For the purposes of this post, "word" is defined as a unit of language such that

  • Has some meaning on its own, often relating to the real world.
  • Complex utterances can be formed simply by transmitting multiple words.
  • There are few enough words for a speaker of the language to know most of them (i.e. less than about a million for humans).

All reasonably bandwidth-efficient, general-purpose forms of communication, real or imagined, seem to have words in this general sense.

  • Human spoken and written languages, including unique ones like Pirahã.
  • Human sign languages usually have distinct signs that are basically words.
  • As far as I know, all constructed languages like Esperanto, Lojban, Ithkuil, Toki Pona, etc.
  • Fictional languages like Klingon, Quenya, Dothraki.
  • Even the circular language from Arrival has words in the general sense (sentences can be split up into symbols with individual meanings).
  • In Max Harms' Crystal Society, there are AI and alien characters that think differently from humans. However, the AIs communicate through concept-representations that are basically words, and the aliens' language uses ~1000 symbols that can be translated into words.
  • Programming languages have variables, keywords, commands, etc., the last two of which have intrinsic meanings. When we create a language made entirely of syntax and variables, it is always by assigning meaning to certain concepts e.g. λfx.x for the number 0.

However, if we relax the requirement that the form of communication is efficient and general-purpose, I can think of several examples of wordless "languages" whose sentences cannot be easily split up into symbols of any kind.

  • Photos and videos are a general-purpose form of communication, but even compressed photos and videos have huge bandwidth requirements, and are thus inefficient.
  • Bees have dance communication where the direction, distance and quality of food are communicated simultaneously, not serially. But this is not general-purpose communication.
  • Human body language for indicating emotions. But as far as I know, human body language used for general-purpose communication basically becomes sign language.


The idea is to create a version of the "starfish" alien language trope in an advanced alien civilization where the language is actually plausible. Most existing examples just have an exotic medium (body language, music, telepathy), or just handwave the unintelligibility. A strange inherent structure for a well-developed language would be far more interesting, and open up narrative possibilities like a Universal Translator being unable to translate anything an alien says until they are finished talking. So, how would a language whose thoughts cannot be broken down into words work, and how would it feel to be an alien communicating in this manner? If you think it's impossible, why?

  • 3
    $\begingroup$ I'm not sure this is a coherent question. A language that cannot be subdivided into distinct components (words) to represent and communicate specific meaning or intent is not a language. Even the most polysynthetic language is built of distinct concepts that can be translated even if not as word:word but as word:sentence. How would such a language user even know how to say anything? $\endgroup$
    – rek
    Commented Apr 1, 2020 at 6:29
  • 10
    $\begingroup$ Your definition of word is what linguists call a morpheme. $\endgroup$
    – KeizerHarm
    Commented Apr 1, 2020 at 6:32
  • 4
    $\begingroup$ @KeizerHarm: Or a lexeme. $\endgroup$
    – AlexP
    Commented Apr 1, 2020 at 6:36
  • 10
    $\begingroup$ If all you want is for the UT being "unable to translate anything an alien says until they are finished talking", that is much easier. Consider Latin or German, languages which have (or at least accept) word orders very unlike that required by English; for example, you cannot begin translating into English animi imperio, corporis servitio magis utimur (we mostly use the mind for government, and the body for service) until the end, because English wants the verb "we use" to come first, whereas Latin is happy placing it at the end. $\endgroup$
    – AlexP
    Commented Apr 1, 2020 at 7:00
  • 3
    $\begingroup$ @rek I have a few bad ideas. E.g. for some weird evolutionary reason, the aliens have a subconscious form of public-key cryptography and communicate through "ciphertext", being unable to understand "plaintext". $\endgroup$
    – lirtosiast
    Commented Apr 1, 2020 at 7:03

20 Answers 20


Shaka, when the walls fell

English has a few good examples that you could use to build up an even more obfuscated language.


You've just read each of those words, but without some sort of context to lead you on, you can't tell which use of the spelling I've chosen. While the language can be split into words, out of context they're just meaningless strings of syllables (apart from read which at least has a fixed context), and written down it's even worse as you don't even know which pronunciation is required.

Develop the language so that the meaning of each word is dependent on every other word in the sentence and none of them have meaning in isolation. Perhaps instead of words having roots, entire sentences or paragraphs have roots and the rest of the "words" are just a series of prefixes and suffices that modify the root.

Darmok and Jalad at Tanagra

Antidisestablishmentarianism is not just a word, it's a movement, a period in history, a whole swathe of social context. To understand the word, its usage, and meaning, is much like understanding the titles I'm using, without the cultural context on which they depend, they're meaningless.

Picard and Dathon at El-Adrel

You could also vary the meaning by social context. Who is saying it to whom? What are their relative social statuses? Again, written down the way we normally would, you're at a loss as you don't know which of the speakers has the higher social status. And why is all their fiction explicit about the size and style of the hat each character wears?

