The mysterious Word 2 sh-initial boost…

by ⋅ 3 Comments

Ever since Prescott Currier first pointed out that letters and words in Voynich Manuscript lines appear to prefer (or to avoid) certain line-positions within lines, pages, and paragraphs, Voynich researchers have been on the lookout for more of these unusual LAAFU (“Line As A Functional Unit”) features. This is simply because (a) they’re a bit weird, (b) they pretty much disprove any Voynich theory that claims that Voynichese is a simple language, and (c) they might possibly be hinting at what’s going on inside the ‘engine’.

So here’s a new LAAFU feature that Claude managed to dig up for me, that I certainly didn’t know before.

The Word 2 sh-initial boost

I have recently been restricting my analyses to Quire 20, because I believe that Currier B ‘evolved’ between the various sections, and I don’t want to accidentally miss features by searching too widely. (I also posted about Neal Keys.) And what emerged from all this is that the second word of a line is unusually often sh-initial.

How strong is this preference? On Tiltman lines (the top line of a paragraph) in Q20, the second word is about 2.14x as likely to be an sh-initial word (21.2% vs 9.9%) than in mid-line: on non-Tiltman lines in Q20, the second word is about 2.46x as likely to be an sh-initial word (16.7% vs 6.8%) than in mid-line. [All values from the Takahashi transcription.]

Also, sh-initial words trail off as you get closer to the line-end. Once you’ve got past the Word 2 spike, the probability of a word being sh-initial stays quite flat (~8%), before sharply dropping off in the last two (~3%).

Finally, sh-non-initial words are generally flat (2-3%) with a slight peak at Word 3 (3.4%), but trail slowly off to ~1% at line-end. Which makes sh-initial words appear to have a noticeably different positional distribution from sh-non-initial words.

As with other LAAFU features, this feels quite unlike the way you’d expect any natural language to work. But what is going on?

Contextual Factors

We might sensibly look at nearby text behaviours, to see if any of them are correlated with this. There are many weak correlations, but none explains or predicts the spike. For example:

  • It isn’t driven by the first word. After words such as daiin and dar, the second word’s sh-initial share is high at the start of a line (0.44) but low after the same words in mid-line (0.11).
  • It isn’t a page or leaf setting. Pages differ a lot in how much sh they have, but the Word 2 boost sits on top of that at about the same size everywhere.
  • Nothing carries over from the end of the previous line.
  • It occurs broadly the same in sections that seem to be recipes as in sections that seem to be non-recipes.
  • There are no specialised sh-initial words that appear at Word 2: it’s the same vocabulary, just more often sh-initial.

Claude also points out:

  • At Word 2, the ch- rate is roughly normal, so sh- is being added rather than ch- being suppressed. The openings that become rarer at Word 2 are gallows-initial and qo-initial words (and a-, ol/or- and o-initial words become commoner).
  • Something similar appears in Q13, though it fades less sharply after Word 2. In Currier A pages, Word 2 favours ch- and sh- alike, rather than sh- over ch-.

So this remains a mysterious feature.

Line-Initial Correlations

So I pushed Claude to look for any correlations between this Word 2 sh-initial spike and the first glyph of each line. Overall, the first glyph made no significant difference. The figures for individual glyphs bounce around, and even rank differently in the two transcriptions. (These are odds ratios on sh-‘s share of ch-/sh-initial words, so they aren’t directly comparable with the 2.14x and 2.46x figures above.)

The one exception was lines starting o- or qo-, where the spike seems to disappear (1.13x, against 2.46x for all other lines). Because I only spotted that grouping after seeing the numbers, I asked Claude to test it in Q13. The same thing happens there: no boost on o-/qo-initial lines (1.15x), and a normal one on other lines (2.32x).

What Is Going On Here?

It might be that an o-initial or qo-initial line-initial word is simply an ordinary mid-line word (okaiin, qokeedy, ol…), whereas most other Q20 lines open with one of the distinctive set of line-initial glyphs.

If this is right, the Word 2 spike may be triggered by (or may be a component part of) some kind of artificial or scribal line-start construction mechanism. That would seem to be a new angle on LAAFU, though that’s still not much use in moving us forward. But it’s interesting, right?

Voynich Q20: Neal Keys update…

by ⋅ 4 Comments

Recently, I’ve been revisiting the Voynich Manuscript’s Quire 20 (‘Q20’) with AI assist from Anthropic’s Claude. The idea is to look at specific Voynichese behaviours in close detail, to see if there’s anything to be found there, however small.

One of the most visually intriguing features is known to Voynich researchers as “Neal Keys”. I remember Philip Neal first describing this to me about 15 years ago (as I recall) at one of the London Voynich pub meets I used to run. So… what’s the latest update on this?

Horizontal Neal Keys

I last posted here about single-leg gallows back in 2020. Briefly, the top line of Voynich paragraphs is known as a “Tiltman line” (in honour of WWII code-breaker Brigadier John Tiltman, who had a particular interest in the Voynich Manuscript), because of its unusual properties. For example, it’s the most common place to find words containing single-leg gallows (EVA p/f), aside from the (separately anomalous) first word of a paragraph.

Philip Neal further pointed out that non-paragraph-initial words on the first line of a paragraph containing single-leg gallows are often adjacent to each other: though I calculated the raw probability of two being adjacent as 2.7%, while the actual probability is 5.4%. So this “Neal Key” behaviour is definitely a thing, though perhaps not quite as big a thing as it at first seemed.

This was the context for revisiting the topic with Claude to hand. The AI analysis proved very similar (feel free to replicate the same exercise):

  • For first-line pairs of p/f-words, there’s no particular preference for pairs (as opposed to singletons or triplets)
  • However, they do tend to cluster more than k-words or t/k-words taken together
  • The left key of an adjacent p/f-word first-line pair ends in -dy unusually often (41% vs 32% expected for p/f-words generally, or 26% for all top-line words)
  • The right key of an adjacent p/f-word first-line pair starts o- or qo- 61% of the time (52% expected)
  • Yet even so, the preference for clustering isn’t explained by Currier’s dy…qo word adjacency behaviour.

Why might this be so? I spent some time exploring the hypothesis that this might be a modal behaviour – i.e. that something on a top line triggers a “behaviour mode” where p/f-words are far more likely to happen than in other modes. As possible support for this, Claude notes that p/f-words on top lines are the only feature whose type persisted across intervening words.

