RE: An interview with Adam Lewis
nablator > 03-03-2018, 12:56 AM
- Koen: Hello and welcome to our fourth Voynich Ninja webchat. In the previous installments we spoke to established figures in Voynich research but we like to hear some new voices as well and that is why today I will be having a chat with Adam Lewis from Puget Sounds University. He is currently writing his senior thesis entitled "An Anatomy of failure: Analysis Attempts to Decode the Voynich Manuscript". So Adam, you're writing your thesis about the Voynich manuscript, that's not an everyday choice. How did you get interested in the manuscript?
- Adam: As far back as I can remember ever since I was a kid, I've always found it interesting. I remember I would oftentimes watch documentaries on History Channel and things that were language or linguistics related always kinda stood out to me and I just remember, sometimes there were things about the Voynich manuscript on and it would always just be really interesting to me, you know. I'd also wonder, you know, what is the mystery behind what's actually in the manuscript. It's just so... I don't know, it just really piqued my interest and then, as I was growing up I just found I continued to have an interest in language, linguistics related things. When I was looking at colleges to attend, one of my considerations was if I go somewhere with a good linguistics program I would still be interested to study this, I actually, when I was touring around, one of my stops was Yale where it is held and I got to actually see it in person which was pretty cool for me. Yeah! I feel like if I ever go to graduate school, that's in the future, I would definitely strongly consider wanting to do more with this from a linguistics area.
- Koen: Right. So how did your teachers, how did your professors react when you proposed this subject? Did you have to discuss it with them, or?
- Adam: It's not, you know, the most well-known thing unless you already kinda know about it but I guess you know that. So I did have to explain a couple of times to different people, you know what it is. I'd say I'm doing my thesis on the Voynich manuscript and they'd be like: "Oh? What's that?" I think maybe one or two people had heard of it but... and then I explained: "It's this manuscript, allegedly a couple hundred of years old, written in a totally unknown script". I gave the whole spiel and they were like: "Oh, it's very interesting".
- Koen: Alright so in your thesis you write about failed attempts to decipher the Voynich and we've seen quite a few of those lately so where do you even start?
- Adam: I actually don't look at so many specific failed attempts. I look at a couple of them, but what I would end up actually doing is: I look at interesting properties for the Voynich manuscript and I actually look at unrelated but successful decryption and translation attempts of/from things that aren't actually related to the Voynich manuscript and... basically look at those and see what are the elements here that led to this being a success... or... not a total success. And basically pull out from those what is important, what is necessary for you to have a successful decryption or translation attempt of anything. And then basically apply those principles to a couple, you know... In this particular situation, the ones I looked at so far are more recent attempts, but basically, where did those recent attempts go wrong. Now of course the recent [un]successful attempts are gonna be biased and saying "oh no this isn't wrong, this is right", but... just looking at... they don't have these certain things.
- Koen: Right. So you've told me that you've had a look at the successful decryptions of Linear B and the CIA Kryptos statue and so, what did you learn from this, what did you take from this?
- Adam: So... what I took from that.... I just kinda went through the story of the decryption of (well, it was mostly decrypted) Kryptos and the entire decipherment of linear B and basically what I learned was: there are three, and potentially more... this hasn't been exhaustive yet but basically there are three (let me see my notes) that are... at least I determined to be necessary for a successful decryption. This is based on both those and I'm sure if I'll continue researching I'll find other elements as well.
The first one is: you need a large corpus size. For Linear B there was... one of the reasons why it took a long time... because for the longest time they didn't have any idea if what was going on was a dead end... because...
- Koen: They didn't have enough material or?
- Adam: Yes, pretty much, that was it. They didn't have a lot of material because there was only, you know, so many different tablets and pieces of pottery that were being discovered. There just wasn't a lot of texts to analyze and you can't... well you can... run a statistical analysis on it but it's not gonna be [what you] expect unless you have more data. And it was a similar thing for Kryptos; the sculpture is divided into four different sections, but each section is short enough that a statistical analysis can't give you much out of it. So, one of the ways that Linear B ended up being deciphered was just you know, over time, as they discovered more text to add to the corpus.
The corpus size is important I think for two different reasons: the first one is, like I said, the bigger it is the better statistical data you're going to get out of it. And the second one is that sometimes you can find something in a different section of the corpus (and I'll give an example) that kind of, is the key, Linear B being a good example of this. It's not necessarily that they didn't have quantitative Linear but basically there ended up being one key tablet that they discovered, kinda nicknamed the tripod tablet, because it had certain key features in it, namely an ideogram that looks like a tripod next to the, I guess, syllabic script for tripod. It's like a real quick overview of that. Basically they added on to the corpus and this [part ?] that they added happened to be the key to solve it.