  • 2
    $\begingroup$ Nice.... Darmok was my first though reading the question. And I scrolled down to see if anyone had already mentioned it. I really like how you demonstrated the concept in the answer itself. $\endgroup$
    – Tonny
    Commented Apr 1, 2020 at 10:30
  • 3
    $\begingroup$ For those that really don't get the references in this answer. "Darmok" is a Star Trek episode that is highly relevant to this question. One of the best Star Trek episodes ever IMHO. $\endgroup$
    – Tonny
    Commented Apr 1, 2020 at 10:41
  • 4
    $\begingroup$ @Matthew,that's half the point! Sentences written in apparently simple English but meaningless without social, or in this case cultural, context. (The cultural context being that they're meaningless without cultural context but that gets a bit meta) $\endgroup$
    – Separatrix
    Commented Apr 1, 2020 at 14:45
  • 4
    $\begingroup$ I don't quite follow the first section. None of those example words are meaningless without context, it's that they have several meanings without a way to choose among them. I can't quite wrap my head around a language that essentially uses exclusively homographs, as any sentence will be extremely difficult to parse - the average human uses ~20k words, but if you need to know each of those words in the context of every other word, that only leaves you with about 150 utterances, each of which have 150 meanings. And I still don't see how those aren't "words". $\endgroup$ Commented Apr 1, 2020 at 15:13
  • 4
    $\begingroup$ Lack of context or changing context doesn't mean the words are inherently without meaning or the meaning can't be communicated. 'Darmok' still has specific meaning to the speaker and, as the episode showed, could eventually be broken down and understood without familiarity with the underlying cultural context. $\endgroup$
    – rek
    Commented Apr 1, 2020 at 16:17

A strange inherent structure for a well-developed language would be far more interesting, and open up narrative possibilities like a Universal Translator being unable to translate anything an alien says until they are finished talking.

Human languages already do this.


In Inuktitut, the root tusaa- "to hear" by itself is meaningless other than the subject has something to do about hearing or listening. It doesn't convey enough information in itself to be useful. You have to listen to the entire utterance to understand what's being conveyed.

tusaatsiaq - to hear well

tusaatsiarunnaq - to be able to hear well

tusaatsiarunnanngittuq - to not be able to hear well

tusaatsiarunnanngittualuk - to not be able to hear very well

tusaatsiarunnanngittualuku - to be in a state of not being able to hear very well

tusaatsiarunnanngittualuujunga - I am in a state of not being able to hear very well (I can't hear very well)

Each affix adds more information, but until you have the complete list of affixes attached to that root, you don't know what the message is.

Any kind of translator, universal or not, would have to wait until the entire word/sentence is completed before it could translate. If you ever watch interpreters in action, you can see this happening. For many European languages, for instance, that generally have a similar grammatical structure with English, they don't have to wait and can translate on the fly.

"Le chien..." "The dog..." "...a chassé..." "...chased..." "...l'auto." "...the car."

For languages like Inuktitut and other agglutinative languages, or those with free word order like Latin due to the case structure, it might be mpossible to do that. If the language is SOV (subject-object-verb), then you can get the dog (subject) did something(verb) to the car(object), but until that last word, that verb, comes out, it can't be properly translated into a language where the verb comes earlier in the sentence, such as in English. The dog chased the car? The dog looked at the car? The dog heard the car? The dog pissed on the car? The dog drove the car? Without the last thing that's said, you don't know which of those it might me.

  • 2
    $\begingroup$ For more information about languages with this structure, check out Polysynthetic languages. $\endgroup$
    – Cecilia
    Commented Apr 1, 2020 at 18:02
  • 1
    $\begingroup$ Latin is also a language like this, highly inflected, ( words are spelled differently for different cases ) and no rules about word order, the endings of the word contain all the information needed. This works great until you have to deal with a large population who speaks said language with an different accent. What usually happens is the word endings which are subtle, get lost, so articles and prepositional phrases are added, and word order becomes important, if the speakers need to talk to non native speakers. Classical Latin's journey to the Romance languages is an excellent example. $\endgroup$
    – chiggsy
    Commented Apr 2, 2020 at 6:09
  • 1
    $\begingroup$ Turkish is highly inflected that you can construct a number of single-word sentences that would break down to crowded ones in other languages like English. "Varlıksızlaştıramayacakmışsınız": You wouldn't have been able to impoverish [them]. $\endgroup$
    – edgerunner
    Commented Apr 2, 2020 at 7:45
  • $\begingroup$ The “needing to wait to the end” isn’t really to do with the agglutination, surely, as about the difference in word/affix order between Inuktitut and English? The issue is just that the information English needs to put first (the subject) is what Inuktitut puts last in that sentence; the same could happen without agglutination (as it does translating German subordinate clauses to English), and wouldn’t happen in translating from Inuktitut to a language whose word order was closer to the ordering of Inuktitut’s affixes. $\endgroup$ Commented Apr 4, 2020 at 7:40
  • 2
    $\begingroup$ @PeterLeFanuLumsdaine, no, because the affixes aren't words in themselves and effect the meaning of both the root and the affixes that were preceding it, not to mention their pronunciation. Also, words can have different translations depending on context: the verbal phrase "she studies" can also mean the noun "student". While "she studies art" can be translated to "student of art", "she studies in the evening" can't be translated that same way as one is translating a verbal phrase to a noun, while the second is translating a verbal phrase to a verbal phrase. $\endgroup$ Commented Apr 6, 2020 at 15:33

