In the text of the Voynich manuscript, certain glyphs frequently occur in isolation. Additionally, there are a few pages that contain a sequence of single glyphs, in a circular band or in the margin like initials. Studying these cases could potentially teach us something about the VM writing system. This post is my first exploration of the phenomenon.
Let us first have a look at which glyphs can occur in isolation in the text, as if they form a “word” by themselves. In modern English, only two glyphs can pull this off: “I” (the personal pronoun) and “a” (the indefinite article). Three if you allow the archaic interjection “o”. However, one can think of situations where other glyphs appear separately as well.
a for
b example
c in
d alphabetic
e lists
Five EVA glyphs appear in isolation over a hundred times; in descending order these are: [s, y, l, o, r]. We might also add [d], which occurs 44 times in isolation. Let’s call these sylord glyphs, a special category with the ability to function as a “word” beyond any reasonable doubt. We can be certain that these, at least, are separate glyphs.
Overall though, this approach won’t take us far. For starters, as in English, not every glyph in a language will appear in isolation. And how do we differentiate between single glyphs, digraphs, ligatures, short words, one-off scribal errors…? And with the VM’s notorious use of ambiguous spaces, can we trust that a free-standing glyph is actually free-standing? And what about minim clusters, which are traditionally transliterated as something like [aiin], but may also be read as one glyph? This leaves us with more problems than solutions.
For example, let us analyze this line from f6v.
In EVA, this would be written something like chockhy.s.or.chy.s.ain.or. The first [s] is clearly separated by spaces. One could argue about the second [s]: on voynichese.com it is attached to “ain”, reading “sain”, while in the ZL file it remains separate. The latter option seems the most likely to me. But now consider the “ain” part: how many glyphs are these? Three, two, or even just one, made of three minims with loops at the ends?
Might we find intentionally single glyphs elsewhere?
In the margins
There are a few pages where series of single glyphs have been added in the margins, almost like initials. The first such example is f49v:
For these to be helpful, we must assume two things:
- They represent single glyphs. One glance at the series suggests that this is very likely the case, since none of these glyphs appear composite, consisting of mostly one or two strokes.
- They were made by someone who “knew” Voynichese. This is impossible to know for sure. In the case of f49v, I see no reasons to doubt the series’ authenticity, but other cases are less clear.
Let us now consider each series in turn. Since we are interested in the behavior of common Voynichese glyphs, the many rare or unique glyphs in these series will be ignored.
f49v
The first series of “single characters” is found on f49v, in the left margin. The sequence is mostly repeating, but since sequences are not the focus of this post, below is just a quick and dirty representation of the glyphs.
This column uses the gallows [k, p, f] and glyphs [d, e, o, r, s, y]. It also includes two rare glyphs, which we will ignore: one resembling the number 2, and another that looks like a mirrored “c” or an open circle. This leaves us with gallows and “sylord” glyphs (only [l] is missing). To me, this suggests that the writer knew what he was doing: gallows have the appearance of initials, and sylord glyphs already occur easily by themselves. If the c-shape is indeed the same as EVA-e, then it is the exception here.
- gallows: [f, k, p]
- sylord: [s, y, o, r, d]
- other: [e]
f66r
A similar column is found on f66r. This one includes more rare glyphs, and no clear repeating sequence. In the image below, I simply cut up the column to make it more presentable. This is no attempt at discovering a sequence.
In this series, we are treated to three different gallows and the complete sylord collection. This is no surprise. Much more interesting are the three other frequent glyphs. It includes EVA-c, which is the first half of the “bench” glyph, if it is indeed the same. But it also included EVA-sh, the bench-with-curl on top. So what is it, are benches made up of parts, or are they single glyphs?
Finally, there is one attestation of EVA-air. If both assumptions are correct – that these represent single glyphs and that the writer knew what he was doing, then the implications are huge. It would mean that [a] is part of the minim sequence, and that [air] was considered a single glyph. Of course, we don’t know if the assumptions are true.
- gallows: [f, p, t]
- sylord: [s, y, l, o, r, d]
- other: [c, sh, air]
f76r
Yet another marginal column is on f76r, but this one contains only 9 tokens, which I awkwardly gathered below.
We get one gallow here, and only [y] is missing to complete the sylord-collection. The surprise guest on this folio is EVA-q, which is notorious for almost always preceding EVA-o.