- Koen: So basically the proverbial Rosetta Stone.
- Adam: Yeah pretty much. so (I guess) applying this to the Voynich manuscript is one of the reasons why that is (I guess) a problem in a lot of attempts is that, at the end of the day we still only have the one manuscript. If we were to discover, you know, anything else written in the Voynichese script, you know, that A. would give us more data, statistically, and we'd be able to manipulate that in all sorts of different ways depending on how we analyze it, and B. maybe there's something in there that's just linguistically or... whatever ends up being the key. But we'd have that, so...
- Koen: Because, if you look at it purely from a quantitative perspective, the Voynich is pretty large for an enciphered document... but what you're saying is that we don't have probably enough variation in the material so our chances of having the key to it is pretty low, still?
- Adam: I think that is definitely part of it, yeah. And when doing any sort of statistical analysis, like is often done for decryption, it's never a bad thing to have more data.
- Koen: Right. And is there anything else that these unsuccessful attempts share?
- Adam: Yes. The second point that I determined was important is a form of external reference: this is sort of related but not entirely. So by external reference I mean something that basically allows us to say with certainty: yes this is correct. To go through the examples with Linear B, as I said the tripod tablet again, the way specifically that it worked was basically even though prior to the discovery of that tablet the translators had a fairly robust well... they had an idea of basically what it was. They said maybe this is a syllabic script of ancient Greek which it would turn out to be. But their findings were based mostly on: if we look at the data this way and arrange it in this way then this seems to kind of makes sense, but... it was at that time still, basically, "this could be a coincidence" kind of thing. And then when the new tablet was discovered, the similarities were basically either an astronomical coincidence or... proof, basically. So that is the external reference to that that basically... the discovered of that confirmed that theory. And, with Kryptos the idea is... the cryptanalysts looked at the code and they deciphered it and they... basically [?] the creator of the sculpture is actually still alive so...
- Koen: So we know that we could know the solution if we asked.
- Adam: Yeah basically. So for the first three parts of Kryptos that were uncovered... they were confirmed by the sculptor: yes this is right, yes this is right, yes this is right, and...
- Koen: That's the best kind of external reference you can hope for, so I'm telling you that you're right.
- Adam: And interestingly one of the three sections actually had a typo in the popular decryption plaintext of it, which was due to just a typo made on the sculpture for aesthetic reasons. The sculptor deleted a character, which messed up the encoding. But that only affected like the last 8 characters or so which coincidentally made it say one phrase instead of another phrase. Specifically it said the phrase "ID BY ROWS" when it should have said "LAYER TWO". And once the sculptor realized that was actually part of what people were saying the plaintext was, he was like "oh oh that was not what it was. In fact that was due to a typo, the actual phrase here should be this." [?] Yes the authoritative of the author is I guess the number one reference we could ever get to prove that it's right and Voynich manuscripts having been written like five hundred years ago, if the radiocarbon date is to be believed -- and I believe the radiocarbon dating -- then the author or authors are dead so... and I suppose the best alternative we could have to that would be, once again, similarly, if we were to discover more text written in the Voynichese script... that would perhaps give us some sort of... confirmation of one of the theories flying about and... it runs into a similar problem. We only have the one manuscript and there's only so much you can pull out of a single source.
- Koen: Right. I think that's a good point. Now the problem you see with so many theories is that it's always based on internal reference inside the manuscript so you say like oh I think this plant is a sunflower so this word must be this but it's just an interpretation sitting on top of another and you're not really getting anywhere.
- Adam: Yeah, and I think that really applies to any good argument in general. You can theorize and theorize all you want but, unless you have some sort of proof to back it up, it's always going to remain, probably, a good coincidence.
- Koen: Right. Right. Is there anything else that successful decryption attempts have in common?
- Adam: Yes. The third elements that I was able to pull out. This is less about the actual properties of a text and word about methodology. And I'm calling this kind of picking a good hypothesis and running with it, based on the foundation of the data, where the idea is basically: if you're going to be working on a proposed solution it always needs to be founded on data. That's just first and foremost. This was in Linear B, this was in Kryptos. Statistical analysis and other data was always: this is a fact of the text and it has to be based off this fact and I think something that is present in some Voynich attempts are... kind of informed by "here is my idea first" and let me see if that form fits what I find to that, which is not the right approach. So found it on the data and then the idea would be to, given what kind of data we have here, where can we go from that. Like, if we assume that this might be the case, what does that tell us, and then does the data support that too. And then the idea being essentially... let me check the phrase... the fewer jumps in reasoning you have to make throughout the process of working on an hypothesis then the stronger it is. And... just something I kind of saw in some Voynich attempts was: a lot of jumps in logic. One specific case that I did look a little deeper into was the one in September of this past year 2017 the Nicholas Gibbs attempt that was written about in the Times Literary Supplement. I could rip into that all day but the idea basically being: there is no basis for thinking "oh, it's Latin".