I don't think you can get rid of words, but you can obfuscate them

It is almost impossible to coherently define a "word", but the smallest possible unit of language with its own meaning is called a morpheme. Like, take the word unbreakable: there's three morphemes in there, un- (negates the word), break (the core meaning), and -able (makes it an adjective). Morphemes can be stems, or they can be affixes, meaning they modify the stem. A suffix (which goes after the word) and a prefix (which goes before the word) are examples of affixes, which modify the stem "break" in the example.

Some language have stems for what others have as affixes. Like "I would die" is three different stems in English, because all three carry their own meaning, but in French it's "Je mourrais", where the word "would" is just an affix for "to die", attached to the verb.

Suffixes and prefixes are common in English; just slap them onto existing words and call it a day. Other languages, French and German among them, make greater use of simulfixes; those are affixes that change something about the word, like turn one vowel into another. An English example is "run -> ran", but e.g. German does it all the time.

You could call that simulfix a pattern which you apply on top of a word to get a different meaning. It is a coherent set of rules for which vowel turns into which other vowel. Sometimes it's not just vowels but consonants too, depending on the language. The constructed language Ithkuil goes nuts with this.

The more affixes a language uses in comparison to its stems, the more synthetic we call that language. The language Yupik is even polysynthetic; they have a word "tuntussuqatarniksaitengqiggtuq", meaning "He had not yet said again that he was going to hunt reindeer". It's all a bunch of modifiers on the stem "tuntu", meaning reindeer.

And now we get to your question. With that many affixes, you could leave out the stem and not much would be lost! Imagine a language where a unit of meaning starts as an array of empty bits, like 000000000000000. Then you sequentially apply patterns to that "words", to convey meaning, the way simulfixes do for natural languages. Like, take a pattern "every odd-numbered non-prime is incremented by 2", and give it the meaning "shy". Apply that and you get a new "word" 000000002000002, meaning shy. Take a couple dozen of these patterns and you get something like 104718361840291; and an experienced listener could deduce which patterns have been applied, and figure out the meaning.

The word itself would appear indivisible. Take one number out and the whole meaning changes, or it becomes gibberish. That way, it fulfils the conditions of the question... mostly.

Because simulfixes and affixes are still morphemes. Each pattern you apply on top of the blank slate is no different from a word in a sentence. Sure, now they all modify each other which is funky to say the least, but there's nothing philosophical unique about this language, only practically. And as with the Latin example; there's plenty of natural languages where you need to hear the entire message before you can translate it, because of word orderings.

It's just that we need words to organise meaning. We need to decompose language into reusable parts, because we cannot make one new word for every new message and expect it to be recognised by the listener. So what you can do is tricks like this to obfuscate your words - which is what the cryptographic answer does - but I think it is impossible to get rid of them.

  • 1
    $\begingroup$ "break" is not an infix "-break-" so it should have no dashes. "un-" is a prefix and "-able" is a suffix, so they should have one dash each to show that they are dependant morphemes. It should be "un- break -able", not "un- -break- -able". $\endgroup$
    – CJ Dennis
    Commented Apr 1, 2020 at 22:39
  • $\begingroup$ @CJDennis corrected $\endgroup$
    – KeizerHarm
    Commented Apr 2, 2020 at 7:08

Kind of.

An example a system for encoding messages so that they are not divisible into words is public key cryptography. The basic idea is then that the sender encodes his message using his private key and the public key of the intended recipient. Recipient then uses their private key and senders public key to decode and authenticate the message.

The messages in transit are indivisible blocks of data. They cannot be meaningfully divided.

Humans would need to convert the blocks to and from some normal language but a species that has such encoding naturally would not need the extra step and no language with discrete words would be needed. They would conceivably be directly encoding their intent with the encoding taking the place of the step where we shape our thoughts into words.

They'd probably have some mechanism similar to our words for shaping thoughts with higher precision and they might even share those constructs with each other but they'd not be part of the language used to communicate.

This only makes sense if the species also uses a communication method where communicating "blocks" makes some kind of sense. High speed bursts of ultrasound or electromagnetic radiation (from radio to light) or pigments or lights over an area of exposed skin would work, I think.