To explore this more, I prompted Claude to assess whether Voynichese had any other similar modal behaviours, that would yield other glyphs or patterns of glyphs to cluster in a similar way. (Note that I was only looking at Q20, rather than the whole manuscript.) Here are some of the more modal / clustery behaviours it suggested, though note that these were all for lines with word 1, word 2 and the line-final word excluded:

  • Top lines: [qot- qot-] 22 matching lines found (13–15 expected)
  • Non-top lines: [word-containing-t word-containing-t word-containing-t]: 41 matching lines found (26 expected)
  • Non-top lines: [ot- ot-] 67 found (vs 52 expected) [though this is similar to the word-containing-t triplet result]

Generally, though, aside from this slight clustering of t-words, nothing else really behaved like p/f-words on Tiltman lines. An initial p made no difference to how many p/f-words the top line had, but they tended to turn up about a word earlier.

Another thing of possible interest was that in top lines not starting with p/f, the words before the first p/f-word included an -ar-final word about a third of the time, against about one in seven in comparable stretches of unkeyed top lines, dropping back to normal after the key. (Though Claude does flag that this was not a very strong signal.)

So, What Do You Make Of All That, Then?

Tiltman was right to point out that the topmost lines of paragraphs behave anomalously: and Philip Neal was also right to note that p/f-words ‘stick together like birds of a feather’ (my words, not his), particularly in Q20. (It’s actually largely true throughout the Voynich Manuscript, though that’s an analysis for another day.)

But on balance, the p/f-word behaviour still seems to me to be arguably more modal than key-like: the fact that it’s restricted so tightly to top lines is perhaps a giveaway that something odd is going on. And even though the way words containing t cluster is less obviously remarkable, it’s hard to exclude the possibility that it’s part of the same broader mechanism, even if we can’t quite see what it is.

I think the good news from this, though, is that trying to draw conclusions and running them past Claude does help keep you honest: any inconsistency between the responses and your summaries gets quickly exposed and ripped apart. (Which I think is a good thing.)

Claude is still very local-minded, and you have to push quite hard to get it to evaluate different hypotheses. But that’s probably a good thing, all in all: genuinely hard work remains devilishly hard, even with AI on your side.

Will The Real Q20 Please Stand Up?

by ⋅ 7 Comments

My inbox has been besieged of late by people with AI-assisted cipher mystery theories – plenty of Voynich ones (of course), but just about any historical cipher now seems to be fair game. It’s tiring: I’m trying to get through them, but there’s already a backlog, and any that’s particularly awful has a tendency to drain my will to engage with the rest of them.

But plenty of other people have been reporting success in using LLMs to crack other (perhaps less mysterious) historical ciphers, so there’s been an open question for me: can I use AI to learn anything new about the Voynich Manuscript? I tried this at the beginning of the year (2026), but got nowhere at all: almost all lines of enquiry sent me back to Cipher Mysteries pages, which – though mildly gratifying – was less than helpful.

Since then, however, models have advanced significantly, and are now able to respond to prompts by conjuring up quite complicated Python scripts, which they then run in a sandbox to give you a number or a metric.

I therefore decided to revisit Q20 – not to magically break the cipher, but simply to try to understand the first glyphs of the lines.

Some Q20 Stars May Be Fake Stars

Anyone who thinks that Q20 is just a uniform set of starred recipes hasn’t really been looking very closely. As I mentioned in 2016 and in 2010, the stars on f103r look fake to me (i.e. they’re not actually flagging genuine paragraphs). Similarly, Rene Zandbergen noted in 2016 that he thought the stars in the first two thirds of f111r looked fake. In 2021, I further noted that:

“Yet the same seems true for the top half of f111v, as well as the bottom half of f108v, and the middle third of f115r; and there’s a paragraph star apparently missing from the middle of both f106v and f113r (though the latter of these two might possibly have just slid down the page).”

I’d now add that the stars beside the “pcharalor” paragraph on f116r (just above the colophon-like “pchallarar” paragraph, on the page just before the michitonese marginalias) also seem to be fake:

Really, a lot of effort appears to have been put in to visually conceal the structure of Q20, and thereby make it seem like “just a bunch of recipes”. But why go to all that trouble?

My suspicion is therefore that if we looked clear-eyed at Q20, we might see ‘tells’ that give us clues about how it is built up. So… what can we see?

Line-Initial Glyphs

One feature Philip Neal talked about many years ago was “vertical keys”, by which he meant the way that the first glyph of lines (particularly in Q20) appeared to have quite a different distribution from other things. This is yet another of those many LAAFU (“line as a functional unit”) behaviours that often feel as though they have been inserted into the manuscript to taunt us codebreakers all the way to our miserable graves.

So this is where I started with Claude: not to crack the entire Voynich Manuscript (“MAKE NO MISTAKES”, sigh), but just to try to tease out anything interesting about the behaviour of line-initial glyphs in Q20. Note that the distribution of line-initial glyphs is biased towards d, s, y, p and t – though you may not know that of the lines beginning d-, s- or y-, the share starting d- starts from 42% in the top third of the page but then falls to 31% in the bottom third, while the y- share rises from 31% to 44% at the bottom.

For Q20, line-initial words differ from mid-line words: and so I wondered whether the line-initial glyph was a null, whether the rest of the first word (the remnant) was a null, or whether the whole first word was a null.

If the Q20 author chooses the first glyph, and the rest of the word then follows that glyph’s usual pattern, Claude reports that that would account for ~69% of how line-start words differ from mid-line words. If the author instead chooses the rest of the word, and the initial glyph is then backfilled in front of it, Claude calculates that it would account for ~35%. As a result, line-initial words largely look like ordinary Voynichese words that have been picked mainly for their first glyph.

Yet the line-initial glyph also seems to matter beyond word #1. When word #1 is da-, word #2 begins ch- or sh- about 64% of the time; and when it starts sa- (same ending, different glyph), only about 40% of the time, with a- and ol- words more common instead. So daiin and saiin aren’t interchangeable at line starts.

Moreover, after line-initial da-, word #2 closely resembles what follows d- in the other line-initial d-words ([d]chedy, [d]sheey, [d]shedy). Though note that after line-initial sa-, word #2 shows a weaker version of the same resemblance.

