How can explorers determine whether strings of alien text is meaningful or just nonsense?

Question

A group of xenoarchaeologists is studying the remains of a bygone civilization with the aim of understanding its culture. So far, isolated artifacts have been found, such as metal plates with symbols inscribed on them and remnants of stone walls of ruined structures. How would they determine whether the shapes they see are attempts to encode something meaningful, purely artistic, art that also represents text, or a meaningless jumble of symbols?

For an introductory problem, here are three strings of English text encoded into a very fancy font.

If these are found engraved onto the wall of an ancient ruin, how would a team of analysts go about firstly determining if any of these strings of symbols are meaningful, and possibly determining what they could mean?

(For anyone wondering, the first line is "Your copper is shit", then "Purple herring canister omega", then a random jumble of letters using the Marker text from the Dead Space games)

As a more complicated example, here is the logo of Al-Jazeera, which I am told is made from arranging the Arabic text into an artistic pattern

Suppose that something like it was encountered among the remains of an alien civilization. How would the exploration team know that there was anything more to it besides looking pretty?

And, finally, if an alien language uses an encoding system unlike anything used by human languages, is there any way to tell if there is any meaning behind the strange symbol(s) or if it's gibberish? As an example, suppose the language's encoding system is somewhat similar to the sigil creation method of New Age enthusiasts, so "EVACUATE IMMEDIATELY" becomes something like this. How would one determine that this is actually an attempt to convey some idea, rather than just a squiggle?

Answer 1

1. If the set of symbols is finite and the markings are found on all sorts of artefacts, then it is pretty safe to conclude that the symbols carry meaning.

2. Carrying a meaning can be done in two ways, namely by representing language and by representing ideas directly.

3. Symbols which carry a meaning by representing language are called writing.

   • If the set of symbols is small, say, from about a dozen different shapes to about forty or fifty different shapes, then the writing is alphabetic.

   • If the set of symbols is of medium size, say, from about fifty to about two hundred different shapes, the writing is syllabic.

   • If the set of symbols is large, from about five hundred to a few thousand different shapes, the writing is logographic.

   Notes:

   • No writing system is pure. For example,

     • In ordinary English writing we use mostly alphabetic writing, but we also use a small number of logographic characters, such as the ten digits 0, 1, ..., 9, and a handful of miscellaneous symbols +, −, °, %, &, and so on.

     • Japanese writing is a mixture of logographic symbols called kanji and syllabic characters called kana.

     • Historically, the Hittite language was written mostly with syllabic characters but mixed with a large number of logographic characters (called sumerograms because they were inherited from Sumerian writing). (Which produces vexing problems, because some words were always written with the logographic symbols so that we don't know how they sounded.)

   • A literate culture will leave lots and lots of text. Literate people have an inescapable tendency to scribble on walls, and a literate culture will automatically place writing on the most diverese objects, such as ceramic containers, bricks, coins, or construction timber. (For an amusing example, we have ancient Greek and Roman scribblings on Egyptian monuments, sign that inconsiderate tourists are a very old phenomenon.)

4. In real life, we use both writing, that is, carrying meaning by representing language, and symbols which represent ideas directly, without representing language. For example,

   • Traffic signs are meaningful, but they do not represent language.

   

   These signs have very definite meanings (and all vehicle drivers must know and obey them), but they do not represent language. They are not writing.

   • Emojis are commonly used little pictures, especially by younger people, which almost carry meaning without representing language.

5. Close study of the strings of symbols is required to distinguish between writing and carrying meaning without representing language. Sometimes it is hard to decide; for example, scholars have not yet reached a definitive conclusion on whether the famous Indus valley symbols are or are not writing.

   Heuristically,

   • If the symbols represent writing, we will find relatively long strings of them:

     

     MANIOS:MED:FHE⁞FHAKED:NUMASIOI

     The text on the 2,700 years old Praeneste fibula, the oldest Latin inscription ever found. (In classical Latin that would be Manius me fecit Numerio -- we see the archaic nominative ending -os instead of classical -us, the archaic accusative in -ed instead of classical -e, very archaic reduplicated perfect of facio, archaic dative ending in -oi, archaic -a- in an internal open syllable which became -e- in the the classical language, and an intervocalic -s- which became -r- in the classical language.) (Picture from Wikimedia by user José M. Ciordia.) Note that the symbol <M> appears three times, the symbol <I> appears three times etc.

   • If the symbols represent writing, we expect to find syntactic patterns. Symbols will appear multiple times in the same inscription, strings of symbols will be repeated, some strings of symbols will be found much more often than others, some strings of symbols will be found in many inscriptions.