The adaptation would really be about encoding thought into blocks for communication not about encryption or authentication. Although having that too would be kind of cool if you can rationalize a evolutionary need or bioengineering.

Note that a block would correspond to a sentence or even a paragraph depending on the block size so at that level the language would still be divisible.

  • 6
    $\begingroup$ Public key cryptography is a red herring here. What you want is simply any block cipher. $\endgroup$
    – AlexP
    Commented Apr 1, 2020 at 7:28
  • $\begingroup$ @AlexP Yes, that is correct. I do say (or try at least) this in second to last paragraph but it probably should have been clearer. Or like it certainly should have been clearer. $\endgroup$ Commented Apr 1, 2020 at 17:18
  • $\begingroup$ Hang on, just because they are aliens does not mean they can ignore information theory. Signals that cannot be broken into pieces do not carry information. They are noise. Static. Crypto turns info into noise to allow it to be broadcast over insecure channels, decrypting it turns the noise back into information. The info comes first, the mathematics comes later, to work out how much info can be sent, and the crypto comes last. A stream of signals that can't be broken into smaller parts is like sending a message with a space heater. The "words" ARE the information, they have to be there. $\endgroup$
    – chiggsy
    Commented Apr 2, 2020 at 7:10
  • 1
    $\begingroup$ @chiggsy Yes, it must be possible to break it into pieces but those parts do not need to be "words". For example in digital encryption those parts are bits. In this case the parts are structures of whatever the communication channel you are using. Those parts do not need to have a one-to-one correspondence to words. So it is not necessary for it to be possible to divide the message into words. You can argue that it will still be possible to do before and after the communication but the question is only about the process of communication it is fine if "words" exist outside of it. $\endgroup$ Commented Apr 2, 2020 at 7:44
  • $\begingroup$ This isn't strictly speaking true. Most encryption algorithms in use today are block ciphers and you can decrypt each block as it arrives, so there's your "words" right there. Sure, the underlying plaintext might need multiple blocks to form a word or sentence, but in either case you don't need to wait until the end of the transmission to make any sense of it. $\endgroup$
    – Tom
    Commented Apr 2, 2020 at 13:52

While I am not sure if it is possible to create a language without morphemes/ discrete concept-representations, plenty of communication can be represented where things are "communicated simultaneously, not serially." For example: music, which is dependent on tone, texture, volume, and harmony in lieu of serially.

An alien language could easily be imagined where communication is based on complex waves such as sound waves in music, where ideas are delivered simultaneously and Fourier decomposed to find meaning. High pitched shrieking perhaps? Humans (both unconsciously and consciously) filter signals like these all the time to understand our senses.


Lossy transmission plus error-correcting codes

The transmission method or medium might be lossy enough that error-correcting codes have evolved naturally.

If in addition the transmission medium provides a natural "blocksize" that may not correspond with the typical size of words, the error correction might happen at the end of each block, thus requiring knowledge at least of the whole block to understand the message.

For example: If the communication medium is 2D (e.g. pixels on a canvas), both columns and rows may carry meaning - and parity bits or more sophisticated techniques could be appended to the end of each row and column. If the area is big enough, the information transmitted in each "batch" would be far more than individual words and due to the lossy transmission, cannot be split into individual parts without losing meaning.


You actually have two problems to solve here:

  1. How does the communication work?
  2. How can I convey that to the reader?

The biggest technical problem with non-word based communication is probably going to be nouns. While it's easy enough to communicate emotions, intentions and actions through other means (see scent/sign/body language) it's very difficult to refer to specific objects without a name/label of some sort. That will inevitably be 'word' based, even if that 'word' is a specific action/signal/telepathic-image.

How to convey it to a reader is going to be even harder! You say "most just handwave the unintelligibility", that's probably because conveying in words a language you've designed to not be conveyed in words is going to be a struggle. At best you'll have a rough translation, which clearly loses some of the meaning and feels no different to a reader than any other low-quality alien language translation. At worst, if the UniversalTranslator(tm) doesn't work, then either:

  • You can't convey to your readers what has happened, in which case why bother to develop the 'language' at all?
  • Your UniversalTranslator is clearly not very good, but you 'the narrator' can understand just fine. You're now back to explaining why your characters can't understand what's going on.
  • $\begingroup$ And abstract nouns would be far harder to communicate than concrete nouns you can physically point to at least. $\endgroup$ Commented Apr 1, 2020 at 14:34

Yes... and no.

Since no one else has done so yet, I'll point out the elephant in the room... You are writing for a human audience. Humans think in words. Therefore, no matter what ideas anyone comes up with, at some point you are going to have to translate your alien's language into words anyway. The alternative is for them to be utterly incomprehensible even to the reader, in which case the reason doesn't matter, just whether you can sell that to the reader.

