Hypotheses on Voynich f67r2


I’ve mentioned before that I have questions about the paints used in the Voynich manuscript, and that outside of the yellows and some of the reds, I see little evidence that they were added by someone who understood Voynichese. In this post I’ll dive into one of the most obvious exceptions to this, as there is one instance where there is definitive proof that whoever added at least some of the red to the Voynich was fluent in Voynichese.

This proof occurs on f67r2, which features a ring of Voynichese words written in red ink and a line of text in red ink. Where it has faded, there is no trace of underlying brown ink, so this red was definitely added by someone who knew what he was doing, as opposed to tracing over another scribe’s letters.  It is the only page in the Voynich where red ink is used for writing. As noted since the earliest days of Voynich research, the text at the bottom of this page also stands out because visible ruling has been used here, and nowhere else in the manuscript.

In medieval manuscripts, writing in red ink is known as rubrication. It is  found in manuscripts from the 5th century onward, although the trend tends to peter out  and disappear with the advent of print. The use of red ink would highlight words which were of special importance, were different, or played a special role in the text.

This post will present some thoughts about what can be learned from the rubricated text on f67r2.


First of all, the presence of rubrication gives us an indication about the timeline of the page’s construction. According to the British Library website’s glossary, rubrics were often added by a separate scribe (the rubricator) but not always. In any case, “rubrication – sometimes done by the scribe – generally followed the laying out and writing of the text”. So the roundel and lines were traced/drawn first, then the brown text was added, then finally the red text.

In the case of 67r2, I don’t think this is a separate scribe. It is also very obvious that the red scribe was not a specialist, and found writing in red ink very challenging. However, I do think this line of text was added after the rest of the brown text on the page had been written. Planning for rubrication during the original page construction stage would justify the use of ruled lines here, as they lay out the space where the rubricated text should later be inserted.

This provides the explanation as to why this is the only place in the Voynich where the text is ruled: because it is the only place where there is rubricated text.

There is a broader aspect of the Voynich construction that can be understood from the presence of rubrication on this page. Rene Zandbergen points out that it is probable that the Q9 foldout was bound incorrectly, and indeed the presence of rubrication on a random middle page doesn’t really make sense. As noted by John Grove on the old mailing list back in 2003, the rubrication suggests that 67r2 was originally intended as the first page of this quire, like this:

Quire 9 (Modified from diagram on voynich.nu)

This suggests that Q9 was intended to be bound like all the other foldout quires in the Voynich, with only one folio to the left of the binding and all the foldouts to the right. Grove further noted that binding it like this makes 67r1 and 68v1 face each other.

It is easy to see that these two pages do appear to be related: they have the same overall design and, when the pages are bound correctly, the faces in the central roundels are actually looking at each other. It is clear that these pages were originally meant to be facing each other once the foldout was open.


The Moons and the Star

On f67r2, red ink is used for the text, but also for the moon-like faces near the middle of the roundel.

The division of the circle into twelve sections, and the fact that each section contains what appears to be a moon face, suggest that each section may represent a month, and that 67r2 might be some sort of moon-related representation of the 12 months of the year, with the presence of 7 “labels” corresponding to the planets, as Rene Zandbergen explains here.

The problem with this is that although it seems like a straightforward explanation, known depictions of 12 moons in a roundel don’t match what we see on f67r2. The 67r2 moons do show a similar alternating colors to the first illustration below, and this parallel almost matches the YRYRYRYYRRYR pattern seen in the Voynich moons.

However, in the above images, and more generally in moon-themed roundel illustrations (such as the Liber Floridus illustration discussed by Marco Ponzi here), the depictions of moons reflect the lunar cycle, but the moons on 67r2 don’t fit that model: they are all crescent moons, with no progression from waxing, to full, to waning, to new moon discernible.

Another factor to be considered is that both of the above illustrations show the Sun at the very top of the circle, behind the new Moon. In 67r2, there is no Sun. One might be tempted to think that here, rather than be shown behind the moons, the Sun might instead be the central figure in the roundel, however, by comparing it to other depictions of the Sun in Q9, we can clearly see that this is not the Sun: the Voynich convention always gives the Sun and Moon a face. The central figure looks like a star.

I removed most of the annoying green bleedthrough from the other side.

Furthermore, I don’t think the dotted line that extends to the left is a text divider. The Voynich divides circular text bands using straight, vertical bars, sometimes ornate, never with a dotted line. Rather, dotted lines in drawings are the Voynich artist’s typical way of showing movement, particularly the movement of things moving/flying/falling through the air. Remember this nymph from f76v?


