In a story I'm writing, I have a group of beings who can learn any language by just listening to voice-recordings* of someone speaking that language over-and-over again** (almost like a highly upgraded version of how all of us learned to speak our parents' languages while we were preschoolers). The beings only have to hear the language samples around 3-5 times (less, if it's one of the more skilled beings among their number) before being able to speak those portions of the language themselves.

This question does not ask how such a thing would be possible; but rather, it asks if beings such as these could still have verbal information hidden from them by someone speaking in verbal cyphers or codewords*** while in their presence, similar to how, even though as a preschooler, I could speak American English fluently, my parents could still keep me from understanding the things they said by verbally spelling them instead of saying them outright). I hope that this is the case, as the story I'm writing involves a fair degree of espionage. If, for some reason, these beings' hyperfast-immersion language learning would also make it so that verbal cyphers and codewords wouldn't work on them, then I would also appreciate a handwavium explanation that I could give to make it so that verbal cyphers and codewords would still work on them.

*The language the beings are trying to learn does not have to be a recording, though. They can also listen to/in on speakers of a language they're trying to learn firsthand.

**If these beings listen to a voice-recording of a conversation that uses only 200 of the words in the language they're trying to learn, then the beings will only know those 200 words; thus, these beings have to listen to large speech samples to get a grasp on the language as a whole. That being said, these beings learn how to speak what portions of a language they've been exposed to within 3–5 repetitions of exposure to it (only 1–2, if it's the more skilled beings among their number).

***That is, spoken cyphers and codewords, like when, in a spy movie, two agents walk up to each other and say things like: "The eagle has the delta worm in its umbrella." "Good, then the sun will rise on an icy tower." For simplicity's sake, we can leave translation/decoding of written material with written cyphers and codewords out of this discussion.

  • 1
    $\begingroup$ Robert Sheckley had a short story about a human with these talents whose job was to go to other civilisations, and open trade with them, so they could be conquered by financial means. He was frustrated by a planet where the language, the gestures, and the social codes changed continuously, so he could never buy or sell anything without finding some new bit of syntax or good manners that he had never met before. The people did not know how this worked either, but they had never known anything else. $\endgroup$ Commented Jan 29, 2023 at 9:41
  • 4
    $\begingroup$ Actual immersion learning consists of hearing particular words and phrases over and over again in context. Remove the context and it needs telepathy or other magic to learn what the random sounds mean. Someone with no knowledge of English or other samples of English could listen to an audio-only recording of the St Crispin's Day speech from Henry V as many times as they liked and they still would not be able to deduce the meaning, even if a large sample like that let some prepositions etc be identified as such. $\endgroup$ Commented Jan 29, 2023 at 18:16
  • $\begingroup$ @KerrAvon2055 the advent of machine language learning models suggests otherwise. The beings could have a complete corpus of syntactic information such that they are able to add semantic information on the fly. If I already know “swim” is a verb, it doesn’t take much for me to figure out what it means as soon as someone uses the word near the activity. The fact that they’re seeing the meaning for the first time would be invisible to observers. $\endgroup$
    – SRM
    Commented May 13, 2023 at 19:57

7 Answers 7


A cipher is an algorithm to transform the letters of the clear text into the letters of a cipher text. A moderately simple cipher would shift the first letter of a word by one place, the second letter by two places, and the third letter by three places, then start again.

The clear text 'hello, hello' becomes 'igomq, kfnop' -- the two instances of 'hello' have different encipherings.

A code is a list of substitutions to transform words of the clear text into the words of the code text. Say the book says that 'hello' becomes 'racoon.'

The clear text 'hello, hello' becomes 'racoon, racoon' -- the two instances of 'hello' have the same encoding.

So your learners should learn codes when they get a sufficiently large sample of the text with the code meaning. But they might not get it. Some verbal codes are deliberately jarring. If 'tiger, tiger, tiger' means 'success' just for this one mission, and animal in all other cases, the learners would at most be confused into thinking that 'tiger' means both 'success' and an animal, but more likely they would simply be confused by that one sentence. And if the recognition code is deliberately designed to blend in ('it might be raining today' 'fortunately I have my umbrella') they will learn just the surface learning.

  • $\begingroup$ Gotcha: Get 'em with the old "once we give them an answer that makes some kind of sense, their brains will be satisfied regardless of whether the information is truly correct, and they're more likely to stop searching" routine.👍 $\endgroup$ Commented Jan 30, 2023 at 23:27
  • $\begingroup$ It works for MLLs like ChatGPT. Takes a lot of counterdata to unseat the first answer. $\endgroup$
    – SRM
    Commented May 13, 2023 at 20:35

Short Answer: Context.

Longer answer:

I'm first going to extensively reference this excellent TNG Episode - Darmok for this answer

So, I don't know how familiar you are with Star Trek - assuming you aren't - Go watch this episode, as it deals with your issue.

Assuming you are - Go watch it anyways cause it's an absolute classic.


Spoiler warning ahead!

(You have been warned).