That said... combining some of the other ideas here, I think it's plausible, and one of the keys is that your aliens don't know how to translate their communication into neat little pieces (even if you, as the Author, can). For that, I'll point you again at treecats, which were exactly like this before they met humans, so there is solid precedent for a species that can't 'wrap their heads around' word-based language.

There are already several ideas here for how they would communicate that you, as the Author, can use to explain to your readers what's going on. The easiest and probably best is that they are telepaths, and rather than exchanging "language", they exchange something more like a gestalt that incorporates whatever feelings, knowledge or desires they want to include in it. For example, while I might say to you "may I have a glass of water", such an alien might send a gestalt that expresses that they are thirsty, when they last had anything to drink, their memory of how water tastes, their imagined image of another alien handing them a glass of water, their imagination of how grateful this will make them feel, and so on. You could even extend this to such gestalt including, or taking the form of, some event from cultural tradition, Darmok-style. All of this could happen in a fraction of a second. (BTW, AI's might communicate this way...)

I'm also less confident that such a species couldn't develop technology. If one such alien can figure out some technological concept (say, how to smelt iron), they can certainly convey that knowledge to another alien. In fact, I could imagine a huge teaching advantage here, since you are directly sharing knowledge rather than having to encode it into words and hope that the listener can decode it correctly. (On the other hand, this might lessen the likelihood of accidental discoveries, so...)


Body language

Look at animals. Most animals I can think of on top of my mind communicate mostly or solely by body language.

Bees communicate to each other inside the hive by dancing. Dogs communicate a lot by their tails. Sharks warn prior to aggression by the way they swim. And look at how birds of paradise flirt.

Even we humans communicate a lot by the way we move and stand.

There are no words in body language, but there are whole sentences and dialogues there.

  • 1
    $\begingroup$ No, there aren't sentences in "body language." It may communicate, but it isn't language, and you can't make propositions in it. $\endgroup$ Commented Apr 1, 2020 at 14:33
  • 2
    $\begingroup$ Its so the Asimov's Second Foundation! - $\endgroup$ Commented Apr 1, 2020 at 14:59
  • 1
    $\begingroup$ I disagree. Bees use specific movements to communicate specific information. I'd call that "words". Even humans do it. Sitting forward is a very recognizable word, as is sitting back with arms crossed. Or rolling ones eyes. There is a clear 1:1 meaning with all body language I can think of. It may not translate to "cat", but it does have a specific meaning. $\endgroup$
    – Dúthomhas
    Commented Apr 2, 2020 at 23:37

I'm approaching this from a different perspective:

It is unlikely that any human or humanoid species, or anyone remotely similar to us would develop such a language, because chunking and object seperation are parts of the human brain that act long, long before even the subconscious processes. In fact, for visual information object recognition happens in the visual cortex, before the information is even handed over to the general processing parts of the brain.

As such, splitting the world we experience into distinct elements is as natural to us as breathing. It is going to be reflected in language in one way or the other, hence "words" (in the most abstract sense).

Your alien species would have to have an entirely different experience of the world, one in which things are no divided naturally. It is difficult to imagine how that would even work, and what kind of thoughts these creatures would have.

There is a second aspect and that is communication channels. Verbal communication is necessarily one-dimensional and sequential. A soundwave is produced and consumed in such a way that you have one bit of information, then the next, then the next. It is highly natural to have pauses in this stream of data, even if just for breathing. Those pauses naturally will get meaning. In fact, our sentence markers - dots, commas, etc. - started out as simple markers for different lengths of breathing pauses back in greek times when writing was just a way to note down speech.

That all means we need to move to a different kind of experience and communication. A species that would communicate visually, say with patterns of colour on a suitable piece of skin/scales/etc. would have a two-dimensional form of communication if those patterns are static (one pattern = one word). But we can extend this into three dimensions by having not the patterns themselves, but the way they shift and change contain the actual meaning.

Like a piece of music, a dance or a play, the meaning of such communication would not be in any single part of it, but in the whole. The more you split off, the more meaning would be lost. Instead of "words", you would have much larger segments of meaning, such as an entire phrase or paragraph.

In fact, a few human scripts come close to something similar. Check the bottom part of this page: http://nihonshock.com/2009/10/crazy-kanji-highest-stroke-count/ for a few examples of complex Kanji. While these are "words" in your sense and don't satisfy your requirement, they point towards the basic principle. If turned into a transmission, you would not be able to understand the entire meaning until the Kanji is complete.

Combine these two concepts - an experience that considers the world as one undivided whole and a three-dimensional communication system that has no pauses or breaks, and you would end up with a language that doesn't have words in any sense, is perfectly continuous and cannot be partially decoded.