- gallows: [k]
- sylord: [s, l, o, r, d]
- other: [q]
Circular diagram: f57v
f57v contains two circular bands of single characters in a series that repeats four times with only minor variations. The image below, made by Sam G, is a good representation of the first band. Several of these characters are rare or unique to this page. I fully acknowledge that these may be important, but they are not the focus of this thread, so let’s look at the rest.
- gallows: [f, k, p, t]
- sylord: [y, l, o, r, d]
- other: [c, m]
This is the inner band:
- gallows: [f, k, t]
- sylord: [s, y, l, o, r, d]
- other: [m, aiin, ar]
In summary, f57v contains all gallows and all sylord glyphs. There is again what looks like the first half of the [ch]-bench, and this time no complete bench to counter it. Interestingly, both [aiin] and [ar] are included in the inner string of single glyphs.
f69r?
Finally, around the centre of f69r, we find six more single (?) characters. At first, I doubted whether these should be considered, since they appear to correspond to points of a diagram. But then it dawned on me. This whole post I’ve been promoting the silly mnemonic sylord, which stands for all free-occurring glyphs in descending order. Now look at this:
In these six points, we get five out of six sylord-glyphs: [s, y, l, o, d]. Note that EVA-r has been present in all other pages we looked at, but here it is the missing one. Even more surprisingly, in its stead we get something that looks like [ed] or [em]. (ZL says “ed”, TT says “em”). Apart from this last combination (?), it looks like this diagram is well aware of Voynichese’s preferred single glyphs.
Discussion
We now have an inventory of the common glyphs used in isolation on five conspicuous folios. The graph below shows on how many of these pages each glyph features in a series of singles:
The graph shows that gallows and sylord glyphs are clear favorites. The gallows [f, k, p] are used in three out of five folios. Common glyphs [l, r, s] occur in all but one folio, and [s, o, d] are found in all five.
Gallows: according to the ZL transliteration file, gallows are not used very frequently in isolation in the normal text. This is not required either, after all most words (in Latin script) consist of multiple letters. Still, gallows are popular choices when the VM needs isolated glyphs. This is no surprise, since the initial-like gallows have a strong air of proper “glyphhood”.
It should be noted that benched gallows do not occur. This may be an argument in favor of “unstacking” benched gallows rather than considering them unique single glyphs.
Sylord: before studying the series of isolated glyphs, I checked which glyphs commonly stood alone in the normal text. In descending order of frequency, these were [s, y , l, o, r, d]. As it turns out, these are also the most popular glyphs in series of singles. There must be something special about these glyphs, making them particularly suited to stand by themselves. Or about other glyphs, making them less likely to stand by themselves. It looks like when they ran out of sylord and gallows, they would resort to “weirdos” rather than other Voynichese glyphs.
Benches: the bench characters are EVA [ch, sh]. Here, the results are ambiguous. The first part of the [ch]-bench appears to be used on two pages. I don’t know whether this is a “weirdo”-glyph that simply looks like half a bench, or it actually hints at benches’ composite nature. Strangely, the bench-with-cap [sh] is used in one of the series, hinting at this being considered a single glyph. Interestingly, [sh] is the most frequently isolated non-sylord glyph. Despite being less than half as frequent as [ch], the capped bench [sh] occurs twice as much in isolation.
e: despite being one of the most frequent glyphs in the MS, [e] does not like to occur in isolation: this only happens on f49v. It is unclear to me whether this was an intentional use of the actual glyph. It looks like this folio, when it had used all but two of the gallows-sylord set, started experimenting with curved lines, since it also includes some weird mirrored version of [e].
Additionally, one page contains [em] as the final member of a series of sylord-glyphs. Given this evidence, I would suggest that [e] has less “glyphhood” than gallows and sylord-glyphs.
q: EVA-q, the glyph that looks like “4”, is usually followed by [o]. In the whole VM, it appears only once in isolation, in the margin of f76r. To me it looks like a clear example of [q], and it is represented as such in both LZ and TT. In the clip below, you can see three examples of [qo] in the main text as well.
Since this is a unique example, it is not clear to what extent this demonstrates that the [q] usually attached to [o] was seen as a separate glyph. If one considers this as evidence, then it probably implies that prefixation of [q] somehow further modifies a word that has already been prefixed with [o]. If I recall correctly, research has been done along these lines, and this is certainly a possibility.
ar, air, ain: Separate glyphs [a, i, n] are entirely absent from the series, yet a-clusters appear on two folios. Granted, the three of them are different and they are not nearly as frequent as gallows or sylord-glyphs, but still a-headed clusters have a clear presence. This should really make us wonder what’s going on with these glyphs. Does [a] exist separately, or is it always part of a larger, combined unit? Or does [a] turn into something else, like [y] when it stands alone? Is [i] a separate character, or merely a stroke in a larger glyph? And why is [n] missing from the separate glyph series? It has a loop, making it more clearly defined for standalone use, yet it is always attached at the end of certain clusters.
