A spaceship is headed for Earth. We detected it with a few months' warning. Messages have been sent to and received from the ship by radio, but we haven't deciphered anything they've said, and don't believe they've deciphered anything we've said. We know, from the radio messages, that their language is human-pronounceable.

The government assembles a team of linguists to prepare for when they land. What type of plan are the linguists likely to come up with for establishing initial communication?

  • 9
    $\begingroup$ The type of plan that the linguists in your story are likely to come up in your story is the type of plan which makes sense for your story. (And if we established communications and yet after a few months we still haven't understood a single word it means that they are either incredibly dumb or positively hostile. Let's see if you understand some Ancient Greek words. Beep, enah. Beep-beep, dyo. Beep-beep, beep, triah. Beep-beep, beep-beep, tesserah. Beep-beep, beep-beep, beep, penteh. What does the word tesserah mean in Ancient Greek?) $\endgroup$
    – AlexP
    Aug 14, 2023 at 0:06
  • 5
    $\begingroup$ @John99 do the people on the ship have to learn how to speak their language, or are they born knowing how to speak it? $\endgroup$
    – Monty Wild
    Aug 14, 2023 at 1:44
  • 3
    $\begingroup$ Moreover, does the alien language not drift at all, over their recorded history? They should have at least some concept that language isn't innate and universal, esp. among aliens. $\endgroup$
    – BMF
    Aug 14, 2023 at 2:24
  • 6
    $\begingroup$ "We know, from the radio messages, that their language is human-pronounceable." -- how? Radio, both analog and digital, doesn't follow some kind of natural law that is trivial to deduct. Here is a quick glimpse into how tricky it gets - and that's from a known-human source: witestlab.poly.edu/blog/capture-and-decode-fm-radio $\endgroup$
    – Tom
    Aug 15, 2023 at 15:55
  • 7
    $\begingroup$ Yeah, if we know that their language is human-pronounceable from their radio transmissions, it's almost certainly because they've worked out our transmission protocols and are deliberately transmitting in the clear to let us know that. $\endgroup$
    – notovny
    Aug 15, 2023 at 22:01

4 Answers 4


Sadly, you're just a couple of months too early to read an upcoming volume of scientific papers specifically addressing the issue of xenolinguistics and theoretical issues with fieldwork on alien languages....

In the meantime, a lot depends on what the aliens look like. The fact that their language is pronounceable by humans is an amazingly unlikely and super convenient coincidence which means that all of the tools of monolingual fieldwork can be brought into play. The movie Arrival does a good job of illustrating the basics, in a much more challenging situation than what you have described. However, the more human-like the aliens are, the smoother it will go, as the linguists involved will be able to rely more heavily on things like assuming that the aliens will have analogous words for analogous body parts and understand pantomime in similar ways as us. If their bodies are radically different, there will have to be a longer stage of figuring out how to understand their body language, and teaching them to understand ours. Note that this is still an issue for human-targeted fieldwork anyway, as gestures are not universal, but it will be a bigger issue when you're not at least starting with bodies of the same shape.

It can be assumed that the aliens can hear at least the same bands of frequencies in which they voices exist, but not outside that. Similarly, no assumptions can be made about their other sensory capacities. Maybe they see a different spectral range from us. Maybe they don't see color. Maybe they don't see at all. Maybe they pay more attention to scent than we do. To cover all of those options in preparation for elicitation work, teams of researchers would be compiling a wide variety of different types of stimulus materials to determine what the aliens can distinguish and what they care about as groundwork for more targeted elicitation later.

Absolutely everything that the aliens say should be recorded, in the highest possible fidelity, so as to capture any distinctions that may not be obvious to not-yet-trained ears. This will be accompanied by video recordings to provide context for each utterance. For everyone interacting directly with the aliens, there will be five or ten who just do data analysis on the recordings that come out of elicitation sessions. While none of these are done in isolation, and elicitation experts will figure stuff out about multiple organizational levels at once, the first issue for analysts will be identifying contrasting phonemes--which might be possible ahead of time based solely on the corpus of transmissions--then establishing a transcription convention, identifying "words" (roots, collocations, idioms, etc.), and then finally building up successively more complex levels of grammar.

And while nobody will strongly expect anything to come of it, someone is going to try zero-shot learning by producing an embedding vector space on the tokenized corpus of recorded alien speech and trying to correlate it with equal-dimensional word-vector spaces for major human languages. It's relatively cheap, and hey, you might get lucky.

Edit, to explain the last paragraph:

Zero-shot learning: learning to classify inputs that belong to categories the learning system has never examples of before, based on correlating knowledge from multiple other sources. I.e., a zero-shot image classifiers might be able to correctly identify pictures of zebras without ever having been trained on zebras because it knows what stripes are, and knows what horses are, and has been told that a zebra looks like a striped horse.

