[Blog Post] Rightward and Downward in the Voynich Manuscript - Patrick Feaster

[pfeaster]

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). 

Koen's answer is correct.  I wanted to wait a few hours before confirming in case anyone else wanted to participate.  Specifically, the order from left to right is:

a) Scribe 3 verso
b) Scribe 2 verso
c) Scribe 3 recto
d) Scribe 2 recto

It's possible that differences in average line and paragraph length will have impacted the appearance of the images in addition to actual differences in distribution.  A larger proportion of short paragraphs could produce more vertical banding, for example.

But some differences still seem fairly decisive.  The sharp contrast in [daiin] between first and second line positions, with a peak in first position, appears for Scribe 3 but not for Scribe 2.  With Scribe 3, the first line position favors vords beginning [o]+gallows; with Scribe 2, it favors vords beginning [qo]+gallows.  The near-end-of-line peak for vords beginning [o]+gallows is stronger and more consistent throughout the paragraph for Scribe 3 than for Scribe 2 (though paragraph lengths might be a factor there).  And so on.

The differences between results for recto and verso page groups attributed to the same scribe seem informative too.  

I went back and repeated the experiment using two groups that alternate between recto and verso pages in order to check whether the recto/verso distinction itself has any impact.  Surprisingly enough, it looks as though it does.  For Scribe 3, [daiin] shows a much stronger preference for first line position on verso pages than on recto pages, and this difference vanishes with scrambled recto/verso groupings.  On the recto pages, [daiin] seems to be about as common in third line position as in first line position; on the verso pages, it seems to be virtually absent there.  If this can be confirmed, I think it would present a challenging hurdle for hypotheses about what positional patterns "mean."

Otherwise, the exact positions of mid-line banding phenomena (as well as the features of the second line position) seem to vary quite a lot both with the recto/verso groups and with the mixed groups.  Maybe that's to be expected with different combinations of line lengths.  So, for example, the near-end-of-line peak of [o]+gallows vords for Scribe 3 occurs a little earlier on average on verso pages than on recto pages, but even when I mix recto/verso pages, its position and vertical consistency still varies noticeably.  I guess the finer structure of banding is likely to be due to random variation and probably isn't trustworthy as a source of insight.

Here are some modest hypotheses I'd like to put forward for critique, based on the foregoing:

1. If the opposite of "flat" is "bumpy," then the text of the VM is fundamentally bumpy, not just superficially bumpy.
2. Its bumpiness isn't limited to beginnings of paragraphs or to beginnings or ends of lines (although it especially affects those places) but extends throughout the whole body of the text.
3. Its bumpiness isn't limited to nonce similarities among nearby vords (basic self-citation) but constrains the text by position according to consistent patterns.
4. Its bumpiness affects the text structure thoroughly enough to preclude any method of text generation (whether natural-language plaintext, cipher, invented language, stochastic hoax, or "other") that is fundamentally flat with only minor tweaks made to the result.

And some questions:

1. The above observations are limited to paragraphic text.  Do labels, circles, and radii show any similar patterning by position (based e.g. on outwardness or angle)?
2. Lines and paragraphs seem to be positionally patterned units, but do larger units show overarching positional patterns too (e.g., multi-paragraph pages, whole bifolios)?
3. Which patterns are tied to absolute positions (e.g., the first vord of a line) and which are tied to relative positions (e.g., the middle third of a line) -- and given chronic uncertainty about vord divisions, how can we reliably tell?
4. Do line length and paragraph length affect positional patterns and, if so, how?