It can be derived from this that the big star has moved (or appears to move) to its current position, that it has entered the circle of moons. In light of this, I looked into medieval texts about shooting-stars, meteors and comets. But if this were a meteor or comet, it has a really puny tail for its size, and observing the tail (or “hair”, coma, from which the word comet derives), and especially its orientation, was a huge part of medieval meteor prognostication, so why minimize the most important part?

Then by chance, I happened upon the following description in a chronicle by Jacobus Malvicius (Brescia, 1380-1452) : “And in those days a star of immense brilliance appeared within the circle of the moon around the first days of its separation from the Sun (Et diebus illis stella fulgoris immensa intra circulum lunae apparuit circa dies primos ipsus separatione a sole). Jacobus is not describing something he personally witnessed, but an event which occurred a full four centuries earlier: According to Umberto Dall’Olmo (1980), this is a description of  the SN1054 supernova. The supernova appeared in April of 1054 and was visible at least into July (some Chinese sources say 21 months). It was brighter than any other object in the sky except for the Moon, and was even visible in daytime for three weeks. This event was visible all over the world, although in Europe, the only surviving records come from these later Italian sources.

The event is also recorded by other Italian chroniclers, who describe the appearance as an “entry” and sometimes use circuitu instead of circulum. In the 1476 Cronaca Ramona (Bologna) we can read: “At this time a very bright star entered the circuit of the new moon” (Stella clarissima in circuitu prime lune ingressa est). The Annales Cavenses describe it in almost identical terms: “At the beginning of the night a very bright star entered the circle of the moon” (Incipiente nocte stella clarissima in circulum lunae primae ingressa est). The same chronicle mentions the event again, saying “The moon, aged 12 days, was obscured, and a very bright star came into the circle of the moon” (stella clarissima venit in circulam lunae). This last entry is said to describe a lunar eclipse, which must certainly have made the new star seem even brighter.

Stephenson & Green, in the 2003 article I am quoting these from, express doubts about whether this is really AD1054, due to discrepancies in the dates given, and prefer to attribute the description to a conjunction of the moon with another bright star such as Venus. The term circuitu/circulum used by the chroniclers puzzles even the modern historians of science (see Williams, 1981). Stephenson & Green suppose it may have meant that this star was obscured by the new moon as it moved behind its disc. But this is doesn’t make sense to me: the description says the star appeared, not disappeared

For my purposes, the identification of the event itself is not of great importance. What I want to highlight is that, at the time of the Voynich’s production, a number of chronicles were written that mentioned a very bright star entering the “circle/circuit” of the Moon when it was in its first quarter. Furthermore, since the event  took place four centuries before the chronicles were written, it is obvious that there is a missing link here: at least one contemporary source must have witnessed and written about the event at some point in order for Malvecius and others to know about it in the early 15th C (see Collins & Claspy, 1999, p.873.)

The memory of the 1054 event was clearly present in 15thC Europe, however, the original text that described it is only present in the form of the fragments which were used to compile chronicles.  Here’s something that might please Koen Gheuens:  a depiction of an elderly HRE Henry III (1017-1056) at Tivoli, pointing at what is described as a new star, in a manuscript produced in 1450 by none other than the Diebold Lauber workshop (Cod. Pal. Germ. 149, f200r). The description of the event on the next page (a most beautiful star appearing around the time of the new moon) appears to be a version of the same original text as the Rampona Chronicle (compare with Collins et al., 1999, p.4)


I see this as further confirmation that we are missing a link: one or several direct accounts of observations of the event must have existed, and must have survived until these 15thC sources which we now have.  Although Stephenson & Green have expressed doubts about these descriptions, I really don’t believe that nobody in Europe looked up at the sky for the whole time SN1054 was visible, nor that everyone in Europe was ignorant enough to confuse Venus with a whole new star for several months. Indeed, Collins & Claspy suggest that the original texts of this observation may have been largely suppressed by the Catholic Church: Comets and other celestial anomalies were seen as bad omens, and SN1054 coincided with the excommunication of the Eastern Patriarch which would lead to the Great Schism, so it is possible that the legates “would have preferred that such arguments and their visible proof not to be subsequently noted in the West” (op.cit, p. 880).

It is very tempting to consider that the circuit of the young moon which is described in these texts could be represented by an artist as a circle of crescent moons representing the Moon’s orbit (Bede, for example, uses the terms circulum and circuitum interchangeably in his texts, whether describing the Moon’s movements or those of the Zodiac).  Such an illustration might show a “new star” (well, actually a dying one, but they wouldn’t have known that) being born into the sky at the time of the new moon, in the path of its orbit.

There is something about the birth of a star that, in my view, fits rather well as a first page introduction to the imagery we see in Q9. Indeed, in some astronomical works, the description of comets, shooting-stars and other celestial mirabilia is placed at the beginning of the section on stars (see for example, Hyginus’ Astronomica).