- Koen: For people who may have forgotten what was it about, in short? He just thought it was abbreviated Latin I think?
- Adam: Yeah, he thought it was abbreviated Latin, that each line was like the equivalent of an abbreviated Latin sentence and each word of Voynichese was actually like...
- Koen: A sentence.
- Adam: A sentence, but each letter represented a certain Latin word, but it just...
- Koen: He had so many possibilities per letter that, per word, you could have like a thousand different translations and it became impossible...
- Adam: Yes and I think the biggest jump in logic he had there was that the foundation of the argument rested on not the data but rather: there is an index. And the index happens to be the [?] that are missing, which there's really no foundation to think that that's the case. And I think that's kind of the core of why the argument doesn't work.
- Koen: But wouldn't just say that if you start from a decent working hypothesis that's completely hypothetical that still you could accumulate so much in support of it that your attempt could be called successful to some extent?
- Adam: I definitely think that that is possible to a degree. If it seems to work and work and work and be internally consistent then I think so, but then we kind of run into the problem of, once again, needing an external reference to essentially verify. So for all we know, maybe, one of the existing decryption attempts has been successful and was very rigorous in its methodology, but we...
- Koen: We can't call it.
- Adam: Yeah.
- Koen: Someone also asked like, you looked at Linear B and the CIA Kryptos statue, and those are things that are completely different from each other and from the Voynich manuscript. Do you think that you may have better chosen examples closer to the 15th century?
- Adam: I think there is definitely room to look at more contemporary examples, because cryptography techniques in the modern day are simply just different from what would have been done before and I think about, to continue researching, I definitely would want to look at more contemporary to the Voynich manuscript sources. So far in my research... I just didn't have the time to do that yet to be totally honest.
- Koen: You are less than half the age of the other people we interviewed so you still have lots of time to catch up with more Voynich research.
- Adam: Yeah and I think definitely that would be a great opportunity to continue to look into... because one piece of information I read when I was researching was that a lot of cryptographic techniques from the 15th century would be very easily sample by computer today, which is worth keeping in mind for sure. Part of the reasons why I chose Linear B and Kryptos specifically out of some of the ones I was looking at are that I wanted to deeper focus was I think that those two are good examples of August the two different approaches to how people view the Voynich manuscript as a text. One is as a translation and the other is as a decryption...
- Koen: Two extremes, in a way.
- Adam: Yeah and I guess my mindset was if we kind of isolate, coming from this approach, what do we learn, and coming from the other approach, what do we learn, and, can we find any similarities in that, since the Voynich manuscript is looked at often in both these kinds of ways.
- Koen: Okay, that makes sense. So you've been studying the Voynich manuscript for about a year nowmaybe...
- Adam: Yeah, I started back in the fall.
- Koen: So, you must have read a lot of opinions and theories about it. Is there one theory that's your favourite at the moment?
- Adam: Yeah, I wouldn't quite call it my favorite...
- Koen: Your least unfavorite?
- Adam: So one of the theories I read was the hoax hypothesis that the Voynich manuscript was created through the use of a text generation algorithm. The particular paper I read went into detail about the proposed forgers making a series of Cardan grilles.
- Koen: Is that Gordon Rugg's theory?
- Adam: Yes it was Gordon Rugg's theory I had the paper up in one of my [?]. It was that one with the Cardan grilles. It was, at least if I was understanding the paper correctly, the theory that Rugg proposes accounts for all of the interesting linguistic properties of the Voynich text, which I think is something that can't be undervalued here. Because the Voynich manuscript has some unique linguistic properties, and, you know, if this account for those then it already has a pretty strong standing in the data. And the, I guess, the less scientific [?] part of his argument being: it might have been created as a financial gain opportunity by hoaxers, but only if the could make it in a quick enough amount of time that it would be worth their time to do. And than him saying and... the researchers were able to use a method that would have been available in the 15th century to create it, to recreate the whole thing in less than a couple of days. That was what the point that really sold me. For me the thing about why recreating the statistical properties is such, I guess, why that of itself is convincing to me at least is because, you know once again, the Voynichese text has such unique properties, very unlike natural languages and, you know, I would be curious to see if it wasn't generated in that way, and it's just some sort of natural language then what would be the explanation to why it has these particular properties when many other languages don't. I'd just want to know what, what about this language, and if that can explain that, and fit in with other things we already know about the text then... I just think it's one of, at least, the more convincing ones that I've read.
- Koen: Okay, Adam, it was very nice talking to you, and thank you for this interview.