or not.

And who said you can't start a sentence with 'And'?

(30-11-2021, 01:38 AM)LisaFaginDavis Wrote: You are not allowed to view links. Register or Login to view.I'd love to hear what you see as the implications of these positional observations. Do different sections reveal different patterns? Are these patterns the result of different topics? Or a sophisticated encipherment? What about the different scribes who write in Currier B (i.e. my scribe 2 vs. 3 vs. 4 vs. 5)? Do they reveal different underlying patterns?

So many great questions!  I wish I had as many great answers.

For the moment, I can at least confirm that different sections/scribes seem to follow different patterns.  I have separate files already on hand for Quires 1-3, Quire 13, and Quire 20 (or 18 in your numbering), so it's easy enough for me to analyze those, and except for f115r I believe they correspond to your scribes 1, 2, and 3, although of course they're not complete for any of them.

Here are the results contrasting vords beginning [qo]+gallows in blue with vords beginning [o]+gallows in green for Quires 1-3 (left), Quire 13 (middle), and Quire 20 a.k.a. 18 (right).

[attachment=6083]

And here's [daiin] for the same three sections.

[attachment=6084]

There are significant-looking differences among all three.  As for implications, I'm still pulling my own thoughts together, but I guess one place to start would be to run through existing hypotheses to see which are and aren't compatible with this kind of patterning.

If you divide a scribe's pages in two parts, are those parts more similar to each other than to different scribes? This would be a good control.

Great idea, Koen!

(30-11-2021, 08:40 PM)Koen G Wrote: You are not allowed to view links. Register or Login to view.If you divide a scribe's pages in two parts, are those parts more similar to each other than to different scribes? This would be a good control.

Great idea -- thanks!  Here are results for scribes 2 and 3 with pages divided into recto and verso groups (which seemed like the fairest way to divide them, although I suppose the recto/verso distinction could conceivably have some bearing of its own on positional patterns).  The same phenomena are plotted as before: [qo]+gallows vords in blue versus [o]+gallows vords in green on the top, [daiin] on the bottom.  Top and bottom images represent consistent combinations of scribe and page group.  Can you tell which top-and-bottom pairs go with which scribes?

[attachment=6085]

(I have some thoughts about how this went but will wait to share them until others have had a chance to weigh in.)

Fine print: Scribe 2 is represented by folios 26, 31, 33, 34, 39, 40, 43, 46, 50, 55, and 75-84, while Scribe 3 is represented by folios 58, 94-95, 103-114, and 116.

Cool! Thanks for checking this so quickly. 3, 2, 3, 2? (if this is incorrect it is because I am reading incorrectly, not because the data is inconclusive). 

I would love to hear your thoughts on this, in my experience performing these kinds of experiments provides one with an understanding that is hard to obtain otherwise.

Each pair of recto and verso pages belong to the same folio. Therefore it would be a surprise if recto and verso groups groups would behave differently. 

Instead I suggest to divide for scribe 1 into Botanical/Pharmaceutical folios and for scribe 2 into Botanical/Biological folios.