03-04-2026, 10:55 AM
I have been arguing that EVA "s" = "und" (and) based on frequency analysis. When I added another Bavarian text to my reference corpus — the "Buch der Natur" (BdN) by Konrad von Megenberg, ca. 1350-75, a natural history text from Regensburg — something jumped out at me. The word "und" is underrepresented in this text compared to the medical recipe books. Here is the frequency table:
The explanation is obvious once you see it: the BdN is literary prose, not a recipe collection. Recipes are full of "take this AND that, mix this AND that" — enumerations. The Bavarian recipe collection hits over 10% "und", the BdN sits at 3.88%. Makes perfect sense.
And then it hit me: if "s" really means "and", then the VMS should show the exact same pattern. The Pharma pages with their ingredient lists should have more "s" than the text-only pages. The prose sections should be low. If the distribution matches what the illustrations tell us about text type, that would be strong hint for s = "und" (and) - and simultaneously confirm the content classification of the manuscript.
Here is the section-by-section count:
And there it is.
Pharma (f87-f102): the pages with jars and herb ingredients: 9.73%! Exactly where you would expect a recipe/ingredient section to land — right between the medical recipe books (5-6%) and the pure recipe collection (10.8%) in my MHD reference.
Astro/Zodiac (f67-f73) is the highest at 10.93%. I have not worked on the Astro section myself, but this makes sense if you think about it — these pages maybe full of star labels arranged in circles. "Star X and star Y and star Z" — pure enumeration. Even more list-like than recipes. But this has to be verified. If anyone here has worked on the Astro section and can tell me whether this makes sense, I would appreciate it — I simply do not have the time to dig into that section as well.
The Herbal section (f1-f57) sits at 6.01% — plant descriptions with some recipe elements mixed in. Right in the middle. Balneo and Cosmo are similar.
And here is a fun one: the so-called "Recipes" section (f58, f103-f116), the text-only pages with star markers in the margin — they come in at 4.16%.
That is possibly not a recipe section. That could be prose. The s-frequency is clear. These pages read like the Buch der Natur (3.88%), not like a recipe collection.
More Tests
Other glyphs also vary across sections. That is typical for the VMS — the variance between sections tends to be higher than in normal texts. So the question is: is s actually special, or is it just one of many glyphs that happen to fluctuate?
I tested all 19 standard EVA characters across all six manuscript sections. Three criteria must be met simultaneously for a conjunction candidate:
Test 1 — Frequency. A conjunction like "and" should account for roughly 3-10% of tokens. Most EVA characters fail this immediately: i sits at 13.6%, t at 15.5%, d at 16.4%, a at 21% — all far too high. m at 2.8% is too low. s is the only standard EVA character that falls in the 3-10% range.
Test 2 — Pattern. A conjunction should peak in enumeration sections and drop in prose sections. Many glyphs fluctuate across sections, but they peak in the wrong places — e peaks in Balneo, t peaks in Astro but drops in Pharma, q peaks in Balneo. s is the only glyph in the conjunction frequency range that peaks in Pharma.
Test 3 — Degree of variation. Most standard EVA characters vary moderately across sections, with max/min ratios between 1.4x and 2.2x. That is normal fluctuation. MHD "und" across five real German text types varies by 2.78x — significantly more, because conjunctions are genuinely text-type sensitive. VMS s comes in at 2.63x — the strongest variation of any common EVA character, and right in line with real "und". No other frequent glyph reaches this level.
One glyph out of 19. Three independent tests. All matching the behavior of "and" in real medieval German text.
The overall max/min ratio — 2.63x for the functional s across VMS sections, compared to 2.78x for "und" across MHD text types — tells the same story. Not identical, but driven by the same mechanism: lists need more conjunctions than prose.
The idea that the “s” could be a simple one—appearing both on its own and at the beginning or end of a word—has become much more likely.
The explanation is obvious once you see it: the BdN is literary prose, not a recipe collection. Recipes are full of "take this AND that, mix this AND that" — enumerations. The Bavarian recipe collection hits over 10% "und", the BdN sits at 3.88%. Makes perfect sense.
And then it hit me: if "s" really means "and", then the VMS should show the exact same pattern. The Pharma pages with their ingredient lists should have more "s" than the text-only pages. The prose sections should be low. If the distribution matches what the illustrations tell us about text type, that would be strong hint for s = "und" (and) - and simultaneously confirm the content classification of the manuscript.
Here is the section-by-section count:
And there it is.
Pharma (f87-f102): the pages with jars and herb ingredients: 9.73%! Exactly where you would expect a recipe/ingredient section to land — right between the medical recipe books (5-6%) and the pure recipe collection (10.8%) in my MHD reference.
Astro/Zodiac (f67-f73) is the highest at 10.93%. I have not worked on the Astro section myself, but this makes sense if you think about it — these pages maybe full of star labels arranged in circles. "Star X and star Y and star Z" — pure enumeration. Even more list-like than recipes. But this has to be verified. If anyone here has worked on the Astro section and can tell me whether this makes sense, I would appreciate it — I simply do not have the time to dig into that section as well.
The Herbal section (f1-f57) sits at 6.01% — plant descriptions with some recipe elements mixed in. Right in the middle. Balneo and Cosmo are similar.
And here is a fun one: the so-called "Recipes" section (f58, f103-f116), the text-only pages with star markers in the margin — they come in at 4.16%.
That is possibly not a recipe section. That could be prose. The s-frequency is clear. These pages read like the Buch der Natur (3.88%), not like a recipe collection.
More Tests
Other glyphs also vary across sections. That is typical for the VMS — the variance between sections tends to be higher than in normal texts. So the question is: is s actually special, or is it just one of many glyphs that happen to fluctuate?
I tested all 19 standard EVA characters across all six manuscript sections. Three criteria must be met simultaneously for a conjunction candidate:
Test 1 — Frequency. A conjunction like "and" should account for roughly 3-10% of tokens. Most EVA characters fail this immediately: i sits at 13.6%, t at 15.5%, d at 16.4%, a at 21% — all far too high. m at 2.8% is too low. s is the only standard EVA character that falls in the 3-10% range.
Test 2 — Pattern. A conjunction should peak in enumeration sections and drop in prose sections. Many glyphs fluctuate across sections, but they peak in the wrong places — e peaks in Balneo, t peaks in Astro but drops in Pharma, q peaks in Balneo. s is the only glyph in the conjunction frequency range that peaks in Pharma.
Test 3 — Degree of variation. Most standard EVA characters vary moderately across sections, with max/min ratios between 1.4x and 2.2x. That is normal fluctuation. MHD "und" across five real German text types varies by 2.78x — significantly more, because conjunctions are genuinely text-type sensitive. VMS s comes in at 2.63x — the strongest variation of any common EVA character, and right in line with real "und". No other frequent glyph reaches this level.
One glyph out of 19. Three independent tests. All matching the behavior of "and" in real medieval German text.
The overall max/min ratio — 2.63x for the functional s across VMS sections, compared to 2.78x for "und" across MHD text types — tells the same story. Not identical, but driven by the same mechanism: lists need more conjunctions than prose.
The idea that the “s” could be a simple one—appearing both on its own and at the beginning or end of a word—has become much more likely.
![[+]](https://www.voynich.ninja/images/collapse_collapsed.png)