Embedding vector space on a tokenized corpus: this is how LLMs, like ChatGPT, encode their inputs. It's a way of being able to do math on words. Basically, you come up with a method of splitting a collection of texts (a corpus) into discrete tokens (letters, words, or whatever happens to work), and then you compute a list of numbers--a vector--that represents each of those tokens based on the other tokens that it occurs in context with. The position of the resulting vectors in higher-dimensional space often correlates with useful semantic features of the tokens.

Correlating word vector spaces: zero-shot learning for machine translation is done by producing embedding vectors for multiple languages, and then looking for clusters of points that have the same shapes in each model. If you assume that the matching points are translation-equivalents, then that gives you a way to convert a semantic vector from one model into a semantic vector from the other model, and start translating languages without ever having seen a parallel text.

This technology is only proven to work at all when starting with extremely large data sets of relatively closely related languages, but it is being seriously researched to see if can be extended to provide cheap machine translation for less well-documented languages and even to decipher animal communication, like whale songs.

  • 5
    $\begingroup$ could you repeat the last paragraph in english please? $\endgroup$
    – ths
    Aug 14, 2023 at 14:02
  • 18
    $\begingroup$ @AlexP No, it isn't. It is a real technology that is being developed for machine translation of human languages, and is being investigated for deciphering cetacean communication. $\endgroup$ Aug 14, 2023 at 20:52
  • 8
    $\begingroup$ The gist is: give a machine a bunch of data, it makes up it's own categories for it and then starts sorting everything into them. At the end you get a string of numbers (called a "vector") that represent how closely something matches each category. To use the zebra example, say there were four categories, black, white, horse and beans. From that, you might get [0.3, 0.3, 1.0, 0.0] for "a bit of black, a bit of white, all horse, and no beans". Then run it in reverse and out pops something human-readable. This is basically ChatGPT in a nutshell (just with waaay less categories). $\endgroup$
    – Samwise
    Aug 14, 2023 at 22:51
  • 5
    $\begingroup$ @AlexP that is not technobabble, but you need to know linguistic theory to make sense of it. In plain English: someone will try to correlate recorded alien speech to recorded human speech in a very mathematical way to see if there is a close match to any human language. $\endgroup$ Aug 14, 2023 at 22:59
  • 3
    $\begingroup$ It means exactly what it says on the tin. But if you want the long answer, wait a couple months and you can read Claire Bowern's chapter in routledge.com/… $\endgroup$ Aug 15, 2023 at 14:45

Start with METI

Take a look at the Arecibo Message, the Voyager Golden Disk, and other METI (Messaging to Extra-Terrestrial Intelligences) that humanity has already done. These projects have already established a baseline on how to attempt communication with an unknown intelligence that shares no common heritage with humanity. There's too much for me to type out here, but try to understand why each of these messages (and each part of each message) were built in the way that they were.

Understand the basics of Cryptology

Linguistics is important for you the author to understand, but you should do at least a brief study on cryptology as well. This is essentially cryptography of the highest order, messages encoded in a medium that you have no context for. Not a cipher of a shared alphabet, not an ASCII message encrypted in binary data, or a secret message hidden in a plain-text letter. Understand the basics of where to start decrypting a message when the encryption method is unknown, and that will help you understand how we would approach this problem.

Establish a method of communication

As other answers have said, you will have to figure out a communal channel to communicate. Maybe they don't hear in a similar way as us, or use audio vibrations through a gaseous medium to 'talk'. Maybe they don't see similar wavelengths or use written symbols to 'write'. Do they feel vibrations like we do? Could we tap on them to create a 'morse code'? Some answers will be evident from the fact that they have a machine (the spaceship), what it looks like, and how they interact with it. At a minimum, they can perceive the material it's made out of, we could find some similar material and arrange pieces of it in a weird order, and assume that they'll be aware of how many of them there are. Consider what other mediums we could assume that they do perceive, or might perceive. How can we confirm that they understand that medium (pro-tip: prime numbers aren't accidents)

With a Cooperative Partner

Assuming that the aliens desire to communicate as well, have established a medium of communication, and are actively engaged in a back-and-forth effort, the best way to start is to establish a shared ontology for the fundamental properties of our universe. Maths for starters; establish communal symbology for numbers, counting, algebra, and primes. With numbers, you can communicate about physics and the elements, and establish a shared symbology for chemistry.

From there you can further abstract. Represent the Earth's position in the Solar System and our planets, map Earth, and discuss geology; planetary science should hold constant across space to where they came from. Discuss the Galaxy, and try to figure out which Solar System they came from. Represent and discuss DNA (do their lifeforms have a similar structure for genetic information?)

At this point, there are enough common references that you're close to learning a second language without knowing a shared language. How would you help an exchange student from Japan to learn English when you don't know any Japanese?

  • $\begingroup$ Once you have math you can extend that to physics and chemistry as the aliens will be operating with the same reality that we do. As for that exchange student--it wasn't exchange, student, or Japan but I've expressed concepts in chemistry many times to overcome a language barrier as it was faster than the dead-tree dictionary. $\endgroup$ Aug 17, 2023 at 2:10

For starters, if they were smart they would talk to people who have deciphered languages of other humans. People from civilization regularly run into primitive tribes in Africa or South America, etc, who have not previously been in contact with the outside world. They usually manage to establish communication pretty quickly.