     For example, in English writing the strings of symbols <the> and <ing> appear very frequently, and it is a rare inscription where none of them appears. In Chinese writing, the strings of symbols 他們 (tāmen, meaning they), 因为 (yīnwèi, because) and 什么 (shénme, what or anything) are frequent.

   • If the symbols represent writing, we expect their distribution to follow a power law. (The specifically linguistic incarnation of a power law is Zipf's law.)

     For example, in English writing the symbol <e> is 175 times as frequent as the symbol <z>.

For a real-life example of reasoning about unkown symbols, see Steve Farmer, Richard Sproat, and Michael Witzel, "The Collapse of the Indus-Script Thesis: The Myth of a Literate Harappan Civilization", in Electronic Journal of Vedic Studies (EJVS) 11 - 2 (2004), pp. 19-57, ISSN 1084-7561. (Their conclusion is that the Indus valley symbols are not writing; but the what's important is the reasoning process, which is explained in detail.)

Answer 2

With short texts, it's really anybody's guess. Although obviously some scripts are more easily confused with cracks in a rock than others, natural patterns can be notoriously complex, if you're only working off a tiny sample. If the sample is short enough, even completely random noise might by chance appear to be meaningful.

With longer texts, you'll probably want to turn to Zipf's law, which describes the expected relationship between how frequently each word is supposed to appear. (This relationship is roughly: the most frequent word appears twice as often as the second most frequent, three times as often as the third and so on.)

Zipf's law is holding surprisingly well for a large range of natural languages, although we don't quite know why exactly, and is therefore less of a law, more like a very good rule of thumb.

But you need a lot more text than just an inscription to apply it and avoid the Runamo "rune" embarrassment.

Answer 3

I would take a look at how they concluded that the Voynich manuscript is not just nonsense symbols stringed together made to look like alien text but it is very probably some unknown language written with unknown symbols.

They have done eg. statistical analysis that shows that the patterns in words and text follow similar rules as texts written in known languages. If it was just nonsense, such patterns would not be found, or they would be sparse and randomly distributed.

Answer 4

Some of the first hints are probably recurrence and order.

Since it is very unlikely that a language is encoded with an infinite set of symbols, sooner or later there will be some repetition.

And together with the repetition some sort of order will stand out, making the thing stand out from a random noise background.

From there to have a meaning associated to the symbols, some context will be needed. Either a translation, a la Rosetta stone, or some other context. E.g. a set of ordered scribbles found in a pottery kiln will likely mean something about the kiln, or in a temple will likely mean something about the worshiping which took place there.

Answer 5

Let's flip this around and imagine aliens are studying human artifacts. How would they prove there is information in them?

A prelude is to prove that the artifacts were created artificially and deliberately. A lot of modern human artifacts obviously use unnatural materials, colors, shapes so this is not hard to show. You could assume that etching a bunch of lines on a metal plate is a lot of work, so it must surely have some purpose to justify that effort. This is the "backdoor" way to show it. Of course, logically, just because someone deliberately drew some lines doesn't necessarily mean they mean something - it's just that it usually does.

To prove information content, you would generally need to:

1. Find patterns that are consistent throughout the known corpus
2. Provide a description of the patterns that is significantly shorter than the text itself (so no 10 page grammar to "explain" a 2 page letter)
3. Show that the patterns are complex enough that they're unlikely to have come from a natural process

With text intended for human consumption, this is easy to do, because human language has very obvious patterns. It was also selected to be intuitive, so that human babies can easily learn it without instruction. This takes care of 1. Also, human brains are quite limited in computational power, and humans are content to speak/read/write at a limited pace, which makes 2 easy. However, observe that there is a minimal amount of sample text you must collect, otherwise 2 and 3 are impossible and 1 is difficult. For English, grammar can probably be inferred from even a single book, but the bigger problem would be the very diverse vocabulary.

If the aliens tried to examine our network transmissions, hard drives, CDs, DVDs, floppies and so forth, they would probably be very confused however. That stuff is all compressed with efficient (max-entropy) schemes, a lot of it looks flat out like random data as a result, and what isn't random has few recognizable patterns. Imagine the task of zero-knowledge reverse engineering something like 7zip compared to reverse engineering English - a completely different ball game. On top of that, much of the information is also encrypted. In fact, the aliens must solve 4 problems at once:

• Guessing the key
• Guessing the encryption scheme
• Guessing the data encoding
• Guessing the purpose of the encoded data

Each one of these on its own is a challenge, and it's basically made possible by making assumptions about the next level down. Solving all at once seems unimaginable. Imagine you're trying to crack an encrypted file - but you don't even know what encryption scheme to plug your putative passwords in, and even if you guessed those right, you get back a JPEG which you don't know how to interpret. So the "correct" output looks just like the gibberish you get from a wrong password. And even if you did figure out how to decompress JPEG, you still don't know if the contents are intended to be text, or sound, or image, or video or a Sid Meier's Alpha Centauri save game.