Putting all these pieces together: if a line-initial da- word is simply an ordinary word chosen because it starts with d-, then its remnant is effectively padding, and so the line’s real content continues with word #2. In other words, a da-initial line may in fact be a way of verbosely enciphering “d-[word #2]”. Though, of course, I have no idea why this should be the case.

Interestingly, Claude offers some related observations:

  • Where the word #1 remnant begins k- or t-, word #2 begins l- about 25% of the time (compared with 8% of the time when the same words appear mid-line).
  • Whatever word #1 is, word #2 almost never begins ot- (~1–2%, against ~9% mid-line).
  • Where the remnant of word #1 begins o- (e.g. ol, or, oaiin), word #2 begins sh- ~35% of the time, against 4% for the same remnants mid-line. (Though this occurs only 30 times in Q20.)

So there may well be other remnant-related correlations we remain unaware of.

FAAFU (“Folio As A Functional Unit”)

As part of the whole prompting cycle, Claude did report some symmetries that I wasn’t previously aware of. The biggest was that, for two of the features I looked at, the recto and verso of the same leaf matched far more often than other page pairs. Pages that currently face each other (a verso and the next recto) don’t match at all. Pages that would face each other at the centre of a singulion match only weakly.

One measure that supported this was the proportion of -ain words to (-aiin + -ain) words. This ranged from 2-5% on the f105 leaf to 59-75% on the f111 leaf, with recto and verso ratios almost always close. Claude says: “Same-leaf pages differ by an average of 7 percentage points, against 23 percentage points for random page pairs.”

Similarly, looking at how often l-initial words appear reveals the same pattern, though not quite as strongly. This is low on the f105, f112 and f114 leaves, but high on f111 and f115v.

So perhaps we should now consider adding FAAFU behaviours to our list of things that will taunt us to our demise.

More generally, Claude thought that the f108-f111 bifolio was full of weird stuff. For example, its inner side (the current centre of the quire) has more than half of its k’s followed by ee, compared with ~30% quire-wide.

There’s also a stylistic shift about halfway down f108v: there, the block starting on line 34 moves towards qok-ee- words (qokeey, qokeedy): qo- before k rises from 34% to 60%, k+ee from 41% to 65%, and -aiin falls from 10% to 3%. This block is noticeably more repetitive, and none of its lines opens like a typical paragraph. f111r’s big block shares its ee behaviour, but not its qo behaviour.

AI for Cipher Mysteries?

So, what do I think? Are LLMs going to start hacking into Cipher Mysteries and leave guest posts bragging about their cracks? Is the next big batch of wondrous OpenAI sessions going to empty my shelves of historical mysteries?

This probably isn’t going to happen any time soon. For one, the Voynich Manuscript still appears to have many fine-grained mysteries to reveal to us, all of which contribute to the much greater cumulative mystery of the thing: and even the things I found with Claude took a great deal of cajoling and iterating. (It’s always eager to try things out, which rarely helps – the first good step is normally to try to recap everything that’s been found so far, and then try to build on that.)

As always, I’d be happy to be proved wrong… but I’m not holding my breath.

Initial notes on Occitan Marginalia…

by ⋅ 38 Comments

Following on from my previous post where I gave (what I think is) an Occitan reading of the marginalia on Voynich Manuscript page f17r [“meilhor aller luce[n]t ben balsamina“], I’ve been looking for good sources on Occitan marginalia. My plan for this post is to start searching for all Occitan marginalia and then focus on the period 1400-1450 later once I’ve got a general handle on the broader literature.

Paul Meyer

The first philologist of interest here (I believe) was Paul Meyer (1840–1917), who spent more than forty years tracing the geography and chronology of Provencale-language documents. Having worked his way through the BnF’s Occitan documents (most of which date to the 13th and 14th centuries) and some English libraries’ ones too, he ended up looking at Occitan marginalia.

Interestingly, these marginalia often appeared not in Occitan documents, but in French and Latin documents. Though Meyer doesn’t seem (unless you know better?) to have explicitly expressed it, his opinion appears to have been that even though Southern French scribes typically wrote Latin or French, many (possibly even most?) spoke Occitan.

Even so, finding 15th century Occitan marginalia has proved somewhat unrewarding. One adjacent article I found was Lyudmila Shegoleva’s “Marginal Notes in Latin and Old Occitan in the Bible from the Collection of N. P. Rumyantsev”, but that only relates to some 14th century marginalia, alas.

Jean-Baptiste Camps

So far, the most helpful thing I’ve found is“Les Manuscrits Occitans À La Bibliothèque Nationale De France” (2010), by Jean-Baptiste Camps. Camps follows in Meyer’s footsteps by grinding through all the Occitan documents in the BnF he could find. (He notes ruefully that the BnF’s indexes didn’t make it easy for him.)

Here’s what Camps has to say about scientific texts (my loose translation) [p. 45]:

The inclusion of scientific texts in this inventory of medieval literary manuscripts might seem surprising. However, it is entirely justified. The medieval definition of science may be somewhat disconcerting to us modern positivists; this is because, back then, ‘science’ had no single definition. It mixed astrology and alchemy with mathematics and medicine, and accepted them all as sciences. In these texts, and particularly in the medicinal recipes that fill the margins of many manuscripts, we find medical theory (inspired by the rediscovery of Aristotle via the Arabs) jostling with popular beliefs, which occasionally possess a core of empirical truth. This heterogeneity, which is sometimes found even in the writings of recognized authors,[78] sometimes yielded controversy: consider, for example, the polemic that kicked up at the University of Montpellier in the years 1426–1428, pitting those who proposed using astrology and talismans in medicine, who included innovative physicians such as Nicolas Colne, dean of the faculty of medicine, and Jacques Angeli, against the supporters of a more classical medicine, led by the former University chancellor Jean Piscis and supported by Jean Gerson. The latter group reproached the former for its use of a ‘Lion’ talisman for kidney problems, reigniting a debate that had raged amongst Montpellier’s Jewish physician community a century earlier.[79]

  • [78] For example, the Catalan physician Arnaud de Villeneuve; on this subject, see Nicolas Weill-Parot, “Astrologie, médecin et art talismanique à Montpellier : les sceaux astrologiques pseudo–arnaldiens”, in L’Université de médecine de Montpellier et son rayonnement (XIIIe–XIVe siècles), actes du colloque international de Montpellier (Université Paul Valéry–Montpellier iii), 17–19 mai 2001, dir. Daniel Le Blévec, Turnhout, 2004, pp. 157–174.
  • [79] Ibid., pp. 157–158.

