In this article about miniature artificial languages on motherboard.com experiments are described where children learn carafully designed miniature artificial languages. It is also said that children give up on learning "unlearnable" languages. A mentioned miniature artificial language is named "Sillyspeak" or "Sillysprochen".

What features of languages make them unlearnable? Is the full documentation of Sillyspeak available?

  • 3
    Possibly not about conlangs, but better in Linguistics SE, as the answer sought for is about language acquisition, which falls into the domain of linguistics more than conlangs.
    – Duncan
    Commented Mar 19, 2018 at 19:03
  • 1
    Indeed I can ask it over there, too. But I wanted to introduce the concept of miniature artificial languages here, and this site also lacks question activity.
    – Sir Cornflakes
    Commented Mar 19, 2018 at 21:09
  • Fair enough. Such invented languages are often referred to as "experimental languages", and Sillyspeak seems to be just such a thing!
    – elemtilas
    Commented Apr 7, 2018 at 1:46

4 Answers 4


If you search for the researchers mentioned in the article, you can find some of their research papers. In those, they describe their methodology. In one paper that I looked at briefly, for example, they mention that they varied the consistency by making the placement of determiners random. (This is actually also mentioned in the Motherboard article).

Their research seems to be focused on finding out how children can learn languages, and what universal features of languages there are, so they up the difficulty of their experimental mini-languages by using rare elements, such as OSV word order (most (Western) languages use SVO). Discovery that some mini-languages are unlearnable is thus a side-effect of pushing the limits as to what are possible human languages. From a cursory glance at several papers I'd think that you can make a language 'unlearnable' if you add some random variation and choose a lot of rare grammatical features.

PS: in the Motherboard article they actually give a reason why no documentation of Sillyspeak is available: they want to avoid it 'escaping' into the wild, as its purpose is to be a completely unknown language to the subjects of the experiments.


Original Fith is a stack-based conlang (LIFO, reverse polish notation) and considered to be unspeakable in real-time since in order to speak good Fith you need to be able to remember more parts and recurse deeper than what is practically possible for a human. It also has a full set stack operators like "swap the order of the two last stack items" and *rotate the top three stack items".


I've seen a paper that children have a lot of trouble learning nonconservative determiners.

For instance, "every" is a conservative determiner because every dog is brown is equivalent to every dog is a brown dog – objects that are brown but not a dog don't count toward the truth of the sentence. In contrast, if we define equi dogs are brown as "the number of dogs is equal to the number of brown objects", then equi would be nonconservative because the sentence above is different from equi dogs are brown dogs.

(Interestingly, my conlang Jbl has nonconservative determiners; for instance, glsh is the equi mentioned above.)


Lojban is often considered unusuable by humans because of its design.

A better example however, may be languages that have features that humans simply can not use. For example, Elkaril has an odd grammatical feature where precise measurements can be given by how long a consonant is held, and mentions that the 'infinite space' in-between vowels can also be used to encode precise information, though I can't seem to recall what purpose that serves. Obviously, humans couldn't possibly use a language like this, and that was intentional.

Personally, I limit myself to features natlangs have. Not that I try to create naturalisitic conlangs, I just do that because natlangs are of course tried and tested beyond what any conlang could be, so I can be certain that the features they have are something a human could actually use. I used to have a problem with creating weird languages that were clearly unusable by humans.

Human minds work a specific way, and despite our diversity there ARE universals (for example, all languages have nouns and verbs, though otherwise the list of parts of speech within a language can vary widely). So there are limits to what ways a language can work. You may have some wiggle room, but you can't just make up whatever you can imagine and expect it to be usable.

  • 3
    "all languages have nouns and verbs" That's not true.
    – Sparksbet
    Commented Mar 26, 2018 at 20:14
  • That language still describes some words as being more 'noun-like' or 'verb-like'. Just because the distinction is somewhat blurred doesn't mean that a language doesn't differentiate the two. Some languages treat adjectives as a sub-class of verbs or adjectives. No language treats nouns and verbs as being a sub-class of an other part of speech.
    – user348
    Commented Mar 27, 2018 at 9:31
  • Languages that treat adjectives as a subclass of verbs can be said to not have verbs, depending on the specifics. This family of languages could be said to have nouns and verbs as part of a single class and many linguists due indeed analyze the languages as lacking nouns entirely. You can try to add artificial distinctions onto another language and claim this language has nouns and verbs because some words seem more noun-like, but that's operating based on your conclusion and your (Eurocentric) beliefs about language rather than on the linguistic evidence.
    – Sparksbet
    Commented Mar 27, 2018 at 12:36
  • As for my previous comment, I made a stupid typo. When I typed 'as a sub-class of verbs or adjectives', I meant to type 'verbs or NOUNS'.
    – user348
    Commented Mar 27, 2018 at 17:51
  • 2
    If word operates as a verb syntactically, it's nonsensical to call it a noun even if it carries what you consider 'noun-like' meaning. Whether something operates sufficiently like a verb syntactically is what actual linguists argue about in natural languages like this -- the meaning doesn't really matter when it comes to these sorts of classes.
    – Sparksbet
    Commented Mar 28, 2018 at 20:53

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.