Both TT and ZL mark three or four single [n] in the main text, but those have a strange shape compared to the normal versions on the same page.
Conclusion
We still have a lot to learn about the way Voynichese script functions. I was surprised to see that the five pages with a separate series of individual glyphs all follow the same logic, relying heavily on gallows and especially sylord-glyphs. These same glyphs are the ones that most often stand alone in the main text as well, which may suggest that the writer(s) of the glyph series made informed choices.
Other glyphs were notably absent: the bench [ch], benched gallows and [a, i, n]. The latter three only appeared in combinations. EVA-e occurred on one page, but in my opinion not entirely reliably. Either way, its absence from the four other pages is strange given the status of [e] as the second most frequent glyph in the main text. Therefore, I would include [e] along the glyphs where something weird is going on.
Focusing on the absence of [a, i, n, e], I can think of two possible “solutions”:
- Are they strokes of larger glyphs? EVA [ain] can be read as a three-minim glyph with a loop on each side.
- Do they have preferred equivalents for standalone situations? For example, maybe [a] becomes [y] in certain contexts, and the [y] version – with tail – is preferred when the glyph stands by itself. Visually, [r] can be seen as [i] with a tail and [s] as [e] with a tail. Apparently this even extends to the bench characters, where the only the tailed version [sh] occurred once.
As so often, this post left me with more questions than answers. And if I learned something, I am not sure what it is.
Nice lists!
Is it possible it is all or partly math-based? I’ve read a number of cipher analyses now but likely only 1%. Still, I don’t think c is an e from what I’ve read. I think it might modify the value of other letters, or symbolically enclose them?
Say, for instance, the glyph 8 had a numeric value of 10. Then c8 might be 20 (2 x 10), whereas 8c might be 5 (10/2).
Just thinking aloud. Thanks for the singles frequency).
The other noticeable thing for me in 57v’s series is how attention seems to be drawn to c’s roundedness or angularity. Best I can do with tablet keyboard: v ^ . The first 8 for instance, had v at the bottom instead of the roundness. And c is shown as < in the first sequence, then rounded to c, and then the bench added. The most curious glyph to me is the c.
LikeLike
Hi Koen,
thank you for the interesting post!
Here are a few ideas/observations (some of which are rather far-fetched, I know):
* I interpret the f49v glyph ‘that looks like a mirrored “c” or an open circle’ as the “plume” that is added to ‘ch’ to form ‘sh’.
* The bottom glyph in the f76r sequence looks to me more like an incomplete ‘sh’ than the typical standalone ‘s’.
* Several of the sequences are ambiguous in nature: it is not always clear if they are single-glyph only or they include words. This is particularly clear in f57v, where all bands with the exception of the second one include both single-glyphs and longer words.
* I tend to read the f79r sequence as forming the word “dolsedy”. According to Stolfi’s and Emma’s grammars, this is not a legal Voynichese word, but “dolshedy” is (it occurs twice in the ms). We can at least say that this is close to being a Voynichese word, which cannot be said of the other sequences.
* I agree with your conclusions. About the second one, Emma mentioned these single-glyph sequences in her 2015 post about the Equivalence of ‘a’ and ‘y’.
She pointed out that ‘y’ appearing in the sequences while ‘a’ doesn’t could support the equivalence between the two. As you say, in f66r ‘air’ is included in what appears to be a list of glyphs. I think this could be compatible with Emma’s idea: ‘y’ represents the basic entity, while ‘a’ only occurs when followed by other glyphs that somehow modify that entity.
If i-sequences are modifiers of ‘a’, ‘e’-sequences might be modifiers of benches and gallows,
LikeLiked by 1 person
Hi Marco
Thank you for your comments. I sometimes find it hard to wrap my head around these things, so it helps to exchange some ideas.
If the “incomplete circle” is a plume, that would mean that at least this writer saw it as a separate “thing”, not just a stroke of a glyph. It is a shame that this is the only page using the isolated curves (EVA-e and the plume), otherwise it might allow for stronger conclusions.