The Rubricated Text

Rubrication was typically used in the following contexts:

  1. Initial letters and Item/paragraph markers
  2. Titles and chapter headings
  3. Colophons
  4. In calendars, to denote special feast days in a vertical list
  5. In diagrams, in alternation with black words, to distinguish nearby labels
  6. To highlight important/special words in the text
  7. To set apart notes, translations/foreign words or instructions to the reader.

Let’s consider the line of text in the bottom section of the page. It might seem like a strange place to start, but please bear with me:


sshey syshees qeykeey ykchey ykchey qokeochy oaiin okalar ol??

The first notable thing about this red line is that it is very probably a meaningful unit. As mentioned in the commentary to the interlinear translation, this is highly likely to be a sentence. This could turn out to be extremely useful, as one of the challenges posed by the lack of punctuation in the Voynich is that we have never been sure of how to isolate a sentence within the text. If there are any good candidates for sentences in the Voynich manuscript, this is the most obvious one. I am not so sure about the brown text above and below. These two lines could also be sentences, but it is less certain.

The rubricated line begins with the understatedsshey“.  This is not at all unusual. As Emma May Smith explains, it is common for linefirst words to begin with EVA s:  “Although one of the strongest linefirst letters (57.5% of all tokens beginning with <s> are linefirst), the kinds of words which come at the beginning of lines are mostly the same as those found in the whole text. There does not seem to be a further pattern except its heavy initial occurrence”.

In light of this, the first word, sshey is interesting because sshey never occurs anywhere else other than in a linefirst position. There are only 7 occurrences of sshey in the Voynich, and they are all linefirst. In fact, 100% of the words that begin with “ss” are linefirst words.

Although it is rubricated, this line of red text does not appear to be the header for a new paragraph or section: It lacks any kind of Grove word or fancy gallows and it begins like any other middle of the paragraph line. This informs us that for the Voynich scribe, the rules of text and line construction do not vary, no matter what color the text is.

It is also worth noting that the words in this red line are very rare in the Voynich text. The most common one occurs only 26 times in the manuscript, two are unique, and the others appear only 6 or 2 times. While it is normal for Voynichese lines of text to contain a few unique words, they also generally include some very frequent words: this one doesn’t.

Keeping this in mind, let’s now look at the other red text on the page, the outer circle of the roundel.


ykchs (ykees) ykchos – ykchyr (ykchys) aram – ykecho ols eesydy – soy shr okar – shekchy ?ykor – ykeody (!)  okchy – dchetdy (!) – rfchykchey (ykchykchey) ykchys – chkchdar  – ykar ykaly – lkshykchy okar – chky chykchr chy.

The base transcription here is again from voynichese.com (Takahashi) but I’ve added some alternative readings in parentheses where I didn’t agree with it. In the first two cases, the alternatives are by Grove. In the third instance, it is by Stolfi, and I am absolutely sure his version is correct: that’s a k, not an f.  Also, I’ve added exclamation marks to words where I think all transcriptions are incorrect: the d in ykeody is probably a g or an m, as it appears to have a tail, and both d’s in dchetdy are wrong, in my view – again I think we might be dealing with g’s, or in the case of the first d, perhaps even an l.

Some of the glyphs are quite hard to ascertain due to ink blobs and other malformations, and it is obvious that the scribe struggled with the red ink here even more than in the straight line. This again confirms that the scribe is not a separate rubrication specialist.

Another limit to the study of this red text is that there is some doubt about whether these red words are individual “labels” or if they are just spread out in a way that gives that impression. As Rene Zandbergen notes in the interlinear transcription, they are more or less aligned with the 12 sections that divide the circle, but not perfectly so. I’ll add that there is no divider drawn between them, nor is there a start/finish marker that would indicate a continuous line of circular text.  Each “label” word (or group of words) begins with a valid, fairly common “linefirst” character: y, s, o… Looking up first/last combinations on Emma May Smith’s blog, the last/first combination frequencies for words within these “labels” are not different than those for words across the “labels”. So, I’ll leave it at “although they don’t seem to be a continuous line of text, we can’t really be sure”.

In spite of these caveats, this circular red text, when considered along with the red line of text below, reveals some interesting things about what is going on in this page.

The first observation is that the red circular text, like the red line at the bottom, gets no special treatment from the scribe. This red text contains no Grove words, and several potential normal linefirst word beginnings.  This confirms the reading order for circular text in the Voynich manuscript: from the inside out. The innermost lines of brown text in the circle features lots of the kind of gallows we might expect for a first line. Therefore, the red line of text is not the title of the diagram, nor of its sections.