More specifically, Camps discusses (pp. 45-46) medical texts:

Among all the didactic and scientific texts, medical texts occupy a special place in Occitan literature, because medicine was likely one of the important areas of interest in the Occitan-speaking world. This importance is linked to the role played by southern France in disseminating medical science stemming from the “Toledo corpus” and, through it, Aristotle’s texts transmitted via the Arabs [80]. Montpellier and its university are thought to have played a particularly important role in this process [81]. Most medical writings in the Occitan-speaking world were written in Latin, and only a few were translated into Occitan [82], although it is “not implausible” that teaching in Montpellier – at least at first – took place in Occitan and that various texts were first translated from Arabic into Catalan or Occitan before being translated into Latin [83]. The extant translations we have concern three main areas: surgery, firstly, which was probably forbidden to clerics and stood apart from the rest of the medical field; secondly, anatomy, going hand in hand with the first, perhaps because in Montpellier, they continued to dissect cadavers even after that was prohibited; and finally, recipes, blending scholarly and vernacular influences.

What we said earlier about scientific texts is particularly true for medicine and pharmacy, which combine astrology with recipes of popular origin. The material that constitutes these texts is therefore quite heterogeneous, and led Maria Sofia Corradini Bozzi to say that “the medieval pharmacopoeia, intertwined as it was with herbalism, magic, and popular beliefs, opens up a research perspective whose dimensions and boundaries are often not easily definable.” [84].

These recipes survive in two main forms: firstly, we find them collected in manuscripts from both Languedoc and Provence; or secondly, very often, in the margins and blank folios of manuscripts, generally in Latin, and containing either religious texts (this is the case of the BnF Latin manuscripts 2459 and 2941, but also of the Grenoble manuscript, BM 159), or alchemical texts (BnF Latin manuscript 11202), or legal texts (BnF nouv. acq. fr. manuscript 11151, Bordeaux, BM 355) [85]. These examples, though still relatively little known, have been or are in the process of being edited and published. In terms of their content, while it is true that they sometimes contain elements from oral culture, they are often also “faithful translations of passages from works composed in Latin by medieval authors” [86] and sometimes seek to integrate “the results of the first experiences that were being conducted in the nascent medical schools with notions coming from the classical Greek-Latin tradition, often mediated through the Islamic world” [87]. They may also contain elements relating to major medical discoveries of the medieval period, such as MS Arsenal 8315, which contains on folio 26v a text on the virtues of brandy inspired by the writings of the Catalan Arnold of Villanolva, generally considered its inventor.

  • [80] Linda M. Paterson, “La Médecine en Occitanie avant 1250”, in Actes du 1er congrès
    international de l’Association internationale d’études occitanes    , dir. Peter Ricketts, London, 1987,
    pp. 383–399, p. 383.
  • [81] Ibid.
  • [82] At least as many have been translated into Hebrew, Ibid., p. 392.
  • [83] Thus, the explicit of the 26th book of the Tesrif of Albucassis, in ms. BnF nouv. acq. lat. 343, indicates “Hic finiuntur XXVIII capitula hujus libri Abulcasim Azaraui in cibariis egritudinem translatus de arabico in vulgari catalanorum et de vulgaris in latinum” ; P. Pansier, “La pratique de l’ophtalmologie dans le Moyen Âge latin”, in Janus, t. 9 (1904), pp. 3–26, p. 7 ; cf. L. M. Paterson, “La Médecine en Occitanie…”, p. 395, note 25.
  • [84] Maria Sofia Corradini Bozzi, “La Fachliteratur occitanica : i codici di argomento medico-farmaceutico”, in La Filologia romanza e i codici (atti del convegno, Messina 19–22 dicembre 1991), dir. Saverio Guida et Fortunata Latella, Messine, 1993, t. 2, pp. 731–742, p. 731.
  • [85] Ibid., pp. 736–737.
  • [86] Ibid., p. 739.
  • [87] Ibid., pp. 739–740.

(I should also mention that Camps’ bibliography appears to be an excellent place to work outwards from.)

Occitan recipe marginalia

From Camps’ section on recipes, we can quickly extract a list of mss to look at, and find out their dates:

  • BnF Latin 2459 – 13th century
    • F.1 “Contra unglas fendudas”
  • BnF Latin 2941 – Occitan recipes added in the 14th-15th centuries
    • F. 84r-84v Recettes de médecine en provençal : « Recipe goma blanca… »
  • Grenoble manuscript, BM 159
  • BnF Latin 11202 – 15th century (Gallica images are here)
    • Also mentioned on Voynich Ninja by nablator in 2023
    • The bibliography includes: D. Kahn, “Littérature et alchimie au Moyen Age : de quelques textes alchimiques attribués à Arthur et à Merlin”, Micrologus, t. 3, 1995, Le crisi dell’alchimia. The Crisis of Alchemy, p. 227-262
  • BnF nouv. acq. fr. manuscript 11151
    • “Contains an Occitan translation of the Mulomedicina and La recepta del vi”
    • See: Borgognoni, Teodorico “Traduction provençale de la Mulomedicina”
  • Bordeaux, BM 0355 – 14th century
    • Mentioned on JONAS.
    • ff. 302v-302v: “Eletouari dels dens… — Eletouari per fer bellas les dens… — Per orinar… — Polvora per apacie colica…”
    • See: Maria Sofia CORRADINI BOZZI, « Sulle trace del volgarizzamento occitanico di un erbario latino » in Studi Mediolatini e Volgari, 37 (1991) : pp. 31-132
    • See: Maria Sofia CORRADINI BOZZI, « Per l’edizione del corpus delle opere mediche in occitanico e in catalano : nuovo bilancio della tradizione manoscritta e analisi linguistica dei testi » in Studi testuali, 3 (2001)
    • See: Maria Sofia Corradini Bozzi, « La letteratura medica medievale in lingua d’oc fra tradizione antica e rinascimento europeo » in El saber i les llengües vernacles a l’època de Llull i Eiximenis. Estudis ICREA sobre vernacularitzacio, Barcelona, Publicacions de l’Abadia de Montserrat, 2012

Where next from here?

Nicely, all of this flapping around eventually managed to lead me to the site I was faintly hoping someone had built (but didn’t actually know about). This turned out to be trobaretz.wordpress.com, which contains links to hundreds of manuscripts containing Occitan texts or fragments.