Seeing the glyphs around the “star” as a separated word “dolsedy” is something I had not thought of yet, though it does make some sense, especially because “ed” is followed by its partner “y”. In this case, we must certainly read the unclear glyph as [d] rather than [m]. Then I find it very interesting that they separated all glyphs, but kept [ed] as a unit!
I agree that a-y equivalence should be the most obvious solution. Words like the frequent “dal” visually suggest that the “a”-part is a separate bit, not just strokes of a larger glyph. Still, I keep some reservations. The fact that i-clusters are so often flanked by “a” and “n” makes it hard for me to believe that “a” is always a separate glyph (as opposed to a cluster-initial form of the minim).
LikeLike
Hi Koen,
these are a couple of examples of circles where words are broken into single glyphs / small groups of glyphs appearing in clearly separated boxes, as for the hypothetical “dolsedy” in f79r.
https://religionnews.com/2015/05/15/readers-solved-mystery-prayer-wheel/
This is said to date to the 12th Century. The third circle reads:
DO, N, A, SA, NCT, I, S[pirit-], US
Here each radial “box” contains groups of variable length, from 1 to 3 glyphs.
https://bildsuche.digitale-sammlungen.de/index.html?c=viewer&l=it&bandnummer=bsb00037155&pimage=00073
This is from a 15th Century book of magic (Munich, CLM 849). The centre reads F,L,O,R,O,N (a demon’s name, I guess). The word “floron” appears in the first and fifth lines of the text above the diagram.
Of course, these examples prove nothing: I was just curious to see if words were sometimes split into single glyphs in similar diagrams.
You undestandably decided to ignore rare symbols in the VMS sequences you discussed. Of course, we can only speculate, but why do you think they are so more frequent in some of those sequences than in paragraph text? What can they be?
The Munich diagram also features unique symbols: these appear to be demonic “sigils”. But, contrary to the VMS, they look very different from the other characters in the page.
LikeLike
Hi Marco
I agree that the “donasancti…” example is good to keep in mind regarding the “dolsedy”, it shows how “ed” could be more like a syllable.
Regarding the rare symbols, I have only one vague hypothesis. Maybe for some reason they needed a series of unique glyphs (like a numbered or alphabetical list). From the VM glyph set, it looks a bit like only “sylord” glyphs are popular picks. They have a strong “standalone quality”. The rest is filled with incidental picks or even unique glyphs.
But this is just my feeling. I’m afraid there will be as many explanations as there are types of VM theories
LikeLiked by 1 person
Hi Koen,
more about the “dolsedy” hypothesis.
https://www.jasondavies.com/voynich/#f67r1_f67r2/0.758/0.378/3.20
In 67r2, the quadrant at 10:30 contains the isolated word “dolchsody”. This is one of seven isolated words. From the description on this page on Rene’s site:
http://voynich.nu/q09/index.html
“Similar illustrations are quite common in other manuscripts. Invariably, the twelve segments represent either the 12 signs of the zodiac or the twelve months of the year. The seven more or less separate words are very suggestive, and could represent the seven planets.”
I think that Nick Pelling suggested that seven of the large wheels in the VMS might correspond to the seven planets. I am not sure that 69r was one of the diagrams he regarded as planet-related.
Supposing that the star in 69r does represent a planet, “dolsedy” and “dolchsody” could be two variants of the name of one of the seven classical planets: since it is represented as a star in 69r, the Sun and the Moon seem less likely than the other five.
I had also totally forgotten that the subject was discussed on the site of Stephen Bax. There I pointed out that there also is a similarly named star (“dolchedy”) in 68r1. However, this would suggest a fixed star, rather than a planet.
https://stephenbax.net/?p=1755#comment-164359
Anyway, the presence of similar labels in two other VMS star-diagrams might give some support to reading the separate glyphs in 69r as a word.
https://imgur.com/gallery/fxgokIk
LikeLike
Ah, I see, so it would be more like a label for the star. My first impression was that the arms of the star divided the area into labelled segments, like a compass. But you make a good case for reading it as one word.
LikeLiked by 1 person
Since Voynich words almost never end in “i” – by my count 8 times – it is tempting to speculate that “n” indicates the last “i”. Similar to “j” being the last “i” in handwritten Latin numerals, e.g. “xviij”, and in Dutch “ij”. Also, words don’t start with “n” but for the few strange isolated ones.
The same might go for “r” and “m”, but adding an extra meaning. For “r” it might be the curl like in “s”, for “m” it might under certain circumstances signify the last word on that line.
The above irrespective of whether the i’s are separate glyphs.
