In 1912, a Polish antique  collector named Wilfrid Voynich found a long-forgotten  chest in an Italian library.

Inside was a well-preserved manuscript,  seemingly from the Middle Ages, with about 240 pages filled with illustrations of plants, stars, people, and animals. But what really set the book apart was its text, which appeared to be written in a totally alien, never-before-seen language.

Voynich spent the rest of his life trying  and failing to unlock the book’s secrets. And to this day, no one has ever  been able to decipher its meaning. But now, a century after its rediscovery, a group of scientists might have finally uncovered a convincing explanation for the truth – that it was an elaborate, medieval hoax. [SciShow Intro] Although it’s usually called  the Voynich Manuscript, we can now trace the book’s  ownership back several centuries, possibly from the court of an Austrian Emperor to its current home at Yale University.

And thanks to carbon dating, we know it was written in the  1400s, likely in central Europe. But we have no idea what the book says. Despite the letters kinda looking  like Latin ones, Voynichese, as it’s known, doesn’t match any  known language, European or otherwise.

So, is it a language? After scientists took interest  in deciphering the manuscript, ey started using fancy math and linguistics  techniques to better understand it. And it turns out that the book seems to  follow a well-known law of linguistics called Zipf’s law.

The rule says that in any language, the nth most frequent word appears  about one over n times less frequently than the first most frequent word. So if the most commonly used  word in English is “the”, then it will be used about a hundred times as often as the hundredth  most common word, “find”, and a thousand times as often as the  one-thousandth most common word, “entry”. And now you know the  one-thousandth most common word in the English language.

Following Zipf’s law is a pretty big hint that whatever ‘words’ the Voynichese language uses do behave like an actual language. And yet, there’s also solid reason to believe that the manuscript is really  a convincing-looking hoax. But to explain that, we need to understand a bit  about information theory.

And since all science needs funding, we need to take a quick break first. Thanks to Brilliant for supporting  this episode of SciShow! Brilliant is an online learning platform where you can develop your  skills and deepen your knowledge.

They’ve got courses on math, science,  data analysis, programming, and more, so you’ll never get bored or  run out of topics to explore. Whether you’re just starting  or you’re already a pro, Brilliant will have a class  that’s perfectly designed to build on your skills while  still offering you a fun challenge. Their courses are designed to let you  play and solve puzzles yourself, k eeping you engaged and  entertained while you learn.

To try Brilliant for free, visit brilliant.org/scishow,  scan the QR code onscreen, or click on the link in the video description. You’ll also get 20% off an  annual premium subscription. The key idea behind the  science of information theory is that some messages have more  ‘information’ in them than others.

Like, compare the phrases  “c” and “I like to swim”. Both have the same number of characters, but clearly one is more  information-rich than the other. Information theory lets us  quantify that difference using something called the conditional entropy.

That measures how disordered  or uncertain the text is. For instance, in English, the letter  “q” is basically always followed by “u”. So if you see a q, you can  reliably predict the next letter – it’s not surprising to see the “u”.

So the entropy of the “u” in “qu” is much lower than the entropy of any other two-letter pair. You can work out entropy values  for every letter in English, or two-letter pair, or  three-letter triple, and so on, and you can do the same for Voynichese. And when you do so, what you see is…kinda weird.

The entropy of pairs of letters in Voynichese is way, way lower than in any known language. Like, you can clearly see it in this graph. Spot the odd one out.

This is kind of a big hint that  Voynichese isn’t a natural language: it acts like it isn’t conveying information. But even more striking evidence  that it isn’t a real language comes from a 2019 paper. In that work, researchers plotted how similar each pair of Voynichese words is, as a function of how far  apart they are in the text.

They found that the closer in the  text two words are to each other, the more similar they are. This really isn’t how a normal book works. For example, in A Game of Thrones, if the word pair “Jon Snow” appears on one page, you can expect to see it again nearby.

But in the Voynich manuscript is as if the words “Jan Snow”, “Don Snow”, and “Jon Snot” are showing up consistently near “Jon Snow”. And this is happening for, like, every word. It’s hard to imagine how this  could happen by accident.

It’s as contrived as the 20th  century American novel “Gadsby”, where the author deliberately  avoided using the letter e. It’s even evidence against  the idea that the manuscript is written in an encoded language, since encryption tends to destroy correlations, but here we find more correlations than usual. So maybe the text is all just… gibberish.

Except, it still follows Zipf’s law. And, like, look at these pages! They clearly look like actual written words!

It’s as if it was written to fool you  into thinking it was a real language, or at least an encryption of one. So the 2019 paper, building on previous works, described how a medieval writer  could have pulled this off, using a basic algorithm. And no, we’re not talking about some advanced AI giving you video recommendations.

This algorithm was a set of  instructions simple enough for a medieval scribe to follow while  writing on parchment by candlelight. The idea was that the writer would start a page by copying out a line from a previous page, with a few letter changes sprinkled in. Then, line by line, they’d fill the  rest of the page with more words by slightly modifying the words already there, or randomly deciding to split or join  up words, or add prefixes, and so on.

So if the line they took from a previous page was “jon snow went to the wall.”, they might write the next line  as “want into jansnow well she” by editing, jumbling, and combining  the previous sentence’s words. Kind of like a medieval ChatGPT. The researchers claimed to  have followed these steps to generate text that follows all the  statistical patterns of Voynichese, from Zipf’s law to the “Jan Snow” effect.

Now, the researchers weren’t saying they’d found the algorithm  that made the manuscript, just that they’d found a plausible  method that the writer could have used. And of course the author didn’t need  to follow any algorithm exactly. They could’ve mixed things up a  bit whenever they felt like it.

But we still don’t know why someone would do this. I mean, this is a lot of  effort for a simple prank. Here, the pictures might give some hints.

They’re organized into different sections. Like, the plant pictures seem  to mostly occur near each other, as if it’s a book with different chapters. So maybe the book was meant to  look like it was about alchemy, the mysterious forerunner to chemistry that was very much in vogue at the time.

After all, alchemists were infamous for  being super secretive about their work. I mean, if you can turn lead  into gold, you don’t just say so. You make other people work for it.

Alchemists often used code words and  highly metaphorical illustrations to simultaneously convey  and conceal their secrets. So maybe the book was made to  trick or fool other alchemists into thinking it had hidden secrets. That could explain why some of the pictures are totally made-up plants and animals.

Maybe the writer was a medieval  Fyre Festival organizer, who wanted to sell the secrets  of the Philosopher’s Stone to a gullible rich sucker. Or maybe it was just an artistic  exercise by a group of bored monks. I mean, it’s not like they  had much else to do back then.

We’ll probably never know the full  truth about who wrote the manuscript, let alone why. But as sad as it is that it probably  doesn’t contain lost alchemical secrets, the quest to learn the manuscript’s  real truth is pretty exciting, too. [ OUTRO ]