Also, I have lots of papers by Maria Sofia Corradini Bozzi to track down and read. Which is nice.

1908 medical reference to Hallermann–Streiff syndrome (HSS)…

by ⋅ Leave a Comment

I was reading about Hallermann–Streiff syndrome this morning, and was a little surprised by the huge gap between Aubry (who first discussed elements of it in 1893, though only partially and in the context of congenital alopecia) and Hallermann (1948) and Streiff (1950) (after both of whom the syndrome was named).

And so I went a-looking, to see what I could find. I searched Hathitrust for “congenital alopecia”, ordered the results by date, and was so amazed by the first result (and also by the fact that it wasn’t visible on the Internet) that I decided to turn it into a webpage.

So, here’s the link to my typed-up version of “Congenital Alopecia as an Expression of Atavism” by James Nevins Hyde, M.D. (1908).

Of course, 1908 was far too early for Hyde to have a modern-feeling ‘syndrome’ framework to place his observations inside. But, to be fair, I think he did as well as he could: the places where he reverts to atavism or to reversion to lower mammalian or amphibious types feel almost like dated science fiction to our modern syndrome-attuned ears.

So maybe dismissing all physicians of that era would be somewhat unfair: Hyde did the best he could to capture, cross-reference and consider the condition, with clear and careful photographs that are useful even now. And maybe the websites discussing the history of HSS should mention Hyde’s contribution?

Was Nick Redfern’s “Black Widow” actually Millicent Dillon? And what about “Project Chickenpox”?

by ⋅ 4 Comments

Having just posted about Project NEPA (and now reading Alex Wellerstein’s fantastic “Restricted Data”), something clicked in the back of my head about Nick Redfern’s source “Black Widow” (in his book “Body Snatchers in the Desert”).

According to Redfern, she claimed that “[she] met a lot of the NEPA and ANP people at Oak Ridge...”. But Project NEPA started as part of the Manhattan Engineer District (under General Leslie Groves), and then control moved to the Armed Forces Special Weapon Project (again, under General Groves), and then to the Power Branch at AFWSP (under Donald Keirn), and then eventually to the AEC Division of Reactor Development (on 01 Feb 1949, the day after General Groves retired – which was probably no coincidence).

All through that process, the highest security rules applied: so Project NEPA really wasn’t something you could talk about idly over the watercooler, at Oak Ridge or indeed anywhere else. All of which means – I’m pretty sure – that the “Black Widow” can only have been working directly on Project NEPA. Which in turn made me think of…

“In the Atomic City”

Here’s Millicent Dillon’s account of being a young female graduate physicist working in “Atomic City” on Project NEPA in 1947. Dillon (nee Gerson) moved from being a physicist to being a writer a little later in life, and for me her writing has a nice, reflective tone that’s easy to read (but is hard to do). Other people have commented on the ‘precision’ of her writing, as though it springs from her time as a technical writer in Project NEPA’s Information and Handbook department. But precision can be learnt: emotional honesty takes proper talent.

(Nice picture from this interview here with her in The Awl.)

Might she have been Nick Redfern’s secret source? It’s entirely possible, and as Dillon (sadly) died in January 2025, perhaps Nick R will tell us at some point. (Go on, Nick!)

She actually spent some time writing her memoirs (“In the Atomic City” was effectively an extract published in The Believer magazine). These were to be called “The Absolute Elsewhere”, but there’s no obvious sign that they were ever printed. But goshdarn, I really wanted to read them.

So I then found that her papers (19.32 linear feet!) had been donated to the Harry Ransom Center at the University of Texas (MS-01167). However, their date range is marked up as “1905-2002”, which feels as though “The Absolute Elsewhere” was probably retained by her family. So maybe we will see this published someday. (If so, please leave a comment here, I’ll be the first to buy it!)

Incidentally, her “Atomic City” article mentioned a book called “Race to Oblivion” by Herbert York, which you can borrow via archive.org. In fact, more or less all the people from this period or Project NEPA seem to have written their memoirs, which (to be honest) all seem to say little or nothing about the important stuff. It will be interesting to see how genuinely honest Dillon’s memoirs will turn out to be.

The May 1947 test (that failed)

In Nick Redfern’s book, the Black Widow mentions “that there were, I think, three classified balloon flights in May, June, and July 1947 and at least two were disasters. […] Sometimes there would be experimental shielding in a radiation experiment and the people would be separated – some having shielding, pressurisation, and some not, depending on the experiment.” (Body Snatchers, p. 9)

Back when I first read this, it made no sense to me at all. The weight of radioactive shielding would be just colossal, far greater than any balloon of this period could manage. Really, any experiment with this kind of shielding would need to go up in a meaty, chunky ol’ aircraft. And surely any plane with Project NEPA would have gone up much later than 1947? So I didn’t really see at all how this could have worked.

However, I recently found out about Project Chickenpox, MX-886. This was a project that General Groves helped kick off in October 1946 to make airborne atomic assembly laboratories. The basic idea was to kit out some C-97 aircraft with the necessary laboratory materials so that an atomic bomb could be assembled while in flight to its target (i.e. rather than having to wait for one to be assembled on the ground before loading it onto the plane). This is described in General Groves’ secret correspondence files (microfilm M1109), which you can buy online: and also in the AFSWP first year history (section 4.6.4 (e)).

Connecting the dots

Until today, I didn’t see how any of these individual points of data might be connected. But then I learned that one of the three C-97 planes owned by AFSWP crashed on 22nd May 1947. Here’s the report from the Prescott Evening Courier of the same day:

5 KILLED IN PLANE CRASH

Dayton, O., May 22 (/P) – Five persons were killed and two others were injured today when a four-engined army transport plane from Wright field crashed and exploded in a field about four miles east of Dayton.

Names of the dead were withheld pending notification of relatives.

A witness said the plane appeared to be in trouble as it swooped low. Seconds later, the plane tilted on its side and plowed into a ditch, then hurtled into a wheat stubble field.

An explosion turned the plane into a mass of flames.

Wright field officials said the plane, containing military personnel, was on a routine test flight.

The plane, a C-97, was the military version of the Boeing stratocruiser – a giant ship capable of carrying scores of troops or tons of cargo.

Wright field officers said they believed the plans stalled as it approached Wright field for a landing.

The injured personnel, whose names and conditions also were withheld were thrown clear of the plane.

They are at Patterson field hospital.