The Trisolarans in the "Three Body Problem" books do not have communication as a seperate process. Instead, their thoughts are visible to all around, that is the closest in literature that I've seen towards such a concept. Their "speech" would not have a beginning or end because they don't actually have "speech" - they just watch each other think.


Multiplex the words

The language has words, but they are spoken syllable by syllable: first syllable of each word first, then second syllables, etc.

This can be taken down to individual letters (or, rather, phonemes).

To make this work, all words of the language would have to have the same number of syllables (or phonemes).

Those aliens might have brain structures that process utterances in parallel. Not sure how evolution might favor this - maybe if language developed in a time where sound was subject to spike noise, so multiplexing would make the communication more resilient (this scheme is being used in CDs and DVDs, where a scratch can damage a specific block but since the information is spread across multiple blocks, you get better error correction).

Make the language heavily modifying

Make the basic words mean very little if at all. Prefixes and suffixes modify the basic meaning.

This kind of stuff is already present in English, e.g. with the "in" prefix which might mean either "inside" ("into", "insert") or "not" ("intransigent").

Bonus points if this is riddled with exceptions for specific combinations.

The language does not have words at all

Communication is via gradients, not impulses.

Bandwidth can be achieved by parallelization, e.g. color patterns on the skin. (Some octopi species do this, as a mating ritual I believe.)


Thoughts are transmitted directly. Maybe they are transmitting brain processes.

This could even be hard sci-fi if you have a high-bandwidth communication - spread-spectrum radio, for example (you'd have to think about getting the radio into and out of a wet environment like the brain though, so maybe their hair is antennae and they connect deeply into the brain? - in that case, shaving would be equivalent to turning somebody deaf and mute, and hair would have to grow to specific lengths to cover the spectrum of interest).

Lieing would be hard, or just as easy as the possibility to split one's thought processes. Since the ability to misinform is one of the most important social tools (a liar can gain evolutionary advantages so this will be present in some form) but the tools for misiformatio would have to be somewhat different, this might induce interesting cultural differences.


In the absence of a formal definition for words, here are some thoughts that come to mind.

You say to want "sentences" without "words", or at least translators who have to wait for the end before beginning (already the case when grammars oppose (German vs English). There are a few ways to bind things together in a way that are not easily separable, but you are stuck with two main results:

  • Making arbitrarily large number of "atomic" (non-compound) sentences, e.g.: all "words" are complex, instead of saying "seventy-nine", they say "ozlwemifwe" which has no linguistic correlation to any other number (as for "one", "two", "eleven").
  • Making portions of meaning interact with each other in a difficult to disentangle fashion.

Some simple solutions might be


If you allow arbitrary medium, then images convey many things, can be psychic or physical. For example, an image from the culmination of "Of Mice and Men" might convey many emotions and words simultaneously. You can think of memes and emojis as a modern expression of language that attempts to harness this combination of cultural meaning.


In the same vein, symbols that intersect in different locations and manners can be given arbitrary meanings that are difficult to disentangle. Taking the example above, you might have "friend" intersecting "coworker" in the relationship section of the symbol, "life" in the contemplation section, and an action area dominated by "end" which intersects "life" and "coworker", but not "friend". There are words, but they become atomic and may be imbued with cultural significance such that the components cannot be disentangled - e.g. "wag the dog" (or "correct battery horse staple").


Telepathy with images and sensations rather than words. Imagine that saying that you saw someone steal something, you instead presented your entire memory of the event. Or if you wanted something to get you something you made a memory of them going away and coming back with the thing you wanted, possibly with an interim scene of them getting it from someone else.

You could then have the memory keepers, those that keep the memories of lessons learnt in the past in an uncorrupted form. I doubt that a society like this would ever get to level of the renaissance let alone industrial or modern era without developing some sort of language. I think this would be because dials and switches are their own form of language. I'd also imagine that it would be hard for them to do anything that involves measurements outside of ratios based on body parts.

Introducing a word using Alien species would almost certainly destroy the society, as being able to communicate complex and abstract theories is much easier with words than it is with video.

  • $\begingroup$ How do you communicate in telepathy whether the vision is real or fictional, literal or metaphorical? $\endgroup$ Commented Apr 1, 2020 at 14:34
  • $\begingroup$ Has someone been reading about treecats? $\endgroup$
    – Matthew
    Commented Apr 1, 2020 at 14:40
  • $\begingroup$ I don't know as I'm not telepathic, but you can just assume that they can tell the difference between a memory and a non-memory. Also, I've never heard of treecats. $\endgroup$ Commented Apr 1, 2020 at 14:53
  • $\begingroup$ Thinking about it, it's probably the amount of detail in the sending, much like the difference between a painting and a photo, except in 3D with sensations rather than just a 2D visual. $\endgroup$ Commented Apr 1, 2020 at 17:31
  • $\begingroup$ Telepathy raises the issue of whether neural net memory is decomposable into words. $\endgroup$ Commented Apr 2, 2020 at 10:38

Transmit your communications across a second dimension

Take the following sentence:

There is a horse in aisle four

That sentence could be described several ways: as a series of letters, or a series of words, but the way your screen describes it is as a series of rows of pixels. If you imagine the letters Th (the first two letters of our sentence) as being represented by the matrix:


and I were to transmit to you:


You could definitely decipher it, but you wouldn't know what the message was until you'd received the entire transmission, because to make it intelligible to you, you would have to stack the "words".