At the other end of the spectrum, sometimes we create texts that are specifically intended to be fool proof, and readable without sufficient understanding of even our simple natural language. For example, https://en.wikipedia.org/wiki/Long-term_nuclear_waste_warning_messages describes attempts to create a message about nuclear waste, that would be comprehensible even to people who are not able or willing to go to the trouble of trying to decipher an unknown text. A more mundane example is various signage for multi-lingual audiences (airports) and young children. These should be the easiest ones to figure out, although the problem is that they tend to have low information content (since nobody wants to write complicated things with symbols) so it's hard to gather enough unique corpus to justify any claims about meaning. For example, how would you prove that the appearance of ⛔ is meaningful, and not just an artifact of the manufacturing process?

So, generally speaking, primitive texts will be easy to figure out because presumably they are intended for biological brains that are hard to upgrade or reprogram. For a technologically advanced civilization, there will be more sophisticated encodings intended for machines, and they will be hard to distinguish from gibberish. The more they use computers, the harder it will be to show that their records contain information. To some extent, this might also be the case if they are simply much more intelligent than humans.

---

To address some specific examples you give:

• The Al-Jazeera logo follows an ancient tradition in Arabic calligraphy. Arabic letters look strange to Westerners even if arranged normally, but if you get past that, the logo is basically a monogram just like the logo used by JRR Tolkien, for example. It is not intended to be "read" but "recognized", and it is rare to write original information in this form (it would be very hard to read). It is more common to use it for texts that are well known, such as verses from the Quran or the names and titles of a Sultan. Therefore the corpus for such calligraphy will be limited - but if it was used more commonly, then analyzing the topology of letters would reveal patterns.
• The made up font is so trivial that you can even tell at a glance it's supposed to be English text. It has overcomplicated letters that are hard to distinguish, but any real analyst would start by labeling each symbol with a number or letter and go from there. It's harder if you have organic variations of the same symbol (handwriting), but the different forms will still be more similar to each other than to other letters. Even if you have alternate forms like Greek σ/ς, they should still be fairly straightforward - otherwise the alphabet itself would be unusable.
• For your triangular sigil, it is not possible to prove that it is meaningful because your corpus is too small. You would need many more sigils like that, so you could look for patterns like the angle of the lines, their curvature, when they cross each other or not. But with just one sigil, I could invent any number of just-so "meanings" for it, all equally correct and plausible. I could claim that the "real information" is that the sigil is black and white as opposed to other colors, and the lines are just decorative. You can't do anything without more samples.

Answer 6

A key point to determine if there is meaning to a marking is where that marking is placed and any relationship to other markings.

For example, markings placed on a rock near a small village but overlaying other markings could just be the doodling of bored teenagers. (If you are up on a high place watching for deer, and it has been three days of nothing, wouldn't you be attacking the nearby rocks with something?) Markings on a door might have meanings. Markings on a piece of clay or metal attached to or near a container might have meaning such as contents or ownership of the container. Markings on buildings in a semi regular manner might indicate something about those buildings. Carvings on a pillar in the middle of town or at the entrance to town on a road might have significance. These might also incorporate artistic expression as Mayan writing allows for significant artistic license.

The problems with trying to discern art is that nearly anything can be called "art", many cultures do not have a concept of "high art" but incorporate artistic expression into their everyday lives, and archaeologists perceive stuff through their own ideas of what art could be. There are multiple cases where the cleaning staff threw away an "art installation" thinking it was trash.

Answer 7

If nothing is known about the beings who have built the walls, refined the metal for the plates and so on, then one has to wonder what "meaningful" really means.

For example, the plates might turn out to be excreted in the same way that bees excrete (?) small plates or scales of wax before forming it into a comb. In this case, one has to wonder whether the existence of the plates implies that their creators had sufficient cognitive ability to conceive of symbolic communication: if they did not, then any statistical tests that suggest that the symbols are meaningful are irrelevant.

If on the other hand the creators did have a reasonable degree of intelligence, and if it can be demonstrated that making the symbols took even a minimal degree of effort, then they must be meaningful e.g. as decoration or a clan mark even if they do not represent a written language: something that might be unknown to even an intelligent race.

Answer 8

Two aspects that haven't been touched upon yet:

1. If the alien language is close enough to human languages to give us any chance of decrypting it, there is no way that they could encode more than one sentence into something like the Al Jazeera calligraphy or the New Age squiggle.
2. Statistical analysis is a good way to decide whether something is a language - provided it is not encrypted or compressed. Who knows, maybe the aliens used a combination of inscriptions in stone with AR to convey locally relevant information? Encryption and compression will make any meaningful communication look like random noise, the more the better they are.