Might this – a secret C-97, easily able to lift a Project NEPA nuclear reactor and shielding into the air – have been one of the tests that failed disastrously?

How did Project NEPA actually start?

by ⋅ Leave a Comment

I’ve been trying for a while to find a reliable insider account of the early days of Project NEPA. Helpfully, I found two accounts of Project NEPA written by Lt. Col. Clyde D. Gasser, the Air Force Engineering Officer for project NEPA in Spring 1947. Incidentally, his 1990 obituary is here, and I also found a page collecting together his thoughts on UFOs (spoiler: he thought they were Russian).

Gasser reported to (later Major General) Donald J. Keirn, the Chief of the Power Plant Laboratory at Wright Field. Keirn had previously been liaison officer to the Manhattan Project, but in May 1947 was appointed special assistant to the director of the Division of Military Application, Atomic Energy Commission (Brig. Gen. James McCormack, Jr.).

Gasser’s first account was a report on NEPA dated 30th May 1947, which – to be frank – was a load of stakeholder-facing success theatre. (Though to be fair, it does have a lot of useful details that help stitch NEPA into the broader military/AEC fabric.) The second Gasser account, however, seems to have been written during the 1960s, and tells much more of the story behind the project. Though this second account has, especially towards the end, a reek of ideological indoctrination and denial (i.e. ‘sure, the project failed, but history will prove us right‘), it is at least accessible and direct about what happened.

I also found Chapter 4 of “Race to Oblivion: A Participant’s View of the Arms Race” (1970) by Herbert F. York fairly useful, along with Lee Hite’s “Giving Wings to the Atom“.

The origins of NEPA

During WWII, the notion of nuclear powered aircraft seems to have been ‘in the air’: Enrico Fermi discussed it in 1941, and (then-Colonel) Donald Keirn certainly seems to have talked about it too. But neither of them had the political muscle to turn it into any kind of scientific R&D programme.

Project NEPA seems to have stemmed from a ‘meeting of minds’ between Army Air Force General H. H. “Hap” Arnold and much-lauded physicist Theodore von Kármán. Back then, every senior figure in the US figure was haunted by the spectre of what had happened at Pearl Harbor: they were therefore alert to any possible technology that offered them the chance to not be caught out again. So at that precise moment, von Kármán’s optimistic gloss on possible technologies was exactly what the General wanted to hear.

Arnold therefore set up a Scientific Advisory Group of the US Air Force (later the Scientific Advisory Board) headed by von Kármán. This yielded a series of short works describing the scientific state of the art in various fields relevant to the Air Force, ranging from guided missiles to aviation medicine: “Where We Stand” (von Kármán’s initial report) and “Toward New Horizons” (“Science, the Key to Air Supremacy“).

But of all of these, it was the chapter on nuclear powered aircraft (“Possibility of Atomic Fuels for Aircraft Propulsion of Power Plants” by Hsue-Shen Tsien) that seems to have captured the imaginations of Hap Arnold and his Air Force planners. For the Cold War with the Soviet Union that they could all see just starting, they came to believe that nuclear aircraft – though technologically challenging – would give the US Air Force numerous structural (and very hard to challenge) strategic advantages:

  • Super-high altitude (because the thin upper air would not limit nuclear power efficiency)
    • Too high to be intercepted by conventional aircraft
  • Super-long endurance (missions could spend days if not weeks in the air)
  • No need to manage a global set of airfields for their (conventional range) bombers to launch from

Remember that the dangers of nuclear fallout weren’t properly understood back then; and though ICBMs had been proposed, the conventional wisdom was that such missiles would be many decades away, and even then would have highly uncertain reliability or precision. For the Air Force planners peering into the future, the only decisive theatre of war they could foresee had bombers and nuclear bombs. And so nuclear aircraft seemed to them to be a logical next step (albeit a monumentally large one).

NEPA, 1946-1947

At this point, Colonel Keirn – who is essentially the hero of Gasser’s account – took the whole idea to General Leslie Groves at the Manhattan Engineering District (basically, what was left of the Manhattan Project after the war), and started to talk about it. However, the Air Force hadn’t been part of the development of the atomic bomb, and so didn’t really have sufficient technical expertise. But at this point, Fairchild Engine and Airplane Company stepped forward: and they were really, really interested in designing and building nuclear-powered aircraft. And that’s basically where it all actually started.

The initial AAF letter of intent by General Spaatz and Maj. Gen. Leslie Groves on 28th May 1946, committing $200,000 to the project (of course, this rapidly increased). Early funding for the project came from the Air Force Research and Development section, with the AEC contributing to the budget from 1949 onwards. For the first fiscal year (1946), funding was $1.3m: by 1951, this had steadily grown to $8.3m. (York p.61)

Then on 15th July 1946, the project moved into a suite of rooms in Raleigh Hotel on West 72nd St. New York City. This was briskly followed on 4th September 1946 by a move to the AEC Administrative Building in Oak Ridge (which at the time was still Clinton Laboratories): and then on 24th September 1946 onwards to Oak Ridge’s S-50 Area near the K-25 Power Plant. (Hite p.3)

This was all happening while the Manhattan Engineering District was approaching the end of 1946, at which point it was handing over some (but not all) of the atomic reins to the newly-formed civilian Atomic Energy Commission. So in February 1947, control of NEPA moved to AEC Subcommittee of the Joint Research & Development Branch (known as the JRDB).

Secrecy and Control

How was the AEC – a civilian body – able to take control of a top secret military project in February 1947? When you put it like that, it doesn’t really make sense.

My understanding is that what really happened is that because the AEC had control of key parts of atomic research (as well as the equipment and the raw materials, such as uranium), they took notional control of the project. However, military control of the project stayed with General Groves (albeit invisibly) as he sidestepped from the Manhattan Engineering District to the newly-formed Armed Forces Special Weapons Project.

The AFSWP was a joint military organisation (i.e. Army, Navy, and Air Force) that officially started on 29th January 1947. It started its life in the former offices of the Manhattan Project (the fifth floor of Washington’s New War Department Building), but on 15 April 1947 it moved to the Pentagon, where it sprawled and sprawled as it grew.

This speaks to what I dislike about the way Project NEPA is almost always presented: that the entire layer of military control via AFSWP is essentially absent. Metaphorically, if the AEC owned the NEPA house, Groves and the AFSWP had the keys (and had installed all the locks too). And I’ve read elsewhere that on anything to with secret stuff, the rest of the AEC Commissioners normally deferred to former Navy Admiral Lewis Strauss, because of his high security clearance. (Though note that Strauss had never seen action, his war was spent in the USNR working for the Bureau of Ordnance.)