Our communications contain layers of meaning that are obvious to us when we know the words. Make your alien language contain layers of words that are obvious once you know the meaning.

I could say "I bought my grandmother flowers after my grandfather's funeral."

Your alien might express something that we'd interpret as

  • "Someone is sad because a mutual loved one died"
  • "I love that other person too."
  • "I wanted to make her happy"
  • "Flowers make her happy"

If you "stack" those meanings together, you might derive my original sentence, but you wouldn't be able to do so until you'd received all of the "meanings" that I was using to communicate.


Mapped philosophy and methods

If those aliens mapped every possible concept on a map (depending on their philosophy, it has great chances to be a two-dimensional disk or a three dimensional sphere, or something complicated depending on how they consider time), the speaker could use a sentence describing the course of its thoughts through functions describing waves, jumps, areas and intersections running through that map.

The words that describe this pattern would have no meaning at all by themselves (or only a few if the speaker starts by telling what the starting point is), and the meaning of those sentences could be infinite, leading to a lot of poetic ways to describe the infinite, and enormous amounts of possible synonyms that could be accessible to knowledgeable aliens and calculators/translators only, but could be roughly understood by those capable of feeling the magnitude of the functions.

Most used groupings of functions would certainly get shortcuts associated to them accessible from the map itself.

New methods of joking could appear, such as describing a concept that once re transcribed into its shape on the map, draws the shape of a banana.

Mathematical and geometrical figures can be described by words, sentences can be turned into geometrical forms and create ideograms.

Elitists can then hide their work into cryptic sentences representing a mathematical function representing another sentence or representing the shape of an atom, molecule, banana... etc.

Protection against incomplete sentences

A language where the first "words" meanings and pronunciation are modified by the last "words" would require the speaker to know before talking the entire structure of his thoughts, and would require the target to hear the entire sentence before understanding its meaning, making it impossible to jump to a conclusion in the middle of an explanation.


Encoding all concepts into numbers

Lets say you have a race of beings which had near infinite memory and very long life spans. Since they can memorize almost everything perfectly, they don't need a way to decompose their language into modular easy to remember parts and can simple convey a whole concept with one number. 12308102398 might mean "I saw a blue dog on my way to work". This race would likely have to spend an enormous amount of time in the beginning of their lives reading through all the number to concept mappings. You could also do something interesting with the ordering of the numbers, the lower numbers being more early concepts (1->"Round rocks are easier to push") or more important concepts (such as religious concepts or moral concepts 1->"Don't harm others").

  • $\begingroup$ The quipu, used by the Inca empire to communicate administrative data, was basically a system of numbers. The numbers are represented by knots in strings, and the tree structure of data is represented by tying strings onto other strings. They had oral language by no written language. It's not clear to me whether quipu language is made up of "words" in the most general sense. $\endgroup$ Commented Apr 2, 2020 at 10:37

Symphonic language

I'm fascinated by the ability of animals that use echolocation to generate signals that can be interpreted by others as the response to an echolocation click. What that means is that, for instance, dolphins can send one another signals that mimic the signature of an object, like a shark or a gap in the ocean floor or a school of fishes. It's not a word, it's a full description of an imaginary part of the world.

Now, imagine a race able to do this and, more, able to combine these messages into a coherent whole. As long as a person is speaking, they're creating a world. Each new voice adds meaning to this world, about its nature, its meaning, its past, its future...

When a person leaves the conversation, they remove all the meaning they brought, unless others have picked up on their "themes" and integrated them into their own messages.

An individual part would make no sense to an alien observer. Only the part plus the context has meaning. This language would be even tougher to translate if these people were able to replay in their heads past contexts. Others would understand that they're just modifying a well-known and very intricate conversation, while outsiders would just hear some random droning.



... by your own definition. Yes, it's possible to imagine a language where complex thoughts are communicated by distinct expressions, but the number of such expressions needed would far exceed the knowledge limit you stipulated regarding human languages:

There are few enough words for a speaker of the language to know most of them (i.e. less than about a million for humans).