In the Star Trek Universe - there's the concept of the Universal translator a bit of Treknology that allows different species to communicate by doing effectively what your super-fast Language learners do. It works with all species.... except The Tamarians. Although the Translator accurately translates the words to English, the meaning is incomprehensible to the crew of the Enterprise.

"The River Tamak! In Winter!"

What on earth does that mean? Well, in the show, we eventually learns that the Tamarian language works entirely on metaphor and example.

A River in winter is Frozen and therefore silent - So, that phrase means 'Be Silent!'

Okay, that's not too hard.

"Shakha, when the walls fell"

Just from the written word - any ideas? Even from just the spoken word, it would not be readily apparent.

But with enough insight into vocal inflection, you might be able to work out it means frustration and disappointment at a task being failed.

But then we get to the phrase from which the episode takes it's name:

"Darmok and Jalad, At Tanagra"

This phrase requires a knowledge of the history and social context of the Tamarian people - that two heroes who didn't know each other were forced into co-operation against a common foe, through which they formed a friendship.

For a real-world example - think of the English Language, specifically those from England. If I was to yell 'It's Behind you!' - almost all native Brits would reply "Oh No it isn't!" from the ubiquitous Christmas Panto. Or if I were to refer to something as 'Pythonesque' - to understand what I'm talking about requires a knowledge of Monty Python's flying circus.

It is in these nuances of language that cannot be learned from just hearing the language alone that such a hyper-fast language learner would be deceived.

  • 2
    $\begingroup$ I've always wondered what dinner table conversation is like on Tama IV. "Gerthan to her husband, upon homecoming?" "Desshok and Valen, by the water cooler." $\endgroup$
    – jdunlop
    Commented Jan 30, 2023 at 6:46

Your aliens understand more than just the words.

If we talk gibberish, anyone listening will know something is up. But we savvy spies think we will fool the aliens by talking about a sandwich I am going to bring you, whether it has mayo, and whether it has spicy mayo. I might legitimately be bringing you lunch. Or we might be talking about something we are calling a sandwich which is not at all a sandwich.

We think we are pretty clever spies. But aliens who can learn like yours do might be listening to aspects of speech that extend beyond words. The right brain is speaking all the time too and there are inflections, volume, voice modulation and other things that give information to words beyond the words themselves. Even though your aliens know what a sandwich is, they know from other clips they have studied about the non-word cues in a spoken exchange. They know from those other clips that what you are receiving from me is more important than lunch and also potentially dangerous to you.


Your beings need to understand a passage according to how most native speakers would understand it if they listened. Consequently, since most people would take,

"It looks like rain but I brought my umbrella"

to mean what it says and not "They're investigating but I will be able to past muster," that's what they learn.


Use ideograms that change meaning all the time.

The thing with most normal languages is that the meaning of each word can vary by context, but not very much. For example, a 'house' is usually a place, sometimes a lineage, and rarely something else.

But not all languages are "normal". My old mother and her old lady friends have developed a language of their own using only emojis. The trick to it is that the meaning of each character is entirely up to context - what an emoji means right now is not what it means one hour from now. In order to keep up, you need to be up to speed with global and local news, social trends, and specially their inside jokes.

For example, one day they were using the skull emoji (💀 if your browser supports it) to refer to young people. Because apparently this is (or was for a time) what gen z uses to express laughter. The next day I asked if they were still talking about gen z but the skull had changed its meaning to... olives. Because one of them had almost choked on one.

If you learn this hellish language by seeing a sample, you only know how to speak the version from an era that spans a handful of hours at most on a slow day. If you intercept a message in this language, you still need to know what each symbol meant at the time it was sent. I don't know how my mother and her baby boomer friends keep track of it, it just seems telepathic to me.


I think your aliens would have a hard time decoding Hawaiian from a few recordings, especially pre-contact Hawaiian, which had many different dialects that used the same phonemes. Although spelling and pronunciation is consistent and straightforward, many words, phrases, and place names have multiple meanings plus hidden meanings. Many compound words (especially names) can be broken into different sets of smaller words, which can have wildly different meanings when taken together.

Finally, the kahakō (the macron found over some vowels in print, indicating a longer-than-normal pronunciation) seems like it would be difficult to grok without a fairly large sample size; see the differences between mana, māna, manā, and mānā, for example: https://hilo.hawaii.edu/wehe/?q=mana


Have code phrases were they say something relevant and true but with a very specific wording...

For example a soldier could say, "I am running out of ammo." and this could mean there are more than 10 guards.

Imagine we are at war and our communication is being monitored by the enemy.

So we agreed to have certain code phrases, which might appear to the enemy as perfectly normal to be said in a conversation, to mean something entirely different than what the usual sentence conveys.

Similar to having someone touch a glass when playing poker to mean he has a face card when card counting in groups.

  • $\begingroup$ "Have code phrases were they something relevant and true"? Can you edit your post to make that make sense, please? $\endgroup$
    – Joachim
    Commented Jan 31, 2023 at 9:57

You must log in to answer this question.

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