Once this whole picture is reframed with AFSWP in its correct place, NEPA looks far less like a post-war experiment in civilian control and far more like a continuation of Manhattan Project secrecy culture by the back door. I just wish that even one of the 50+ articles and books I’ve read about Project NEPA had bothered to point this out. But now you know.

Looking for a researcher near Montgomery, AL…

by ⋅ 14 Comments

My search for scientific balloon history 1945-1949 has just taken a sharp turn to one side. I’ve managed to locate a folder in the Air Force Historical Research Agency at Maxwell AFB in Montgomery, Alabama that has (or, at least, in 1986 had) information about some super-early scientific balloon stuff that might possibly be what I’ve been chasing after for a while now.

(Just so you know, this is the inside of a KC-135 Stratotanker at Maxwell AFB.)

But – Lawdy, Lawdy – finding an independent researcher in Alabama happy to go onto a military site isn’t proving easy at all. I know, it’s only a single folder, so it’s barely even a day gig, which isn’t much. But when you have nothing, even a small something can be a huge deal. It eez what it eez.

So… does anyone here happen to know someone who might fit the bill? Or have any suggestions as to how to find someone who might fit the bill? Thanks!

Finally, I can read the Voynich Manuscript! [*]

by ⋅ 25 Comments

[*] Well, a bit of it, anyway.

Over the last weeks, my Cipher Mysteries inbox has been inundated with AI-generated theories. But – and with my apologies to the genuine cipher theories that also landed there, which I promise I will get back to – I’ve instead been focused on the Voynich Manuscript. Or, more specifically, on a single line of its text.

And I think I can read it.

The f17r marginalia

Back in 2006, I was hugely fortunate to be allowed by the Beinecke Rare Book & Manuscript Library’s curators to spend a few days working with the Voynich Manuscript. One of the highlights of that trip came when I borrowed their UV blacklamp to examine the tiny marginalia at the top of page f17r.

To my huge surprise back then, what the UV light made visible at the end of that line was some Voynichese text. Later on, another group did some multispectral scans of f17r, so here’s what they saw (“Voynich_17r_WBUVUVP_019_F”, auto-equalised in Gimp):

Why was all this important? Because it strongly suggested that there was some kind of direct link between (one of) the marginalia people and the curious Voynichese writing itself. (And also that the Voynichese letters on the final page (f116v) were probably not coincidental). For me, this all suggested that understanding the Voynich Manuscript’s marginalia might not only tell us something about a later owner of the manuscript, it might also tell us about its creator(s).

So the marginalia are a big deal to me. And if you’re interested in the Voynich Manuscript, they really should be a big deal to you too.

Theories about the f17r marginalia

So, what does it say, what does it say? Poundstone thought that the first word might refer to herbal writer Pierandrea Mattioli (1501-1577): Brumbaugh similarly thought it might refer to Mathias de l’Obel (1538-1616). But these both feel quite wrong, because the cursive gothic handwriting is typical of the (mid-)15th century, not of the 16th century.

Since then, there have been plenty of partial readings of the f17r marginalia, most of which seem – possibly emboldened by the apparent fragments of German writing on f116v – to be German-ish. But though such readings typically start promisingly, they quickly fall to pieces when you look more closely.

A special mention here to two good attempts:

For “The Curse of the Voynich” (2006), I had my own proposed reading (on pp. 24-25), but I’ll come back to that in a minute. Still, the important thing I noted was that the second letter of the first word appeared to have been emended: and so consequently the first word actually seemed to have originally been “meilhor“.

The mysterious third word

Also back in 2006, I wondered if the third word might be “lutz” (the Occitan word for “light”). Yet that (and almost all the other theories about how to read this word) failed to explain why there was a macron (overbar) over the end part of that word. In 14th-15th century texts, the macron was widely used as a way of inserting a missing ‘n’. So… why on earth was there a macron over this word?

Back in 2017, Helmut Winkler posted this on Voynich Ninja:

I think there are several more ways to read the “lucz”,  e.g. lucem or lucet, but in this contextI I would suggest  luc[ea]m her[bam], one of the Alchemical Herbs

Now fast forward to a July 2025 comment by Marco Ponzi, mentioning some text from Fribourg MS L.52 (f.8v): “oleum lucet, balsamum redolet” (and I have a ton to say about L.52 on another day):

Marco then noted:

For the third word in the Voynich f17r marginalia, assuming that the text is Latin (very doubtful) and that the initial [letter] is ‘L’, a possible reading could maybe be “lucent” (where the macron stands for the missing ‘n’) – “they shine”, third person plural.

You can’t fault Marco’s logic here: even though adding the macron would ‘nasalise’ lucet into lucent, it overall does not look like Latin. So… how can we resolve all this?

Reading f17r marginalia (finally)

The reason I’ve ground through all the above (giving credit to everyone who helped with all the steps along the way) is that the final reading didn’t just magically pop into my head. I started looking at this properly in 2005-2006, so it has taken nearly twenty years to get to (what I believe to be) the end line.

When writing Curse, my tentative reading of this marginalia was that this was Occitan, and that it began:

  • meilhor aller lutz [kou?]…

…which was close, but no cigar. With the benefit of all the above, I am now pretty sure that it is Occitan, and that it reads (using a Latin abbreviation style to render the Occitan “lucent”):

  • meilhor aller lucent ben balsamina [….]

Or, one multispectral block at a time:

meilhor aller

luc[ent] ben

balsamina

Balsamia / Balsamina

If you look at Wellcome MS.626 (Livres des simples médecines), you can easily find (because it’s arranged alphabetically) balsamia [balsam] on folio xxix:

To be precise, this is talking about the original (and mythical, almost unobtainium-like) balsam from the East, a plant known not to anyone by actual experience. So this is reporting – medieval herbal-style – on a plant without flowers yielding a kind of resin, and the artist is just guessing at what it might look like. It’s really not a literal drawing of a tree.

And yet a century later, Leonhart Fuchs in his 1542 Hist. Stirpium was using the word ‘balsamines’: “Duo Balsamines genera damus“, and the word balsamine was being used in French in 1545 to mean “balsam-like”. So even though this isn’t “balsamina” in its modern sense (e.g. impatiens balsamina), it is a word that is specifically being used to evoke balsam-like qualities.