In the language you imagine, these distinct expressions become the irreducible units, but the number of such expression would be astronomical. A quick search suggests that research shows that in English (for example) there are ~3,000 words used for 90% of everyday usage. Let's pick 8 words as an arbitrary number of words in a complex thought. 3,000 ^ 8 is 6.6e+27. I fully realize that a great many such combinations make no sense whatsoever, but the shear magnitude here is telling. e.g., let's say only 1/1,000,000th of those possible combinations makes a coherent thought; great! now we are down to 6.6e+21! How about 1/100,000,000,000,000 (1 in 100 trillion)? Now we are down to 6.6e+13. We are trying to get to 1e+6 (your "about a million"). That's the power of orders of magnitude for you.

(In reality, I think the number of words an average person knows is more like 20,000. There aren't even 200,000 words in English. [1])

I've made some wild assumptions here, but it's enough (I feel) to confidently say there would be too many expressions to know. Why? Because I know it takes a lot more than 8 words to express some coherent, complex, individual thoughts. And because I haven't even considered numbers. There are lots of those, and a language without distinct words would need, gosh, lots of unique expressions for any thought that involved even one number. Then, let's talk about irrational numbers like pi, and how many digits of precision are needed for an advanced (by which I assume you mean star-faring, since you mention a UT and such) civilization. This language would need a unique expression for every additional decimal of precision of pi -- for every coherent thought that involved pi.

So, while I've been around the block enough times to avoid using the word "impossible" when talking about what we might find out there elsewhere in the galaxy or the universe, the members of this civilization you imagine would need to have an impractically-capable brain, capable of keeping track of an astronomical number of unique expressions.

... Besides which, if they can do that, and meet your 3rd criterion, ... well, then those become the words of that language, because such a body of expression would certainly meet your 1st and 2nd criteria for words, being irreducible, and certainly being able to be used together to communicate even more complex ideas.

[1] "The Second Edition of the 20-volume Oxford English Dictionary, published in 1989, contains full entries for 171,476 words in current use, and 47,156 obsolete words. To this may be added around 9,500 derivative words included as subentries."




Humans already use this to some extend - maybe your aliens have taken it to the extreme.

If I were to tell you about (rather than show a picture of) "a dog with a hat, sitting in a room on fire, saying this is fine" you could probably piece it together if you knew the words making up that sentence.

If I were to go further, I might use memes that require knowledge of their context to properly parse. Maybe something like "we don't need to prepare for a pandemic", followed by those vegetables labeled "corona" - you may or may not know them, so if I don't include the "allow us to introduce ourselves" text, only people who already know that this is what the vegetables are supposed to say will understand it.

Another example at that level: Advice animals. If I give you a piece of advice, you should take it quite differently if it comes with an "actual advice mallard" (duck picture) compared to the same advice coming from a "foul bachelor frog" or "bad advice dog".

Other memes may require knowledge of TV shows, cultural events, or even a history of misusing a meme in a specific context. Translating a meme thus would require the text, the picture, and the context fully parsed - trying to deliver a translation for "you should definitely bring homemade pot brownies to the reception..." before hearing the part "... says bad advice dog" would be disastrous.


Let's say we have a race of intelligent camouflaging cuttlefish. They have no vocal chords, so they communicate by changing the colors of their body, sort of like our real cephalopods actually do, but way more advanced. They still have not enough resolution to show clear pictures, nor enough FPS to show movies, so they use more abstract communication signals, like differently shaped and colored spots (possibly differently textured and positioned as well, etc.).

It takes time to fully form a single finished spot, but they take advantage of the fact that they can keep changing colors on their entire back (or whatever visual speech organ they have developed) at the same time. They start with a configuration of very fuzzy grayish spots, gradually turning them into sharper, clearer, brighter spots. An initial fuzzy spot is already meaningful, but that meaning is very vague and abstract. By changing it towards its final color and shape, they gradually add more attributes to it, until it becomes something clear and unambiguous.

When we, humans, speak, the moment we start speaking we have a clear mental image, but only of a small part of the "big picture", so for us it's like assembling a jigsaw puzzle. When the cuttlefish "speak", the moment they start speaking they already have the entire "big picture", but it's very vague, so for them it's like creating a watercolor painting - the background, the basic shapes, more and more details on top of that. Or maybe like developing a film.

For example, let's say a cuttlefish is communicating this sentence: "A man killed a fly with a slipper." If your Universal Translator is running at the moment they start "speaking", the translation would be: "A (grayish blob) (grayish blob)-bed a (grayish blob) with a (grayish blob)." A split of second later, more information is added, so it becomes: "A (living creature) (harmed) a (living creature) with a (footwear)." And so on, and so on. Although you can start translating at the very beginning of the "speech", it doesn't make much sense to us until the "speech" is finished, because that's not how we talk.

To make things even more complicated, let's say you have to carefully watch the entire metamorphosis of each blob to understand its full and final meaning - so you can't just wait till the end and then take a picture of the last "frame" with your smartphone. For example, a "grayish blob" first evolves into a "creature", then into a "man", then splits in two - the "man" and his "slipper", something like that.


You must log in to answer this question.

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