Since “ii” was written “ij” to prevent confusion with “ü” and it is not clear how in Voynich there would be confusion, it might be an ornamental curl there.
LikeLike
All the [s, y, l, o, r, d] glyphs occur frequently at the end of words. The only other glyphs which occur at the end of words with the same frequency are [n, m]. Both of these are context-dependent: [n] to the environment immediately following [i], and [m] to the end of lines (though not exclusively.
The two phenomena (of occurring at the end of words and alone) are likely related to a third: words can be truncated more easily from left to right than right to left. (So removing a glyph from either end of a word to result in a new valid word.) While not a firm rule, more words conform to it than not. Words which fail truncation from the left often do so due to the presence of [e] or [i], which also prevents truncation from the right – neither [e] or [i] are common at the start or end of a word.
What this suggests to me is a certain level of glyph dependency from left to right, so that when the right glyph is removed the left glyph “loses” something of itself, such as context or an expected effect. It may be that the [s, y, l, o, r, d] glyphs are not those can appear alone, but rather what glyphs look like when they are without any context.
(All this said, I would be careful not to dismiss the possibility that lone glyphs are used for numeric value. A numeral system similar to the Greek script would more commonly use some glyphs rather than others. I believe (though with no firm evidence) that the glyph repetitions on f57v are likely numeric. The only glyph which definitely(?) changes is the central gallows of a 17 glyph sequence, leaving repeating 8-glyph sequences to either side. (Though [r] occurs twice in the sequence, providing a potential piece of evidence against my belief.))
LikeLiked by 3 people
Hi Emma,
the gallows also are frequent in the single-glyph sequences discussed in Koen’s post. They behave as a rather homogeneous group and in Stolfi’s model, they form the “core” of words. Differently from the [s, y, l, o, r, d] glyphs, they rarely appear at the end of words. Any ideas about why they can appear as stand-alone in single-glyph sequences but not in the text?
I also agree about the hypothesis that some of the sequences might correspond to numbers. In this light, it seems interesting that there are so many different glyph-types and so many isolated glyph-tokens. The Roman, Greek and Arabic systems tend to use more than a single glyph-token to represent each number, e.g. they all write numbers in the 11-19 range using at least two tokens. On the other hand, this hypothetical system has many different glyph-types each apparently corresponding to a different number.
LikeLike
Hi Marco, glyphs in single glyph sequences might simply take the same form as when they stand alone in the text, rather than only being those which can occur alone. So gallows have the same form no matter what, whereas [s, y, l, o, r, d] glyphs are the standalone versions of other glyphs which don’t appear.
On numerals: you’re quite right to question how the system would work. I don’t know and I’ll admit to that. This is only my guess and good evidence is wanting.
LikeLike
Hi Koen,
thanks also for pointing out stand-alone ‘n’!
The ZL and TT (voynichese.com) transliterations agree on three occurrences: 14v, 32r, 35r.
Takahashi has a fourth one in f57v, but that one should be ignored as a ghost due to poor images. In that position there only are two faint parallel lines marking the start of the circular text.
http://www.voynichese.com/#/f57v/exa:n/352
https://www.jasondavies.com/voynich/#f57v/0.4/0.267/4.40
The character appears to be different both in shape and behaviour from regular EVA:n. All three occurrences are at the end of lines, so this bizarre character appears to be one more special line-effect, something like EVA:m/g, but much rarer.
LikeLike
I was surprised to see the behavior and shape of the supposed standalone “n” as well. If this glyph occurred more frequently, I wonder if it should have gotten its own character in EVA. If we say that this is the same as EVA-n, then we confirm the existence of positional variation.
LikeLike
A related glyph occurs at the start of the last two lines on 10r. This time, however, the glyph bears a tail/plume similar to [r] rather than [n]. Four of the lines on that page start quite weirdly and I’m not sure what’s going on there.
Overall, it feels like, in addition to the [e] and [i] strokes, there’s this third kind of stroke, which is a backward curve. It is rarely seen with the same tail or plume as [n] and [r], but could be more common. What if, just as [a] is a combination of [e] and [i], that [o] is a combination of [e] and this stroke?
LikeLike
Hi Koen.
I would like to ask you if that sylord-sequence occurs just in one particular section or in all of them.
Now that you have observed that pattern, it is intriguing what linguistic values or roles could play.
Thanks.
LikeLike
Ciao,
Grande lavoro, ma secondo me l’alfabeto EVA è una strada senza uscita.
LikeLike