I once talked to a missionary who was the first to contact such a tribe. He said that when he walked into their village, he was carrying various equipment they weren't familiar with, canteen and tent and radio and whatever, and so they crowded around and said "what's this? what's that?" So then he went around the village pointing at various things and asking "what's that?" and was able to learn words for many common items.

Of course a lot depends on how similar the aliens are to us. If they eat food and wear shoes and so on, building up a list of words for common objects might be fairly easy. But if their physiology, or even more important, their way of thinking, is very different from us, the challenges quickly mount. Does their language have nouns and verbs and adjectives? Or do they just not think in those terms at all.

I'm reminded of a science fiction story I read once where a human meets aliens. Somehow they are able to talk -- that wasn't explained. But the aliens ask him, "How do you hear?" He replies, "With my ears," and points to his ears, and tries to explain how they work. The aliens are baffled by this response. So he asks them how they hear. And one of them says, "I hear of my home world, of ground and sky."

I thought it was a good scene. Perhaps aliens would be so alien that they would just not think like us, and such basic questions would get totally different answers from what we would expect.

Or not. One could also speculate that logic is inherent in the nature of the universe and if aliens are capable of building technological devices like spaceships, they just MUST think about science and technology in essentially the same way we do.

As we have no examples of intelligent aliens to work with, we just have no empirical knowledge and can only speculate.

  • $\begingroup$ I was about to make a comment along the lines of the second to last paragraph, until I got to the second to last paragraph... Good job. $\endgroup$
    – Jedediah
    Aug 15, 2023 at 3:04
  • $\begingroup$ 1/10001001 is a surprisingly good entry for math assuming the aliens have starships. $\endgroup$
    – Joshua
    Aug 15, 2023 at 19:16
  • $\begingroup$ Is the missionary you talked to Daniel Everett? His book, Don't Sleep, there are Snakes, could be of great interest to OP. $\endgroup$ Aug 16, 2023 at 9:28
  • $\begingroup$ @Lorraine It was decades ago, I'm afraid I don't remember his name or even where I met him. But I'd guess there are dozens, maybe hundreds, of missionaries and anthropologists who could share similar stories. $\endgroup$
    – Jay
    Aug 18, 2023 at 4:44

Start with maths and work from there. 1+1=2 will be the same to an alien, just different symbols. So it gives us a point of reference and a bunch of concepts known to both sides to begin with.

Then look at getting hold of their version of a childrens primer and move forwards.

Unlike earth languages, I don't think the audio methods would be better than the visual. It's unlikely we'd produce similar sounds, and those are just a beginning anyway. Whereas we can communicate in mutually defined symbols.

Chinese script is an example. Old Chinese each character was a word. It doesn't matter what language you applied it to it had the same meaning in any language. So widely different languages can use the same symbols for the same concept.

  • $\begingroup$ Chinese script is useful for writing Chinese. It cannot be used to write a different language unless specifically adapted. For example, Japanese script uses many of the same characters as Chinese, but when writing Japanese those characters (1) are supplemented with home-grown kana, and (2) they have different phonetic and semantic meanings. Only some very few ideographic characters can be used trans-lingually; for example, 93 means than same thing regardless if it is read ninety-three, dreiundneunzig, or quatre-vingt-treize. But the vast majority of words cannot be neatly mapped like this. $\endgroup$
    – AlexP
    Aug 14, 2023 at 15:52
  • $\begingroup$ @AlexP China has hundreds of languages not just one. From at least 9 distinct language groups. $\endgroup$
    – Kilisi
    Aug 14, 2023 at 19:23
  • 1
    $\begingroup$ @AlexP from Wikipedia on Chinese languages "share the same writing system (Hanzi) and are mutually intelligible in written form." This isn't all, your example is right, but enough to prove my premise is sound. $\endgroup$
    – Kilisi
    Aug 14, 2023 at 19:52
  • 4
    $\begingroup$ @AlexP Mainstream Chinese languages are mutually intelligible too. I've been learning mandarin for a few years and at some point I was looking for correspondents; one of my correspondents turned out to speak Cantonese, not Mandarin. So we had text-message conversations where she wrote Cantonese and I wrote Mandarin, and we understood each-other. Of course I could never read a newspaper article meant to be read by Cantonese speakers, but a text-message conversation is fine. $\endgroup$
    – Stef
    Aug 14, 2023 at 21:50
  • 1
    $\begingroup$ Beijing standardized the writing system long ago. When two Chinese speakers run into a dialect issue (most commonly Mandarin/Cantonese) they resort to writing because it will be the same even if the sounds are different. Finger on palm, it doesn't even need a piece of paper. Chinese movies are frequently subtitled in Chinese for this reason, also--it works in both the mainland and Hong Kong. Furthermore, my wife has taught herself to get by in Cantonese from such subtitles--she's a native Mandarin speaker. $\endgroup$ Aug 17, 2023 at 2:18

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .