Typically, a serializable language can be expressed by strings.
Can a natural language be non-serializable?

Fictional source which motivated me to ask this question: a character in C. Stross' novel "Accelerando" is mentioned to speak several non-serializable languages.

  • 5
    "Story of Your Life", a Ted Chiang's fiction story, describes an alien's language that has no progression in time; its clauses are non-serializable and perceived as a whole. Dec 3, 2019 at 17:33
  • 2
    @bytebuster But it's also not fully described. I'm not aware of any conlang that's not serializable.
    – curiousdannii
    Dec 4, 2019 at 13:21
  • They're certainly not separate, parallel channels, but would any supersegmental features qualify?
    – anomaly
    Dec 4, 2019 at 14:17
  • 1
    Is this "non-serialisable" just by claim of the author, or is some reasoning given how and why it isn't serialisable? Dec 6, 2019 at 13:49
  • 1
    @jknappen-ReinstateMonica more like a claim
    – J. Doe
    Dec 6, 2019 at 13:50

4 Answers 4


In one sense, every language is serializable: record someone speaking or signing it, then encode that video into whatever format you like, and now it's been turned into a string of bits. But that's not a very interesting answer.

According to many theories, all spoken languages are fundamentally linear: they're made up of a linear sequence of phonemes, one after another. There are never multiple independent phonemes happening at once (*). So these languages are always serializable.

Signed languages are more complicated. In most signed languages, you can have multiple "phonemes" (cheremes?) happening at a time: different movements, hand positions, and facial gestures can all be happening at once. So figuring out how to serialize these languages is much more difficult. It can still be done—the most famous example is Sutton SignWriting—but it's significantly more complicated, since the language isn't already mostly serialized for you.

(*) The biggest exception being tone, in autosegmental theories. So you need to come up with a convention for this, like "write the tone for a syllable immediately after the syllable nucleus", which is what the IPA did.

  • 3
    An interesting spin on the first paragraph: the number of strings is countably infinite. A language which can convey an uncountably infinite number of concepts would, in theory, not map to strings. A language with those properties is actually a rather interesting thing to explore.
    – Cort Ammon
    Dec 4, 2019 at 5:01
  • In theory being linear does not necessarily imply that they are serializable in the usual meaning, as there could be an uncountable number of slight variations of length or tone of vowels. However, in practice no man could notice the tiny differences. Dec 4, 2019 at 14:01
  • @CortAmmon-ReinstateMonica Natural languages exist to transfer information about something specific (I mean the naturally evolve from something like that). Can one convey information about something specific if language designed to convey an uncountable infinite number of concepts?
    – user28434
    Dec 4, 2019 at 14:59
  • 1
    @user28434 The difficulty of answering that question is why I find it so interesting. It tests the limits of my own creativity regarding what language could and could not be and what could or could not be conveyed.... and what does it mean to "convey" something anyways? (it quickly becomes one of those "turtles all the way down" sorts of things)
    – Cort Ammon
    Dec 4, 2019 at 16:41
  • 1
    @CortAmmon-ReinstateMonica: you may be interested in "The Vastness of Natural Languages D. Terence Langendoen & Paul M. Postal Linguistics and Philosophy 9 (2):225-243 (1986) ". I never read the paper, but I remember enjoying some of the discussions about it.
    – Colin Fine
    Dec 6, 2019 at 13:47

For spoken languages that make the vast majority of natural languages, the answer is a plain no. By speaking we are forced to serialise our thoughts.

Sign languages are a bit more difficult to analyse because there are several things happening simultaneously. However, they can be written in a kind of partitur format, and in the very last be serialised, but the alphabets for the resulting strings like Sutton sign writing can become rather complex.

TL;DR: All natural languages are serialisable.

  • 1
    Ah, so you mean that because time seems to be one-dimensional, serialisable, so does language? Dec 4, 2019 at 5:41
  • 1
    Our linguistic theories of human natural language, some of them, make it to be serial. However, that doesn't make it true that it is serial.
    – Greg Lee
    Dec 4, 2019 at 19:12
  • "Being serial" and "being serialisable" are two different things, the latter being weaker than the former. It is really hard to think of things that aren't serialisable with some effort (maybe the complex numbers from mathematics apply here) Dec 6, 2019 at 13:45

This second answer of mine is much more speculative, and it says

any language—natural or not—is serialisable

The function of language is to convey some information from a sender to a recipient. The amount of information is always finite and can be measured in bits or a similar unit for measuring information. We also know from quantum mechanics, that information is bound to a minimal amount of energy. So, any language uses some physical encoding of a finite amount of bits. This encoding of bits can always be serialised.


Could an ET's language count as a natural language? I think so, and so that is what I will assume. Then imagine an ET (such as VOR's color display creature) which uses a non-serializable channel for language communication, then, sure, there is no guarantee that ET's utterances in his non-serial language could be adequately translated into a serial human language. So the answer is "yes".

VOR is a novel by James Blish. VOR

  • Well, only observed languages of observed extraterrestial's count as natural here. And show me a non-serialisable channel of communication: TV shows very well that an optical channel is serialisable. Dec 6, 2019 at 13:47
  • @jknappen-ReinstateMonica, Is what is natural really confined to what is observed?
    – Greg Lee
    Dec 7, 2019 at 17:05
  • At least it should be observable to count as natural. Dec 7, 2019 at 18:07
  • @jknappen-ReinstateMonica, Perhaps you're confusing already observed with observable. There will be nothing unobservable about the ET languages when we encounter them. Probably.
    – Greg Lee
    Dec 7, 2019 at 21:22

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.