The 67r2 Text: a Strange Divide

As is well-known by Voynich researchers, the Voynich manuscript’s text roughly falls into two big “Currier languages“: A and B. Running through the list of criteria for each, Quire 9, and therefore 67r2, appears to generally be considered neither, or somewhere in between the two. Julian Bunn, in a study of glyph frequencies across the Voynich, unfortunately does not map any folios from Q9.

Taken as a whole, 67r2 presents characteristics of both “languages”: for example, it contains 24 instances of “ar”, 17 of “ol” and 9 of “ey”, which would make it more of a Currier B page. However, it also presents plenty of “chy” which is more characteristic of A, and only one lonely “ed”, the rarity of which is a hallmark of A “language” pages.

Examining f67r2 however, I noticed something unusual. While the words on the page display properties of both Currier A and B, this only holds true when we consider the page as a whole. Upon closer inspection, the characteristics of the texts in brown ink are actually quite clearly different from those of the text in red ink.


For instance, while the whole page shows 22 occurrences of aiin, only one of these is in the red text (it is in the lower line of red text; for contrast, the brown lines of text right above and below it contain a healthy 4 and 3 instances of aiin). There are 7 instances of ain, all in the brown text.  This also applies to oiin: 4 instances, all in brown. There are 8 instances of “chol” again all in the brown text, and if we expand that to “ol” then there are 17 occurrences, 16 of which are in the brown text.

The red text is devoid of any gallows other than k (except for one lone t), whereas the rest of the page abounds with t (22), f (7) and p (7). While the red text does contain similarities to Currier B, such as relatively high e’s and a standalone r, it lacks the typical “edy” combinations that would be expected for Currier B. Instead, the red text presents sequences dominated by strings of combinations of the characters k, ch, and y, especially in the circular red text where extreme strings such as ykchykchey ykchys and chky chykchr chy can be found.  While similar (but not so extreme) sequences are present elsewhere in the Voynich in Currier A folios, they don’t occur at all in the brown text on the same page.

I want to insist here on the fact that what is odd about this is not the fact that the red text should present a given mix of A and B, or even, really, the combinations of characters themselves. What is very unexpected is that it presents such different properties from the brown text on the same page.

What could explain this? The scribe doesn’t apply different rules to the construction of red text units: he used the same system to generate both. In my opinion, this might reflect the fact that something about the underlying plaintext causes the resulting red Voynichese to be clearly unlike the brown text on the same page.

A multi-lingual page?

It has been suggested that the difference between Currier A and B could reflect different plaintext dialects. However the fact that the two appear to blend in various parts of the manuscript which display properties of both and neither, could argue instead in favor of A, B, and their variants reflecting different methods of expressing the underlying text.

But on 67v2, the presence of a difference between two statistically distinct units of text on a single folio, and furthermore, the fact that one of these is written in rubricated text, leads me to the hypothesis that these just might be two actual plaintext languages, which are expressed in the same Currier “language”.

This brings me to one of the possibilities held within option 7 in the list: rubrication to set apart translated text or text in a different language. We can find this in numerous medieval manuscripts, either in the form of notes, which are generally marginal or squeezed in between the lines of the original text by a later owner, or as an integral part of a bilingual work. In this latter case, the rubricated text has been planned for: it is not just an addendum.

For example, the late 14thC manuscript below, is in three languages: English, Latin and French. In the rubricated text, John Gower apologizes in Latin, explaining that he is English and that his command of French is imperfect. A longer version of the apology follows, in French.

Pardon my French! Gower’s Traité..., WLC/LM/8, f.203v

In the example above, we can see that a couple of words in both versions are very similar, as Latin and French are related languages. This is what we observe in f67r2: the red text is clearly different from the brown one, but still related in many ways.

Another example can be found in these bilingual psalters from England, where the lines of Latin text alternate with rubricated lines in Old English throughout the manuscript

This alternation is reminiscent of what we see in the block of text at the bottom of 67r2, which raises the possibility that the red text here could be a translation or version of the text in one or both of the brown lines of text.

Similar uses of rubrication do not signal a direct translation, but differentiate incorporated text written in different languages. This can also be found in medieval scientific works, such as this example from Southern Germany c 1410, which features several illustrated folios in which the German text is in black and the Latin is in red (LoC, Rosenwald Ms 4):

Another possibility might be that one of the languages on this page actually expresses a large proportion of numbers, and the other mostly real words.

I will readily admit that analyzing the text is not my strong suit, but in this post I really wanted to look at every aspect of this page. I wouldn’t be surprised if there are flaws and errors in my analysis. I welcome any comments and critiques from anyone with a better understanding of the Currier languages, and indeed a better grasp of the properties of Voynichese in general!


P.S: The lunar phases might however be depicted in a non-sequential way in Q9. If we rebind it following Grove’s suggestion, the images below could perhaps show the moon in four different states, respectively: waxing/waning, full and new:

Q9lunar phases