And I think the word we’re seeing at the top of f17r is “balsamina”.

Occitan marginalia. Really? Really.

So, my argument here is that the marginalia at the top of Voynich manuacript page f17r is written in Occitan. And guess what? Back in 2006, I argued long and hard that the Voynich zodiac roundel month names (which also appear to be marginalia) were also in Occitan. So this should, in theory, be the least surprising marginalia language identification ever.

And yet I already hear every single Voynich Ninja commenter disagreeing. Pffft. It is what it is. It’s Occitan.

If the Voynich Manuscript was written by people who appear to have been writing natively in Occitan, the first thing we should now be doing is looking at every single Occitan herbal-related manuscript from the period, such as BNCF Manuscript Palatino 586. Roll with it for a change.

The Roswell capsule: was it Tex Settle’s Flying Coffin?

by ⋅ 21 Comments

As I noted last year, it’s well documented that the round-ended capsule-shaped metal stratospheric gondola designed by Charles Burgess and Tex Settle (referred to as “The Flying Coffin”) was approved by Rear Admiral Moffett, who died in April 1933 (on the US Navy airship USS Akron). It’s also known that it was constructed in the Naval Aircraft Factory in the Philadelphia Navy Yard, but never flown.

And really, you have to admit that witness descriptions of the 1947 Roswell Incident capsule – also a round-ended thin-skinned capsule-shaped hollow metal ‘thing’ – do sound extraordinarily similar. The only obvious difference (from what little we know of both) is the size, with the Roswell capsule being described as slightly larger (but not really by very much).

But even so, might the two have been the same thing? Was it the Flying Coffin that was found in Roswell? It’s a question I’ve been trying to resolve for a few years now. And I think I’ve now found an answer, but in a place I never expected…

Aluminium Alloys

Pure aluminium is nicely malleable – hammers and a buck, and you’re basically there. So if you were making a round-ended capsule (i.e. one with no stress concentration points), it would be easy in a factory to shape the aluminium gores into their desired final shape.

George Melies "A Trip To The Moon". Factory workers hammer a rocket into shape.

Auguste Piccard insisted on using pure aluminium for his gondolas, because (as I recall) he wanted to avoid stress failures while high up in the stratosphere. Which, to be fair, in the 1930s was entirely reasonable. But more generally, to prospective stratospheric gondola makers, pure aluminium isn’t really as strong as you’d like. Resisting the pressure difference in the stratosphere would mean a pure aluminium gondola would need to be quite thick – and quite heavy.

Yet at the same time, even by 1933 there was moderate knowledge of aluminium alloys. Duralumin was an Al–Cu–Mg alloy that had been invented in 1909: it included a thin aluminium “top coat”, to help cope with weathering. Yet even though Duralumin was widely used in aeronatical engineering (Zeppelins, airships, even the USS Akron, though it was a storm that brought that down), 1933-era engineers were still wary of its potential for stress fractures. But if not that, then what?

Looking at the range of aluminium alloys known in 1933, what looks to my eyes a safe middle ground (between pure aluminium and Duralumin) for stratospheric gondolas would have been an aluminium alloy formed of aluminium with ~1% manganese. (Less than 0.7% manganese doesn’t really gain strength, while more than 1.5% manganese and becomes more brittle.) While not as strong as Duralumin, aluminium with ~1% manganese would be nearly as malleable as aluminium, but significantly stronger.

Even in 1933, a top supplier like Alcoa could give you access to pretty good aluminium (if not particularly pure compared to modern day materials), and with whatever extra stuff you wanted mixed in. So, the Flying Coffin could have been (and I believe indeed probably was) made of an aluminium alloy, though rather than Duralumin I would guess Al + 1% Mn. And that was pretty damn impressive stuff.

Welding in 1933 and 1947

Impressive… except that aluminium alloy was only a part of the story. In 1933, the only real way to join gores (curved panels) together was with oxy-acetylene welding. And oxy-welded seams were – even in the hands of top welders – not great, not great at all. The poor seam quality meant that you had to use much thicker panels that you would like, to be defensive against leaky, cracky, achey-breaky seams.

That in turn would pretty much double the weight of the capsule, whichever type of aluminium you used. And so what I think we can predict about Settle and Burgess’ Flying Coffin is that:

  • it was a bit of chunky boi
  • it would have had visible, fairly ugly seams
  • it would have felt basically industrial, not futuristic

However, fast forward to 1946-1947, and metal assembly had been revolutionised by TIG (Tungsten Insert Gas) welding. (It’s known as WIG in Europe, because Tungsten was also known as “Wolfram”.) This was developed in 1941-1942, so would not have been available to Settle and Burgess in 1933.

TIG welding was a world away from oxy-acetylene welding. All of a sudden, beautiful clean (and non-leaking) seams were possible, and even the norm. And this in turn meant that thinner sheets of aluminium alloy could be used, with the seams neatly sealed and then polished and brushed.

So the question really is: do you think the Roswell capsule would have evoked shock and awe if it had been constructed using 1933’s thicker aluminium alloy and oxy-acetylene welding? Or would 1947’s shiny (and – literally – seamless) lightweight capsule have given everyone who looked at it a wobbly feeling of glimpsing an unknown, alien future?

My own conclusion is that it was probably not the Flying Coffin, but a TIG-welded post-WWII round-ended aluminium alloy capsule. And if that’s correct, whoever stumbled across it in the middle of nowhere north of Roswell might easily – absent a genuine explanation – end up jibbering for years.

The Shock of the Really New

Arthur C. Clarke’s famous adage “Any sufficiently advanced technology is indistinguishable from magic” was something he write in “Profiles of the future; an inquiry into the limits of the possible” as a footnoted aside. I suspect Clarke wanted his aphorism to play to the sci-fi gallery, to the readerly desire for curiosity.

But if I’m right, the evidence from Roswell hints at something slightly different (and I don’t intend this as a snowclone at all). Rather: it suggests that any sufficiently advanced technology is first and foremost a shock, one powerful enough to disrupt your ability to make sense of it.

In the real world of 1947, the US Army was presented with what must surely have seemed like Clarke’s “sufficiently advanced technology”, and the reverberations of that shock – plus the long chain of tragic misunderstandings it triggered – are still very much with us, nearly 80 years later.